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 that determines how pointers should be extended from
1008 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
1009 greater than zero if pointers should be zero-extended, zero if they
1010 should be sign-extended, and negative if some other sort of conversion
1011 is needed. In the last case, the extension is done by the target's
1012 @code{ptr_extend} instruction.
1014 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
1015 and @code{word_mode} are all the same width.
1018 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
1019 A macro to update @var{m} and @var{unsignedp} when an object whose type
1020 is @var{type} and which has the specified mode and signedness is to be
1021 stored in a register. This macro is only called when @var{type} is a
1024 On most RISC machines, which only have operations that operate on a full
1025 register, define this macro to set @var{m} to @code{word_mode} if
1026 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
1027 cases, only integer modes should be widened because wider-precision
1028 floating-point operations are usually more expensive than their narrower
1031 For most machines, the macro definition does not change @var{unsignedp}.
1032 However, some machines, have instructions that preferentially handle
1033 either signed or unsigned quantities of certain modes. For example, on
1034 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
1035 sign-extend the result to 64 bits. On such machines, set
1036 @var{unsignedp} according to which kind of extension is more efficient.
1038 Do not define this macro if it would never modify @var{m}.
1041 @defmac PROMOTE_FUNCTION_MODE
1042 Like @code{PROMOTE_MODE}, but is applied to outgoing function arguments or
1043 function return values, as specified by @code{TARGET_PROMOTE_FUNCTION_ARGS}
1044 and @code{TARGET_PROMOTE_FUNCTION_RETURN}, respectively.
1046 The default is @code{PROMOTE_MODE}.
1049 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_ARGS (tree @var{fntype})
1050 This target hook should return @code{true} if the promotion described by
1051 @code{PROMOTE_FUNCTION_MODE} should be done for outgoing function
1055 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_RETURN (tree @var{fntype})
1056 This target hook should return @code{true} if the promotion described by
1057 @code{PROMOTE_FUNCTION_MODE} should be done for the return value of
1060 If this target hook returns @code{true}, @code{TARGET_FUNCTION_VALUE}
1061 must perform the same promotions done by @code{PROMOTE_FUNCTION_MODE}.
1064 @defmac PARM_BOUNDARY
1065 Normal alignment required for function parameters on the stack, in
1066 bits. All stack parameters receive at least this much alignment
1067 regardless of data type. On most machines, this is the same as the
1071 @defmac STACK_BOUNDARY
1072 Define this macro to the minimum alignment enforced by hardware for the
1073 stack pointer on this machine. The definition is a C expression for the
1074 desired alignment (measured in bits). This value is used as a default
1075 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1076 this should be the same as @code{PARM_BOUNDARY}.
1079 @defmac PREFERRED_STACK_BOUNDARY
1080 Define this macro if you wish to preserve a certain alignment for the
1081 stack pointer, greater than what the hardware enforces. The definition
1082 is a C expression for the desired alignment (measured in bits). This
1083 macro must evaluate to a value equal to or larger than
1084 @code{STACK_BOUNDARY}.
1087 @defmac FUNCTION_BOUNDARY
1088 Alignment required for a function entry point, in bits.
1091 @defmac BIGGEST_ALIGNMENT
1092 Biggest alignment that any data type can require on this machine, in
1093 bits. Note that this is not the biggest alignment that is supported,
1094 just the biggest alignment that, when violated, may cause a fault.
1097 @defmac MINIMUM_ATOMIC_ALIGNMENT
1098 If defined, the smallest alignment, in bits, that can be given to an
1099 object that can be referenced in one operation, without disturbing any
1100 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1101 on machines that don't have byte or half-word store operations.
1104 @defmac BIGGEST_FIELD_ALIGNMENT
1105 Biggest alignment that any structure or union field can require on this
1106 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1107 structure and union fields only, unless the field alignment has been set
1108 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1111 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1112 An expression for the alignment of a structure field @var{field} if the
1113 alignment computed in the usual way (including applying of
1114 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1115 alignment) is @var{computed}. It overrides alignment only if the
1116 field alignment has not been set by the
1117 @code{__attribute__ ((aligned (@var{n})))} construct.
1120 @defmac MAX_OFILE_ALIGNMENT
1121 Biggest alignment supported by the object file format of this machine.
1122 Use this macro to limit the alignment which can be specified using the
1123 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1124 the default value is @code{BIGGEST_ALIGNMENT}.
1126 On systems that use ELF, the default (in @file{config/elfos.h}) is
1127 the largest supported 32-bit ELF section alignment representable on
1128 a 32-bit host e.g. @samp{(((unsigned HOST_WIDEST_INT) 1 << 28) * 8)}.
1129 On 32-bit ELF the largest supported section alignment in bits is
1130 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1133 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1134 If defined, a C expression to compute the alignment for a variable in
1135 the static store. @var{type} is the data type, and @var{basic-align} is
1136 the alignment that the object would ordinarily have. The value of this
1137 macro is used instead of that alignment to align the object.
1139 If this macro is not defined, then @var{basic-align} is used.
1142 One use of this macro is to increase alignment of medium-size data to
1143 make it all fit in fewer cache lines. Another is to cause character
1144 arrays to be word-aligned so that @code{strcpy} calls that copy
1145 constants to character arrays can be done inline.
1148 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1149 If defined, a C expression to compute the alignment given to a constant
1150 that is being placed in memory. @var{constant} is the constant and
1151 @var{basic-align} is the alignment that the object would ordinarily
1152 have. The value of this macro is used instead of that alignment to
1155 If this macro is not defined, then @var{basic-align} is used.
1157 The typical use of this macro is to increase alignment for string
1158 constants to be word aligned so that @code{strcpy} calls that copy
1159 constants can be done inline.
1162 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1163 If defined, a C expression to compute the alignment for a variable in
1164 the local store. @var{type} is the data type, and @var{basic-align} is
1165 the alignment that the object would ordinarily have. The value of this
1166 macro is used instead of that alignment to align the object.
1168 If this macro is not defined, then @var{basic-align} is used.
1170 One use of this macro is to increase alignment of medium-size data to
1171 make it all fit in fewer cache lines.
1174 @defmac EMPTY_FIELD_BOUNDARY
1175 Alignment in bits to be given to a structure bit-field that follows an
1176 empty field such as @code{int : 0;}.
1178 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1181 @defmac STRUCTURE_SIZE_BOUNDARY
1182 Number of bits which any structure or union's size must be a multiple of.
1183 Each structure or union's size is rounded up to a multiple of this.
1185 If you do not define this macro, the default is the same as
1186 @code{BITS_PER_UNIT}.
1189 @defmac STRICT_ALIGNMENT
1190 Define this macro to be the value 1 if instructions will fail to work
1191 if given data not on the nominal alignment. If instructions will merely
1192 go slower in that case, define this macro as 0.
1195 @defmac PCC_BITFIELD_TYPE_MATTERS
1196 Define this if you wish to imitate the way many other C compilers handle
1197 alignment of bit-fields and the structures that contain them.
1199 The behavior is that the type written for a named bit-field (@code{int},
1200 @code{short}, or other integer type) imposes an alignment for the entire
1201 structure, as if the structure really did contain an ordinary field of
1202 that type. In addition, the bit-field is placed within the structure so
1203 that it would fit within such a field, not crossing a boundary for it.
1205 Thus, on most machines, a named bit-field whose type is written as
1206 @code{int} would not cross a four-byte boundary, and would force
1207 four-byte alignment for the whole structure. (The alignment used may
1208 not be four bytes; it is controlled by the other alignment parameters.)
1210 An unnamed bit-field will not affect the alignment of the containing
1213 If the macro is defined, its definition should be a C expression;
1214 a nonzero value for the expression enables this behavior.
1216 Note that if this macro is not defined, or its value is zero, some
1217 bit-fields may cross more than one alignment boundary. The compiler can
1218 support such references if there are @samp{insv}, @samp{extv}, and
1219 @samp{extzv} insns that can directly reference memory.
1221 The other known way of making bit-fields work is to define
1222 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1223 Then every structure can be accessed with fullwords.
1225 Unless the machine has bit-field instructions or you define
1226 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1227 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1229 If your aim is to make GCC use the same conventions for laying out
1230 bit-fields as are used by another compiler, here is how to investigate
1231 what the other compiler does. Compile and run this program:
1250 printf ("Size of foo1 is %d\n",
1251 sizeof (struct foo1));
1252 printf ("Size of foo2 is %d\n",
1253 sizeof (struct foo2));
1258 If this prints 2 and 5, then the compiler's behavior is what you would
1259 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1262 @defmac BITFIELD_NBYTES_LIMITED
1263 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1264 to aligning a bit-field within the structure.
1267 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELDS (void)
1268 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1269 whether unnamed bitfields affect the alignment of the containing
1270 structure. The hook should return true if the structure should inherit
1271 the alignment requirements of an unnamed bitfield's type.
1274 @deftypefn {Target Hook} bool TARGET_NARROW_VOLATILE_BITFIELDS (void)
1275 This target hook should return @code{true} if accesses to volatile bitfields
1276 should use the narrowest mode possible. It should return @code{false} if
1277 these accesses should use the bitfield container type.
1279 The default is @code{!TARGET_STRICT_ALIGN}.
1282 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1283 Return 1 if a structure or array containing @var{field} should be accessed using
1286 If @var{field} is the only field in the structure, @var{mode} is its
1287 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1288 case where structures of one field would require the structure's mode to
1289 retain the field's mode.
1291 Normally, this is not needed. See the file @file{c4x.h} for an example
1292 of how to use this macro to prevent a structure having a floating point
1293 field from being accessed in an integer mode.
1296 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1297 Define this macro as an expression for the alignment of a type (given
1298 by @var{type} as a tree node) if the alignment computed in the usual
1299 way is @var{computed} and the alignment explicitly specified was
1302 The default is to use @var{specified} if it is larger; otherwise, use
1303 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1306 @defmac MAX_FIXED_MODE_SIZE
1307 An integer expression for the size in bits of the largest integer
1308 machine mode that should actually be used. All integer machine modes of
1309 this size or smaller can be used for structures and unions with the
1310 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1311 (DImode)} is assumed.
1314 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1315 If defined, an expression of type @code{enum machine_mode} that
1316 specifies the mode of the save area operand of a
1317 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1318 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1319 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1320 having its mode specified.
1322 You need not define this macro if it always returns @code{Pmode}. You
1323 would most commonly define this macro if the
1324 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1328 @defmac STACK_SIZE_MODE
1329 If defined, an expression of type @code{enum machine_mode} that
1330 specifies the mode of the size increment operand of an
1331 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1333 You need not define this macro if it always returns @code{word_mode}.
1334 You would most commonly define this macro if the @code{allocate_stack}
1335 pattern needs to support both a 32- and a 64-bit mode.
1338 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_CMP_RETURN_MODE ()
1339 This target hook should return the mode to be used for the return value
1340 of compare instructions expanded to libgcc calls. If not defined
1341 @code{word_mode} is returned which is the right choice for a majority of
1345 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_SHIFT_COUNT_MODE ()
1346 This target hook should return the mode to be used for the shift count operand
1347 of shift instructions expanded to libgcc calls. If not defined
1348 @code{word_mode} is returned which is the right choice for a majority of
1352 @defmac TARGET_FLOAT_FORMAT
1353 A code distinguishing the floating point format of the target machine.
1354 There are four defined values:
1357 @item IEEE_FLOAT_FORMAT
1358 This code indicates IEEE floating point. It is the default; there is no
1359 need to define @code{TARGET_FLOAT_FORMAT} when the format is IEEE@.
1361 @item VAX_FLOAT_FORMAT
1362 This code indicates the ``F float'' (for @code{float}) and ``D float''
1363 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
1365 @item C4X_FLOAT_FORMAT
1366 This code indicates the format used on the TMS320C3x/C4x.
1369 If your target uses a floating point format other than these, you must
1370 define a new @var{name}_FLOAT_FORMAT code for it, and add support for
1371 it to @file{real.c}.
1373 The ordering of the component words of floating point values stored in
1374 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1377 @defmac MODE_HAS_NANS (@var{mode})
1378 When defined, this macro should be true if @var{mode} has a NaN
1379 representation. The compiler assumes that NaNs are not equal to
1380 anything (including themselves) and that addition, subtraction,
1381 multiplication and division all return NaNs when one operand is
1384 By default, this macro is true if @var{mode} is a floating-point
1385 mode and the target floating-point format is IEEE@.
1388 @defmac MODE_HAS_INFINITIES (@var{mode})
1389 This macro should be true if @var{mode} can represent infinity. At
1390 present, the compiler uses this macro to decide whether @samp{x - x}
1391 is always defined. By default, the macro is true when @var{mode}
1392 is a floating-point mode and the target format is IEEE@.
1395 @defmac MODE_HAS_SIGNED_ZEROS (@var{mode})
1396 True if @var{mode} distinguishes between positive and negative zero.
1397 The rules are expected to follow the IEEE standard:
1401 @samp{x + x} has the same sign as @samp{x}.
1404 If the sum of two values with opposite sign is zero, the result is
1405 positive for all rounding modes expect towards @minus{}infinity, for
1406 which it is negative.
1409 The sign of a product or quotient is negative when exactly one
1410 of the operands is negative.
1413 The default definition is true if @var{mode} is a floating-point
1414 mode and the target format is IEEE@.
1417 @defmac MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1418 If defined, this macro should be true for @var{mode} if it has at
1419 least one rounding mode in which @samp{x} and @samp{-x} can be
1420 rounded to numbers of different magnitude. Two such modes are
1421 towards @minus{}infinity and towards +infinity.
1423 The default definition of this macro is true if @var{mode} is
1424 a floating-point mode and the target format is IEEE@.
1427 @defmac ROUND_TOWARDS_ZERO
1428 If defined, this macro should be true if the prevailing rounding
1429 mode is towards zero. A true value has the following effects:
1433 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1436 @file{libgcc.a}'s floating-point emulator will round towards zero
1437 rather than towards nearest.
1440 The compiler's floating-point emulator will round towards zero after
1441 doing arithmetic, and when converting from the internal float format to
1445 The macro does not affect the parsing of string literals. When the
1446 primary rounding mode is towards zero, library functions like
1447 @code{strtod} might still round towards nearest, and the compiler's
1448 parser should behave like the target's @code{strtod} where possible.
1450 Not defining this macro is equivalent to returning zero.
1453 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1454 This macro should return true if floats with @var{size}
1455 bits do not have a NaN or infinity representation, but use the largest
1456 exponent for normal numbers instead.
1458 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1459 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1460 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1461 floating-point arithmetic.
1463 The default definition of this macro returns false for all sizes.
1466 @deftypefn {Target Hook} bool TARGET_VECTOR_OPAQUE_P (tree @var{type})
1467 This target hook should return @code{true} a vector is opaque. That
1468 is, if no cast is needed when copying a vector value of type
1469 @var{type} into another vector lvalue of the same size. Vector opaque
1470 types cannot be initialized. The default is that there are no such
1474 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1475 This target hook returns @code{true} if bit-fields in the given
1476 @var{record_type} are to be laid out following the rules of Microsoft
1477 Visual C/C++, namely: (i) a bit-field won't share the same storage
1478 unit with the previous bit-field if their underlying types have
1479 different sizes, and the bit-field will be aligned to the highest
1480 alignment of the underlying types of itself and of the previous
1481 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1482 the whole enclosing structure, even if it is unnamed; except that
1483 (iii) a zero-sized bit-field will be disregarded unless it follows
1484 another bit-field of nonzero size. If this hook returns @code{true},
1485 other macros that control bit-field layout are ignored.
1487 When a bit-field is inserted into a packed record, the whole size
1488 of the underlying type is used by one or more same-size adjacent
1489 bit-fields (that is, if its long:3, 32 bits is used in the record,
1490 and any additional adjacent long bit-fields are packed into the same
1491 chunk of 32 bits. However, if the size changes, a new field of that
1492 size is allocated). In an unpacked record, this is the same as using
1493 alignment, but not equivalent when packing.
1495 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1496 the latter will take precedence. If @samp{__attribute__((packed))} is
1497 used on a single field when MS bit-fields are in use, it will take
1498 precedence for that field, but the alignment of the rest of the structure
1499 may affect its placement.
1502 @deftypefn {Target Hook} {bool} TARGET_DECIMAL_FLOAT_SUPPORTED_P (void)
1503 Returns true if the target supports decimal floating point.
1506 @deftypefn {Target Hook} {bool} TARGET_FIXED_POINT_SUPPORTED_P (void)
1507 Returns true if the target supports fixed-point arithmetic.
1510 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_TYPE (tree @var{type})
1511 If your target defines any fundamental types, or any types your target
1512 uses should be mangled differently from the default, define this hook
1513 to return the appropriate encoding for these types as part of a C++
1514 mangled name. The @var{type} argument is the tree structure representing
1515 the type to be mangled. The hook may be applied to trees which are
1516 not target-specific fundamental types; it should return @code{NULL}
1517 for all such types, as well as arguments it does not recognize. If the
1518 return value is not @code{NULL}, it must point to a statically-allocated
1521 Target-specific fundamental types might be new fundamental types or
1522 qualified versions of ordinary fundamental types. Encode new
1523 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1524 is the name used for the type in source code, and @var{n} is the
1525 length of @var{name} in decimal. Encode qualified versions of
1526 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1527 @var{name} is the name used for the type qualifier in source code,
1528 @var{n} is the length of @var{name} as above, and @var{code} is the
1529 code used to represent the unqualified version of this type. (See
1530 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1531 codes.) In both cases the spaces are for clarity; do not include any
1532 spaces in your string.
1534 This hook is applied to types prior to typedef resolution. If the mangled
1535 name for a particular type depends only on that type's main variant, you
1536 can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT}
1539 The default version of this hook always returns @code{NULL}, which is
1540 appropriate for a target that does not define any new fundamental
1545 @section Layout of Source Language Data Types
1547 These macros define the sizes and other characteristics of the standard
1548 basic data types used in programs being compiled. Unlike the macros in
1549 the previous section, these apply to specific features of C and related
1550 languages, rather than to fundamental aspects of storage layout.
1552 @defmac INT_TYPE_SIZE
1553 A C expression for the size in bits of the type @code{int} on the
1554 target machine. If you don't define this, the default is one word.
1557 @defmac SHORT_TYPE_SIZE
1558 A C expression for the size in bits of the type @code{short} on the
1559 target machine. If you don't define this, the default is half a word.
1560 (If this would be less than one storage unit, it is rounded up to one
1564 @defmac LONG_TYPE_SIZE
1565 A C expression for the size in bits of the type @code{long} on the
1566 target machine. If you don't define this, the default is one word.
1569 @defmac ADA_LONG_TYPE_SIZE
1570 On some machines, the size used for the Ada equivalent of the type
1571 @code{long} by a native Ada compiler differs from that used by C@. In
1572 that situation, define this macro to be a C expression to be used for
1573 the size of that type. If you don't define this, the default is the
1574 value of @code{LONG_TYPE_SIZE}.
1577 @defmac LONG_LONG_TYPE_SIZE
1578 A C expression for the size in bits of the type @code{long long} on the
1579 target machine. If you don't define this, the default is two
1580 words. If you want to support GNU Ada on your machine, the value of this
1581 macro must be at least 64.
1584 @defmac CHAR_TYPE_SIZE
1585 A C expression for the size in bits of the type @code{char} on the
1586 target machine. If you don't define this, the default is
1587 @code{BITS_PER_UNIT}.
1590 @defmac BOOL_TYPE_SIZE
1591 A C expression for the size in bits of the C++ type @code{bool} and
1592 C99 type @code{_Bool} on the target machine. If you don't define
1593 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1596 @defmac FLOAT_TYPE_SIZE
1597 A C expression for the size in bits of the type @code{float} on the
1598 target machine. If you don't define this, the default is one word.
1601 @defmac DOUBLE_TYPE_SIZE
1602 A C expression for the size in bits of the type @code{double} on the
1603 target machine. If you don't define this, the default is two
1607 @defmac LONG_DOUBLE_TYPE_SIZE
1608 A C expression for the size in bits of the type @code{long double} on
1609 the target machine. If you don't define this, the default is two
1613 @defmac SHORT_FRACT_TYPE_SIZE
1614 A C expression for the size in bits of the type @code{short _Fract} on
1615 the target machine. If you don't define this, the default is
1616 @code{BITS_PER_UNIT}.
1619 @defmac FRACT_TYPE_SIZE
1620 A C expression for the size in bits of the type @code{_Fract} on
1621 the target machine. If you don't define this, the default is
1622 @code{BITS_PER_UNIT * 2}.
1625 @defmac LONG_FRACT_TYPE_SIZE
1626 A C expression for the size in bits of the type @code{long _Fract} on
1627 the target machine. If you don't define this, the default is
1628 @code{BITS_PER_UNIT * 4}.
1631 @defmac LONG_LONG_FRACT_TYPE_SIZE
1632 A C expression for the size in bits of the type @code{long long _Fract} on
1633 the target machine. If you don't define this, the default is
1634 @code{BITS_PER_UNIT * 8}.
1637 @defmac SHORT_ACCUM_TYPE_SIZE
1638 A C expression for the size in bits of the type @code{short _Accum} on
1639 the target machine. If you don't define this, the default is
1640 @code{BITS_PER_UNIT * 2}.
1643 @defmac ACCUM_TYPE_SIZE
1644 A C expression for the size in bits of the type @code{_Accum} on
1645 the target machine. If you don't define this, the default is
1646 @code{BITS_PER_UNIT * 4}.
1649 @defmac LONG_ACCUM_TYPE_SIZE
1650 A C expression for the size in bits of the type @code{long _Accum} on
1651 the target machine. If you don't define this, the default is
1652 @code{BITS_PER_UNIT * 8}.
1655 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1656 A C expression for the size in bits of the type @code{long long _Accum} on
1657 the target machine. If you don't define this, the default is
1658 @code{BITS_PER_UNIT * 16}.
1661 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1662 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1663 if you want routines in @file{libgcc2.a} for a size other than
1664 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1665 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1668 @defmac LIBGCC2_HAS_DF_MODE
1669 Define this macro if neither @code{LIBGCC2_DOUBLE_TYPE_SIZE} nor
1670 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1671 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1672 anyway. If you don't define this and either @code{LIBGCC2_DOUBLE_TYPE_SIZE}
1673 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1677 @defmac LIBGCC2_HAS_XF_MODE
1678 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1679 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1680 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1681 is 80 then the default is 1, otherwise it is 0.
1684 @defmac LIBGCC2_HAS_TF_MODE
1685 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1686 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1687 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1688 is 128 then the default is 1, otherwise it is 0.
1695 Define these macros to be the size in bits of the mantissa of
1696 @code{SFmode}, @code{DFmode}, @code{XFmode} and @code{TFmode} values,
1697 if the defaults in @file{libgcc2.h} are inappropriate. By default,
1698 @code{FLT_MANT_DIG} is used for @code{SF_SIZE}, @code{LDBL_MANT_DIG}
1699 for @code{XF_SIZE} and @code{TF_SIZE}, and @code{DBL_MANT_DIG} or
1700 @code{LDBL_MANT_DIG} for @code{DF_SIZE} according to whether
1701 @code{LIBGCC2_DOUBLE_TYPE_SIZE} or
1702 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64.
1705 @defmac TARGET_FLT_EVAL_METHOD
1706 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1707 assuming, if applicable, that the floating-point control word is in its
1708 default state. If you do not define this macro the value of
1709 @code{FLT_EVAL_METHOD} will be zero.
1712 @defmac WIDEST_HARDWARE_FP_SIZE
1713 A C expression for the size in bits of the widest floating-point format
1714 supported by the hardware. If you define this macro, you must specify a
1715 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1716 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1720 @defmac DEFAULT_SIGNED_CHAR
1721 An expression whose value is 1 or 0, according to whether the type
1722 @code{char} should be signed or unsigned by default. The user can
1723 always override this default with the options @option{-fsigned-char}
1724 and @option{-funsigned-char}.
1727 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1728 This target hook should return true if the compiler should give an
1729 @code{enum} type only as many bytes as it takes to represent the range
1730 of possible values of that type. It should return false if all
1731 @code{enum} types should be allocated like @code{int}.
1733 The default is to return false.
1737 A C expression for a string describing the name of the data type to use
1738 for size values. The typedef name @code{size_t} is defined using the
1739 contents of the string.
1741 The string can contain more than one keyword. If so, separate them with
1742 spaces, and write first any length keyword, then @code{unsigned} if
1743 appropriate, and finally @code{int}. The string must exactly match one
1744 of the data type names defined in the function
1745 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1746 omit @code{int} or change the order---that would cause the compiler to
1749 If you don't define this macro, the default is @code{"long unsigned
1753 @defmac PTRDIFF_TYPE
1754 A C expression for a string describing the name of the data type to use
1755 for the result of subtracting two pointers. The typedef name
1756 @code{ptrdiff_t} is defined using the contents of the string. See
1757 @code{SIZE_TYPE} above for more information.
