1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,
2 @c 2002, 2003, 2004, 2005, 2006, 2007 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
7 @chapter Target Description Macros and Functions
8 @cindex machine description macros
9 @cindex target description macros
10 @cindex macros, target description
11 @cindex @file{tm.h} macros
13 In addition to the file @file{@var{machine}.md}, a machine description
14 includes a C header file conventionally given the name
15 @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}.
16 The header file defines numerous macros that convey the information
17 about the target machine that does not fit into the scheme of the
18 @file{.md} file. The file @file{tm.h} should be a link to
19 @file{@var{machine}.h}. The header file @file{config.h} includes
20 @file{tm.h} and most compiler source files include @file{config.h}. The
21 source file defines a variable @code{targetm}, which is a structure
22 containing pointers to functions and data relating to the target
23 machine. @file{@var{machine}.c} should also contain their definitions,
24 if they are not defined elsewhere in GCC, and other functions called
25 through the macros defined in the @file{.h} file.
28 * Target Structure:: The @code{targetm} variable.
29 * Driver:: Controlling how the driver runs the compilation passes.
30 * Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}.
31 * Per-Function Data:: Defining data structures for per-function information.
32 * Storage Layout:: Defining sizes and alignments of data.
33 * Type Layout:: Defining sizes and properties of basic user data types.
34 * Registers:: Naming and describing the hardware registers.
35 * Register Classes:: Defining the classes of hardware registers.
36 * Old Constraints:: The old way to define machine-specific constraints.
37 * Stack and Calling:: Defining which way the stack grows and by how much.
38 * Varargs:: Defining the varargs macros.
39 * Trampolines:: Code set up at run time to enter a nested function.
40 * Library Calls:: Controlling how library routines are implicitly called.
41 * Addressing Modes:: Defining addressing modes valid for memory operands.
42 * Anchored Addresses:: Defining how @option{-fsection-anchors} should work.
43 * Condition Code:: Defining how insns update the condition code.
44 * Costs:: Defining relative costs of different operations.
45 * Scheduling:: Adjusting the behavior of the instruction scheduler.
46 * Sections:: Dividing storage into text, data, and other sections.
47 * PIC:: Macros for position independent code.
48 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
49 * Debugging Info:: Defining the format of debugging output.
50 * Floating Point:: Handling floating point for cross-compilers.
51 * Mode Switching:: Insertion of mode-switching instructions.
52 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
53 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
54 * PCH Target:: Validity checking for precompiled headers.
55 * C++ ABI:: Controlling C++ ABI changes.
56 * Misc:: Everything else.
59 @node Target Structure
60 @section The Global @code{targetm} Variable
62 @cindex target functions
64 @deftypevar {struct gcc_target} targetm
65 The target @file{.c} file must define the global @code{targetm} variable
66 which contains pointers to functions and data relating to the target
67 machine. The variable is declared in @file{target.h};
68 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
69 used to initialize the variable, and macros for the default initializers
70 for elements of the structure. The @file{.c} file should override those
71 macros for which the default definition is inappropriate. For example:
74 #include "target-def.h"
76 /* @r{Initialize the GCC target structure.} */
78 #undef TARGET_COMP_TYPE_ATTRIBUTES
79 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
81 struct gcc_target targetm = TARGET_INITIALIZER;
85 Where a macro should be defined in the @file{.c} file in this manner to
86 form part of the @code{targetm} structure, it is documented below as a
87 ``Target Hook'' with a prototype. Many macros will change in future
88 from being defined in the @file{.h} file to being part of the
89 @code{targetm} structure.
92 @section Controlling the Compilation Driver, @file{gcc}
94 @cindex controlling the compilation driver
96 @c prevent bad page break with this line
97 You can control the compilation driver.
99 @defmac SWITCH_TAKES_ARG (@var{char})
100 A C expression which determines whether the option @option{-@var{char}}
101 takes arguments. The value should be the number of arguments that
102 option takes--zero, for many options.
104 By default, this macro is defined as
105 @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options
106 properly. You need not define @code{SWITCH_TAKES_ARG} unless you
107 wish to add additional options which take arguments. Any redefinition
108 should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for
112 @defmac WORD_SWITCH_TAKES_ARG (@var{name})
113 A C expression which determines whether the option @option{-@var{name}}
114 takes arguments. The value should be the number of arguments that
115 option takes--zero, for many options. This macro rather than
116 @code{SWITCH_TAKES_ARG} is used for multi-character option names.
118 By default, this macro is defined as
119 @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
120 properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
121 wish to add additional options which take arguments. Any redefinition
122 should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
126 @defmac SWITCH_CURTAILS_COMPILATION (@var{char})
127 A C expression which determines whether the option @option{-@var{char}}
128 stops compilation before the generation of an executable. The value is
129 boolean, nonzero if the option does stop an executable from being
130 generated, zero otherwise.
132 By default, this macro is defined as
133 @code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard
134 options properly. You need not define
135 @code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional
136 options which affect the generation of an executable. Any redefinition
137 should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check
138 for additional options.
141 @defmac SWITCHES_NEED_SPACES
142 A string-valued C expression which enumerates the options for which
143 the linker needs a space between the option and its argument.
145 If this macro is not defined, the default value is @code{""}.
148 @defmac TARGET_OPTION_TRANSLATE_TABLE
149 If defined, a list of pairs of strings, the first of which is a
150 potential command line target to the @file{gcc} driver program, and the
151 second of which is a space-separated (tabs and other whitespace are not
152 supported) list of options with which to replace the first option. The
153 target defining this list is responsible for assuring that the results
154 are valid. Replacement options may not be the @code{--opt} style, they
155 must be the @code{-opt} style. It is the intention of this macro to
156 provide a mechanism for substitution that affects the multilibs chosen,
157 such as one option that enables many options, some of which select
158 multilibs. Example nonsensical definition, where @option{-malt-abi},
159 @option{-EB}, and @option{-mspoo} cause different multilibs to be chosen:
162 #define TARGET_OPTION_TRANSLATE_TABLE \
163 @{ "-fast", "-march=fast-foo -malt-abi -I/usr/fast-foo" @}, \
164 @{ "-compat", "-EB -malign=4 -mspoo" @}
168 @defmac DRIVER_SELF_SPECS
169 A list of specs for the driver itself. It should be a suitable
170 initializer for an array of strings, with no surrounding braces.
172 The driver applies these specs to its own command line between loading
173 default @file{specs} files (but not command-line specified ones) and
174 choosing the multilib directory or running any subcommands. It
175 applies them in the order given, so each spec can depend on the
176 options added by earlier ones. It is also possible to remove options
177 using @samp{%<@var{option}} in the usual way.
179 This macro can be useful when a port has several interdependent target
180 options. It provides a way of standardizing the command line so
181 that the other specs are easier to write.
183 Do not define this macro if it does not need to do anything.
186 @defmac OPTION_DEFAULT_SPECS
187 A list of specs used to support configure-time default options (i.e.@:
188 @option{--with} options) in the driver. It should be a suitable initializer
189 for an array of structures, each containing two strings, without the
190 outermost pair of surrounding braces.
192 The first item in the pair is the name of the default. This must match
193 the code in @file{config.gcc} for the target. The second item is a spec
194 to apply if a default with this name was specified. The string
195 @samp{%(VALUE)} in the spec will be replaced by the value of the default
196 everywhere it occurs.
198 The driver will apply these specs to its own command line between loading
199 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
200 the same mechanism as @code{DRIVER_SELF_SPECS}.
202 Do not define this macro if it does not need to do anything.
206 A C string constant that tells the GCC driver program options to
207 pass to CPP@. It can also specify how to translate options you
208 give to GCC into options for GCC to pass to the CPP@.
210 Do not define this macro if it does not need to do anything.
213 @defmac CPLUSPLUS_CPP_SPEC
214 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
215 than C@. If you do not define this macro, then the value of
216 @code{CPP_SPEC} (if any) will be used instead.
220 A C string constant that tells the GCC driver program options to
221 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
223 It can also specify how to translate options you give to GCC into options
224 for GCC to pass to front ends.
226 Do not define this macro if it does not need to do anything.
230 A C string constant that tells the GCC driver program options to
231 pass to @code{cc1plus}. It can also specify how to translate options you
232 give to GCC into options for GCC to pass to the @code{cc1plus}.
234 Do not define this macro if it does not need to do anything.
235 Note that everything defined in CC1_SPEC is already passed to
236 @code{cc1plus} so there is no need to duplicate the contents of
237 CC1_SPEC in CC1PLUS_SPEC@.
241 A C string constant that tells the GCC driver program options to
242 pass to the assembler. It can also specify how to translate options
243 you give to GCC into options for GCC to pass to the assembler.
244 See the file @file{sun3.h} for an example of this.
246 Do not define this macro if it does not need to do anything.
249 @defmac ASM_FINAL_SPEC
250 A C string constant that tells the GCC driver program how to
251 run any programs which cleanup after the normal assembler.
252 Normally, this is not needed. See the file @file{mips.h} for
255 Do not define this macro if it does not need to do anything.
258 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
259 Define this macro, with no value, if the driver should give the assembler
260 an argument consisting of a single dash, @option{-}, to instruct it to
261 read from its standard input (which will be a pipe connected to the
262 output of the compiler proper). This argument is given after any
263 @option{-o} option specifying the name of the output file.
265 If you do not define this macro, the assembler is assumed to read its
266 standard input if given no non-option arguments. If your assembler
267 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
268 see @file{mips.h} for instance.
272 A C string constant that tells the GCC driver program options to
273 pass to the linker. It can also specify how to translate options you
274 give to GCC into options for GCC to pass to the linker.
276 Do not define this macro if it does not need to do anything.
280 Another C string constant used much like @code{LINK_SPEC}. The difference
281 between the two is that @code{LIB_SPEC} is used at the end of the
282 command given to the linker.
284 If this macro is not defined, a default is provided that
285 loads the standard C library from the usual place. See @file{gcc.c}.
289 Another C string constant that tells the GCC driver program
290 how and when to place a reference to @file{libgcc.a} into the
291 linker command line. This constant is placed both before and after
292 the value of @code{LIB_SPEC}.
294 If this macro is not defined, the GCC driver provides a default that
295 passes the string @option{-lgcc} to the linker.
298 @defmac REAL_LIBGCC_SPEC
299 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
300 @code{LIBGCC_SPEC} is not directly used by the driver program but is
301 instead modified to refer to different versions of @file{libgcc.a}
302 depending on the values of the command line flags @option{-static},
303 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
304 targets where these modifications are inappropriate, define
305 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
306 driver how to place a reference to @file{libgcc} on the link command
307 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
310 @defmac USE_LD_AS_NEEDED
311 A macro that controls the modifications to @code{LIBGCC_SPEC}
312 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
313 generated that uses --as-needed and the shared libgcc in place of the
314 static exception handler library, when linking without any of
315 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
319 If defined, this C string constant is added to @code{LINK_SPEC}.
320 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
321 the modifications to @code{LIBGCC_SPEC} mentioned in
322 @code{REAL_LIBGCC_SPEC}.
325 @defmac STARTFILE_SPEC
326 Another C string constant used much like @code{LINK_SPEC}. The
327 difference between the two is that @code{STARTFILE_SPEC} is used at
328 the very beginning of the command given to the linker.
330 If this macro is not defined, a default is provided that loads the
331 standard C startup file from the usual place. See @file{gcc.c}.
335 Another C string constant used much like @code{LINK_SPEC}. The
336 difference between the two is that @code{ENDFILE_SPEC} is used at
337 the very end of the command given to the linker.
339 Do not define this macro if it does not need to do anything.
342 @defmac THREAD_MODEL_SPEC
343 GCC @code{-v} will print the thread model GCC was configured to use.
344 However, this doesn't work on platforms that are multilibbed on thread
345 models, such as AIX 4.3. On such platforms, define
346 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
347 blanks that names one of the recognized thread models. @code{%*}, the
348 default value of this macro, will expand to the value of
349 @code{thread_file} set in @file{config.gcc}.
352 @defmac SYSROOT_SUFFIX_SPEC
353 Define this macro to add a suffix to the target sysroot when GCC is
354 configured with a sysroot. This will cause GCC to search for usr/lib,
355 et al, within sysroot+suffix.
358 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
359 Define this macro to add a headers_suffix to the target sysroot when
360 GCC is configured with a sysroot. This will cause GCC to pass the
361 updated sysroot+headers_suffix to CPP, causing it to search for
362 usr/include, et al, within sysroot+headers_suffix.
366 Define this macro to provide additional specifications to put in the
367 @file{specs} file that can be used in various specifications like
370 The definition should be an initializer for an array of structures,
371 containing a string constant, that defines the specification name, and a
372 string constant that provides the specification.
374 Do not define this macro if it does not need to do anything.
376 @code{EXTRA_SPECS} is useful when an architecture contains several
377 related targets, which have various @code{@dots{}_SPECS} which are similar
378 to each other, and the maintainer would like one central place to keep
381 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
382 define either @code{_CALL_SYSV} when the System V calling sequence is
383 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
386 The @file{config/rs6000/rs6000.h} target file defines:
389 #define EXTRA_SPECS \
390 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
392 #define CPP_SYS_DEFAULT ""
395 The @file{config/rs6000/sysv.h} target file defines:
399 "%@{posix: -D_POSIX_SOURCE @} \
400 %@{mcall-sysv: -D_CALL_SYSV @} \
401 %@{!mcall-sysv: %(cpp_sysv_default) @} \
402 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
404 #undef CPP_SYSV_DEFAULT
405 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
408 while the @file{config/rs6000/eabiaix.h} target file defines
409 @code{CPP_SYSV_DEFAULT} as:
412 #undef CPP_SYSV_DEFAULT
413 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
417 @defmac LINK_LIBGCC_SPECIAL_1
418 Define this macro if the driver program should find the library
419 @file{libgcc.a}. If you do not define this macro, the driver program will pass
420 the argument @option{-lgcc} to tell the linker to do the search.
423 @defmac LINK_GCC_C_SEQUENCE_SPEC
424 The sequence in which libgcc and libc are specified to the linker.
425 By default this is @code{%G %L %G}.
428 @defmac LINK_COMMAND_SPEC
429 A C string constant giving the complete command line need to execute the
430 linker. When you do this, you will need to update your port each time a
431 change is made to the link command line within @file{gcc.c}. Therefore,
432 define this macro only if you need to completely redefine the command
433 line for invoking the linker and there is no other way to accomplish
434 the effect you need. Overriding this macro may be avoidable by overriding
435 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
438 @defmac LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
439 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
440 directories from linking commands. Do not give it a nonzero value if
441 removing duplicate search directories changes the linker's semantics.
444 @defmac MULTILIB_DEFAULTS
445 Define this macro as a C expression for the initializer of an array of
446 string to tell the driver program which options are defaults for this
447 target and thus do not need to be handled specially when using
448 @code{MULTILIB_OPTIONS}.
450 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
451 the target makefile fragment or if none of the options listed in
452 @code{MULTILIB_OPTIONS} are set by default.
453 @xref{Target Fragment}.
456 @defmac RELATIVE_PREFIX_NOT_LINKDIR
457 Define this macro to tell @command{gcc} that it should only translate
458 a @option{-B} prefix into a @option{-L} linker option if the prefix
459 indicates an absolute file name.
462 @defmac MD_EXEC_PREFIX
463 If defined, this macro is an additional prefix to try after
464 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
465 when the @option{-b} option is used, or the compiler is built as a cross
466 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
467 to the list of directories used to find the assembler in @file{configure.in}.
470 @defmac STANDARD_STARTFILE_PREFIX
471 Define this macro as a C string constant if you wish to override the
472 standard choice of @code{libdir} as the default prefix to
473 try when searching for startup files such as @file{crt0.o}.
474 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
475 is built as a cross compiler.
478 @defmac STANDARD_STARTFILE_PREFIX_1
479 Define this macro as a C string constant if you wish to override the
480 standard choice of @code{/lib} as a prefix to try after the default prefix
481 when searching for startup files such as @file{crt0.o}.
482 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
483 is built as a cross compiler.
486 @defmac STANDARD_STARTFILE_PREFIX_2
487 Define this macro as a C string constant if you wish to override the
488 standard choice of @code{/lib} as yet another prefix to try after the
489 default prefix when searching for startup files such as @file{crt0.o}.
490 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
491 is built as a cross compiler.
494 @defmac MD_STARTFILE_PREFIX
495 If defined, this macro supplies an additional prefix to try after the
496 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
497 @option{-b} option is used, or when the compiler is built as a cross
501 @defmac MD_STARTFILE_PREFIX_1
502 If defined, this macro supplies yet another prefix to try after the
503 standard prefixes. It is not searched when the @option{-b} option is
504 used, or when the compiler is built as a cross compiler.
507 @defmac INIT_ENVIRONMENT
508 Define this macro as a C string constant if you wish to set environment
509 variables for programs called by the driver, such as the assembler and
510 loader. The driver passes the value of this macro to @code{putenv} to
511 initialize the necessary environment variables.
514 @defmac LOCAL_INCLUDE_DIR
515 Define this macro as a C string constant if you wish to override the
516 standard choice of @file{/usr/local/include} as the default prefix to
517 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
518 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
520 Cross compilers do not search either @file{/usr/local/include} or its
524 @defmac MODIFY_TARGET_NAME
525 Define this macro if you wish to define command-line switches that
526 modify the default target name.
528 For each switch, you can include a string to be appended to the first
529 part of the configuration name or a string to be deleted from the
530 configuration name, if present. The definition should be an initializer
531 for an array of structures. Each array element should have three
532 elements: the switch name (a string constant, including the initial
533 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
534 indicate whether the string should be inserted or deleted, and the string
535 to be inserted or deleted (a string constant).
537 For example, on a machine where @samp{64} at the end of the
538 configuration name denotes a 64-bit target and you want the @option{-32}
539 and @option{-64} switches to select between 32- and 64-bit targets, you would
543 #define MODIFY_TARGET_NAME \
544 @{ @{ "-32", DELETE, "64"@}, \
545 @{"-64", ADD, "64"@}@}
549 @defmac SYSTEM_INCLUDE_DIR
550 Define this macro as a C string constant if you wish to specify a
551 system-specific directory to search for header files before the standard
552 directory. @code{SYSTEM_INCLUDE_DIR} comes before
553 @code{STANDARD_INCLUDE_DIR} in the search order.
555 Cross compilers do not use this macro and do not search the directory
559 @defmac STANDARD_INCLUDE_DIR
560 Define this macro as a C string constant if you wish to override the
561 standard choice of @file{/usr/include} as the default prefix to
562 try when searching for header files.
564 Cross compilers ignore this macro and do not search either
565 @file{/usr/include} or its replacement.
568 @defmac STANDARD_INCLUDE_COMPONENT
569 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
570 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
571 If you do not define this macro, no component is used.
574 @defmac INCLUDE_DEFAULTS
575 Define this macro if you wish to override the entire default search path
576 for include files. For a native compiler, the default search path
577 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
578 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
579 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
580 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
581 and specify private search areas for GCC@. The directory
582 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
584 The definition should be an initializer for an array of structures.
585 Each array element should have four elements: the directory name (a
586 string constant), the component name (also a string constant), a flag
587 for C++-only directories,
588 and a flag showing that the includes in the directory don't need to be
589 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
590 the array with a null element.
592 The component name denotes what GNU package the include file is part of,
593 if any, in all uppercase letters. For example, it might be @samp{GCC}
594 or @samp{BINUTILS}. If the package is part of a vendor-supplied
595 operating system, code the component name as @samp{0}.
597 For example, here is the definition used for VAX/VMS:
600 #define INCLUDE_DEFAULTS \
602 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
603 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
604 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
611 Here is the order of prefixes tried for exec files:
615 Any prefixes specified by the user with @option{-B}.
618 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
619 is not set and the compiler has not been installed in the configure-time
620 @var{prefix}, the location in which the compiler has actually been installed.
623 The directories specified by the environment variable @code{COMPILER_PATH}.
626 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
627 in the configured-time @var{prefix}.
630 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
633 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
636 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
640 Here is the order of prefixes tried for startfiles:
644 Any prefixes specified by the user with @option{-B}.
647 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
648 value based on the installed toolchain location.
651 The directories specified by the environment variable @code{LIBRARY_PATH}
652 (or port-specific name; native only, cross compilers do not use this).
655 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
656 in the configured @var{prefix} or this is a native compiler.
659 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
662 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
666 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
667 native compiler, or we have a target system root.
670 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
671 native compiler, or we have a target system root.
674 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
675 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
676 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
679 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
680 compiler, or we have a target system root. The default for this macro is
684 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
685 compiler, or we have a target system root. The default for this macro is
689 @node Run-time Target
690 @section Run-time Target Specification
691 @cindex run-time target specification
692 @cindex predefined macros
693 @cindex target specifications
695 @c prevent bad page break with this line
696 Here are run-time target specifications.
698 @defmac TARGET_CPU_CPP_BUILTINS ()
699 This function-like macro expands to a block of code that defines
700 built-in preprocessor macros and assertions for the target CPU, using
701 the functions @code{builtin_define}, @code{builtin_define_std} and
702 @code{builtin_assert}. When the front end
703 calls this macro it provides a trailing semicolon, and since it has
704 finished command line option processing your code can use those
707 @code{builtin_assert} takes a string in the form you pass to the
708 command-line option @option{-A}, such as @code{cpu=mips}, and creates
709 the assertion. @code{builtin_define} takes a string in the form
710 accepted by option @option{-D} and unconditionally defines the macro.
712 @code{builtin_define_std} takes a string representing the name of an
713 object-like macro. If it doesn't lie in the user's namespace,
714 @code{builtin_define_std} defines it unconditionally. Otherwise, it
715 defines a version with two leading underscores, and another version
716 with two leading and trailing underscores, and defines the original
717 only if an ISO standard was not requested on the command line. For
718 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
719 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
720 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
721 defines only @code{_ABI64}.
723 You can also test for the C dialect being compiled. The variable
724 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
725 or @code{clk_objective_c}. Note that if we are preprocessing
726 assembler, this variable will be @code{clk_c} but the function-like
727 macro @code{preprocessing_asm_p()} will return true, so you might want
728 to check for that first. If you need to check for strict ANSI, the
729 variable @code{flag_iso} can be used. The function-like macro
730 @code{preprocessing_trad_p()} can be used to check for traditional
734 @defmac TARGET_OS_CPP_BUILTINS ()
735 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
736 and is used for the target operating system instead.
739 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
740 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
741 and is used for the target object format. @file{elfos.h} uses this
742 macro to define @code{__ELF__}, so you probably do not need to define
746 @deftypevar {extern int} target_flags
747 This variable is declared in @file{options.h}, which is included before
748 any target-specific headers.
751 @deftypevar {Target Hook} int TARGET_DEFAULT_TARGET_FLAGS
752 This variable specifies the initial value of @code{target_flags}.
753 Its default setting is 0.
756 @cindex optional hardware or system features
757 @cindex features, optional, in system conventions
759 @deftypefn {Target Hook} bool TARGET_HANDLE_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
760 This hook is called whenever the user specifies one of the
761 target-specific options described by the @file{.opt} definition files
762 (@pxref{Options}). It has the opportunity to do some option-specific
763 processing and should return true if the option is valid. The default
764 definition does nothing but return true.
766 @var{code} specifies the @code{OPT_@var{name}} enumeration value
767 associated with the selected option; @var{name} is just a rendering of
768 the option name in which non-alphanumeric characters are replaced by
769 underscores. @var{arg} specifies the string argument and is null if
770 no argument was given. If the option is flagged as a @code{UInteger}
771 (@pxref{Option properties}), @var{value} is the numeric value of the
772 argument. Otherwise @var{value} is 1 if the positive form of the
773 option was used and 0 if the ``no-'' form was.
776 @deftypefn {Target Hook} bool TARGET_HANDLE_C_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
777 This target hook is called whenever the user specifies one of the
778 target-specific C language family options described by the @file{.opt}
779 definition files(@pxref{Options}). It has the opportunity to do some
780 option-specific processing and should return true if the option is
781 valid. The default definition does nothing but return false.
783 In general, you should use @code{TARGET_HANDLE_OPTION} to handle
784 options. However, if processing an option requires routines that are
785 only available in the C (and related language) front ends, then you
786 should use @code{TARGET_HANDLE_C_OPTION} instead.
789 @defmac TARGET_VERSION
790 This macro is a C statement to print on @code{stderr} a string
791 describing the particular machine description choice. Every machine
792 description should define @code{TARGET_VERSION}. For example:
796 #define TARGET_VERSION \
797 fprintf (stderr, " (68k, Motorola syntax)");
799 #define TARGET_VERSION \
800 fprintf (stderr, " (68k, MIT syntax)");
805 @defmac OVERRIDE_OPTIONS
806 Sometimes certain combinations of command options do not make sense on
807 a particular target machine. You can define a macro
808 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
809 defined, is executed once just after all the command options have been
812 Don't use this macro to turn on various extra optimizations for
813 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
816 @defmac C_COMMON_OVERRIDE_OPTIONS
817 This is similar to @code{OVERRIDE_OPTIONS} but is only used in the C
818 language frontends (C, Objective-C, C++, Objective-C++) and so can be
819 used to alter option flag variables which only exist in those
823 @defmac OPTIMIZATION_OPTIONS (@var{level}, @var{size})
824 Some machines may desire to change what optimizations are performed for
825 various optimization levels. This macro, if defined, is executed once
826 just after the optimization level is determined and before the remainder
827 of the command options have been parsed. Values set in this macro are
828 used as the default values for the other command line options.
830 @var{level} is the optimization level specified; 2 if @option{-O2} is
831 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
833 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
835 You should not use this macro to change options that are not
836 machine-specific. These should uniformly selected by the same
837 optimization level on all supported machines. Use this macro to enable
838 machine-specific optimizations.
840 @strong{Do not examine @code{write_symbols} in
841 this macro!} The debugging options are not supposed to alter the
845 @deftypefn {Target Hook} bool TARGET_HELP (void)
846 This hook is called in response to the user invoking
847 @option{--target-help} on the command line. It gives the target a
848 chance to display extra information on the target specific command
849 line options found in its @file{.opt} file.
852 @defmac CAN_DEBUG_WITHOUT_FP
853 Define this macro if debugging can be performed even without a frame
854 pointer. If this macro is defined, GCC will turn on the
855 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
858 @node Per-Function Data
859 @section Defining data structures for per-function information.
860 @cindex per-function data
861 @cindex data structures
863 If the target needs to store information on a per-function basis, GCC
864 provides a macro and a couple of variables to allow this. Note, just
865 using statics to store the information is a bad idea, since GCC supports
866 nested functions, so you can be halfway through encoding one function
867 when another one comes along.
869 GCC defines a data structure called @code{struct function} which
870 contains all of the data specific to an individual function. This
871 structure contains a field called @code{machine} whose type is
872 @code{struct machine_function *}, which can be used by targets to point
873 to their own specific data.
875 If a target needs per-function specific data it should define the type
876 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
877 This macro should be used to initialize the function pointer
878 @code{init_machine_status}. This pointer is explained below.
880 One typical use of per-function, target specific data is to create an
881 RTX to hold the register containing the function's return address. This
882 RTX can then be used to implement the @code{__builtin_return_address}
883 function, for level 0.
885 Note---earlier implementations of GCC used a single data area to hold
886 all of the per-function information. Thus when processing of a nested
887 function began the old per-function data had to be pushed onto a
888 stack, and when the processing was finished, it had to be popped off the
889 stack. GCC used to provide function pointers called
890 @code{save_machine_status} and @code{restore_machine_status} to handle
891 the saving and restoring of the target specific information. Since the
892 single data area approach is no longer used, these pointers are no
895 @defmac INIT_EXPANDERS
896 Macro called to initialize any target specific information. This macro
897 is called once per function, before generation of any RTL has begun.
898 The intention of this macro is to allow the initialization of the
899 function pointer @code{init_machine_status}.
902 @deftypevar {void (*)(struct function *)} init_machine_status
903 If this function pointer is non-@code{NULL} it will be called once per
904 function, before function compilation starts, in order to allow the
905 target to perform any target specific initialization of the
906 @code{struct function} structure. It is intended that this would be
907 used to initialize the @code{machine} of that structure.
909 @code{struct machine_function} structures are expected to be freed by GC@.
910 Generally, any memory that they reference must be allocated by using
911 @code{ggc_alloc}, including the structure itself.
915 @section Storage Layout
916 @cindex storage layout
918 Note that the definitions of the macros in this table which are sizes or
919 alignments measured in bits do not need to be constant. They can be C
920 expressions that refer to static variables, such as the @code{target_flags}.
921 @xref{Run-time Target}.
923 @defmac BITS_BIG_ENDIAN
924 Define this macro to have the value 1 if the most significant bit in a
925 byte has the lowest number; otherwise define it to have the value zero.
926 This means that bit-field instructions count from the most significant
927 bit. If the machine has no bit-field instructions, then this must still
928 be defined, but it doesn't matter which value it is defined to. This
929 macro need not be a constant.
931 This macro does not affect the way structure fields are packed into
932 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
935 @defmac BYTES_BIG_ENDIAN
936 Define this macro to have the value 1 if the most significant byte in a
937 word has the lowest number. This macro need not be a constant.
940 @defmac WORDS_BIG_ENDIAN
941 Define this macro to have the value 1 if, in a multiword object, the
942 most significant word has the lowest number. This applies to both
943 memory locations and registers; GCC fundamentally assumes that the
944 order of words in memory is the same as the order in registers. This
945 macro need not be a constant.
948 @defmac LIBGCC2_WORDS_BIG_ENDIAN
949 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
950 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
951 used only when compiling @file{libgcc2.c}. Typically the value will be set
952 based on preprocessor defines.
955 @defmac FLOAT_WORDS_BIG_ENDIAN
956 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
957 @code{TFmode} floating point numbers are stored in memory with the word
958 containing the sign bit at the lowest address; otherwise define it to
959 have the value 0. This macro need not be a constant.
961 You need not define this macro if the ordering is the same as for
965 @defmac BITS_PER_UNIT
966 Define this macro to be the number of bits in an addressable storage
967 unit (byte). If you do not define this macro the default is 8.
970 @defmac BITS_PER_WORD
971 Number of bits in a word. If you do not define this macro, the default
972 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
975 @defmac MAX_BITS_PER_WORD
976 Maximum number of bits in a word. If this is undefined, the default is
977 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
978 largest value that @code{BITS_PER_WORD} can have at run-time.
981 @defmac UNITS_PER_WORD
982 Number of storage units in a word; normally the size of a general-purpose
983 register, a power of two from 1 or 8.
986 @defmac MIN_UNITS_PER_WORD
987 Minimum number of units in a word. If this is undefined, the default is
988 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
989 smallest value that @code{UNITS_PER_WORD} can have at run-time.
992 @defmac UNITS_PER_SIMD_WORD
993 Number of units in the vectors that the vectorizer can produce.
994 The default is equal to @code{UNITS_PER_WORD}, because the vectorizer
995 can do some transformations even in absence of specialized @acronym{SIMD}
1000 Width of a pointer, in bits. You must specify a value no wider than the
1001 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
1002 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
1003 a value the default is @code{BITS_PER_WORD}.
1006 @defmac POINTERS_EXTEND_UNSIGNED
1007 A C expression whose value is greater than zero if pointers that need to be
1008 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
1009 be zero-extended and zero if they are to be sign-extended. If the value
1010 is less then zero then there must be an "ptr_extend" instruction that
1011 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
1013 You need not define this macro if the @code{POINTER_SIZE} is equal
1014 to the width of @code{Pmode}.
1017 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
1018 A macro to update @var{m} and @var{unsignedp} when an object whose type
1019 is @var{type} and which has the specified mode and signedness is to be
1020 stored in a register. This macro is only called when @var{type} is a
1023 On most RISC machines, which only have operations that operate on a full
1024 register, define this macro to set @var{m} to @code{word_mode} if
1025 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
1026 cases, only integer modes should be widened because wider-precision
1027 floating-point operations are usually more expensive than their narrower
1030 For most machines, the macro definition does not change @var{unsignedp}.
1031 However, some machines, have instructions that preferentially handle
1032 either signed or unsigned quantities of certain modes. For example, on
1033 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
1034 sign-extend the result to 64 bits. On such machines, set
1035 @var{unsignedp} according to which kind of extension is more efficient.
1037 Do not define this macro if it would never modify @var{m}.
1040 @defmac PROMOTE_FUNCTION_MODE
1041 Like @code{PROMOTE_MODE}, but is applied to outgoing function arguments or
1042 function return values, as specified by @code{TARGET_PROMOTE_FUNCTION_ARGS}
1043 and @code{TARGET_PROMOTE_FUNCTION_RETURN}, respectively.
1045 The default is @code{PROMOTE_MODE}.
1048 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_ARGS (tree @var{fntype})
1049 This target hook should return @code{true} if the promotion described by
1050 @code{PROMOTE_FUNCTION_MODE} should be done for outgoing function
1054 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_RETURN (tree @var{fntype})
1055 This target hook should return @code{true} if the promotion described by
1056 @code{PROMOTE_FUNCTION_MODE} should be done for the return value of
1059 If this target hook returns @code{true}, @code{TARGET_FUNCTION_VALUE}
1060 must perform the same promotions done by @code{PROMOTE_FUNCTION_MODE}.
1063 @defmac PARM_BOUNDARY
1064 Normal alignment required for function parameters on the stack, in
1065 bits. All stack parameters receive at least this much alignment
1066 regardless of data type. On most machines, this is the same as the
1070 @defmac STACK_BOUNDARY
1071 Define this macro to the minimum alignment enforced by hardware for the
1072 stack pointer on this machine. The definition is a C expression for the
1073 desired alignment (measured in bits). This value is used as a default
1074 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1075 this should be the same as @code{PARM_BOUNDARY}.
1078 @defmac PREFERRED_STACK_BOUNDARY
1079 Define this macro if you wish to preserve a certain alignment for the
1080 stack pointer, greater than what the hardware enforces. The definition
1081 is a C expression for the desired alignment (measured in bits). This
1082 macro must evaluate to a value equal to or larger than
1083 @code{STACK_BOUNDARY}.
1086 @defmac FUNCTION_BOUNDARY
1087 Alignment required for a function entry point, in bits.
1090 @defmac BIGGEST_ALIGNMENT
1091 Biggest alignment that any data type can require on this machine, in
1092 bits. Note that this is not the biggest alignment that is supported,
1093 just the biggest alignment that, when violated, may cause a fault.
1096 @defmac MINIMUM_ATOMIC_ALIGNMENT
1097 If defined, the smallest alignment, in bits, that can be given to an
1098 object that can be referenced in one operation, without disturbing any
1099 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1100 on machines that don't have byte or half-word store operations.
1103 @defmac BIGGEST_FIELD_ALIGNMENT
1104 Biggest alignment that any structure or union field can require on this
1105 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1106 structure and union fields only, unless the field alignment has been set
1107 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1110 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1111 An expression for the alignment of a structure field @var{field} if the
1112 alignment computed in the usual way (including applying of
1113 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1114 alignment) is @var{computed}. It overrides alignment only if the
1115 field alignment has not been set by the
1116 @code{__attribute__ ((aligned (@var{n})))} construct.
1119 @defmac MAX_OFILE_ALIGNMENT
1120 Biggest alignment supported by the object file format of this machine.
1121 Use this macro to limit the alignment which can be specified using the
1122 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1123 the default value is @code{BIGGEST_ALIGNMENT}.
1125 On systems that use ELF, the default (in @file{config/elfos.h}) is
1126 the largest supported 32-bit ELF section alignment representable on
1127 a 32-bit host e.g. @samp{(((unsigned HOST_WIDEST_INT) 1 << 28) * 8)}.
1128 On 32-bit ELF the largest supported section alignment in bits is
1129 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1132 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1133 If defined, a C expression to compute the alignment for a variable in
1134 the static store. @var{type} is the data type, and @var{basic-align} is
1135 the alignment that the object would ordinarily have. The value of this
1136 macro is used instead of that alignment to align the object.
1138 If this macro is not defined, then @var{basic-align} is used.
1141 One use of this macro is to increase alignment of medium-size data to
1142 make it all fit in fewer cache lines. Another is to cause character
1143 arrays to be word-aligned so that @code{strcpy} calls that copy
1144 constants to character arrays can be done inline.
1147 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1148 If defined, a C expression to compute the alignment given to a constant
1149 that is being placed in memory. @var{constant} is the constant and
1150 @var{basic-align} is the alignment that the object would ordinarily
1151 have. The value of this macro is used instead of that alignment to
1154 If this macro is not defined, then @var{basic-align} is used.
1156 The typical use of this macro is to increase alignment for string
1157 constants to be word aligned so that @code{strcpy} calls that copy
1158 constants can be done inline.
1161 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1162 If defined, a C expression to compute the alignment for a variable in
1163 the local store. @var{type} is the data type, and @var{basic-align} is
1164 the alignment that the object would ordinarily have. The value of this
1165 macro is used instead of that alignment to align the object.
1167 If this macro is not defined, then @var{basic-align} is used.
1169 One use of this macro is to increase alignment of medium-size data to
1170 make it all fit in fewer cache lines.
1173 @defmac EMPTY_FIELD_BOUNDARY
1174 Alignment in bits to be given to a structure bit-field that follows an
1175 empty field such as @code{int : 0;}.
1177 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1180 @defmac STRUCTURE_SIZE_BOUNDARY
1181 Number of bits which any structure or union's size must be a multiple of.
1182 Each structure or union's size is rounded up to a multiple of this.
1184 If you do not define this macro, the default is the same as
1185 @code{BITS_PER_UNIT}.
1188 @defmac STRICT_ALIGNMENT
1189 Define this macro to be the value 1 if instructions will fail to work
1190 if given data not on the nominal alignment. If instructions will merely
1191 go slower in that case, define this macro as 0.
1194 @defmac PCC_BITFIELD_TYPE_MATTERS
1195 Define this if you wish to imitate the way many other C compilers handle
1196 alignment of bit-fields and the structures that contain them.
1198 The behavior is that the type written for a named bit-field (@code{int},
1199 @code{short}, or other integer type) imposes an alignment for the entire
1200 structure, as if the structure really did contain an ordinary field of
1201 that type. In addition, the bit-field is placed within the structure so
1202 that it would fit within such a field, not crossing a boundary for it.
1204 Thus, on most machines, a named bit-field whose type is written as
1205 @code{int} would not cross a four-byte boundary, and would force
1206 four-byte alignment for the whole structure. (The alignment used may
1207 not be four bytes; it is controlled by the other alignment parameters.)
1209 An unnamed bit-field will not affect the alignment of the containing
1212 If the macro is defined, its definition should be a C expression;
1213 a nonzero value for the expression enables this behavior.
1215 Note that if this macro is not defined, or its value is zero, some
1216 bit-fields may cross more than one alignment boundary. The compiler can
1217 support such references if there are @samp{insv}, @samp{extv}, and
1218 @samp{extzv} insns that can directly reference memory.
1220 The other known way of making bit-fields work is to define
1221 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1222 Then every structure can be accessed with fullwords.
1224 Unless the machine has bit-field instructions or you define
1225 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1226 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1228 If your aim is to make GCC use the same conventions for laying out
1229 bit-fields as are used by another compiler, here is how to investigate
1230 what the other compiler does. Compile and run this program:
1249 printf ("Size of foo1 is %d\n",
1250 sizeof (struct foo1));
1251 printf ("Size of foo2 is %d\n",
1252 sizeof (struct foo2));
1257 If this prints 2 and 5, then the compiler's behavior is what you would
1258 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1261 @defmac BITFIELD_NBYTES_LIMITED
1262 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1263 to aligning a bit-field within the structure.
1266 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELDS (void)
1267 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1268 whether unnamed bitfields affect the alignment of the containing
1269 structure. The hook should return true if the structure should inherit
1270 the alignment requirements of an unnamed bitfield's type.
1273 @deftypefn {Target Hook} bool TARGET_NARROW_VOLATILE_BITFIELDS (void)
1274 This target hook should return @code{true} if accesses to volatile bitfields
1275 should use the narrowest mode possible. It should return @code{false} if
1276 these accesses should use the bitfield container type.
1278 The default is @code{!TARGET_STRICT_ALIGN}.
1281 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1282 Return 1 if a structure or array containing @var{field} should be accessed using
1285 If @var{field} is the only field in the structure, @var{mode} is its
1286 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1287 case where structures of one field would require the structure's mode to
1288 retain the field's mode.
1290 Normally, this is not needed. See the file @file{c4x.h} for an example
1291 of how to use this macro to prevent a structure having a floating point
1292 field from being accessed in an integer mode.
1295 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1296 Define this macro as an expression for the alignment of a type (given
1297 by @var{type} as a tree node) if the alignment computed in the usual
1298 way is @var{computed} and the alignment explicitly specified was
1301 The default is to use @var{specified} if it is larger; otherwise, use
1302 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1305 @defmac MAX_FIXED_MODE_SIZE
1306 An integer expression for the size in bits of the largest integer
1307 machine mode that should actually be used. All integer machine modes of
1308 this size or smaller can be used for structures and unions with the
1309 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1310 (DImode)} is assumed.
1313 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1314 If defined, an expression of type @code{enum machine_mode} that
1315 specifies the mode of the save area operand of a
1316 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1317 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1318 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1319 having its mode specified.
1321 You need not define this macro if it always returns @code{Pmode}. You
1322 would most commonly define this macro if the
1323 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1327 @defmac STACK_SIZE_MODE
1328 If defined, an expression of type @code{enum machine_mode} that
1329 specifies the mode of the size increment operand of an
1330 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1332 You need not define this macro if it always returns @code{word_mode}.
1333 You would most commonly define this macro if the @code{allocate_stack}
1334 pattern needs to support both a 32- and a 64-bit mode.
1337 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_CMP_RETURN_MODE ()
1338 This target hook should return the mode to be used for the return value
1339 of compare instructions expanded to libgcc calls. If not defined
1340 @code{word_mode} is returned which is the right choice for a majority of
1344 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_SHIFT_COUNT_MODE ()
1345 This target hook should return the mode to be used for the shift count operand
1346 of shift instructions expanded to libgcc calls. If not defined
1347 @code{word_mode} is returned which is the right choice for a majority of
1351 @defmac TARGET_FLOAT_FORMAT
1352 A code distinguishing the floating point format of the target machine.
1353 There are four defined values:
1356 @item IEEE_FLOAT_FORMAT
1357 This code indicates IEEE floating point. It is the default; there is no
1358 need to define @code{TARGET_FLOAT_FORMAT} when the format is IEEE@.