1759 If you don't define this macro, the default is @code{"long int"}.
1763 A C expression for a string describing the name of the data type to use
1764 for wide characters. The typedef name @code{wchar_t} is defined using
1765 the contents of the string. See @code{SIZE_TYPE} above for more
1768 If you don't define this macro, the default is @code{"int"}.
1771 @defmac WCHAR_TYPE_SIZE
1772 A C expression for the size in bits of the data type for wide
1773 characters. This is used in @code{cpp}, which cannot make use of
1778 A C expression for a string describing the name of the data type to
1779 use for wide characters passed to @code{printf} and returned from
1780 @code{getwc}. The typedef name @code{wint_t} is defined using the
1781 contents of the string. See @code{SIZE_TYPE} above for more
1784 If you don't define this macro, the default is @code{"unsigned int"}.
1788 A C expression for a string describing the name of the data type that
1789 can represent any value of any standard or extended signed integer type.
1790 The typedef name @code{intmax_t} is defined using the contents of the
1791 string. See @code{SIZE_TYPE} above for more information.
1793 If you don't define this macro, the default is the first of
1794 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1795 much precision as @code{long long int}.
1798 @defmac UINTMAX_TYPE
1799 A C expression for a string describing the name of the data type that
1800 can represent any value of any standard or extended unsigned integer
1801 type. The typedef name @code{uintmax_t} is defined using the contents
1802 of the string. See @code{SIZE_TYPE} above for more information.
1804 If you don't define this macro, the default is the first of
1805 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1806 unsigned int"} that has as much precision as @code{long long unsigned
1810 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1811 The C++ compiler represents a pointer-to-member-function with a struct
1818 ptrdiff_t vtable_index;
1825 The C++ compiler must use one bit to indicate whether the function that
1826 will be called through a pointer-to-member-function is virtual.
1827 Normally, we assume that the low-order bit of a function pointer must
1828 always be zero. Then, by ensuring that the vtable_index is odd, we can
1829 distinguish which variant of the union is in use. But, on some
1830 platforms function pointers can be odd, and so this doesn't work. In
1831 that case, we use the low-order bit of the @code{delta} field, and shift
1832 the remainder of the @code{delta} field to the left.
1834 GCC will automatically make the right selection about where to store
1835 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1836 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1837 set such that functions always start at even addresses, but the lowest
1838 bit of pointers to functions indicate whether the function at that
1839 address is in ARM or Thumb mode. If this is the case of your
1840 architecture, you should define this macro to
1841 @code{ptrmemfunc_vbit_in_delta}.
1843 In general, you should not have to define this macro. On architectures
1844 in which function addresses are always even, according to
1845 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1846 @code{ptrmemfunc_vbit_in_pfn}.
1849 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1850 Normally, the C++ compiler uses function pointers in vtables. This
1851 macro allows the target to change to use ``function descriptors''
1852 instead. Function descriptors are found on targets for whom a
1853 function pointer is actually a small data structure. Normally the
1854 data structure consists of the actual code address plus a data
1855 pointer to which the function's data is relative.
1857 If vtables are used, the value of this macro should be the number
1858 of words that the function descriptor occupies.
1861 @defmac TARGET_VTABLE_ENTRY_ALIGN
1862 By default, the vtable entries are void pointers, the so the alignment
1863 is the same as pointer alignment. The value of this macro specifies
1864 the alignment of the vtable entry in bits. It should be defined only
1865 when special alignment is necessary. */
1868 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1869 There are a few non-descriptor entries in the vtable at offsets below
1870 zero. If these entries must be padded (say, to preserve the alignment
1871 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1872 of words in each data entry.
1876 @section Register Usage
1877 @cindex register usage
1879 This section explains how to describe what registers the target machine
1880 has, and how (in general) they can be used.
1882 The description of which registers a specific instruction can use is
1883 done with register classes; see @ref{Register Classes}. For information
1884 on using registers to access a stack frame, see @ref{Frame Registers}.
1885 For passing values in registers, see @ref{Register Arguments}.
1886 For returning values in registers, see @ref{Scalar Return}.
1889 * Register Basics:: Number and kinds of registers.
1890 * Allocation Order:: Order in which registers are allocated.
1891 * Values in Registers:: What kinds of values each reg can hold.
1892 * Leaf Functions:: Renumbering registers for leaf functions.
1893 * Stack Registers:: Handling a register stack such as 80387.
1896 @node Register Basics
1897 @subsection Basic Characteristics of Registers
1899 @c prevent bad page break with this line
1900 Registers have various characteristics.
1902 @defmac FIRST_PSEUDO_REGISTER
1903 Number of hardware registers known to the compiler. They receive
1904 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1905 pseudo register's number really is assigned the number
1906 @code{FIRST_PSEUDO_REGISTER}.
1909 @defmac FIXED_REGISTERS
1910 @cindex fixed register
1911 An initializer that says which registers are used for fixed purposes
1912 all throughout the compiled code and are therefore not available for
1913 general allocation. These would include the stack pointer, the frame
1914 pointer (except on machines where that can be used as a general
1915 register when no frame pointer is needed), the program counter on
1916 machines where that is considered one of the addressable registers,
1917 and any other numbered register with a standard use.
1919 This information is expressed as a sequence of numbers, separated by
1920 commas and surrounded by braces. The @var{n}th number is 1 if
1921 register @var{n} is fixed, 0 otherwise.
1923 The table initialized from this macro, and the table initialized by
1924 the following one, may be overridden at run time either automatically,
1925 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1926 the user with the command options @option{-ffixed-@var{reg}},
1927 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1930 @defmac CALL_USED_REGISTERS
1931 @cindex call-used register
1932 @cindex call-clobbered register
1933 @cindex call-saved register
1934 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1935 clobbered (in general) by function calls as well as for fixed
1936 registers. This macro therefore identifies the registers that are not
1937 available for general allocation of values that must live across
1940 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1941 automatically saves it on function entry and restores it on function
1942 exit, if the register is used within the function.
1945 @defmac CALL_REALLY_USED_REGISTERS
1946 @cindex call-used register
1947 @cindex call-clobbered register
1948 @cindex call-saved register
1949 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1950 that the entire set of @code{FIXED_REGISTERS} be included.
1951 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1952 This macro is optional. If not specified, it defaults to the value
1953 of @code{CALL_USED_REGISTERS}.
1956 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1957 @cindex call-used register
1958 @cindex call-clobbered register
1959 @cindex call-saved register
1960 A C expression that is nonzero if it is not permissible to store a
1961 value of mode @var{mode} in hard register number @var{regno} across a
1962 call without some part of it being clobbered. For most machines this
1963 macro need not be defined. It is only required for machines that do not
1964 preserve the entire contents of a register across a call.
1968 @findex call_used_regs
1971 @findex reg_class_contents
1972 @defmac CONDITIONAL_REGISTER_USAGE
1973 Zero or more C statements that may conditionally modify five variables
1974 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1975 @code{reg_names}, and @code{reg_class_contents}, to take into account
1976 any dependence of these register sets on target flags. The first three
1977 of these are of type @code{char []} (interpreted as Boolean vectors).
1978 @code{global_regs} is a @code{const char *[]}, and
1979 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1980 called, @code{fixed_regs}, @code{call_used_regs},
1981 @code{reg_class_contents}, and @code{reg_names} have been initialized
1982 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1983 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1984 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1985 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1986 command options have been applied.
1988 You need not define this macro if it has no work to do.
1990 @cindex disabling certain registers
1991 @cindex controlling register usage
1992 If the usage of an entire class of registers depends on the target
1993 flags, you may indicate this to GCC by using this macro to modify
1994 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1995 registers in the classes which should not be used by GCC@. Also define
1996 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1997 to return @code{NO_REGS} if it
1998 is called with a letter for a class that shouldn't be used.
2000 (However, if this class is not included in @code{GENERAL_REGS} and all
2001 of the insn patterns whose constraints permit this class are
2002 controlled by target switches, then GCC will automatically avoid using
2003 these registers when the target switches are opposed to them.)
2006 @defmac INCOMING_REGNO (@var{out})
2007 Define this macro if the target machine has register windows. This C
2008 expression returns the register number as seen by the called function
2009 corresponding to the register number @var{out} as seen by the calling
2010 function. Return @var{out} if register number @var{out} is not an
2014 @defmac OUTGOING_REGNO (@var{in})
2015 Define this macro if the target machine has register windows. This C
2016 expression returns the register number as seen by the calling function
2017 corresponding to the register number @var{in} as seen by the called
2018 function. Return @var{in} if register number @var{in} is not an inbound
2022 @defmac LOCAL_REGNO (@var{regno})
2023 Define this macro if the target machine has register windows. This C
2024 expression returns true if the register is call-saved but is in the
2025 register window. Unlike most call-saved registers, such registers
2026 need not be explicitly restored on function exit or during non-local
2031 If the program counter has a register number, define this as that
2032 register number. Otherwise, do not define it.
2035 @node Allocation Order
2036 @subsection Order of Allocation of Registers
2037 @cindex order of register allocation
2038 @cindex register allocation order
2040 @c prevent bad page break with this line
2041 Registers are allocated in order.
2043 @defmac REG_ALLOC_ORDER
2044 If defined, an initializer for a vector of integers, containing the
2045 numbers of hard registers in the order in which GCC should prefer
2046 to use them (from most preferred to least).
2048 If this macro is not defined, registers are used lowest numbered first
2049 (all else being equal).
2051 One use of this macro is on machines where the highest numbered
2052 registers must always be saved and the save-multiple-registers
2053 instruction supports only sequences of consecutive registers. On such
2054 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
2055 the highest numbered allocable register first.
2058 @defmac ORDER_REGS_FOR_LOCAL_ALLOC
2059 A C statement (sans semicolon) to choose the order in which to allocate
2060 hard registers for pseudo-registers local to a basic block.
2062 Store the desired register order in the array @code{reg_alloc_order}.
2063 Element 0 should be the register to allocate first; element 1, the next
2064 register; and so on.
2066 The macro body should not assume anything about the contents of
2067 @code{reg_alloc_order} before execution of the macro.
2069 On most machines, it is not necessary to define this macro.
2072 @node Values in Registers
2073 @subsection How Values Fit in Registers
2075 This section discusses the macros that describe which kinds of values
2076 (specifically, which machine modes) each register can hold, and how many
2077 consecutive registers are needed for a given mode.
2079 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2080 A C expression for the number of consecutive hard registers, starting
2081 at register number @var{regno}, required to hold a value of mode
2084 On a machine where all registers are exactly one word, a suitable
2085 definition of this macro is
2088 #define HARD_REGNO_NREGS(REGNO, MODE) \
2089 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2094 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2095 A C expression that is nonzero if a value of mode @var{mode}, stored
2096 in memory, ends with padding that causes it to take up more space than
2097 in registers starting at register number @var{regno} (as determined by
2098 multiplying GCC's notion of the size of the register when containing
2099 this mode by the number of registers returned by
2100 @code{HARD_REGNO_NREGS}). By default this is zero.
2102 For example, if a floating-point value is stored in three 32-bit
2103 registers but takes up 128 bits in memory, then this would be
2106 This macros only needs to be defined if there are cases where
2107 @code{subreg_get_info}
2108 would otherwise wrongly determine that a @code{subreg} can be
2109 represented by an offset to the register number, when in fact such a
2110 @code{subreg} would contain some of the padding not stored in
2111 registers and so not be representable.
2114 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2115 For values of @var{regno} and @var{mode} for which
2116 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2117 returning the greater number of registers required to hold the value
2118 including any padding. In the example above, the value would be four.
2121 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2122 Define this macro if the natural size of registers that hold values
2123 of mode @var{mode} is not the word size. It is a C expression that
2124 should give the natural size in bytes for the specified mode. It is
2125 used by the register allocator to try to optimize its results. This
2126 happens for example on SPARC 64-bit where the natural size of
2127 floating-point registers is still 32-bit.
2130 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2131 A C expression that is nonzero if it is permissible to store a value
2132 of mode @var{mode} in hard register number @var{regno} (or in several
2133 registers starting with that one). For a machine where all registers
2134 are equivalent, a suitable definition is
2137 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2140 You need not include code to check for the numbers of fixed registers,
2141 because the allocation mechanism considers them to be always occupied.
2143 @cindex register pairs
2144 On some machines, double-precision values must be kept in even/odd
2145 register pairs. You can implement that by defining this macro to reject
2146 odd register numbers for such modes.
2148 The minimum requirement for a mode to be OK in a register is that the
2149 @samp{mov@var{mode}} instruction pattern support moves between the
2150 register and other hard register in the same class and that moving a
2151 value into the register and back out not alter it.
2153 Since the same instruction used to move @code{word_mode} will work for
2154 all narrower integer modes, it is not necessary on any machine for
2155 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2156 you define patterns @samp{movhi}, etc., to take advantage of this. This
2157 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2158 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2161 Many machines have special registers for floating point arithmetic.
2162 Often people assume that floating point machine modes are allowed only
2163 in floating point registers. This is not true. Any registers that
2164 can hold integers can safely @emph{hold} a floating point machine
2165 mode, whether or not floating arithmetic can be done on it in those
2166 registers. Integer move instructions can be used to move the values.
2168 On some machines, though, the converse is true: fixed-point machine
2169 modes may not go in floating registers. This is true if the floating
2170 registers normalize any value stored in them, because storing a
2171 non-floating value there would garble it. In this case,
2172 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2173 floating registers. But if the floating registers do not automatically
2174 normalize, if you can store any bit pattern in one and retrieve it
2175 unchanged without a trap, then any machine mode may go in a floating
2176 register, so you can define this macro to say so.
2178 The primary significance of special floating registers is rather that
2179 they are the registers acceptable in floating point arithmetic
2180 instructions. However, this is of no concern to
2181 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2182 constraints for those instructions.
2184 On some machines, the floating registers are especially slow to access,
2185 so that it is better to store a value in a stack frame than in such a
2186 register if floating point arithmetic is not being done. As long as the
2187 floating registers are not in class @code{GENERAL_REGS}, they will not
2188 be used unless some pattern's constraint asks for one.
2191 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2192 A C expression that is nonzero if it is OK to rename a hard register
2193 @var{from} to another hard register @var{to}.
2195 One common use of this macro is to prevent renaming of a register to
2196 another register that is not saved by a prologue in an interrupt
2199 The default is always nonzero.
2202 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2203 A C expression that is nonzero if a value of mode
2204 @var{mode1} is accessible in mode @var{mode2} without copying.
2206 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2207 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2208 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2209 should be nonzero. If they differ for any @var{r}, you should define
2210 this macro to return zero unless some other mechanism ensures the
2211 accessibility of the value in a narrower mode.
2213 You should define this macro to return nonzero in as many cases as
2214 possible since doing so will allow GCC to perform better register
2218 @defmac AVOID_CCMODE_COPIES
2219 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2220 registers. You should only define this macro if support for copying to/from
2221 @code{CCmode} is incomplete.
2224 @node Leaf Functions
2225 @subsection Handling Leaf Functions
2227 @cindex leaf functions
2228 @cindex functions, leaf
2229 On some machines, a leaf function (i.e., one which makes no calls) can run
2230 more efficiently if it does not make its own register window. Often this
2231 means it is required to receive its arguments in the registers where they
2232 are passed by the caller, instead of the registers where they would
2235 The special treatment for leaf functions generally applies only when
2236 other conditions are met; for example, often they may use only those
2237 registers for its own variables and temporaries. We use the term ``leaf
2238 function'' to mean a function that is suitable for this special
2239 handling, so that functions with no calls are not necessarily ``leaf
2242 GCC assigns register numbers before it knows whether the function is
2243 suitable for leaf function treatment. So it needs to renumber the
2244 registers in order to output a leaf function. The following macros
2247 @defmac LEAF_REGISTERS
2248 Name of a char vector, indexed by hard register number, which
2249 contains 1 for a register that is allowable in a candidate for leaf
2252 If leaf function treatment involves renumbering the registers, then the
2253 registers marked here should be the ones before renumbering---those that
2254 GCC would ordinarily allocate. The registers which will actually be
2255 used in the assembler code, after renumbering, should not be marked with 1
2258 Define this macro only if the target machine offers a way to optimize
2259 the treatment of leaf functions.
2262 @defmac LEAF_REG_REMAP (@var{regno})
2263 A C expression whose value is the register number to which @var{regno}
2264 should be renumbered, when a function is treated as a leaf function.
2266 If @var{regno} is a register number which should not appear in a leaf
2267 function before renumbering, then the expression should yield @minus{}1, which
2268 will cause the compiler to abort.
2270 Define this macro only if the target machine offers a way to optimize the
2271 treatment of leaf functions, and registers need to be renumbered to do
2275 @findex current_function_is_leaf
2276 @findex current_function_uses_only_leaf_regs
2277 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2278 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2279 specially. They can test the C variable @code{current_function_is_leaf}
2280 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2281 set prior to local register allocation and is valid for the remaining
2282 compiler passes. They can also test the C variable
2283 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2284 functions which only use leaf registers.
2285 @code{current_function_uses_only_leaf_regs} is valid after all passes
2286 that modify the instructions have been run and is only useful if
2287 @code{LEAF_REGISTERS} is defined.
2288 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2289 @c of the next paragraph?! --mew 2feb93
2291 @node Stack Registers
2292 @subsection Registers That Form a Stack
2294 There are special features to handle computers where some of the
2295 ``registers'' form a stack. Stack registers are normally written by
2296 pushing onto the stack, and are numbered relative to the top of the
2299 Currently, GCC can only handle one group of stack-like registers, and
2300 they must be consecutively numbered. Furthermore, the existing
2301 support for stack-like registers is specific to the 80387 floating
2302 point coprocessor. If you have a new architecture that uses
2303 stack-like registers, you will need to do substantial work on
2304 @file{reg-stack.c} and write your machine description to cooperate
2305 with it, as well as defining these macros.
2308 Define this if the machine has any stack-like registers.
2311 @defmac FIRST_STACK_REG
2312 The number of the first stack-like register. This one is the top
2316 @defmac LAST_STACK_REG
2317 The number of the last stack-like register. This one is the bottom of
2321 @node Register Classes
2322 @section Register Classes
2323 @cindex register class definitions
2324 @cindex class definitions, register
2326 On many machines, the numbered registers are not all equivalent.
2327 For example, certain registers may not be allowed for indexed addressing;
2328 certain registers may not be allowed in some instructions. These machine
2329 restrictions are described to the compiler using @dfn{register classes}.
2331 You define a number of register classes, giving each one a name and saying
2332 which of the registers belong to it. Then you can specify register classes
2333 that are allowed as operands to particular instruction patterns.
2337 In general, each register will belong to several classes. In fact, one
2338 class must be named @code{ALL_REGS} and contain all the registers. Another
2339 class must be named @code{NO_REGS} and contain no registers. Often the
2340 union of two classes will be another class; however, this is not required.
2342 @findex GENERAL_REGS
2343 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2344 terribly special about the name, but the operand constraint letters
2345 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2346 the same as @code{ALL_REGS}, just define it as a macro which expands
2349 Order the classes so that if class @var{x} is contained in class @var{y}
2350 then @var{x} has a lower class number than @var{y}.
2352 The way classes other than @code{GENERAL_REGS} are specified in operand
2353 constraints is through machine-dependent operand constraint letters.
2354 You can define such letters to correspond to various classes, then use
2355 them in operand constraints.
2357 You should define a class for the union of two classes whenever some
2358 instruction allows both classes. For example, if an instruction allows
2359 either a floating point (coprocessor) register or a general register for a
2360 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2361 which includes both of them. Otherwise you will get suboptimal code.
2363 You must also specify certain redundant information about the register
2364 classes: for each class, which classes contain it and which ones are
2365 contained in it; for each pair of classes, the largest class contained
2368 When a value occupying several consecutive registers is expected in a
2369 certain class, all the registers used must belong to that class.
2370 Therefore, register classes cannot be used to enforce a requirement for
2371 a register pair to start with an even-numbered register. The way to
2372 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2374 Register classes used for input-operands of bitwise-and or shift
2375 instructions have a special requirement: each such class must have, for
2376 each fixed-point machine mode, a subclass whose registers can transfer that
2377 mode to or from memory. For example, on some machines, the operations for
2378 single-byte values (@code{QImode}) are limited to certain registers. When
2379 this is so, each register class that is used in a bitwise-and or shift
2380 instruction must have a subclass consisting of registers from which
2381 single-byte values can be loaded or stored. This is so that
2382 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2384 @deftp {Data type} {enum reg_class}
2385 An enumerated type that must be defined with all the register class names
2386 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2387 must be the last register class, followed by one more enumerated value,
2388 @code{LIM_REG_CLASSES}, which is not a register class but rather
2389 tells how many classes there are.
2391 Each register class has a number, which is the value of casting
2392 the class name to type @code{int}. The number serves as an index
2393 in many of the tables described below.
2396 @defmac N_REG_CLASSES
2397 The number of distinct register classes, defined as follows:
2400 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2404 @defmac REG_CLASS_NAMES
2405 An initializer containing the names of the register classes as C string
2406 constants. These names are used in writing some of the debugging dumps.
2409 @defmac REG_CLASS_CONTENTS
2410 An initializer containing the contents of the register classes, as integers
2411 which are bit masks. The @var{n}th integer specifies the contents of class
2412 @var{n}. The way the integer @var{mask} is interpreted is that
2413 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2415 When the machine has more than 32 registers, an integer does not suffice.
2416 Then the integers are replaced by sub-initializers, braced groupings containing
2417 several integers. Each sub-initializer must be suitable as an initializer
2418 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2419 In this situation, the first integer in each sub-initializer corresponds to
2420 registers 0 through 31, the second integer to registers 32 through 63, and
2424 @defmac REGNO_REG_CLASS (@var{regno})
2425 A C expression whose value is a register class containing hard register
2426 @var{regno}. In general there is more than one such class; choose a class
2427 which is @dfn{minimal}, meaning that no smaller class also contains the
2431 @defmac BASE_REG_CLASS
2432 A macro whose definition is the name of the class to which a valid
2433 base register must belong. A base register is one used in an address
2434 which is the register value plus a displacement.
2437 @defmac MODE_BASE_REG_CLASS (@var{mode})
2438 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2439 the selection of a base register in a mode dependent manner. If
2440 @var{mode} is VOIDmode then it should return the same value as
2441 @code{BASE_REG_CLASS}.
2444 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2445 A C expression whose value is the register class to which a valid
2446 base register must belong in order to be used in a base plus index
2447 register address. You should define this macro if base plus index
2448 addresses have different requirements than other base register uses.
2451 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{outer_code}, @var{index_code})
2452 A C expression whose value is the register class to which a valid
2453 base register must belong. @var{outer_code} and @var{index_code} define the
2454 context in which the base register occurs. @var{outer_code} is the code of
2455 the immediately enclosing expression (@code{MEM} for the top level of an
2456 address, @code{ADDRESS} for something that occurs in an
2457 @code{address_operand}). @var{index_code} is the code of the corresponding
2458 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2461 @defmac INDEX_REG_CLASS
2462 A macro whose definition is the name of the class to which a valid
2463 index register must belong. An index register is one used in an
2464 address where its value is either multiplied by a scale factor or
2465 added to another register (as well as added to a displacement).
2468 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2469 A C expression which is nonzero if register number @var{num} is
2470 suitable for use as a base register in operand addresses. It may be
2471 either a suitable hard register or a pseudo register that has been
2472 allocated such a hard register.
2475 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2476 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2477 that expression may examine the mode of the memory reference in
2478 @var{mode}. You should define this macro if the mode of the memory
2479 reference affects whether a register may be used as a base register. If
2480 you define this macro, the compiler will use it instead of
2481 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2482 addresses that appear outside a @code{MEM}, i.e., as an
2483 @code{address_operand}.
2487 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2488 A C expression which is nonzero if register number @var{num} is suitable for
2489 use as a base register in base plus index operand addresses, accessing
2490 memory in mode @var{mode}. It may be either a suitable hard register or a
2491 pseudo register that has been allocated such a hard register. You should
2492 define this macro if base plus index addresses have different requirements
2493 than other base register uses.
2495 Use of this macro is deprecated; please use the more general
2496 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2499 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{outer_code}, @var{index_code})
2500 A C expression that is just like @code{REGNO_MODE_OK_FOR_BASE_P}, except
2501 that that expression may examine the context in which the register
2502 appears in the memory reference. @var{outer_code} is the code of the
2503 immediately enclosing expression (@code{MEM} if at the top level of the
2504 address, @code{ADDRESS} for something that occurs in an
2505 @code{address_operand}). @var{index_code} is the code of the
2506 corresponding index expression if @var{outer_code} is @code{PLUS};
2507 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2508 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2511 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2512 A C expression which is nonzero if register number @var{num} is
2513 suitable for use as an index register in operand addresses. It may be
2514 either a suitable hard register or a pseudo register that has been
2515 allocated such a hard register.
2517 The difference between an index register and a base register is that
2518 the index register may be scaled. If an address involves the sum of
2519 two registers, neither one of them scaled, then either one may be
2520 labeled the ``base'' and the other the ``index''; but whichever
2521 labeling is used must fit the machine's constraints of which registers
2522 may serve in each capacity. The compiler will try both labelings,
2523 looking for one that is valid, and will reload one or both registers
2524 only if neither labeling works.