1360 @item VAX_FLOAT_FORMAT
1361 This code indicates the ``F float'' (for @code{float}) and ``D float''
1362 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
1364 @item C4X_FLOAT_FORMAT
1365 This code indicates the format used on the TMS320C3x/C4x.
1368 If your target uses a floating point format other than these, you must
1369 define a new @var{name}_FLOAT_FORMAT code for it, and add support for
1370 it to @file{real.c}.
1372 The ordering of the component words of floating point values stored in
1373 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1376 @defmac MODE_HAS_NANS (@var{mode})
1377 When defined, this macro should be true if @var{mode} has a NaN
1378 representation. The compiler assumes that NaNs are not equal to
1379 anything (including themselves) and that addition, subtraction,
1380 multiplication and division all return NaNs when one operand is
1383 By default, this macro is true if @var{mode} is a floating-point
1384 mode and the target floating-point format is IEEE@.
1387 @defmac MODE_HAS_INFINITIES (@var{mode})
1388 This macro should be true if @var{mode} can represent infinity. At
1389 present, the compiler uses this macro to decide whether @samp{x - x}
1390 is always defined. By default, the macro is true when @var{mode}
1391 is a floating-point mode and the target format is IEEE@.
1394 @defmac MODE_HAS_SIGNED_ZEROS (@var{mode})
1395 True if @var{mode} distinguishes between positive and negative zero.
1396 The rules are expected to follow the IEEE standard:
1400 @samp{x + x} has the same sign as @samp{x}.
1403 If the sum of two values with opposite sign is zero, the result is
1404 positive for all rounding modes expect towards @minus{}infinity, for
1405 which it is negative.
1408 The sign of a product or quotient is negative when exactly one
1409 of the operands is negative.
1412 The default definition is true if @var{mode} is a floating-point
1413 mode and the target format is IEEE@.
1416 @defmac MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1417 If defined, this macro should be true for @var{mode} if it has at
1418 least one rounding mode in which @samp{x} and @samp{-x} can be
1419 rounded to numbers of different magnitude. Two such modes are
1420 towards @minus{}infinity and towards +infinity.
1422 The default definition of this macro is true if @var{mode} is
1423 a floating-point mode and the target format is IEEE@.
1426 @defmac ROUND_TOWARDS_ZERO
1427 If defined, this macro should be true if the prevailing rounding
1428 mode is towards zero. A true value has the following effects:
1432 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1435 @file{libgcc.a}'s floating-point emulator will round towards zero
1436 rather than towards nearest.
1439 The compiler's floating-point emulator will round towards zero after
1440 doing arithmetic, and when converting from the internal float format to
1444 The macro does not affect the parsing of string literals. When the
1445 primary rounding mode is towards zero, library functions like
1446 @code{strtod} might still round towards nearest, and the compiler's
1447 parser should behave like the target's @code{strtod} where possible.
1449 Not defining this macro is equivalent to returning zero.
1452 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1453 This macro should return true if floats with @var{size}
1454 bits do not have a NaN or infinity representation, but use the largest
1455 exponent for normal numbers instead.
1457 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1458 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1459 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1460 floating-point arithmetic.
1462 The default definition of this macro returns false for all sizes.
1465 @deftypefn {Target Hook} bool TARGET_VECTOR_OPAQUE_P (tree @var{type})
1466 This target hook should return @code{true} a vector is opaque. That
1467 is, if no cast is needed when copying a vector value of type
1468 @var{type} into another vector lvalue of the same size. Vector opaque
1469 types cannot be initialized. The default is that there are no such
1473 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1474 This target hook returns @code{true} if bit-fields in the given
1475 @var{record_type} are to be laid out following the rules of Microsoft
1476 Visual C/C++, namely: (i) a bit-field won't share the same storage
1477 unit with the previous bit-field if their underlying types have
1478 different sizes, and the bit-field will be aligned to the highest
1479 alignment of the underlying types of itself and of the previous
1480 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1481 the whole enclosing structure, even if it is unnamed; except that
1482 (iii) a zero-sized bit-field will be disregarded unless it follows
1483 another bit-field of nonzero size. If this hook returns @code{true},
1484 other macros that control bit-field layout are ignored.
1486 When a bit-field is inserted into a packed record, the whole size
1487 of the underlying type is used by one or more same-size adjacent
1488 bit-fields (that is, if its long:3, 32 bits is used in the record,
1489 and any additional adjacent long bit-fields are packed into the same
1490 chunk of 32 bits. However, if the size changes, a new field of that
1491 size is allocated). In an unpacked record, this is the same as using
1492 alignment, but not equivalent when packing.
1494 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1495 the latter will take precedence. If @samp{__attribute__((packed))} is
1496 used on a single field when MS bit-fields are in use, it will take
1497 precedence for that field, but the alignment of the rest of the structure
1498 may affect its placement.
1501 @deftypefn {Target Hook} {bool} TARGET_DECIMAL_FLOAT_SUPPORTED_P (void)
1502 Returns true if the target supports decimal floating point.
1505 @deftypefn {Target Hook} {bool} TARGET_FIXED_POINT_SUPPORTED_P (void)
1506 Returns true if the target supports fixed-point arithmetic.
1509 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_TYPE (tree @var{type})
1510 If your target defines any fundamental types, or any types your target
1511 uses should be mangled differently from the default, define this hook
1512 to return the appropriate encoding for these types as part of a C++
1513 mangled name. The @var{type} argument is the tree structure representing
1514 the type to be mangled. The hook may be applied to trees which are
1515 not target-specific fundamental types; it should return @code{NULL}
1516 for all such types, as well as arguments it does not recognize. If the
1517 return value is not @code{NULL}, it must point to a statically-allocated
1520 Target-specific fundamental types might be new fundamental types or
1521 qualified versions of ordinary fundamental types. Encode new
1522 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1523 is the name used for the type in source code, and @var{n} is the
1524 length of @var{name} in decimal. Encode qualified versions of
1525 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1526 @var{name} is the name used for the type qualifier in source code,
1527 @var{n} is the length of @var{name} as above, and @var{code} is the
1528 code used to represent the unqualified version of this type. (See
1529 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1530 codes.) In both cases the spaces are for clarity; do not include any
1531 spaces in your string.
1533 This hook is applied to types prior to typedef resolution. If the mangled
1534 name for a particular type depends only on that type's main variant, you
1535 can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT}
1538 The default version of this hook always returns @code{NULL}, which is
1539 appropriate for a target that does not define any new fundamental
1544 @section Layout of Source Language Data Types
1546 These macros define the sizes and other characteristics of the standard
1547 basic data types used in programs being compiled. Unlike the macros in
1548 the previous section, these apply to specific features of C and related
1549 languages, rather than to fundamental aspects of storage layout.
1551 @defmac INT_TYPE_SIZE
1552 A C expression for the size in bits of the type @code{int} on the
1553 target machine. If you don't define this, the default is one word.
1556 @defmac SHORT_TYPE_SIZE
1557 A C expression for the size in bits of the type @code{short} on the
1558 target machine. If you don't define this, the default is half a word.
1559 (If this would be less than one storage unit, it is rounded up to one
1563 @defmac LONG_TYPE_SIZE
1564 A C expression for the size in bits of the type @code{long} on the
1565 target machine. If you don't define this, the default is one word.
1568 @defmac ADA_LONG_TYPE_SIZE
1569 On some machines, the size used for the Ada equivalent of the type
1570 @code{long} by a native Ada compiler differs from that used by C@. In
1571 that situation, define this macro to be a C expression to be used for
1572 the size of that type. If you don't define this, the default is the
1573 value of @code{LONG_TYPE_SIZE}.
1576 @defmac LONG_LONG_TYPE_SIZE
1577 A C expression for the size in bits of the type @code{long long} on the
1578 target machine. If you don't define this, the default is two
1579 words. If you want to support GNU Ada on your machine, the value of this
1580 macro must be at least 64.
1583 @defmac CHAR_TYPE_SIZE
1584 A C expression for the size in bits of the type @code{char} on the
1585 target machine. If you don't define this, the default is
1586 @code{BITS_PER_UNIT}.
1589 @defmac BOOL_TYPE_SIZE
1590 A C expression for the size in bits of the C++ type @code{bool} and
1591 C99 type @code{_Bool} on the target machine. If you don't define
1592 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1595 @defmac FLOAT_TYPE_SIZE
1596 A C expression for the size in bits of the type @code{float} on the
1597 target machine. If you don't define this, the default is one word.
1600 @defmac DOUBLE_TYPE_SIZE
1601 A C expression for the size in bits of the type @code{double} on the
1602 target machine. If you don't define this, the default is two
1606 @defmac LONG_DOUBLE_TYPE_SIZE
1607 A C expression for the size in bits of the type @code{long double} on
1608 the target machine. If you don't define this, the default is two
1612 @defmac SHORT_FRACT_TYPE_SIZE
1613 A C expression for the size in bits of the type @code{short _Fract} on
1614 the target machine. If you don't define this, the default is
1615 @code{BITS_PER_UNIT}.
1618 @defmac FRACT_TYPE_SIZE
1619 A C expression for the size in bits of the type @code{_Fract} on
1620 the target machine. If you don't define this, the default is
1621 @code{BITS_PER_UNIT * 2}.
1624 @defmac LONG_FRACT_TYPE_SIZE
1625 A C expression for the size in bits of the type @code{long _Fract} on
1626 the target machine. If you don't define this, the default is
1627 @code{BITS_PER_UNIT * 4}.
1630 @defmac LONG_LONG_FRACT_TYPE_SIZE
1631 A C expression for the size in bits of the type @code{long long _Fract} on
1632 the target machine. If you don't define this, the default is
1633 @code{BITS_PER_UNIT * 8}.
1636 @defmac SHORT_ACCUM_TYPE_SIZE
1637 A C expression for the size in bits of the type @code{short _Accum} on
1638 the target machine. If you don't define this, the default is
1639 @code{BITS_PER_UNIT * 2}.
1642 @defmac ACCUM_TYPE_SIZE
1643 A C expression for the size in bits of the type @code{_Accum} on
1644 the target machine. If you don't define this, the default is
1645 @code{BITS_PER_UNIT * 4}.
1648 @defmac LONG_ACCUM_TYPE_SIZE
1649 A C expression for the size in bits of the type @code{long _Accum} on
1650 the target machine. If you don't define this, the default is
1651 @code{BITS_PER_UNIT * 8}.
1654 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1655 A C expression for the size in bits of the type @code{long long _Accum} on
1656 the target machine. If you don't define this, the default is
1657 @code{BITS_PER_UNIT * 16}.
1660 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1661 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1662 if you want routines in @file{libgcc2.a} for a size other than
1663 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1664 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1667 @defmac LIBGCC2_HAS_DF_MODE
1668 Define this macro if neither @code{LIBGCC2_DOUBLE_TYPE_SIZE} nor
1669 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1670 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1671 anyway. If you don't define this and either @code{LIBGCC2_DOUBLE_TYPE_SIZE}
1672 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1676 @defmac LIBGCC2_HAS_XF_MODE
1677 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1678 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1679 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1680 is 80 then the default is 1, otherwise it is 0.
1683 @defmac LIBGCC2_HAS_TF_MODE
1684 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1685 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1686 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1687 is 128 then the default is 1, otherwise it is 0.
1694 Define these macros to be the size in bits of the mantissa of
1695 @code{SFmode}, @code{DFmode}, @code{XFmode} and @code{TFmode} values,
1696 if the defaults in @file{libgcc2.h} are inappropriate. By default,
1697 @code{FLT_MANT_DIG} is used for @code{SF_SIZE}, @code{LDBL_MANT_DIG}
1698 for @code{XF_SIZE} and @code{TF_SIZE}, and @code{DBL_MANT_DIG} or
1699 @code{LDBL_MANT_DIG} for @code{DF_SIZE} according to whether
1700 @code{LIBGCC2_DOUBLE_TYPE_SIZE} or
1701 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64.
1704 @defmac TARGET_FLT_EVAL_METHOD
1705 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1706 assuming, if applicable, that the floating-point control word is in its
1707 default state. If you do not define this macro the value of
1708 @code{FLT_EVAL_METHOD} will be zero.
1711 @defmac WIDEST_HARDWARE_FP_SIZE
1712 A C expression for the size in bits of the widest floating-point format
1713 supported by the hardware. If you define this macro, you must specify a
1714 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1715 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1719 @defmac DEFAULT_SIGNED_CHAR
1720 An expression whose value is 1 or 0, according to whether the type
1721 @code{char} should be signed or unsigned by default. The user can
1722 always override this default with the options @option{-fsigned-char}
1723 and @option{-funsigned-char}.
1726 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1727 This target hook should return true if the compiler should give an
1728 @code{enum} type only as many bytes as it takes to represent the range
1729 of possible values of that type. It should return false if all
1730 @code{enum} types should be allocated like @code{int}.
1732 The default is to return false.
1736 A C expression for a string describing the name of the data type to use
1737 for size values. The typedef name @code{size_t} is defined using the
1738 contents of the string.
1740 The string can contain more than one keyword. If so, separate them with
1741 spaces, and write first any length keyword, then @code{unsigned} if
1742 appropriate, and finally @code{int}. The string must exactly match one
1743 of the data type names defined in the function
1744 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1745 omit @code{int} or change the order---that would cause the compiler to
1748 If you don't define this macro, the default is @code{"long unsigned
1752 @defmac PTRDIFF_TYPE
1753 A C expression for a string describing the name of the data type to use
1754 for the result of subtracting two pointers. The typedef name
1755 @code{ptrdiff_t} is defined using the contents of the string. See
1756 @code{SIZE_TYPE} above for more information.
1758 If you don't define this macro, the default is @code{"long int"}.
1762 A C expression for a string describing the name of the data type to use
1763 for wide characters. The typedef name @code{wchar_t} is defined using
1764 the contents of the string. See @code{SIZE_TYPE} above for more
1767 If you don't define this macro, the default is @code{"int"}.
1770 @defmac WCHAR_TYPE_SIZE
1771 A C expression for the size in bits of the data type for wide
1772 characters. This is used in @code{cpp}, which cannot make use of
1777 A C expression for a string describing the name of the data type to
1778 use for wide characters passed to @code{printf} and returned from
1779 @code{getwc}. The typedef name @code{wint_t} is defined using the
1780 contents of the string. See @code{SIZE_TYPE} above for more
1783 If you don't define this macro, the default is @code{"unsigned int"}.
1787 A C expression for a string describing the name of the data type that
1788 can represent any value of any standard or extended signed integer type.
1789 The typedef name @code{intmax_t} is defined using the contents of the
1790 string. See @code{SIZE_TYPE} above for more information.
1792 If you don't define this macro, the default is the first of
1793 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1794 much precision as @code{long long int}.
1797 @defmac UINTMAX_TYPE
1798 A C expression for a string describing the name of the data type that
1799 can represent any value of any standard or extended unsigned integer
1800 type. The typedef name @code{uintmax_t} is defined using the contents
1801 of the string. See @code{SIZE_TYPE} above for more information.
1803 If you don't define this macro, the default is the first of
1804 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1805 unsigned int"} that has as much precision as @code{long long unsigned
1809 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1810 The C++ compiler represents a pointer-to-member-function with a struct
1817 ptrdiff_t vtable_index;
1824 The C++ compiler must use one bit to indicate whether the function that
1825 will be called through a pointer-to-member-function is virtual.
1826 Normally, we assume that the low-order bit of a function pointer must
1827 always be zero. Then, by ensuring that the vtable_index is odd, we can
1828 distinguish which variant of the union is in use. But, on some
1829 platforms function pointers can be odd, and so this doesn't work. In
1830 that case, we use the low-order bit of the @code{delta} field, and shift
1831 the remainder of the @code{delta} field to the left.
1833 GCC will automatically make the right selection about where to store
1834 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1835 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1836 set such that functions always start at even addresses, but the lowest
1837 bit of pointers to functions indicate whether the function at that
1838 address is in ARM or Thumb mode. If this is the case of your
1839 architecture, you should define this macro to
1840 @code{ptrmemfunc_vbit_in_delta}.
1842 In general, you should not have to define this macro. On architectures
1843 in which function addresses are always even, according to
1844 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1845 @code{ptrmemfunc_vbit_in_pfn}.
1848 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1849 Normally, the C++ compiler uses function pointers in vtables. This
1850 macro allows the target to change to use ``function descriptors''
1851 instead. Function descriptors are found on targets for whom a
1852 function pointer is actually a small data structure. Normally the
1853 data structure consists of the actual code address plus a data
1854 pointer to which the function's data is relative.
1856 If vtables are used, the value of this macro should be the number
1857 of words that the function descriptor occupies.
1860 @defmac TARGET_VTABLE_ENTRY_ALIGN
1861 By default, the vtable entries are void pointers, the so the alignment
1862 is the same as pointer alignment. The value of this macro specifies
1863 the alignment of the vtable entry in bits. It should be defined only
1864 when special alignment is necessary. */
1867 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1868 There are a few non-descriptor entries in the vtable at offsets below
1869 zero. If these entries must be padded (say, to preserve the alignment
1870 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1871 of words in each data entry.
1875 @section Register Usage
1876 @cindex register usage
1878 This section explains how to describe what registers the target machine
1879 has, and how (in general) they can be used.
1881 The description of which registers a specific instruction can use is
1882 done with register classes; see @ref{Register Classes}. For information
1883 on using registers to access a stack frame, see @ref{Frame Registers}.
1884 For passing values in registers, see @ref{Register Arguments}.
1885 For returning values in registers, see @ref{Scalar Return}.
1888 * Register Basics:: Number and kinds of registers.
1889 * Allocation Order:: Order in which registers are allocated.
1890 * Values in Registers:: What kinds of values each reg can hold.
1891 * Leaf Functions:: Renumbering registers for leaf functions.
1892 * Stack Registers:: Handling a register stack such as 80387.
1895 @node Register Basics
1896 @subsection Basic Characteristics of Registers
1898 @c prevent bad page break with this line
1899 Registers have various characteristics.
1901 @defmac FIRST_PSEUDO_REGISTER
1902 Number of hardware registers known to the compiler. They receive
1903 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1904 pseudo register's number really is assigned the number
1905 @code{FIRST_PSEUDO_REGISTER}.
1908 @defmac FIXED_REGISTERS
1909 @cindex fixed register
1910 An initializer that says which registers are used for fixed purposes
1911 all throughout the compiled code and are therefore not available for
1912 general allocation. These would include the stack pointer, the frame
1913 pointer (except on machines where that can be used as a general
1914 register when no frame pointer is needed), the program counter on
1915 machines where that is considered one of the addressable registers,
1916 and any other numbered register with a standard use.
1918 This information is expressed as a sequence of numbers, separated by
1919 commas and surrounded by braces. The @var{n}th number is 1 if
1920 register @var{n} is fixed, 0 otherwise.
1922 The table initialized from this macro, and the table initialized by
1923 the following one, may be overridden at run time either automatically,
1924 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1925 the user with the command options @option{-ffixed-@var{reg}},
1926 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1929 @defmac CALL_USED_REGISTERS
1930 @cindex call-used register
1931 @cindex call-clobbered register
1932 @cindex call-saved register
1933 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1934 clobbered (in general) by function calls as well as for fixed
1935 registers. This macro therefore identifies the registers that are not
1936 available for general allocation of values that must live across
1939 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1940 automatically saves it on function entry and restores it on function
1941 exit, if the register is used within the function.
1944 @defmac CALL_REALLY_USED_REGISTERS
1945 @cindex call-used register
1946 @cindex call-clobbered register
1947 @cindex call-saved register
1948 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1949 that the entire set of @code{FIXED_REGISTERS} be included.
1950 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1951 This macro is optional. If not specified, it defaults to the value
1952 of @code{CALL_USED_REGISTERS}.
1955 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1956 @cindex call-used register
1957 @cindex call-clobbered register
1958 @cindex call-saved register
1959 A C expression that is nonzero if it is not permissible to store a
1960 value of mode @var{mode} in hard register number @var{regno} across a
1961 call without some part of it being clobbered. For most machines this
1962 macro need not be defined. It is only required for machines that do not
1963 preserve the entire contents of a register across a call.
1967 @findex call_used_regs
1970 @findex reg_class_contents
1971 @defmac CONDITIONAL_REGISTER_USAGE
1972 Zero or more C statements that may conditionally modify five variables
1973 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1974 @code{reg_names}, and @code{reg_class_contents}, to take into account
1975 any dependence of these register sets on target flags. The first three
1976 of these are of type @code{char []} (interpreted as Boolean vectors).
1977 @code{global_regs} is a @code{const char *[]}, and
1978 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1979 called, @code{fixed_regs}, @code{call_used_regs},
1980 @code{reg_class_contents}, and @code{reg_names} have been initialized
1981 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1982 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1983 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1984 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1985 command options have been applied.
1987 You need not define this macro if it has no work to do.
1989 @cindex disabling certain registers
1990 @cindex controlling register usage
1991 If the usage of an entire class of registers depends on the target
1992 flags, you may indicate this to GCC by using this macro to modify
1993 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1994 registers in the classes which should not be used by GCC@. Also define
1995 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1996 to return @code{NO_REGS} if it
1997 is called with a letter for a class that shouldn't be used.
1999 (However, if this class is not included in @code{GENERAL_REGS} and all
2000 of the insn patterns whose constraints permit this class are
2001 controlled by target switches, then GCC will automatically avoid using
2002 these registers when the target switches are opposed to them.)
2005 @defmac INCOMING_REGNO (@var{out})
2006 Define this macro if the target machine has register windows. This C
2007 expression returns the register number as seen by the called function
2008 corresponding to the register number @var{out} as seen by the calling
2009 function. Return @var{out} if register number @var{out} is not an
2013 @defmac OUTGOING_REGNO (@var{in})
2014 Define this macro if the target machine has register windows. This C
2015 expression returns the register number as seen by the calling function
2016 corresponding to the register number @var{in} as seen by the called
2017 function. Return @var{in} if register number @var{in} is not an inbound
2021 @defmac LOCAL_REGNO (@var{regno})
2022 Define this macro if the target machine has register windows. This C
2023 expression returns true if the register is call-saved but is in the
2024 register window. Unlike most call-saved registers, such registers
2025 need not be explicitly restored on function exit or during non-local
2030 If the program counter has a register number, define this as that
2031 register number. Otherwise, do not define it.
2034 @node Allocation Order
2035 @subsection Order of Allocation of Registers
2036 @cindex order of register allocation
2037 @cindex register allocation order
2039 @c prevent bad page break with this line
2040 Registers are allocated in order.
2042 @defmac REG_ALLOC_ORDER
2043 If defined, an initializer for a vector of integers, containing the
2044 numbers of hard registers in the order in which GCC should prefer
2045 to use them (from most preferred to least).
2047 If this macro is not defined, registers are used lowest numbered first
2048 (all else being equal).
2050 One use of this macro is on machines where the highest numbered
2051 registers must always be saved and the save-multiple-registers
2052 instruction supports only sequences of consecutive registers. On such
2053 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
2054 the highest numbered allocable register first.
2057 @defmac ORDER_REGS_FOR_LOCAL_ALLOC
2058 A C statement (sans semicolon) to choose the order in which to allocate
2059 hard registers for pseudo-registers local to a basic block.
2061 Store the desired register order in the array @code{reg_alloc_order}.
2062 Element 0 should be the register to allocate first; element 1, the next
2063 register; and so on.
2065 The macro body should not assume anything about the contents of
2066 @code{reg_alloc_order} before execution of the macro.
2068 On most machines, it is not necessary to define this macro.
2071 @node Values in Registers
2072 @subsection How Values Fit in Registers
2074 This section discusses the macros that describe which kinds of values
2075 (specifically, which machine modes) each register can hold, and how many
2076 consecutive registers are needed for a given mode.
2078 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2079 A C expression for the number of consecutive hard registers, starting
2080 at register number @var{regno}, required to hold a value of mode
2083 On a machine where all registers are exactly one word, a suitable
2084 definition of this macro is
2087 #define HARD_REGNO_NREGS(REGNO, MODE) \
2088 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2093 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2094 A C expression that is nonzero if a value of mode @var{mode}, stored
2095 in memory, ends with padding that causes it to take up more space than
2096 in registers starting at register number @var{regno} (as determined by
2097 multiplying GCC's notion of the size of the register when containing
2098 this mode by the number of registers returned by
2099 @code{HARD_REGNO_NREGS}). By default this is zero.
2101 For example, if a floating-point value is stored in three 32-bit
2102 registers but takes up 128 bits in memory, then this would be
2105 This macros only needs to be defined if there are cases where
2106 @code{subreg_get_info}
2107 would otherwise wrongly determine that a @code{subreg} can be
2108 represented by an offset to the register number, when in fact such a
2109 @code{subreg} would contain some of the padding not stored in
2110 registers and so not be representable.
2113 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2114 For values of @var{regno} and @var{mode} for which
2115 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2116 returning the greater number of registers required to hold the value
2117 including any padding. In the example above, the value would be four.
2120 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2121 Define this macro if the natural size of registers that hold values
2122 of mode @var{mode} is not the word size. It is a C expression that
2123 should give the natural size in bytes for the specified mode. It is
2124 used by the register allocator to try to optimize its results. This
2125 happens for example on SPARC 64-bit where the natural size of
2126 floating-point registers is still 32-bit.
2129 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2130 A C expression that is nonzero if it is permissible to store a value
2131 of mode @var{mode} in hard register number @var{regno} (or in several
2132 registers starting with that one). For a machine where all registers
2133 are equivalent, a suitable definition is
2136 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2139 You need not include code to check for the numbers of fixed registers,
2140 because the allocation mechanism considers them to be always occupied.
2142 @cindex register pairs
2143 On some machines, double-precision values must be kept in even/odd
2144 register pairs. You can implement that by defining this macro to reject
2145 odd register numbers for such modes.
2147 The minimum requirement for a mode to be OK in a register is that the
2148 @samp{mov@var{mode}} instruction pattern support moves between the
2149 register and other hard register in the same class and that moving a
2150 value into the register and back out not alter it.
2152 Since the same instruction used to move @code{word_mode} will work for
2153 all narrower integer modes, it is not necessary on any machine for
2154 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2155 you define patterns @samp{movhi}, etc., to take advantage of this. This
2156 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2157 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2160 Many machines have special registers for floating point arithmetic.
2161 Often people assume that floating point machine modes are allowed only
2162 in floating point registers. This is not true. Any registers that
2163 can hold integers can safely @emph{hold} a floating point machine
2164 mode, whether or not floating arithmetic can be done on it in those
2165 registers. Integer move instructions can be used to move the values.
2167 On some machines, though, the converse is true: fixed-point machine
2168 modes may not go in floating registers. This is true if the floating
2169 registers normalize any value stored in them, because storing a
2170 non-floating value there would garble it. In this case,
2171 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2172 floating registers. But if the floating registers do not automatically
2173 normalize, if you can store any bit pattern in one and retrieve it
2174 unchanged without a trap, then any machine mode may go in a floating
2175 register, so you can define this macro to say so.
2177 The primary significance of special floating registers is rather that
2178 they are the registers acceptable in floating point arithmetic
2179 instructions. However, this is of no concern to
2180 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2181 constraints for those instructions.
2183 On some machines, the floating registers are especially slow to access,
2184 so that it is better to store a value in a stack frame than in such a
2185 register if floating point arithmetic is not being done. As long as the
2186 floating registers are not in class @code{GENERAL_REGS}, they will not
2187 be used unless some pattern's constraint asks for one.
2190 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2191 A C expression that is nonzero if it is OK to rename a hard register
2192 @var{from} to another hard register @var{to}.
2194 One common use of this macro is to prevent renaming of a register to
2195 another register that is not saved by a prologue in an interrupt
2198 The default is always nonzero.
2201 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2202 A C expression that is nonzero if a value of mode
2203 @var{mode1} is accessible in mode @var{mode2} without copying.
2205 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2206 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2207 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2208 should be nonzero. If they differ for any @var{r}, you should define
2209 this macro to return zero unless some other mechanism ensures the
2210 accessibility of the value in a narrower mode.
2212 You should define this macro to return nonzero in as many cases as
2213 possible since doing so will allow GCC to perform better register
2217 @defmac AVOID_CCMODE_COPIES
2218 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2219 registers. You should only define this macro if support for copying to/from
2220 @code{CCmode} is incomplete.
2223 @node Leaf Functions
2224 @subsection Handling Leaf Functions
2226 @cindex leaf functions
2227 @cindex functions, leaf
2228 On some machines, a leaf function (i.e., one which makes no calls) can run
2229 more efficiently if it does not make its own register window. Often this
2230 means it is required to receive its arguments in the registers where they
2231 are passed by the caller, instead of the registers where they would
2234 The special treatment for leaf functions generally applies only when
2235 other conditions are met; for example, often they may use only those
2236 registers for its own variables and temporaries. We use the term ``leaf
2237 function'' to mean a function that is suitable for this special
2238 handling, so that functions with no calls are not necessarily ``leaf
2241 GCC assigns register numbers before it knows whether the function is
2242 suitable for leaf function treatment. So it needs to renumber the
2243 registers in order to output a leaf function. The following macros
2246 @defmac LEAF_REGISTERS
2247 Name of a char vector, indexed by hard register number, which
2248 contains 1 for a register that is allowable in a candidate for leaf
2251 If leaf function treatment involves renumbering the registers, then the
2252 registers marked here should be the ones before renumbering---those that
2253 GCC would ordinarily allocate. The registers which will actually be
2254 used in the assembler code, after renumbering, should not be marked with 1
2257 Define this macro only if the target machine offers a way to optimize
2258 the treatment of leaf functions.
2261 @defmac LEAF_REG_REMAP (@var{regno})
2262 A C expression whose value is the register number to which @var{regno}
2263 should be renumbered, when a function is treated as a leaf function.
2265 If @var{regno} is a register number which should not appear in a leaf
2266 function before renumbering, then the expression should yield @minus{}1, which
2267 will cause the compiler to abort.
2269 Define this macro only if the target machine offers a way to optimize the
2270 treatment of leaf functions, and registers need to be renumbered to do
2274 @findex current_function_is_leaf
2275 @findex current_function_uses_only_leaf_regs
2276 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2277 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2278 specially. They can test the C variable @code{current_function_is_leaf}
2279 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2280 set prior to local register allocation and is valid for the remaining
2281 compiler passes. They can also test the C variable
2282 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2283 functions which only use leaf registers.
2284 @code{current_function_uses_only_leaf_regs} is valid after all passes
2285 that modify the instructions have been run and is only useful if
2286 @code{LEAF_REGISTERS} is defined.
2287 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2288 @c of the next paragraph?! --mew 2feb93
2290 @node Stack Registers
2291 @subsection Registers That Form a Stack
2293 There are special features to handle computers where some of the
2294 ``registers'' form a stack. Stack registers are normally written by
2295 pushing onto the stack, and are numbered relative to the top of the
2298 Currently, GCC can only handle one group of stack-like registers, and
2299 they must be consecutively numbered. Furthermore, the existing
2300 support for stack-like registers is specific to the 80387 floating
2301 point coprocessor. If you have a new architecture that uses
2302 stack-like registers, you will need to do substantial work on
2303 @file{reg-stack.c} and write your machine description to cooperate
2304 with it, as well as defining these macros.
2307 Define this if the machine has any stack-like registers.
2310 @defmac FIRST_STACK_REG
2311 The number of the first stack-like register. This one is the top
2315 @defmac LAST_STACK_REG
2316 The number of the last stack-like register. This one is the bottom of
2320 @node Register Classes
2321 @section Register Classes
2322 @cindex register class definitions
2323 @cindex class definitions, register
2325 On many machines, the numbered registers are not all equivalent.
2326 For example, certain registers may not be allowed for indexed addressing;
2327 certain registers may not be allowed in some instructions. These machine
2328 restrictions are described to the compiler using @dfn{register classes}.
2330 You define a number of register classes, giving each one a name and saying
2331 which of the registers belong to it. Then you can specify register classes
2332 that are allowed as operands to particular instruction patterns.
2336 In general, each register will belong to several classes. In fact, one
2337 class must be named @code{ALL_REGS} and contain all the registers. Another
2338 class must be named @code{NO_REGS} and contain no registers. Often the
2339 union of two classes will be another class; however, this is not required.
2341 @findex GENERAL_REGS
2342 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2343 terribly special about the name, but the operand constraint letters
2344 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2345 the same as @code{ALL_REGS}, just define it as a macro which expands
2348 Order the classes so that if class @var{x} is contained in class @var{y}
2349 then @var{x} has a lower class number than @var{y}.
2351 The way classes other than @code{GENERAL_REGS} are specified in operand
2352 constraints is through machine-dependent operand constraint letters.
2353 You can define such letters to correspond to various classes, then use
2354 them in operand constraints.
2356 You should define a class for the union of two classes whenever some
2357 instruction allows both classes. For example, if an instruction allows
2358 either a floating point (coprocessor) register or a general register for a
2359 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2360 which includes both of them. Otherwise you will get suboptimal code.
2362 You must also specify certain redundant information about the register
2363 classes: for each class, which classes contain it and which ones are
2364 contained in it; for each pair of classes, the largest class contained
2367 When a value occupying several consecutive registers is expected in a
2368 certain class, all the registers used must belong to that class.
2369 Therefore, register classes cannot be used to enforce a requirement for
2370 a register pair to start with an even-numbered register. The way to
2371 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2373 Register classes used for input-operands of bitwise-and or shift
2374 instructions have a special requirement: each such class must have, for
2375 each fixed-point machine mode, a subclass whose registers can transfer that
2376 mode to or from memory. For example, on some machines, the operations for
2377 single-byte values (@code{QImode}) are limited to certain registers. When
2378 this is so, each register class that is used in a bitwise-and or shift
2379 instruction must have a subclass consisting of registers from which
2380 single-byte values can be loaded or stored. This is so that
2381 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2383 @deftp {Data type} {enum reg_class}
2384 An enumerated type that must be defined with all the register class names
2385 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2386 must be the last register class, followed by one more enumerated value,
2387 @code{LIM_REG_CLASSES}, which is not a register class but rather
2388 tells how many classes there are.
2390 Each register class has a number, which is the value of casting
2391 the class name to type @code{int}. The number serves as an index
2392 in many of the tables described below.
2395 @defmac N_REG_CLASSES
2396 The number of distinct register classes, defined as follows:
2399 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2403 @defmac REG_CLASS_NAMES
2404 An initializer containing the names of the register classes as C string
2405 constants. These names are used in writing some of the debugging dumps.
2408 @defmac REG_CLASS_CONTENTS
2409 An initializer containing the contents of the register classes, as integers
2410 which are bit masks. The @var{n}th integer specifies the contents of class
2411 @var{n}. The way the integer @var{mask} is interpreted is that
2412 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2414 When the machine has more than 32 registers, an integer does not suffice.
2415 Then the integers are replaced by sub-initializers, braced groupings containing
2416 several integers. Each sub-initializer must be suitable as an initializer
2417 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2418 In this situation, the first integer in each sub-initializer corresponds to
2419 registers 0 through 31, the second integer to registers 32 through 63, and
2423 @defmac REGNO_REG_CLASS (@var{regno})
2424 A C expression whose value is a register class containing hard register
2425 @var{regno}. In general there is more than one such class; choose a class
2426 which is @dfn{minimal}, meaning that no smaller class also contains the
2430 @defmac BASE_REG_CLASS
2431 A macro whose definition is the name of the class to which a valid
2432 base register must belong. A base register is one used in an address
2433 which is the register value plus a displacement.
2436 @defmac MODE_BASE_REG_CLASS (@var{mode})
2437 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2438 the selection of a base register in a mode dependent manner. If
2439 @var{mode} is VOIDmode then it should return the same value as
2440 @code{BASE_REG_CLASS}.
2443 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2444 A C expression whose value is the register class to which a valid
2445 base register must belong in order to be used in a base plus index
2446 register address. You should define this macro if base plus index
2447 addresses have different requirements than other base register uses.
2450 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{outer_code}, @var{index_code})
2451 A C expression whose value is the register class to which a valid
2452 base register must belong. @var{outer_code} and @var{index_code} define the
2453 context in which the base register occurs. @var{outer_code} is the code of
2454 the immediately enclosing expression (@code{MEM} for the top level of an
2455 address, @code{ADDRESS} for something that occurs in an
2456 @code{address_operand}). @var{index_code} is the code of the corresponding
2457 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2460 @defmac INDEX_REG_CLASS
2461 A macro whose definition is the name of the class to which a valid
2462 index register must belong. An index register is one used in an
2463 address where its value is either multiplied by a scale factor or
2464 added to another register (as well as added to a displacement).
2467 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2468 A C expression which is nonzero if register number @var{num} is
2469 suitable for use as a base register in operand addresses. It may be
2470 either a suitable hard register or a pseudo register that has been
2471 allocated such a hard register.
2474 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2475 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2476 that expression may examine the mode of the memory reference in
2477 @var{mode}. You should define this macro if the mode of the memory
2478 reference affects whether a register may be used as a base register. If
2479 you define this macro, the compiler will use it instead of
2480 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2481 addresses that appear outside a @code{MEM}, i.e., as an
2482 @code{address_operand}.
2486 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2487 A C expression which is nonzero if register number @var{num} is suitable for
2488 use as a base register in base plus index operand addresses, accessing
2489 memory in mode @var{mode}. It may be either a suitable hard register or a
2490 pseudo register that has been allocated such a hard register. You should
2491 define this macro if base plus index addresses have different requirements
2492 than other base register uses.
2494 Use of this macro is deprecated; please use the more general
2495 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2498 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{outer_code}, @var{index_code})
2499 A C expression that is just like @code{REGNO_MODE_OK_FOR_BASE_P}, except
2500 that that expression may examine the context in which the register
2501 appears in the memory reference. @var{outer_code} is the code of the
2502 immediately enclosing expression (@code{MEM} if at the top level of the
2503 address, @code{ADDRESS} for something that occurs in an
2504 @code{address_operand}). @var{index_code} is the code of the
2505 corresponding index expression if @var{outer_code} is @code{PLUS};
2506 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2507 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2510 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2511 A C expression which is nonzero if register number @var{num} is
2512 suitable for use as an index register in operand addresses. It may be
2513 either a suitable hard register or a pseudo register that has been
2514 allocated such a hard register.
2516 The difference between an index register and a base register is that
2517 the index register may be scaled. If an address involves the sum of
2518 two registers, neither one of them scaled, then either one may be
2519 labeled the ``base'' and the other the ``index''; but whichever
2520 labeling is used must fit the machine's constraints of which registers
2521 may serve in each capacity. The compiler will try both labelings,
2522 looking for one that is valid, and will reload one or both registers
2523 only if neither labeling works.
2526 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2527 A C expression that places additional restrictions on the register class
2528 to use when it is necessary to copy value @var{x} into a register in class
2529 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2530 another, smaller class. On many machines, the following definition is
2534 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2537 Sometimes returning a more restrictive class makes better code. For
2538 example, on the 68000, when @var{x} is an integer constant that is in range
2539 for a @samp{moveq} instruction, the value of this macro is always
2540 @code{DATA_REGS} as long as @var{class} includes the data registers.
2541 Requiring a data register guarantees that a @samp{moveq} will be used.
2543 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2544 @var{class} is if @var{x} is a legitimate constant which cannot be
2545 loaded into some register class. By returning @code{NO_REGS} you can
2546 force @var{x} into a memory location. For example, rs6000 can load
2547 immediate values into general-purpose registers, but does not have an
2548 instruction for loading an immediate value into a floating-point
2549 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2550 @var{x} is a floating-point constant. If the constant can't be loaded
2551 into any kind of register, code generation will be better if
2552 @code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2553 of using @code{PREFERRED_RELOAD_CLASS}.
2555 If an insn has pseudos in it after register allocation, reload will go
2556 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2557 to find the best one. Returning @code{NO_REGS}, in this case, makes
2558 reload add a @code{!} in front of the constraint: the x86 back-end uses
2559 this feature to discourage usage of 387 registers when math is done in
2560 the SSE registers (and vice versa).
2563 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2564 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2565 input reloads. If you don't define this macro, the default is to use
2566 @var{class}, unchanged.
2568 You can also use @code{PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2569 reload from using some alternatives, like @code{PREFERRED_RELOAD_CLASS}.
2572 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2573 A C expression that places additional restrictions on the register class
2574 to use when it is necessary to be able to hold a value of mode
2575 @var{mode} in a reload register for which class @var{class} would
2578 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2579 there are certain modes that simply can't go in certain reload classes.
2581 The value is a register class; perhaps @var{class}, or perhaps another,
2584 Don't define this macro unless the target machine has limitations which
2585 require the macro to do something nontrivial.
2588 @deftypefn {Target Hook} {enum reg_class} TARGET_SECONDARY_RELOAD (bool @var{in_p}, rtx @var{x}, enum reg_class @var{reload_class}, enum machine_mode @var{reload_mode}, secondary_reload_info *@var{sri})
2589 Many machines have some registers that cannot be copied directly to or
2590 from memory or even from other types of registers. An example is the
2591 @samp{MQ} register, which on most machines, can only be copied to or
2592 from general registers, but not memory. Below, we shall be using the
2593 term 'intermediate register' when a move operation cannot be performed
2594 directly, but has to be done by copying the source into the intermediate
2595 register first, and then copying the intermediate register to the
2596 destination. An intermediate register always has the same mode as
2597 source and destination. Since it holds the actual value being copied,
2598 reload might apply optimizations to re-use an intermediate register
2599 and eliding the copy from the source when it can determine that the
2600 intermediate register still holds the required value.
2602 Another kind of secondary reload is required on some machines which
2603 allow copying all registers to and from memory, but require a scratch
2604 register for stores to some memory locations (e.g., those with symbolic
2605 address on the RT, and those with certain symbolic address on the SPARC
2606 when compiling PIC)@. Scratch registers need not have the same mode
2607 as the value being copied, and usually hold a different value that
2608 that being copied. Special patterns in the md file are needed to
2609 describe how the copy is performed with the help of the scratch register;
2610 these patterns also describe the number, register class(es) and mode(s)
2611 of the scratch register(s).
2613 In some cases, both an intermediate and a scratch register are required.
2615 For input reloads, this target hook is called with nonzero @var{in_p},
2616 and @var{x} is an rtx that needs to be copied to a register in of class
2617 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2618 hook is called with zero @var{in_p}, and a register of class @var{reload_mode}
2619 needs to be copied to rtx @var{x} in @var{reload_mode}.
2621 If copying a register of @var{reload_class} from/to @var{x} requires
2622 an intermediate register, the hook @code{secondary_reload} should
2623 return the register class required for this intermediate register.
2624 If no intermediate register is required, it should return NO_REGS.
2625 If more than one intermediate register is required, describe the one
2626 that is closest in the copy chain to the reload register.
2628 If scratch registers are needed, you also have to describe how to
2629 perform the copy from/to the reload register to/from this
2630 closest intermediate register. Or if no intermediate register is
2631 required, but still a scratch register is needed, describe the
2632 copy from/to the reload register to/from the reload operand @var{x}.
2634 You do this by setting @code{sri->icode} to the instruction code of a pattern
2635 in the md file which performs the move. Operands 0 and 1 are the output
2636 and input of this copy, respectively. Operands from operand 2 onward are
2637 for scratch operands. These scratch operands must have a mode, and a
2638 single-register-class
2639 @c [later: or memory]
2642 When an intermediate register is used, the @code{secondary_reload}
2643 hook will be called again to determine how to copy the intermediate
2644 register to/from the reload operand @var{x}, so your hook must also
2645 have code to handle the register class of the intermediate operand.
2647 @c [For later: maybe we'll allow multi-alternative reload patterns -
2648 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2649 @c and match the constraints of input and output to determine the required
2650 @c alternative. A restriction would be that constraints used to match
2651 @c against reloads registers would have to be written as register class
2652 @c constraints, or we need a new target macro / hook that tells us if an
2653 @c arbitrary constraint can match an unknown register of a given class.
2654 @c Such a macro / hook would also be useful in other places.]
2657 @var{x} might be a pseudo-register or a @code{subreg} of a
2658 pseudo-register, which could either be in a hard register or in memory.
2659 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2660 in memory and the hard register number if it is in a register.
2662 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2663 currently not supported. For the time being, you will have to continue
2664 to use @code{SECONDARY_MEMORY_NEEDED} for that purpose.
2666 @code{copy_cost} also uses this target hook to find out how values are
2667 copied. If you want it to include some extra cost for the need to allocate
2668 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2669 Or if two dependent moves are supposed to have a lower cost than the sum
2670 of the individual moves due to expected fortuitous scheduling and/or special
2671 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2674 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2675 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2676 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2677 These macros are obsolete, new ports should use the target hook
2678 @code{TARGET_SECONDARY_RELOAD} instead.
2680 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2681 target hook. Older ports still define these macros to indicate to the
2682 reload phase that it may
2683 need to allocate at least one register for a reload in addition to the
2684 register to contain the data. Specifically, if copying @var{x} to a
2685 register @var{class} in @var{mode} requires an intermediate register,
2686 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2687 largest register class all of whose registers can be used as
2688 intermediate registers or scratch registers.
2690 If copying a register @var{class} in @var{mode} to @var{x} requires an
2691 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2692 was supposed to be defined be defined to return the largest register
2693 class required. If the
2694 requirements for input and output reloads were the same, the macro
2695 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2698 The values returned by these macros are often @code{GENERAL_REGS}.
2699 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2700 can be directly copied to or from a register of @var{class} in
2701 @var{mode} without requiring a scratch register. Do not define this
2702 macro if it would always return @code{NO_REGS}.
2704 If a scratch register is required (either with or without an
2705 intermediate register), you were supposed to define patterns for
2706 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2707 (@pxref{Standard Names}. These patterns, which were normally
2708 implemented with a @code{define_expand}, should be similar to the
2709 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2712 These patterns need constraints for the reload register and scratch
2714 contain a single register class. If the original reload register (whose
2715 class is @var{class}) can meet the constraint given in the pattern, the
2716 value returned by these macros is used for the class of the scratch
2717 register. Otherwise, two additional reload registers are required.
2718 Their classes are obtained from the constraints in the insn pattern.
2720 @var{x} might be a pseudo-register or a @code{subreg} of a
2721 pseudo-register, which could either be in a hard register or in memory.
2722 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2723 in memory and the hard register number if it is in a register.
2725 These macros should not be used in the case where a particular class of
2726 registers can only be copied to memory and not to another class of
2727 registers. In that case, secondary reload registers are not needed and
2728 would not be helpful. Instead, a stack location must be used to perform
2729 the copy and the @code{mov@var{m}} pattern should use memory as an
2730 intermediate storage. This case often occurs between floating-point and
2734 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2735 Certain machines have the property that some registers cannot be copied
2736 to some other registers without using memory. Define this macro on
2737 those machines to be a C expression that is nonzero if objects of mode
2738 @var{m} in registers of @var{class1} can only be copied to registers of
2739 class @var{class2} by storing a register of @var{class1} into memory
2740 and loading that memory location into a register of @var{class2}.
2742 Do not define this macro if its value would always be zero.
2745 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2746 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2747 allocates a stack slot for a memory location needed for register copies.
2748 If this macro is defined, the compiler instead uses the memory location
2749 defined by this macro.
2751 Do not define this macro if you do not define
2752 @code{SECONDARY_MEMORY_NEEDED}.
2755 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2756 When the compiler needs a secondary memory location to copy between two
2757 registers of mode @var{mode}, it normally allocates sufficient memory to
2758 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2759 load operations in a mode that many bits wide and whose class is the
2760 same as that of @var{mode}.
2762 This is right thing to do on most machines because it ensures that all
2763 bits of the register are copied and prevents accesses to the registers
2764 in a narrower mode, which some machines prohibit for floating-point
2767 However, this default behavior is not correct on some machines, such as
2768 the DEC Alpha, that store short integers in floating-point registers
2769 differently than in integer registers. On those machines, the default
2770 widening will not work correctly and you must define this macro to
2771 suppress that widening in some cases. See the file @file{alpha.h} for
2774 Do not define this macro if you do not define
2775 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2776 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2779 @defmac SMALL_REGISTER_CLASSES
2780 On some machines, it is risky to let hard registers live across arbitrary
2781 insns. Typically, these machines have instructions that require values
2782 to be in specific registers (like an accumulator), and reload will fail
2783 if the required hard register is used for another purpose across such an
2786 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2787 value on these machines. When this macro has a nonzero value, the
2788 compiler will try to minimize the lifetime of hard registers.
2790 It is always safe to define this macro with a nonzero value, but if you
2791 unnecessarily define it, you will reduce the amount of optimizations
2792 that can be performed in some cases. If you do not define this macro
2793 with a nonzero value when it is required, the compiler will run out of
2794 spill registers and print a fatal error message. For most machines, you
2795 should not define this macro at all.
2798 @defmac CLASS_LIKELY_SPILLED_P (@var{class})
2799 A C expression whose value is nonzero if pseudos that have been assigned
2800 to registers of class @var{class} would likely be spilled because
2801 registers of @var{class} are needed for spill registers.
2803 The default value of this macro returns 1 if @var{class} has exactly one
2804 register and zero otherwise. On most machines, this default should be
2805 used. Only define this macro to some other expression if pseudos
2806 allocated by @file{local-alloc.c} end up in memory because their hard
2807 registers were needed for spill registers. If this macro returns nonzero
2808 for those classes, those pseudos will only be allocated by
2809 @file{global.c}, which knows how to reallocate the pseudo to another
2810 register. If there would not be another register available for
2811 reallocation, you should not change the definition of this macro since
2812 the only effect of such a definition would be to slow down register
2816 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2817 A C expression for the maximum number of consecutive registers
2818 of class @var{class} needed to hold a value of mode @var{mode}.
2820 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2821 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2822 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2823 @var{mode})} for all @var{regno} values in the class @var{class}.
2825 This macro helps control the handling of multiple-word values
2829 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2830 If defined, a C expression that returns nonzero for a @var{class} for which
2831 a change from mode @var{from} to mode @var{to} is invalid.
2833 For the example, loading 32-bit integer or floating-point objects into
2834 floating-point registers on the Alpha extends them to 64 bits.
2835 Therefore loading a 64-bit object and then storing it as a 32-bit object
2836 does not store the low-order 32 bits, as would be the case for a normal
2837 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2841 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2842 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2843 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2847 @node Old Constraints
2848 @section Obsolete Macros for Defining Constraints
2849 @cindex defining constraints, obsolete method
2850 @cindex constraints, defining, obsolete method
2852 Machine-specific constraints can be defined with these macros instead
2853 of the machine description constructs described in @ref{Define
2854 Constraints}. This mechanism is obsolete. New ports should not use
2855 it; old ports should convert to the new mechanism.
2857 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2858 For the constraint at the start of @var{str}, which starts with the letter
2859 @var{c}, return the length. This allows you to have register class /
2860 constant / extra constraints that are longer than a single letter;
2861 you don't need to define this macro if you can do with single-letter
2862 constraints only. The definition of this macro should use
2863 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2864 to handle specially.
2865 There are some sanity checks in genoutput.c that check the constraint lengths
2866 for the md file, so you can also use this macro to help you while you are
2867 transitioning from a byzantine single-letter-constraint scheme: when you
2868 return a negative length for a constraint you want to re-use, genoutput
2869 will complain about every instance where it is used in the md file.
2872 @defmac REG_CLASS_FROM_LETTER (@var{char})
2873 A C expression which defines the machine-dependent operand constraint
2874 letters for register classes. If @var{char} is such a letter, the
2875 value should be the register class corresponding to it. Otherwise,
2876 the value should be @code{NO_REGS}. The register letter @samp{r},
2877 corresponding to class @code{GENERAL_REGS}, will not be passed
2878 to this macro; you do not need to handle it.
2881 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2882 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2883 passed in @var{str}, so that you can use suffixes to distinguish between
2887 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2888 A C expression that defines the machine-dependent operand constraint
2889 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2890 particular ranges of integer values. If @var{c} is one of those
2891 letters, the expression should check that @var{value}, an integer, is in
2892 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2893 not one of those letters, the value should be 0 regardless of
2897 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2898 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2899 string passed in @var{str}, so that you can use suffixes to distinguish
2900 between different variants.
2903 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2904 A C expression that defines the machine-dependent operand constraint
2905 letters that specify particular ranges of @code{const_double} values
2906 (@samp{G} or @samp{H}).
2908 If @var{c} is one of those letters, the expression should check that
2909 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2910 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2911 letters, the value should be 0 regardless of @var{value}.
2913 @code{const_double} is used for all floating-point constants and for
2914 @code{DImode} fixed-point constants. A given letter can accept either
2915 or both kinds of values. It can use @code{GET_MODE} to distinguish
2916 between these kinds.
2919 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2920 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2921 string passed in @var{str}, so that you can use suffixes to distinguish
2922 between different variants.
2925 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2926 A C expression that defines the optional machine-dependent constraint
2927 letters that can be used to segregate specific types of operands, usually
2928 memory references, for the target machine. Any letter that is not
2929 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2930 @code{REG_CLASS_FROM_CONSTRAINT}
2931 may be used. Normally this macro will not be defined.
2933 If it is required for a particular target machine, it should return 1
2934 if @var{value} corresponds to the operand type represented by the
2935 constraint letter @var{c}. If @var{c} is not defined as an extra
2936 constraint, the value returned should be 0 regardless of @var{value}.
2938 For example, on the ROMP, load instructions cannot have their output
2939 in r0 if the memory reference contains a symbolic address. Constraint
2940 letter @samp{Q} is defined as representing a memory address that does
2941 @emph{not} contain a symbolic address. An alternative is specified with
2942 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2943 alternative specifies @samp{m} on the input and a register class that
2944 does not include r0 on the output.
2947 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2948 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2949 in @var{str}, so that you can use suffixes to distinguish between different
2953 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2954 A C expression that defines the optional machine-dependent constraint
2955 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2956 be treated like memory constraints by the reload pass.
2958 It should return 1 if the operand type represented by the constraint
2959 at the start of @var{str}, the first letter of which is the letter @var{c},
2960 comprises a subset of all memory references including
2961 all those whose address is simply a base register. This allows the reload
2962 pass to reload an operand, if it does not directly correspond to the operand
2963 type of @var{c}, by copying its address into a base register.
2965 For example, on the S/390, some instructions do not accept arbitrary
2966 memory references, but only those that do not make use of an index
2967 register. The constraint letter @samp{Q} is defined via
2968 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2969 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2970 a @samp{Q} constraint can handle any memory operand, because the
2971 reload pass knows it can be reloaded by copying the memory address
2972 into a base register if required. This is analogous to the way
2973 a @samp{o} constraint can handle any memory operand.
2976 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2977 A C expression that defines the optional machine-dependent constraint
2978 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2979 @code{EXTRA_CONSTRAINT_STR}, that should
2980 be treated like address constraints by the reload pass.
2982 It should return 1 if the operand type represented by the constraint
2983 at the start of @var{str}, which starts with the letter @var{c}, comprises
2984 a subset of all memory addresses including
2985 all those that consist of just a base register. This allows the reload
2986 pass to reload an operand, if it does not directly correspond to the operand
2987 type of @var{str}, by copying it into a base register.
2989 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2990 be used with the @code{address_operand} predicate. It is treated
2991 analogously to the @samp{p} constraint.
2994 @node Stack and Calling
2995 @section Stack Layout and Calling Conventions
2996 @cindex calling conventions
2998 @c prevent bad page break with this line
2999 This describes the stack layout and calling conventions.
3003 * Exception Handling::
3008 * Register Arguments::
3010 * Aggregate Return::
3015 * Stack Smashing Protection::
3019 @subsection Basic Stack Layout
3020 @cindex stack frame layout
3021 @cindex frame layout
3023 @c prevent bad page break with this line
3024 Here is the basic stack layout.
3026 @defmac STACK_GROWS_DOWNWARD
3027 Define this macro if pushing a word onto the stack moves the stack
3028 pointer to a smaller address.
3030 When we say, ``define this macro if @dots{}'', it means that the
3031 compiler checks this macro only with @code{#ifdef} so the precise
3032 definition used does not matter.
3035 @defmac STACK_PUSH_CODE
3036 This macro defines the operation used when something is pushed
3037 on the stack. In RTL, a push operation will be
3038 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
3040 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
3041 and @code{POST_INC}. Which of these is correct depends on
3042 the stack direction and on whether the stack pointer points
3043 to the last item on the stack or whether it points to the
3044 space for the next item on the stack.
3046 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
3047 defined, which is almost always right, and @code{PRE_INC} otherwise,
3048 which is often wrong.
3051 @defmac FRAME_GROWS_DOWNWARD
3052 Define this macro to nonzero value if the addresses of local variable slots
3053 are at negative offsets from the frame pointer.
3056 @defmac ARGS_GROW_DOWNWARD
3057 Define this macro if successive arguments to a function occupy decreasing
3058 addresses on the stack.
3061 @defmac STARTING_FRAME_OFFSET
3062 Offset from the frame pointer to the first local variable slot to be allocated.
3064 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
3065 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
3066 Otherwise, it is found by adding the length of the first slot to the
3067 value @code{STARTING_FRAME_OFFSET}.
3068 @c i'm not sure if the above is still correct.. had to change it to get
3069 @c rid of an overfull. --mew 2feb93
3072 @defmac STACK_ALIGNMENT_NEEDED
3073 Define to zero to disable final alignment of the stack during reload.
3074 The nonzero default for this macro is suitable for most ports.
3076 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
3077 is a register save block following the local block that doesn't require
3078 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
3079 stack alignment and do it in the backend.
3082 @defmac STACK_POINTER_OFFSET
3083 Offset from the stack pointer register to the first location at which
3084 outgoing arguments are placed. If not specified, the default value of
3085 zero is used. This is the proper value for most machines.
3087 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3088 the first location at which outgoing arguments are placed.
3091 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3092 Offset from the argument pointer register to the first argument's
3093 address. On some machines it may depend on the data type of the
3096 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3097 the first argument's address.
3100 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3101 Offset from the stack pointer register to an item dynamically allocated
3102 on the stack, e.g., by @code{alloca}.
3104 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3105 length of the outgoing arguments. The default is correct for most
3106 machines. See @file{function.c} for details.
3109 @defmac INITIAL_FRAME_ADDRESS_RTX
3110 A C expression whose value is RTL representing the address of the initial
3111 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3112 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3113 default value will be used. Define this macro in order to make frame pointer
3114 elimination work in the presence of @code{__builtin_frame_address (count)} and
3115 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3118 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3119 A C expression whose value is RTL representing the address in a stack
3120 frame where the pointer to the caller's frame is stored. Assume that
3121 @var{frameaddr} is an RTL expression for the address of the stack frame
3124 If you don't define this macro, the default is to return the value
3125 of @var{frameaddr}---that is, the stack frame address is also the
3126 address of the stack word that points to the previous frame.
3129 @defmac SETUP_FRAME_ADDRESSES
3130 If defined, a C expression that produces the machine-specific code to
3131 setup the stack so that arbitrary frames can be accessed. For example,
3132 on the SPARC, we must flush all of the register windows to the stack
3133 before we can access arbitrary stack frames. You will seldom need to
3137 @deftypefn {Target Hook} bool TARGET_BUILTIN_SETJMP_FRAME_VALUE ()
3138 This target hook should return an rtx that is used to store
3139 the address of the current frame into the built in @code{setjmp} buffer.
3140 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3141 machines. One reason you may need to define this target hook is if
3142 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3145 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3146 A C expression whose value is RTL representing the value of the frame
3147 address for the current frame. @var{frameaddr} is the frame pointer
3148 of the current frame. This is used for __builtin_frame_address.
3149 You need only define this macro if the frame address is not the same
3150 as the frame pointer. Most machines do not need to define it.
3153 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3154 A C expression whose value is RTL representing the value of the return
3155 address for the frame @var{count} steps up from the current frame, after
3156 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3157 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3158 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
3160 The value of the expression must always be the correct address when
3161 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
3162 determine the return address of other frames.
3165 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3166 Define this if the return address of a particular stack frame is accessed
3167 from the frame pointer of the previous stack frame.
3170 @defmac INCOMING_RETURN_ADDR_RTX
3171 A C expression whose value is RTL representing the location of the
3172 incoming return address at the beginning of any function, before the
3173 prologue. This RTL is either a @code{REG}, indicating that the return
3174 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3177 You only need to define this macro if you want to support call frame
3178 debugging information like that provided by DWARF 2.
3180 If this RTL is a @code{REG}, you should also define
3181 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3184 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3185 A C expression whose value is an integer giving a DWARF 2 column
3186 number that may be used as an alternative return column. The column
3187 must not correspond to any gcc hard register (that is, it must not
3188 be in the range of @code{DWARF_FRAME_REGNUM}).
3190 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3191 general register, but an alternative column needs to be used for signal
3192 frames. Some targets have also used different frame return columns
3196 @defmac DWARF_ZERO_REG
3197 A C expression whose value is an integer giving a DWARF 2 register
3198 number that is considered to always have the value zero. This should
3199 only be defined if the target has an architected zero register, and
3200 someone decided it was a good idea to use that register number to
3201 terminate the stack backtrace. New ports should avoid this.
3204 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3205 This target hook allows the backend to emit frame-related insns that
3206 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3207 info engine will invoke it on insns of the form
3209 (set (reg) (unspec [...] UNSPEC_INDEX))
3213 (set (reg) (unspec_volatile [...] UNSPECV_INDEX)).
3215 to let the backend emit the call frame instructions. @var{label} is
3216 the CFI label attached to the insn, @var{pattern} is the pattern of
3217 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3220 @defmac INCOMING_FRAME_SP_OFFSET
3221 A C expression whose value is an integer giving the offset, in bytes,
3222 from the value of the stack pointer register to the top of the stack
3223 frame at the beginning of any function, before the prologue. The top of
3224 the frame is defined to be the value of the stack pointer in the
3225 previous frame, just before the call instruction.
3227 You only need to define this macro if you want to support call frame
3228 debugging information like that provided by DWARF 2.
3231 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3232 A C expression whose value is an integer giving the offset, in bytes,
3233 from the argument pointer to the canonical frame address (cfa). The
3234 final value should coincide with that calculated by
3235 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3236 during virtual register instantiation.
3238 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
3239 which is correct for most machines; in general, the arguments are found
3240 immediately before the stack frame. Note that this is not the case on
3241 some targets that save registers into the caller's frame, such as SPARC
3242 and rs6000, and so such targets need to define this macro.
3244 You only need to define this macro if the default is incorrect, and you
3245 want to support call frame debugging information like that provided by
3249 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3250 If defined, a C expression whose value is an integer giving the offset
3251 in bytes from the frame pointer to the canonical frame address (cfa).
3252 The final value should coincide with that calculated by
3253 @code{INCOMING_FRAME_SP_OFFSET}.
3255 Normally the CFA is calculated as an offset from the argument pointer,
3256 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3257 variable due to the ABI, this may not be possible. If this macro is
3258 defined, it implies that the virtual register instantiation should be
3259 based on the frame pointer instead of the argument pointer. Only one
3260 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3264 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3265 If defined, a C expression whose value is an integer giving the offset
3266 in bytes from the canonical frame address (cfa) to the frame base used
3267 in DWARF 2 debug information. The default is zero. A different value
3268 may reduce the size of debug information on some ports.
3271 @node Exception Handling
3272 @subsection Exception Handling Support
3273 @cindex exception handling
3275 @defmac EH_RETURN_DATA_REGNO (@var{N})
3276 A C expression whose value is the @var{N}th register number used for
3277 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3278 @var{N} registers are usable.
3280 The exception handling library routines communicate with the exception
3281 handlers via a set of agreed upon registers. Ideally these registers
3282 should be call-clobbered; it is possible to use call-saved registers,
3283 but may negatively impact code size. The target must support at least
3284 2 data registers, but should define 4 if there are enough free registers.
3286 You must define this macro if you want to support call frame exception
3287 handling like that provided by DWARF 2.
3290 @defmac EH_RETURN_STACKADJ_RTX
3291 A C expression whose value is RTL representing a location in which
3292 to store a stack adjustment to be applied before function return.
3293 This is used to unwind the stack to an exception handler's call frame.
3294 It will be assigned zero on code paths that return normally.
3296 Typically this is a call-clobbered hard register that is otherwise
3297 untouched by the epilogue, but could also be a stack slot.
3299 Do not define this macro if the stack pointer is saved and restored
3300 by the regular prolog and epilog code in the call frame itself; in
3301 this case, the exception handling library routines will update the
3302 stack location to be restored in place. Otherwise, you must define
3303 this macro if you want to support call frame exception handling like
3304 that provided by DWARF 2.
3307 @defmac EH_RETURN_HANDLER_RTX
3308 A C expression whose value is RTL representing a location in which
3309 to store the address of an exception handler to which we should
3310 return. It will not be assigned on code paths that return normally.
3312 Typically this is the location in the call frame at which the normal
3313 return address is stored. For targets that return by popping an
3314 address off the stack, this might be a memory address just below
3315 the @emph{target} call frame rather than inside the current call
3316 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3317 been assigned, so it may be used to calculate the location of the
3320 Some targets have more complex requirements than storing to an
3321 address calculable during initial code generation. In that case
3322 the @code{eh_return} instruction pattern should be used instead.
3324 If you want to support call frame exception handling, you must
3325 define either this macro or the @code{eh_return} instruction pattern.
3328 @defmac RETURN_ADDR_OFFSET
3329 If defined, an integer-valued C expression for which rtl will be generated
3330 to add it to the exception handler address before it is searched in the
3331 exception handling tables, and to subtract it again from the address before
3332 using it to return to the exception handler.
3335 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3336 This macro chooses the encoding of pointers embedded in the exception
3337 handling sections. If at all possible, this should be defined such
3338 that the exception handling section will not require dynamic relocations,
3339 and so may be read-only.
3341 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3342 @var{global} is true if the symbol may be affected by dynamic relocations.
3343 The macro should return a combination of the @code{DW_EH_PE_*} defines
3344 as found in @file{dwarf2.h}.
3346 If this macro is not defined, pointers will not be encoded but
3347 represented directly.
3350 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3351 This macro allows the target to emit whatever special magic is required
3352 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3353 Generic code takes care of pc-relative and indirect encodings; this must
3354 be defined if the target uses text-relative or data-relative encodings.
3356 This is a C statement that branches to @var{done} if the format was
3357 handled. @var{encoding} is the format chosen, @var{size} is the number
3358 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3362 @defmac MD_UNWIND_SUPPORT
3363 A string specifying a file to be #include'd in unwind-dw2.c. The file
3364 so included typically defines @code{MD_FALLBACK_FRAME_STATE_FOR}.
3367 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3368 This macro allows the target to add CPU and operating system specific
3369 code to the call-frame unwinder for use when there is no unwind data
3370 available. The most common reason to implement this macro is to unwind
3371 through signal frames.
3373 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
3374 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3375 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3376 for the address of the code being executed and @code{context->cfa} for
3377 the stack pointer value. If the frame can be decoded, the register save
3378 addresses should be updated in @var{fs} and the macro should evaluate to
3379 @code{_URC_NO_REASON}. If the frame cannot be decoded, the macro should
3380 evaluate to @code{_URC_END_OF_STACK}.
3382 For proper signal handling in Java this macro is accompanied by
3383 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3386 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3387 This macro allows the target to add operating system specific code to the
3388 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3389 usually used for signal or interrupt frames.
3391 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3392 @var{context} is an @code{_Unwind_Context};
3393 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3394 for the abi and context in the @code{.unwabi} directive. If the
3395 @code{.unwabi} directive can be handled, the register save addresses should
3396 be updated in @var{fs}.
3399 @defmac TARGET_USES_WEAK_UNWIND_INFO
3400 A C expression that evaluates to true if the target requires unwind
3401 info to be given comdat linkage. Define it to be @code{1} if comdat
3402 linkage is necessary. The default is @code{0}.
3405 @node Stack Checking
3406 @subsection Specifying How Stack Checking is Done
3408 GCC will check that stack references are within the boundaries of
3409 the stack, if the @option{-fstack-check} is specified, in one of three ways:
3413 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3414 will assume that you have arranged for stack checking to be done at
3415 appropriate places in the configuration files, e.g., in
3416 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
3420 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
3421 called @code{check_stack} in your @file{md} file, GCC will call that
3422 pattern with one argument which is the address to compare the stack
3423 value against. You must arrange for this pattern to report an error if
3424 the stack pointer is out of range.
3427 If neither of the above are true, GCC will generate code to periodically
3428 ``probe'' the stack pointer using the values of the macros defined below.
3431 Normally, you will use the default values of these macros, so GCC
3432 will use the third approach.
3434 @defmac STACK_CHECK_BUILTIN
3435 A nonzero value if stack checking is done by the configuration files in a
3436 machine-dependent manner. You should define this macro if stack checking
3437 is require by the ABI of your machine or if you would like to have to stack
3438 checking in some more efficient way than GCC's portable approach.
3439 The default value of this macro is zero.
3442 @defmac STACK_CHECK_PROBE_INTERVAL
3443 An integer representing the interval at which GCC must generate stack
3444 probe instructions. You will normally define this macro to be no larger
3445 than the size of the ``guard pages'' at the end of a stack area. The
3446 default value of 4096 is suitable for most systems.
3449 @defmac STACK_CHECK_PROBE_LOAD
3450 A integer which is nonzero if GCC should perform the stack probe
3451 as a load instruction and zero if GCC should use a store instruction.
3452 The default is zero, which is the most efficient choice on most systems.
3455 @defmac STACK_CHECK_PROTECT
3456 The number of bytes of stack needed to recover from a stack overflow,
3457 for languages where such a recovery is supported. The default value of
3458 75 words should be adequate for most machines.
3461 @defmac STACK_CHECK_MAX_FRAME_SIZE
3462 The maximum size of a stack frame, in bytes. GCC will generate probe
3463 instructions in non-leaf functions to ensure at least this many bytes of
3464 stack are available. If a stack frame is larger than this size, stack
3465 checking will not be reliable and GCC will issue a warning. The
3466 default is chosen so that GCC only generates one instruction on most
3467 systems. You should normally not change the default value of this macro.
3470 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3471 GCC uses this value to generate the above warning message. It
3472 represents the amount of fixed frame used by a function, not including
3473 space for any callee-saved registers, temporaries and user variables.
3474 You need only specify an upper bound for this amount and will normally
3475 use the default of four words.
3478 @defmac STACK_CHECK_MAX_VAR_SIZE
3479 The maximum size, in bytes, of an object that GCC will place in the
3480 fixed area of the stack frame when the user specifies
3481 @option{-fstack-check}.
3482 GCC computed the default from the values of the above macros and you will
3483 normally not need to override that default.
3487 @node Frame Registers
3488 @subsection Registers That Address the Stack Frame
3490 @c prevent bad page break with this line
3491 This discusses registers that address the stack frame.
3493 @defmac STACK_POINTER_REGNUM
3494 The register number of the stack pointer register, which must also be a
3495 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3496 the hardware determines which register this is.
3499 @defmac FRAME_POINTER_REGNUM
3500 The register number of the frame pointer register, which is used to
3501 access automatic variables in the stack frame. On some machines, the
3502 hardware determines which register this is. On other machines, you can
3503 choose any register you wish for this purpose.
3506 @defmac HARD_FRAME_POINTER_REGNUM
3507 On some machines the offset between the frame pointer and starting
3508 offset of the automatic variables is not known until after register
3509 allocation has been done (for example, because the saved registers are
3510 between these two locations). On those machines, define
3511 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3512 be used internally until the offset is known, and define
3513 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3514 used for the frame pointer.
3516 You should define this macro only in the very rare circumstances when it
3517 is not possible to calculate the offset between the frame pointer and
3518 the automatic variables until after register allocation has been
3519 completed. When this macro is defined, you must also indicate in your
3520 definition of @code{ELIMINABLE_REGS} how to eliminate
3521 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3522 or @code{STACK_POINTER_REGNUM}.
3524 Do not define this macro if it would be the same as
3525 @code{FRAME_POINTER_REGNUM}.
3528 @defmac ARG_POINTER_REGNUM
3529 The register number of the arg pointer register, which is used to access
3530 the function's argument list. On some machines, this is the same as the
3531 frame pointer register. On some machines, the hardware determines which
3532 register this is. On other machines, you can choose any register you
3533 wish for this purpose. If this is not the same register as the frame
3534 pointer register, then you must mark it as a fixed register according to
3535 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3536 (@pxref{Elimination}).
3539 @defmac RETURN_ADDRESS_POINTER_REGNUM
3540 The register number of the return address pointer register, which is used to
3541 access the current function's return address from the stack. On some
3542 machines, the return address is not at a fixed offset from the frame
3543 pointer or stack pointer or argument pointer. This register can be defined
3544 to point to the return address on the stack, and then be converted by
3545 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3547 Do not define this macro unless there is no other way to get the return
3548 address from the stack.
3551 @defmac STATIC_CHAIN_REGNUM
3552 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3553 Register numbers used for passing a function's static chain pointer. If
3554 register windows are used, the register number as seen by the called
3555 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3556 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3557 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3560 The static chain register need not be a fixed register.
3562 If the static chain is passed in memory, these macros should not be
3563 defined; instead, the next two macros should be defined.
3566 @defmac STATIC_CHAIN
3567 @defmacx STATIC_CHAIN_INCOMING
3568 If the static chain is passed in memory, these macros provide rtx giving
3569 @code{mem} expressions that denote where they are stored.
3570 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3571 as seen by the calling and called functions, respectively. Often the former
3572 will be at an offset from the stack pointer and the latter at an offset from
3575 @findex stack_pointer_rtx
3576 @findex frame_pointer_rtx
3577 @findex arg_pointer_rtx
3578 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3579 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3580 macros and should be used to refer to those items.
3582 If the static chain is passed in a register, the two previous macros should
3586 @defmac DWARF_FRAME_REGISTERS
3587 This macro specifies the maximum number of hard registers that can be
3588 saved in a call frame. This is used to size data structures used in
3589 DWARF2 exception handling.
3591 Prior to GCC 3.0, this macro was needed in order to establish a stable
3592 exception handling ABI in the face of adding new hard registers for ISA
3593 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3594 in the number of hard registers. Nevertheless, this macro can still be
3595 used to reduce the runtime memory requirements of the exception handling
3596 routines, which can be substantial if the ISA contains a lot of
3597 registers that are not call-saved.
3599 If this macro is not defined, it defaults to
3600 @code{FIRST_PSEUDO_REGISTER}.
3603 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3605 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3606 for backward compatibility in pre GCC 3.0 compiled code.
3608 If this macro is not defined, it defaults to
3609 @code{DWARF_FRAME_REGISTERS}.
3612 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3614 Define this macro if the target's representation for dwarf registers
3615 is different than the internal representation for unwind column.
3616 Given a dwarf register, this macro should return the internal unwind
3617 column number to use instead.
3619 See the PowerPC's SPE target for an example.
3622 @defmac DWARF_FRAME_REGNUM (@var{regno})
3624 Define this macro if the target's representation for dwarf registers
3625 used in .eh_frame or .debug_frame is different from that used in other
3626 debug info sections. Given a GCC hard register number, this macro
3627 should return the .eh_frame register number. The default is
3628 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3632 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3634 Define this macro to map register numbers held in the call frame info
3635 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3636 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3637 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3638 return @code{@var{regno}}.
3643 @subsection Eliminating Frame Pointer and Arg Pointer
3645 @c prevent bad page break with this line
3646 This is about eliminating the frame pointer and arg pointer.
3648 @defmac FRAME_POINTER_REQUIRED
3649 A C expression which is nonzero if a function must have and use a frame
3650 pointer. This expression is evaluated in the reload pass. If its value is
3651 nonzero the function will have a frame pointer.
3653 The expression can in principle examine the current function and decide
3654 according to the facts, but on most machines the constant 0 or the
3655 constant 1 suffices. Use 0 when the machine allows code to be generated
3656 with no frame pointer, and doing so saves some time or space. Use 1
3657 when there is no possible advantage to avoiding a frame pointer.
3659 In certain cases, the compiler does not know how to produce valid code
3660 without a frame pointer. The compiler recognizes those cases and
3661 automatically gives the function a frame pointer regardless of what
3662 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3665 In a function that does not require a frame pointer, the frame pointer
3666 register can be allocated for ordinary usage, unless you mark it as a
3667 fixed register. See @code{FIXED_REGISTERS} for more information.
3670 @findex get_frame_size
3671 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3672 A C statement to store in the variable @var{depth-var} the difference
3673 between the frame pointer and the stack pointer values immediately after
3674 the function prologue. The value would be computed from information
3675 such as the result of @code{get_frame_size ()} and the tables of
3676 registers @code{regs_ever_live} and @code{call_used_regs}.
3678 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3679 need not be defined. Otherwise, it must be defined even if
3680 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3681 case, you may set @var{depth-var} to anything.
3684 @defmac ELIMINABLE_REGS
3685 If defined, this macro specifies a table of register pairs used to
3686 eliminate unneeded registers that point into the stack frame. If it is not
3687 defined, the only elimination attempted by the compiler is to replace
3688 references to the frame pointer with references to the stack pointer.
3690 The definition of this macro is a list of structure initializations, each
3691 of which specifies an original and replacement register.
3693 On some machines, the position of the argument pointer is not known until
3694 the compilation is completed. In such a case, a separate hard register
3695 must be used for the argument pointer. This register can be eliminated by
3696 replacing it with either the frame pointer or the argument pointer,
3697 depending on whether or not the frame pointer has been eliminated.
3699 In this case, you might specify:
3701 #define ELIMINABLE_REGS \
3702 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3703 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3704 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3707 Note that the elimination of the argument pointer with the stack pointer is
3708 specified first since that is the preferred elimination.
3711 @defmac CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3712 A C expression that returns nonzero if the compiler is allowed to try
3713 to replace register number @var{from-reg} with register number
3714 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3715 is defined, and will usually be the constant 1, since most of the cases
3716 preventing register elimination are things that the compiler already
3720 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3721 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3722 specifies the initial difference between the specified pair of
3723 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3727 @node Stack Arguments
3728 @subsection Passing Function Arguments on the Stack
3729 @cindex arguments on stack
3730 @cindex stack arguments
3732 The macros in this section control how arguments are passed
3733 on the stack. See the following section for other macros that
3734 control passing certain arguments in registers.
3736 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (tree @var{fntype})
3737 This target hook returns @code{true} if an argument declared in a
3738 prototype as an integral type smaller than @code{int} should actually be
3739 passed as an @code{int}. In addition to avoiding errors in certain
3740 cases of mismatch, it also makes for better code on certain machines.
3741 The default is to not promote prototypes.
3745 A C expression. If nonzero, push insns will be used to pass
3747 If the target machine does not have a push instruction, set it to zero.
3748 That directs GCC to use an alternate strategy: to
3749 allocate the entire argument block and then store the arguments into
3750 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3753 @defmac PUSH_ARGS_REVERSED
3754 A C expression. If nonzero, function arguments will be evaluated from
3755 last to first, rather than from first to last. If this macro is not
3756 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3757 and args grow in opposite directions, and 0 otherwise.
3760 @defmac PUSH_ROUNDING (@var{npushed})
3761 A C expression that is the number of bytes actually pushed onto the
3762 stack when an instruction attempts to push @var{npushed} bytes.
3764 On some machines, the definition
3767 #define PUSH_ROUNDING(BYTES) (BYTES)
3771 will suffice. But on other machines, instructions that appear
3772 to push one byte actually push two bytes in an attempt to maintain
3773 alignment. Then the definition should be
3776 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3780 @findex current_function_outgoing_args_size
3781 @defmac ACCUMULATE_OUTGOING_ARGS
3782 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3783 will be computed and placed into the variable
3784 @code{current_function_outgoing_args_size}. No space will be pushed
3785 onto the stack for each call; instead, the function prologue should
3786 increase the stack frame size by this amount.
3788 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3792 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3793 Define this macro if functions should assume that stack space has been
3794 allocated for arguments even when their values are passed in
3797 The value of this macro is the size, in bytes, of the area reserved for
3798 arguments passed in registers for the function represented by @var{fndecl},
3799 which can be zero if GCC is calling a library function.
3801 This space can be allocated by the caller, or be a part of the
3802 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3805 @c above is overfull. not sure what to do. --mew 5feb93 did
3806 @c something, not sure if it looks good. --mew 10feb93
3808 @defmac OUTGOING_REG_PARM_STACK_SPACE
3809 Define this to a nonzero value if it is the responsibility of the caller
3810 to allocate the area reserved for arguments passed in registers.
3812 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3813 whether the space for these arguments counts in the value of
3814 @code{current_function_outgoing_args_size}.
3817 @defmac STACK_PARMS_IN_REG_PARM_AREA
3818 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3819 stack parameters don't skip the area specified by it.
3820 @c i changed this, makes more sens and it should have taken care of the
3821 @c overfull.. not as specific, tho. --mew 5feb93
3823 Normally, when a parameter is not passed in registers, it is placed on the
3824 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3825 suppresses this behavior and causes the parameter to be passed on the
3826 stack in its natural location.
3829 @defmac RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3830 A C expression that should indicate the number of bytes of its own
3831 arguments that a function pops on returning, or 0 if the
3832 function pops no arguments and the caller must therefore pop them all
3833 after the function returns.
3835 @var{fundecl} is a C variable whose value is a tree node that describes
3836 the function in question. Normally it is a node of type
3837 @code{FUNCTION_DECL} that describes the declaration of the function.
3838 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3840 @var{funtype} is a C variable whose value is a tree node that
3841 describes the function in question. Normally it is a node of type
3842 @code{FUNCTION_TYPE} that describes the data type of the function.
3843 From this it is possible to obtain the data types of the value and
3844 arguments (if known).
3846 When a call to a library function is being considered, @var{fundecl}
3847 will contain an identifier node for the library function. Thus, if
3848 you need to distinguish among various library functions, you can do so
3849 by their names. Note that ``library function'' in this context means
3850 a function used to perform arithmetic, whose name is known specially
3851 in the compiler and was not mentioned in the C code being compiled.
3853 @var{stack-size} is the number of bytes of arguments passed on the
3854 stack. If a variable number of bytes is passed, it is zero, and
3855 argument popping will always be the responsibility of the calling function.
3857 On the VAX, all functions always pop their arguments, so the definition
3858 of this macro is @var{stack-size}. On the 68000, using the standard
3859 calling convention, no functions pop their arguments, so the value of
3860 the macro is always 0 in this case. But an alternative calling
3861 convention is available in which functions that take a fixed number of
3862 arguments pop them but other functions (such as @code{printf}) pop
3863 nothing (the caller pops all). When this convention is in use,
3864 @var{funtype} is examined to determine whether a function takes a fixed
3865 number of arguments.
3868 @defmac CALL_POPS_ARGS (@var{cum})
3869 A C expression that should indicate the number of bytes a call sequence
3870 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3871 when compiling a function call.
3873 @var{cum} is the variable in which all arguments to the called function
3874 have been accumulated.
3876 On certain architectures, such as the SH5, a call trampoline is used
3877 that pops certain registers off the stack, depending on the arguments
3878 that have been passed to the function. Since this is a property of the
3879 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3883 @node Register Arguments
3884 @subsection Passing Arguments in Registers
3885 @cindex arguments in registers
3886 @cindex registers arguments
3888 This section describes the macros which let you control how various
3889 types of arguments are passed in registers or how they are arranged in
3892 @defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3893 A C expression that controls whether a function argument is passed
3894 in a register, and which register.
3896 The arguments are @var{cum}, which summarizes all the previous
3897 arguments; @var{mode}, the machine mode of the argument; @var{type},
3898 the data type of the argument as a tree node or 0 if that is not known
3899 (which happens for C support library functions); and @var{named},
3900 which is 1 for an ordinary argument and 0 for nameless arguments that
3901 correspond to @samp{@dots{}} in the called function's prototype.
3902 @var{type} can be an incomplete type if a syntax error has previously
3905 The value of the expression is usually either a @code{reg} RTX for the
3906 hard register in which to pass the argument, or zero to pass the
3907 argument on the stack.
3909 For machines like the VAX and 68000, where normally all arguments are
3910 pushed, zero suffices as a definition.
3912 The value of the expression can also be a @code{parallel} RTX@. This is
3913 used when an argument is passed in multiple locations. The mode of the
3914 @code{parallel} should be the mode of the entire argument. The
3915 @code{parallel} holds any number of @code{expr_list} pairs; each one
3916 describes where part of the argument is passed. In each
3917 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3918 register in which to pass this part of the argument, and the mode of the
3919 register RTX indicates how large this part of the argument is. The
3920 second operand of the @code{expr_list} is a @code{const_int} which gives
3921 the offset in bytes into the entire argument of where this part starts.
3922 As a special exception the first @code{expr_list} in the @code{parallel}
3923 RTX may have a first operand of zero. This indicates that the entire
3924 argument is also stored on the stack.
3926 The last time this macro is called, it is called with @code{MODE ==
3927 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3928 pattern as operands 2 and 3 respectively.