2527 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2528 A C expression that places additional restrictions on the register class
2529 to use when it is necessary to copy value @var{x} into a register in class
2530 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2531 another, smaller class. On many machines, the following definition is
2535 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2538 Sometimes returning a more restrictive class makes better code. For
2539 example, on the 68000, when @var{x} is an integer constant that is in range
2540 for a @samp{moveq} instruction, the value of this macro is always
2541 @code{DATA_REGS} as long as @var{class} includes the data registers.
2542 Requiring a data register guarantees that a @samp{moveq} will be used.
2544 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2545 @var{class} is if @var{x} is a legitimate constant which cannot be
2546 loaded into some register class. By returning @code{NO_REGS} you can
2547 force @var{x} into a memory location. For example, rs6000 can load
2548 immediate values into general-purpose registers, but does not have an
2549 instruction for loading an immediate value into a floating-point
2550 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2551 @var{x} is a floating-point constant. If the constant can't be loaded
2552 into any kind of register, code generation will be better if
2553 @code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2554 of using @code{PREFERRED_RELOAD_CLASS}.
2556 If an insn has pseudos in it after register allocation, reload will go
2557 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2558 to find the best one. Returning @code{NO_REGS}, in this case, makes
2559 reload add a @code{!} in front of the constraint: the x86 back-end uses
2560 this feature to discourage usage of 387 registers when math is done in
2561 the SSE registers (and vice versa).
2564 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2565 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2566 input reloads. If you don't define this macro, the default is to use
2567 @var{class}, unchanged.
2569 You can also use @code{PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2570 reload from using some alternatives, like @code{PREFERRED_RELOAD_CLASS}.
2573 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2574 A C expression that places additional restrictions on the register class
2575 to use when it is necessary to be able to hold a value of mode
2576 @var{mode} in a reload register for which class @var{class} would
2579 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2580 there are certain modes that simply can't go in certain reload classes.
2582 The value is a register class; perhaps @var{class}, or perhaps another,
2585 Don't define this macro unless the target machine has limitations which
2586 require the macro to do something nontrivial.
2589 @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})
2590 Many machines have some registers that cannot be copied directly to or
2591 from memory or even from other types of registers. An example is the
2592 @samp{MQ} register, which on most machines, can only be copied to or
2593 from general registers, but not memory. Below, we shall be using the
2594 term 'intermediate register' when a move operation cannot be performed
2595 directly, but has to be done by copying the source into the intermediate
2596 register first, and then copying the intermediate register to the
2597 destination. An intermediate register always has the same mode as
2598 source and destination. Since it holds the actual value being copied,
2599 reload might apply optimizations to re-use an intermediate register
2600 and eliding the copy from the source when it can determine that the
2601 intermediate register still holds the required value.
2603 Another kind of secondary reload is required on some machines which
2604 allow copying all registers to and from memory, but require a scratch
2605 register for stores to some memory locations (e.g., those with symbolic
2606 address on the RT, and those with certain symbolic address on the SPARC
2607 when compiling PIC)@. Scratch registers need not have the same mode
2608 as the value being copied, and usually hold a different value that
2609 that being copied. Special patterns in the md file are needed to
2610 describe how the copy is performed with the help of the scratch register;
2611 these patterns also describe the number, register class(es) and mode(s)
2612 of the scratch register(s).
2614 In some cases, both an intermediate and a scratch register are required.
2616 For input reloads, this target hook is called with nonzero @var{in_p},
2617 and @var{x} is an rtx that needs to be copied to a register in of class
2618 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2619 hook is called with zero @var{in_p}, and a register of class @var{reload_mode}
2620 needs to be copied to rtx @var{x} in @var{reload_mode}.
2622 If copying a register of @var{reload_class} from/to @var{x} requires
2623 an intermediate register, the hook @code{secondary_reload} should
2624 return the register class required for this intermediate register.
2625 If no intermediate register is required, it should return NO_REGS.
2626 If more than one intermediate register is required, describe the one
2627 that is closest in the copy chain to the reload register.
2629 If scratch registers are needed, you also have to describe how to
2630 perform the copy from/to the reload register to/from this
2631 closest intermediate register. Or if no intermediate register is
2632 required, but still a scratch register is needed, describe the
2633 copy from/to the reload register to/from the reload operand @var{x}.
2635 You do this by setting @code{sri->icode} to the instruction code of a pattern
2636 in the md file which performs the move. Operands 0 and 1 are the output
2637 and input of this copy, respectively. Operands from operand 2 onward are
2638 for scratch operands. These scratch operands must have a mode, and a
2639 single-register-class
2640 @c [later: or memory]
2643 When an intermediate register is used, the @code{secondary_reload}
2644 hook will be called again to determine how to copy the intermediate
2645 register to/from the reload operand @var{x}, so your hook must also
2646 have code to handle the register class of the intermediate operand.
2648 @c [For later: maybe we'll allow multi-alternative reload patterns -
2649 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2650 @c and match the constraints of input and output to determine the required
2651 @c alternative. A restriction would be that constraints used to match
2652 @c against reloads registers would have to be written as register class
2653 @c constraints, or we need a new target macro / hook that tells us if an
2654 @c arbitrary constraint can match an unknown register of a given class.
2655 @c Such a macro / hook would also be useful in other places.]
2658 @var{x} might be a pseudo-register or a @code{subreg} of a
2659 pseudo-register, which could either be in a hard register or in memory.
2660 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2661 in memory and the hard register number if it is in a register.
2663 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2664 currently not supported. For the time being, you will have to continue
2665 to use @code{SECONDARY_MEMORY_NEEDED} for that purpose.
2667 @code{copy_cost} also uses this target hook to find out how values are
2668 copied. If you want it to include some extra cost for the need to allocate
2669 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2670 Or if two dependent moves are supposed to have a lower cost than the sum
2671 of the individual moves due to expected fortuitous scheduling and/or special
2672 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2675 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2676 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2677 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2678 These macros are obsolete, new ports should use the target hook
2679 @code{TARGET_SECONDARY_RELOAD} instead.
2681 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2682 target hook. Older ports still define these macros to indicate to the
2683 reload phase that it may
2684 need to allocate at least one register for a reload in addition to the
2685 register to contain the data. Specifically, if copying @var{x} to a
2686 register @var{class} in @var{mode} requires an intermediate register,
2687 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2688 largest register class all of whose registers can be used as
2689 intermediate registers or scratch registers.
2691 If copying a register @var{class} in @var{mode} to @var{x} requires an
2692 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2693 was supposed to be defined be defined to return the largest register
2694 class required. If the
2695 requirements for input and output reloads were the same, the macro
2696 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2699 The values returned by these macros are often @code{GENERAL_REGS}.
2700 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2701 can be directly copied to or from a register of @var{class} in
2702 @var{mode} without requiring a scratch register. Do not define this
2703 macro if it would always return @code{NO_REGS}.
2705 If a scratch register is required (either with or without an
2706 intermediate register), you were supposed to define patterns for
2707 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2708 (@pxref{Standard Names}. These patterns, which were normally
2709 implemented with a @code{define_expand}, should be similar to the
2710 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2713 These patterns need constraints for the reload register and scratch
2715 contain a single register class. If the original reload register (whose
2716 class is @var{class}) can meet the constraint given in the pattern, the
2717 value returned by these macros is used for the class of the scratch
2718 register. Otherwise, two additional reload registers are required.
2719 Their classes are obtained from the constraints in the insn pattern.
2721 @var{x} might be a pseudo-register or a @code{subreg} of a
2722 pseudo-register, which could either be in a hard register or in memory.
2723 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2724 in memory and the hard register number if it is in a register.
2726 These macros should not be used in the case where a particular class of
2727 registers can only be copied to memory and not to another class of
2728 registers. In that case, secondary reload registers are not needed and
2729 would not be helpful. Instead, a stack location must be used to perform
2730 the copy and the @code{mov@var{m}} pattern should use memory as an
2731 intermediate storage. This case often occurs between floating-point and
2735 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2736 Certain machines have the property that some registers cannot be copied
2737 to some other registers without using memory. Define this macro on
2738 those machines to be a C expression that is nonzero if objects of mode
2739 @var{m} in registers of @var{class1} can only be copied to registers of
2740 class @var{class2} by storing a register of @var{class1} into memory
2741 and loading that memory location into a register of @var{class2}.
2743 Do not define this macro if its value would always be zero.
2746 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2747 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2748 allocates a stack slot for a memory location needed for register copies.
2749 If this macro is defined, the compiler instead uses the memory location
2750 defined by this macro.
2752 Do not define this macro if you do not define
2753 @code{SECONDARY_MEMORY_NEEDED}.
2756 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2757 When the compiler needs a secondary memory location to copy between two
2758 registers of mode @var{mode}, it normally allocates sufficient memory to
2759 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2760 load operations in a mode that many bits wide and whose class is the
2761 same as that of @var{mode}.
2763 This is right thing to do on most machines because it ensures that all
2764 bits of the register are copied and prevents accesses to the registers
2765 in a narrower mode, which some machines prohibit for floating-point
2768 However, this default behavior is not correct on some machines, such as
2769 the DEC Alpha, that store short integers in floating-point registers
2770 differently than in integer registers. On those machines, the default
2771 widening will not work correctly and you must define this macro to
2772 suppress that widening in some cases. See the file @file{alpha.h} for
2775 Do not define this macro if you do not define
2776 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2777 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2780 @defmac SMALL_REGISTER_CLASSES
2781 On some machines, it is risky to let hard registers live across arbitrary
2782 insns. Typically, these machines have instructions that require values
2783 to be in specific registers (like an accumulator), and reload will fail
2784 if the required hard register is used for another purpose across such an
2787 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2788 value on these machines. When this macro has a nonzero value, the
2789 compiler will try to minimize the lifetime of hard registers.
2791 It is always safe to define this macro with a nonzero value, but if you
2792 unnecessarily define it, you will reduce the amount of optimizations
2793 that can be performed in some cases. If you do not define this macro
2794 with a nonzero value when it is required, the compiler will run out of
2795 spill registers and print a fatal error message. For most machines, you
2796 should not define this macro at all.
2799 @defmac CLASS_LIKELY_SPILLED_P (@var{class})
2800 A C expression whose value is nonzero if pseudos that have been assigned
2801 to registers of class @var{class} would likely be spilled because
2802 registers of @var{class} are needed for spill registers.
2804 The default value of this macro returns 1 if @var{class} has exactly one
2805 register and zero otherwise. On most machines, this default should be
2806 used. Only define this macro to some other expression if pseudos
2807 allocated by @file{local-alloc.c} end up in memory because their hard
2808 registers were needed for spill registers. If this macro returns nonzero
2809 for those classes, those pseudos will only be allocated by
2810 @file{global.c}, which knows how to reallocate the pseudo to another
2811 register. If there would not be another register available for
2812 reallocation, you should not change the definition of this macro since
2813 the only effect of such a definition would be to slow down register
2817 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2818 A C expression for the maximum number of consecutive registers
2819 of class @var{class} needed to hold a value of mode @var{mode}.
2821 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2822 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2823 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2824 @var{mode})} for all @var{regno} values in the class @var{class}.
2826 This macro helps control the handling of multiple-word values
2830 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2831 If defined, a C expression that returns nonzero for a @var{class} for which
2832 a change from mode @var{from} to mode @var{to} is invalid.
2834 For the example, loading 32-bit integer or floating-point objects into
2835 floating-point registers on the Alpha extends them to 64 bits.
2836 Therefore loading a 64-bit object and then storing it as a 32-bit object
2837 does not store the low-order 32 bits, as would be the case for a normal
2838 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2842 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2843 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2844 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2848 @node Old Constraints
2849 @section Obsolete Macros for Defining Constraints
2850 @cindex defining constraints, obsolete method
2851 @cindex constraints, defining, obsolete method
2853 Machine-specific constraints can be defined with these macros instead
2854 of the machine description constructs described in @ref{Define
2855 Constraints}. This mechanism is obsolete. New ports should not use
2856 it; old ports should convert to the new mechanism.
2858 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2859 For the constraint at the start of @var{str}, which starts with the letter
2860 @var{c}, return the length. This allows you to have register class /
2861 constant / extra constraints that are longer than a single letter;
2862 you don't need to define this macro if you can do with single-letter
2863 constraints only. The definition of this macro should use
2864 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2865 to handle specially.
2866 There are some sanity checks in genoutput.c that check the constraint lengths
2867 for the md file, so you can also use this macro to help you while you are
2868 transitioning from a byzantine single-letter-constraint scheme: when you
2869 return a negative length for a constraint you want to re-use, genoutput
2870 will complain about every instance where it is used in the md file.
2873 @defmac REG_CLASS_FROM_LETTER (@var{char})
2874 A C expression which defines the machine-dependent operand constraint
2875 letters for register classes. If @var{char} is such a letter, the
2876 value should be the register class corresponding to it. Otherwise,
2877 the value should be @code{NO_REGS}. The register letter @samp{r},
2878 corresponding to class @code{GENERAL_REGS}, will not be passed
2879 to this macro; you do not need to handle it.
2882 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2883 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2884 passed in @var{str}, so that you can use suffixes to distinguish between
2888 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2889 A C expression that defines the machine-dependent operand constraint
2890 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2891 particular ranges of integer values. If @var{c} is one of those
2892 letters, the expression should check that @var{value}, an integer, is in
2893 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2894 not one of those letters, the value should be 0 regardless of
2898 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2899 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2900 string passed in @var{str}, so that you can use suffixes to distinguish
2901 between different variants.
2904 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2905 A C expression that defines the machine-dependent operand constraint
2906 letters that specify particular ranges of @code{const_double} values
2907 (@samp{G} or @samp{H}).
2909 If @var{c} is one of those letters, the expression should check that
2910 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2911 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2912 letters, the value should be 0 regardless of @var{value}.
2914 @code{const_double} is used for all floating-point constants and for
2915 @code{DImode} fixed-point constants. A given letter can accept either
2916 or both kinds of values. It can use @code{GET_MODE} to distinguish
2917 between these kinds.
2920 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2921 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2922 string passed in @var{str}, so that you can use suffixes to distinguish
2923 between different variants.
2926 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2927 A C expression that defines the optional machine-dependent constraint
2928 letters that can be used to segregate specific types of operands, usually
2929 memory references, for the target machine. Any letter that is not
2930 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2931 @code{REG_CLASS_FROM_CONSTRAINT}
2932 may be used. Normally this macro will not be defined.
2934 If it is required for a particular target machine, it should return 1
2935 if @var{value} corresponds to the operand type represented by the
2936 constraint letter @var{c}. If @var{c} is not defined as an extra
2937 constraint, the value returned should be 0 regardless of @var{value}.
2939 For example, on the ROMP, load instructions cannot have their output
2940 in r0 if the memory reference contains a symbolic address. Constraint
2941 letter @samp{Q} is defined as representing a memory address that does
2942 @emph{not} contain a symbolic address. An alternative is specified with
2943 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2944 alternative specifies @samp{m} on the input and a register class that
2945 does not include r0 on the output.
2948 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2949 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2950 in @var{str}, so that you can use suffixes to distinguish between different
2954 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2955 A C expression that defines the optional machine-dependent constraint
2956 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2957 be treated like memory constraints by the reload pass.
2959 It should return 1 if the operand type represented by the constraint
2960 at the start of @var{str}, the first letter of which is the letter @var{c},
2961 comprises a subset of all memory references including
2962 all those whose address is simply a base register. This allows the reload
2963 pass to reload an operand, if it does not directly correspond to the operand
2964 type of @var{c}, by copying its address into a base register.
2966 For example, on the S/390, some instructions do not accept arbitrary
2967 memory references, but only those that do not make use of an index
2968 register. The constraint letter @samp{Q} is defined via
2969 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2970 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2971 a @samp{Q} constraint can handle any memory operand, because the
2972 reload pass knows it can be reloaded by copying the memory address
2973 into a base register if required. This is analogous to the way
2974 a @samp{o} constraint can handle any memory operand.
2977 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2978 A C expression that defines the optional machine-dependent constraint
2979 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2980 @code{EXTRA_CONSTRAINT_STR}, that should
2981 be treated like address constraints by the reload pass.
2983 It should return 1 if the operand type represented by the constraint
2984 at the start of @var{str}, which starts with the letter @var{c}, comprises
2985 a subset of all memory addresses including
2986 all those that consist of just a base register. This allows the reload
2987 pass to reload an operand, if it does not directly correspond to the operand
2988 type of @var{str}, by copying it into a base register.
2990 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2991 be used with the @code{address_operand} predicate. It is treated
2992 analogously to the @samp{p} constraint.
2995 @node Stack and Calling
2996 @section Stack Layout and Calling Conventions
2997 @cindex calling conventions
2999 @c prevent bad page break with this line
3000 This describes the stack layout and calling conventions.
3004 * Exception Handling::
3009 * Register Arguments::
3011 * Aggregate Return::
3016 * Stack Smashing Protection::
3020 @subsection Basic Stack Layout
3021 @cindex stack frame layout
3022 @cindex frame layout
3024 @c prevent bad page break with this line
3025 Here is the basic stack layout.
3027 @defmac STACK_GROWS_DOWNWARD
3028 Define this macro if pushing a word onto the stack moves the stack
3029 pointer to a smaller address.
3031 When we say, ``define this macro if @dots{}'', it means that the
3032 compiler checks this macro only with @code{#ifdef} so the precise
3033 definition used does not matter.
3036 @defmac STACK_PUSH_CODE
3037 This macro defines the operation used when something is pushed
3038 on the stack. In RTL, a push operation will be
3039 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
3041 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
3042 and @code{POST_INC}. Which of these is correct depends on
3043 the stack direction and on whether the stack pointer points
3044 to the last item on the stack or whether it points to the
3045 space for the next item on the stack.
3047 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
3048 defined, which is almost always right, and @code{PRE_INC} otherwise,
3049 which is often wrong.
3052 @defmac FRAME_GROWS_DOWNWARD
3053 Define this macro to nonzero value if the addresses of local variable slots
3054 are at negative offsets from the frame pointer.
3057 @defmac ARGS_GROW_DOWNWARD
3058 Define this macro if successive arguments to a function occupy decreasing
3059 addresses on the stack.
3062 @defmac STARTING_FRAME_OFFSET
3063 Offset from the frame pointer to the first local variable slot to be allocated.
3065 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
3066 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
3067 Otherwise, it is found by adding the length of the first slot to the
3068 value @code{STARTING_FRAME_OFFSET}.
3069 @c i'm not sure if the above is still correct.. had to change it to get
3070 @c rid of an overfull. --mew 2feb93
3073 @defmac STACK_ALIGNMENT_NEEDED
3074 Define to zero to disable final alignment of the stack during reload.
3075 The nonzero default for this macro is suitable for most ports.
3077 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
3078 is a register save block following the local block that doesn't require
3079 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
3080 stack alignment and do it in the backend.
3083 @defmac STACK_POINTER_OFFSET
3084 Offset from the stack pointer register to the first location at which
3085 outgoing arguments are placed. If not specified, the default value of
3086 zero is used. This is the proper value for most machines.
3088 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3089 the first location at which outgoing arguments are placed.
3092 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3093 Offset from the argument pointer register to the first argument's
3094 address. On some machines it may depend on the data type of the
3097 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3098 the first argument's address.
3101 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3102 Offset from the stack pointer register to an item dynamically allocated
3103 on the stack, e.g., by @code{alloca}.
3105 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3106 length of the outgoing arguments. The default is correct for most
3107 machines. See @file{function.c} for details.
3110 @defmac INITIAL_FRAME_ADDRESS_RTX
3111 A C expression whose value is RTL representing the address of the initial
3112 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3113 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3114 default value will be used. Define this macro in order to make frame pointer
3115 elimination work in the presence of @code{__builtin_frame_address (count)} and
3116 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3119 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3120 A C expression whose value is RTL representing the address in a stack
3121 frame where the pointer to the caller's frame is stored. Assume that
3122 @var{frameaddr} is an RTL expression for the address of the stack frame
3125 If you don't define this macro, the default is to return the value
3126 of @var{frameaddr}---that is, the stack frame address is also the
3127 address of the stack word that points to the previous frame.
3130 @defmac SETUP_FRAME_ADDRESSES
3131 If defined, a C expression that produces the machine-specific code to
3132 setup the stack so that arbitrary frames can be accessed. For example,
3133 on the SPARC, we must flush all of the register windows to the stack
3134 before we can access arbitrary stack frames. You will seldom need to
3138 @deftypefn {Target Hook} bool TARGET_BUILTIN_SETJMP_FRAME_VALUE ()
3139 This target hook should return an rtx that is used to store
3140 the address of the current frame into the built in @code{setjmp} buffer.
3141 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3142 machines. One reason you may need to define this target hook is if
3143 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3146 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3147 A C expression whose value is RTL representing the value of the frame
3148 address for the current frame. @var{frameaddr} is the frame pointer
3149 of the current frame. This is used for __builtin_frame_address.
3150 You need only define this macro if the frame address is not the same
3151 as the frame pointer. Most machines do not need to define it.
3154 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3155 A C expression whose value is RTL representing the value of the return
3156 address for the frame @var{count} steps up from the current frame, after
3157 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3158 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3159 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
3161 The value of the expression must always be the correct address when
3162 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
3163 determine the return address of other frames.
3166 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3167 Define this if the return address of a particular stack frame is accessed
3168 from the frame pointer of the previous stack frame.
3171 @defmac INCOMING_RETURN_ADDR_RTX
3172 A C expression whose value is RTL representing the location of the
3173 incoming return address at the beginning of any function, before the
3174 prologue. This RTL is either a @code{REG}, indicating that the return
3175 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3178 You only need to define this macro if you want to support call frame
3179 debugging information like that provided by DWARF 2.
3181 If this RTL is a @code{REG}, you should also define
3182 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3185 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3186 A C expression whose value is an integer giving a DWARF 2 column
3187 number that may be used as an alternative return column. The column
3188 must not correspond to any gcc hard register (that is, it must not
3189 be in the range of @code{DWARF_FRAME_REGNUM}).
3191 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3192 general register, but an alternative column needs to be used for signal
3193 frames. Some targets have also used different frame return columns
3197 @defmac DWARF_ZERO_REG
3198 A C expression whose value is an integer giving a DWARF 2 register
3199 number that is considered to always have the value zero. This should
3200 only be defined if the target has an architected zero register, and
3201 someone decided it was a good idea to use that register number to
3202 terminate the stack backtrace. New ports should avoid this.
3205 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3206 This target hook allows the backend to emit frame-related insns that
3207 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3208 info engine will invoke it on insns of the form
3210 (set (reg) (unspec [...] UNSPEC_INDEX))
3214 (set (reg) (unspec_volatile [...] UNSPECV_INDEX)).
3216 to let the backend emit the call frame instructions. @var{label} is
3217 the CFI label attached to the insn, @var{pattern} is the pattern of
3218 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3221 @defmac INCOMING_FRAME_SP_OFFSET
3222 A C expression whose value is an integer giving the offset, in bytes,
3223 from the value of the stack pointer register to the top of the stack
3224 frame at the beginning of any function, before the prologue. The top of
3225 the frame is defined to be the value of the stack pointer in the
3226 previous frame, just before the call instruction.
3228 You only need to define this macro if you want to support call frame
3229 debugging information like that provided by DWARF 2.
3232 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3233 A C expression whose value is an integer giving the offset, in bytes,
3234 from the argument pointer to the canonical frame address (cfa). The
3235 final value should coincide with that calculated by
3236 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3237 during virtual register instantiation.
3239 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
3240 which is correct for most machines; in general, the arguments are found
3241 immediately before the stack frame. Note that this is not the case on
3242 some targets that save registers into the caller's frame, such as SPARC
3243 and rs6000, and so such targets need to define this macro.
3245 You only need to define this macro if the default is incorrect, and you
3246 want to support call frame debugging information like that provided by
3250 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3251 If defined, a C expression whose value is an integer giving the offset
3252 in bytes from the frame pointer to the canonical frame address (cfa).
3253 The final value should coincide with that calculated by
3254 @code{INCOMING_FRAME_SP_OFFSET}.
3256 Normally the CFA is calculated as an offset from the argument pointer,
3257 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3258 variable due to the ABI, this may not be possible. If this macro is
3259 defined, it implies that the virtual register instantiation should be
3260 based on the frame pointer instead of the argument pointer. Only one
3261 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3265 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3266 If defined, a C expression whose value is an integer giving the offset
3267 in bytes from the canonical frame address (cfa) to the frame base used
3268 in DWARF 2 debug information. The default is zero. A different value
3269 may reduce the size of debug information on some ports.
3272 @node Exception Handling
3273 @subsection Exception Handling Support
3274 @cindex exception handling
3276 @defmac EH_RETURN_DATA_REGNO (@var{N})
3277 A C expression whose value is the @var{N}th register number used for
3278 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3279 @var{N} registers are usable.
3281 The exception handling library routines communicate with the exception
3282 handlers via a set of agreed upon registers. Ideally these registers
3283 should be call-clobbered; it is possible to use call-saved registers,
3284 but may negatively impact code size. The target must support at least
3285 2 data registers, but should define 4 if there are enough free registers.
3287 You must define this macro if you want to support call frame exception
3288 handling like that provided by DWARF 2.