3930 @cindex @file{stdarg.h} and register arguments
3931 The usual way to make the ISO library @file{stdarg.h} work on a machine
3932 where some arguments are usually passed in registers, is to cause
3933 nameless arguments to be passed on the stack instead. This is done
3934 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3936 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3937 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3938 You may use the hook @code{targetm.calls.must_pass_in_stack}
3939 in the definition of this macro to determine if this argument is of a
3940 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3941 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3942 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3943 defined, the argument will be computed in the stack and then loaded into
3947 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, tree @var{type})
3948 This target hook should return @code{true} if we should not pass @var{type}
3949 solely in registers. The file @file{expr.h} defines a
3950 definition that is usually appropriate, refer to @file{expr.h} for additional
3954 @defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3955 Define this macro if the target machine has ``register windows'', so
3956 that the register in which a function sees an arguments is not
3957 necessarily the same as the one in which the caller passed the
3960 For such machines, @code{FUNCTION_ARG} computes the register in which
3961 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3962 be defined in a similar fashion to tell the function being called
3963 where the arguments will arrive.
3965 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3966 serves both purposes.
3969 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3970 This target hook returns the number of bytes at the beginning of an
3971 argument that must be put in registers. The value must be zero for
3972 arguments that are passed entirely in registers or that are entirely
3973 pushed on the stack.
3975 On some machines, certain arguments must be passed partially in
3976 registers and partially in memory. On these machines, typically the
3977 first few words of arguments are passed in registers, and the rest
3978 on the stack. If a multi-word argument (a @code{double} or a
3979 structure) crosses that boundary, its first few words must be passed
3980 in registers and the rest must be pushed. This macro tells the
3981 compiler when this occurs, and how many bytes should go in registers.
3983 @code{FUNCTION_ARG} for these arguments should return the first
3984 register to be used by the caller for this argument; likewise
3985 @code{FUNCTION_INCOMING_ARG}, for the called function.
3988 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3989 This target hook should return @code{true} if an argument at the
3990 position indicated by @var{cum} should be passed by reference. This
3991 predicate is queried after target independent reasons for being
3992 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
3994 If the hook returns true, a copy of that argument is made in memory and a
3995 pointer to the argument is passed instead of the argument itself.
3996 The pointer is passed in whatever way is appropriate for passing a pointer
4000 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
4001 The function argument described by the parameters to this hook is
4002 known to be passed by reference. The hook should return true if the
4003 function argument should be copied by the callee instead of copied
4006 For any argument for which the hook returns true, if it can be
4007 determined that the argument is not modified, then a copy need
4010 The default version of this hook always returns false.
4013 @defmac CUMULATIVE_ARGS
4014 A C type for declaring a variable that is used as the first argument of
4015 @code{FUNCTION_ARG} and other related values. For some target machines,
4016 the type @code{int} suffices and can hold the number of bytes of
4019 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
4020 arguments that have been passed on the stack. The compiler has other
4021 variables to keep track of that. For target machines on which all
4022 arguments are passed on the stack, there is no need to store anything in
4023 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
4024 should not be empty, so use @code{int}.
4027 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
4028 A C statement (sans semicolon) for initializing the variable
4029 @var{cum} for the state at the beginning of the argument list. The
4030 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
4031 is the tree node for the data type of the function which will receive
4032 the args, or 0 if the args are to a compiler support library function.
4033 For direct calls that are not libcalls, @var{fndecl} contain the
4034 declaration node of the function. @var{fndecl} is also set when
4035 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
4036 being compiled. @var{n_named_args} is set to the number of named
4037 arguments, including a structure return address if it is passed as a
4038 parameter, when making a call. When processing incoming arguments,
4039 @var{n_named_args} is set to @minus{}1.
4041 When processing a call to a compiler support library function,
4042 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
4043 contains the name of the function, as a string. @var{libname} is 0 when
4044 an ordinary C function call is being processed. Thus, each time this
4045 macro is called, either @var{libname} or @var{fntype} is nonzero, but
4046 never both of them at once.
4049 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
4050 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
4051 it gets a @code{MODE} argument instead of @var{fntype}, that would be
4052 @code{NULL}. @var{indirect} would always be zero, too. If this macro
4053 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
4054 0)} is used instead.
4057 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
4058 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
4059 finding the arguments for the function being compiled. If this macro is
4060 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
4062 The value passed for @var{libname} is always 0, since library routines
4063 with special calling conventions are never compiled with GCC@. The
4064 argument @var{libname} exists for symmetry with
4065 @code{INIT_CUMULATIVE_ARGS}.
4066 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
4067 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
4070 @defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
4071 A C statement (sans semicolon) to update the summarizer variable
4072 @var{cum} to advance past an argument in the argument list. The
4073 values @var{mode}, @var{type} and @var{named} describe that argument.
4074 Once this is done, the variable @var{cum} is suitable for analyzing
4075 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
4077 This macro need not do anything if the argument in question was passed
4078 on the stack. The compiler knows how to track the amount of stack space
4079 used for arguments without any special help.
4082 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
4083 If defined, a C expression which determines whether, and in which direction,
4084 to pad out an argument with extra space. The value should be of type
4085 @code{enum direction}: either @code{upward} to pad above the argument,
4086 @code{downward} to pad below, or @code{none} to inhibit padding.
4088 The @emph{amount} of padding is always just enough to reach the next
4089 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
4092 This macro has a default definition which is right for most systems.
4093 For little-endian machines, the default is to pad upward. For
4094 big-endian machines, the default is to pad downward for an argument of
4095 constant size shorter than an @code{int}, and upward otherwise.
4098 @defmac PAD_VARARGS_DOWN
4099 If defined, a C expression which determines whether the default
4100 implementation of va_arg will attempt to pad down before reading the
4101 next argument, if that argument is smaller than its aligned space as
4102 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4103 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4106 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4107 Specify padding for the last element of a block move between registers and
4108 memory. @var{first} is nonzero if this is the only element. Defining this
4109 macro allows better control of register function parameters on big-endian
4110 machines, without using @code{PARALLEL} rtl. In particular,
4111 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4112 registers, as there is no longer a "wrong" part of a register; For example,
4113 a three byte aggregate may be passed in the high part of a register if so
4117 @defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
4118 If defined, a C expression that gives the alignment boundary, in bits,
4119 of an argument with the specified mode and type. If it is not defined,
4120 @code{PARM_BOUNDARY} is used for all arguments.
4123 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4124 A C expression that is nonzero if @var{regno} is the number of a hard
4125 register in which function arguments are sometimes passed. This does
4126 @emph{not} include implicit arguments such as the static chain and
4127 the structure-value address. On many machines, no registers can be
4128 used for this purpose since all function arguments are pushed on the
4132 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (tree @var{type})
4133 This hook should return true if parameter of type @var{type} are passed
4134 as two scalar parameters. By default, GCC will attempt to pack complex
4135 arguments into the target's word size. Some ABIs require complex arguments
4136 to be split and treated as their individual components. For example, on
4137 AIX64, complex floats should be passed in a pair of floating point
4138 registers, even though a complex float would fit in one 64-bit floating
4141 The default value of this hook is @code{NULL}, which is treated as always
4145 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
4146 This hook returns a type node for @code{va_list} for the target.
4147 The default version of the hook returns @code{void*}.
4150 @deftypefn {Target Hook} tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree @var{valist}, tree @var{type}, tree *@var{pre_p}, tree *@var{post_p})
4151 This hook performs target-specific gimplification of
4152 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4153 arguments to @code{va_arg}; the latter two are as in
4154 @code{gimplify.c:gimplify_expr}.
4157 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (enum machine_mode @var{mode})
4158 Define this to return nonzero if the port can handle pointers
4159 with machine mode @var{mode}. The default version of this
4160 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4163 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4164 Define this to return nonzero if the port is prepared to handle
4165 insns involving scalar mode @var{mode}. For a scalar mode to be
4166 considered supported, all the basic arithmetic and comparisons
4169 The default version of this hook returns true for any mode
4170 required to handle the basic C types (as defined by the port).
4171 Included here are the double-word arithmetic supported by the
4172 code in @file{optabs.c}.
4175 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4176 Define this to return nonzero if the port is prepared to handle
4177 insns involving vector mode @var{mode}. At the very least, it
4178 must have move patterns for this mode.
4182 @subsection How Scalar Function Values Are Returned
4183 @cindex return values in registers
4184 @cindex values, returned by functions
4185 @cindex scalars, returned as values
4187 This section discusses the macros that control returning scalars as
4188 values---values that can fit in registers.
4190 @deftypefn {Target Hook} rtx TARGET_FUNCTION_VALUE (tree @var{ret_type}, tree @var{fn_decl_or_type}, bool @var{outgoing})
4192 Define this to return an RTX representing the place where a function
4193 returns or receives a value of data type @var{ret_type}, a tree node
4194 node representing a data type. @var{fn_decl_or_type} is a tree node
4195 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4196 function being called. If @var{outgoing} is false, the hook should
4197 compute the register in which the caller will see the return value.
4198 Otherwise, the hook should return an RTX representing the place where
4199 a function returns a value.
4201 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4202 (Actually, on most machines, scalar values are returned in the same
4203 place regardless of mode.) The value of the expression is usually a
4204 @code{reg} RTX for the hard register where the return value is stored.
4205 The value can also be a @code{parallel} RTX, if the return value is in
4206 multiple places. See @code{FUNCTION_ARG} for an explanation of the
4207 @code{parallel} form. Note that the callee will populate every
4208 location specified in the @code{parallel}, but if the first element of
4209 the @code{parallel} contains the whole return value, callers will use
4210 that element as the canonical location and ignore the others. The m68k
4211 port uses this type of @code{parallel} to return pointers in both
4212 @samp{%a0} (the canonical location) and @samp{%d0}.
4214 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4215 the same promotion rules specified in @code{PROMOTE_MODE} if
4216 @var{valtype} is a scalar type.
4218 If the precise function being called is known, @var{func} is a tree
4219 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4220 pointer. This makes it possible to use a different value-returning
4221 convention for specific functions when all their calls are
4224 Some target machines have ``register windows'' so that the register in
4225 which a function returns its value is not the same as the one in which
4226 the caller sees the value. For such machines, you should return
4227 different RTX depending on @var{outgoing}.
4229 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4230 aggregate data types, because these are returned in another way. See
4231 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4234 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4235 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4236 a new target instead.
4239 @defmac FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
4240 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4241 a new target instead.
4244 @defmac LIBCALL_VALUE (@var{mode})
4245 A C expression to create an RTX representing the place where a library
4246 function returns a value of mode @var{mode}. If the precise function
4247 being called is known, @var{func} is a tree node
4248 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4249 pointer. This makes it possible to use a different value-returning
4250 convention for specific functions when all their calls are
4253 Note that ``library function'' in this context means a compiler
4254 support routine, used to perform arithmetic, whose name is known
4255 specially by the compiler and was not mentioned in the C code being
4258 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
4259 data types, because none of the library functions returns such types.
4262 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4263 A C expression that is nonzero if @var{regno} is the number of a hard
4264 register in which the values of called function may come back.
4266 A register whose use for returning values is limited to serving as the
4267 second of a pair (for a value of type @code{double}, say) need not be
4268 recognized by this macro. So for most machines, this definition
4272 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4275 If the machine has register windows, so that the caller and the called
4276 function use different registers for the return value, this macro
4277 should recognize only the caller's register numbers.
4280 @defmac APPLY_RESULT_SIZE
4281 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4282 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4283 saving and restoring an arbitrary return value.
4286 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (tree @var{type})
4287 This hook should return true if values of type @var{type} are returned
4288 at the most significant end of a register (in other words, if they are
4289 padded at the least significant end). You can assume that @var{type}
4290 is returned in a register; the caller is required to check this.
4292 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4293 be able to hold the complete return value. For example, if a 1-, 2-
4294 or 3-byte structure is returned at the most significant end of a
4295 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4299 @node Aggregate Return
4300 @subsection How Large Values Are Returned
4301 @cindex aggregates as return values
4302 @cindex large return values
4303 @cindex returning aggregate values
4304 @cindex structure value address
4306 When a function value's mode is @code{BLKmode} (and in some other
4307 cases), the value is not returned according to
4308 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4309 caller passes the address of a block of memory in which the value
4310 should be stored. This address is called the @dfn{structure value
4313 This section describes how to control returning structure values in
4316 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (tree @var{type}, tree @var{fntype})
4317 This target hook should return a nonzero value to say to return the
4318 function value in memory, just as large structures are always returned.
4319 Here @var{type} will be the data type of the value, and @var{fntype}
4320 will be the type of the function doing the returning, or @code{NULL} for
4323 Note that values of mode @code{BLKmode} must be explicitly handled
4324 by this function. Also, the option @option{-fpcc-struct-return}
4325 takes effect regardless of this macro. On most systems, it is
4326 possible to leave the hook undefined; this causes a default
4327 definition to be used, whose value is the constant 1 for @code{BLKmode}
4328 values, and 0 otherwise.
4330 Do not use this hook to indicate that structures and unions should always
4331 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4335 @defmac DEFAULT_PCC_STRUCT_RETURN
4336 Define this macro to be 1 if all structure and union return values must be
4337 in memory. Since this results in slower code, this should be defined
4338 only if needed for compatibility with other compilers or with an ABI@.
4339 If you define this macro to be 0, then the conventions used for structure
4340 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4343 If not defined, this defaults to the value 1.
4346 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4347 This target hook should return the location of the structure value
4348 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4349 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4350 be @code{NULL}, for libcalls. You do not need to define this target
4351 hook if the address is always passed as an ``invisible'' first
4354 On some architectures the place where the structure value address
4355 is found by the called function is not the same place that the
4356 caller put it. This can be due to register windows, or it could
4357 be because the function prologue moves it to a different place.
4358 @var{incoming} is @code{1} or @code{2} when the location is needed in
4359 the context of the called function, and @code{0} in the context of
4362 If @var{incoming} is nonzero and the address is to be found on the
4363 stack, return a @code{mem} which refers to the frame pointer. If
4364 @var{incoming} is @code{2}, the result is being used to fetch the
4365 structure value address at the beginning of a function. If you need
4366 to emit adjusting code, you should do it at this point.
4369 @defmac PCC_STATIC_STRUCT_RETURN
4370 Define this macro if the usual system convention on the target machine
4371 for returning structures and unions is for the called function to return
4372 the address of a static variable containing the value.
4374 Do not define this if the usual system convention is for the caller to
4375 pass an address to the subroutine.
4377 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4378 nothing when you use @option{-freg-struct-return} mode.
4382 @subsection Caller-Saves Register Allocation
4384 If you enable it, GCC can save registers around function calls. This
4385 makes it possible to use call-clobbered registers to hold variables that
4386 must live across calls.
4388 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4389 A C expression to determine whether it is worthwhile to consider placing
4390 a pseudo-register in a call-clobbered hard register and saving and
4391 restoring it around each function call. The expression should be 1 when
4392 this is worth doing, and 0 otherwise.
4394 If you don't define this macro, a default is used which is good on most
4395 machines: @code{4 * @var{calls} < @var{refs}}.
4398 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4399 A C expression specifying which mode is required for saving @var{nregs}
4400 of a pseudo-register in call-clobbered hard register @var{regno}. If
4401 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4402 returned. For most machines this macro need not be defined since GCC
4403 will select the smallest suitable mode.
4406 @node Function Entry
4407 @subsection Function Entry and Exit
4408 @cindex function entry and exit
4412 This section describes the macros that output function entry
4413 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4415 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4416 If defined, a function that outputs the assembler code for entry to a
4417 function. The prologue is responsible for setting up the stack frame,
4418 initializing the frame pointer register, saving registers that must be
4419 saved, and allocating @var{size} additional bytes of storage for the
4420 local variables. @var{size} is an integer. @var{file} is a stdio
4421 stream to which the assembler code should be output.
4423 The label for the beginning of the function need not be output by this
4424 macro. That has already been done when the macro is run.
4426 @findex regs_ever_live
4427 To determine which registers to save, the macro can refer to the array
4428 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4429 @var{r} is used anywhere within the function. This implies the function
4430 prologue should save register @var{r}, provided it is not one of the
4431 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4432 @code{regs_ever_live}.)
4434 On machines that have ``register windows'', the function entry code does
4435 not save on the stack the registers that are in the windows, even if
4436 they are supposed to be preserved by function calls; instead it takes
4437 appropriate steps to ``push'' the register stack, if any non-call-used
4438 registers are used in the function.
4440 @findex frame_pointer_needed
4441 On machines where functions may or may not have frame-pointers, the
4442 function entry code must vary accordingly; it must set up the frame
4443 pointer if one is wanted, and not otherwise. To determine whether a
4444 frame pointer is in wanted, the macro can refer to the variable
4445 @code{frame_pointer_needed}. The variable's value will be 1 at run
4446 time in a function that needs a frame pointer. @xref{Elimination}.
4448 The function entry code is responsible for allocating any stack space
4449 required for the function. This stack space consists of the regions
4450 listed below. In most cases, these regions are allocated in the
4451 order listed, with the last listed region closest to the top of the
4452 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4453 the highest address if it is not defined). You can use a different order
4454 for a machine if doing so is more convenient or required for
4455 compatibility reasons. Except in cases where required by standard
4456 or by a debugger, there is no reason why the stack layout used by GCC
4457 need agree with that used by other compilers for a machine.
4460 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4461 If defined, a function that outputs assembler code at the end of a
4462 prologue. This should be used when the function prologue is being
4463 emitted as RTL, and you have some extra assembler that needs to be
4464 emitted. @xref{prologue instruction pattern}.
4467 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4468 If defined, a function that outputs assembler code at the start of an
4469 epilogue. This should be used when the function epilogue is being
4470 emitted as RTL, and you have some extra assembler that needs to be
4471 emitted. @xref{epilogue instruction pattern}.
4474 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4475 If defined, a function that outputs the assembler code for exit from a
4476 function. The epilogue is responsible for restoring the saved
4477 registers and stack pointer to their values when the function was
4478 called, and returning control to the caller. This macro takes the
4479 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4480 registers to restore are determined from @code{regs_ever_live} and
4481 @code{CALL_USED_REGISTERS} in the same way.
4483 On some machines, there is a single instruction that does all the work
4484 of returning from the function. On these machines, give that
4485 instruction the name @samp{return} and do not define the macro
4486 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4488 Do not define a pattern named @samp{return} if you want the
4489 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4490 switches to control whether return instructions or epilogues are used,
4491 define a @samp{return} pattern with a validity condition that tests the
4492 target switches appropriately. If the @samp{return} pattern's validity
4493 condition is false, epilogues will be used.
4495 On machines where functions may or may not have frame-pointers, the
4496 function exit code must vary accordingly. Sometimes the code for these
4497 two cases is completely different. To determine whether a frame pointer
4498 is wanted, the macro can refer to the variable
4499 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4500 a function that needs a frame pointer.
4502 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4503 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4504 The C variable @code{current_function_is_leaf} is nonzero for such a
4505 function. @xref{Leaf Functions}.
4507 On some machines, some functions pop their arguments on exit while
4508 others leave that for the caller to do. For example, the 68020 when
4509 given @option{-mrtd} pops arguments in functions that take a fixed
4510 number of arguments.
4512 @findex current_function_pops_args
4513 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4514 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4515 needs to know what was decided. The variable that is called
4516 @code{current_function_pops_args} is the number of bytes of its
4517 arguments that a function should pop. @xref{Scalar Return}.
4518 @c what is the "its arguments" in the above sentence referring to, pray
4519 @c tell? --mew 5feb93
4524 @findex current_function_pretend_args_size
4525 A region of @code{current_function_pretend_args_size} bytes of
4526 uninitialized space just underneath the first argument arriving on the
4527 stack. (This may not be at the very start of the allocated stack region
4528 if the calling sequence has pushed anything else since pushing the stack
4529 arguments. But usually, on such machines, nothing else has been pushed
4530 yet, because the function prologue itself does all the pushing.) This
4531 region is used on machines where an argument may be passed partly in
4532 registers and partly in memory, and, in some cases to support the
4533 features in @code{<stdarg.h>}.
4536 An area of memory used to save certain registers used by the function.
4537 The size of this area, which may also include space for such things as
4538 the return address and pointers to previous stack frames, is
4539 machine-specific and usually depends on which registers have been used
4540 in the function. Machines with register windows often do not require
4544 A region of at least @var{size} bytes, possibly rounded up to an allocation
4545 boundary, to contain the local variables of the function. On some machines,
4546 this region and the save area may occur in the opposite order, with the
4547 save area closer to the top of the stack.
4550 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4551 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4552 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4553 argument lists of the function. @xref{Stack Arguments}.
4556 @defmac EXIT_IGNORE_STACK
4557 Define this macro as a C expression that is nonzero if the return
4558 instruction or the function epilogue ignores the value of the stack
4559 pointer; in other words, if it is safe to delete an instruction to
4560 adjust the stack pointer before a return from the function. The
4563 Note that this macro's value is relevant only for functions for which
4564 frame pointers are maintained. It is never safe to delete a final
4565 stack adjustment in a function that has no frame pointer, and the
4566 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4569 @defmac EPILOGUE_USES (@var{regno})
4570 Define this macro as a C expression that is nonzero for registers that are
4571 used by the epilogue or the @samp{return} pattern. The stack and frame
4572 pointer registers are already assumed to be used as needed.
4575 @defmac EH_USES (@var{regno})
4576 Define this macro as a C expression that is nonzero for registers that are
4577 used by the exception handling mechanism, and so should be considered live
4578 on entry to an exception edge.
4581 @defmac DELAY_SLOTS_FOR_EPILOGUE
4582 Define this macro if the function epilogue contains delay slots to which
4583 instructions from the rest of the function can be ``moved''. The
4584 definition should be a C expression whose value is an integer
4585 representing the number of delay slots there.
4588 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4589 A C expression that returns 1 if @var{insn} can be placed in delay
4590 slot number @var{n} of the epilogue.
4592 The argument @var{n} is an integer which identifies the delay slot now
4593 being considered (since different slots may have different rules of
4594 eligibility). It is never negative and is always less than the number
4595 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4596 If you reject a particular insn for a given delay slot, in principle, it
4597 may be reconsidered for a subsequent delay slot. Also, other insns may
4598 (at least in principle) be considered for the so far unfilled delay
4601 @findex current_function_epilogue_delay_list
4602 @findex final_scan_insn
4603 The insns accepted to fill the epilogue delay slots are put in an RTL
4604 list made with @code{insn_list} objects, stored in the variable
4605 @code{current_function_epilogue_delay_list}. The insn for the first
4606 delay slot comes first in the list. Your definition of the macro
4607 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4608 outputting the insns in this list, usually by calling
4609 @code{final_scan_insn}.
4611 You need not define this macro if you did not define
4612 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4615 @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})
4616 A function that outputs the assembler code for a thunk
4617 function, used to implement C++ virtual function calls with multiple
4618 inheritance. The thunk acts as a wrapper around a virtual function,
4619 adjusting the implicit object parameter before handing control off to
4622 First, emit code to add the integer @var{delta} to the location that
4623 contains the incoming first argument. Assume that this argument
4624 contains a pointer, and is the one used to pass the @code{this} pointer
4625 in C++. This is the incoming argument @emph{before} the function prologue,
4626 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4627 all other incoming arguments.
4629 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4630 made after adding @code{delta}. In particular, if @var{p} is the
4631 adjusted pointer, the following adjustment should be made:
4634 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4637 After the additions, emit code to jump to @var{function}, which is a
4638 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4639 not touch the return address. Hence returning from @var{FUNCTION} will
4640 return to whoever called the current @samp{thunk}.
4642 The effect must be as if @var{function} had been called directly with
4643 the adjusted first argument. This macro is responsible for emitting all
4644 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4645 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4647 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4648 have already been extracted from it.) It might possibly be useful on
4649 some targets, but probably not.
4651 If you do not define this macro, the target-independent code in the C++
4652 front end will generate a less efficient heavyweight thunk that calls
4653 @var{function} instead of jumping to it. The generic approach does
4654 not support varargs.
4657 @deftypefn {Target Hook} bool TARGET_ASM_CAN_OUTPUT_MI_THUNK (tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, HOST_WIDE_INT @var{vcall_offset}, tree @var{function})
4658 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4659 to output the assembler code for the thunk function specified by the
4660 arguments it is passed, and false otherwise. In the latter case, the
4661 generic approach will be used by the C++ front end, with the limitations
4666 @subsection Generating Code for Profiling
4667 @cindex profiling, code generation
4669 These macros will help you generate code for profiling.
4671 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4672 A C statement or compound statement to output to @var{file} some
4673 assembler code to call the profiling subroutine @code{mcount}.
4676 The details of how @code{mcount} expects to be called are determined by
4677 your operating system environment, not by GCC@. To figure them out,
4678 compile a small program for profiling using the system's installed C
4679 compiler and look at the assembler code that results.
4681 Older implementations of @code{mcount} expect the address of a counter
4682 variable to be loaded into some register. The name of this variable is
4683 @samp{LP} followed by the number @var{labelno}, so you would generate
4684 the name using @samp{LP%d} in a @code{fprintf}.
4687 @defmac PROFILE_HOOK
4688 A C statement or compound statement to output to @var{file} some assembly
4689 code to call the profiling subroutine @code{mcount} even the target does
4690 not support profiling.
4693 @defmac NO_PROFILE_COUNTERS
4694 Define this macro to be an expression with a nonzero value if the
4695 @code{mcount} subroutine on your system does not need a counter variable
4696 allocated for each function. This is true for almost all modern
4697 implementations. If you define this macro, you must not use the
4698 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4701 @defmac PROFILE_BEFORE_PROLOGUE
4702 Define this macro if the code for function profiling should come before
4703 the function prologue. Normally, the profiling code comes after.
4707 @subsection Permitting tail calls
4710 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4711 True if it is ok to do sibling call optimization for the specified
4712 call expression @var{exp}. @var{decl} will be the called function,
4713 or @code{NULL} if this is an indirect call.
4715 It is not uncommon for limitations of calling conventions to prevent
4716 tail calls to functions outside the current unit of translation, or
4717 during PIC compilation. The hook is used to enforce these restrictions,
4718 as the @code{sibcall} md pattern can not fail, or fall over to a
4719 ``normal'' call. The criteria for successful sibling call optimization
4720 may vary greatly between different architectures.
4723 @deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap *@var{regs})
4724 Add any hard registers to @var{regs} that are live on entry to the
4725 function. This hook only needs to be defined to provide registers that
4726 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4727 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4728 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4729 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4732 @node Stack Smashing Protection
4733 @subsection Stack smashing protection
4734 @cindex stack smashing protection
4736 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void)
4737 This hook returns a @code{DECL} node for the external variable to use
4738 for the stack protection guard. This variable is initialized by the
4739 runtime to some random value and is used to initialize the guard value
4740 that is placed at the top of the local stack frame. The type of this
4741 variable must be @code{ptr_type_node}.
4743 The default version of this hook creates a variable called
4744 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
4747 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void)
4748 This hook returns a tree expression that alerts the runtime that the
4749 stack protect guard variable has been modified. This expression should
4750 involve a call to a @code{noreturn} function.
4752 The default version of this hook invokes a function called
4753 @samp{__stack_chk_fail}, taking no arguments. This function is
4754 normally defined in @file{libgcc2.c}.
4758 @section Implementing the Varargs Macros
4759 @cindex varargs implementation
4761 GCC comes with an implementation of @code{<varargs.h>} and
4762 @code{<stdarg.h>} that work without change on machines that pass arguments
4763 on the stack. Other machines require their own implementations of
4764 varargs, and the two machine independent header files must have
4765 conditionals to include it.
4767 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4768 the calling convention for @code{va_start}. The traditional
4769 implementation takes just one argument, which is the variable in which
4770 to store the argument pointer. The ISO implementation of
4771 @code{va_start} takes an additional second argument. The user is
4772 supposed to write the last named argument of the function here.
4774 However, @code{va_start} should not use this argument. The way to find
4775 the end of the named arguments is with the built-in functions described
4778 @defmac __builtin_saveregs ()
4779 Use this built-in function to save the argument registers in memory so
4780 that the varargs mechanism can access them. Both ISO and traditional
4781 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4782 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4784 On some machines, @code{__builtin_saveregs} is open-coded under the
4785 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4786 other machines, it calls a routine written in assembler language,
4787 found in @file{libgcc2.c}.
4789 Code generated for the call to @code{__builtin_saveregs} appears at the
4790 beginning of the function, as opposed to where the call to
4791 @code{__builtin_saveregs} is written, regardless of what the code is.
4792 This is because the registers must be saved before the function starts
4793 to use them for its own purposes.
4794 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4798 @defmac __builtin_args_info (@var{category})
4799 Use this built-in function to find the first anonymous arguments in
4802 In general, a machine may have several categories of registers used for
4803 arguments, each for a particular category of data types. (For example,
4804 on some machines, floating-point registers are used for floating-point
4805 arguments while other arguments are passed in the general registers.)
4806 To make non-varargs functions use the proper calling convention, you
4807 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4808 registers in each category have been used so far
4810 @code{__builtin_args_info} accesses the same data structure of type
4811 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4812 with it, with @var{category} specifying which word to access. Thus, the
4813 value indicates the first unused register in a given category.
4815 Normally, you would use @code{__builtin_args_info} in the implementation
4816 of @code{va_start}, accessing each category just once and storing the
4817 value in the @code{va_list} object. This is because @code{va_list} will
4818 have to update the values, and there is no way to alter the
4819 values accessed by @code{__builtin_args_info}.
4822 @defmac __builtin_next_arg (@var{lastarg})
4823 This is the equivalent of @code{__builtin_args_info}, for stack
4824 arguments. It returns the address of the first anonymous stack
4825 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4826 returns the address of the location above the first anonymous stack
4827 argument. Use it in @code{va_start} to initialize the pointer for
4828 fetching arguments from the stack. Also use it in @code{va_start} to
4829 verify that the second parameter @var{lastarg} is the last named argument
4830 of the current function.
4833 @defmac __builtin_classify_type (@var{object})
4834 Since each machine has its own conventions for which data types are
4835 passed in which kind of register, your implementation of @code{va_arg}
4836 has to embody these conventions. The easiest way to categorize the
4837 specified data type is to use @code{__builtin_classify_type} together
4838 with @code{sizeof} and @code{__alignof__}.
4840 @code{__builtin_classify_type} ignores the value of @var{object},
4841 considering only its data type. It returns an integer describing what
4842 kind of type that is---integer, floating, pointer, structure, and so on.
4844 The file @file{typeclass.h} defines an enumeration that you can use to
4845 interpret the values of @code{__builtin_classify_type}.
4848 These machine description macros help implement varargs:
4850 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
4851 If defined, this hook produces the machine-specific code for a call to
4852 @code{__builtin_saveregs}. This code will be moved to the very
4853 beginning of the function, before any parameter access are made. The
4854 return value of this function should be an RTX that contains the value
4855 to use as the return of @code{__builtin_saveregs}.
4858 @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})
4859 This target hook offers an alternative to using
4860 @code{__builtin_saveregs} and defining the hook
4861 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
4862 register arguments into the stack so that all the arguments appear to
4863 have been passed consecutively on the stack. Once this is done, you can
4864 use the standard implementation of varargs that works for machines that
4865 pass all their arguments on the stack.
4867 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
4868 structure, containing the values that are obtained after processing the
4869 named arguments. The arguments @var{mode} and @var{type} describe the
4870 last named argument---its machine mode and its data type as a tree node.
4872 The target hook should do two things: first, push onto the stack all the
4873 argument registers @emph{not} used for the named arguments, and second,
4874 store the size of the data thus pushed into the @code{int}-valued
4875 variable pointed to by @var{pretend_args_size}. The value that you
4876 store here will serve as additional offset for setting up the stack
4879 Because you must generate code to push the anonymous arguments at
4880 compile time without knowing their data types,
4881 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
4882 have just a single category of argument register and use it uniformly
4885 If the argument @var{second_time} is nonzero, it means that the
4886 arguments of the function are being analyzed for the second time. This
4887 happens for an inline function, which is not actually compiled until the
4888 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
4889 not generate any instructions in this case.
4892 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca})
4893 Define this hook to return @code{true} if the location where a function
4894 argument is passed depends on whether or not it is a named argument.
4896 This hook controls how the @var{named} argument to @code{FUNCTION_ARG}
4897 is set for varargs and stdarg functions. If this hook returns
4898 @code{true}, the @var{named} argument is always true for named
4899 arguments, and false for unnamed arguments. If it returns @code{false},
4900 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
4901 then all arguments are treated as named. Otherwise, all named arguments
4902 except the last are treated as named.
4904 You need not define this hook if it always returns zero.
4907 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED
4908 If you need to conditionally change ABIs so that one works with
4909 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
4910 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
4911 defined, then define this hook to return @code{true} if
4912 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
4913 Otherwise, you should not define this hook.
4917 @section Trampolines for Nested Functions
4918 @cindex trampolines for nested functions
4919 @cindex nested functions, trampolines for
4921 A @dfn{trampoline} is a small piece of code that is created at run time
4922 when the address of a nested function is taken. It normally resides on
4923 the stack, in the stack frame of the containing function. These macros
4924 tell GCC how to generate code to allocate and initialize a
4927 The instructions in the trampoline must do two things: load a constant
4928 address into the static chain register, and jump to the real address of
4929 the nested function. On CISC machines such as the m68k, this requires
4930 two instructions, a move immediate and a jump. Then the two addresses
4931 exist in the trampoline as word-long immediate operands. On RISC
4932 machines, it is often necessary to load each address into a register in
4933 two parts. Then pieces of each address form separate immediate
4936 The code generated to initialize the trampoline must store the variable
4937 parts---the static chain value and the function address---into the
4938 immediate operands of the instructions. On a CISC machine, this is
4939 simply a matter of copying each address to a memory reference at the
4940 proper offset from the start of the trampoline. On a RISC machine, it
4941 may be necessary to take out pieces of the address and store them
4944 @defmac TRAMPOLINE_TEMPLATE (@var{file})
4945 A C statement to output, on the stream @var{file}, assembler code for a
4946 block of data that contains the constant parts of a trampoline. This
4947 code should not include a label---the label is taken care of
4950 If you do not define this macro, it means no template is needed
4951 for the target. Do not define this macro on systems where the block move
4952 code to copy the trampoline into place would be larger than the code
4953 to generate it on the spot.
4956 @defmac TRAMPOLINE_SECTION
4957 Return the section into which the trampoline template is to be placed
4958 (@pxref{Sections}). The default value is @code{readonly_data_section}.
4961 @defmac TRAMPOLINE_SIZE
4962 A C expression for the size in bytes of the trampoline, as an integer.
4965 @defmac TRAMPOLINE_ALIGNMENT
4966 Alignment required for trampolines, in bits.
4968 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4969 is used for aligning trampolines.
4972 @defmac INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4973 A C statement to initialize the variable parts of a trampoline.
4974 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4975 an RTX for the address of the nested function; @var{static_chain} is an
4976 RTX for the static chain value that should be passed to the function
4980 @defmac TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4981 A C statement that should perform any machine-specific adjustment in
4982 the address of the trampoline. Its argument contains the address that
4983 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4984 used for a function call should be different from the address in which
4985 the template was stored, the different address should be assigned to
4986 @var{addr}. If this macro is not defined, @var{addr} will be used for
4989 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4990 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4991 If this macro is not defined, by default the trampoline is allocated as
4992 a stack slot. This default is right for most machines. The exceptions
4993 are machines where it is impossible to execute instructions in the stack
4994 area. On such machines, you may have to implement a separate stack,
4995 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4996 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4998 @var{fp} points to a data structure, a @code{struct function}, which
4999 describes the compilation status of the immediate containing function of
5000 the function which the trampoline is for. The stack slot for the
5001 trampoline is in the stack frame of this containing function. Other
5002 allocation strategies probably must do something analogous with this
5006 Implementing trampolines is difficult on many machines because they have
5007 separate instruction and data caches. Writing into a stack location
5008 fails to clear the memory in the instruction cache, so when the program
5009 jumps to that location, it executes the old contents.
5011 Here are two possible solutions. One is to clear the relevant parts of
5012 the instruction cache whenever a trampoline is set up. The other is to
5013 make all trampolines identical, by having them jump to a standard
5014 subroutine. The former technique makes trampoline execution faster; the
5015 latter makes initialization faster.
5017 To clear the instruction cache when a trampoline is initialized, define
5018 the following macro.
5020 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
5021 If defined, expands to a C expression clearing the @emph{instruction
5022 cache} in the specified interval. The definition of this macro would
5023 typically be a series of @code{asm} statements. Both @var{beg} and
5024 @var{end} are both pointer expressions.
5027 The operating system may also require the stack to be made executable
5028 before calling the trampoline. To implement this requirement, define
5029 the following macro.
5031 @defmac ENABLE_EXECUTE_STACK
5032 Define this macro if certain operations must be performed before executing
5033 code located on the stack. The macro should expand to a series of C
5034 file-scope constructs (e.g.@: functions) and provide a unique entry point
5035 named @code{__enable_execute_stack}. The target is responsible for
5036 emitting calls to the entry point in the code, for example from the
5037 @code{INITIALIZE_TRAMPOLINE} macro.
5040 To use a standard subroutine, define the following macro. In addition,
5041 you must make sure that the instructions in a trampoline fill an entire
5042 cache line with identical instructions, or else ensure that the
5043 beginning of the trampoline code is always aligned at the same point in
5044 its cache line. Look in @file{m68k.h} as a guide.
5046 @defmac TRANSFER_FROM_TRAMPOLINE
5047 Define this macro if trampolines need a special subroutine to do their
5048 work. The macro should expand to a series of @code{asm} statements
5049 which will be compiled with GCC@. They go in a library function named
5050 @code{__transfer_from_trampoline}.
5052 If you need to avoid executing the ordinary prologue code of a compiled
5053 C function when you jump to the subroutine, you can do so by placing a
5054 special label of your own in the assembler code. Use one @code{asm}
5055 statement to generate an assembler label, and another to make the label
5056 global. Then trampolines can use that label to jump directly to your
5057 special assembler code.
5061 @section Implicit Calls to Library Routines
5062 @cindex library subroutine names
5063 @cindex @file{libgcc.a}
5065 @c prevent bad page break with this line
5066 Here is an explanation of implicit calls to library routines.
5068 @defmac DECLARE_LIBRARY_RENAMES
5069 This macro, if defined, should expand to a piece of C code that will get
5070 expanded when compiling functions for libgcc.a. It can be used to
5071 provide alternate names for GCC's internal library functions if there
5072 are ABI-mandated names that the compiler should provide.
5075 @findex init_one_libfunc
5076 @findex set_optab_libfunc
5077 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
5078 This hook should declare additional library routines or rename
5079 existing ones, using the functions @code{set_optab_libfunc} and
5080 @code{init_one_libfunc} defined in @file{optabs.c}.
5081 @code{init_optabs} calls this macro after initializing all the normal
5084 The default is to do nothing. Most ports don't need to define this hook.
5087 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
5088 This macro should return @code{true} if the library routine that
5089 implements the floating point comparison operator @var{comparison} in
5090 mode @var{mode} will return a boolean, and @var{false} if it will
5093 GCC's own floating point libraries return tristates from the
5094 comparison operators, so the default returns false always. Most ports
5095 don't need to define this macro.
5098 @defmac TARGET_LIB_INT_CMP_BIASED
5099 This macro should evaluate to @code{true} if the integer comparison
5100 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5101 operand is smaller than the second, 1 to indicate that they are equal,
5102 and 2 to indicate that the first operand is greater than the second.
5103 If this macro evaluates to @code{false} the comparison functions return
5104 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5105 in @file{libgcc.a}, you do not need to define this macro.
5108 @cindex US Software GOFAST, floating point emulation library
5109 @cindex floating point emulation library, US Software GOFAST
5110 @cindex GOFAST, floating point emulation library
5111 @findex gofast_maybe_init_libfuncs
5112 @defmac US_SOFTWARE_GOFAST
5113 Define this macro if your system C library uses the US Software GOFAST
5114 library to provide floating point emulation.
5116 In addition to defining this macro, your architecture must set
5117 @code{TARGET_INIT_LIBFUNCS} to @code{gofast_maybe_init_libfuncs}, or
5118 else call that function from its version of that hook. It is defined
5119 in @file{config/gofast.h}, which must be included by your
5120 architecture's @file{@var{cpu}.c} file. See @file{sparc/sparc.c} for
5123 If this macro is defined, the
5124 @code{TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL} target hook must return
5125 false for @code{SFmode} and @code{DFmode} comparisons.
5128 @cindex @code{EDOM}, implicit usage
5131 The value of @code{EDOM} on the target machine, as a C integer constant
5132 expression. If you don't define this macro, GCC does not attempt to
5133 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5134 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5137 If you do not define @code{TARGET_EDOM}, then compiled code reports
5138 domain errors by calling the library function and letting it report the
5139 error. If mathematical functions on your system use @code{matherr} when
5140 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5141 that @code{matherr} is used normally.
5144 @cindex @code{errno}, implicit usage
5145 @defmac GEN_ERRNO_RTX
5146 Define this macro as a C expression to create an rtl expression that
5147 refers to the global ``variable'' @code{errno}. (On certain systems,
5148 @code{errno} may not actually be a variable.) If you don't define this
5149 macro, a reasonable default is used.
5152 @cindex C99 math functions, implicit usage
5153 @defmac TARGET_C99_FUNCTIONS
5154 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
5155 @code{sinf} and similarly for other functions defined by C99 standard. The
5156 default is nonzero that should be proper value for most modern systems, however
5157 number of existing systems lacks support for these functions in the runtime so
5158 they needs this macro to be redefined to 0.
5161 @cindex sincos math function, implicit usage
5162 @defmac TARGET_HAS_SINCOS
5163 When this macro is nonzero, GCC will implicitly optimize calls to @code{sin}
5164 and @code{cos} with the same argument to a call to @code{sincos}. The
5165 default is zero. The target has to provide the following functions:
5167 void sincos(double x, double *sin, double *cos);
5168 void sincosf(float x, float *sin, float *cos);
5169 void sincosl(long double x, long double *sin, long double *cos);
5173 @defmac NEXT_OBJC_RUNTIME
5174 Define this macro to generate code for Objective-C message sending using
5175 the calling convention of the NeXT system. This calling convention
5176 involves passing the object, the selector and the method arguments all
5177 at once to the method-lookup library function.
5179 The default calling convention passes just the object and the selector
5180 to the lookup function, which returns a pointer to the method.
5183 @node Addressing Modes
5184 @section Addressing Modes
5185 @cindex addressing modes
5187 @c prevent bad page break with this line
5188 This is about addressing modes.
5190 @defmac HAVE_PRE_INCREMENT
5191 @defmacx HAVE_PRE_DECREMENT
5192 @defmacx HAVE_POST_INCREMENT
5193 @defmacx HAVE_POST_DECREMENT
5194 A C expression that is nonzero if the machine supports pre-increment,
5195 pre-decrement, post-increment, or post-decrement addressing respectively.
5198 @defmac HAVE_PRE_MODIFY_DISP
5199 @defmacx HAVE_POST_MODIFY_DISP
5200 A C expression that is nonzero if the machine supports pre- or
5201 post-address side-effect generation involving constants other than
5202 the size of the memory operand.
5205 @defmac HAVE_PRE_MODIFY_REG
5206 @defmacx HAVE_POST_MODIFY_REG
5207 A C expression that is nonzero if the machine supports pre- or
5208 post-address side-effect generation involving a register displacement.
5211 @defmac CONSTANT_ADDRESS_P (@var{x})
5212 A C expression that is 1 if the RTX @var{x} is a constant which
5213 is a valid address. On most machines, this can be defined as
5214 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
5215 in which constant addresses are supported.
5218 @defmac CONSTANT_P (@var{x})
5219 @code{CONSTANT_P}, which is defined by target-independent code,
5220 accepts integer-values expressions whose values are not explicitly
5221 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5222 expressions and @code{const} arithmetic expressions, in addition to
5223 @code{const_int} and @code{const_double} expressions.
5226 @defmac MAX_REGS_PER_ADDRESS
5227 A number, the maximum number of registers that can appear in a valid
5228 memory address. Note that it is up to you to specify a value equal to
5229 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
5233 @defmac GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5234 A C compound statement with a conditional @code{goto @var{label};}
5235 executed if @var{x} (an RTX) is a legitimate memory address on the
5236 target machine for a memory operand of mode @var{mode}.
5238 It usually pays to define several simpler macros to serve as
5239 subroutines for this one. Otherwise it may be too complicated to
5242 This macro must exist in two variants: a strict variant and a
5243 non-strict one. The strict variant is used in the reload pass. It
5244 must be defined so that any pseudo-register that has not been
5245 allocated a hard register is considered a memory reference. In
5246 contexts where some kind of register is required, a pseudo-register
5247 with no hard register must be rejected.
5249 The non-strict variant is used in other passes. It must be defined to
5250 accept all pseudo-registers in every context where some kind of
5251 register is required.
5253 @findex REG_OK_STRICT
5254 Compiler source files that want to use the strict variant of this
5255 macro define the macro @code{REG_OK_STRICT}. You should use an
5256 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
5257 in that case and the non-strict variant otherwise.
5259 Subroutines to check for acceptable registers for various purposes (one
5260 for base registers, one for index registers, and so on) are typically
5261 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
5262 Then only these subroutine macros need have two variants; the higher
5263 levels of macros may be the same whether strict or not.
5265 Normally, constant addresses which are the sum of a @code{symbol_ref}
5266 and an integer are stored inside a @code{const} RTX to mark them as
5267 constant. Therefore, there is no need to recognize such sums
5268 specifically as legitimate addresses. Normally you would simply
5269 recognize any @code{const} as legitimate.
5271 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5272 sums that are not marked with @code{const}. It assumes that a naked
5273 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5274 naked constant sums as illegitimate addresses, so that none of them will
5275 be given to @code{PRINT_OPERAND_ADDRESS}.
5277 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5278 On some machines, whether a symbolic address is legitimate depends on
5279 the section that the address refers to. On these machines, define the
5280 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5281 into the @code{symbol_ref}, and then check for it here. When you see a
5282 @code{const}, you will have to look inside it to find the
5283 @code{symbol_ref} in order to determine the section. @xref{Assembler
5287 @defmac FIND_BASE_TERM (@var{x})
5288 A C expression to determine the base term of address @var{x}.
5289 This macro is used in only one place: `find_base_term' in alias.c.
5291 It is always safe for this macro to not be defined. It exists so
5292 that alias analysis can understand machine-dependent addresses.
5294 The typical use of this macro is to handle addresses containing
5295 a label_ref or symbol_ref within an UNSPEC@.
5298 @defmac LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
5299 A C compound statement that attempts to replace @var{x} with a valid
5300 memory address for an operand of mode @var{mode}. @var{win} will be a
5301 C statement label elsewhere in the code; the macro definition may use
5304 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
5308 to avoid further processing if the address has become legitimate.
5310 @findex break_out_memory_refs
5311 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5312 and @var{oldx} will be the operand that was given to that function to produce
5315 The code generated by this macro should not alter the substructure of
5316 @var{x}. If it transforms @var{x} into a more legitimate form, it
5317 should assign @var{x} (which will always be a C variable) a new value.
5319 It is not necessary for this macro to come up with a legitimate
5320 address. The compiler has standard ways of doing so in all cases. In
5321 fact, it is safe to omit this macro. But often a
5322 machine-dependent strategy can generate better code.
5325 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5326 A C compound statement that attempts to replace @var{x}, which is an address
5327 that needs reloading, with a valid memory address for an operand of mode
5328 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5329 It is not necessary to define this macro, but it might be useful for
5330 performance reasons.
5332 For example, on the i386, it is sometimes possible to use a single
5333 reload register instead of two by reloading a sum of two pseudo
5334 registers into a register. On the other hand, for number of RISC
5335 processors offsets are limited so that often an intermediate address
5336 needs to be generated in order to address a stack slot. By defining
5337 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5338 generated for adjacent some stack slots can be made identical, and thus
5341 @emph{Note}: This macro should be used with caution. It is necessary
5342 to know something of how reload works in order to effectively use this,
5343 and it is quite easy to produce macros that build in too much knowledge
5344 of reload internals.
5346 @emph{Note}: This macro must be able to reload an address created by a
5347 previous invocation of this macro. If it fails to handle such addresses
5348 then the compiler may generate incorrect code or abort.
5351 The macro definition should use @code{push_reload} to indicate parts that
5352 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5353 suitable to be passed unaltered to @code{push_reload}.
5355 The code generated by this macro must not alter the substructure of
5356 @var{x}. If it transforms @var{x} into a more legitimate form, it
5357 should assign @var{x} (which will always be a C variable) a new value.
5358 This also applies to parts that you change indirectly by calling
5361 @findex strict_memory_address_p
5362 The macro definition may use @code{strict_memory_address_p} to test if
5363 the address has become legitimate.
5366 If you want to change only a part of @var{x}, one standard way of doing
5367 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5368 single level of rtl. Thus, if the part to be changed is not at the
5369 top level, you'll need to replace first the top level.
5370 It is not necessary for this macro to come up with a legitimate
5371 address; but often a machine-dependent strategy can generate better code.
5374 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5375 A C statement or compound statement with a conditional @code{goto
5376 @var{label};} executed if memory address @var{x} (an RTX) can have
5377 different meanings depending on the machine mode of the memory
5378 reference it is used for or if the address is valid for some modes
5381 Autoincrement and autodecrement addresses typically have mode-dependent
5382 effects because the amount of the increment or decrement is the size
5383 of the operand being addressed. Some machines have other mode-dependent
5384 addresses. Many RISC machines have no mode-dependent addresses.
5386 You may assume that @var{addr} is a valid address for the machine.
5389 @defmac LEGITIMATE_CONSTANT_P (@var{x})
5390 A C expression that is nonzero if @var{x} is a legitimate constant for
5391 an immediate operand on the target machine. You can assume that
5392 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5393 @samp{1} is a suitable definition for this macro on machines where
5394 anything @code{CONSTANT_P} is valid.
5397 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5398 This hook is used to undo the possibly obfuscating effects of the
5399 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5400 macros. Some backend implementations of these macros wrap symbol
5401 references inside an @code{UNSPEC} rtx to represent PIC or similar
5402 addressing modes. This target hook allows GCC's optimizers to understand
5403 the semantics of these opaque @code{UNSPEC}s by converting them back
5404 into their original form.
5407 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (rtx @var{x})
5408 This hook should return true if @var{x} is of a form that cannot (or
5409 should not) be spilled to the constant pool. The default version of
5410 this hook returns false.
5412 The primary reason to define this hook is to prevent reload from
5413 deciding that a non-legitimate constant would be better reloaded
5414 from the constant pool instead of spilling and reloading a register
5415 holding the constant. This restriction is often true of addresses
5416 of TLS symbols for various targets.
5419 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (enum machine_mode @var{mode}, rtx @var{x})
5420 This hook should return true if pool entries for constant @var{x} can
5421 be placed in an @code{object_block} structure. @var{mode} is the mode
5424 The default version returns false for all constants.
5427 @deftypefn {Target Hook} tree TARGET_BUILTIN_RECIPROCAL (enum tree_code @var{fn}, bool @var{tm_fn}, bool @var{sqrt})
5428 This hook should return the DECL of a function that implements reciprocal of
5429 the builtin function with builtin function code @var{fn}, or
5430 @code{NULL_TREE} if such a function is not available. @var{tm_fn} is true
5431 when @var{fn} is a code of a machine-dependent builtin function. When
5432 @var{sqrt} is true, additional optimizations that apply only to the reciprocal
5433 of a square root function are performed, and only reciprocals of @code{sqrt}
5437 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5438 This hook should return the DECL of a function @var{f} that given an
5439 address @var{addr} as an argument returns a mask @var{m} that can be
5440 used to extract from two vectors the relevant data that resides in
5441 @var{addr} in case @var{addr} is not properly aligned.
5443 The autovectorizer, when vectorizing a load operation from an address
5444 @var{addr} that may be unaligned, will generate two vector loads from
5445 the two aligned addresses around @var{addr}. It then generates a
5446 @code{REALIGN_LOAD} operation to extract the relevant data from the
5447 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5448 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5449 the third argument, @var{OFF}, defines how the data will be extracted
5450 from these two vectors: if @var{OFF} is 0, then the returned vector is
5451 @var{v2}; otherwise, the returned vector is composed from the last
5452 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5453 @var{OFF} elements of @var{v2}.
5455 If this hook is defined, the autovectorizer will generate a call
5456 to @var{f} (using the DECL tree that this hook returns) and will
5457 use the return value of @var{f} as the argument @var{OFF} to
5458 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5459 should comply with the semantics expected by @code{REALIGN_LOAD}
5461 If this hook is not defined, then @var{addr} will be used as
5462 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5463 log2(@var{VS})-1 bits of @var{addr} will be considered.
5466 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN (tree @var{x})
5467 This hook should return the DECL of a function @var{f} that implements
5468 widening multiplication of the even elements of two input vectors of type @var{x}.
5470 If this hook is defined, the autovectorizer will use it along with the
5471 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD} target hook when vectorizing
5472 widening multiplication in cases that the order of the results does not have to be
5473 preserved (e.g. used only by a reduction computation). Otherwise, the
5474 @code{widen_mult_hi/lo} idioms will be used.
5477 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD (tree @var{x})
5478 This hook should return the DECL of a function @var{f} that implements
5479 widening multiplication of the odd elements of two input vectors of type @var{x}.
5481 If this hook is defined, the autovectorizer will use it along with the
5482 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN} target hook when vectorizing
5483 widening multiplication in cases that the order of the results does not have to be
5484 preserved (e.g. used only by a reduction computation). Otherwise, the
5485 @code{widen_mult_hi/lo} idioms will be used.
5488 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_CONVERSION (enum tree_code @var{code}, tree @var{type})
5489 This hook should return the DECL of a function that implements conversion of the
5490 input vector of type @var{type}.
5491 If @var{type} is an integral type, the result of the conversion is a vector of
5492 floating-point type of the same size.
5493 If @var{type} is a floating-point type, the result of the conversion is a vector
5494 of integral type of the same size.
5495 @var{code} specifies how the conversion is to be applied
5496 (truncation, rounding, etc.).
5498 If this hook is defined, the autovectorizer will use the
5499 @code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing
5500 conversion. Otherwise, it will return @code{NULL_TREE}.
5503 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION (enum built_in_function @var{code}, tree @var{vec_type_out}, tree @var{vec_type_in})
5504 This hook should return the decl of a function that implements the vectorized
5505 variant of the builtin function with builtin function code @var{code} or
5506 @code{NULL_TREE} if such a function is not available. The return type of
5507 the vectorized function shall be of vector type @var{vec_type_out} and the
5508 argument types should be @var{vec_type_in}.
5511 @node Anchored Addresses
5512 @section Anchored Addresses
5513 @cindex anchored addresses
5514 @cindex @option{-fsection-anchors}
5516 GCC usually addresses every static object as a separate entity.
5517 For example, if we have:
5521 int foo (void) @{ return a + b + c; @}
5524 the code for @code{foo} will usually calculate three separate symbolic
5525 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
5526 it would be better to calculate just one symbolic address and access
5527 the three variables relative to it. The equivalent pseudocode would
5533 register int *xr = &x;
5534 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5538 (which isn't valid C). We refer to shared addresses like @code{x} as
5539 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
5541 The hooks below describe the target properties that GCC needs to know
5542 in order to make effective use of section anchors. It won't use
5543 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
5544 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
5546 @deftypevar {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET
5547 The minimum offset that should be applied to a section anchor.
5548 On most targets, it should be the smallest offset that can be
5549 applied to a base register while still giving a legitimate address
5550 for every mode. The default value is 0.
5553 @deftypevar {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET
5554 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
5555 offset that should be applied to section anchors. The default
5559 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x})
5560 Write the assembly code to define section anchor @var{x}, which is a
5561 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
5562 The hook is called with the assembly output position set to the beginning
5563 of @code{SYMBOL_REF_BLOCK (@var{x})}.
5565 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
5566 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
5567 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
5568 is @code{NULL}, which disables the use of section anchors altogether.
5571 @deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (rtx @var{x})
5572 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
5573 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
5574 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
5576 The default version is correct for most targets, but you might need to
5577 intercept this hook to handle things like target-specific attributes
5578 or target-specific sections.
5581 @node Condition Code
5582 @section Condition Code Status
5583 @cindex condition code status
5585 @c prevent bad page break with this line
5586 This describes the condition code status.
5589 The file @file{conditions.h} defines a variable @code{cc_status} to
5590 describe how the condition code was computed (in case the interpretation of
5591 the condition code depends on the instruction that it was set by). This
5592 variable contains the RTL expressions on which the condition code is
5593 currently based, and several standard flags.
5595 Sometimes additional machine-specific flags must be defined in the machine
5596 description header file. It can also add additional machine-specific
5597 information by defining @code{CC_STATUS_MDEP}.
5599 @defmac CC_STATUS_MDEP
5600 C code for a data type which is used for declaring the @code{mdep}
5601 component of @code{cc_status}. It defaults to @code{int}.
5603 This macro is not used on machines that do not use @code{cc0}.
5606 @defmac CC_STATUS_MDEP_INIT
5607 A C expression to initialize the @code{mdep} field to ``empty''.
5608 The default definition does nothing, since most machines don't use
5609 the field anyway. If you want to use the field, you should probably
5610 define this macro to initialize it.
5612 This macro is not used on machines that do not use @code{cc0}.
5615 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5616 A C compound statement to set the components of @code{cc_status}
5617 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5618 this macro's responsibility to recognize insns that set the condition
5619 code as a byproduct of other activity as well as those that explicitly
5622 This macro is not used on machines that do not use @code{cc0}.
5624 If there are insns that do not set the condition code but do alter
5625 other machine registers, this macro must check to see whether they
5626 invalidate the expressions that the condition code is recorded as
5627 reflecting. For example, on the 68000, insns that store in address
5628 registers do not set the condition code, which means that usually
5629 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5630 insns. But suppose that the previous insn set the condition code
5631 based on location @samp{a4@@(102)} and the current insn stores a new
5632 value in @samp{a4}. Although the condition code is not changed by
5633 this, it will no longer be true that it reflects the contents of
5634 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5635 @code{cc_status} in this case to say that nothing is known about the
5636 condition code value.
5638 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5639 with the results of peephole optimization: insns whose patterns are
5640 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5641 constants which are just the operands. The RTL structure of these
5642 insns is not sufficient to indicate what the insns actually do. What
5643 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5644 @code{CC_STATUS_INIT}.
5646 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5647 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5648 @samp{cc}. This avoids having detailed information about patterns in
5649 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5652 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5653 Returns a mode from class @code{MODE_CC} to be used when comparison
5654 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5655 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5656 @pxref{Jump Patterns} for a description of the reason for this
5660 #define SELECT_CC_MODE(OP,X,Y) \
5661 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5662 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5663 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5664 || GET_CODE (X) == NEG) \
5665 ? CC_NOOVmode : CCmode))
5668 You should define this macro if and only if you define extra CC modes
5669 in @file{@var{machine}-modes.def}.
5672 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5673 On some machines not all possible comparisons are defined, but you can
5674 convert an invalid comparison into a valid one. For example, the Alpha
5675 does not have a @code{GT} comparison, but you can use an @code{LT}
5676 comparison instead and swap the order of the operands.
5678 On such machines, define this macro to be a C statement to do any
5679 required conversions. @var{code} is the initial comparison code
5680 and @var{op0} and @var{op1} are the left and right operands of the
5681 comparison, respectively. You should modify @var{code}, @var{op0}, and
5682 @var{op1} as required.
5684 GCC will not assume that the comparison resulting from this macro is
5685 valid but will see if the resulting insn matches a pattern in the
5688 You need not define this macro if it would never change the comparison
5692 @defmac REVERSIBLE_CC_MODE (@var{mode})
5693 A C expression whose value is one if it is always safe to reverse a
5694 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5695 can ever return @var{mode} for a floating-point inequality comparison,
5696 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5698 You need not define this macro if it would always returns zero or if the
5699 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5700 For example, here is the definition used on the SPARC, where floating-point
5701 inequality comparisons are always given @code{CCFPEmode}:
5704 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5708 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
5709 A C expression whose value is reversed condition code of the @var{code} for
5710 comparison done in CC_MODE @var{mode}. The macro is used only in case
5711 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5712 machine has some non-standard way how to reverse certain conditionals. For
5713 instance in case all floating point conditions are non-trapping, compiler may
5714 freely convert unordered compares to ordered one. Then definition may look
5718 #define REVERSE_CONDITION(CODE, MODE) \
5719 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5720 : reverse_condition_maybe_unordered (CODE))
5724 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
5725 A C expression that returns true if the conditional execution predicate
5726 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
5727 versa. Define this to return 0 if the target has conditional execution
5728 predicates that cannot be reversed safely. There is no need to validate
5729 that the arguments of op1 and op2 are the same, this is done separately.
5730 If no expansion is specified, this macro is defined as follows:
5733 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5734 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
5738 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *, unsigned int *)
5739 On targets which do not use @code{(cc0)}, and which use a hard
5740 register rather than a pseudo-register to hold condition codes, the
5741 regular CSE passes are often not able to identify cases in which the
5742 hard register is set to a common value. Use this hook to enable a
5743 small pass which optimizes such cases. This hook should return true
5744 to enable this pass, and it should set the integers to which its
5745 arguments point to the hard register numbers used for condition codes.
5746 When there is only one such register, as is true on most systems, the
5747 integer pointed to by the second argument should be set to
5748 @code{INVALID_REGNUM}.
5750 The default version of this hook returns false.
5753 @deftypefn {Target Hook} enum machine_mode TARGET_CC_MODES_COMPATIBLE (enum machine_mode, enum machine_mode)
5754 On targets which use multiple condition code modes in class
5755 @code{MODE_CC}, it is sometimes the case that a comparison can be
5756 validly done in more than one mode. On such a system, define this
5757 target hook to take two mode arguments and to return a mode in which
5758 both comparisons may be validly done. If there is no such mode,
5759 return @code{VOIDmode}.
5761 The default version of this hook checks whether the modes are the
5762 same. If they are, it returns that mode. If they are different, it
5763 returns @code{VOIDmode}.
5767 @section Describing Relative Costs of Operations
5768 @cindex costs of instructions
5769 @cindex relative costs
5770 @cindex speed of instructions
5772 These macros let you describe the relative speed of various operations
5773 on the target machine.
5775 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5776 A C expression for the cost of moving data of mode @var{mode} from a
5777 register in class @var{from} to one in class @var{to}. The classes are
5778 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5779 value of 2 is the default; other values are interpreted relative to
5782 It is not required that the cost always equal 2 when @var{from} is the
5783 same as @var{to}; on some machines it is expensive to move between
5784 registers if they are not general registers.
5786 If reload sees an insn consisting of a single @code{set} between two
5787 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5788 classes returns a value of 2, reload does not check to ensure that the
5789 constraints of the insn are met. Setting a cost of other than 2 will
5790 allow reload to verify that the constraints are met. You should do this
5791 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5794 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5795 A C expression for the cost of moving data of mode @var{mode} between a
5796 register of class @var{class} and memory; @var{in} is zero if the value
5797 is to be written to memory, nonzero if it is to be read in. This cost
5798 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5799 registers and memory is more expensive than between two registers, you
5800 should define this macro to express the relative cost.
5802 If you do not define this macro, GCC uses a default cost of 4 plus
5803 the cost of copying via a secondary reload register, if one is
5804 needed. If your machine requires a secondary reload register to copy
5805 between memory and a register of @var{class} but the reload mechanism is
5806 more complex than copying via an intermediate, define this macro to
5807 reflect the actual cost of the move.
5809 GCC defines the function @code{memory_move_secondary_cost} if
5810 secondary reloads are needed. It computes the costs due to copying via
5811 a secondary register. If your machine copies from memory using a
5812 secondary register in the conventional way but the default base value of
5813 4 is not correct for your machine, define this macro to add some other
5814 value to the result of that function. The arguments to that function
5815 are the same as to this macro.
5819 A C expression for the cost of a branch instruction. A value of 1 is
5820 the default; other values are interpreted relative to that.
5823 Here are additional macros which do not specify precise relative costs,
5824 but only that certain actions are more expensive than GCC would
5827 @defmac SLOW_BYTE_ACCESS
5828 Define this macro as a C expression which is nonzero if accessing less
5829 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5830 faster than accessing a word of memory, i.e., if such access
5831 require more than one instruction or if there is no difference in cost
5832 between byte and (aligned) word loads.
5834 When this macro is not defined, the compiler will access a field by
5835 finding the smallest containing object; when it is defined, a fullword
5836 load will be used if alignment permits. Unless bytes accesses are
5837 faster than word accesses, using word accesses is preferable since it
5838 may eliminate subsequent memory access if subsequent accesses occur to
5839 other fields in the same word of the structure, but to different bytes.
5842 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5843 Define this macro to be the value 1 if memory accesses described by the
5844 @var{mode} and @var{alignment} parameters have a cost many times greater
5845 than aligned accesses, for example if they are emulated in a trap
5848 When this macro is nonzero, the compiler will act as if
5849 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5850 moves. This can cause significantly more instructions to be produced.
5851 Therefore, do not set this macro nonzero if unaligned accesses only add a
5852 cycle or two to the time for a memory access.
5854 If the value of this macro is always zero, it need not be defined. If
5855 this macro is defined, it should produce a nonzero value when
5856 @code{STRICT_ALIGNMENT} is nonzero.
5860 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5861 which a sequence of insns should be generated instead of a
5862 string move insn or a library call. Increasing the value will always
5863 make code faster, but eventually incurs high cost in increased code size.
5865 Note that on machines where the corresponding move insn is a
5866 @code{define_expand} that emits a sequence of insns, this macro counts
5867 the number of such sequences.
5869 If you don't define this, a reasonable default is used.
5872 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5873 A C expression used to determine whether @code{move_by_pieces} will be used to
5874 copy a chunk of memory, or whether some other block move mechanism
5875 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5876 than @code{MOVE_RATIO}.
5879 @defmac MOVE_MAX_PIECES
5880 A C expression used by @code{move_by_pieces} to determine the largest unit
5881 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5885 The threshold of number of scalar move insns, @emph{below} which a sequence
5886 of insns should be generated to clear memory instead of a string clear insn
5887 or a library call. Increasing the value will always make code faster, but
5888 eventually incurs high cost in increased code size.
5890 If you don't define this, a reasonable default is used.
5893 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5894 A C expression used to determine whether @code{clear_by_pieces} will be used
5895 to clear a chunk of memory, or whether some other block clear mechanism
5896 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5897 than @code{CLEAR_RATIO}.
5901 The threshold of number of scalar move insns, @emph{below} which a sequence
5902 of insns should be generated to set memory to a constant value, instead of
5903 a block set insn or a library call.
5904 Increasing the value will always make code faster, but
5905 eventually incurs high cost in increased code size.
5907 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
5910 @defmac SET_BY_PIECES_P (@var{size}, @var{alignment})
5911 A C expression used to determine whether @code{store_by_pieces} will be
5912 used to set a chunk of memory to a constant value, or whether some
5913 other mechanism will be used. Used by @code{__builtin_memset} when
5914 storing values other than constant zero.
5915 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5916 than @code{SET_RATIO}.
5919 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
5920 A C expression used to determine whether @code{store_by_pieces} will be
5921 used to set a chunk of memory to a constant string value, or whether some
5922 other mechanism will be used. Used by @code{__builtin_strcpy} when
5923 called with a constant source string.
5924 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5925 than @code{MOVE_RATIO}.
5928 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
5929 A C expression used to determine whether a load postincrement is a good
5930 thing to use for a given mode. Defaults to the value of
5931 @code{HAVE_POST_INCREMENT}.
5934 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
5935 A C expression used to determine whether a load postdecrement is a good
5936 thing to use for a given mode. Defaults to the value of
5937 @code{HAVE_POST_DECREMENT}.
5940 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
5941 A C expression used to determine whether a load preincrement is a good
5942 thing to use for a given mode. Defaults to the value of
5943 @code{HAVE_PRE_INCREMENT}.
5946 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
5947 A C expression used to determine whether a load predecrement is a good
5948 thing to use for a given mode. Defaults to the value of
5949 @code{HAVE_PRE_DECREMENT}.
5952 @defmac USE_STORE_POST_INCREMENT (@var{mode})
5953 A C expression used to determine whether a store postincrement is a good
5954 thing to use for a given mode. Defaults to the value of
5955 @code{HAVE_POST_INCREMENT}.
5958 @defmac USE_STORE_POST_DECREMENT (@var{mode})
5959 A C expression used to determine whether a store postdecrement is a good
5960 thing to use for a given mode. Defaults to the value of
5961 @code{HAVE_POST_DECREMENT}.
5964 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
5965 This macro is used to determine whether a store preincrement is a good
5966 thing to use for a given mode. Defaults to the value of
5967 @code{HAVE_PRE_INCREMENT}.
5970 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
5971 This macro is used to determine whether a store predecrement is a good
5972 thing to use for a given mode. Defaults to the value of
5973 @code{HAVE_PRE_DECREMENT}.
5976 @defmac NO_FUNCTION_CSE
5977 Define this macro if it is as good or better to call a constant
5978 function address than to call an address kept in a register.
5981 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
5982 Define this macro if a non-short-circuit operation produced by
5983 @samp{fold_range_test ()} is optimal. This macro defaults to true if
5984 @code{BRANCH_COST} is greater than or equal to the value 2.
5987 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total})
5988 This target hook describes the relative costs of RTL expressions.
5990 The cost may depend on the precise form of the expression, which is
5991 available for examination in @var{x}, and the rtx code of the expression
5992 in which it is contained, found in @var{outer_code}. @var{code} is the
5993 expression code---redundant, since it can be obtained with
5994 @code{GET_CODE (@var{x})}.
5996 In implementing this hook, you can use the construct
5997 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6000 On entry to the hook, @code{*@var{total}} contains a default estimate
6001 for the cost of the expression. The hook should modify this value as
6002 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
6003 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
6004 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
6006 When optimizing for code size, i.e.@: when @code{optimize_size} is
6007 nonzero, this target hook should be used to estimate the relative
6008 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6010 The hook returns true when all subexpressions of @var{x} have been
6011 processed, and false when @code{rtx_cost} should recurse.
6014 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address})
6015 This hook computes the cost of an addressing mode that contains
6016 @var{address}. If not defined, the cost is computed from
6017 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
6019 For most CISC machines, the default cost is a good approximation of the
6020 true cost of the addressing mode. However, on RISC machines, all
6021 instructions normally have the same length and execution time. Hence
6022 all addresses will have equal costs.
6024 In cases where more than one form of an address is known, the form with
6025 the lowest cost will be used. If multiple forms have the same, lowest,
6026 cost, the one that is the most complex will be used.
6028 For example, suppose an address that is equal to the sum of a register
6029 and a constant is used twice in the same basic block. When this macro
6030 is not defined, the address will be computed in a register and memory
6031 references will be indirect through that register. On machines where
6032 the cost of the addressing mode containing the sum is no higher than
6033 that of a simple indirect reference, this will produce an additional
6034 instruction and possibly require an additional register. Proper
6035 specification of this macro eliminates this overhead for such machines.
6037 This hook is never called with an invalid address.
6039 On machines where an address involving more than one register is as
6040 cheap as an address computation involving only one register, defining
6041 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
6042 be live over a region of code where only one would have been if
6043 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
6044 should be considered in the definition of this macro. Equivalent costs
6045 should probably only be given to addresses with different numbers of
6046 registers on machines with lots of registers.
6050 @section Adjusting the Instruction Scheduler
6052 The instruction scheduler may need a fair amount of machine-specific
6053 adjustment in order to produce good code. GCC provides several target
6054 hooks for this purpose. It is usually enough to define just a few of
6055 them: try the first ones in this list first.
6057 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
6058 This hook returns the maximum number of instructions that can ever
6059 issue at the same time on the target machine. The default is one.
6060 Although the insn scheduler can define itself the possibility of issue
6061 an insn on the same cycle, the value can serve as an additional
6062 constraint to issue insns on the same simulated processor cycle (see
6063 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
6064 This value must be constant over the entire compilation. If you need
6065 it to vary depending on what the instructions are, you must use
6066 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
6069 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
6070 This hook is executed by the scheduler after it has scheduled an insn
6071 from the ready list. It should return the number of insns which can
6072 still be issued in the current cycle. The default is
6073 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
6074 @code{USE}, which normally are not counted against the issue rate.
6075 You should define this hook if some insns take more machine resources
6076 than others, so that fewer insns can follow them in the same cycle.
6077 @var{file} is either a null pointer, or a stdio stream to write any
6078 debug output to. @var{verbose} is the verbose level provided by
6079 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
6083 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
6084 This function corrects the value of @var{cost} based on the
6085 relationship between @var{insn} and @var{dep_insn} through the
6086 dependence @var{link}. It should return the new value. The default
6087 is to make no adjustment to @var{cost}. This can be used for example
6088 to specify to the scheduler using the traditional pipeline description
6089 that an output- or anti-dependence does not incur the same cost as a
6090 data-dependence. If the scheduler using the automaton based pipeline
6091 description, the cost of anti-dependence is zero and the cost of
6092 output-dependence is maximum of one and the difference of latency
6093 times of the first and the second insns. If these values are not
6094 acceptable, you could use the hook to modify them too. See also
6095 @pxref{Processor pipeline description}.
6098 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
6099 This hook adjusts the integer scheduling priority @var{priority} of
6100 @var{insn}. It should return the new priority. Increase the priority to
6101 execute @var{insn} earlier, reduce the priority to execute @var{insn}
6102 later. Do not define this hook if you do not need to adjust the
6103 scheduling priorities of insns.
6106 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
6107 This hook is executed by the scheduler after it has scheduled the ready
6108 list, to allow the machine description to reorder it (for example to
6109 combine two small instructions together on @samp{VLIW} machines).
6110 @var{file} is either a null pointer, or a stdio stream to write any
6111 debug output to. @var{verbose} is the verbose level provided by
6112 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
6113 list of instructions that are ready to be scheduled. @var{n_readyp} is
6114 a pointer to the number of elements in the ready list. The scheduler
6115 reads the ready list in reverse order, starting with
6116 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
6117 is the timer tick of the scheduler. You may modify the ready list and
6118 the number of ready insns. The return value is the number of insns that
6119 can issue this cycle; normally this is just @code{issue_rate}. See also
6120 @samp{TARGET_SCHED_REORDER2}.
6123 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
6124 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
6125 function is called whenever the scheduler starts a new cycle. This one
6126 is called once per iteration over a cycle, immediately after
6127 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
6128 return the number of insns to be scheduled in the same cycle. Defining
6129 this hook can be useful if there are frequent situations where
6130 scheduling one insn causes other insns to become ready in the same
6131 cycle. These other insns can then be taken into account properly.
6134 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
6135 This hook is called after evaluation forward dependencies of insns in
6136 chain given by two parameter values (@var{head} and @var{tail}
6137 correspondingly) but before insns scheduling of the insn chain. For
6138 example, it can be used for better insn classification if it requires
6139 analysis of dependencies. This hook can use backward and forward
6140 dependencies of the insn scheduler because they are already
6144 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
6145 This hook is executed by the scheduler at the beginning of each block of
6146 instructions that are to be scheduled. @var{file} is either a null
6147 pointer, or a stdio stream to write any debug output to. @var{verbose}
6148 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6149 @var{max_ready} is the maximum number of insns in the current scheduling
6150 region that can be live at the same time. This can be used to allocate
6151 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
6154 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
6155 This hook is executed by the scheduler at the end of each block of
6156 instructions that are to be scheduled. It can be used to perform
6157 cleanup of any actions done by the other scheduling hooks. @var{file}
6158 is either a null pointer, or a stdio stream to write any debug output
6159 to. @var{verbose} is the verbose level provided by
6160 @option{-fsched-verbose-@var{n}}.
6163 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
6164 This hook is executed by the scheduler after function level initializations.
6165 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6166 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6167 @var{old_max_uid} is the maximum insn uid when scheduling begins.
6170 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
6171 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
6172 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6173 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6176 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
6177 The hook returns an RTL insn. The automaton state used in the
6178 pipeline hazard recognizer is changed as if the insn were scheduled
6179 when the new simulated processor cycle starts. Usage of the hook may
6180 simplify the automaton pipeline description for some @acronym{VLIW}
6181 processors. If the hook is defined, it is used only for the automaton
6182 based pipeline description. The default is not to change the state
6183 when the new simulated processor cycle starts.
6186 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
6187 The hook can be used to initialize data used by the previous hook.
6190 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
6191 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6192 to changed the state as if the insn were scheduled when the new
6193 simulated processor cycle finishes.
6196 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
6197 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
6198 used to initialize data used by the previous hook.
6201 @deftypefn {Target Hook} void TARGET_SCHED_DFA_PRE_CYCLE_ADVANCE (void)
6202 The hook to notify target that the current simulated cycle is about to finish.
6203 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6204 to change the state in more complicated situations - e.g. when advancing
6205 state on a single insn is not enough.
6208 @deftypefn {Target Hook} void TARGET_SCHED_DFA_POST_CYCLE_ADVANCE (void)
6209 The hook to notify target that new simulated cycle has just started.
6210 The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used
6211 to change the state in more complicated situations - e.g. when advancing
6212 state on a single insn is not enough.
6215 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
6216 This hook controls better choosing an insn from the ready insn queue
6217 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
6218 chooses the first insn from the queue. If the hook returns a positive
6219 value, an additional scheduler code tries all permutations of
6220 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
6221 subsequent ready insns to choose an insn whose issue will result in
6222 maximal number of issued insns on the same cycle. For the
6223 @acronym{VLIW} processor, the code could actually solve the problem of
6224 packing simple insns into the @acronym{VLIW} insn. Of course, if the
6225 rules of @acronym{VLIW} packing are described in the automaton.
6227 This code also could be used for superscalar @acronym{RISC}
6228 processors. Let us consider a superscalar @acronym{RISC} processor
6229 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
6230 @var{B}, some insns can be executed only in pipelines @var{B} or
6231 @var{C}, and one insn can be executed in pipeline @var{B}. The
6232 processor may issue the 1st insn into @var{A} and the 2nd one into
6233 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
6234 until the next cycle. If the scheduler issues the 3rd insn the first,
6235 the processor could issue all 3 insns per cycle.
6237 Actually this code demonstrates advantages of the automaton based
6238 pipeline hazard recognizer. We try quickly and easy many insn
6239 schedules to choose the best one.
6241 The default is no multipass scheduling.
6244 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
6246 This hook controls what insns from the ready insn queue will be
6247 considered for the multipass insn scheduling. If the hook returns
6248 zero for insn passed as the parameter, the insn will be not chosen to
6251 The default is that any ready insns can be chosen to be issued.
6254 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
6256 This hook is called by the insn scheduler before issuing insn passed
6257 as the third parameter on given cycle. If the hook returns nonzero,
6258 the insn is not issued on given processors cycle. Instead of that,
6259 the processor cycle is advanced. If the value passed through the last
6260 parameter is zero, the insn ready queue is not sorted on the new cycle
6261 start as usually. The first parameter passes file for debugging
6262 output. The second one passes the scheduler verbose level of the
6263 debugging output. The forth and the fifth parameter values are
6264 correspondingly processor cycle on which the previous insn has been
6265 issued and the current processor cycle.
6268 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct dep_def *@var{_dep}, int @var{cost}, int @var{distance})
6269 This hook is used to define which dependences are considered costly by
6270 the target, so costly that it is not advisable to schedule the insns that
6271 are involved in the dependence too close to one another. The parameters
6272 to this hook are as follows: The first parameter @var{_dep} is the dependence
6273 being evaluated. The second parameter @var{cost} is the cost of the
6274 dependence, and the third
6275 parameter @var{distance} is the distance in cycles between the two insns.
6276 The hook returns @code{true} if considering the distance between the two
6277 insns the dependence between them is considered costly by the target,
6278 and @code{false} otherwise.
6280 Defining this hook can be useful in multiple-issue out-of-order machines,
6281 where (a) it's practically hopeless to predict the actual data/resource
6282 delays, however: (b) there's a better chance to predict the actual grouping
6283 that will be formed, and (c) correctly emulating the grouping can be very
6284 important. In such targets one may want to allow issuing dependent insns
6285 closer to one another---i.e., closer than the dependence distance; however,
6286 not in cases of "costly dependences", which this hooks allows to define.
6289 @deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void)
6290 This hook is called by the insn scheduler after emitting a new instruction to
6291 the instruction stream. The hook notifies a target backend to extend its
6292 per instruction data structures.
6295 @deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx @var{insn}, int @var{request}, rtx *@var{new_pat})
6296 This hook is called by the insn scheduler when @var{insn} has only
6297 speculative dependencies and therefore can be scheduled speculatively.
6298 The hook is used to check if the pattern of @var{insn} has a speculative
6299 version and, in case of successful check, to generate that speculative
6300 pattern. The hook should return 1, if the instruction has a speculative form,
6301 or -1, if it doesn't. @var{request} describes the type of requested
6302 speculation. If the return value equals 1 then @var{new_pat} is assigned
6303 the generated speculative pattern.
6306 @deftypefn {Target Hook} int TARGET_SCHED_NEEDS_BLOCK_P (rtx @var{insn})
6307 This hook is called by the insn scheduler during generation of recovery code
6308 for @var{insn}. It should return nonzero, if the corresponding check
6309 instruction should branch to recovery code, or zero otherwise.
6312 @deftypefn {Target Hook} rtx TARGET_SCHED_GEN_CHECK (rtx @var{insn}, rtx @var{label}, int @var{mutate_p})
6313 This hook is called by the insn scheduler to generate a pattern for recovery
6314 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
6315 speculative instruction for which the check should be generated.
6316 @var{label} is either a label of a basic block, where recovery code should
6317 be emitted, or a null pointer, when requested check doesn't branch to
6318 recovery code (a simple check). If @var{mutate_p} is nonzero, then
6319 a pattern for a branchy check corresponding to a simple check denoted by
6320 @var{insn} should be generated. In this case @var{label} can't be null.
6323 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC (rtx @var{insn})
6324 This hook is used as a workaround for
6325 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD} not being
6326 called on the first instruction of the ready list. The hook is used to
6327 discard speculative instruction that stand first in the ready list from
6328 being scheduled on the current cycle. For non-speculative instructions,
6329 the hook should always return nonzero. For example, in the ia64 backend
6330 the hook is used to cancel data speculative insns when the ALAT table
6334 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (unsigned int *@var{flags}, spec_info_t @var{spec_info})
6335 This hook is used by the insn scheduler to find out what features should be
6336 enabled/used. @var{flags} initially may have either the SCHED_RGN or SCHED_EBB
6337 bit set. This denotes the scheduler pass for which the data should be
6338 provided. The target backend should modify @var{flags} by modifying
6339 the bits corresponding to the following features: USE_DEPS_LIST, USE_GLAT,
6340 DETACH_LIFE_INFO, and DO_SPECULATION. For the DO_SPECULATION feature
6341 an additional structure @var{spec_info} should be filled by the target.
6342 The structure describes speculation types that can be used in the scheduler.
6345 @deftypefn {Target Hook} int TARGET_SCHED_SMS_RES_MII (struct ddg *@var{g})
6346 This hook is called by the swing modulo scheduler to calculate a
6347 resource-based lower bound which is based on the resources available in
6348 the machine and the resources required by each instruction. The target
6349 backend can use @var{g} to calculate such bound. A very simple lower
6350 bound will be used in case this hook is not implemented: the total number
6351 of instructions divided by the issue rate.
6355 @section Dividing the Output into Sections (Texts, Data, @dots{})
6356 @c the above section title is WAY too long. maybe cut the part between
6357 @c the (...)? --mew 10feb93
6359 An object file is divided into sections containing different types of
6360 data. In the most common case, there are three sections: the @dfn{text
6361 section}, which holds instructions and read-only data; the @dfn{data
6362 section}, which holds initialized writable data; and the @dfn{bss
6363 section}, which holds uninitialized data. Some systems have other kinds
6366 @file{varasm.c} provides several well-known sections, such as
6367 @code{text_section}, @code{data_section} and @code{bss_section}.
6368 The normal way of controlling a @code{@var{foo}_section} variable
6369 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
6370 as described below. The macros are only read once, when @file{varasm.c}
6371 initializes itself, so their values must be run-time constants.
6372 They may however depend on command-line flags.
6374 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
6375 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
6376 to be string literals.
6378 Some assemblers require a different string to be written every time a
6379 section is selected. If your assembler falls into this category, you
6380 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
6381 @code{get_unnamed_section} to set up the sections.
6383 You must always create a @code{text_section}, either by defining
6384 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
6385 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
6386 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
6387 create a distinct @code{readonly_data_section}, the default is to
6388 reuse @code{text_section}.
6390 All the other @file{varasm.c} sections are optional, and are null
6391 if the target does not provide them.
6393 @defmac TEXT_SECTION_ASM_OP
6394 A C expression whose value is a string, including spacing, containing the
6395 assembler operation that should precede instructions and read-only data.
6396 Normally @code{"\t.text"} is right.
6399 @defmac HOT_TEXT_SECTION_NAME
6400 If defined, a C string constant for the name of the section containing most
6401 frequently executed functions of the program. If not defined, GCC will provide
6402 a default definition if the target supports named sections.
6405 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
6406 If defined, a C string constant for the name of the section containing unlikely
6407 executed functions in the program.
6410 @defmac DATA_SECTION_ASM_OP
6411 A C expression whose value is a string, including spacing, containing the
6412 assembler operation to identify the following data as writable initialized
6413 data. Normally @code{"\t.data"} is right.