3291 @defmac EH_RETURN_STACKADJ_RTX
3292 A C expression whose value is RTL representing a location in which
3293 to store a stack adjustment to be applied before function return.
3294 This is used to unwind the stack to an exception handler's call frame.
3295 It will be assigned zero on code paths that return normally.
3297 Typically this is a call-clobbered hard register that is otherwise
3298 untouched by the epilogue, but could also be a stack slot.
3300 Do not define this macro if the stack pointer is saved and restored
3301 by the regular prolog and epilog code in the call frame itself; in
3302 this case, the exception handling library routines will update the
3303 stack location to be restored in place. Otherwise, you must define
3304 this macro if you want to support call frame exception handling like
3305 that provided by DWARF 2.
3308 @defmac EH_RETURN_HANDLER_RTX
3309 A C expression whose value is RTL representing a location in which
3310 to store the address of an exception handler to which we should
3311 return. It will not be assigned on code paths that return normally.
3313 Typically this is the location in the call frame at which the normal
3314 return address is stored. For targets that return by popping an
3315 address off the stack, this might be a memory address just below
3316 the @emph{target} call frame rather than inside the current call
3317 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3318 been assigned, so it may be used to calculate the location of the
3321 Some targets have more complex requirements than storing to an
3322 address calculable during initial code generation. In that case
3323 the @code{eh_return} instruction pattern should be used instead.
3325 If you want to support call frame exception handling, you must
3326 define either this macro or the @code{eh_return} instruction pattern.
3329 @defmac RETURN_ADDR_OFFSET
3330 If defined, an integer-valued C expression for which rtl will be generated
3331 to add it to the exception handler address before it is searched in the
3332 exception handling tables, and to subtract it again from the address before
3333 using it to return to the exception handler.
3336 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3337 This macro chooses the encoding of pointers embedded in the exception
3338 handling sections. If at all possible, this should be defined such
3339 that the exception handling section will not require dynamic relocations,
3340 and so may be read-only.
3342 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3343 @var{global} is true if the symbol may be affected by dynamic relocations.
3344 The macro should return a combination of the @code{DW_EH_PE_*} defines
3345 as found in @file{dwarf2.h}.
3347 If this macro is not defined, pointers will not be encoded but
3348 represented directly.
3351 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3352 This macro allows the target to emit whatever special magic is required
3353 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3354 Generic code takes care of pc-relative and indirect encodings; this must
3355 be defined if the target uses text-relative or data-relative encodings.
3357 This is a C statement that branches to @var{done} if the format was
3358 handled. @var{encoding} is the format chosen, @var{size} is the number
3359 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3363 @defmac MD_UNWIND_SUPPORT
3364 A string specifying a file to be #include'd in unwind-dw2.c. The file
3365 so included typically defines @code{MD_FALLBACK_FRAME_STATE_FOR}.
3368 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3369 This macro allows the target to add CPU and operating system specific
3370 code to the call-frame unwinder for use when there is no unwind data
3371 available. The most common reason to implement this macro is to unwind
3372 through signal frames.
3374 This macro is called from @code{uw_frame_state_for} in
3375 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
3376 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3377 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3378 for the address of the code being executed and @code{context->cfa} for
3379 the stack pointer value. If the frame can be decoded, the register
3380 save addresses should be updated in @var{fs} and the macro should
3381 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
3382 the macro should evaluate to @code{_URC_END_OF_STACK}.
3384 For proper signal handling in Java this macro is accompanied by
3385 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3388 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3389 This macro allows the target to add operating system specific code to the
3390 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3391 usually used for signal or interrupt frames.
3393 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3394 @var{context} is an @code{_Unwind_Context};
3395 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3396 for the abi and context in the @code{.unwabi} directive. If the
3397 @code{.unwabi} directive can be handled, the register save addresses should
3398 be updated in @var{fs}.
3401 @defmac TARGET_USES_WEAK_UNWIND_INFO
3402 A C expression that evaluates to true if the target requires unwind
3403 info to be given comdat linkage. Define it to be @code{1} if comdat
3404 linkage is necessary. The default is @code{0}.
3407 @node Stack Checking
3408 @subsection Specifying How Stack Checking is Done
3410 GCC will check that stack references are within the boundaries of
3411 the stack, if the @option{-fstack-check} is specified, in one of three ways:
3415 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3416 will assume that you have arranged for stack checking to be done at
3417 appropriate places in the configuration files, e.g., in
3418 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
3422 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
3423 called @code{check_stack} in your @file{md} file, GCC will call that
3424 pattern with one argument which is the address to compare the stack
3425 value against. You must arrange for this pattern to report an error if
3426 the stack pointer is out of range.
3429 If neither of the above are true, GCC will generate code to periodically
3430 ``probe'' the stack pointer using the values of the macros defined below.
3433 Normally, you will use the default values of these macros, so GCC
3434 will use the third approach.
3436 @defmac STACK_CHECK_BUILTIN
3437 A nonzero value if stack checking is done by the configuration files in a
3438 machine-dependent manner. You should define this macro if stack checking
3439 is require by the ABI of your machine or if you would like to have to stack
3440 checking in some more efficient way than GCC's portable approach.
3441 The default value of this macro is zero.
3444 @defmac STACK_CHECK_PROBE_INTERVAL
3445 An integer representing the interval at which GCC must generate stack
3446 probe instructions. You will normally define this macro to be no larger
3447 than the size of the ``guard pages'' at the end of a stack area. The
3448 default value of 4096 is suitable for most systems.
3451 @defmac STACK_CHECK_PROBE_LOAD
3452 A integer which is nonzero if GCC should perform the stack probe
3453 as a load instruction and zero if GCC should use a store instruction.
3454 The default is zero, which is the most efficient choice on most systems.
3457 @defmac STACK_CHECK_PROTECT
3458 The number of bytes of stack needed to recover from a stack overflow,
3459 for languages where such a recovery is supported. The default value of
3460 75 words should be adequate for most machines.
3463 @defmac STACK_CHECK_MAX_FRAME_SIZE
3464 The maximum size of a stack frame, in bytes. GCC will generate probe
3465 instructions in non-leaf functions to ensure at least this many bytes of
3466 stack are available. If a stack frame is larger than this size, stack
3467 checking will not be reliable and GCC will issue a warning. The
3468 default is chosen so that GCC only generates one instruction on most
3469 systems. You should normally not change the default value of this macro.
3472 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3473 GCC uses this value to generate the above warning message. It
3474 represents the amount of fixed frame used by a function, not including
3475 space for any callee-saved registers, temporaries and user variables.
3476 You need only specify an upper bound for this amount and will normally
3477 use the default of four words.
3480 @defmac STACK_CHECK_MAX_VAR_SIZE
3481 The maximum size, in bytes, of an object that GCC will place in the
3482 fixed area of the stack frame when the user specifies
3483 @option{-fstack-check}.
3484 GCC computed the default from the values of the above macros and you will
3485 normally not need to override that default.
3489 @node Frame Registers
3490 @subsection Registers That Address the Stack Frame
3492 @c prevent bad page break with this line
3493 This discusses registers that address the stack frame.
3495 @defmac STACK_POINTER_REGNUM
3496 The register number of the stack pointer register, which must also be a
3497 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3498 the hardware determines which register this is.
3501 @defmac FRAME_POINTER_REGNUM
3502 The register number of the frame pointer register, which is used to
3503 access automatic variables in the stack frame. On some machines, the
3504 hardware determines which register this is. On other machines, you can
3505 choose any register you wish for this purpose.
3508 @defmac HARD_FRAME_POINTER_REGNUM
3509 On some machines the offset between the frame pointer and starting
3510 offset of the automatic variables is not known until after register
3511 allocation has been done (for example, because the saved registers are
3512 between these two locations). On those machines, define
3513 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3514 be used internally until the offset is known, and define
3515 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3516 used for the frame pointer.
3518 You should define this macro only in the very rare circumstances when it
3519 is not possible to calculate the offset between the frame pointer and
3520 the automatic variables until after register allocation has been
3521 completed. When this macro is defined, you must also indicate in your
3522 definition of @code{ELIMINABLE_REGS} how to eliminate
3523 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3524 or @code{STACK_POINTER_REGNUM}.
3526 Do not define this macro if it would be the same as
3527 @code{FRAME_POINTER_REGNUM}.
3530 @defmac ARG_POINTER_REGNUM
3531 The register number of the arg pointer register, which is used to access
3532 the function's argument list. On some machines, this is the same as the
3533 frame pointer register. On some machines, the hardware determines which
3534 register this is. On other machines, you can choose any register you
3535 wish for this purpose. If this is not the same register as the frame
3536 pointer register, then you must mark it as a fixed register according to
3537 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3538 (@pxref{Elimination}).
3541 @defmac RETURN_ADDRESS_POINTER_REGNUM
3542 The register number of the return address pointer register, which is used to
3543 access the current function's return address from the stack. On some
3544 machines, the return address is not at a fixed offset from the frame
3545 pointer or stack pointer or argument pointer. This register can be defined
3546 to point to the return address on the stack, and then be converted by
3547 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3549 Do not define this macro unless there is no other way to get the return
3550 address from the stack.
3553 @defmac STATIC_CHAIN_REGNUM
3554 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3555 Register numbers used for passing a function's static chain pointer. If
3556 register windows are used, the register number as seen by the called
3557 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3558 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3559 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3562 The static chain register need not be a fixed register.
3564 If the static chain is passed in memory, these macros should not be
3565 defined; instead, the next two macros should be defined.
3568 @defmac STATIC_CHAIN
3569 @defmacx STATIC_CHAIN_INCOMING
3570 If the static chain is passed in memory, these macros provide rtx giving
3571 @code{mem} expressions that denote where they are stored.
3572 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3573 as seen by the calling and called functions, respectively. Often the former
3574 will be at an offset from the stack pointer and the latter at an offset from
3577 @findex stack_pointer_rtx
3578 @findex frame_pointer_rtx
3579 @findex arg_pointer_rtx
3580 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3581 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3582 macros and should be used to refer to those items.
3584 If the static chain is passed in a register, the two previous macros should
3588 @defmac DWARF_FRAME_REGISTERS
3589 This macro specifies the maximum number of hard registers that can be
3590 saved in a call frame. This is used to size data structures used in
3591 DWARF2 exception handling.
3593 Prior to GCC 3.0, this macro was needed in order to establish a stable
3594 exception handling ABI in the face of adding new hard registers for ISA
3595 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3596 in the number of hard registers. Nevertheless, this macro can still be
3597 used to reduce the runtime memory requirements of the exception handling
3598 routines, which can be substantial if the ISA contains a lot of
3599 registers that are not call-saved.
3601 If this macro is not defined, it defaults to
3602 @code{FIRST_PSEUDO_REGISTER}.
3605 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3607 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3608 for backward compatibility in pre GCC 3.0 compiled code.
3610 If this macro is not defined, it defaults to
3611 @code{DWARF_FRAME_REGISTERS}.
3614 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3616 Define this macro if the target's representation for dwarf registers
3617 is different than the internal representation for unwind column.
3618 Given a dwarf register, this macro should return the internal unwind
3619 column number to use instead.
3621 See the PowerPC's SPE target for an example.
3624 @defmac DWARF_FRAME_REGNUM (@var{regno})
3626 Define this macro if the target's representation for dwarf registers
3627 used in .eh_frame or .debug_frame is different from that used in other
3628 debug info sections. Given a GCC hard register number, this macro
3629 should return the .eh_frame register number. The default is
3630 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3634 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3636 Define this macro to map register numbers held in the call frame info
3637 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3638 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3639 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3640 return @code{@var{regno}}.
3645 @subsection Eliminating Frame Pointer and Arg Pointer
3647 @c prevent bad page break with this line
3648 This is about eliminating the frame pointer and arg pointer.
3650 @defmac FRAME_POINTER_REQUIRED
3651 A C expression which is nonzero if a function must have and use a frame
3652 pointer. This expression is evaluated in the reload pass. If its value is
3653 nonzero the function will have a frame pointer.
3655 The expression can in principle examine the current function and decide
3656 according to the facts, but on most machines the constant 0 or the
3657 constant 1 suffices. Use 0 when the machine allows code to be generated
3658 with no frame pointer, and doing so saves some time or space. Use 1
3659 when there is no possible advantage to avoiding a frame pointer.
3661 In certain cases, the compiler does not know how to produce valid code
3662 without a frame pointer. The compiler recognizes those cases and
3663 automatically gives the function a frame pointer regardless of what
3664 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3667 In a function that does not require a frame pointer, the frame pointer
3668 register can be allocated for ordinary usage, unless you mark it as a
3669 fixed register. See @code{FIXED_REGISTERS} for more information.
3672 @findex get_frame_size
3673 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3674 A C statement to store in the variable @var{depth-var} the difference
3675 between the frame pointer and the stack pointer values immediately after
3676 the function prologue. The value would be computed from information
3677 such as the result of @code{get_frame_size ()} and the tables of
3678 registers @code{regs_ever_live} and @code{call_used_regs}.
3680 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3681 need not be defined. Otherwise, it must be defined even if
3682 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3683 case, you may set @var{depth-var} to anything.
3686 @defmac ELIMINABLE_REGS
3687 If defined, this macro specifies a table of register pairs used to
3688 eliminate unneeded registers that point into the stack frame. If it is not
3689 defined, the only elimination attempted by the compiler is to replace
3690 references to the frame pointer with references to the stack pointer.
3692 The definition of this macro is a list of structure initializations, each
3693 of which specifies an original and replacement register.
3695 On some machines, the position of the argument pointer is not known until
3696 the compilation is completed. In such a case, a separate hard register
3697 must be used for the argument pointer. This register can be eliminated by
3698 replacing it with either the frame pointer or the argument pointer,
3699 depending on whether or not the frame pointer has been eliminated.
3701 In this case, you might specify:
3703 #define ELIMINABLE_REGS \
3704 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3705 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3706 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3709 Note that the elimination of the argument pointer with the stack pointer is
3710 specified first since that is the preferred elimination.
3713 @defmac CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3714 A C expression that returns nonzero if the compiler is allowed to try
3715 to replace register number @var{from-reg} with register number
3716 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3717 is defined, and will usually be the constant 1, since most of the cases
3718 preventing register elimination are things that the compiler already
3722 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3723 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3724 specifies the initial difference between the specified pair of
3725 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3729 @node Stack Arguments
3730 @subsection Passing Function Arguments on the Stack
3731 @cindex arguments on stack
3732 @cindex stack arguments
3734 The macros in this section control how arguments are passed
3735 on the stack. See the following section for other macros that
3736 control passing certain arguments in registers.
3738 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (tree @var{fntype})
3739 This target hook returns @code{true} if an argument declared in a
3740 prototype as an integral type smaller than @code{int} should actually be
3741 passed as an @code{int}. In addition to avoiding errors in certain
3742 cases of mismatch, it also makes for better code on certain machines.
3743 The default is to not promote prototypes.
3747 A C expression. If nonzero, push insns will be used to pass
3749 If the target machine does not have a push instruction, set it to zero.
3750 That directs GCC to use an alternate strategy: to
3751 allocate the entire argument block and then store the arguments into
3752 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3755 @defmac PUSH_ARGS_REVERSED
3756 A C expression. If nonzero, function arguments will be evaluated from
3757 last to first, rather than from first to last. If this macro is not
3758 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3759 and args grow in opposite directions, and 0 otherwise.
3762 @defmac PUSH_ROUNDING (@var{npushed})
3763 A C expression that is the number of bytes actually pushed onto the
3764 stack when an instruction attempts to push @var{npushed} bytes.
3766 On some machines, the definition
3769 #define PUSH_ROUNDING(BYTES) (BYTES)
3773 will suffice. But on other machines, instructions that appear
3774 to push one byte actually push two bytes in an attempt to maintain
3775 alignment. Then the definition should be
3778 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3782 @findex current_function_outgoing_args_size
3783 @defmac ACCUMULATE_OUTGOING_ARGS
3784 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3785 will be computed and placed into the variable
3786 @code{current_function_outgoing_args_size}. No space will be pushed
3787 onto the stack for each call; instead, the function prologue should
3788 increase the stack frame size by this amount.
3790 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3794 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3795 Define this macro if functions should assume that stack space has been
3796 allocated for arguments even when their values are passed in
3799 The value of this macro is the size, in bytes, of the area reserved for
3800 arguments passed in registers for the function represented by @var{fndecl},
3801 which can be zero if GCC is calling a library function.
3803 This space can be allocated by the caller, or be a part of the
3804 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3807 @c above is overfull. not sure what to do. --mew 5feb93 did
3808 @c something, not sure if it looks good. --mew 10feb93
3810 @defmac OUTGOING_REG_PARM_STACK_SPACE
3811 Define this to a nonzero value if it is the responsibility of the caller
3812 to allocate the area reserved for arguments passed in registers.
3814 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3815 whether the space for these arguments counts in the value of
3816 @code{current_function_outgoing_args_size}.
3819 @defmac STACK_PARMS_IN_REG_PARM_AREA
3820 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3821 stack parameters don't skip the area specified by it.
3822 @c i changed this, makes more sens and it should have taken care of the
3823 @c overfull.. not as specific, tho. --mew 5feb93
3825 Normally, when a parameter is not passed in registers, it is placed on the
3826 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3827 suppresses this behavior and causes the parameter to be passed on the
3828 stack in its natural location.
3831 @defmac RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3832 A C expression that should indicate the number of bytes of its own
3833 arguments that a function pops on returning, or 0 if the
3834 function pops no arguments and the caller must therefore pop them all
3835 after the function returns.
3837 @var{fundecl} is a C variable whose value is a tree node that describes
3838 the function in question. Normally it is a node of type
3839 @code{FUNCTION_DECL} that describes the declaration of the function.
3840 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3842 @var{funtype} is a C variable whose value is a tree node that
3843 describes the function in question. Normally it is a node of type
3844 @code{FUNCTION_TYPE} that describes the data type of the function.
3845 From this it is possible to obtain the data types of the value and
3846 arguments (if known).
3848 When a call to a library function is being considered, @var{fundecl}
3849 will contain an identifier node for the library function. Thus, if
3850 you need to distinguish among various library functions, you can do so
3851 by their names. Note that ``library function'' in this context means
3852 a function used to perform arithmetic, whose name is known specially
3853 in the compiler and was not mentioned in the C code being compiled.
3855 @var{stack-size} is the number of bytes of arguments passed on the
3856 stack. If a variable number of bytes is passed, it is zero, and
3857 argument popping will always be the responsibility of the calling function.
3859 On the VAX, all functions always pop their arguments, so the definition
3860 of this macro is @var{stack-size}. On the 68000, using the standard
3861 calling convention, no functions pop their arguments, so the value of
3862 the macro is always 0 in this case. But an alternative calling
3863 convention is available in which functions that take a fixed number of
3864 arguments pop them but other functions (such as @code{printf}) pop
3865 nothing (the caller pops all). When this convention is in use,
3866 @var{funtype} is examined to determine whether a function takes a fixed
3867 number of arguments.
3870 @defmac CALL_POPS_ARGS (@var{cum})
3871 A C expression that should indicate the number of bytes a call sequence
3872 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3873 when compiling a function call.
3875 @var{cum} is the variable in which all arguments to the called function
3876 have been accumulated.
3878 On certain architectures, such as the SH5, a call trampoline is used
3879 that pops certain registers off the stack, depending on the arguments
3880 that have been passed to the function. Since this is a property of the
3881 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3885 @node Register Arguments
3886 @subsection Passing Arguments in Registers
3887 @cindex arguments in registers
3888 @cindex registers arguments
3890 This section describes the macros which let you control how various
3891 types of arguments are passed in registers or how they are arranged in
3894 @defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3895 A C expression that controls whether a function argument is passed
3896 in a register, and which register.
3898 The arguments are @var{cum}, which summarizes all the previous
3899 arguments; @var{mode}, the machine mode of the argument; @var{type},
3900 the data type of the argument as a tree node or 0 if that is not known
3901 (which happens for C support library functions); and @var{named},
3902 which is 1 for an ordinary argument and 0 for nameless arguments that
3903 correspond to @samp{@dots{}} in the called function's prototype.
3904 @var{type} can be an incomplete type if a syntax error has previously
3907 The value of the expression is usually either a @code{reg} RTX for the
3908 hard register in which to pass the argument, or zero to pass the
3909 argument on the stack.
3911 For machines like the VAX and 68000, where normally all arguments are
3912 pushed, zero suffices as a definition.
3914 The value of the expression can also be a @code{parallel} RTX@. This is
3915 used when an argument is passed in multiple locations. The mode of the
3916 @code{parallel} should be the mode of the entire argument. The
3917 @code{parallel} holds any number of @code{expr_list} pairs; each one
3918 describes where part of the argument is passed. In each
3919 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3920 register in which to pass this part of the argument, and the mode of the
3921 register RTX indicates how large this part of the argument is. The
3922 second operand of the @code{expr_list} is a @code{const_int} which gives
3923 the offset in bytes into the entire argument of where this part starts.
3924 As a special exception the first @code{expr_list} in the @code{parallel}
3925 RTX may have a first operand of zero. This indicates that the entire
3926 argument is also stored on the stack.
3928 The last time this macro is called, it is called with @code{MODE ==
3929 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3930 pattern as operands 2 and 3 respectively.
3932 @cindex @file{stdarg.h} and register arguments
3933 The usual way to make the ISO library @file{stdarg.h} work on a machine
3934 where some arguments are usually passed in registers, is to cause
3935 nameless arguments to be passed on the stack instead. This is done
3936 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3938 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3939 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3940 You may use the hook @code{targetm.calls.must_pass_in_stack}
3941 in the definition of this macro to determine if this argument is of a
3942 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3943 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3944 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3945 defined, the argument will be computed in the stack and then loaded into
3949 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, tree @var{type})
3950 This target hook should return @code{true} if we should not pass @var{type}
3951 solely in registers. The file @file{expr.h} defines a
3952 definition that is usually appropriate, refer to @file{expr.h} for additional
3956 @defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3957 Define this macro if the target machine has ``register windows'', so
3958 that the register in which a function sees an arguments is not
3959 necessarily the same as the one in which the caller passed the
3962 For such machines, @code{FUNCTION_ARG} computes the register in which
3963 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3964 be defined in a similar fashion to tell the function being called
3965 where the arguments will arrive.
3967 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3968 serves both purposes.
3971 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3972 This target hook returns the number of bytes at the beginning of an
3973 argument that must be put in registers. The value must be zero for
3974 arguments that are passed entirely in registers or that are entirely
3975 pushed on the stack.
3977 On some machines, certain arguments must be passed partially in
3978 registers and partially in memory. On these machines, typically the
3979 first few words of arguments are passed in registers, and the rest
3980 on the stack. If a multi-word argument (a @code{double} or a
3981 structure) crosses that boundary, its first few words must be passed
3982 in registers and the rest must be pushed. This macro tells the
3983 compiler when this occurs, and how many bytes should go in registers.
3985 @code{FUNCTION_ARG} for these arguments should return the first
3986 register to be used by the caller for this argument; likewise
3987 @code{FUNCTION_INCOMING_ARG}, for the called function.
3990 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3991 This target hook should return @code{true} if an argument at the
3992 position indicated by @var{cum} should be passed by reference. This
3993 predicate is queried after target independent reasons for being
3994 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
3996 If the hook returns true, a copy of that argument is made in memory and a
3997 pointer to the argument is passed instead of the argument itself.
3998 The pointer is passed in whatever way is appropriate for passing a pointer
4002 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
4003 The function argument described by the parameters to this hook is
4004 known to be passed by reference. The hook should return true if the
4005 function argument should be copied by the callee instead of copied
4008 For any argument for which the hook returns true, if it can be
4009 determined that the argument is not modified, then a copy need
4012 The default version of this hook always returns false.
4015 @defmac CUMULATIVE_ARGS
4016 A C type for declaring a variable that is used as the first argument of
4017 @code{FUNCTION_ARG} and other related values. For some target machines,
4018 the type @code{int} suffices and can hold the number of bytes of
4021 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
4022 arguments that have been passed on the stack. The compiler has other
4023 variables to keep track of that. For target machines on which all
4024 arguments are passed on the stack, there is no need to store anything in
4025 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
4026 should not be empty, so use @code{int}.
4029 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
4030 A C statement (sans semicolon) for initializing the variable
4031 @var{cum} for the state at the beginning of the argument list. The
4032 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
4033 is the tree node for the data type of the function which will receive
4034 the args, or 0 if the args are to a compiler support library function.
4035 For direct calls that are not libcalls, @var{fndecl} contain the
4036 declaration node of the function. @var{fndecl} is also set when
4037 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
4038 being compiled. @var{n_named_args} is set to the number of named
4039 arguments, including a structure return address if it is passed as a
4040 parameter, when making a call. When processing incoming arguments,
4041 @var{n_named_args} is set to @minus{}1.
4043 When processing a call to a compiler support library function,
4044 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
4045 contains the name of the function, as a string. @var{libname} is 0 when
4046 an ordinary C function call is being processed. Thus, each time this
4047 macro is called, either @var{libname} or @var{fntype} is nonzero, but
4048 never both of them at once.