6416 @defmac SDATA_SECTION_ASM_OP
6417 If defined, a C expression whose value is a string, including spacing,
6418 containing the assembler operation to identify the following data as
6419 initialized, writable small data.
6422 @defmac READONLY_DATA_SECTION_ASM_OP
6423 A C expression whose value is a string, including spacing, containing the
6424 assembler operation to identify the following data as read-only initialized
6428 @defmac BSS_SECTION_ASM_OP
6429 If defined, a C expression whose value is a string, including spacing,
6430 containing the assembler operation to identify the following data as
6431 uninitialized global data. If not defined, and neither
6432 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
6433 uninitialized global data will be output in the data section if
6434 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
6438 @defmac SBSS_SECTION_ASM_OP
6439 If defined, a C expression whose value is a string, including spacing,
6440 containing the assembler operation to identify the following data as
6441 uninitialized, writable small data.
6444 @defmac INIT_SECTION_ASM_OP
6445 If defined, a C expression whose value is a string, including spacing,
6446 containing the assembler operation to identify the following data as
6447 initialization code. If not defined, GCC will assume such a section does
6448 not exist. This section has no corresponding @code{init_section}
6449 variable; it is used entirely in runtime code.
6452 @defmac FINI_SECTION_ASM_OP
6453 If defined, a C expression whose value is a string, including spacing,
6454 containing the assembler operation to identify the following data as
6455 finalization code. If not defined, GCC will assume such a section does
6456 not exist. This section has no corresponding @code{fini_section}
6457 variable; it is used entirely in runtime code.
6460 @defmac INIT_ARRAY_SECTION_ASM_OP
6461 If defined, a C expression whose value is a string, including spacing,
6462 containing the assembler operation to identify the following data as
6463 part of the @code{.init_array} (or equivalent) section. If not
6464 defined, GCC will assume such a section does not exist. Do not define
6465 both this macro and @code{INIT_SECTION_ASM_OP}.
6468 @defmac FINI_ARRAY_SECTION_ASM_OP
6469 If defined, a C expression whose value is a string, including spacing,
6470 containing the assembler operation to identify the following data as
6471 part of the @code{.fini_array} (or equivalent) section. If not
6472 defined, GCC will assume such a section does not exist. Do not define
6473 both this macro and @code{FINI_SECTION_ASM_OP}.
6476 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
6477 If defined, an ASM statement that switches to a different section
6478 via @var{section_op}, calls @var{function}, and switches back to
6479 the text section. This is used in @file{crtstuff.c} if
6480 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
6481 to initialization and finalization functions from the init and fini
6482 sections. By default, this macro uses a simple function call. Some
6483 ports need hand-crafted assembly code to avoid dependencies on
6484 registers initialized in the function prologue or to ensure that
6485 constant pools don't end up too far way in the text section.
6488 @defmac TARGET_LIBGCC_SDATA_SECTION
6489 If defined, a string which names the section into which small
6490 variables defined in crtstuff and libgcc should go. This is useful
6491 when the target has options for optimizing access to small data, and
6492 you want the crtstuff and libgcc routines to be conservative in what
6493 they expect of your application yet liberal in what your application
6494 expects. For example, for targets with a @code{.sdata} section (like
6495 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
6496 require small data support from your application, but use this macro
6497 to put small data into @code{.sdata} so that your application can
6498 access these variables whether it uses small data or not.
6501 @defmac FORCE_CODE_SECTION_ALIGN
6502 If defined, an ASM statement that aligns a code section to some
6503 arbitrary boundary. This is used to force all fragments of the
6504 @code{.init} and @code{.fini} sections to have to same alignment
6505 and thus prevent the linker from having to add any padding.
6508 @defmac JUMP_TABLES_IN_TEXT_SECTION
6509 Define this macro to be an expression with a nonzero value if jump
6510 tables (for @code{tablejump} insns) should be output in the text
6511 section, along with the assembler instructions. Otherwise, the
6512 readonly data section is used.
6514 This macro is irrelevant if there is no separate readonly data section.
6517 @deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void)
6518 Define this hook if you need to do something special to set up the
6519 @file{varasm.c} sections, or if your target has some special sections
6520 of its own that you need to create.
6522 GCC calls this hook after processing the command line, but before writing
6523 any assembly code, and before calling any of the section-returning hooks
6527 @deftypefn {Target Hook} TARGET_ASM_RELOC_RW_MASK (void)
6528 Return a mask describing how relocations should be treated when
6529 selecting sections. Bit 1 should be set if global relocations
6530 should be placed in a read-write section; bit 0 should be set if
6531 local relocations should be placed in a read-write section.
6533 The default version of this function returns 3 when @option{-fpic}
6534 is in effect, and 0 otherwise. The hook is typically redefined
6535 when the target cannot support (some kinds of) dynamic relocations
6536 in read-only sections even in executables.
6539 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
6540 Return the section into which @var{exp} should be placed. You can
6541 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
6542 some sort. @var{reloc} indicates whether the initial value of @var{exp}
6543 requires link-time relocations. Bit 0 is set when variable contains
6544 local relocations only, while bit 1 is set for global relocations.
6545 @var{align} is the constant alignment in bits.
6547 The default version of this function takes care of putting read-only
6548 variables in @code{readonly_data_section}.
6550 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
6553 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
6554 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
6555 for @code{FUNCTION_DECL}s as well as for variables and constants.
6557 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
6558 function has been determined to be likely to be called, and nonzero if
6559 it is unlikely to be called.
6562 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
6563 Build up a unique section name, expressed as a @code{STRING_CST} node,
6564 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
6565 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
6566 the initial value of @var{exp} requires link-time relocations.
6568 The default version of this function appends the symbol name to the
6569 ELF section name that would normally be used for the symbol. For
6570 example, the function @code{foo} would be placed in @code{.text.foo}.
6571 Whatever the actual target object format, this is often good enough.
6574 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
6575 Return the readonly data section associated with
6576 @samp{DECL_SECTION_NAME (@var{decl})}.
6577 The default version of this function selects @code{.gnu.linkonce.r.name} if
6578 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
6579 if function is in @code{.text.name}, and the normal readonly-data section
6583 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
6584 Return the section into which a constant @var{x}, of mode @var{mode},
6585 should be placed. You can assume that @var{x} is some kind of
6586 constant in RTL@. The argument @var{mode} is redundant except in the
6587 case of a @code{const_int} rtx. @var{align} is the constant alignment
6590 The default version of this function takes care of putting symbolic
6591 constants in @code{flag_pic} mode in @code{data_section} and everything
6592 else in @code{readonly_data_section}.
6595 @deftypefn {Target Hook} void TARGET_MANGLE_DECL_ASSEMBLER_NAME (tree @var{decl}, tree @var{id})
6596 Define this hook if you need to postprocess the assembler name generated
6597 by target-independent code. The @var{id} provided to this hook will be
6598 the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C,
6599 or the mangled name of the @var{decl} in C++). The return value of the
6600 hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on
6601 your target system. The default implementation of this hook just
6602 returns the @var{id} provided.
6605 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
6606 Define this hook if references to a symbol or a constant must be
6607 treated differently depending on something about the variable or
6608 function named by the symbol (such as what section it is in).
6610 The hook is executed immediately after rtl has been created for
6611 @var{decl}, which may be a variable or function declaration or
6612 an entry in the constant pool. In either case, @var{rtl} is the
6613 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
6614 in this hook; that field may not have been initialized yet.
6616 In the case of a constant, it is safe to assume that the rtl is
6617 a @code{mem} whose address is a @code{symbol_ref}. Most decls
6618 will also have this form, but that is not guaranteed. Global
6619 register variables, for instance, will have a @code{reg} for their
6620 rtl. (Normally the right thing to do with such unusual rtl is
6623 The @var{new_decl_p} argument will be true if this is the first time
6624 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
6625 be false for subsequent invocations, which will happen for duplicate
6626 declarations. Whether or not anything must be done for the duplicate
6627 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
6628 @var{new_decl_p} is always true when the hook is called for a constant.
6630 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
6631 The usual thing for this hook to do is to record flags in the
6632 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
6633 Historically, the name string was modified if it was necessary to
6634 encode more than one bit of information, but this practice is now
6635 discouraged; use @code{SYMBOL_REF_FLAGS}.
6637 The default definition of this hook, @code{default_encode_section_info}
6638 in @file{varasm.c}, sets a number of commonly-useful bits in
6639 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
6640 before overriding it.
6643 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
6644 Decode @var{name} and return the real name part, sans
6645 the characters that @code{TARGET_ENCODE_SECTION_INFO}
6649 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
6650 Returns true if @var{exp} should be placed into a ``small data'' section.
6651 The default version of this hook always returns false.
6654 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
6655 Contains the value true if the target places read-only
6656 ``small data'' into a separate section. The default value is false.
6659 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
6660 Returns true if @var{exp} names an object for which name resolution
6661 rules must resolve to the current ``module'' (dynamic shared library
6662 or executable image).
6664 The default version of this hook implements the name resolution rules
6665 for ELF, which has a looser model of global name binding than other
6666 currently supported object file formats.
6669 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
6670 Contains the value true if the target supports thread-local storage.
6671 The default value is false.
6676 @section Position Independent Code
6677 @cindex position independent code
6680 This section describes macros that help implement generation of position
6681 independent code. Simply defining these macros is not enough to
6682 generate valid PIC; you must also add support to the macros
6683 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
6684 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
6685 @samp{movsi} to do something appropriate when the source operand
6686 contains a symbolic address. You may also need to alter the handling of
6687 switch statements so that they use relative addresses.
6688 @c i rearranged the order of the macros above to try to force one of
6689 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
6691 @defmac PIC_OFFSET_TABLE_REGNUM
6692 The register number of the register used to address a table of static
6693 data addresses in memory. In some cases this register is defined by a
6694 processor's ``application binary interface'' (ABI)@. When this macro
6695 is defined, RTL is generated for this register once, as with the stack
6696 pointer and frame pointer registers. If this macro is not defined, it
6697 is up to the machine-dependent files to allocate such a register (if
6698 necessary). Note that this register must be fixed when in use (e.g.@:
6699 when @code{flag_pic} is true).
6702 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6703 Define this macro if the register defined by
6704 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
6705 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
6708 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
6709 A C expression that is nonzero if @var{x} is a legitimate immediate
6710 operand on the target machine when generating position independent code.
6711 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6712 check this. You can also assume @var{flag_pic} is true, so you need not
6713 check it either. You need not define this macro if all constants
6714 (including @code{SYMBOL_REF}) can be immediate operands when generating
6715 position independent code.
6718 @node Assembler Format
6719 @section Defining the Output Assembler Language
6721 This section describes macros whose principal purpose is to describe how
6722 to write instructions in assembler language---rather than what the
6726 * File Framework:: Structural information for the assembler file.
6727 * Data Output:: Output of constants (numbers, strings, addresses).
6728 * Uninitialized Data:: Output of uninitialized variables.
6729 * Label Output:: Output and generation of labels.
6730 * Initialization:: General principles of initialization
6731 and termination routines.
6732 * Macros for Initialization::
6733 Specific macros that control the handling of
6734 initialization and termination routines.
6735 * Instruction Output:: Output of actual instructions.
6736 * Dispatch Tables:: Output of jump tables.
6737 * Exception Region Output:: Output of exception region code.
6738 * Alignment Output:: Pseudo ops for alignment and skipping data.
6741 @node File Framework
6742 @subsection The Overall Framework of an Assembler File
6743 @cindex assembler format
6744 @cindex output of assembler code
6746 @c prevent bad page break with this line
6747 This describes the overall framework of an assembly file.
6749 @deftypefn {Target Hook} void TARGET_ASM_FILE_START ()
6750 @findex default_file_start
6751 Output to @code{asm_out_file} any text which the assembler expects to
6752 find at the beginning of a file. The default behavior is controlled
6753 by two flags, documented below. Unless your target's assembler is
6754 quite unusual, if you override the default, you should call
6755 @code{default_file_start} at some point in your target hook. This
6756 lets other target files rely on these variables.
6759 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
6760 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
6761 printed as the very first line in the assembly file, unless
6762 @option{-fverbose-asm} is in effect. (If that macro has been defined
6763 to the empty string, this variable has no effect.) With the normal
6764 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
6765 assembler that it need not bother stripping comments or extra
6766 whitespace from its input. This allows it to work a bit faster.
6768 The default is false. You should not set it to true unless you have
6769 verified that your port does not generate any extra whitespace or
6770 comments that will cause GAS to issue errors in NO_APP mode.
6773 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
6774 If this flag is true, @code{output_file_directive} will be called
6775 for the primary source file, immediately after printing
6776 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
6777 this to be done. The default is false.
6780 @deftypefn {Target Hook} void TARGET_ASM_FILE_END ()
6781 Output to @code{asm_out_file} any text which the assembler expects
6782 to find at the end of a file. The default is to output nothing.
6785 @deftypefun void file_end_indicate_exec_stack ()
6786 Some systems use a common convention, the @samp{.note.GNU-stack}
6787 special section, to indicate whether or not an object file relies on
6788 the stack being executable. If your system uses this convention, you
6789 should define @code{TARGET_ASM_FILE_END} to this function. If you
6790 need to do other things in that hook, have your hook function call
6794 @defmac ASM_COMMENT_START
6795 A C string constant describing how to begin a comment in the target
6796 assembler language. The compiler assumes that the comment will end at
6797 the end of the line.
6801 A C string constant for text to be output before each @code{asm}
6802 statement or group of consecutive ones. Normally this is
6803 @code{"#APP"}, which is a comment that has no effect on most
6804 assemblers but tells the GNU assembler that it must check the lines
6805 that follow for all valid assembler constructs.
6809 A C string constant for text to be output after each @code{asm}
6810 statement or group of consecutive ones. Normally this is
6811 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6812 time-saving assumptions that are valid for ordinary compiler output.
6815 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6816 A C statement to output COFF information or DWARF debugging information
6817 which indicates that filename @var{name} is the current source file to
6818 the stdio stream @var{stream}.
6820 This macro need not be defined if the standard form of output
6821 for the file format in use is appropriate.
6824 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6825 A C statement to output the string @var{string} to the stdio stream
6826 @var{stream}. If you do not call the function @code{output_quoted_string}
6827 in your config files, GCC will only call it to output filenames to
6828 the assembler source. So you can use it to canonicalize the format
6829 of the filename using this macro.
6832 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6833 A C statement to output something to the assembler file to handle a
6834 @samp{#ident} directive containing the text @var{string}. If this
6835 macro is not defined, nothing is output for a @samp{#ident} directive.
6838 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6839 Output assembly directives to switch to section @var{name}. The section
6840 should have attributes as specified by @var{flags}, which is a bit mask
6841 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6842 is nonzero, it contains an alignment in bytes to be used for the section,
6843 otherwise some target default should be used. Only targets that must
6844 specify an alignment within the section directive need pay attention to
6845 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6848 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6849 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6852 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
6853 @deftypefn {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
6854 This flag is true if we can create zeroed data by switching to a BSS
6855 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
6856 This is true on most ELF targets.
6859 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6860 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6861 based on a variable or function decl, a section name, and whether or not the
6862 declaration's initializer may contain runtime relocations. @var{decl} may be
6863 null, in which case read-write data should be assumed.
6865 The default version of this function handles choosing code vs data,
6866 read-only vs read-write data, and @code{flag_pic}. You should only
6867 need to override this if your target has special flags that might be
6868 set via @code{__attribute__}.
6871 @deftypefn {Target Hook} {int} TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type @var{type}, const char * @var{text})
6872 Provides the target with the ability to record the gcc command line
6873 switches that have been passed to the compiler, and options that are
6874 enabled. The @var{type} argument specifies what is being recorded.
6875 It can take the following values:
6878 @item SWITCH_TYPE_PASSED
6879 @var{text} is a command line switch that has been set by the user.
6881 @item SWITCH_TYPE_ENABLED
6882 @var{text} is an option which has been enabled. This might be as a
6883 direct result of a command line switch, or because it is enabled by
6884 default or because it has been enabled as a side effect of a different
6885 command line switch. For example, the @option{-O2} switch enables
6886 various different individual optimization passes.
6888 @item SWITCH_TYPE_DESCRIPTIVE
6889 @var{text} is either NULL or some descriptive text which should be
6890 ignored. If @var{text} is NULL then it is being used to warn the
6891 target hook that either recording is starting or ending. The first
6892 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
6893 warning is for start up and the second time the warning is for
6894 wind down. This feature is to allow the target hook to make any
6895 necessary preparations before it starts to record switches and to
6896 perform any necessary tidying up after it has finished recording
6899 @item SWITCH_TYPE_LINE_START
6900 This option can be ignored by this target hook.
6902 @item SWITCH_TYPE_LINE_END
6903 This option can be ignored by this target hook.
6906 The hook's return value must be zero. Other return values may be
6907 supported in the future.
6909 By default this hook is set to NULL, but an example implementation is
6910 provided for ELF based targets. Called @var{elf_record_gcc_switches},
6911 it records the switches as ASCII text inside a new, string mergeable
6912 section in the assembler output file. The name of the new section is
6913 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
6917 @deftypefn {Target Hook} {const char *} TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
6918 This is the name of the section that will be created by the example
6919 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
6925 @subsection Output of Data
6928 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6929 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6930 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6931 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6932 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6933 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6934 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6935 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6936 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6937 These hooks specify assembly directives for creating certain kinds
6938 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6939 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6940 aligned two-byte object, and so on. Any of the hooks may be
6941 @code{NULL}, indicating that no suitable directive is available.
6943 The compiler will print these strings at the start of a new line,
6944 followed immediately by the object's initial value. In most cases,
6945 the string should contain a tab, a pseudo-op, and then another tab.
6948 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6949 The @code{assemble_integer} function uses this hook to output an
6950 integer object. @var{x} is the object's value, @var{size} is its size
6951 in bytes and @var{aligned_p} indicates whether it is aligned. The
6952 function should return @code{true} if it was able to output the
6953 object. If it returns false, @code{assemble_integer} will try to
6954 split the object into smaller parts.
6956 The default implementation of this hook will use the
6957 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6958 when the relevant string is @code{NULL}.
6961 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6962 A C statement to recognize @var{rtx} patterns that
6963 @code{output_addr_const} can't deal with, and output assembly code to
6964 @var{stream} corresponding to the pattern @var{x}. This may be used to
6965 allow machine-dependent @code{UNSPEC}s to appear within constants.
6967 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6968 @code{goto fail}, so that a standard error message is printed. If it
6969 prints an error message itself, by calling, for example,
6970 @code{output_operand_lossage}, it may just complete normally.
6973 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6974 A C statement to output to the stdio stream @var{stream} an assembler
6975 instruction to assemble a string constant containing the @var{len}
6976 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6977 @code{char *} and @var{len} a C expression of type @code{int}.
6979 If the assembler has a @code{.ascii} pseudo-op as found in the
6980 Berkeley Unix assembler, do not define the macro
6981 @code{ASM_OUTPUT_ASCII}.
6984 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6985 A C statement to output word @var{n} of a function descriptor for
6986 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6987 is defined, and is otherwise unused.
6990 @defmac CONSTANT_POOL_BEFORE_FUNCTION
6991 You may define this macro as a C expression. You should define the
6992 expression to have a nonzero value if GCC should output the constant
6993 pool for a function before the code for the function, or a zero value if
6994 GCC should output the constant pool after the function. If you do
6995 not define this macro, the usual case, GCC will output the constant
6996 pool before the function.
6999 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
7000 A C statement to output assembler commands to define the start of the
7001 constant pool for a function. @var{funname} is a string giving
7002 the name of the function. Should the return type of the function
7003 be required, it can be obtained via @var{fundecl}. @var{size}
7004 is the size, in bytes, of the constant pool that will be written
7005 immediately after this call.
7007 If no constant-pool prefix is required, the usual case, this macro need
7011 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
7012 A C statement (with or without semicolon) to output a constant in the
7013 constant pool, if it needs special treatment. (This macro need not do
7014 anything for RTL expressions that can be output normally.)
7016 The argument @var{file} is the standard I/O stream to output the
7017 assembler code on. @var{x} is the RTL expression for the constant to
7018 output, and @var{mode} is the machine mode (in case @var{x} is a
7019 @samp{const_int}). @var{align} is the required alignment for the value
7020 @var{x}; you should output an assembler directive to force this much
7023 The argument @var{labelno} is a number to use in an internal label for
7024 the address of this pool entry. The definition of this macro is
7025 responsible for outputting the label definition at the proper place.
7026 Here is how to do this:
7029 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
7032 When you output a pool entry specially, you should end with a
7033 @code{goto} to the label @var{jumpto}. This will prevent the same pool
7034 entry from being output a second time in the usual manner.
7036 You need not define this macro if it would do nothing.
7039 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
7040 A C statement to output assembler commands to at the end of the constant
7041 pool for a function. @var{funname} is a string giving the name of the
7042 function. Should the return type of the function be required, you can
7043 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
7044 constant pool that GCC wrote immediately before this call.
7046 If no constant-pool epilogue is required, the usual case, you need not
7050 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
7051 Define this macro as a C expression which is nonzero if @var{C} is
7052 used as a logical line separator by the assembler. @var{STR} points
7053 to the position in the string where @var{C} was found; this can be used if
7054 a line separator uses multiple characters.
7056 If you do not define this macro, the default is that only
7057 the character @samp{;} is treated as a logical line separator.
7060 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
7061 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
7062 These target hooks are C string constants, describing the syntax in the
7063 assembler for grouping arithmetic expressions. If not overridden, they
7064 default to normal parentheses, which is correct for most assemblers.
7067 These macros are provided by @file{real.h} for writing the definitions
7068 of @code{ASM_OUTPUT_DOUBLE} and the like:
7070 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
7071 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
7072 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
7073 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
7074 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
7075 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
7076 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
7077 target's floating point representation, and store its bit pattern in
7078 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
7079 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
7080 simple @code{long int}. For the others, it should be an array of
7081 @code{long int}. The number of elements in this array is determined
7082 by the size of the desired target floating point data type: 32 bits of
7083 it go in each @code{long int} array element. Each array element holds
7084 32 bits of the result, even if @code{long int} is wider than 32 bits
7085 on the host machine.
7087 The array element values are designed so that you can print them out
7088 using @code{fprintf} in the order they should appear in the target
7092 @node Uninitialized Data
7093 @subsection Output of Uninitialized Variables
7095 Each of the macros in this section is used to do the whole job of
7096 outputting a single uninitialized variable.
7098 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
7099 A C statement (sans semicolon) to output to the stdio stream
7100 @var{stream} the assembler definition of a common-label named
7101 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7102 is the size rounded up to whatever alignment the caller wants.
7104 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7105 output the name itself; before and after that, output the additional
7106 assembler syntax for defining the name, and a newline.
7108 This macro controls how the assembler definitions of uninitialized
7109 common global variables are output.
7112 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
7113 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
7114 separate, explicit argument. If you define this macro, it is used in
7115 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
7116 handling the required alignment of the variable. The alignment is specified
7117 as the number of bits.
7120 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7121 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
7122 variable to be output, if there is one, or @code{NULL_TREE} if there
7123 is no corresponding variable. If you define this macro, GCC will use it
7124 in place of both @code{ASM_OUTPUT_COMMON} and
7125 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
7126 the variable's decl in order to chose what to output.
7129 @defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
7130 A C statement (sans semicolon) to output to the stdio stream
7131 @var{stream} the assembler definition of uninitialized global @var{decl} named
7132 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7133 is the size rounded up to whatever alignment the caller wants.
7135 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
7136 defining this macro. If unable, use the expression
7137 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
7138 before and after that, output the additional assembler syntax for defining
7139 the name, and a newline.
7141 There are two ways of handling global BSS. One is to define either
7142 this macro or its aligned counterpart, @code{ASM_OUTPUT_ALIGNED_BSS}.
7143 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
7144 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
7145 You do not need to do both.
7147 Some languages do not have @code{common} data, and require a
7148 non-common form of global BSS in order to handle uninitialized globals
7149 efficiently. C++ is one example of this. However, if the target does
7150 not support global BSS, the front end may choose to make globals
7151 common in order to save space in the object file.
7154 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7155 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
7156 separate, explicit argument. If you define this macro, it is used in
7157 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
7158 handling the required alignment of the variable. The alignment is specified
7159 as the number of bits.
7161 Try to use function @code{asm_output_aligned_bss} defined in file
7162 @file{varasm.c} when defining this macro.
7165 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
7166 A C statement (sans semicolon) to output to the stdio stream
7167 @var{stream} the assembler definition of a local-common-label named
7168 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7169 is the size rounded up to whatever alignment the caller wants.
7171 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7172 output the name itself; before and after that, output the additional
7173 assembler syntax for defining the name, and a newline.
7175 This macro controls how the assembler definitions of uninitialized
7176 static variables are output.
7179 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
7180 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
7181 separate, explicit argument. If you define this macro, it is used in
7182 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
7183 handling the required alignment of the variable. The alignment is specified
7184 as the number of bits.
7187 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7188 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
7189 variable to be output, if there is one, or @code{NULL_TREE} if there
7190 is no corresponding variable. If you define this macro, GCC will use it
7191 in place of both @code{ASM_OUTPUT_DECL} and
7192 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
7193 the variable's decl in order to chose what to output.
7197 @subsection Output and Generation of Labels
7199 @c prevent bad page break with this line
7200 This is about outputting labels.
7202 @findex assemble_name
7203 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
7204 A C statement (sans semicolon) to output to the stdio stream
7205 @var{stream} the assembler definition of a label named @var{name}.
7206 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7207 output the name itself; before and after that, output the additional
7208 assembler syntax for defining the name, and a newline. A default
7209 definition of this macro is provided which is correct for most systems.
7212 @findex assemble_name_raw
7213 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
7214 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
7215 to refer to a compiler-generated label. The default definition uses
7216 @code{assemble_name_raw}, which is like @code{assemble_name} except
7217 that it is more efficient.
7221 A C string containing the appropriate assembler directive to specify the
7222 size of a symbol, without any arguments. On systems that use ELF, the
7223 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
7224 systems, the default is not to define this macro.
7226 Define this macro only if it is correct to use the default definitions
7227 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
7228 for your system. If you need your own custom definitions of those
7229 macros, or if you do not need explicit symbol sizes at all, do not
7233 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
7234 A C statement (sans semicolon) to output to the stdio stream
7235 @var{stream} a directive telling the assembler that the size of the
7236 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
7237 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7241 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
7242 A C statement (sans semicolon) to output to the stdio stream
7243 @var{stream} a directive telling the assembler to calculate the size of
7244 the symbol @var{name} by subtracting its address from the current
7247 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7248 provided. The default assumes that the assembler recognizes a special
7249 @samp{.} symbol as referring to the current address, and can calculate
7250 the difference between this and another symbol. If your assembler does
7251 not recognize @samp{.} or cannot do calculations with it, you will need
7252 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
7256 A C string containing the appropriate assembler directive to specify the
7257 type of a symbol, without any arguments. On systems that use ELF, the
7258 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
7259 systems, the default is not to define this macro.
7261 Define this macro only if it is correct to use the default definition of
7262 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7263 custom definition of this macro, or if you do not need explicit symbol
7264 types at all, do not define this macro.
7267 @defmac TYPE_OPERAND_FMT
7268 A C string which specifies (using @code{printf} syntax) the format of
7269 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
7270 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
7271 the default is not to define this macro.
7273 Define this macro only if it is correct to use the default definition of
7274 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7275 custom definition of this macro, or if you do not need explicit symbol
7276 types at all, do not define this macro.
7279 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
7280 A C statement (sans semicolon) to output to the stdio stream
7281 @var{stream} a directive telling the assembler that the type of the
7282 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
7283 that string is always either @samp{"function"} or @samp{"object"}, but
7284 you should not count on this.
7286 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
7287 definition of this macro is provided.
7290 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
7291 A C statement (sans semicolon) to output to the stdio stream
7292 @var{stream} any text necessary for declaring the name @var{name} of a
7293 function which is being defined. This macro is responsible for
7294 outputting the label definition (perhaps using
7295 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
7296 @code{FUNCTION_DECL} tree node representing the function.
7298 If this macro is not defined, then the function name is defined in the
7299 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7301 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7305 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
7306 A C statement (sans semicolon) to output to the stdio stream
7307 @var{stream} any text necessary for declaring the size of a function
7308 which is being defined. The argument @var{name} is the name of the
7309 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
7310 representing the function.
7312 If this macro is not defined, then the function size is not defined.
7314 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
7318 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
7319 A C statement (sans semicolon) to output to the stdio stream
7320 @var{stream} any text necessary for declaring the name @var{name} of an
7321 initialized variable which is being defined. This macro must output the
7322 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
7323 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
7325 If this macro is not defined, then the variable name is defined in the
7326 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7328 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
7329 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
7332 @defmac ASM_DECLARE_CONSTANT_NAME (@var{stream}, @var{name}, @var{exp}, @var{size})
7333 A C statement (sans semicolon) to output to the stdio stream
7334 @var{stream} any text necessary for declaring the name @var{name} of a
7335 constant which is being defined. This macro is responsible for
7336 outputting the label definition (perhaps using
7337 @code{ASM_OUTPUT_LABEL}). The argument @var{exp} is the
7338 value of the constant, and @var{size} is the size of the constant
7339 in bytes. @var{name} will be an internal label.
7341 If this macro is not defined, then the @var{name} is defined in the
7342 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7344 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7348 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
7349 A C statement (sans semicolon) to output to the stdio stream
7350 @var{stream} any text necessary for claiming a register @var{regno}
7351 for a global variable @var{decl} with name @var{name}.
7353 If you don't define this macro, that is equivalent to defining it to do
7357 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
7358 A C statement (sans semicolon) to finish up declaring a variable name
7359 once the compiler has processed its initializer fully and thus has had a
7360 chance to determine the size of an array when controlled by an
7361 initializer. This is used on systems where it's necessary to declare
7362 something about the size of the object.
7364 If you don't define this macro, that is equivalent to defining it to do
7367 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
7368 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
7371 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
7372 This target hook is a function to output to the stdio stream
7373 @var{stream} some commands that will make the label @var{name} global;
7374 that is, available for reference from other files.
7376 The default implementation relies on a proper definition of
7377 @code{GLOBAL_ASM_OP}.
7380 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *@var{stream}, tree @var{decl})
7381 This target hook is a function to output to the stdio stream
7382 @var{stream} some commands that will make the name associated with @var{decl}
7383 global; that is, available for reference from other files.
7385 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
7388 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
7389 A C statement (sans semicolon) to output to the stdio stream
7390 @var{stream} some commands that will make the label @var{name} weak;
7391 that is, available for reference from other files but only used if
7392 no other definition is available. Use the expression
7393 @code{assemble_name (@var{stream}, @var{name})} to output the name
7394 itself; before and after that, output the additional assembler syntax
7395 for making that name weak, and a newline.
7397 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
7398 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
7402 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
7403 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
7404 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
7405 or variable decl. If @var{value} is not @code{NULL}, this C statement
7406 should output to the stdio stream @var{stream} assembler code which
7407 defines (equates) the weak symbol @var{name} to have the value
7408 @var{value}. If @var{value} is @code{NULL}, it should output commands
7409 to make @var{name} weak.
7412 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
7413 Outputs a directive that enables @var{name} to be used to refer to
7414 symbol @var{value} with weak-symbol semantics. @code{decl} is the
7415 declaration of @code{name}.
7418 @defmac SUPPORTS_WEAK
7419 A C expression which evaluates to true if the target supports weak symbols.
7421 If you don't define this macro, @file{defaults.h} provides a default
7422 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
7423 is defined, the default definition is @samp{1}; otherwise, it is
7424 @samp{0}. Define this macro if you want to control weak symbol support
7425 with a compiler flag such as @option{-melf}.
7428 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
7429 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
7430 public symbol such that extra copies in multiple translation units will
7431 be discarded by the linker. Define this macro if your object file
7432 format provides support for this concept, such as the @samp{COMDAT}
7433 section flags in the Microsoft Windows PE/COFF format, and this support
7434 requires changes to @var{decl}, such as putting it in a separate section.
7437 @defmac SUPPORTS_ONE_ONLY
7438 A C expression which evaluates to true if the target supports one-only
7441 If you don't define this macro, @file{varasm.c} provides a default
7442 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
7443 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
7444 you want to control one-only symbol support with a compiler flag, or if
7445 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
7446 be emitted as one-only.
7449 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
7450 This target hook is a function to output to @var{asm_out_file} some
7451 commands that will make the symbol(s) associated with @var{decl} have
7452 hidden, protected or internal visibility as specified by @var{visibility}.
7455 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
7456 A C expression that evaluates to true if the target's linker expects
7457 that weak symbols do not appear in a static archive's table of contents.
7458 The default is @code{0}.
7460 Leaving weak symbols out of an archive's table of contents means that,
7461 if a symbol will only have a definition in one translation unit and
7462 will have undefined references from other translation units, that
7463 symbol should not be weak. Defining this macro to be nonzero will
7464 thus have the effect that certain symbols that would normally be weak
7465 (explicit template instantiations, and vtables for polymorphic classes
7466 with noninline key methods) will instead be nonweak.
7468 The C++ ABI requires this macro to be zero. Define this macro for
7469 targets where full C++ ABI compliance is impossible and where linker
7470 restrictions require weak symbols to be left out of a static archive's
7474 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
7475 A C statement (sans semicolon) to output to the stdio stream
7476 @var{stream} any text necessary for declaring the name of an external
7477 symbol named @var{name} which is referenced in this compilation but
7478 not defined. The value of @var{decl} is the tree node for the
7481 This macro need not be defined if it does not need to output anything.
7482 The GNU assembler and most Unix assemblers don't require anything.
7485 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
7486 This target hook is a function to output to @var{asm_out_file} an assembler
7487 pseudo-op to declare a library function name external. The name of the
7488 library function is given by @var{symref}, which is a @code{symbol_ref}.
7491 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (tree @var{decl})
7492 This target hook is a function to output to @var{asm_out_file} an assembler
7493 directive to annotate used symbol. Darwin target use .no_dead_code_strip
7497 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
7498 A C statement (sans semicolon) to output to the stdio stream
7499 @var{stream} a reference in assembler syntax to a label named
7500 @var{name}. This should add @samp{_} to the front of the name, if that
7501 is customary on your operating system, as it is in most Berkeley Unix
7502 systems. This macro is used in @code{assemble_name}.
7505 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
7506 A C statement (sans semicolon) to output a reference to
7507 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
7508 will be used to output the name of the symbol. This macro may be used
7509 to modify the way a symbol is referenced depending on information
7510 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
7513 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
7514 A C statement (sans semicolon) to output a reference to @var{buf}, the
7515 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
7516 @code{assemble_name} will be used to output the name of the symbol.
7517 This macro is not used by @code{output_asm_label}, or the @code{%l}
7518 specifier that calls it; the intention is that this macro should be set
7519 when it is necessary to output a label differently when its address is
7523 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
7524 A function to output to the stdio stream @var{stream} a label whose
7525 name is made from the string @var{prefix} and the number @var{labelno}.
7527 It is absolutely essential that these labels be distinct from the labels
7528 used for user-level functions and variables. Otherwise, certain programs
7529 will have name conflicts with internal labels.
7531 It is desirable to exclude internal labels from the symbol table of the
7532 object file. Most assemblers have a naming convention for labels that
7533 should be excluded; on many systems, the letter @samp{L} at the
7534 beginning of a label has this effect. You should find out what
7535 convention your system uses, and follow it.
7537 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
7540 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
7541 A C statement to output to the stdio stream @var{stream} a debug info
7542 label whose name is made from the string @var{prefix} and the number
7543 @var{num}. This is useful for VLIW targets, where debug info labels
7544 may need to be treated differently than branch target labels. On some
7545 systems, branch target labels must be at the beginning of instruction
7546 bundles, but debug info labels can occur in the middle of instruction
7549 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
7553 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
7554 A C statement to store into the string @var{string} a label whose name
7555 is made from the string @var{prefix} and the number @var{num}.
7557 This string, when output subsequently by @code{assemble_name}, should
7558 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
7559 with the same @var{prefix} and @var{num}.
7561 If the string begins with @samp{*}, then @code{assemble_name} will
7562 output the rest of the string unchanged. It is often convenient for
7563 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
7564 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
7565 to output the string, and may change it. (Of course,
7566 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
7567 you should know what it does on your machine.)
7570 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
7571 A C expression to assign to @var{outvar} (which is a variable of type
7572 @code{char *}) a newly allocated string made from the string
7573 @var{name} and the number @var{number}, with some suitable punctuation
7574 added. Use @code{alloca} to get space for the string.
7576 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
7577 produce an assembler label for an internal static variable whose name is
7578 @var{name}. Therefore, the string must be such as to result in valid
7579 assembler code. The argument @var{number} is different each time this
7580 macro is executed; it prevents conflicts between similarly-named
7581 internal static variables in different scopes.
7583 Ideally this string should not be a valid C identifier, to prevent any
7584 conflict with the user's own symbols. Most assemblers allow periods
7585 or percent signs in assembler symbols; putting at least one of these
7586 between the name and the number will suffice.
7588 If this macro is not defined, a default definition will be provided
7589 which is correct for most systems.
7592 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
7593 A C statement to output to the stdio stream @var{stream} assembler code
7594 which defines (equates) the symbol @var{name} to have the value @var{value}.
7597 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7598 correct for most systems.
7601 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
7602 A C statement to output to the stdio stream @var{stream} assembler code
7603 which defines (equates) the symbol whose tree node is @var{decl_of_name}
7604 to have the value of the tree node @var{decl_of_value}. This macro will
7605 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
7606 the tree nodes are available.
7609 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7610 correct for most systems.
7613 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
7614 A C statement that evaluates to true if the assembler code which defines
7615 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
7616 of the tree node @var{decl_of_value} should be emitted near the end of the
7617 current compilation unit. The default is to not defer output of defines.
7618 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
7619 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
7622 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
7623 A C statement to output to the stdio stream @var{stream} assembler code
7624 which defines (equates) the weak symbol @var{name} to have the value
7625 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
7626 an undefined weak symbol.
7628 Define this macro if the target only supports weak aliases; define
7629 @code{ASM_OUTPUT_DEF} instead if possible.
7632 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
7633 Define this macro to override the default assembler names used for
7634 Objective-C methods.
7636 The default name is a unique method number followed by the name of the
7637 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
7638 the category is also included in the assembler name (e.g.@:
7641 These names are safe on most systems, but make debugging difficult since
7642 the method's selector is not present in the name. Therefore, particular
7643 systems define other ways of computing names.
7645 @var{buf} is an expression of type @code{char *} which gives you a
7646 buffer in which to store the name; its length is as long as
7647 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
7648 50 characters extra.
7650 The argument @var{is_inst} specifies whether the method is an instance
7651 method or a class method; @var{class_name} is the name of the class;
7652 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
7653 in a category); and @var{sel_name} is the name of the selector.
7655 On systems where the assembler can handle quoted names, you can use this
7656 macro to provide more human-readable names.
7659 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
7660 A C statement (sans semicolon) to output to the stdio stream
7661 @var{stream} commands to declare that the label @var{name} is an
7662 Objective-C class reference. This is only needed for targets whose
7663 linkers have special support for NeXT-style runtimes.
7666 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
7667 A C statement (sans semicolon) to output to the stdio stream
7668 @var{stream} commands to declare that the label @var{name} is an
7669 unresolved Objective-C class reference. This is only needed for targets
7670 whose linkers have special support for NeXT-style runtimes.
7673 @node Initialization
7674 @subsection How Initialization Functions Are Handled
7675 @cindex initialization routines
7676 @cindex termination routines
7677 @cindex constructors, output of
7678 @cindex destructors, output of
7680 The compiled code for certain languages includes @dfn{constructors}
7681 (also called @dfn{initialization routines})---functions to initialize
7682 data in the program when the program is started. These functions need
7683 to be called before the program is ``started''---that is to say, before
7684 @code{main} is called.
7686 Compiling some languages generates @dfn{destructors} (also called
7687 @dfn{termination routines}) that should be called when the program
7690 To make the initialization and termination functions work, the compiler
7691 must output something in the assembler code to cause those functions to
7692 be called at the appropriate time. When you port the compiler to a new
7693 system, you need to specify how to do this.
7695 There are two major ways that GCC currently supports the execution of
7696 initialization and termination functions. Each way has two variants.
7697 Much of the structure is common to all four variations.
7699 @findex __CTOR_LIST__
7700 @findex __DTOR_LIST__
7701 The linker must build two lists of these functions---a list of
7702 initialization functions, called @code{__CTOR_LIST__}, and a list of
7703 termination functions, called @code{__DTOR_LIST__}.
7705 Each list always begins with an ignored function pointer (which may hold
7706 0, @minus{}1, or a count of the function pointers after it, depending on
7707 the environment). This is followed by a series of zero or more function
7708 pointers to constructors (or destructors), followed by a function
7709 pointer containing zero.
7711 Depending on the operating system and its executable file format, either
7712 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
7713 time and exit time. Constructors are called in reverse order of the
7714 list; destructors in forward order.
7716 The best way to handle static constructors works only for object file
7717 formats which provide arbitrarily-named sections. A section is set
7718 aside for a list of constructors, and another for a list of destructors.
7719 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
7720 object file that defines an initialization function also puts a word in
7721 the constructor section to point to that function. The linker
7722 accumulates all these words into one contiguous @samp{.ctors} section.
7723 Termination functions are handled similarly.
7725 This method will be chosen as the default by @file{target-def.h} if
7726 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
7727 support arbitrary sections, but does support special designated
7728 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
7729 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
7731 When arbitrary sections are available, there are two variants, depending
7732 upon how the code in @file{crtstuff.c} is called. On systems that
7733 support a @dfn{.init} section which is executed at program startup,
7734 parts of @file{crtstuff.c} are compiled into that section. The
7735 program is linked by the @command{gcc} driver like this:
7738 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
7741 The prologue of a function (@code{__init}) appears in the @code{.init}
7742 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
7743 for the function @code{__fini} in the @dfn{.fini} section. Normally these
7744 files are provided by the operating system or by the GNU C library, but
7745 are provided by GCC for a few targets.
7747 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
7748 compiled from @file{crtstuff.c}. They contain, among other things, code
7749 fragments within the @code{.init} and @code{.fini} sections that branch
7750 to routines in the @code{.text} section. The linker will pull all parts
7751 of a section together, which results in a complete @code{__init} function
7752 that invokes the routines we need at startup.
7754 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
7757 If no init section is available, when GCC compiles any function called
7758 @code{main} (or more accurately, any function designated as a program
7759 entry point by the language front end calling @code{expand_main_function}),
7760 it inserts a procedure call to @code{__main} as the first executable code
7761 after the function prologue. The @code{__main} function is defined
7762 in @file{libgcc2.c} and runs the global constructors.