4051 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
4052 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
4053 it gets a @code{MODE} argument instead of @var{fntype}, that would be
4054 @code{NULL}. @var{indirect} would always be zero, too. If this macro
4055 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
4056 0)} is used instead.
4059 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
4060 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
4061 finding the arguments for the function being compiled. If this macro is
4062 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
4064 The value passed for @var{libname} is always 0, since library routines
4065 with special calling conventions are never compiled with GCC@. The
4066 argument @var{libname} exists for symmetry with
4067 @code{INIT_CUMULATIVE_ARGS}.
4068 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
4069 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
4072 @defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
4073 A C statement (sans semicolon) to update the summarizer variable
4074 @var{cum} to advance past an argument in the argument list. The
4075 values @var{mode}, @var{type} and @var{named} describe that argument.
4076 Once this is done, the variable @var{cum} is suitable for analyzing
4077 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
4079 This macro need not do anything if the argument in question was passed
4080 on the stack. The compiler knows how to track the amount of stack space
4081 used for arguments without any special help.
4084 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
4085 If defined, a C expression which determines whether, and in which direction,
4086 to pad out an argument with extra space. The value should be of type
4087 @code{enum direction}: either @code{upward} to pad above the argument,
4088 @code{downward} to pad below, or @code{none} to inhibit padding.
4090 The @emph{amount} of padding is always just enough to reach the next
4091 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
4094 This macro has a default definition which is right for most systems.
4095 For little-endian machines, the default is to pad upward. For
4096 big-endian machines, the default is to pad downward for an argument of
4097 constant size shorter than an @code{int}, and upward otherwise.
4100 @defmac PAD_VARARGS_DOWN
4101 If defined, a C expression which determines whether the default
4102 implementation of va_arg will attempt to pad down before reading the
4103 next argument, if that argument is smaller than its aligned space as
4104 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4105 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4108 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4109 Specify padding for the last element of a block move between registers and
4110 memory. @var{first} is nonzero if this is the only element. Defining this
4111 macro allows better control of register function parameters on big-endian
4112 machines, without using @code{PARALLEL} rtl. In particular,
4113 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4114 registers, as there is no longer a "wrong" part of a register; For example,
4115 a three byte aggregate may be passed in the high part of a register if so
4119 @defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
4120 If defined, a C expression that gives the alignment boundary, in bits,
4121 of an argument with the specified mode and type. If it is not defined,
4122 @code{PARM_BOUNDARY} is used for all arguments.
4125 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4126 A C expression that is nonzero if @var{regno} is the number of a hard
4127 register in which function arguments are sometimes passed. This does
4128 @emph{not} include implicit arguments such as the static chain and
4129 the structure-value address. On many machines, no registers can be
4130 used for this purpose since all function arguments are pushed on the
4134 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (tree @var{type})
4135 This hook should return true if parameter of type @var{type} are passed
4136 as two scalar parameters. By default, GCC will attempt to pack complex
4137 arguments into the target's word size. Some ABIs require complex arguments
4138 to be split and treated as their individual components. For example, on
4139 AIX64, complex floats should be passed in a pair of floating point
4140 registers, even though a complex float would fit in one 64-bit floating
4143 The default value of this hook is @code{NULL}, which is treated as always
4147 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
4148 This hook returns a type node for @code{va_list} for the target.
4149 The default version of the hook returns @code{void*}.
4152 @deftypefn {Target Hook} tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree @var{valist}, tree @var{type}, tree *@var{pre_p}, tree *@var{post_p})
4153 This hook performs target-specific gimplification of
4154 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4155 arguments to @code{va_arg}; the latter two are as in
4156 @code{gimplify.c:gimplify_expr}.
4159 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (enum machine_mode @var{mode})
4160 Define this to return nonzero if the port can handle pointers
4161 with machine mode @var{mode}. The default version of this
4162 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4165 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4166 Define this to return nonzero if the port is prepared to handle
4167 insns involving scalar mode @var{mode}. For a scalar mode to be
4168 considered supported, all the basic arithmetic and comparisons
4171 The default version of this hook returns true for any mode
4172 required to handle the basic C types (as defined by the port).
4173 Included here are the double-word arithmetic supported by the
4174 code in @file{optabs.c}.
4177 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4178 Define this to return nonzero if the port is prepared to handle
4179 insns involving vector mode @var{mode}. At the very least, it
4180 must have move patterns for this mode.
4184 @subsection How Scalar Function Values Are Returned
4185 @cindex return values in registers
4186 @cindex values, returned by functions
4187 @cindex scalars, returned as values
4189 This section discusses the macros that control returning scalars as
4190 values---values that can fit in registers.
4192 @deftypefn {Target Hook} rtx TARGET_FUNCTION_VALUE (tree @var{ret_type}, tree @var{fn_decl_or_type}, bool @var{outgoing})
4194 Define this to return an RTX representing the place where a function
4195 returns or receives a value of data type @var{ret_type}, a tree node
4196 node representing a data type. @var{fn_decl_or_type} is a tree node
4197 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4198 function being called. If @var{outgoing} is false, the hook should
4199 compute the register in which the caller will see the return value.
4200 Otherwise, the hook should return an RTX representing the place where
4201 a function returns a value.
4203 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4204 (Actually, on most machines, scalar values are returned in the same
4205 place regardless of mode.) The value of the expression is usually a
4206 @code{reg} RTX for the hard register where the return value is stored.
4207 The value can also be a @code{parallel} RTX, if the return value is in
4208 multiple places. See @code{FUNCTION_ARG} for an explanation of the
4209 @code{parallel} form. Note that the callee will populate every
4210 location specified in the @code{parallel}, but if the first element of
4211 the @code{parallel} contains the whole return value, callers will use
4212 that element as the canonical location and ignore the others. The m68k
4213 port uses this type of @code{parallel} to return pointers in both
4214 @samp{%a0} (the canonical location) and @samp{%d0}.
4216 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4217 the same promotion rules specified in @code{PROMOTE_MODE} if
4218 @var{valtype} is a scalar type.
4220 If the precise function being called is known, @var{func} is a tree
4221 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4222 pointer. This makes it possible to use a different value-returning
4223 convention for specific functions when all their calls are
4226 Some target machines have ``register windows'' so that the register in
4227 which a function returns its value is not the same as the one in which
4228 the caller sees the value. For such machines, you should return
4229 different RTX depending on @var{outgoing}.
4231 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4232 aggregate data types, because these are returned in another way. See
4233 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4236 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4237 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4238 a new target instead.
4241 @defmac FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
4242 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4243 a new target instead.
4246 @defmac LIBCALL_VALUE (@var{mode})
4247 A C expression to create an RTX representing the place where a library
4248 function returns a value of mode @var{mode}. If the precise function
4249 being called is known, @var{func} is a tree node
4250 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4251 pointer. This makes it possible to use a different value-returning
4252 convention for specific functions when all their calls are
4255 Note that ``library function'' in this context means a compiler
4256 support routine, used to perform arithmetic, whose name is known
4257 specially by the compiler and was not mentioned in the C code being
4260 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
4261 data types, because none of the library functions returns such types.
4264 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4265 A C expression that is nonzero if @var{regno} is the number of a hard
4266 register in which the values of called function may come back.
4268 A register whose use for returning values is limited to serving as the
4269 second of a pair (for a value of type @code{double}, say) need not be
4270 recognized by this macro. So for most machines, this definition
4274 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4277 If the machine has register windows, so that the caller and the called
4278 function use different registers for the return value, this macro
4279 should recognize only the caller's register numbers.
4282 @defmac APPLY_RESULT_SIZE
4283 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4284 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4285 saving and restoring an arbitrary return value.
4288 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (tree @var{type})
4289 This hook should return true if values of type @var{type} are returned
4290 at the most significant end of a register (in other words, if they are
4291 padded at the least significant end). You can assume that @var{type}
4292 is returned in a register; the caller is required to check this.
4294 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4295 be able to hold the complete return value. For example, if a 1-, 2-
4296 or 3-byte structure is returned at the most significant end of a
4297 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4301 @node Aggregate Return
4302 @subsection How Large Values Are Returned
4303 @cindex aggregates as return values
4304 @cindex large return values
4305 @cindex returning aggregate values
4306 @cindex structure value address
4308 When a function value's mode is @code{BLKmode} (and in some other
4309 cases), the value is not returned according to
4310 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4311 caller passes the address of a block of memory in which the value
4312 should be stored. This address is called the @dfn{structure value
4315 This section describes how to control returning structure values in
4318 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (tree @var{type}, tree @var{fntype})
4319 This target hook should return a nonzero value to say to return the
4320 function value in memory, just as large structures are always returned.
4321 Here @var{type} will be the data type of the value, and @var{fntype}
4322 will be the type of the function doing the returning, or @code{NULL} for
4325 Note that values of mode @code{BLKmode} must be explicitly handled
4326 by this function. Also, the option @option{-fpcc-struct-return}
4327 takes effect regardless of this macro. On most systems, it is
4328 possible to leave the hook undefined; this causes a default
4329 definition to be used, whose value is the constant 1 for @code{BLKmode}
4330 values, and 0 otherwise.
4332 Do not use this hook to indicate that structures and unions should always
4333 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4337 @defmac DEFAULT_PCC_STRUCT_RETURN
4338 Define this macro to be 1 if all structure and union return values must be
4339 in memory. Since this results in slower code, this should be defined
4340 only if needed for compatibility with other compilers or with an ABI@.
4341 If you define this macro to be 0, then the conventions used for structure
4342 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4345 If not defined, this defaults to the value 1.
4348 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4349 This target hook should return the location of the structure value
4350 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4351 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4352 be @code{NULL}, for libcalls. You do not need to define this target
4353 hook if the address is always passed as an ``invisible'' first
4356 On some architectures the place where the structure value address
4357 is found by the called function is not the same place that the
4358 caller put it. This can be due to register windows, or it could
4359 be because the function prologue moves it to a different place.
4360 @var{incoming} is @code{1} or @code{2} when the location is needed in
4361 the context of the called function, and @code{0} in the context of
4364 If @var{incoming} is nonzero and the address is to be found on the
4365 stack, return a @code{mem} which refers to the frame pointer. If
4366 @var{incoming} is @code{2}, the result is being used to fetch the
4367 structure value address at the beginning of a function. If you need
4368 to emit adjusting code, you should do it at this point.
4371 @defmac PCC_STATIC_STRUCT_RETURN
4372 Define this macro if the usual system convention on the target machine
4373 for returning structures and unions is for the called function to return
4374 the address of a static variable containing the value.
4376 Do not define this if the usual system convention is for the caller to
4377 pass an address to the subroutine.
4379 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4380 nothing when you use @option{-freg-struct-return} mode.
4384 @subsection Caller-Saves Register Allocation
4386 If you enable it, GCC can save registers around function calls. This
4387 makes it possible to use call-clobbered registers to hold variables that
4388 must live across calls.
4390 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4391 A C expression to determine whether it is worthwhile to consider placing
4392 a pseudo-register in a call-clobbered hard register and saving and
4393 restoring it around each function call. The expression should be 1 when
4394 this is worth doing, and 0 otherwise.
4396 If you don't define this macro, a default is used which is good on most
4397 machines: @code{4 * @var{calls} < @var{refs}}.
4400 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4401 A C expression specifying which mode is required for saving @var{nregs}
4402 of a pseudo-register in call-clobbered hard register @var{regno}. If
4403 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4404 returned. For most machines this macro need not be defined since GCC
4405 will select the smallest suitable mode.
4408 @node Function Entry
4409 @subsection Function Entry and Exit
4410 @cindex function entry and exit
4414 This section describes the macros that output function entry
4415 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4417 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4418 If defined, a function that outputs the assembler code for entry to a
4419 function. The prologue is responsible for setting up the stack frame,
4420 initializing the frame pointer register, saving registers that must be
4421 saved, and allocating @var{size} additional bytes of storage for the
4422 local variables. @var{size} is an integer. @var{file} is a stdio
4423 stream to which the assembler code should be output.
4425 The label for the beginning of the function need not be output by this
4426 macro. That has already been done when the macro is run.
4428 @findex regs_ever_live
4429 To determine which registers to save, the macro can refer to the array
4430 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4431 @var{r} is used anywhere within the function. This implies the function
4432 prologue should save register @var{r}, provided it is not one of the
4433 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4434 @code{regs_ever_live}.)
4436 On machines that have ``register windows'', the function entry code does
4437 not save on the stack the registers that are in the windows, even if
4438 they are supposed to be preserved by function calls; instead it takes
4439 appropriate steps to ``push'' the register stack, if any non-call-used
4440 registers are used in the function.
4442 @findex frame_pointer_needed
4443 On machines where functions may or may not have frame-pointers, the
4444 function entry code must vary accordingly; it must set up the frame
4445 pointer if one is wanted, and not otherwise. To determine whether a
4446 frame pointer is in wanted, the macro can refer to the variable
4447 @code{frame_pointer_needed}. The variable's value will be 1 at run
4448 time in a function that needs a frame pointer. @xref{Elimination}.
4450 The function entry code is responsible for allocating any stack space
4451 required for the function. This stack space consists of the regions
4452 listed below. In most cases, these regions are allocated in the
4453 order listed, with the last listed region closest to the top of the
4454 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4455 the highest address if it is not defined). You can use a different order
4456 for a machine if doing so is more convenient or required for
4457 compatibility reasons. Except in cases where required by standard
4458 or by a debugger, there is no reason why the stack layout used by GCC
4459 need agree with that used by other compilers for a machine.
4462 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4463 If defined, a function that outputs assembler code at the end of a
4464 prologue. This should be used when the function prologue is being
4465 emitted as RTL, and you have some extra assembler that needs to be
4466 emitted. @xref{prologue instruction pattern}.
4469 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4470 If defined, a function that outputs assembler code at the start of an
4471 epilogue. This should be used when the function epilogue is being
4472 emitted as RTL, and you have some extra assembler that needs to be
4473 emitted. @xref{epilogue instruction pattern}.
4476 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4477 If defined, a function that outputs the assembler code for exit from a
4478 function. The epilogue is responsible for restoring the saved
4479 registers and stack pointer to their values when the function was
4480 called, and returning control to the caller. This macro takes the
4481 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4482 registers to restore are determined from @code{regs_ever_live} and
4483 @code{CALL_USED_REGISTERS} in the same way.
4485 On some machines, there is a single instruction that does all the work
4486 of returning from the function. On these machines, give that
4487 instruction the name @samp{return} and do not define the macro
4488 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4490 Do not define a pattern named @samp{return} if you want the
4491 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4492 switches to control whether return instructions or epilogues are used,
4493 define a @samp{return} pattern with a validity condition that tests the
4494 target switches appropriately. If the @samp{return} pattern's validity
4495 condition is false, epilogues will be used.
4497 On machines where functions may or may not have frame-pointers, the
4498 function exit code must vary accordingly. Sometimes the code for these
4499 two cases is completely different. To determine whether a frame pointer
4500 is wanted, the macro can refer to the variable
4501 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4502 a function that needs a frame pointer.
4504 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4505 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4506 The C variable @code{current_function_is_leaf} is nonzero for such a
4507 function. @xref{Leaf Functions}.
4509 On some machines, some functions pop their arguments on exit while
4510 others leave that for the caller to do. For example, the 68020 when
4511 given @option{-mrtd} pops arguments in functions that take a fixed
4512 number of arguments.
4514 @findex current_function_pops_args
4515 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4516 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4517 needs to know what was decided. The variable that is called
4518 @code{current_function_pops_args} is the number of bytes of its
4519 arguments that a function should pop. @xref{Scalar Return}.
4520 @c what is the "its arguments" in the above sentence referring to, pray
4521 @c tell? --mew 5feb93
4526 @findex current_function_pretend_args_size
4527 A region of @code{current_function_pretend_args_size} bytes of
4528 uninitialized space just underneath the first argument arriving on the
4529 stack. (This may not be at the very start of the allocated stack region
4530 if the calling sequence has pushed anything else since pushing the stack
4531 arguments. But usually, on such machines, nothing else has been pushed
4532 yet, because the function prologue itself does all the pushing.) This
4533 region is used on machines where an argument may be passed partly in
4534 registers and partly in memory, and, in some cases to support the
4535 features in @code{<stdarg.h>}.
4538 An area of memory used to save certain registers used by the function.
4539 The size of this area, which may also include space for such things as
4540 the return address and pointers to previous stack frames, is
4541 machine-specific and usually depends on which registers have been used
4542 in the function. Machines with register windows often do not require
4546 A region of at least @var{size} bytes, possibly rounded up to an allocation
4547 boundary, to contain the local variables of the function. On some machines,
4548 this region and the save area may occur in the opposite order, with the
4549 save area closer to the top of the stack.
4552 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4553 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4554 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4555 argument lists of the function. @xref{Stack Arguments}.
4558 @defmac EXIT_IGNORE_STACK
4559 Define this macro as a C expression that is nonzero if the return
4560 instruction or the function epilogue ignores the value of the stack
4561 pointer; in other words, if it is safe to delete an instruction to
4562 adjust the stack pointer before a return from the function. The
4565 Note that this macro's value is relevant only for functions for which
4566 frame pointers are maintained. It is never safe to delete a final
4567 stack adjustment in a function that has no frame pointer, and the
4568 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4571 @defmac EPILOGUE_USES (@var{regno})
4572 Define this macro as a C expression that is nonzero for registers that are
4573 used by the epilogue or the @samp{return} pattern. The stack and frame
4574 pointer registers are already assumed to be used as needed.
4577 @defmac EH_USES (@var{regno})
4578 Define this macro as a C expression that is nonzero for registers that are
4579 used by the exception handling mechanism, and so should be considered live
4580 on entry to an exception edge.
4583 @defmac DELAY_SLOTS_FOR_EPILOGUE
4584 Define this macro if the function epilogue contains delay slots to which
4585 instructions from the rest of the function can be ``moved''. The
4586 definition should be a C expression whose value is an integer
4587 representing the number of delay slots there.
4590 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4591 A C expression that returns 1 if @var{insn} can be placed in delay
4592 slot number @var{n} of the epilogue.
4594 The argument @var{n} is an integer which identifies the delay slot now
4595 being considered (since different slots may have different rules of
4596 eligibility). It is never negative and is always less than the number
4597 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4598 If you reject a particular insn for a given delay slot, in principle, it
4599 may be reconsidered for a subsequent delay slot. Also, other insns may
4600 (at least in principle) be considered for the so far unfilled delay
4603 @findex current_function_epilogue_delay_list
4604 @findex final_scan_insn
4605 The insns accepted to fill the epilogue delay slots are put in an RTL
4606 list made with @code{insn_list} objects, stored in the variable
4607 @code{current_function_epilogue_delay_list}. The insn for the first
4608 delay slot comes first in the list. Your definition of the macro
4609 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4610 outputting the insns in this list, usually by calling
4611 @code{final_scan_insn}.
4613 You need not define this macro if you did not define
4614 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4617 @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})
4618 A function that outputs the assembler code for a thunk
4619 function, used to implement C++ virtual function calls with multiple
4620 inheritance. The thunk acts as a wrapper around a virtual function,
4621 adjusting the implicit object parameter before handing control off to
4624 First, emit code to add the integer @var{delta} to the location that
4625 contains the incoming first argument. Assume that this argument
4626 contains a pointer, and is the one used to pass the @code{this} pointer
4627 in C++. This is the incoming argument @emph{before} the function prologue,
4628 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4629 all other incoming arguments.
4631 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4632 made after adding @code{delta}. In particular, if @var{p} is the
4633 adjusted pointer, the following adjustment should be made:
4636 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4639 After the additions, emit code to jump to @var{function}, which is a
4640 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4641 not touch the return address. Hence returning from @var{FUNCTION} will
4642 return to whoever called the current @samp{thunk}.
4644 The effect must be as if @var{function} had been called directly with
4645 the adjusted first argument. This macro is responsible for emitting all
4646 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4647 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4649 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4650 have already been extracted from it.) It might possibly be useful on
4651 some targets, but probably not.
4653 If you do not define this macro, the target-independent code in the C++
4654 front end will generate a less efficient heavyweight thunk that calls
4655 @var{function} instead of jumping to it. The generic approach does
4656 not support varargs.
4659 @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})
4660 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4661 to output the assembler code for the thunk function specified by the
4662 arguments it is passed, and false otherwise. In the latter case, the
4663 generic approach will be used by the C++ front end, with the limitations
4668 @subsection Generating Code for Profiling
4669 @cindex profiling, code generation
4671 These macros will help you generate code for profiling.
4673 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4674 A C statement or compound statement to output to @var{file} some
4675 assembler code to call the profiling subroutine @code{mcount}.
4678 The details of how @code{mcount} expects to be called are determined by
4679 your operating system environment, not by GCC@. To figure them out,
4680 compile a small program for profiling using the system's installed C
4681 compiler and look at the assembler code that results.
4683 Older implementations of @code{mcount} expect the address of a counter
4684 variable to be loaded into some register. The name of this variable is
4685 @samp{LP} followed by the number @var{labelno}, so you would generate
4686 the name using @samp{LP%d} in a @code{fprintf}.
4689 @defmac PROFILE_HOOK
4690 A C statement or compound statement to output to @var{file} some assembly
4691 code to call the profiling subroutine @code{mcount} even the target does
4692 not support profiling.
4695 @defmac NO_PROFILE_COUNTERS
4696 Define this macro to be an expression with a nonzero value if the
4697 @code{mcount} subroutine on your system does not need a counter variable
4698 allocated for each function. This is true for almost all modern
4699 implementations. If you define this macro, you must not use the
4700 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4703 @defmac PROFILE_BEFORE_PROLOGUE
4704 Define this macro if the code for function profiling should come before
4705 the function prologue. Normally, the profiling code comes after.
4709 @subsection Permitting tail calls
4712 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4713 True if it is ok to do sibling call optimization for the specified
4714 call expression @var{exp}. @var{decl} will be the called function,
4715 or @code{NULL} if this is an indirect call.
4717 It is not uncommon for limitations of calling conventions to prevent
4718 tail calls to functions outside the current unit of translation, or
4719 during PIC compilation. The hook is used to enforce these restrictions,
4720 as the @code{sibcall} md pattern can not fail, or fall over to a
4721 ``normal'' call. The criteria for successful sibling call optimization
4722 may vary greatly between different architectures.
4725 @deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap *@var{regs})
4726 Add any hard registers to @var{regs} that are live on entry to the
4727 function. This hook only needs to be defined to provide registers that
4728 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4729 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4730 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4731 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4734 @node Stack Smashing Protection
4735 @subsection Stack smashing protection
4736 @cindex stack smashing protection
4738 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void)
4739 This hook returns a @code{DECL} node for the external variable to use
4740 for the stack protection guard. This variable is initialized by the
4741 runtime to some random value and is used to initialize the guard value
4742 that is placed at the top of the local stack frame. The type of this
4743 variable must be @code{ptr_type_node}.
4745 The default version of this hook creates a variable called
4746 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
4749 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void)
4750 This hook returns a tree expression that alerts the runtime that the
4751 stack protect guard variable has been modified. This expression should
4752 involve a call to a @code{noreturn} function.
4754 The default version of this hook invokes a function called
4755 @samp{__stack_chk_fail}, taking no arguments. This function is
4756 normally defined in @file{libgcc2.c}.
4760 @section Implementing the Varargs Macros
4761 @cindex varargs implementation
4763 GCC comes with an implementation of @code{<varargs.h>} and
4764 @code{<stdarg.h>} that work without change on machines that pass arguments
4765 on the stack. Other machines require their own implementations of
4766 varargs, and the two machine independent header files must have
4767 conditionals to include it.
4769 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4770 the calling convention for @code{va_start}. The traditional
4771 implementation takes just one argument, which is the variable in which
4772 to store the argument pointer. The ISO implementation of
4773 @code{va_start} takes an additional second argument. The user is
4774 supposed to write the last named argument of the function here.
4776 However, @code{va_start} should not use this argument. The way to find
4777 the end of the named arguments is with the built-in functions described
4780 @defmac __builtin_saveregs ()
4781 Use this built-in function to save the argument registers in memory so
4782 that the varargs mechanism can access them. Both ISO and traditional
4783 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4784 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4786 On some machines, @code{__builtin_saveregs} is open-coded under the
4787 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4788 other machines, it calls a routine written in assembler language,
4789 found in @file{libgcc2.c}.
4791 Code generated for the call to @code{__builtin_saveregs} appears at the
4792 beginning of the function, as opposed to where the call to
4793 @code{__builtin_saveregs} is written, regardless of what the code is.
4794 This is because the registers must be saved before the function starts
4795 to use them for its own purposes.
4796 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4800 @defmac __builtin_args_info (@var{category})
4801 Use this built-in function to find the first anonymous arguments in
4804 In general, a machine may have several categories of registers used for
4805 arguments, each for a particular category of data types. (For example,
4806 on some machines, floating-point registers are used for floating-point
4807 arguments while other arguments are passed in the general registers.)
4808 To make non-varargs functions use the proper calling convention, you
4809 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4810 registers in each category have been used so far
4812 @code{__builtin_args_info} accesses the same data structure of type
4813 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4814 with it, with @var{category} specifying which word to access. Thus, the
4815 value indicates the first unused register in a given category.