7764 In file formats that don't support arbitrary sections, there are again
7765 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
7766 and an `a.out' format must be used. In this case,
7767 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
7768 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
7769 and with the address of the void function containing the initialization
7770 code as its value. The GNU linker recognizes this as a request to add
7771 the value to a @dfn{set}; the values are accumulated, and are eventually
7772 placed in the executable as a vector in the format described above, with
7773 a leading (ignored) count and a trailing zero element.
7774 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
7775 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
7776 the compilation of @code{main} to call @code{__main} as above, starting
7777 the initialization process.
7779 The last variant uses neither arbitrary sections nor the GNU linker.
7780 This is preferable when you want to do dynamic linking and when using
7781 file formats which the GNU linker does not support, such as `ECOFF'@. In
7782 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
7783 termination functions are recognized simply by their names. This requires
7784 an extra program in the linkage step, called @command{collect2}. This program
7785 pretends to be the linker, for use with GCC; it does its job by running
7786 the ordinary linker, but also arranges to include the vectors of
7787 initialization and termination functions. These functions are called
7788 via @code{__main} as described above. In order to use this method,
7789 @code{use_collect2} must be defined in the target in @file{config.gcc}.
7792 The following section describes the specific macros that control and
7793 customize the handling of initialization and termination functions.
7796 @node Macros for Initialization
7797 @subsection Macros Controlling Initialization Routines
7799 Here are the macros that control how the compiler handles initialization
7800 and termination functions:
7802 @defmac INIT_SECTION_ASM_OP
7803 If defined, a C string constant, including spacing, for the assembler
7804 operation to identify the following data as initialization code. If not
7805 defined, GCC will assume such a section does not exist. When you are
7806 using special sections for initialization and termination functions, this
7807 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7808 run the initialization functions.
7811 @defmac HAS_INIT_SECTION
7812 If defined, @code{main} will not call @code{__main} as described above.
7813 This macro should be defined for systems that control start-up code
7814 on a symbol-by-symbol basis, such as OSF/1, and should not
7815 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7818 @defmac LD_INIT_SWITCH
7819 If defined, a C string constant for a switch that tells the linker that
7820 the following symbol is an initialization routine.
7823 @defmac LD_FINI_SWITCH
7824 If defined, a C string constant for a switch that tells the linker that
7825 the following symbol is a finalization routine.
7828 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7829 If defined, a C statement that will write a function that can be
7830 automatically called when a shared library is loaded. The function
7831 should call @var{func}, which takes no arguments. If not defined, and
7832 the object format requires an explicit initialization function, then a
7833 function called @code{_GLOBAL__DI} will be generated.
7835 This function and the following one are used by collect2 when linking a
7836 shared library that needs constructors or destructors, or has DWARF2
7837 exception tables embedded in the code.
7840 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7841 If defined, a C statement that will write a function that can be
7842 automatically called when a shared library is unloaded. The function
7843 should call @var{func}, which takes no arguments. If not defined, and
7844 the object format requires an explicit finalization function, then a
7845 function called @code{_GLOBAL__DD} will be generated.
7848 @defmac INVOKE__main
7849 If defined, @code{main} will call @code{__main} despite the presence of
7850 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7851 where the init section is not actually run automatically, but is still
7852 useful for collecting the lists of constructors and destructors.
7855 @defmac SUPPORTS_INIT_PRIORITY
7856 If nonzero, the C++ @code{init_priority} attribute is supported and the
7857 compiler should emit instructions to control the order of initialization
7858 of objects. If zero, the compiler will issue an error message upon
7859 encountering an @code{init_priority} attribute.
7862 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7863 This value is true if the target supports some ``native'' method of
7864 collecting constructors and destructors to be run at startup and exit.
7865 It is false if we must use @command{collect2}.
7868 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7869 If defined, a function that outputs assembler code to arrange to call
7870 the function referenced by @var{symbol} at initialization time.
7872 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7873 no arguments and with no return value. If the target supports initialization
7874 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7875 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7877 If this macro is not defined by the target, a suitable default will
7878 be chosen if (1) the target supports arbitrary section names, (2) the
7879 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7883 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7884 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7885 functions rather than initialization functions.
7888 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7889 generated for the generated object file will have static linkage.
7891 If your system uses @command{collect2} as the means of processing
7892 constructors, then that program normally uses @command{nm} to scan
7893 an object file for constructor functions to be called.
7895 On certain kinds of systems, you can define this macro to make
7896 @command{collect2} work faster (and, in some cases, make it work at all):
7898 @defmac OBJECT_FORMAT_COFF
7899 Define this macro if the system uses COFF (Common Object File Format)
7900 object files, so that @command{collect2} can assume this format and scan
7901 object files directly for dynamic constructor/destructor functions.
7903 This macro is effective only in a native compiler; @command{collect2} as
7904 part of a cross compiler always uses @command{nm} for the target machine.
7907 @defmac REAL_NM_FILE_NAME
7908 Define this macro as a C string constant containing the file name to use
7909 to execute @command{nm}. The default is to search the path normally for
7912 If your system supports shared libraries and has a program to list the
7913 dynamic dependencies of a given library or executable, you can define
7914 these macros to enable support for running initialization and
7915 termination functions in shared libraries:
7919 Define this macro to a C string constant containing the name of the program
7920 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7923 @defmac PARSE_LDD_OUTPUT (@var{ptr})
7924 Define this macro to be C code that extracts filenames from the output
7925 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7926 of type @code{char *} that points to the beginning of a line of output
7927 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7928 code must advance @var{ptr} to the beginning of the filename on that
7929 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7932 @node Instruction Output
7933 @subsection Output of Assembler Instructions
7935 @c prevent bad page break with this line
7936 This describes assembler instruction output.
7938 @defmac REGISTER_NAMES
7939 A C initializer containing the assembler's names for the machine
7940 registers, each one as a C string constant. This is what translates
7941 register numbers in the compiler into assembler language.
7944 @defmac ADDITIONAL_REGISTER_NAMES
7945 If defined, a C initializer for an array of structures containing a name
7946 and a register number. This macro defines additional names for hard
7947 registers, thus allowing the @code{asm} option in declarations to refer
7948 to registers using alternate names.
7951 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7952 Define this macro if you are using an unusual assembler that
7953 requires different names for the machine instructions.
7955 The definition is a C statement or statements which output an
7956 assembler instruction opcode to the stdio stream @var{stream}. The
7957 macro-operand @var{ptr} is a variable of type @code{char *} which
7958 points to the opcode name in its ``internal'' form---the form that is
7959 written in the machine description. The definition should output the
7960 opcode name to @var{stream}, performing any translation you desire, and
7961 increment the variable @var{ptr} to point at the end of the opcode
7962 so that it will not be output twice.
7964 In fact, your macro definition may process less than the entire opcode
7965 name, or more than the opcode name; but if you want to process text
7966 that includes @samp{%}-sequences to substitute operands, you must take
7967 care of the substitution yourself. Just be sure to increment
7968 @var{ptr} over whatever text should not be output normally.
7970 @findex recog_data.operand
7971 If you need to look at the operand values, they can be found as the
7972 elements of @code{recog_data.operand}.
7974 If the macro definition does nothing, the instruction is output
7978 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7979 If defined, a C statement to be executed just prior to the output of
7980 assembler code for @var{insn}, to modify the extracted operands so
7981 they will be output differently.
7983 Here the argument @var{opvec} is the vector containing the operands
7984 extracted from @var{insn}, and @var{noperands} is the number of
7985 elements of the vector which contain meaningful data for this insn.
7986 The contents of this vector are what will be used to convert the insn
7987 template into assembler code, so you can change the assembler output
7988 by changing the contents of the vector.
7990 This macro is useful when various assembler syntaxes share a single
7991 file of instruction patterns; by defining this macro differently, you
7992 can cause a large class of instructions to be output differently (such
7993 as with rearranged operands). Naturally, variations in assembler
7994 syntax affecting individual insn patterns ought to be handled by
7995 writing conditional output routines in those patterns.
7997 If this macro is not defined, it is equivalent to a null statement.
8000 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
8001 A C compound statement to output to stdio stream @var{stream} the
8002 assembler syntax for an instruction operand @var{x}. @var{x} is an
8005 @var{code} is a value that can be used to specify one of several ways
8006 of printing the operand. It is used when identical operands must be
8007 printed differently depending on the context. @var{code} comes from
8008 the @samp{%} specification that was used to request printing of the
8009 operand. If the specification was just @samp{%@var{digit}} then
8010 @var{code} is 0; if the specification was @samp{%@var{ltr}
8011 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
8014 If @var{x} is a register, this macro should print the register's name.
8015 The names can be found in an array @code{reg_names} whose type is
8016 @code{char *[]}. @code{reg_names} is initialized from
8017 @code{REGISTER_NAMES}.
8019 When the machine description has a specification @samp{%@var{punct}}
8020 (a @samp{%} followed by a punctuation character), this macro is called
8021 with a null pointer for @var{x} and the punctuation character for
8025 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
8026 A C expression which evaluates to true if @var{code} is a valid
8027 punctuation character for use in the @code{PRINT_OPERAND} macro. If
8028 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
8029 punctuation characters (except for the standard one, @samp{%}) are used
8033 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
8034 A C compound statement to output to stdio stream @var{stream} the
8035 assembler syntax for an instruction operand that is a memory reference
8036 whose address is @var{x}. @var{x} is an RTL expression.
8038 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
8039 On some machines, the syntax for a symbolic address depends on the
8040 section that the address refers to. On these machines, define the hook
8041 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
8042 @code{symbol_ref}, and then check for it here. @xref{Assembler
8046 @findex dbr_sequence_length
8047 @defmac DBR_OUTPUT_SEQEND (@var{file})
8048 A C statement, to be executed after all slot-filler instructions have
8049 been output. If necessary, call @code{dbr_sequence_length} to
8050 determine the number of slots filled in a sequence (zero if not
8051 currently outputting a sequence), to decide how many no-ops to output,
8054 Don't define this macro if it has nothing to do, but it is helpful in
8055 reading assembly output if the extent of the delay sequence is made
8056 explicit (e.g.@: with white space).
8059 @findex final_sequence
8060 Note that output routines for instructions with delay slots must be
8061 prepared to deal with not being output as part of a sequence
8062 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
8063 found.) The variable @code{final_sequence} is null when not
8064 processing a sequence, otherwise it contains the @code{sequence} rtx
8068 @defmac REGISTER_PREFIX
8069 @defmacx LOCAL_LABEL_PREFIX
8070 @defmacx USER_LABEL_PREFIX
8071 @defmacx IMMEDIATE_PREFIX
8072 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
8073 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
8074 @file{final.c}). These are useful when a single @file{md} file must
8075 support multiple assembler formats. In that case, the various @file{tm.h}
8076 files can define these macros differently.
8079 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
8080 If defined this macro should expand to a series of @code{case}
8081 statements which will be parsed inside the @code{switch} statement of
8082 the @code{asm_fprintf} function. This allows targets to define extra
8083 printf formats which may useful when generating their assembler
8084 statements. Note that uppercase letters are reserved for future
8085 generic extensions to asm_fprintf, and so are not available to target
8086 specific code. The output file is given by the parameter @var{file}.
8087 The varargs input pointer is @var{argptr} and the rest of the format
8088 string, starting the character after the one that is being switched
8089 upon, is pointed to by @var{format}.
8092 @defmac ASSEMBLER_DIALECT
8093 If your target supports multiple dialects of assembler language (such as
8094 different opcodes), define this macro as a C expression that gives the
8095 numeric index of the assembler language dialect to use, with zero as the
8098 If this macro is defined, you may use constructs of the form
8100 @samp{@{option0|option1|option2@dots{}@}}
8103 in the output templates of patterns (@pxref{Output Template}) or in the
8104 first argument of @code{asm_fprintf}. This construct outputs
8105 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
8106 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
8107 within these strings retain their usual meaning. If there are fewer
8108 alternatives within the braces than the value of
8109 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
8111 If you do not define this macro, the characters @samp{@{}, @samp{|} and
8112 @samp{@}} do not have any special meaning when used in templates or
8113 operands to @code{asm_fprintf}.
8115 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
8116 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
8117 the variations in assembler language syntax with that mechanism. Define
8118 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
8119 if the syntax variant are larger and involve such things as different
8120 opcodes or operand order.
8123 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
8124 A C expression to output to @var{stream} some assembler code
8125 which will push hard register number @var{regno} onto the stack.
8126 The code need not be optimal, since this macro is used only when
8130 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
8131 A C expression to output to @var{stream} some assembler code
8132 which will pop hard register number @var{regno} off of the stack.
8133 The code need not be optimal, since this macro is used only when
8137 @node Dispatch Tables
8138 @subsection Output of Dispatch Tables
8140 @c prevent bad page break with this line
8141 This concerns dispatch tables.
8143 @cindex dispatch table
8144 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
8145 A C statement to output to the stdio stream @var{stream} an assembler
8146 pseudo-instruction to generate a difference between two labels.
8147 @var{value} and @var{rel} are the numbers of two internal labels. The
8148 definitions of these labels are output using
8149 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
8150 way here. For example,
8153 fprintf (@var{stream}, "\t.word L%d-L%d\n",
8154 @var{value}, @var{rel})
8157 You must provide this macro on machines where the addresses in a
8158 dispatch table are relative to the table's own address. If defined, GCC
8159 will also use this macro on all machines when producing PIC@.
8160 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
8161 mode and flags can be read.
8164 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
8165 This macro should be provided on machines where the addresses
8166 in a dispatch table are absolute.
8168 The definition should be a C statement to output to the stdio stream
8169 @var{stream} an assembler pseudo-instruction to generate a reference to
8170 a label. @var{value} is the number of an internal label whose
8171 definition is output using @code{(*targetm.asm_out.internal_label)}.
8175 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
8179 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
8180 Define this if the label before a jump-table needs to be output
8181 specially. The first three arguments are the same as for
8182 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
8183 jump-table which follows (a @code{jump_insn} containing an
8184 @code{addr_vec} or @code{addr_diff_vec}).
8186 This feature is used on system V to output a @code{swbeg} statement
8189 If this macro is not defined, these labels are output with
8190 @code{(*targetm.asm_out.internal_label)}.
8193 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
8194 Define this if something special must be output at the end of a
8195 jump-table. The definition should be a C statement to be executed
8196 after the assembler code for the table is written. It should write
8197 the appropriate code to stdio stream @var{stream}. The argument
8198 @var{table} is the jump-table insn, and @var{num} is the label-number
8199 of the preceding label.
8201 If this macro is not defined, nothing special is output at the end of
8205 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (@var{stream}, @var{decl}, @var{for_eh}, @var{empty})
8206 This target hook emits a label at the beginning of each FDE@. It
8207 should be defined on targets where FDEs need special labels, and it
8208 should write the appropriate label, for the FDE associated with the
8209 function declaration @var{decl}, to the stdio stream @var{stream}.
8210 The third argument, @var{for_eh}, is a boolean: true if this is for an
8211 exception table. The fourth argument, @var{empty}, is a boolean:
8212 true if this is a placeholder label for an omitted FDE@.
8214 The default is that FDEs are not given nonlocal labels.
8217 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (@var{stream})
8218 This target hook emits a label at the beginning of the exception table.
8219 It should be defined on targets where it is desirable for the table
8220 to be broken up according to function.
8222 The default is that no label is emitted.
8225 @deftypefn {Target Hook} void TARGET_UNWIND_EMIT (FILE * @var{stream}, rtx @var{insn})
8226 This target hook emits and assembly directives required to unwind the
8227 given instruction. This is only used when TARGET_UNWIND_INFO is set.
8230 @node Exception Region Output
8231 @subsection Assembler Commands for Exception Regions
8233 @c prevent bad page break with this line
8235 This describes commands marking the start and the end of an exception
8238 @defmac EH_FRAME_SECTION_NAME
8239 If defined, a C string constant for the name of the section containing
8240 exception handling frame unwind information. If not defined, GCC will
8241 provide a default definition if the target supports named sections.
8242 @file{crtstuff.c} uses this macro to switch to the appropriate section.
8244 You should define this symbol if your target supports DWARF 2 frame
8245 unwind information and the default definition does not work.
8248 @defmac EH_FRAME_IN_DATA_SECTION
8249 If defined, DWARF 2 frame unwind information will be placed in the
8250 data section even though the target supports named sections. This
8251 might be necessary, for instance, if the system linker does garbage
8252 collection and sections cannot be marked as not to be collected.
8254 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
8258 @defmac EH_TABLES_CAN_BE_READ_ONLY
8259 Define this macro to 1 if your target is such that no frame unwind
8260 information encoding used with non-PIC code will ever require a
8261 runtime relocation, but the linker may not support merging read-only
8262 and read-write sections into a single read-write section.
8265 @defmac MASK_RETURN_ADDR
8266 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
8267 that it does not contain any extraneous set bits in it.
8270 @defmac DWARF2_UNWIND_INFO
8271 Define this macro to 0 if your target supports DWARF 2 frame unwind
8272 information, but it does not yet work with exception handling.
8273 Otherwise, if your target supports this information (if it defines
8274 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
8275 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of 1.
8277 If @code{TARGET_UNWIND_INFO} is defined, the target specific unwinder
8278 will be used in all cases. Defining this macro will enable the generation
8279 of DWARF 2 frame debugging information.
8281 If @code{TARGET_UNWIND_INFO} is not defined, and this macro is defined to 1,
8282 the DWARF 2 unwinder will be the default exception handling mechanism;
8283 otherwise, the @code{setjmp}/@code{longjmp}-based scheme will be used by
8287 @defmac TARGET_UNWIND_INFO
8288 Define this macro if your target has ABI specified unwind tables. Usually
8289 these will be output by @code{TARGET_UNWIND_EMIT}.
8292 @deftypevar {Target Hook} bool TARGET_UNWIND_TABLES_DEFAULT
8293 This variable should be set to @code{true} if the target ABI requires unwinding
8294 tables even when exceptions are not used.
8297 @defmac MUST_USE_SJLJ_EXCEPTIONS
8298 This macro need only be defined if @code{DWARF2_UNWIND_INFO} is
8299 runtime-variable. In that case, @file{except.h} cannot correctly
8300 determine the corresponding definition of @code{MUST_USE_SJLJ_EXCEPTIONS},
8301 so the target must provide it directly.
8304 @defmac DONT_USE_BUILTIN_SETJMP
8305 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
8306 should use the @code{setjmp}/@code{longjmp} functions from the C library
8307 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
8310 @defmac DWARF_CIE_DATA_ALIGNMENT
8311 This macro need only be defined if the target might save registers in the
8312 function prologue at an offset to the stack pointer that is not aligned to
8313 @code{UNITS_PER_WORD}. The definition should be the negative minimum
8314 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
8315 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
8316 the target supports DWARF 2 frame unwind information.
8319 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
8320 Contains the value true if the target should add a zero word onto the
8321 end of a Dwarf-2 frame info section when used for exception handling.
8322 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
8326 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
8327 Given a register, this hook should return a parallel of registers to
8328 represent where to find the register pieces. Define this hook if the
8329 register and its mode are represented in Dwarf in non-contiguous
8330 locations, or if the register should be represented in more than one
8331 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
8332 If not defined, the default is to return @code{NULL_RTX}.
8335 @deftypefn {Target Hook} void TARGET_INIT_DWARF_REG_SIZES_EXTRA (tree @var{address})
8336 If some registers are represented in Dwarf-2 unwind information in
8337 multiple pieces, define this hook to fill in information about the
8338 sizes of those pieces in the table used by the unwinder at runtime.
8339 It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after
8340 filling in a single size corresponding to each hard register;
8341 @var{address} is the address of the table.
8344 @deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym})
8345 This hook is used to output a reference from a frame unwinding table to
8346 the type_info object identified by @var{sym}. It should return @code{true}
8347 if the reference was output. Returning @code{false} will cause the
8348 reference to be output using the normal Dwarf2 routines.
8351 @deftypefn {Target Hook} bool TARGET_ARM_EABI_UNWINDER
8352 This hook should be set to @code{true} on targets that use an ARM EABI
8353 based unwinding library, and @code{false} on other targets. This effects
8354 the format of unwinding tables, and how the unwinder in entered after
8355 running a cleanup. The default is @code{false}.
8358 @node Alignment Output
8359 @subsection Assembler Commands for Alignment
8361 @c prevent bad page break with this line
8362 This describes commands for alignment.
8364 @defmac JUMP_ALIGN (@var{label})
8365 The alignment (log base 2) to put in front of @var{label}, which is
8366 a common destination of jumps and has no fallthru incoming edge.
8368 This macro need not be defined if you don't want any special alignment
8369 to be done at such a time. Most machine descriptions do not currently
8372 Unless it's necessary to inspect the @var{label} parameter, it is better
8373 to set the variable @var{align_jumps} in the target's
8374 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8375 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
8378 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
8379 The alignment (log base 2) to put in front of @var{label}, which follows
8382 This macro need not be defined if you don't want any special alignment
8383 to be done at such a time. Most machine descriptions do not currently
8387 @defmac LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
8388 The maximum number of bytes to skip when applying
8389 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
8390 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8393 @defmac LOOP_ALIGN (@var{label})
8394 The alignment (log base 2) to put in front of @var{label}, which follows
8395 a @code{NOTE_INSN_LOOP_BEG} note.
8397 This macro need not be defined if you don't want any special alignment
8398 to be done at such a time. Most machine descriptions do not currently
8401 Unless it's necessary to inspect the @var{label} parameter, it is better
8402 to set the variable @code{align_loops} in the target's
8403 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8404 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
8407 @defmac LOOP_ALIGN_MAX_SKIP
8408 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
8409 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8412 @defmac LABEL_ALIGN (@var{label})
8413 The alignment (log base 2) to put in front of @var{label}.
8414 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
8415 the maximum of the specified values is used.
8417 Unless it's necessary to inspect the @var{label} parameter, it is better
8418 to set the variable @code{align_labels} in the target's
8419 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8420 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
8423 @defmac LABEL_ALIGN_MAX_SKIP
8424 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
8425 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8428 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
8429 A C statement to output to the stdio stream @var{stream} an assembler
8430 instruction to advance the location counter by @var{nbytes} bytes.
8431 Those bytes should be zero when loaded. @var{nbytes} will be a C
8432 expression of type @code{unsigned HOST_WIDE_INT}.
8435 @defmac ASM_NO_SKIP_IN_TEXT
8436 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
8437 text section because it fails to put zeros in the bytes that are skipped.
8438 This is true on many Unix systems, where the pseudo--op to skip bytes
8439 produces no-op instructions rather than zeros when used in the text
8443 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
8444 A C statement to output to the stdio stream @var{stream} an assembler
8445 command to advance the location counter to a multiple of 2 to the
8446 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
8449 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
8450 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
8451 for padding, if necessary.
8454 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
8455 A C statement to output to the stdio stream @var{stream} an assembler
8456 command to advance the location counter to a multiple of 2 to the
8457 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
8458 satisfy the alignment request. @var{power} and @var{max_skip} will be
8459 a C expression of type @code{int}.
8463 @node Debugging Info
8464 @section Controlling Debugging Information Format
8466 @c prevent bad page break with this line
8467 This describes how to specify debugging information.
8470 * All Debuggers:: Macros that affect all debugging formats uniformly.
8471 * DBX Options:: Macros enabling specific options in DBX format.
8472 * DBX Hooks:: Hook macros for varying DBX format.
8473 * File Names and DBX:: Macros controlling output of file names in DBX format.
8474 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
8475 * VMS Debug:: Macros for VMS debug format.
8479 @subsection Macros Affecting All Debugging Formats
8481 @c prevent bad page break with this line
8482 These macros affect all debugging formats.
8484 @defmac DBX_REGISTER_NUMBER (@var{regno})
8485 A C expression that returns the DBX register number for the compiler
8486 register number @var{regno}. In the default macro provided, the value
8487 of this expression will be @var{regno} itself. But sometimes there are
8488 some registers that the compiler knows about and DBX does not, or vice
8489 versa. In such cases, some register may need to have one number in the
8490 compiler and another for DBX@.
8492 If two registers have consecutive numbers inside GCC, and they can be
8493 used as a pair to hold a multiword value, then they @emph{must} have
8494 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
8495 Otherwise, debuggers will be unable to access such a pair, because they
8496 expect register pairs to be consecutive in their own numbering scheme.
8498 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
8499 does not preserve register pairs, then what you must do instead is
8500 redefine the actual register numbering scheme.
8503 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
8504 A C expression that returns the integer offset value for an automatic
8505 variable having address @var{x} (an RTL expression). The default
8506 computation assumes that @var{x} is based on the frame-pointer and
8507 gives the offset from the frame-pointer. This is required for targets
8508 that produce debugging output for DBX or COFF-style debugging output
8509 for SDB and allow the frame-pointer to be eliminated when the
8510 @option{-g} options is used.
8513 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
8514 A C expression that returns the integer offset value for an argument
8515 having address @var{x} (an RTL expression). The nominal offset is
8519 @defmac PREFERRED_DEBUGGING_TYPE
8520 A C expression that returns the type of debugging output GCC should
8521 produce when the user specifies just @option{-g}. Define
8522 this if you have arranged for GCC to support more than one format of
8523 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
8524 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
8525 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
8527 When the user specifies @option{-ggdb}, GCC normally also uses the
8528 value of this macro to select the debugging output format, but with two
8529 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
8530 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
8531 defined, GCC uses @code{DBX_DEBUG}.
8533 The value of this macro only affects the default debugging output; the
8534 user can always get a specific type of output by using @option{-gstabs},
8535 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
8539 @subsection Specific Options for DBX Output
8541 @c prevent bad page break with this line
8542 These are specific options for DBX output.
8544 @defmac DBX_DEBUGGING_INFO
8545 Define this macro if GCC should produce debugging output for DBX
8546 in response to the @option{-g} option.
8549 @defmac XCOFF_DEBUGGING_INFO
8550 Define this macro if GCC should produce XCOFF format debugging output
8551 in response to the @option{-g} option. This is a variant of DBX format.
8554 @defmac DEFAULT_GDB_EXTENSIONS
8555 Define this macro to control whether GCC should by default generate
8556 GDB's extended version of DBX debugging information (assuming DBX-format
8557 debugging information is enabled at all). If you don't define the
8558 macro, the default is 1: always generate the extended information
8559 if there is any occasion to.
8562 @defmac DEBUG_SYMS_TEXT
8563 Define this macro if all @code{.stabs} commands should be output while
8564 in the text section.
8567 @defmac ASM_STABS_OP
8568 A C string constant, including spacing, naming the assembler pseudo op to
8569 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
8570 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
8571 applies only to DBX debugging information format.
8574 @defmac ASM_STABD_OP
8575 A C string constant, including spacing, naming the assembler pseudo op to
8576 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
8577 value is the current location. If you don't define this macro,
8578 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
8582 @defmac ASM_STABN_OP
8583 A C string constant, including spacing, naming the assembler pseudo op to
8584 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
8585 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
8586 macro applies only to DBX debugging information format.
8589 @defmac DBX_NO_XREFS
8590 Define this macro if DBX on your system does not support the construct
8591 @samp{xs@var{tagname}}. On some systems, this construct is used to
8592 describe a forward reference to a structure named @var{tagname}.
8593 On other systems, this construct is not supported at all.
8596 @defmac DBX_CONTIN_LENGTH
8597 A symbol name in DBX-format debugging information is normally
8598 continued (split into two separate @code{.stabs} directives) when it
8599 exceeds a certain length (by default, 80 characters). On some
8600 operating systems, DBX requires this splitting; on others, splitting
8601 must not be done. You can inhibit splitting by defining this macro
8602 with the value zero. You can override the default splitting-length by
8603 defining this macro as an expression for the length you desire.
8606 @defmac DBX_CONTIN_CHAR
8607 Normally continuation is indicated by adding a @samp{\} character to
8608 the end of a @code{.stabs} string when a continuation follows. To use
8609 a different character instead, define this macro as a character
8610 constant for the character you want to use. Do not define this macro
8611 if backslash is correct for your system.
8614 @defmac DBX_STATIC_STAB_DATA_SECTION
8615 Define this macro if it is necessary to go to the data section before
8616 outputting the @samp{.stabs} pseudo-op for a non-global static
8620 @defmac DBX_TYPE_DECL_STABS_CODE
8621 The value to use in the ``code'' field of the @code{.stabs} directive
8622 for a typedef. The default is @code{N_LSYM}.
8625 @defmac DBX_STATIC_CONST_VAR_CODE
8626 The value to use in the ``code'' field of the @code{.stabs} directive
8627 for a static variable located in the text section. DBX format does not
8628 provide any ``right'' way to do this. The default is @code{N_FUN}.
8631 @defmac DBX_REGPARM_STABS_CODE
8632 The value to use in the ``code'' field of the @code{.stabs} directive
8633 for a parameter passed in registers. DBX format does not provide any
8634 ``right'' way to do this. The default is @code{N_RSYM}.
8637 @defmac DBX_REGPARM_STABS_LETTER
8638 The letter to use in DBX symbol data to identify a symbol as a parameter
8639 passed in registers. DBX format does not customarily provide any way to
8640 do this. The default is @code{'P'}.
8643 @defmac DBX_FUNCTION_FIRST
8644 Define this macro if the DBX information for a function and its
8645 arguments should precede the assembler code for the function. Normally,
8646 in DBX format, the debugging information entirely follows the assembler
8650 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
8651 Define this macro, with value 1, if the value of a symbol describing
8652 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
8653 relative to the start of the enclosing function. Normally, GCC uses
8654 an absolute address.
8657 @defmac DBX_LINES_FUNCTION_RELATIVE
8658 Define this macro, with value 1, if the value of a symbol indicating
8659 the current line number (@code{N_SLINE}) should be relative to the
8660 start of the enclosing function. Normally, GCC uses an absolute address.
8663 @defmac DBX_USE_BINCL
8664 Define this macro if GCC should generate @code{N_BINCL} and
8665 @code{N_EINCL} stabs for included header files, as on Sun systems. This
8666 macro also directs GCC to output a type number as a pair of a file
8667 number and a type number within the file. Normally, GCC does not
8668 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
8669 number for a type number.
8673 @subsection Open-Ended Hooks for DBX Format
8675 @c prevent bad page break with this line
8676 These are hooks for DBX format.
8678 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
8679 Define this macro to say how to output to @var{stream} the debugging
8680 information for the start of a scope level for variable names. The
8681 argument @var{name} is the name of an assembler symbol (for use with
8682 @code{assemble_name}) whose value is the address where the scope begins.
8685 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
8686 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
8689 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
8690 Define this macro if the target machine requires special handling to
8691 output an @code{N_FUN} entry for the function @var{decl}.
8694 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
8695 A C statement to output DBX debugging information before code for line
8696 number @var{line} of the current source file to the stdio stream
8697 @var{stream}. @var{counter} is the number of time the macro was
8698 invoked, including the current invocation; it is intended to generate
8699 unique labels in the assembly output.
8701 This macro should not be defined if the default output is correct, or
8702 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
8705 @defmac NO_DBX_FUNCTION_END
8706 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
8707 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
8708 On those machines, define this macro to turn this feature off without
8709 disturbing the rest of the gdb extensions.
8712 @defmac NO_DBX_BNSYM_ENSYM
8713 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
8714 extension construct. On those machines, define this macro to turn this
8715 feature off without disturbing the rest of the gdb extensions.
8718 @node File Names and DBX
8719 @subsection File Names in DBX Format
8721 @c prevent bad page break with this line
8722 This describes file names in DBX format.
8724 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
8725 A C statement to output DBX debugging information to the stdio stream
8726 @var{stream}, which indicates that file @var{name} is the main source
8727 file---the file specified as the input file for compilation.
8728 This macro is called only once, at the beginning of compilation.
8730 This macro need not be defined if the standard form of output
8731 for DBX debugging information is appropriate.
8733 It may be necessary to refer to a label equal to the beginning of the
8734 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
8735 to do so. If you do this, you must also set the variable
8736 @var{used_ltext_label_name} to @code{true}.
8739 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
8740 Define this macro, with value 1, if GCC should not emit an indication
8741 of the current directory for compilation and current source language at
8742 the beginning of the file.
8745 @defmac NO_DBX_GCC_MARKER
8746 Define this macro, with value 1, if GCC should not emit an indication
8747 that this object file was compiled by GCC@. The default is to emit
8748 an @code{N_OPT} stab at the beginning of every source file, with
8749 @samp{gcc2_compiled.} for the string and value 0.
8752 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
8753 A C statement to output DBX debugging information at the end of
8754 compilation of the main source file @var{name}. Output should be
8755 written to the stdio stream @var{stream}.
8757 If you don't define this macro, nothing special is output at the end
8758 of compilation, which is correct for most machines.
8761 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
8762 Define this macro @emph{instead of} defining
8763 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
8764 the end of compilation is a @code{N_SO} stab with an empty string,
8765 whose value is the highest absolute text address in the file.
8770 @subsection Macros for SDB and DWARF Output
8772 @c prevent bad page break with this line
8773 Here are macros for SDB and DWARF output.
8775 @defmac SDB_DEBUGGING_INFO
8776 Define this macro if GCC should produce COFF-style debugging output
8777 for SDB in response to the @option{-g} option.
8780 @defmac DWARF2_DEBUGGING_INFO
8781 Define this macro if GCC should produce dwarf version 2 format
8782 debugging output in response to the @option{-g} option.
8784 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (tree @var{function})
8785 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
8786 be emitted for each function. Instead of an integer return the enum
8787 value for the @code{DW_CC_} tag.
8790 To support optional call frame debugging information, you must also
8791 define @code{INCOMING_RETURN_ADDR_RTX} and either set
8792 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
8793 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
8794 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
8797 @defmac DWARF2_FRAME_INFO
8798 Define this macro to a nonzero value if GCC should always output
8799 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
8800 (@pxref{Exception Region Output} is nonzero, GCC will output this
8801 information not matter how you define @code{DWARF2_FRAME_INFO}.
8804 @defmac DWARF2_ASM_LINE_DEBUG_INFO
8805 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8806 line debug info sections. This will result in much more compact line number
8807 tables, and hence is desirable if it works.
8810 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8811 A C statement to issue assembly directives that create a difference
8812 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
8815 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
8816 A C statement to issue assembly directives that create a
8817 section-relative reference to the given @var{label}, using an integer of the
8818 given @var{size}. The label is known to be defined in the given @var{section}.
8821 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8822 A C statement to issue assembly directives that create a self-relative
8823 reference to the given @var{label}, using an integer of the given @var{size}.
8826 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{FILE}, int @var{size}, rtx @var{x})
8827 If defined, this target hook is a function which outputs a DTP-relative
8828 reference to the given TLS symbol of the specified size.
8831 @defmac PUT_SDB_@dots{}
8832 Define these macros to override the assembler syntax for the special
8833 SDB assembler directives. See @file{sdbout.c} for a list of these
8834 macros and their arguments. If the standard syntax is used, you need
8835 not define them yourself.
8839 Some assemblers do not support a semicolon as a delimiter, even between
8840 SDB assembler directives. In that case, define this macro to be the
8841 delimiter to use (usually @samp{\n}). It is not necessary to define
8842 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8846 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
8847 Define this macro to allow references to unknown structure,
8848 union, or enumeration tags to be emitted. Standard COFF does not
8849 allow handling of unknown references, MIPS ECOFF has support for
8853 @defmac SDB_ALLOW_FORWARD_REFERENCES
8854 Define this macro to allow references to structure, union, or
8855 enumeration tags that have not yet been seen to be handled. Some
8856 assemblers choke if forward tags are used, while some require it.
8859 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
8860 A C statement to output SDB debugging information before code for line
8861 number @var{line} of the current source file to the stdio stream
8862 @var{stream}. The default is to emit an @code{.ln} directive.
8867 @subsection Macros for VMS Debug Format
8869 @c prevent bad page break with this line
8870 Here are macros for VMS debug format.
8872 @defmac VMS_DEBUGGING_INFO
8873 Define this macro if GCC should produce debugging output for VMS
8874 in response to the @option{-g} option. The default behavior for VMS
8875 is to generate minimal debug info for a traceback in the absence of
8876 @option{-g} unless explicitly overridden with @option{-g0}. This
8877 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8878 @code{OVERRIDE_OPTIONS}.
8881 @node Floating Point
8882 @section Cross Compilation and Floating Point
8883 @cindex cross compilation and floating point
8884 @cindex floating point and cross compilation
8886 While all modern machines use twos-complement representation for integers,
8887 there are a variety of representations for floating point numbers. This
8888 means that in a cross-compiler the representation of floating point numbers
8889 in the compiled program may be different from that used in the machine
8890 doing the compilation.
8892 Because different representation systems may offer different amounts of
8893 range and precision, all floating point constants must be represented in
8894 the target machine's format. Therefore, the cross compiler cannot
8895 safely use the host machine's floating point arithmetic; it must emulate
8896 the target's arithmetic. To ensure consistency, GCC always uses
8897 emulation to work with floating point values, even when the host and
8898 target floating point formats are identical.
8900 The following macros are provided by @file{real.h} for the compiler to
8901 use. All parts of the compiler which generate or optimize
8902 floating-point calculations must use these macros. They may evaluate
8903 their operands more than once, so operands must not have side effects.
8905 @defmac REAL_VALUE_TYPE
8906 The C data type to be used to hold a floating point value in the target
8907 machine's format. Typically this is a @code{struct} containing an
8908 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8912 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8913 Compares for equality the two values, @var{x} and @var{y}. If the target
8914 floating point format supports negative zeroes and/or NaNs,
8915 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8916 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8919 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8920 Tests whether @var{x} is less than @var{y}.
8923 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8924 Truncates @var{x} to a signed integer, rounding toward zero.
8927 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8928 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8929 @var{x} is negative, returns zero.
8932 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8933 Converts @var{string} into a floating point number in the target machine's
8934 representation for mode @var{mode}. This routine can handle both
8935 decimal and hexadecimal floating point constants, using the syntax
8936 defined by the C language for both.
8939 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8940 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8943 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8944 Determines whether @var{x} represents infinity (positive or negative).
8947 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8948 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8951 @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})
8952 Calculates an arithmetic operation on the two floating point values
8953 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8956 The operation to be performed is specified by @var{code}. Only the
8957 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8958 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8960 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8961 target's floating point format cannot represent infinity, it will call
8962 @code{abort}. Callers should check for this situation first, using
8963 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8966 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8967 Returns the negative of the floating point value @var{x}.
8970 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8971 Returns the absolute value of @var{x}.
8974 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8975 Truncates the floating point value @var{x} to fit in @var{mode}. The
8976 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8977 appropriate bit pattern to be output as a floating constant whose
8978 precision accords with mode @var{mode}.
8981 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8982 Converts a floating point value @var{x} into a double-precision integer
8983 which is then stored into @var{low} and @var{high}. If the value is not
8984 integral, it is truncated.
8987 @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})
8988 Converts a double-precision integer found in @var{low} and @var{high},
8989 into a floating point value which is then stored into @var{x}. The
8990 value is truncated to fit in mode @var{mode}.
8993 @node Mode Switching
8994 @section Mode Switching Instructions
8995 @cindex mode switching
8996 The following macros control mode switching optimizations:
8998 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
8999 Define this macro if the port needs extra instructions inserted for mode
9000 switching in an optimizing compilation.
9002 For an example, the SH4 can perform both single and double precision
9003 floating point operations, but to perform a single precision operation,
9004 the FPSCR PR bit has to be cleared, while for a double precision
9005 operation, this bit has to be set. Changing the PR bit requires a general
9006 purpose register as a scratch register, hence these FPSCR sets have to
9007 be inserted before reload, i.e.@: you can't put this into instruction emitting
9008 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
9010 You can have multiple entities that are mode-switched, and select at run time
9011 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
9012 return nonzero for any @var{entity} that needs mode-switching.
9013 If you define this macro, you also have to define
9014 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
9015 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
9016 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
9020 @defmac NUM_MODES_FOR_MODE_SWITCHING
9021 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
9022 initializer for an array of integers. Each initializer element
9023 N refers to an entity that needs mode switching, and specifies the number
9024 of different modes that might need to be set for this entity.
9025 The position of the initializer in the initializer---starting counting at
9026 zero---determines the integer that is used to refer to the mode-switched
9028 In macros that take mode arguments / yield a mode result, modes are
9029 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
9030 switch is needed / supplied.
9033 @defmac MODE_NEEDED (@var{entity}, @var{insn})
9034 @var{entity} is an integer specifying a mode-switched entity. If
9035 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
9036 return an integer value not larger than the corresponding element in
9037 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
9038 be switched into prior to the execution of @var{insn}.
9041 @defmac MODE_AFTER (@var{mode}, @var{insn})
9042 If this macro is defined, it is evaluated for every @var{insn} during
9043 mode switching. It determines the mode that an insn results in (if
9044 different from the incoming mode).
9047 @defmac MODE_ENTRY (@var{entity})
9048 If this macro is defined, it is evaluated for every @var{entity} that needs
9049 mode switching. It should evaluate to an integer, which is a mode that
9050 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
9051 is defined then @code{MODE_EXIT} must be defined.
9054 @defmac MODE_EXIT (@var{entity})
9055 If this macro is defined, it is evaluated for every @var{entity} that needs
9056 mode switching. It should evaluate to an integer, which is a mode that
9057 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
9058 is defined then @code{MODE_ENTRY} must be defined.
9061 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
9062 This macro specifies the order in which modes for @var{entity} are processed.
9063 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
9064 lowest. The value of the macro should be an integer designating a mode
9065 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
9066 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
9067 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
9070 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
9071 Generate one or more insns to set @var{entity} to @var{mode}.
9072 @var{hard_reg_live} is the set of hard registers live at the point where
9073 the insn(s) are to be inserted.
9076 @node Target Attributes
9077 @section Defining target-specific uses of @code{__attribute__}
9078 @cindex target attributes
9079 @cindex machine attributes
9080 @cindex attributes, target-specific
9082 Target-specific attributes may be defined for functions, data and types.
9083 These are described using the following target hooks; they also need to
9084 be documented in @file{extend.texi}.
9086 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
9087 If defined, this target hook points to an array of @samp{struct
9088 attribute_spec} (defined in @file{tree.h}) specifying the machine
9089 specific attributes for this target and some of the restrictions on the
9090 entities to which these attributes are applied and the arguments they
9094 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
9095 If defined, this target hook is a function which returns zero if the attributes on
9096 @var{type1} and @var{type2} are incompatible, one if they are compatible,
9097 and two if they are nearly compatible (which causes a warning to be
9098 generated). If this is not defined, machine-specific attributes are
9099 supposed always to be compatible.
9102 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
9103 If defined, this target hook is a function which assigns default attributes to
9104 newly defined @var{type}.