4817 Normally, you would use @code{__builtin_args_info} in the implementation
4818 of @code{va_start}, accessing each category just once and storing the
4819 value in the @code{va_list} object. This is because @code{va_list} will
4820 have to update the values, and there is no way to alter the
4821 values accessed by @code{__builtin_args_info}.
4824 @defmac __builtin_next_arg (@var{lastarg})
4825 This is the equivalent of @code{__builtin_args_info}, for stack
4826 arguments. It returns the address of the first anonymous stack
4827 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4828 returns the address of the location above the first anonymous stack
4829 argument. Use it in @code{va_start} to initialize the pointer for
4830 fetching arguments from the stack. Also use it in @code{va_start} to
4831 verify that the second parameter @var{lastarg} is the last named argument
4832 of the current function.
4835 @defmac __builtin_classify_type (@var{object})
4836 Since each machine has its own conventions for which data types are
4837 passed in which kind of register, your implementation of @code{va_arg}
4838 has to embody these conventions. The easiest way to categorize the
4839 specified data type is to use @code{__builtin_classify_type} together
4840 with @code{sizeof} and @code{__alignof__}.
4842 @code{__builtin_classify_type} ignores the value of @var{object},
4843 considering only its data type. It returns an integer describing what
4844 kind of type that is---integer, floating, pointer, structure, and so on.
4846 The file @file{typeclass.h} defines an enumeration that you can use to
4847 interpret the values of @code{__builtin_classify_type}.
4850 These machine description macros help implement varargs:
4852 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
4853 If defined, this hook produces the machine-specific code for a call to
4854 @code{__builtin_saveregs}. This code will be moved to the very
4855 beginning of the function, before any parameter access are made. The
4856 return value of this function should be an RTX that contains the value
4857 to use as the return of @code{__builtin_saveregs}.
4860 @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})
4861 This target hook offers an alternative to using
4862 @code{__builtin_saveregs} and defining the hook
4863 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
4864 register arguments into the stack so that all the arguments appear to
4865 have been passed consecutively on the stack. Once this is done, you can
4866 use the standard implementation of varargs that works for machines that
4867 pass all their arguments on the stack.
4869 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
4870 structure, containing the values that are obtained after processing the
4871 named arguments. The arguments @var{mode} and @var{type} describe the
4872 last named argument---its machine mode and its data type as a tree node.
4874 The target hook should do two things: first, push onto the stack all the
4875 argument registers @emph{not} used for the named arguments, and second,
4876 store the size of the data thus pushed into the @code{int}-valued
4877 variable pointed to by @var{pretend_args_size}. The value that you
4878 store here will serve as additional offset for setting up the stack
4881 Because you must generate code to push the anonymous arguments at
4882 compile time without knowing their data types,
4883 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
4884 have just a single category of argument register and use it uniformly
4887 If the argument @var{second_time} is nonzero, it means that the
4888 arguments of the function are being analyzed for the second time. This
4889 happens for an inline function, which is not actually compiled until the
4890 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
4891 not generate any instructions in this case.
4894 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca})
4895 Define this hook to return @code{true} if the location where a function
4896 argument is passed depends on whether or not it is a named argument.
4898 This hook controls how the @var{named} argument to @code{FUNCTION_ARG}
4899 is set for varargs and stdarg functions. If this hook returns
4900 @code{true}, the @var{named} argument is always true for named
4901 arguments, and false for unnamed arguments. If it returns @code{false},
4902 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
4903 then all arguments are treated as named. Otherwise, all named arguments
4904 except the last are treated as named.
4906 You need not define this hook if it always returns zero.
4909 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED
4910 If you need to conditionally change ABIs so that one works with
4911 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
4912 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
4913 defined, then define this hook to return @code{true} if
4914 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
4915 Otherwise, you should not define this hook.
4919 @section Trampolines for Nested Functions
4920 @cindex trampolines for nested functions
4921 @cindex nested functions, trampolines for
4923 A @dfn{trampoline} is a small piece of code that is created at run time
4924 when the address of a nested function is taken. It normally resides on
4925 the stack, in the stack frame of the containing function. These macros
4926 tell GCC how to generate code to allocate and initialize a
4929 The instructions in the trampoline must do two things: load a constant
4930 address into the static chain register, and jump to the real address of
4931 the nested function. On CISC machines such as the m68k, this requires
4932 two instructions, a move immediate and a jump. Then the two addresses
4933 exist in the trampoline as word-long immediate operands. On RISC
4934 machines, it is often necessary to load each address into a register in
4935 two parts. Then pieces of each address form separate immediate
4938 The code generated to initialize the trampoline must store the variable
4939 parts---the static chain value and the function address---into the
4940 immediate operands of the instructions. On a CISC machine, this is
4941 simply a matter of copying each address to a memory reference at the
4942 proper offset from the start of the trampoline. On a RISC machine, it
4943 may be necessary to take out pieces of the address and store them
4946 @defmac TRAMPOLINE_TEMPLATE (@var{file})
4947 A C statement to output, on the stream @var{file}, assembler code for a
4948 block of data that contains the constant parts of a trampoline. This
4949 code should not include a label---the label is taken care of
4952 If you do not define this macro, it means no template is needed
4953 for the target. Do not define this macro on systems where the block move
4954 code to copy the trampoline into place would be larger than the code
4955 to generate it on the spot.
4958 @defmac TRAMPOLINE_SECTION
4959 Return the section into which the trampoline template is to be placed
4960 (@pxref{Sections}). The default value is @code{readonly_data_section}.
4963 @defmac TRAMPOLINE_SIZE
4964 A C expression for the size in bytes of the trampoline, as an integer.
4967 @defmac TRAMPOLINE_ALIGNMENT
4968 Alignment required for trampolines, in bits.
4970 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4971 is used for aligning trampolines.
4974 @defmac INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4975 A C statement to initialize the variable parts of a trampoline.
4976 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4977 an RTX for the address of the nested function; @var{static_chain} is an
4978 RTX for the static chain value that should be passed to the function
4982 @defmac TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4983 A C statement that should perform any machine-specific adjustment in
4984 the address of the trampoline. Its argument contains the address that
4985 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4986 used for a function call should be different from the address in which
4987 the template was stored, the different address should be assigned to
4988 @var{addr}. If this macro is not defined, @var{addr} will be used for
4991 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4992 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4993 If this macro is not defined, by default the trampoline is allocated as
4994 a stack slot. This default is right for most machines. The exceptions
4995 are machines where it is impossible to execute instructions in the stack
4996 area. On such machines, you may have to implement a separate stack,
4997 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4998 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
5000 @var{fp} points to a data structure, a @code{struct function}, which
5001 describes the compilation status of the immediate containing function of
5002 the function which the trampoline is for. The stack slot for the
5003 trampoline is in the stack frame of this containing function. Other
5004 allocation strategies probably must do something analogous with this
5008 Implementing trampolines is difficult on many machines because they have
5009 separate instruction and data caches. Writing into a stack location
5010 fails to clear the memory in the instruction cache, so when the program
5011 jumps to that location, it executes the old contents.
5013 Here are two possible solutions. One is to clear the relevant parts of
5014 the instruction cache whenever a trampoline is set up. The other is to
5015 make all trampolines identical, by having them jump to a standard
5016 subroutine. The former technique makes trampoline execution faster; the
5017 latter makes initialization faster.
5019 To clear the instruction cache when a trampoline is initialized, define
5020 the following macro.
5022 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
5023 If defined, expands to a C expression clearing the @emph{instruction
5024 cache} in the specified interval. The definition of this macro would
5025 typically be a series of @code{asm} statements. Both @var{beg} and
5026 @var{end} are both pointer expressions.
5029 The operating system may also require the stack to be made executable
5030 before calling the trampoline. To implement this requirement, define
5031 the following macro.
5033 @defmac ENABLE_EXECUTE_STACK
5034 Define this macro if certain operations must be performed before executing
5035 code located on the stack. The macro should expand to a series of C
5036 file-scope constructs (e.g.@: functions) and provide a unique entry point
5037 named @code{__enable_execute_stack}. The target is responsible for
5038 emitting calls to the entry point in the code, for example from the
5039 @code{INITIALIZE_TRAMPOLINE} macro.
5042 To use a standard subroutine, define the following macro. In addition,
5043 you must make sure that the instructions in a trampoline fill an entire
5044 cache line with identical instructions, or else ensure that the
5045 beginning of the trampoline code is always aligned at the same point in
5046 its cache line. Look in @file{m68k.h} as a guide.
5048 @defmac TRANSFER_FROM_TRAMPOLINE
5049 Define this macro if trampolines need a special subroutine to do their
5050 work. The macro should expand to a series of @code{asm} statements
5051 which will be compiled with GCC@. They go in a library function named
5052 @code{__transfer_from_trampoline}.
5054 If you need to avoid executing the ordinary prologue code of a compiled
5055 C function when you jump to the subroutine, you can do so by placing a
5056 special label of your own in the assembler code. Use one @code{asm}
5057 statement to generate an assembler label, and another to make the label
5058 global. Then trampolines can use that label to jump directly to your
5059 special assembler code.
5063 @section Implicit Calls to Library Routines
5064 @cindex library subroutine names
5065 @cindex @file{libgcc.a}
5067 @c prevent bad page break with this line
5068 Here is an explanation of implicit calls to library routines.
5070 @defmac DECLARE_LIBRARY_RENAMES
5071 This macro, if defined, should expand to a piece of C code that will get
5072 expanded when compiling functions for libgcc.a. It can be used to
5073 provide alternate names for GCC's internal library functions if there
5074 are ABI-mandated names that the compiler should provide.
5077 @findex init_one_libfunc
5078 @findex set_optab_libfunc
5079 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
5080 This hook should declare additional library routines or rename
5081 existing ones, using the functions @code{set_optab_libfunc} and
5082 @code{init_one_libfunc} defined in @file{optabs.c}.
5083 @code{init_optabs} calls this macro after initializing all the normal
5086 The default is to do nothing. Most ports don't need to define this hook.
5089 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
5090 This macro should return @code{true} if the library routine that
5091 implements the floating point comparison operator @var{comparison} in
5092 mode @var{mode} will return a boolean, and @var{false} if it will
5095 GCC's own floating point libraries return tristates from the
5096 comparison operators, so the default returns false always. Most ports
5097 don't need to define this macro.
5100 @defmac TARGET_LIB_INT_CMP_BIASED
5101 This macro should evaluate to @code{true} if the integer comparison
5102 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5103 operand is smaller than the second, 1 to indicate that they are equal,
5104 and 2 to indicate that the first operand is greater than the second.
5105 If this macro evaluates to @code{false} the comparison functions return
5106 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5107 in @file{libgcc.a}, you do not need to define this macro.
5110 @cindex US Software GOFAST, floating point emulation library
5111 @cindex floating point emulation library, US Software GOFAST
5112 @cindex GOFAST, floating point emulation library
5113 @findex gofast_maybe_init_libfuncs
5114 @defmac US_SOFTWARE_GOFAST
5115 Define this macro if your system C library uses the US Software GOFAST
5116 library to provide floating point emulation.
5118 In addition to defining this macro, your architecture must set
5119 @code{TARGET_INIT_LIBFUNCS} to @code{gofast_maybe_init_libfuncs}, or
5120 else call that function from its version of that hook. It is defined
5121 in @file{config/gofast.h}, which must be included by your
5122 architecture's @file{@var{cpu}.c} file. See @file{sparc/sparc.c} for
5125 If this macro is defined, the
5126 @code{TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL} target hook must return
5127 false for @code{SFmode} and @code{DFmode} comparisons.
5130 @cindex @code{EDOM}, implicit usage
5133 The value of @code{EDOM} on the target machine, as a C integer constant
5134 expression. If you don't define this macro, GCC does not attempt to
5135 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5136 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5139 If you do not define @code{TARGET_EDOM}, then compiled code reports
5140 domain errors by calling the library function and letting it report the
5141 error. If mathematical functions on your system use @code{matherr} when
5142 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5143 that @code{matherr} is used normally.
5146 @cindex @code{errno}, implicit usage
5147 @defmac GEN_ERRNO_RTX
5148 Define this macro as a C expression to create an rtl expression that
5149 refers to the global ``variable'' @code{errno}. (On certain systems,
5150 @code{errno} may not actually be a variable.) If you don't define this
5151 macro, a reasonable default is used.
5154 @cindex C99 math functions, implicit usage
5155 @defmac TARGET_C99_FUNCTIONS
5156 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
5157 @code{sinf} and similarly for other functions defined by C99 standard. The
5158 default is nonzero that should be proper value for most modern systems, however
5159 number of existing systems lacks support for these functions in the runtime so
5160 they needs this macro to be redefined to 0.
5163 @cindex sincos math function, implicit usage
5164 @defmac TARGET_HAS_SINCOS
5165 When this macro is nonzero, GCC will implicitly optimize calls to @code{sin}
5166 and @code{cos} with the same argument to a call to @code{sincos}. The
5167 default is zero. The target has to provide the following functions:
5169 void sincos(double x, double *sin, double *cos);
5170 void sincosf(float x, float *sin, float *cos);
5171 void sincosl(long double x, long double *sin, long double *cos);
5175 @defmac NEXT_OBJC_RUNTIME
5176 Define this macro to generate code for Objective-C message sending using
5177 the calling convention of the NeXT system. This calling convention
5178 involves passing the object, the selector and the method arguments all
5179 at once to the method-lookup library function.
5181 The default calling convention passes just the object and the selector
5182 to the lookup function, which returns a pointer to the method.
5185 @node Addressing Modes
5186 @section Addressing Modes
5187 @cindex addressing modes
5189 @c prevent bad page break with this line
5190 This is about addressing modes.
5192 @defmac HAVE_PRE_INCREMENT
5193 @defmacx HAVE_PRE_DECREMENT
5194 @defmacx HAVE_POST_INCREMENT
5195 @defmacx HAVE_POST_DECREMENT
5196 A C expression that is nonzero if the machine supports pre-increment,
5197 pre-decrement, post-increment, or post-decrement addressing respectively.
5200 @defmac HAVE_PRE_MODIFY_DISP
5201 @defmacx HAVE_POST_MODIFY_DISP
5202 A C expression that is nonzero if the machine supports pre- or
5203 post-address side-effect generation involving constants other than
5204 the size of the memory operand.
5207 @defmac HAVE_PRE_MODIFY_REG
5208 @defmacx HAVE_POST_MODIFY_REG
5209 A C expression that is nonzero if the machine supports pre- or
5210 post-address side-effect generation involving a register displacement.
5213 @defmac CONSTANT_ADDRESS_P (@var{x})
5214 A C expression that is 1 if the RTX @var{x} is a constant which
5215 is a valid address. On most machines, this can be defined as
5216 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
5217 in which constant addresses are supported.
5220 @defmac CONSTANT_P (@var{x})
5221 @code{CONSTANT_P}, which is defined by target-independent code,
5222 accepts integer-values expressions whose values are not explicitly
5223 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5224 expressions and @code{const} arithmetic expressions, in addition to
5225 @code{const_int} and @code{const_double} expressions.
5228 @defmac MAX_REGS_PER_ADDRESS
5229 A number, the maximum number of registers that can appear in a valid
5230 memory address. Note that it is up to you to specify a value equal to
5231 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
5235 @defmac GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5236 A C compound statement with a conditional @code{goto @var{label};}
5237 executed if @var{x} (an RTX) is a legitimate memory address on the
5238 target machine for a memory operand of mode @var{mode}.
5240 It usually pays to define several simpler macros to serve as
5241 subroutines for this one. Otherwise it may be too complicated to
5244 This macro must exist in two variants: a strict variant and a
5245 non-strict one. The strict variant is used in the reload pass. It
5246 must be defined so that any pseudo-register that has not been
5247 allocated a hard register is considered a memory reference. In
5248 contexts where some kind of register is required, a pseudo-register
5249 with no hard register must be rejected.
5251 The non-strict variant is used in other passes. It must be defined to
5252 accept all pseudo-registers in every context where some kind of
5253 register is required.
5255 @findex REG_OK_STRICT
5256 Compiler source files that want to use the strict variant of this
5257 macro define the macro @code{REG_OK_STRICT}. You should use an
5258 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
5259 in that case and the non-strict variant otherwise.
5261 Subroutines to check for acceptable registers for various purposes (one
5262 for base registers, one for index registers, and so on) are typically
5263 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
5264 Then only these subroutine macros need have two variants; the higher
5265 levels of macros may be the same whether strict or not.
5267 Normally, constant addresses which are the sum of a @code{symbol_ref}
5268 and an integer are stored inside a @code{const} RTX to mark them as
5269 constant. Therefore, there is no need to recognize such sums
5270 specifically as legitimate addresses. Normally you would simply
5271 recognize any @code{const} as legitimate.
5273 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5274 sums that are not marked with @code{const}. It assumes that a naked
5275 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5276 naked constant sums as illegitimate addresses, so that none of them will
5277 be given to @code{PRINT_OPERAND_ADDRESS}.
5279 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5280 On some machines, whether a symbolic address is legitimate depends on
5281 the section that the address refers to. On these machines, define the
5282 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5283 into the @code{symbol_ref}, and then check for it here. When you see a
5284 @code{const}, you will have to look inside it to find the
5285 @code{symbol_ref} in order to determine the section. @xref{Assembler
5289 @defmac FIND_BASE_TERM (@var{x})
5290 A C expression to determine the base term of address @var{x}.
5291 This macro is used in only one place: `find_base_term' in alias.c.
5293 It is always safe for this macro to not be defined. It exists so
5294 that alias analysis can understand machine-dependent addresses.
5296 The typical use of this macro is to handle addresses containing
5297 a label_ref or symbol_ref within an UNSPEC@.
5300 @defmac LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
5301 A C compound statement that attempts to replace @var{x} with a valid
5302 memory address for an operand of mode @var{mode}. @var{win} will be a
5303 C statement label elsewhere in the code; the macro definition may use
5306 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
5310 to avoid further processing if the address has become legitimate.
5312 @findex break_out_memory_refs
5313 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5314 and @var{oldx} will be the operand that was given to that function to produce
5317 The code generated by this macro should not alter the substructure of
5318 @var{x}. If it transforms @var{x} into a more legitimate form, it
5319 should assign @var{x} (which will always be a C variable) a new value.
5321 It is not necessary for this macro to come up with a legitimate
5322 address. The compiler has standard ways of doing so in all cases. In
5323 fact, it is safe to omit this macro. But often a
5324 machine-dependent strategy can generate better code.
5327 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5328 A C compound statement that attempts to replace @var{x}, which is an address
5329 that needs reloading, with a valid memory address for an operand of mode
5330 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5331 It is not necessary to define this macro, but it might be useful for
5332 performance reasons.
5334 For example, on the i386, it is sometimes possible to use a single
5335 reload register instead of two by reloading a sum of two pseudo
5336 registers into a register. On the other hand, for number of RISC
5337 processors offsets are limited so that often an intermediate address
5338 needs to be generated in order to address a stack slot. By defining
5339 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5340 generated for adjacent some stack slots can be made identical, and thus
5343 @emph{Note}: This macro should be used with caution. It is necessary
5344 to know something of how reload works in order to effectively use this,
5345 and it is quite easy to produce macros that build in too much knowledge
5346 of reload internals.
5348 @emph{Note}: This macro must be able to reload an address created by a
5349 previous invocation of this macro. If it fails to handle such addresses
5350 then the compiler may generate incorrect code or abort.
5353 The macro definition should use @code{push_reload} to indicate parts that
5354 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5355 suitable to be passed unaltered to @code{push_reload}.
5357 The code generated by this macro must not alter the substructure of
5358 @var{x}. If it transforms @var{x} into a more legitimate form, it
5359 should assign @var{x} (which will always be a C variable) a new value.
5360 This also applies to parts that you change indirectly by calling
5363 @findex strict_memory_address_p
5364 The macro definition may use @code{strict_memory_address_p} to test if
5365 the address has become legitimate.
5368 If you want to change only a part of @var{x}, one standard way of doing
5369 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5370 single level of rtl. Thus, if the part to be changed is not at the
5371 top level, you'll need to replace first the top level.
5372 It is not necessary for this macro to come up with a legitimate
5373 address; but often a machine-dependent strategy can generate better code.
5376 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5377 A C statement or compound statement with a conditional @code{goto
5378 @var{label};} executed if memory address @var{x} (an RTX) can have
5379 different meanings depending on the machine mode of the memory
5380 reference it is used for or if the address is valid for some modes
5383 Autoincrement and autodecrement addresses typically have mode-dependent
5384 effects because the amount of the increment or decrement is the size
5385 of the operand being addressed. Some machines have other mode-dependent
5386 addresses. Many RISC machines have no mode-dependent addresses.
5388 You may assume that @var{addr} is a valid address for the machine.
5391 @defmac LEGITIMATE_CONSTANT_P (@var{x})
5392 A C expression that is nonzero if @var{x} is a legitimate constant for
5393 an immediate operand on the target machine. You can assume that
5394 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5395 @samp{1} is a suitable definition for this macro on machines where
5396 anything @code{CONSTANT_P} is valid.
5399 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5400 This hook is used to undo the possibly obfuscating effects of the
5401 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5402 macros. Some backend implementations of these macros wrap symbol
5403 references inside an @code{UNSPEC} rtx to represent PIC or similar
5404 addressing modes. This target hook allows GCC's optimizers to understand
5405 the semantics of these opaque @code{UNSPEC}s by converting them back
5406 into their original form.
5409 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (rtx @var{x})
5410 This hook should return true if @var{x} is of a form that cannot (or
5411 should not) be spilled to the constant pool. The default version of
5412 this hook returns false.
5414 The primary reason to define this hook is to prevent reload from
5415 deciding that a non-legitimate constant would be better reloaded
5416 from the constant pool instead of spilling and reloading a register
5417 holding the constant. This restriction is often true of addresses
5418 of TLS symbols for various targets.
5421 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (enum machine_mode @var{mode}, rtx @var{x})
5422 This hook should return true if pool entries for constant @var{x} can
5423 be placed in an @code{object_block} structure. @var{mode} is the mode
5426 The default version returns false for all constants.
5429 @deftypefn {Target Hook} tree TARGET_BUILTIN_RECIPROCAL (enum tree_code @var{fn}, bool @var{tm_fn}, bool @var{sqrt})
5430 This hook should return the DECL of a function that implements reciprocal of
5431 the builtin function with builtin function code @var{fn}, or
5432 @code{NULL_TREE} if such a function is not available. @var{tm_fn} is true
5433 when @var{fn} is a code of a machine-dependent builtin function. When
5434 @var{sqrt} is true, additional optimizations that apply only to the reciprocal
5435 of a square root function are performed, and only reciprocals of @code{sqrt}
5439 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5440 This hook should return the DECL of a function @var{f} that given an
5441 address @var{addr} as an argument returns a mask @var{m} that can be
5442 used to extract from two vectors the relevant data that resides in
5443 @var{addr} in case @var{addr} is not properly aligned.
5445 The autovectorizer, when vectorizing a load operation from an address
5446 @var{addr} that may be unaligned, will generate two vector loads from
5447 the two aligned addresses around @var{addr}. It then generates a
5448 @code{REALIGN_LOAD} operation to extract the relevant data from the
5449 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5450 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5451 the third argument, @var{OFF}, defines how the data will be extracted
5452 from these two vectors: if @var{OFF} is 0, then the returned vector is
5453 @var{v2}; otherwise, the returned vector is composed from the last
5454 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5455 @var{OFF} elements of @var{v2}.
5457 If this hook is defined, the autovectorizer will generate a call
5458 to @var{f} (using the DECL tree that this hook returns) and will
5459 use the return value of @var{f} as the argument @var{OFF} to
5460 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5461 should comply with the semantics expected by @code{REALIGN_LOAD}
5463 If this hook is not defined, then @var{addr} will be used as
5464 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5465 log2(@var{VS})-1 bits of @var{addr} will be considered.
5468 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN (tree @var{x})
5469 This hook should return the DECL of a function @var{f} that implements
5470 widening multiplication of the even elements of two input vectors of type @var{x}.
5472 If this hook is defined, the autovectorizer will use it along with the
5473 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD} target hook when vectorizing
5474 widening multiplication in cases that the order of the results does not have to be
5475 preserved (e.g. used only by a reduction computation). Otherwise, the
5476 @code{widen_mult_hi/lo} idioms will be used.
5479 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD (tree @var{x})
5480 This hook should return the DECL of a function @var{f} that implements
5481 widening multiplication of the odd elements of two input vectors of type @var{x}.
5483 If this hook is defined, the autovectorizer will use it along with the
5484 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN} target hook when vectorizing
5485 widening multiplication in cases that the order of the results does not have to be
5486 preserved (e.g. used only by a reduction computation). Otherwise, the
5487 @code{widen_mult_hi/lo} idioms will be used.
5490 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_CONVERSION (enum tree_code @var{code}, tree @var{type})
5491 This hook should return the DECL of a function that implements conversion of the
5492 input vector of type @var{type}.
5493 If @var{type} is an integral type, the result of the conversion is a vector of
5494 floating-point type of the same size.
5495 If @var{type} is a floating-point type, the result of the conversion is a vector
5496 of integral type of the same size.