9107 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
9108 Define this target hook if the merging of type attributes needs special
9109 handling. If defined, the result is a list of the combined
9110 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
9111 that @code{comptypes} has already been called and returned 1. This
9112 function may call @code{merge_attributes} to handle machine-independent
9116 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
9117 Define this target hook if the merging of decl attributes needs special
9118 handling. If defined, the result is a list of the combined
9119 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
9120 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
9121 when this is needed are when one attribute overrides another, or when an
9122 attribute is nullified by a subsequent definition. This function may
9123 call @code{merge_attributes} to handle machine-independent merging.
9125 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
9126 If the only target-specific handling you require is @samp{dllimport}
9127 for Microsoft Windows targets, you should define the macro
9128 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
9129 will then define a function called
9130 @code{merge_dllimport_decl_attributes} which can then be defined as
9131 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
9132 add @code{handle_dll_attribute} in the attribute table for your port
9133 to perform initial processing of the @samp{dllimport} and
9134 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
9135 @file{i386/i386.c}, for example.
9138 @deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (tree @var{decl})
9139 @var{decl} is a variable or function with @code{__attribute__((dllimport))}
9140 specified. Use this hook if the target needs to add extra validation
9141 checks to @code{handle_dll_attribute}.
9144 @defmac TARGET_DECLSPEC
9145 Define this macro to a nonzero value if you want to treat
9146 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
9147 default, this behavior is enabled only for targets that define
9148 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
9149 of @code{__declspec} is via a built-in macro, but you should not rely
9150 on this implementation detail.
9153 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
9154 Define this target hook if you want to be able to add attributes to a decl
9155 when it is being created. This is normally useful for back ends which
9156 wish to implement a pragma by using the attributes which correspond to
9157 the pragma's effect. The @var{node} argument is the decl which is being
9158 created. The @var{attr_ptr} argument is a pointer to the attribute list
9159 for this decl. The list itself should not be modified, since it may be
9160 shared with other decls, but attributes may be chained on the head of
9161 the list and @code{*@var{attr_ptr}} modified to point to the new
9162 attributes, or a copy of the list may be made if further changes are
9166 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
9168 This target hook returns @code{true} if it is ok to inline @var{fndecl}
9169 into the current function, despite its having target-specific
9170 attributes, @code{false} otherwise. By default, if a function has a
9171 target specific attribute attached to it, it will not be inlined.
9174 @node MIPS Coprocessors
9175 @section Defining coprocessor specifics for MIPS targets.
9176 @cindex MIPS coprocessor-definition macros
9178 The MIPS specification allows MIPS implementations to have as many as 4
9179 coprocessors, each with as many as 32 private registers. GCC supports
9180 accessing these registers and transferring values between the registers
9181 and memory using asm-ized variables. For example:
9184 register unsigned int cp0count asm ("c0r1");
9190 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
9191 names may be added as described below, or the default names may be
9192 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
9194 Coprocessor registers are assumed to be epilogue-used; sets to them will
9195 be preserved even if it does not appear that the register is used again
9196 later in the function.
9198 Another note: according to the MIPS spec, coprocessor 1 (if present) is
9199 the FPU@. One accesses COP1 registers through standard mips
9200 floating-point support; they are not included in this mechanism.
9202 There is one macro used in defining the MIPS coprocessor interface which
9203 you may want to override in subtargets; it is described below.
9205 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
9206 A comma-separated list (with leading comma) of pairs describing the
9207 alternate names of coprocessor registers. The format of each entry should be
9209 @{ @var{alternatename}, @var{register_number}@}
9215 @section Parameters for Precompiled Header Validity Checking
9216 @cindex parameters, precompiled headers
9218 @deftypefn {Target Hook} void *TARGET_GET_PCH_VALIDITY (size_t *@var{sz})
9219 This hook returns the data needed by @code{TARGET_PCH_VALID_P} and sets
9220 @samp{*@var{sz}} to the size of the data in bytes.
9223 @deftypefn {Target Hook} const char *TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz})
9224 This hook checks whether the options used to create a PCH file are
9225 compatible with the current settings. It returns @code{NULL}
9226 if so and a suitable error message if not. Error messages will
9227 be presented to the user and must be localized using @samp{_(@var{msg})}.
9229 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
9230 when the PCH file was created and @var{sz} is the size of that data in bytes.
9231 It's safe to assume that the data was created by the same version of the
9232 compiler, so no format checking is needed.
9234 The default definition of @code{default_pch_valid_p} should be
9235 suitable for most targets.
9238 @deftypefn {Target Hook} const char *TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags})
9239 If this hook is nonnull, the default implementation of
9240 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
9241 of @code{target_flags}. @var{pch_flags} specifies the value that
9242 @code{target_flags} had when the PCH file was created. The return
9243 value is the same as for @code{TARGET_PCH_VALID_P}.
9247 @section C++ ABI parameters
9248 @cindex parameters, c++ abi
9250 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
9251 Define this hook to override the integer type used for guard variables.
9252 These are used to implement one-time construction of static objects. The
9253 default is long_long_integer_type_node.
9256 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
9257 This hook determines how guard variables are used. It should return
9258 @code{false} (the default) if first byte should be used. A return value of
9259 @code{true} indicates the least significant bit should be used.
9262 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
9263 This hook returns the size of the cookie to use when allocating an array
9264 whose elements have the indicated @var{type}. Assumes that it is already
9265 known that a cookie is needed. The default is
9266 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
9267 IA64/Generic C++ ABI@.
9270 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
9271 This hook should return @code{true} if the element size should be stored in
9272 array cookies. The default is to return @code{false}.
9275 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
9276 If defined by a backend this hook allows the decision made to export
9277 class @var{type} to be overruled. Upon entry @var{import_export}
9278 will contain 1 if the class is going to be exported, @minus{}1 if it is going
9279 to be imported and 0 otherwise. This function should return the
9280 modified value and perform any other actions necessary to support the
9281 backend's targeted operating system.
9284 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
9285 This hook should return @code{true} if constructors and destructors return
9286 the address of the object created/destroyed. The default is to return
9290 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
9291 This hook returns true if the key method for a class (i.e., the method
9292 which, if defined in the current translation unit, causes the virtual
9293 table to be emitted) may be an inline function. Under the standard
9294 Itanium C++ ABI the key method may be an inline function so long as
9295 the function is not declared inline in the class definition. Under
9296 some variants of the ABI, an inline function can never be the key
9297 method. The default is to return @code{true}.
9300 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
9301 @var{decl} is a virtual table, virtual table table, typeinfo object,
9302 or other similar implicit class data object that will be emitted with
9303 external linkage in this translation unit. No ELF visibility has been
9304 explicitly specified. If the target needs to specify a visibility
9305 other than that of the containing class, use this hook to set
9306 @code{DECL_VISIBILITY} and @code{DECL_VISIBILITY_SPECIFIED}.
9309 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
9310 This hook returns true (the default) if virtual tables and other
9311 similar implicit class data objects are always COMDAT if they have
9312 external linkage. If this hook returns false, then class data for
9313 classes whose virtual table will be emitted in only one translation
9314 unit will not be COMDAT.
9317 @deftypefn {Target Hook} bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void)
9318 This hook returns true (the default) if the RTTI information for
9319 the basic types which is defined in the C++ runtime should always
9320 be COMDAT, false if it should not be COMDAT.
9323 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
9324 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
9325 should be used to register static destructors when @option{-fuse-cxa-atexit}
9326 is in effect. The default is to return false to use @code{__cxa_atexit}.
9329 @deftypefn {Target Hook} bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void)
9330 This hook returns true if the target @code{atexit} function can be used
9331 in the same manner as @code{__cxa_atexit} to register C++ static
9332 destructors. This requires that @code{atexit}-registered functions in
9333 shared libraries are run in the correct order when the libraries are
9334 unloaded. The default is to return false.
9337 @deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type})
9338 @var{type} is a C++ class (i.e., RECORD_TYPE or UNION_TYPE) that has just been
9339 defined. Use this hook to make adjustments to the class (eg, tweak
9340 visibility or perform any other required target modifications).
9344 @section Miscellaneous Parameters
9345 @cindex parameters, miscellaneous
9347 @c prevent bad page break with this line
9348 Here are several miscellaneous parameters.
9350 @defmac HAS_LONG_COND_BRANCH
9351 Define this boolean macro to indicate whether or not your architecture
9352 has conditional branches that can span all of memory. It is used in
9353 conjunction with an optimization that partitions hot and cold basic
9354 blocks into separate sections of the executable. If this macro is
9355 set to false, gcc will convert any conditional branches that attempt
9356 to cross between sections into unconditional branches or indirect jumps.
9359 @defmac HAS_LONG_UNCOND_BRANCH
9360 Define this boolean macro to indicate whether or not your architecture
9361 has unconditional branches that can span all of memory. It is used in
9362 conjunction with an optimization that partitions hot and cold basic
9363 blocks into separate sections of the executable. If this macro is
9364 set to false, gcc will convert any unconditional branches that attempt
9365 to cross between sections into indirect jumps.
9368 @defmac CASE_VECTOR_MODE
9369 An alias for a machine mode name. This is the machine mode that
9370 elements of a jump-table should have.
9373 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
9374 Optional: return the preferred mode for an @code{addr_diff_vec}
9375 when the minimum and maximum offset are known. If you define this,
9376 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
9377 To make this work, you also have to define @code{INSN_ALIGN} and
9378 make the alignment for @code{addr_diff_vec} explicit.
9379 The @var{body} argument is provided so that the offset_unsigned and scale
9380 flags can be updated.
9383 @defmac CASE_VECTOR_PC_RELATIVE
9384 Define this macro to be a C expression to indicate when jump-tables
9385 should contain relative addresses. You need not define this macro if
9386 jump-tables never contain relative addresses, or jump-tables should
9387 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
9391 @defmac CASE_VALUES_THRESHOLD
9392 Define this to be the smallest number of different values for which it
9393 is best to use a jump-table instead of a tree of conditional branches.
9394 The default is four for machines with a @code{casesi} instruction and
9395 five otherwise. This is best for most machines.
9398 @defmac CASE_USE_BIT_TESTS
9399 Define this macro to be a C expression to indicate whether C switch
9400 statements may be implemented by a sequence of bit tests. This is
9401 advantageous on processors that can efficiently implement left shift
9402 of 1 by the number of bits held in a register, but inappropriate on
9403 targets that would require a loop. By default, this macro returns
9404 @code{true} if the target defines an @code{ashlsi3} pattern, and
9405 @code{false} otherwise.
9408 @defmac WORD_REGISTER_OPERATIONS
9409 Define this macro if operations between registers with integral mode
9410 smaller than a word are always performed on the entire register.
9411 Most RISC machines have this property and most CISC machines do not.
9414 @defmac LOAD_EXTEND_OP (@var{mem_mode})
9415 Define this macro to be a C expression indicating when insns that read
9416 memory in @var{mem_mode}, an integral mode narrower than a word, set the
9417 bits outside of @var{mem_mode} to be either the sign-extension or the
9418 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
9419 of @var{mem_mode} for which the
9420 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
9421 @code{UNKNOWN} for other modes.
9423 This macro is not called with @var{mem_mode} non-integral or with a width
9424 greater than or equal to @code{BITS_PER_WORD}, so you may return any
9425 value in this case. Do not define this macro if it would always return
9426 @code{UNKNOWN}. On machines where this macro is defined, you will normally
9427 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
9429 You may return a non-@code{UNKNOWN} value even if for some hard registers
9430 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
9431 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
9432 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
9433 integral mode larger than this but not larger than @code{word_mode}.
9435 You must return @code{UNKNOWN} if for some hard registers that allow this
9436 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
9437 @code{word_mode}, but that they can change to another integral mode that
9438 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
9441 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
9442 Define this macro if loading short immediate values into registers sign
9446 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
9447 Define this macro if the same instructions that convert a floating
9448 point number to a signed fixed point number also convert validly to an
9452 @deftypefn {Target Hook} int TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (enum machine_mode @var{mode})
9453 When @option{-ffast-math} is in effect, GCC tries to optimize
9454 divisions by the same divisor, by turning them into multiplications by
9455 the reciprocal. This target hook specifies the minimum number of divisions
9456 that should be there for GCC to perform the optimization for a variable
9457 of mode @var{mode}. The default implementation returns 3 if the machine
9458 has an instruction for the division, and 2 if it does not.
9462 The maximum number of bytes that a single instruction can move quickly
9463 between memory and registers or between two memory locations.
9466 @defmac MAX_MOVE_MAX
9467 The maximum number of bytes that a single instruction can move quickly
9468 between memory and registers or between two memory locations. If this
9469 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
9470 constant value that is the largest value that @code{MOVE_MAX} can have
9474 @defmac SHIFT_COUNT_TRUNCATED
9475 A C expression that is nonzero if on this machine the number of bits
9476 actually used for the count of a shift operation is equal to the number
9477 of bits needed to represent the size of the object being shifted. When
9478 this macro is nonzero, the compiler will assume that it is safe to omit
9479 a sign-extend, zero-extend, and certain bitwise `and' instructions that
9480 truncates the count of a shift operation. On machines that have
9481 instructions that act on bit-fields at variable positions, which may
9482 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
9483 also enables deletion of truncations of the values that serve as
9484 arguments to bit-field instructions.
9486 If both types of instructions truncate the count (for shifts) and
9487 position (for bit-field operations), or if no variable-position bit-field
9488 instructions exist, you should define this macro.
9490 However, on some machines, such as the 80386 and the 680x0, truncation
9491 only applies to shift operations and not the (real or pretended)
9492 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
9493 such machines. Instead, add patterns to the @file{md} file that include
9494 the implied truncation of the shift instructions.
9496 You need not define this macro if it would always have the value of zero.
9499 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
9500 @deftypefn {Target Hook} int TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode})
9501 This function describes how the standard shift patterns for @var{mode}
9502 deal with shifts by negative amounts or by more than the width of the mode.
9503 @xref{shift patterns}.
9505 On many machines, the shift patterns will apply a mask @var{m} to the
9506 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
9507 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
9508 this is true for mode @var{mode}, the function should return @var{m},
9509 otherwise it should return 0. A return value of 0 indicates that no
9510 particular behavior is guaranteed.
9512 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
9513 @emph{not} apply to general shift rtxes; it applies only to instructions
9514 that are generated by the named shift patterns.
9516 The default implementation of this function returns
9517 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
9518 and 0 otherwise. This definition is always safe, but if
9519 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
9520 nevertheless truncate the shift count, you may get better code
9524 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
9525 A C expression which is nonzero if on this machine it is safe to
9526 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
9527 bits (where @var{outprec} is smaller than @var{inprec}) by merely
9528 operating on it as if it had only @var{outprec} bits.
9530 On many machines, this expression can be 1.
9532 @c rearranged this, removed the phrase "it is reported that". this was
9533 @c to fix an overfull hbox. --mew 10feb93
9534 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
9535 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
9536 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
9537 such cases may improve things.
9540 @deftypefn {Target Hook} int TARGET_MODE_REP_EXTENDED (enum machine_mode @var{mode}, enum machine_mode @var{rep_mode})
9541 The representation of an integral mode can be such that the values
9542 are always extended to a wider integral mode. Return
9543 @code{SIGN_EXTEND} if values of @var{mode} are represented in
9544 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
9545 otherwise. (Currently, none of the targets use zero-extended
9546 representation this way so unlike @code{LOAD_EXTEND_OP},
9547 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
9548 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
9549 @var{mode} to @var{mode_rep} so that @var{mode_rep} is not the next
9550 widest integral mode and currently we take advantage of this fact.)
9552 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
9553 value even if the extension is not performed on certain hard registers
9554 as long as for the @code{REGNO_REG_CLASS} of these hard registers
9555 @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero.
9557 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
9558 describe two related properties. If you define
9559 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
9560 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
9563 In order to enforce the representation of @code{mode},
9564 @code{TRULY_NOOP_TRUNCATION} should return false when truncating to
9568 @defmac STORE_FLAG_VALUE
9569 A C expression describing the value returned by a comparison operator
9570 with an integral mode and stored by a store-flag instruction
9571 (@samp{s@var{cond}}) when the condition is true. This description must
9572 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
9573 comparison operators whose results have a @code{MODE_INT} mode.
9575 A value of 1 or @minus{}1 means that the instruction implementing the
9576 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
9577 and 0 when the comparison is false. Otherwise, the value indicates
9578 which bits of the result are guaranteed to be 1 when the comparison is
9579 true. This value is interpreted in the mode of the comparison
9580 operation, which is given by the mode of the first operand in the
9581 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
9582 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
9585 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
9586 generate code that depends only on the specified bits. It can also
9587 replace comparison operators with equivalent operations if they cause
9588 the required bits to be set, even if the remaining bits are undefined.
9589 For example, on a machine whose comparison operators return an
9590 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
9591 @samp{0x80000000}, saying that just the sign bit is relevant, the
9595 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
9602 (ashift:SI @var{x} (const_int @var{n}))
9606 where @var{n} is the appropriate shift count to move the bit being
9607 tested into the sign bit.
9609 There is no way to describe a machine that always sets the low-order bit
9610 for a true value, but does not guarantee the value of any other bits,
9611 but we do not know of any machine that has such an instruction. If you
9612 are trying to port GCC to such a machine, include an instruction to
9613 perform a logical-and of the result with 1 in the pattern for the
9614 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
9616 Often, a machine will have multiple instructions that obtain a value
9617 from a comparison (or the condition codes). Here are rules to guide the
9618 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
9623 Use the shortest sequence that yields a valid definition for
9624 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
9625 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
9626 comparison operators to do so because there may be opportunities to
9627 combine the normalization with other operations.
9630 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
9631 slightly preferred on machines with expensive jumps and 1 preferred on
9635 As a second choice, choose a value of @samp{0x80000001} if instructions
9636 exist that set both the sign and low-order bits but do not define the
9640 Otherwise, use a value of @samp{0x80000000}.
9643 Many machines can produce both the value chosen for
9644 @code{STORE_FLAG_VALUE} and its negation in the same number of
9645 instructions. On those machines, you should also define a pattern for
9646 those cases, e.g., one matching
9649 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
9652 Some machines can also perform @code{and} or @code{plus} operations on
9653 condition code values with less instructions than the corresponding
9654 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
9655 machines, define the appropriate patterns. Use the names @code{incscc}
9656 and @code{decscc}, respectively, for the patterns which perform
9657 @code{plus} or @code{minus} operations on condition code values. See
9658 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
9659 find such instruction sequences on other machines.
9661 If this macro is not defined, the default value, 1, is used. You need
9662 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
9663 instructions, or if the value generated by these instructions is 1.
9666 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
9667 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
9668 returned when comparison operators with floating-point results are true.
9669 Define this macro on machines that have comparison operations that return
9670 floating-point values. If there are no such operations, do not define
9674 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
9675 A C expression that gives a rtx representing the nonzero true element
9676 for vector comparisons. The returned rtx should be valid for the inner
9677 mode of @var{mode} which is guaranteed to be a vector mode. Define
9678 this macro on machines that have vector comparison operations that
9679 return a vector result. If there are no such operations, do not define
9680 this macro. Typically, this macro is defined as @code{const1_rtx} or
9681 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
9682 the compiler optimizing such vector comparison operations for the
9686 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
9687 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
9688 A C expression that indicates whether the architecture defines a value
9689 for @code{clz} or @code{ctz} with a zero operand.
9690 A result of @code{0} indicates the value is undefined.
9691 If the value is defined for only the RTL expression, the macro should
9692 evaluate to @code{1}; if the value applies also to the corresponding optab
9693 entry (which is normally the case if it expands directly into
9694 the corresponding RTL), then the macro should evaluate to @code{2}.
9695 In the cases where the value is defined, @var{value} should be set to
9698 If this macro is not defined, the value of @code{clz} or
9699 @code{ctz} at zero is assumed to be undefined.
9701 This macro must be defined if the target's expansion for @code{ffs}
9702 relies on a particular value to get correct results. Otherwise it
9703 is not necessary, though it may be used to optimize some corner cases, and
9704 to provide a default expansion for the @code{ffs} optab.
9706 Note that regardless of this macro the ``definedness'' of @code{clz}
9707 and @code{ctz} at zero do @emph{not} extend to the builtin functions
9708 visible to the user. Thus one may be free to adjust the value at will
9709 to match the target expansion of these operations without fear of
9714 An alias for the machine mode for pointers. On most machines, define
9715 this to be the integer mode corresponding to the width of a hardware
9716 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
9717 On some machines you must define this to be one of the partial integer
9718 modes, such as @code{PSImode}.
9720 The width of @code{Pmode} must be at least as large as the value of
9721 @code{POINTER_SIZE}. If it is not equal, you must define the macro
9722 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
9726 @defmac FUNCTION_MODE
9727 An alias for the machine mode used for memory references to functions
9728 being called, in @code{call} RTL expressions. On most CISC machines,
9729 where an instruction can begin at any byte address, this should be
9730 @code{QImode}. On most RISC machines, where all instructions have fixed
9731 size and alignment, this should be a mode with the same size and alignment
9732 as the machine instruction words - typically @code{SImode} or @code{HImode}.
9735 @defmac STDC_0_IN_SYSTEM_HEADERS
9736 In normal operation, the preprocessor expands @code{__STDC__} to the
9737 constant 1, to signify that GCC conforms to ISO Standard C@. On some
9738 hosts, like Solaris, the system compiler uses a different convention,
9739 where @code{__STDC__} is normally 0, but is 1 if the user specifies
9740 strict conformance to the C Standard.
9742 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
9743 convention when processing system header files, but when processing user
9744 files @code{__STDC__} will always expand to 1.
9747 @defmac NO_IMPLICIT_EXTERN_C
9748 Define this macro if the system header files support C++ as well as C@.
9749 This macro inhibits the usual method of using system header files in
9750 C++, which is to pretend that the file's contents are enclosed in
9751 @samp{extern "C" @{@dots{}@}}.
9756 @defmac REGISTER_TARGET_PRAGMAS ()
9757 Define this macro if you want to implement any target-specific pragmas.
9758 If defined, it is a C expression which makes a series of calls to
9759 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
9760 for each pragma. The macro may also do any
9761 setup required for the pragmas.
9763 The primary reason to define this macro is to provide compatibility with
9764 other compilers for the same target. In general, we discourage
9765 definition of target-specific pragmas for GCC@.
9767 If the pragma can be implemented by attributes then you should consider
9768 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
9770 Preprocessor macros that appear on pragma lines are not expanded. All
9771 @samp{#pragma} directives that do not match any registered pragma are
9772 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
9775 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9776 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9778 Each call to @code{c_register_pragma} or
9779 @code{c_register_pragma_with_expansion} establishes one pragma. The
9780 @var{callback} routine will be called when the preprocessor encounters a
9784 #pragma [@var{space}] @var{name} @dots{}
9787 @var{space} is the case-sensitive namespace of the pragma, or
9788 @code{NULL} to put the pragma in the global namespace. The callback
9789 routine receives @var{pfile} as its first argument, which can be passed
9790 on to cpplib's functions if necessary. You can lex tokens after the
9791 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
9792 callback will be silently ignored. The end of the line is indicated by
9793 a token of type @code{CPP_EOF}. Macro expansion occurs on the
9794 arguments of pragmas registered with
9795 @code{c_register_pragma_with_expansion} but not on the arguments of
9796 pragmas registered with @code{c_register_pragma}.
9798 For an example use of this routine, see @file{c4x.h} and the callback
9799 routines defined in @file{c4x-c.c}.
9801 Note that the use of @code{pragma_lex} is specific to the C and C++
9802 compilers. It will not work in the Java or Fortran compilers, or any
9803 other language compilers for that matter. Thus if @code{pragma_lex} is going
9804 to be called from target-specific code, it must only be done so when
9805 building the C and C++ compilers. This can be done by defining the
9806 variables @code{c_target_objs} and @code{cxx_target_objs} in the
9807 target entry in the @file{config.gcc} file. These variables should name
9808 the target-specific, language-specific object file which contains the
9809 code that uses @code{pragma_lex}. Note it will also be necessary to add a
9810 rule to the makefile fragment pointed to by @code{tmake_file} that shows
9811 how to build this object file.
9816 @defmac HANDLE_SYSV_PRAGMA
9817 Define this macro (to a value of 1) if you want the System V style
9818 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
9819 [=<value>]} to be supported by gcc.
9821 The pack pragma specifies the maximum alignment (in bytes) of fields
9822 within a structure, in much the same way as the @samp{__aligned__} and
9823 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
9824 the behavior to the default.
9826 A subtlety for Microsoft Visual C/C++ style bit-field packing
9827 (e.g.@: -mms-bitfields) for targets that support it:
9828 When a bit-field is inserted into a packed record, the whole size
9829 of the underlying type is used by one or more same-size adjacent
9830 bit-fields (that is, if its long:3, 32 bits is used in the record,
9831 and any additional adjacent long bit-fields are packed into the same
9832 chunk of 32 bits. However, if the size changes, a new field of that
9835 If both MS bit-fields and @samp{__attribute__((packed))} are used,
9836 the latter will take precedence. If @samp{__attribute__((packed))} is
9837 used on a single field when MS bit-fields are in use, it will take
9838 precedence for that field, but the alignment of the rest of the structure
9839 may affect its placement.
9841 The weak pragma only works if @code{SUPPORTS_WEAK} and
9842 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
9843 of specifically named weak labels, optionally with a value.
9848 @defmac HANDLE_PRAGMA_PACK_PUSH_POP
9849 Define this macro (to a value of 1) if you want to support the Win32
9850 style pragmas @samp{#pragma pack(push[,@var{n}])} and @samp{#pragma
9851 pack(pop)}. The @samp{pack(push,[@var{n}])} pragma specifies the maximum
9852 alignment (in bytes) of fields within a structure, in much the same way as
9853 the @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
9854 pack value of zero resets the behavior to the default. Successive
9855 invocations of this pragma cause the previous values to be stacked, so
9856 that invocations of @samp{#pragma pack(pop)} will return to the previous
9860 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
9861 Define this macro, as well as
9862 @code{HANDLE_SYSV_PRAGMA}, if macros should be expanded in the
9863 arguments of @samp{#pragma pack}.
9866 @defmac TARGET_DEFAULT_PACK_STRUCT
9867 If your target requires a structure packing default other than 0 (meaning
9868 the machine default), define this macro to the necessary value (in bytes).
9869 This must be a value that would also be valid to use with
9870 @samp{#pragma pack()} (that is, a small power of two).
9875 @defmac HANDLE_PRAGMA_PUSH_POP_MACRO
9876 Define this macro if you want to support the Win32 style pragmas
9877 @samp{#pragma push_macro(macro-name-as-string)} and @samp{#pragma
9878 pop_macro(macro-name-as-string)}. The @samp{#pragma push_macro(
9879 macro-name-as-string)} pragma saves the named macro and via
9880 @samp{#pragma pop_macro(macro-name-as-string)} it will return to the
9885 @defmac DOLLARS_IN_IDENTIFIERS
9886 Define this macro to control use of the character @samp{$} in
9887 identifier names for the C family of languages. 0 means @samp{$} is
9888 not allowed by default; 1 means it is allowed. 1 is the default;
9889 there is no need to define this macro in that case.
9892 @defmac NO_DOLLAR_IN_LABEL
9893 Define this macro if the assembler does not accept the character
9894 @samp{$} in label names. By default constructors and destructors in
9895 G++ have @samp{$} in the identifiers. If this macro is defined,
9896 @samp{.} is used instead.
9899 @defmac NO_DOT_IN_LABEL
9900 Define this macro if the assembler does not accept the character
9901 @samp{.} in label names. By default constructors and destructors in G++
9902 have names that use @samp{.}. If this macro is defined, these names
9903 are rewritten to avoid @samp{.}.
9906 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
9907 Define this macro as a C expression that is nonzero if it is safe for the
9908 delay slot scheduler to place instructions in the delay slot of @var{insn},
9909 even if they appear to use a resource set or clobbered in @var{insn}.
9910 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
9911 every @code{call_insn} has this behavior. On machines where some @code{insn}
9912 or @code{jump_insn} is really a function call and hence has this behavior,
9913 you should define this macro.
9915 You need not define this macro if it would always return zero.
9918 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
9919 Define this macro as a C expression that is nonzero if it is safe for the
9920 delay slot scheduler to place instructions in the delay slot of @var{insn},
9921 even if they appear to set or clobber a resource referenced in @var{insn}.
9922 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
9923 some @code{insn} or @code{jump_insn} is really a function call and its operands
9924 are registers whose use is actually in the subroutine it calls, you should
9925 define this macro. Doing so allows the delay slot scheduler to move
9926 instructions which copy arguments into the argument registers into the delay
9929 You need not define this macro if it would always return zero.
9932 @defmac MULTIPLE_SYMBOL_SPACES
9933 Define this macro as a C expression that is nonzero if, in some cases,
9934 global symbols from one translation unit may not be bound to undefined
9935 symbols in another translation unit without user intervention. For
9936 instance, under Microsoft Windows symbols must be explicitly imported
9937 from shared libraries (DLLs).
9939 You need not define this macro if it would always evaluate to zero.
9942 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{outputs}, tree @var{inputs}, tree @var{clobbers})
9943 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
9944 any hard regs the port wishes to automatically clobber for an asm.
9945 It should return the result of the last @code{tree_cons} used to add a
9946 clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the
9947 corresponding parameters to the asm and may be inspected to avoid
9948 clobbering a register that is an input or output of the asm. You can use
9949 @code{tree_overlaps_hard_reg_set}, declared in @file{tree.h}, to test
9950 for overlap with regards to asm-declared registers.
9953 @defmac MATH_LIBRARY
9954 Define this macro as a C string constant for the linker argument to link
9955 in the system math library, or @samp{""} if the target does not have a
9956 separate math library.
9958 You need only define this macro if the default of @samp{"-lm"} is wrong.
9961 @defmac LIBRARY_PATH_ENV
9962 Define this macro as a C string constant for the environment variable that
9963 specifies where the linker should look for libraries.
9965 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
9969 @defmac TARGET_POSIX_IO
9970 Define this macro if the target supports the following POSIX@ file
9971 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
9972 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
9973 to use file locking when exiting a program, which avoids race conditions
9974 if the program has forked. It will also create directories at run-time
9975 for cross-profiling.
9978 @defmac MAX_CONDITIONAL_EXECUTE
9980 A C expression for the maximum number of instructions to execute via
9981 conditional execution instructions instead of a branch. A value of
9982 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
9983 1 if it does use cc0.
9986 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
9987 Used if the target needs to perform machine-dependent modifications on the
9988 conditionals used for turning basic blocks into conditionally executed code.
9989 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
9990 contains information about the currently processed blocks. @var{true_expr}
9991 and @var{false_expr} are the tests that are used for converting the
9992 then-block and the else-block, respectively. Set either @var{true_expr} or
9993 @var{false_expr} to a null pointer if the tests cannot be converted.
9996 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
9997 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
9998 if-statements into conditions combined by @code{and} and @code{or} operations.
9999 @var{bb} contains the basic block that contains the test that is currently
10000 being processed and about to be turned into a condition.
10003 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
10004 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
10005 be converted to conditional execution format. @var{ce_info} points to
10006 a data structure, @code{struct ce_if_block}, which contains information
10007 about the currently processed blocks.
10010 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
10011 A C expression to perform any final machine dependent modifications in
10012 converting code to conditional execution. The involved basic blocks
10013 can be found in the @code{struct ce_if_block} structure that is pointed
10014 to by @var{ce_info}.
10017 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
10018 A C expression to cancel any machine dependent modifications in
10019 converting code to conditional execution. The involved basic blocks
10020 can be found in the @code{struct ce_if_block} structure that is pointed
10021 to by @var{ce_info}.
10024 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
10025 A C expression to initialize any extra fields in a @code{struct ce_if_block}
10026 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
10029 @defmac IFCVT_EXTRA_FIELDS
10030 If defined, it should expand to a set of field declarations that will be
10031 added to the @code{struct ce_if_block} structure. These should be initialized
10032 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
10035 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG ()
10036 If non-null, this hook performs a target-specific pass over the
10037 instruction stream. The compiler will run it at all optimization levels,
10038 just before the point at which it normally does delayed-branch scheduling.
10040 The exact purpose of the hook varies from target to target. Some use
10041 it to do transformations that are necessary for correctness, such as
10042 laying out in-function constant pools or avoiding hardware hazards.
10043 Others use it as an opportunity to do some machine-dependent optimizations.
10045 You need not implement the hook if it has nothing to do. The default
10046 definition is null.
10049 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
10050 Define this hook if you have any machine-specific built-in functions
10051 that need to be defined. It should be a function that performs the
10054 Machine specific built-in functions can be useful to expand special machine
10055 instructions that would otherwise not normally be generated because
10056 they have no equivalent in the source language (for example, SIMD vector
10057 instructions or prefetch instructions).
10059 To create a built-in function, call the function
10060 @code{lang_hooks.builtin_function}
10061 which is defined by the language front end. You can use any type nodes set
10062 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
10063 only language front ends that use those two functions will call
10064 @samp{TARGET_INIT_BUILTINS}.
10067 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
10069 Expand a call to a machine specific built-in function that was set up by
10070 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
10071 function call; the result should go to @var{target} if that is
10072 convenient, and have mode @var{mode} if that is convenient.
10073 @var{subtarget} may be used as the target for computing one of
10074 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
10075 ignored. This function should return the result of the call to the
10079 @deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (tree @var{fndecl}, tree @var{arglist})
10081 Select a replacement for a machine specific built-in function that
10082 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
10083 @emph{before} regular type checking, and so allows the target to
10084 implement a crude form of function overloading. @var{fndecl} is the
10085 declaration of the built-in function. @var{arglist} is the list of
10086 arguments passed to the built-in function. The result is a
10087 complete expression that implements the operation, usually
10088 another @code{CALL_EXPR}.
10091 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, tree @var{arglist}, bool @var{ignore})
10093 Fold a call to a machine specific built-in function that was set up by
10094 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
10095 built-in function. @var{arglist} is the list of arguments passed to
10096 the built-in function. The result is another tree containing a
10097 simplified expression for the call's result. If @var{ignore} is true
10098 the value will be ignored.
10101 @deftypefn {Target Hook} const char * TARGET_INVALID_WITHIN_DOLOOP (rtx @var{insn})
10103 Take an instruction in @var{insn} and return NULL if it is valid within a
10104 low-overhead loop, otherwise return a string why doloop could not be applied.
10106 Many targets use special registers for low-overhead looping. For any
10107 instruction that clobbers these this function should return a string indicating
10108 the reason why the doloop could not be applied.
10109 By default, the RTL loop optimizer does not use a present doloop pattern for
10110 loops containing function calls or branch on table instructions.
10113 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
10115 Take a branch insn in @var{branch1} and another in @var{branch2}.
10116 Return true if redirecting @var{branch1} to the destination of
10117 @var{branch2} is possible.
10119 On some targets, branches may have a limited range. Optimizing the
10120 filling of delay slots can result in branches being redirected, and this
10121 may in turn cause a branch offset to overflow.
10124 @deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (rtx @var{x}, @var{outer_code})
10125 This target hook returns @code{true} if @var{x} is considered to be commutative.
10126 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
10127 PLUS to be commutative inside a MEM. @var{outer_code} is the rtx code
10128 of the enclosing rtl, if known, otherwise it is UNKNOWN.
10131 @deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg})
10133 When the initial value of a hard register has been copied in a pseudo
10134 register, it is often not necessary to actually allocate another register
10135 to this pseudo register, because the original hard register or a stack slot
10136 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
10137 is called at the start of register allocation once for each hard register
10138 that had its initial value copied by using
10139 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
10140 Possible values are @code{NULL_RTX}, if you don't want
10141 to do any special allocation, a @code{REG} rtx---that would typically be
10142 the hard register itself, if it is known not to be clobbered---or a
10144 If you are returning a @code{MEM}, this is only a hint for the allocator;
10145 it might decide to use another register anyways.
10146 You may use @code{current_function_leaf_function} in the hook, functions
10147 that use @code{REG_N_SETS}, to determine if the hard
10148 register in question will not be clobbered.
10149 The default value of this hook is @code{NULL}, which disables any special
10153 @deftypefn {Target Hook} void TARGET_SET_CURRENT_FUNCTION (tree @var{decl})
10154 The compiler invokes this hook whenever it changes its current function
10155 context (@code{cfun}). You can define this function if
10156 the back end needs to perform any initialization or reset actions on a
10157 per-function basis. For example, it may be used to implement function
10158 attributes that affect register usage or code generation patterns.
10159 The argument @var{decl} is the declaration for the new function context,
10160 and may be null to indicate that the compiler has left a function context
10161 and is returning to processing at the top level.
10162 The default hook function does nothing.
10164 GCC sets @code{cfun} to a dummy function context during initialization of
10165 some parts of the back end. The hook function is not invoked in this
10166 situation; you need not worry about the hook being invoked recursively,
10167 or when the back end is in a partially-initialized state.
10170 @defmac TARGET_OBJECT_SUFFIX
10171 Define this macro to be a C string representing the suffix for object
10172 files on your target machine. If you do not define this macro, GCC will
10173 use @samp{.o} as the suffix for object files.
10176 @defmac TARGET_EXECUTABLE_SUFFIX
10177 Define this macro to be a C string representing the suffix to be
10178 automatically added to executable files on your target machine. If you
10179 do not define this macro, GCC will use the null string as the suffix for
10183 @defmac COLLECT_EXPORT_LIST
10184 If defined, @code{collect2} will scan the individual object files
10185 specified on its command line and create an export list for the linker.
10186 Define this macro for systems like AIX, where the linker discards
10187 object files that are not referenced from @code{main} and uses export
10191 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
10192 Define this macro to a C expression representing a variant of the
10193 method call @var{mdecl}, if Java Native Interface (JNI) methods
10194 must be invoked differently from other methods on your target.
10195 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
10196 the @code{stdcall} calling convention and this macro is then
10197 defined as this expression:
10200 build_type_attribute_variant (@var{mdecl},
10202 (get_identifier ("stdcall"),
10207 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
10208 This target hook returns @code{true} past the point in which new jump
10209 instructions could be created. On machines that require a register for
10210 every jump such as the SHmedia ISA of SH5, this point would typically be
10211 reload, so this target hook should be defined to a function such as:
10215 cannot_modify_jumps_past_reload_p ()
10217 return (reload_completed || reload_in_progress);
10222 @deftypefn {Target Hook} int TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
10223 This target hook returns a register class for which branch target register
10224 optimizations should be applied. All registers in this class should be
10225 usable interchangeably. After reload, registers in this class will be
10226 re-allocated and loads will be hoisted out of loops and be subjected
10227 to inter-block scheduling.
10230 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
10231 Branch target register optimization will by default exclude callee-saved
10233 that are not already live during the current function; if this target hook
10234 returns true, they will be included. The target code must than make sure
10235 that all target registers in the class returned by
10236 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
10237 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
10238 epilogues have already been generated. Note, even if you only return
10239 true when @var{after_prologue_epilogue_gen} is false, you still are likely
10240 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
10241 to reserve space for caller-saved target registers.
10244 @defmac POWI_MAX_MULTS
10245 If defined, this macro is interpreted as a signed integer C expression
10246 that specifies the maximum number of floating point multiplications
10247 that should be emitted when expanding exponentiation by an integer
10248 constant inline. When this value is defined, exponentiation requiring
10249 more than this number of multiplications is implemented by calling the
10250 system library's @code{pow}, @code{powf} or @code{powl} routines.
10251 The default value places no upper bound on the multiplication count.
10254 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
10255 This target hook should register any extra include files for the
10256 target. The parameter @var{stdinc} indicates if normal include files
10257 are present. The parameter @var{sysroot} is the system root directory.
10258 The parameter @var{iprefix} is the prefix for the gcc directory.
10261 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
10262 This target hook should register any extra include files for the
10263 target before any standard headers. The parameter @var{stdinc}
10264 indicates if normal include files are present. The parameter
10265 @var{sysroot} is the system root directory. The parameter
10266 @var{iprefix} is the prefix for the gcc directory.
10269 @deftypefn Macro void TARGET_OPTF (char *@var{path})
10270 This target hook should register special include paths for the target.
10271 The parameter @var{path} is the include to register. On Darwin
10272 systems, this is used for Framework includes, which have semantics
10273 that are different from @option{-I}.
10276 @deftypefn {Target Hook} bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
10277 This target hook returns @code{true} if it is safe to use a local alias
10278 for a virtual function @var{fndecl} when constructing thunks,
10279 @code{false} otherwise. By default, the hook returns @code{true} for all
10280 functions, if a target supports aliases (i.e.@: defines
10281 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
10284 @defmac TARGET_FORMAT_TYPES
10285 If defined, this macro is the name of a global variable containing
10286 target-specific format checking information for the @option{-Wformat}
10287 option. The default is to have no target-specific format checks.
10290 @defmac TARGET_N_FORMAT_TYPES
10291 If defined, this macro is the number of entries in
10292 @code{TARGET_FORMAT_TYPES}.
10295 @deftypefn {Target Hook} bool TARGET_RELAXED_ORDERING
10296 If set to @code{true}, means that the target's memory model does not
10297 guarantee that loads which do not depend on one another will access
10298 main memory in the order of the instruction stream; if ordering is
10299 important, an explicit memory barrier must be used. This is true of
10300 many recent processors which implement a policy of ``relaxed,''
10301 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
10302 and ia64. The default is @code{false}.
10305 @deftypefn {Target Hook} const char *TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (tree @var{typelist}, tree @var{funcdecl}, tree @var{val})
10306 If defined, this macro returns the diagnostic message when it is
10307 illegal to pass argument @var{val} to function @var{funcdecl}
10308 with prototype @var{typelist}.
10311 @deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (tree @var{fromtype}, tree @var{totype})
10312 If defined, this macro returns the diagnostic message when it is
10313 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
10314 if validity should be determined by the front end.
10317 @deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, tree @var{type})
10318 If defined, this macro returns the diagnostic message when it is
10319 invalid to apply operation @var{op} (where unary plus is denoted by
10320 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
10321 if validity should be determined by the front end.
10324 @deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, tree @var{type1}, tree @var{type2})
10325 If defined, this macro returns the diagnostic message when it is
10326 invalid to apply operation @var{op} to operands of types @var{type1}
10327 and @var{type2}, or @code{NULL} if validity should be determined by
10331 @defmac TARGET_USE_JCR_SECTION
10332 This macro determines whether to use the JCR section to register Java
10333 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
10334 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.
10338 This macro determines the size of the objective C jump buffer for the
10339 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
10342 @defmac LIBGCC2_UNWIND_ATTRIBUTE
10343 Define this macro if any target-specific attributes need to be attached
10344 to the functions in @file{libgcc} that provide low-level support for
10345 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
10346 and the associated definitions of those functions.