5497 @var{code} specifies how the conversion is to be applied
5498 (truncation, rounding, etc.).
5500 If this hook is defined, the autovectorizer will use the
5501 @code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing
5502 conversion. Otherwise, it will return @code{NULL_TREE}.
5505 @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})
5506 This hook should return the decl of a function that implements the vectorized
5507 variant of the builtin function with builtin function code @var{code} or
5508 @code{NULL_TREE} if such a function is not available. The return type of
5509 the vectorized function shall be of vector type @var{vec_type_out} and the
5510 argument types should be @var{vec_type_in}.
5513 @node Anchored Addresses
5514 @section Anchored Addresses
5515 @cindex anchored addresses
5516 @cindex @option{-fsection-anchors}
5518 GCC usually addresses every static object as a separate entity.
5519 For example, if we have:
5523 int foo (void) @{ return a + b + c; @}
5526 the code for @code{foo} will usually calculate three separate symbolic
5527 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
5528 it would be better to calculate just one symbolic address and access
5529 the three variables relative to it. The equivalent pseudocode would
5535 register int *xr = &x;
5536 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5540 (which isn't valid C). We refer to shared addresses like @code{x} as
5541 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
5543 The hooks below describe the target properties that GCC needs to know
5544 in order to make effective use of section anchors. It won't use
5545 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
5546 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
5548 @deftypevar {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET
5549 The minimum offset that should be applied to a section anchor.
5550 On most targets, it should be the smallest offset that can be
5551 applied to a base register while still giving a legitimate address
5552 for every mode. The default value is 0.
5555 @deftypevar {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET
5556 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
5557 offset that should be applied to section anchors. The default
5561 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x})
5562 Write the assembly code to define section anchor @var{x}, which is a
5563 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
5564 The hook is called with the assembly output position set to the beginning
5565 of @code{SYMBOL_REF_BLOCK (@var{x})}.
5567 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
5568 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
5569 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
5570 is @code{NULL}, which disables the use of section anchors altogether.
5573 @deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (rtx @var{x})
5574 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
5575 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
5576 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
5578 The default version is correct for most targets, but you might need to
5579 intercept this hook to handle things like target-specific attributes
5580 or target-specific sections.
5583 @node Condition Code
5584 @section Condition Code Status
5585 @cindex condition code status
5587 @c prevent bad page break with this line
5588 This describes the condition code status.
5591 The file @file{conditions.h} defines a variable @code{cc_status} to
5592 describe how the condition code was computed (in case the interpretation of
5593 the condition code depends on the instruction that it was set by). This
5594 variable contains the RTL expressions on which the condition code is
5595 currently based, and several standard flags.
5597 Sometimes additional machine-specific flags must be defined in the machine
5598 description header file. It can also add additional machine-specific
5599 information by defining @code{CC_STATUS_MDEP}.
5601 @defmac CC_STATUS_MDEP
5602 C code for a data type which is used for declaring the @code{mdep}
5603 component of @code{cc_status}. It defaults to @code{int}.
5605 This macro is not used on machines that do not use @code{cc0}.
5608 @defmac CC_STATUS_MDEP_INIT
5609 A C expression to initialize the @code{mdep} field to ``empty''.
5610 The default definition does nothing, since most machines don't use
5611 the field anyway. If you want to use the field, you should probably
5612 define this macro to initialize it.
5614 This macro is not used on machines that do not use @code{cc0}.
5617 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5618 A C compound statement to set the components of @code{cc_status}
5619 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5620 this macro's responsibility to recognize insns that set the condition
5621 code as a byproduct of other activity as well as those that explicitly
5624 This macro is not used on machines that do not use @code{cc0}.
5626 If there are insns that do not set the condition code but do alter
5627 other machine registers, this macro must check to see whether they
5628 invalidate the expressions that the condition code is recorded as
5629 reflecting. For example, on the 68000, insns that store in address
5630 registers do not set the condition code, which means that usually
5631 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5632 insns. But suppose that the previous insn set the condition code
5633 based on location @samp{a4@@(102)} and the current insn stores a new
5634 value in @samp{a4}. Although the condition code is not changed by
5635 this, it will no longer be true that it reflects the contents of
5636 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5637 @code{cc_status} in this case to say that nothing is known about the
5638 condition code value.
5640 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5641 with the results of peephole optimization: insns whose patterns are
5642 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5643 constants which are just the operands. The RTL structure of these
5644 insns is not sufficient to indicate what the insns actually do. What
5645 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5646 @code{CC_STATUS_INIT}.
5648 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5649 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5650 @samp{cc}. This avoids having detailed information about patterns in
5651 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5654 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5655 Returns a mode from class @code{MODE_CC} to be used when comparison
5656 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5657 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5658 @pxref{Jump Patterns} for a description of the reason for this
5662 #define SELECT_CC_MODE(OP,X,Y) \
5663 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5664 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5665 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5666 || GET_CODE (X) == NEG) \
5667 ? CC_NOOVmode : CCmode))
5670 You should define this macro if and only if you define extra CC modes
5671 in @file{@var{machine}-modes.def}.
5674 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5675 On some machines not all possible comparisons are defined, but you can
5676 convert an invalid comparison into a valid one. For example, the Alpha
5677 does not have a @code{GT} comparison, but you can use an @code{LT}
5678 comparison instead and swap the order of the operands.
5680 On such machines, define this macro to be a C statement to do any
5681 required conversions. @var{code} is the initial comparison code
5682 and @var{op0} and @var{op1} are the left and right operands of the
5683 comparison, respectively. You should modify @var{code}, @var{op0}, and
5684 @var{op1} as required.
5686 GCC will not assume that the comparison resulting from this macro is
5687 valid but will see if the resulting insn matches a pattern in the
5690 You need not define this macro if it would never change the comparison
5694 @defmac REVERSIBLE_CC_MODE (@var{mode})
5695 A C expression whose value is one if it is always safe to reverse a
5696 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5697 can ever return @var{mode} for a floating-point inequality comparison,
5698 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5700 You need not define this macro if it would always returns zero or if the
5701 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5702 For example, here is the definition used on the SPARC, where floating-point
5703 inequality comparisons are always given @code{CCFPEmode}:
5706 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5710 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
5711 A C expression whose value is reversed condition code of the @var{code} for
5712 comparison done in CC_MODE @var{mode}. The macro is used only in case
5713 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5714 machine has some non-standard way how to reverse certain conditionals. For
5715 instance in case all floating point conditions are non-trapping, compiler may
5716 freely convert unordered compares to ordered one. Then definition may look
5720 #define REVERSE_CONDITION(CODE, MODE) \
5721 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5722 : reverse_condition_maybe_unordered (CODE))
5726 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
5727 A C expression that returns true if the conditional execution predicate
5728 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
5729 versa. Define this to return 0 if the target has conditional execution
5730 predicates that cannot be reversed safely. There is no need to validate
5731 that the arguments of op1 and op2 are the same, this is done separately.
5732 If no expansion is specified, this macro is defined as follows:
5735 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5736 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
5740 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *, unsigned int *)
5741 On targets which do not use @code{(cc0)}, and which use a hard
5742 register rather than a pseudo-register to hold condition codes, the
5743 regular CSE passes are often not able to identify cases in which the
5744 hard register is set to a common value. Use this hook to enable a
5745 small pass which optimizes such cases. This hook should return true
5746 to enable this pass, and it should set the integers to which its
5747 arguments point to the hard register numbers used for condition codes.
5748 When there is only one such register, as is true on most systems, the
5749 integer pointed to by the second argument should be set to
5750 @code{INVALID_REGNUM}.
5752 The default version of this hook returns false.
5755 @deftypefn {Target Hook} enum machine_mode TARGET_CC_MODES_COMPATIBLE (enum machine_mode, enum machine_mode)
5756 On targets which use multiple condition code modes in class
5757 @code{MODE_CC}, it is sometimes the case that a comparison can be
5758 validly done in more than one mode. On such a system, define this
5759 target hook to take two mode arguments and to return a mode in which
5760 both comparisons may be validly done. If there is no such mode,
5761 return @code{VOIDmode}.
5763 The default version of this hook checks whether the modes are the
5764 same. If they are, it returns that mode. If they are different, it
5765 returns @code{VOIDmode}.
5769 @section Describing Relative Costs of Operations
5770 @cindex costs of instructions
5771 @cindex relative costs
5772 @cindex speed of instructions
5774 These macros let you describe the relative speed of various operations
5775 on the target machine.
5777 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5778 A C expression for the cost of moving data of mode @var{mode} from a
5779 register in class @var{from} to one in class @var{to}. The classes are
5780 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5781 value of 2 is the default; other values are interpreted relative to
5784 It is not required that the cost always equal 2 when @var{from} is the
5785 same as @var{to}; on some machines it is expensive to move between
5786 registers if they are not general registers.
5788 If reload sees an insn consisting of a single @code{set} between two
5789 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5790 classes returns a value of 2, reload does not check to ensure that the
5791 constraints of the insn are met. Setting a cost of other than 2 will
5792 allow reload to verify that the constraints are met. You should do this
5793 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5796 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5797 A C expression for the cost of moving data of mode @var{mode} between a
5798 register of class @var{class} and memory; @var{in} is zero if the value
5799 is to be written to memory, nonzero if it is to be read in. This cost
5800 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5801 registers and memory is more expensive than between two registers, you
5802 should define this macro to express the relative cost.
5804 If you do not define this macro, GCC uses a default cost of 4 plus
5805 the cost of copying via a secondary reload register, if one is
5806 needed. If your machine requires a secondary reload register to copy
5807 between memory and a register of @var{class} but the reload mechanism is
5808 more complex than copying via an intermediate, define this macro to
5809 reflect the actual cost of the move.
5811 GCC defines the function @code{memory_move_secondary_cost} if
5812 secondary reloads are needed. It computes the costs due to copying via
5813 a secondary register. If your machine copies from memory using a
5814 secondary register in the conventional way but the default base value of
5815 4 is not correct for your machine, define this macro to add some other
5816 value to the result of that function. The arguments to that function
5817 are the same as to this macro.
5821 A C expression for the cost of a branch instruction. A value of 1 is
5822 the default; other values are interpreted relative to that.
5825 Here are additional macros which do not specify precise relative costs,
5826 but only that certain actions are more expensive than GCC would
5829 @defmac SLOW_BYTE_ACCESS
5830 Define this macro as a C expression which is nonzero if accessing less
5831 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5832 faster than accessing a word of memory, i.e., if such access
5833 require more than one instruction or if there is no difference in cost
5834 between byte and (aligned) word loads.
5836 When this macro is not defined, the compiler will access a field by
5837 finding the smallest containing object; when it is defined, a fullword
5838 load will be used if alignment permits. Unless bytes accesses are
5839 faster than word accesses, using word accesses is preferable since it
5840 may eliminate subsequent memory access if subsequent accesses occur to
5841 other fields in the same word of the structure, but to different bytes.
5844 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5845 Define this macro to be the value 1 if memory accesses described by the
5846 @var{mode} and @var{alignment} parameters have a cost many times greater
5847 than aligned accesses, for example if they are emulated in a trap
5850 When this macro is nonzero, the compiler will act as if
5851 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5852 moves. This can cause significantly more instructions to be produced.
5853 Therefore, do not set this macro nonzero if unaligned accesses only add a
5854 cycle or two to the time for a memory access.
5856 If the value of this macro is always zero, it need not be defined. If
5857 this macro is defined, it should produce a nonzero value when
5858 @code{STRICT_ALIGNMENT} is nonzero.
5862 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5863 which a sequence of insns should be generated instead of a
5864 string move insn or a library call. Increasing the value will always
5865 make code faster, but eventually incurs high cost in increased code size.
5867 Note that on machines where the corresponding move insn is a
5868 @code{define_expand} that emits a sequence of insns, this macro counts
5869 the number of such sequences.
5871 If you don't define this, a reasonable default is used.
5874 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5875 A C expression used to determine whether @code{move_by_pieces} will be used to
5876 copy a chunk of memory, or whether some other block move mechanism
5877 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5878 than @code{MOVE_RATIO}.
5881 @defmac MOVE_MAX_PIECES
5882 A C expression used by @code{move_by_pieces} to determine the largest unit
5883 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5887 The threshold of number of scalar move insns, @emph{below} which a sequence
5888 of insns should be generated to clear memory instead of a string clear insn
5889 or a library call. Increasing the value will always make code faster, but
5890 eventually incurs high cost in increased code size.
5892 If you don't define this, a reasonable default is used.
5895 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5896 A C expression used to determine whether @code{clear_by_pieces} will be used
5897 to clear a chunk of memory, or whether some other block clear mechanism
5898 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5899 than @code{CLEAR_RATIO}.
5903 The threshold of number of scalar move insns, @emph{below} which a sequence
5904 of insns should be generated to set memory to a constant value, instead of
5905 a block set insn or a library call.
5906 Increasing the value will always make code faster, but
5907 eventually incurs high cost in increased code size.
5909 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
5912 @defmac SET_BY_PIECES_P (@var{size}, @var{alignment})
5913 A C expression used to determine whether @code{store_by_pieces} will be
5914 used to set a chunk of memory to a constant value, or whether some
5915 other mechanism will be used. Used by @code{__builtin_memset} when
5916 storing values other than constant zero.
5917 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5918 than @code{SET_RATIO}.
5921 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
5922 A C expression used to determine whether @code{store_by_pieces} will be
5923 used to set a chunk of memory to a constant string value, or whether some
5924 other mechanism will be used. Used by @code{__builtin_strcpy} when
5925 called with a constant source string.
5926 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5927 than @code{MOVE_RATIO}.
5930 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
5931 A C expression used to determine whether a load postincrement is a good
5932 thing to use for a given mode. Defaults to the value of
5933 @code{HAVE_POST_INCREMENT}.
5936 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
5937 A C expression used to determine whether a load postdecrement is a good
5938 thing to use for a given mode. Defaults to the value of
5939 @code{HAVE_POST_DECREMENT}.
5942 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
5943 A C expression used to determine whether a load preincrement is a good
5944 thing to use for a given mode. Defaults to the value of
5945 @code{HAVE_PRE_INCREMENT}.
5948 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
5949 A C expression used to determine whether a load predecrement is a good
5950 thing to use for a given mode. Defaults to the value of
5951 @code{HAVE_PRE_DECREMENT}.
5954 @defmac USE_STORE_POST_INCREMENT (@var{mode})
5955 A C expression used to determine whether a store postincrement is a good
5956 thing to use for a given mode. Defaults to the value of
5957 @code{HAVE_POST_INCREMENT}.
5960 @defmac USE_STORE_POST_DECREMENT (@var{mode})
5961 A C expression used to determine whether a store postdecrement is a good
5962 thing to use for a given mode. Defaults to the value of
5963 @code{HAVE_POST_DECREMENT}.
5966 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
5967 This macro is used to determine whether a store preincrement is a good
5968 thing to use for a given mode. Defaults to the value of
5969 @code{HAVE_PRE_INCREMENT}.
5972 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
5973 This macro is used to determine whether a store predecrement is a good
5974 thing to use for a given mode. Defaults to the value of
5975 @code{HAVE_PRE_DECREMENT}.
5978 @defmac NO_FUNCTION_CSE
5979 Define this macro if it is as good or better to call a constant
5980 function address than to call an address kept in a register.
5983 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
5984 Define this macro if a non-short-circuit operation produced by
5985 @samp{fold_range_test ()} is optimal. This macro defaults to true if
5986 @code{BRANCH_COST} is greater than or equal to the value 2.
5989 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total})
5990 This target hook describes the relative costs of RTL expressions.
5992 The cost may depend on the precise form of the expression, which is
5993 available for examination in @var{x}, and the rtx code of the expression
5994 in which it is contained, found in @var{outer_code}. @var{code} is the
5995 expression code---redundant, since it can be obtained with
5996 @code{GET_CODE (@var{x})}.
5998 In implementing this hook, you can use the construct
5999 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6002 On entry to the hook, @code{*@var{total}} contains a default estimate
6003 for the cost of the expression. The hook should modify this value as
6004 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
6005 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
6006 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
6008 When optimizing for code size, i.e.@: when @code{optimize_size} is
6009 nonzero, this target hook should be used to estimate the relative
6010 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6012 The hook returns true when all subexpressions of @var{x} have been
6013 processed, and false when @code{rtx_cost} should recurse.
6016 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address})
6017 This hook computes the cost of an addressing mode that contains
6018 @var{address}. If not defined, the cost is computed from
6019 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
6021 For most CISC machines, the default cost is a good approximation of the
6022 true cost of the addressing mode. However, on RISC machines, all
6023 instructions normally have the same length and execution time. Hence
6024 all addresses will have equal costs.
6026 In cases where more than one form of an address is known, the form with
6027 the lowest cost will be used. If multiple forms have the same, lowest,
6028 cost, the one that is the most complex will be used.
6030 For example, suppose an address that is equal to the sum of a register
6031 and a constant is used twice in the same basic block. When this macro
6032 is not defined, the address will be computed in a register and memory
6033 references will be indirect through that register. On machines where
6034 the cost of the addressing mode containing the sum is no higher than
6035 that of a simple indirect reference, this will produce an additional
6036 instruction and possibly require an additional register. Proper
6037 specification of this macro eliminates this overhead for such machines.
6039 This hook is never called with an invalid address.
6041 On machines where an address involving more than one register is as
6042 cheap as an address computation involving only one register, defining
6043 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
6044 be live over a region of code where only one would have been if
6045 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
6046 should be considered in the definition of this macro. Equivalent costs
6047 should probably only be given to addresses with different numbers of
6048 registers on machines with lots of registers.
6052 @section Adjusting the Instruction Scheduler
6054 The instruction scheduler may need a fair amount of machine-specific
6055 adjustment in order to produce good code. GCC provides several target
6056 hooks for this purpose. It is usually enough to define just a few of
6057 them: try the first ones in this list first.
6059 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
6060 This hook returns the maximum number of instructions that can ever
6061 issue at the same time on the target machine. The default is one.
6062 Although the insn scheduler can define itself the possibility of issue
6063 an insn on the same cycle, the value can serve as an additional
6064 constraint to issue insns on the same simulated processor cycle (see
6065 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
6066 This value must be constant over the entire compilation. If you need
6067 it to vary depending on what the instructions are, you must use
6068 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
6071 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
6072 This hook is executed by the scheduler after it has scheduled an insn
6073 from the ready list. It should return the number of insns which can
6074 still be issued in the current cycle. The default is
6075 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
6076 @code{USE}, which normally are not counted against the issue rate.
6077 You should define this hook if some insns take more machine resources
6078 than others, so that fewer insns can follow them in the same cycle.
6079 @var{file} is either a null pointer, or a stdio stream to write any
6080 debug output to. @var{verbose} is the verbose level provided by
6081 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
6085 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
6086 This function corrects the value of @var{cost} based on the
6087 relationship between @var{insn} and @var{dep_insn} through the
6088 dependence @var{link}. It should return the new value. The default
6089 is to make no adjustment to @var{cost}. This can be used for example
6090 to specify to the scheduler using the traditional pipeline description
6091 that an output- or anti-dependence does not incur the same cost as a
6092 data-dependence. If the scheduler using the automaton based pipeline
6093 description, the cost of anti-dependence is zero and the cost of
6094 output-dependence is maximum of one and the difference of latency
6095 times of the first and the second insns. If these values are not
6096 acceptable, you could use the hook to modify them too. See also
6097 @pxref{Processor pipeline description}.
6100 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
6101 This hook adjusts the integer scheduling priority @var{priority} of
6102 @var{insn}. It should return the new priority. Increase the priority to
6103 execute @var{insn} earlier, reduce the priority to execute @var{insn}
6104 later. Do not define this hook if you do not need to adjust the
6105 scheduling priorities of insns.
6108 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
6109 This hook is executed by the scheduler after it has scheduled the ready
6110 list, to allow the machine description to reorder it (for example to
6111 combine two small instructions together on @samp{VLIW} machines).
6112 @var{file} is either a null pointer, or a stdio stream to write any
6113 debug output to. @var{verbose} is the verbose level provided by
6114 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
6115 list of instructions that are ready to be scheduled. @var{n_readyp} is
6116 a pointer to the number of elements in the ready list. The scheduler
6117 reads the ready list in reverse order, starting with
6118 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
6119 is the timer tick of the scheduler. You may modify the ready list and
6120 the number of ready insns. The return value is the number of insns that
6121 can issue this cycle; normally this is just @code{issue_rate}. See also
6122 @samp{TARGET_SCHED_REORDER2}.
6125 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
6126 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
6127 function is called whenever the scheduler starts a new cycle. This one
6128 is called once per iteration over a cycle, immediately after
6129 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
6130 return the number of insns to be scheduled in the same cycle. Defining
6131 this hook can be useful if there are frequent situations where
6132 scheduling one insn causes other insns to become ready in the same
6133 cycle. These other insns can then be taken into account properly.
6136 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
6137 This hook is called after evaluation forward dependencies of insns in
6138 chain given by two parameter values (@var{head} and @var{tail}
6139 correspondingly) but before insns scheduling of the insn chain. For
6140 example, it can be used for better insn classification if it requires
6141 analysis of dependencies. This hook can use backward and forward
6142 dependencies of the insn scheduler because they are already
6146 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
6147 This hook is executed by the scheduler at the beginning of each block of
6148 instructions that are to be scheduled. @var{file} is either a null
6149 pointer, or a stdio stream to write any debug output to. @var{verbose}
6150 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6151 @var{max_ready} is the maximum number of insns in the current scheduling
6152 region that can be live at the same time. This can be used to allocate
6153 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
6156 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
6157 This hook is executed by the scheduler at the end of each block of
6158 instructions that are to be scheduled. It can be used to perform
6159 cleanup of any actions done by the other scheduling hooks. @var{file}
6160 is either a null pointer, or a stdio stream to write any debug output
6161 to. @var{verbose} is the verbose level provided by
6162 @option{-fsched-verbose-@var{n}}.
6165 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
6166 This hook is executed by the scheduler after function level initializations.
6167 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6168 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6169 @var{old_max_uid} is the maximum insn uid when scheduling begins.
6172 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
6173 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
6174 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6175 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6178 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
6179 The hook returns an RTL insn. The automaton state used in the
6180 pipeline hazard recognizer is changed as if the insn were scheduled
6181 when the new simulated processor cycle starts. Usage of the hook may
6182 simplify the automaton pipeline description for some @acronym{VLIW}
6183 processors. If the hook is defined, it is used only for the automaton
6184 based pipeline description. The default is not to change the state
6185 when the new simulated processor cycle starts.
6188 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
6189 The hook can be used to initialize data used by the previous hook.
6192 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
6193 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6194 to changed the state as if the insn were scheduled when the new
6195 simulated processor cycle finishes.
6198 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
6199 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
6200 used to initialize data used by the previous hook.
6203 @deftypefn {Target Hook} void TARGET_SCHED_DFA_PRE_CYCLE_ADVANCE (void)
6204 The hook to notify target that the current simulated cycle is about to finish.
6205 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6206 to change the state in more complicated situations - e.g. when advancing
6207 state on a single insn is not enough.
6210 @deftypefn {Target Hook} void TARGET_SCHED_DFA_POST_CYCLE_ADVANCE (void)
6211 The hook to notify target that new simulated cycle has just started.
6212 The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used
6213 to change the state in more complicated situations - e.g. when advancing
6214 state on a single insn is not enough.
6217 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
6218 This hook controls better choosing an insn from the ready insn queue
6219 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
6220 chooses the first insn from the queue. If the hook returns a positive
6221 value, an additional scheduler code tries all permutations of
6222 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
6223 subsequent ready insns to choose an insn whose issue will result in
6224 maximal number of issued insns on the same cycle. For the
6225 @acronym{VLIW} processor, the code could actually solve the problem of
6226 packing simple insns into the @acronym{VLIW} insn. Of course, if the
6227 rules of @acronym{VLIW} packing are described in the automaton.
6229 This code also could be used for superscalar @acronym{RISC}
6230 processors. Let us consider a superscalar @acronym{RISC} processor
6231 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
6232 @var{B}, some insns can be executed only in pipelines @var{B} or
6233 @var{C}, and one insn can be executed in pipeline @var{B}. The
6234 processor may issue the 1st insn into @var{A} and the 2nd one into
6235 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
6236 until the next cycle. If the scheduler issues the 3rd insn the first,
6237 the processor could issue all 3 insns per cycle.
6239 Actually this code demonstrates advantages of the automaton based
6240 pipeline hazard recognizer. We try quickly and easy many insn
6241 schedules to choose the best one.
6243 The default is no multipass scheduling.
6246 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
6248 This hook controls what insns from the ready insn queue will be
6249 considered for the multipass insn scheduling. If the hook returns
6250 zero for insn passed as the parameter, the insn will be not chosen to
6253 The default is that any ready insns can be chosen to be issued.
6256 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
6258 This hook is called by the insn scheduler before issuing insn passed
6259 as the third parameter on given cycle. If the hook returns nonzero,
6260 the insn is not issued on given processors cycle. Instead of that,
6261 the processor cycle is advanced. If the value passed through the last
6262 parameter is zero, the insn ready queue is not sorted on the new cycle
6263 start as usually. The first parameter passes file for debugging
6264 output. The second one passes the scheduler verbose level of the
6265 debugging output. The forth and the fifth parameter values are
6266 correspondingly processor cycle on which the previous insn has been
6267 issued and the current processor cycle.
6270 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct dep_def *@var{_dep}, int @var{cost}, int @var{distance})
6271 This hook is used to define which dependences are considered costly by
6272 the target, so costly that it is not advisable to schedule the insns that
6273 are involved in the dependence too close to one another. The parameters
6274 to this hook are as follows: The first parameter @var{_dep} is the dependence
6275 being evaluated. The second parameter @var{cost} is the cost of the
6276 dependence, and the third
6277 parameter @var{distance} is the distance in cycles between the two insns.
6278 The hook returns @code{true} if considering the distance between the two
6279 insns the dependence between them is considered costly by the target,
6280 and @code{false} otherwise.
6282 Defining this hook can be useful in multiple-issue out-of-order machines,
6283 where (a) it's practically hopeless to predict the actual data/resource
6284 delays, however: (b) there's a better chance to predict the actual grouping
6285 that will be formed, and (c) correctly emulating the grouping can be very
6286 important. In such targets one may want to allow issuing dependent insns
6287 closer to one another---i.e., closer than the dependence distance; however,
6288 not in cases of "costly dependences", which this hooks allows to define.
6291 @deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void)
6292 This hook is called by the insn scheduler after emitting a new instruction to
6293 the instruction stream. The hook notifies a target backend to extend its
6294 per instruction data structures.
6297 @deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx @var{insn}, int @var{request}, rtx *@var{new_pat})
6298 This hook is called by the insn scheduler when @var{insn} has only
6299 speculative dependencies and therefore can be scheduled speculatively.
6300 The hook is used to check if the pattern of @var{insn} has a speculative
6301 version and, in case of successful check, to generate that speculative
6302 pattern. The hook should return 1, if the instruction has a speculative form,
6303 or -1, if it doesn't. @var{request} describes the type of requested
6304 speculation. If the return value equals 1 then @var{new_pat} is assigned
6305 the generated speculative pattern.
6308 @deftypefn {Target Hook} int TARGET_SCHED_NEEDS_BLOCK_P (rtx @var{insn})
6309 This hook is called by the insn scheduler during generation of recovery code
6310 for @var{insn}. It should return nonzero, if the corresponding check
6311 instruction should branch to recovery code, or zero otherwise.
6314 @deftypefn {Target Hook} rtx TARGET_SCHED_GEN_CHECK (rtx @var{insn}, rtx @var{label}, int @var{mutate_p})
6315 This hook is called by the insn scheduler to generate a pattern for recovery
6316 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
6317 speculative instruction for which the check should be generated.
6318 @var{label} is either a label of a basic block, where recovery code should
6319 be emitted, or a null pointer, when requested check doesn't branch to
6320 recovery code (a simple check). If @var{mutate_p} is nonzero, then
6321 a pattern for a branchy check corresponding to a simple check denoted by
6322 @var{insn} should be generated. In this case @var{label} can't be null.
6325 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC (rtx @var{insn})
6326 This hook is used as a workaround for
6327 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD} not being
6328 called on the first instruction of the ready list. The hook is used to
6329 discard speculative instruction that stand first in the ready list from
6330 being scheduled on the current cycle. For non-speculative instructions,
6331 the hook should always return nonzero. For example, in the ia64 backend
6332 the hook is used to cancel data speculative insns when the ALAT table
6336 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (unsigned int *@var{flags}, spec_info_t @var{spec_info})
6337 This hook is used by the insn scheduler to find out what features should be
6338 enabled/used. @var{flags} initially may have either the SCHED_RGN or SCHED_EBB
6339 bit set. This denotes the scheduler pass for which the data should be
6340 provided. The target backend should modify @var{flags} by modifying
6341 the bits corresponding to the following features: USE_DEPS_LIST, USE_GLAT,
6342 DETACH_LIFE_INFO, and DO_SPECULATION. For the DO_SPECULATION feature
6343 an additional structure @var{spec_info} should be filled by the target.
6344 The structure describes speculation types that can be used in the scheduler.
6347 @deftypefn {Target Hook} int TARGET_SCHED_SMS_RES_MII (struct ddg *@var{g})
6348 This hook is called by the swing modulo scheduler to calculate a
6349 resource-based lower bound which is based on the resources available in
6350 the machine and the resources required by each instruction. The target
6351 backend can use @var{g} to calculate such bound. A very simple lower
6352 bound will be used in case this hook is not implemented: the total number
6353 of instructions divided by the issue rate.
6357 @section Dividing the Output into Sections (Texts, Data, @dots{})
6358 @c the above section title is WAY too long. maybe cut the part between
6359 @c the (...)? --mew 10feb93
6361 An object file is divided into sections containing different types of
6362 data. In the most common case, there are three sections: the @dfn{text
6363 section}, which holds instructions and read-only data; the @dfn{data
6364 section}, which holds initialized writable data; and the @dfn{bss
6365 section}, which holds uninitialized data. Some systems have other kinds
6368 @file{varasm.c} provides several well-known sections, such as
6369 @code{text_section}, @code{data_section} and @code{bss_section}.
6370 The normal way of controlling a @code{@var{foo}_section} variable
6371 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
6372 as described below. The macros are only read once, when @file{varasm.c}
6373 initializes itself, so their values must be run-time constants.
6374 They may however depend on command-line flags.
6376 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
6377 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
6378 to be string literals.
6380 Some assemblers require a different string to be written every time a
6381 section is selected. If your assembler falls into this category, you
6382 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
6383 @code{get_unnamed_section} to set up the sections.
6385 You must always create a @code{text_section}, either by defining
6386 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
6387 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
6388 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
6389 create a distinct @code{readonly_data_section}, the default is to
6390 reuse @code{text_section}.
6392 All the other @file{varasm.c} sections are optional, and are null
6393 if the target does not provide them.
6395 @defmac TEXT_SECTION_ASM_OP
6396 A C expression whose value is a string, including spacing, containing the
6397 assembler operation that should precede instructions and read-only data.
6398 Normally @code{"\t.text"} is right.
6401 @defmac HOT_TEXT_SECTION_NAME
6402 If defined, a C string constant for the name of the section containing most
6403 frequently executed functions of the program. If not defined, GCC will provide
6404 a default definition if the target supports named sections.
6407 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
6408 If defined, a C string constant for the name of the section containing unlikely
6409 executed functions in the program.
6412 @defmac DATA_SECTION_ASM_OP
6413 A C expression whose value is a string, including spacing, containing the
6414 assembler operation to identify the following data as writable initialized
6415 data. Normally @code{"\t.data"} is right.
6418 @defmac SDATA_SECTION_ASM_OP
6419 If defined, a C expression whose value is a string, including spacing,
6420 containing the assembler operation to identify the following data as
6421 initialized, writable small data.
6424 @defmac READONLY_DATA_SECTION_ASM_OP
6425 A C expression whose value is a string, including spacing, containing the
6426 assembler operation to identify the following data as read-only initialized
6430 @defmac BSS_SECTION_ASM_OP
6431 If defined, a C expression whose value is a string, including spacing,
6432 containing the assembler operation to identify the following data as
6433 uninitialized global data. If not defined, and neither
6434 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
6435 uninitialized global data will be output in the data section if
6436 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
6440 @defmac SBSS_SECTION_ASM_OP
6441 If defined, a C expression whose value is a string, including spacing,
6442 containing the assembler operation to identify the following data as
6443 uninitialized, writable small data.
6446 @defmac INIT_SECTION_ASM_OP
6447 If defined, a C expression whose value is a string, including spacing,
6448 containing the assembler operation to identify the following data as
6449 initialization code. If not defined, GCC will assume such a section does
6450 not exist. This section has no corresponding @code{init_section}
6451 variable; it is used entirely in runtime code.
6454 @defmac FINI_SECTION_ASM_OP
6455 If defined, a C expression whose value is a string, including spacing,
6456 containing the assembler operation to identify the following data as
6457 finalization code. If not defined, GCC will assume such a section does
6458 not exist. This section has no corresponding @code{fini_section}
6459 variable; it is used entirely in runtime code.
6462 @defmac INIT_ARRAY_SECTION_ASM_OP
6463 If defined, a C expression whose value is a string, including spacing,
6464 containing the assembler operation to identify the following data as
6465 part of the @code{.init_array} (or equivalent) section. If not
6466 defined, GCC will assume such a section does not exist. Do not define
6467 both this macro and @code{INIT_SECTION_ASM_OP}.
6470 @defmac FINI_ARRAY_SECTION_ASM_OP
6471 If defined, a C expression whose value is a string, including spacing,
6472 containing the assembler operation to identify the following data as
6473 part of the @code{.fini_array} (or equivalent) section. If not
6474 defined, GCC will assume such a section does not exist. Do not define
6475 both this macro and @code{FINI_SECTION_ASM_OP}.
6478 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
6479 If defined, an ASM statement that switches to a different section
6480 via @var{section_op}, calls @var{function}, and switches back to
6481 the text section. This is used in @file{crtstuff.c} if
6482 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
6483 to initialization and finalization functions from the init and fini
6484 sections. By default, this macro uses a simple function call. Some
6485 ports need hand-crafted assembly code to avoid dependencies on
6486 registers initialized in the function prologue or to ensure that
6487 constant pools don't end up too far way in the text section.
6490 @defmac TARGET_LIBGCC_SDATA_SECTION
6491 If defined, a string which names the section into which small
6492 variables defined in crtstuff and libgcc should go. This is useful
6493 when the target has options for optimizing access to small data, and
6494 you want the crtstuff and libgcc routines to be conservative in what
6495 they expect of your application yet liberal in what your application
6496 expects. For example, for targets with a @code{.sdata} section (like
6497 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
6498 require small data support from your application, but use this macro
6499 to put small data into @code{.sdata} so that your application can
6500 access these variables whether it uses small data or not.
6503 @defmac FORCE_CODE_SECTION_ALIGN
6504 If defined, an ASM statement that aligns a code section to some
6505 arbitrary boundary. This is used to force all fragments of the
6506 @code{.init} and @code{.fini} sections to have to same alignment
6507 and thus prevent the linker from having to add any padding.
6510 @defmac JUMP_TABLES_IN_TEXT_SECTION
6511 Define this macro to be an expression with a nonzero value if jump
6512 tables (for @code{tablejump} insns) should be output in the text
6513 section, along with the assembler instructions. Otherwise, the
6514 readonly data section is used.
6516 This macro is irrelevant if there is no separate readonly data section.
6519 @deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void)
6520 Define this hook if you need to do something special to set up the
6521 @file{varasm.c} sections, or if your target has some special sections
6522 of its own that you need to create.
6524 GCC calls this hook after processing the command line, but before writing
6525 any assembly code, and before calling any of the section-returning hooks
6529 @deftypefn {Target Hook} TARGET_ASM_RELOC_RW_MASK (void)
6530 Return a mask describing how relocations should be treated when
6531 selecting sections. Bit 1 should be set if global relocations
6532 should be placed in a read-write section; bit 0 should be set if
6533 local relocations should be placed in a read-write section.
6535 The default version of this function returns 3 when @option{-fpic}
6536 is in effect, and 0 otherwise. The hook is typically redefined
6537 when the target cannot support (some kinds of) dynamic relocations
6538 in read-only sections even in executables.
6541 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
6542 Return the section into which @var{exp} should be placed. You can
6543 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
6544 some sort. @var{reloc} indicates whether the initial value of @var{exp}
6545 requires link-time relocations. Bit 0 is set when variable contains
6546 local relocations only, while bit 1 is set for global relocations.
6547 @var{align} is the constant alignment in bits.
6549 The default version of this function takes care of putting read-only
6550 variables in @code{readonly_data_section}.
6552 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
6555 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
6556 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
6557 for @code{FUNCTION_DECL}s as well as for variables and constants.
6559 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
6560 function has been determined to be likely to be called, and nonzero if
6561 it is unlikely to be called.
6564 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
6565 Build up a unique section name, expressed as a @code{STRING_CST} node,
6566 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
6567 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
6568 the initial value of @var{exp} requires link-time relocations.
6570 The default version of this function appends the symbol name to the
6571 ELF section name that would normally be used for the symbol. For
6572 example, the function @code{foo} would be placed in @code{.text.foo}.
6573 Whatever the actual target object format, this is often good enough.
6576 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
6577 Return the readonly data section associated with
6578 @samp{DECL_SECTION_NAME (@var{decl})}.
6579 The default version of this function selects @code{.gnu.linkonce.r.name} if
6580 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
6581 if function is in @code{.text.name}, and the normal readonly-data section
6585 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
6586 Return the section into which a constant @var{x}, of mode @var{mode},
6587 should be placed. You can assume that @var{x} is some kind of
6588 constant in RTL@. The argument @var{mode} is redundant except in the
6589 case of a @code{const_int} rtx. @var{align} is the constant alignment
6592 The default version of this function takes care of putting symbolic
6593 constants in @code{flag_pic} mode in @code{data_section} and everything
6594 else in @code{readonly_data_section}.
6597 @deftypefn {Target Hook} void TARGET_MANGLE_DECL_ASSEMBLER_NAME (tree @var{decl}, tree @var{id})
6598 Define this hook if you need to postprocess the assembler name generated
6599 by target-independent code. The @var{id} provided to this hook will be
6600 the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C,
6601 or the mangled name of the @var{decl} in C++). The return value of the
6602 hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on
6603 your target system. The default implementation of this hook just
6604 returns the @var{id} provided.
6607 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
6608 Define this hook if references to a symbol or a constant must be
6609 treated differently depending on something about the variable or
6610 function named by the symbol (such as what section it is in).
6612 The hook is executed immediately after rtl has been created for
6613 @var{decl}, which may be a variable or function declaration or
6614 an entry in the constant pool. In either case, @var{rtl} is the
6615 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
6616 in this hook; that field may not have been initialized yet.
6618 In the case of a constant, it is safe to assume that the rtl is
6619 a @code{mem} whose address is a @code{symbol_ref}. Most decls
6620 will also have this form, but that is not guaranteed. Global
6621 register variables, for instance, will have a @code{reg} for their
6622 rtl. (Normally the right thing to do with such unusual rtl is
6625 The @var{new_decl_p} argument will be true if this is the first time
6626 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
6627 be false for subsequent invocations, which will happen for duplicate
6628 declarations. Whether or not anything must be done for the duplicate
6629 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
6630 @var{new_decl_p} is always true when the hook is called for a constant.
6632 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
6633 The usual thing for this hook to do is to record flags in the
6634 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
6635 Historically, the name string was modified if it was necessary to
6636 encode more than one bit of information, but this practice is now
6637 discouraged; use @code{SYMBOL_REF_FLAGS}.
6639 The default definition of this hook, @code{default_encode_section_info}
6640 in @file{varasm.c}, sets a number of commonly-useful bits in
6641 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
6642 before overriding it.
6645 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
6646 Decode @var{name} and return the real name part, sans
6647 the characters that @code{TARGET_ENCODE_SECTION_INFO}
6651 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
6652 Returns true if @var{exp} should be placed into a ``small data'' section.
6653 The default version of this hook always returns false.
6656 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
6657 Contains the value true if the target places read-only
6658 ``small data'' into a separate section. The default value is false.
6661 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
6662 Returns true if @var{exp} names an object for which name resolution
6663 rules must resolve to the current ``module'' (dynamic shared library
6664 or executable image).
6666 The default version of this hook implements the name resolution rules
6667 for ELF, which has a looser model of global name binding than other
6668 currently supported object file formats.
6671 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
6672 Contains the value true if the target supports thread-local storage.
6673 The default value is false.
6678 @section Position Independent Code
6679 @cindex position independent code
6682 This section describes macros that help implement generation of position
6683 independent code. Simply defining these macros is not enough to
6684 generate valid PIC; you must also add support to the macros
6685 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
6686 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
6687 @samp{movsi} to do something appropriate when the source operand
6688 contains a symbolic address. You may also need to alter the handling of
6689 switch statements so that they use relative addresses.
6690 @c i rearranged the order of the macros above to try to force one of
6691 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
6693 @defmac PIC_OFFSET_TABLE_REGNUM
6694 The register number of the register used to address a table of static
6695 data addresses in memory. In some cases this register is defined by a
6696 processor's ``application binary interface'' (ABI)@. When this macro
6697 is defined, RTL is generated for this register once, as with the stack
6698 pointer and frame pointer registers. If this macro is not defined, it
6699 is up to the machine-dependent files to allocate such a register (if
6700 necessary). Note that this register must be fixed when in use (e.g.@:
6701 when @code{flag_pic} is true).
6704 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6705 Define this macro if the register defined by
6706 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
6707 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
6710 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
6711 A C expression that is nonzero if @var{x} is a legitimate immediate
6712 operand on the target machine when generating position independent code.
6713 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6714 check this. You can also assume @var{flag_pic} is true, so you need not
6715 check it either. You need not define this macro if all constants
6716 (including @code{SYMBOL_REF}) can be immediate operands when generating
6717 position independent code.
6720 @node Assembler Format
6721 @section Defining the Output Assembler Language
6723 This section describes macros whose principal purpose is to describe how
6724 to write instructions in assembler language---rather than what the
6728 * File Framework:: Structural information for the assembler file.
6729 * Data Output:: Output of constants (numbers, strings, addresses).
6730 * Uninitialized Data:: Output of uninitialized variables.
6731 * Label Output:: Output and generation of labels.
6732 * Initialization:: General principles of initialization
6733 and termination routines.
6734 * Macros for Initialization::
6735 Specific macros that control the handling of
6736 initialization and termination routines.
6737 * Instruction Output:: Output of actual instructions.
6738 * Dispatch Tables:: Output of jump tables.
6739 * Exception Region Output:: Output of exception region code.
6740 * Alignment Output:: Pseudo ops for alignment and skipping data.
6743 @node File Framework
6744 @subsection The Overall Framework of an Assembler File
6745 @cindex assembler format
6746 @cindex output of assembler code
6748 @c prevent bad page break with this line
6749 This describes the overall framework of an assembly file.
6751 @deftypefn {Target Hook} void TARGET_ASM_FILE_START ()
6752 @findex default_file_start
6753 Output to @code{asm_out_file} any text which the assembler expects to
6754 find at the beginning of a file. The default behavior is controlled
6755 by two flags, documented below. Unless your target's assembler is
6756 quite unusual, if you override the default, you should call
6757 @code{default_file_start} at some point in your target hook. This
6758 lets other target files rely on these variables.
6761 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
6762 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
6763 printed as the very first line in the assembly file, unless
6764 @option{-fverbose-asm} is in effect. (If that macro has been defined
6765 to the empty string, this variable has no effect.) With the normal
6766 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
6767 assembler that it need not bother stripping comments or extra
6768 whitespace from its input. This allows it to work a bit faster.
6770 The default is false. You should not set it to true unless you have
6771 verified that your port does not generate any extra whitespace or
6772 comments that will cause GAS to issue errors in NO_APP mode.
6775 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
6776 If this flag is true, @code{output_file_directive} will be called
6777 for the primary source file, immediately after printing
6778 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
6779 this to be done. The default is false.
6782 @deftypefn {Target Hook} void TARGET_ASM_FILE_END ()
6783 Output to @code{asm_out_file} any text which the assembler expects
6784 to find at the end of a file. The default is to output nothing.
6787 @deftypefun void file_end_indicate_exec_stack ()
6788 Some systems use a common convention, the @samp{.note.GNU-stack}
6789 special section, to indicate whether or not an object file relies on
6790 the stack being executable. If your system uses this convention, you
6791 should define @code{TARGET_ASM_FILE_END} to this function. If you
6792 need to do other things in that hook, have your hook function call
6796 @defmac ASM_COMMENT_START
6797 A C string constant describing how to begin a comment in the target
6798 assembler language. The compiler assumes that the comment will end at
6799 the end of the line.
6803 A C string constant for text to be output before each @code{asm}
6804 statement or group of consecutive ones. Normally this is
6805 @code{"#APP"}, which is a comment that has no effect on most
6806 assemblers but tells the GNU assembler that it must check the lines
6807 that follow for all valid assembler constructs.
6811 A C string constant for text to be output after each @code{asm}
6812 statement or group of consecutive ones. Normally this is
6813 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6814 time-saving assumptions that are valid for ordinary compiler output.
6817 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6818 A C statement to output COFF information or DWARF debugging information
6819 which indicates that filename @var{name} is the current source file to
6820 the stdio stream @var{stream}.
6822 This macro need not be defined if the standard form of output
6823 for the file format in use is appropriate.
6826 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6827 A C statement to output the string @var{string} to the stdio stream
6828 @var{stream}. If you do not call the function @code{output_quoted_string}
6829 in your config files, GCC will only call it to output filenames to
6830 the assembler source. So you can use it to canonicalize the format
6831 of the filename using this macro.
6834 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6835 A C statement to output something to the assembler file to handle a
6836 @samp{#ident} directive containing the text @var{string}. If this
6837 macro is not defined, nothing is output for a @samp{#ident} directive.
6840 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6841 Output assembly directives to switch to section @var{name}. The section
6842 should have attributes as specified by @var{flags}, which is a bit mask
6843 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6844 is nonzero, it contains an alignment in bytes to be used for the section,
6845 otherwise some target default should be used. Only targets that must
6846 specify an alignment within the section directive need pay attention to
6847 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6850 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6851 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6854 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
6855 @deftypefn {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
6856 This flag is true if we can create zeroed data by switching to a BSS
6857 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
6858 This is true on most ELF targets.
6861 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6862 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6863 based on a variable or function decl, a section name, and whether or not the
6864 declaration's initializer may contain runtime relocations. @var{decl} may be
6865 null, in which case read-write data should be assumed.
6867 The default version of this function handles choosing code vs data,
6868 read-only vs read-write data, and @code{flag_pic}. You should only
6869 need to override this if your target has special flags that might be
6870 set via @code{__attribute__}.
6873 @deftypefn {Target Hook} {int} TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type @var{type}, const char * @var{text})
6874 Provides the target with the ability to record the gcc command line
6875 switches that have been passed to the compiler, and options that are
6876 enabled. The @var{type} argument specifies what is being recorded.
6877 It can take the following values:
6880 @item SWITCH_TYPE_PASSED
6881 @var{text} is a command line switch that has been set by the user.
6883 @item SWITCH_TYPE_ENABLED
6884 @var{text} is an option which has been enabled. This might be as a
6885 direct result of a command line switch, or because it is enabled by
6886 default or because it has been enabled as a side effect of a different
6887 command line switch. For example, the @option{-O2} switch enables
6888 various different individual optimization passes.
6890 @item SWITCH_TYPE_DESCRIPTIVE
6891 @var{text} is either NULL or some descriptive text which should be
6892 ignored. If @var{text} is NULL then it is being used to warn the
6893 target hook that either recording is starting or ending. The first
6894 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
6895 warning is for start up and the second time the warning is for
6896 wind down. This feature is to allow the target hook to make any
6897 necessary preparations before it starts to record switches and to
6898 perform any necessary tidying up after it has finished recording
6901 @item SWITCH_TYPE_LINE_START
6902 This option can be ignored by this target hook.
6904 @item SWITCH_TYPE_LINE_END
6905 This option can be ignored by this target hook.
6908 The hook's return value must be zero. Other return values may be
6909 supported in the future.
6911 By default this hook is set to NULL, but an example implementation is
6912 provided for ELF based targets. Called @var{elf_record_gcc_switches},
6913 it records the switches as ASCII text inside a new, string mergeable
6914 section in the assembler output file. The name of the new section is
6915 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
6919 @deftypefn {Target Hook} {const char *} TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
6920 This is the name of the section that will be created by the example
6921 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
6927 @subsection Output of Data
6930 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6931 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6932 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6933 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6934 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6935 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6936 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6937 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6938 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6939 These hooks specify assembly directives for creating certain kinds
6940 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6941 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6942 aligned two-byte object, and so on. Any of the hooks may be
6943 @code{NULL}, indicating that no suitable directive is available.
6945 The compiler will print these strings at the start of a new line,
6946 followed immediately by the object's initial value. In most cases,
6947 the string should contain a tab, a pseudo-op, and then another tab.
6950 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6951 The @code{assemble_integer} function uses this hook to output an
6952 integer object. @var{x} is the object's value, @var{size} is its size
6953 in bytes and @var{aligned_p} indicates whether it is aligned. The
6954 function should return @code{true} if it was able to output the
6955 object. If it returns false, @code{assemble_integer} will try to
6956 split the object into smaller parts.
6958 The default implementation of this hook will use the
6959 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6960 when the relevant string is @code{NULL}.
6963 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6964 A C statement to recognize @var{rtx} patterns that
6965 @code{output_addr_const} can't deal with, and output assembly code to
6966 @var{stream} corresponding to the pattern @var{x}. This may be used to
6967 allow machine-dependent @code{UNSPEC}s to appear within constants.
6969 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6970 @code{goto fail}, so that a standard error message is printed. If it
6971 prints an error message itself, by calling, for example,
6972 @code{output_operand_lossage}, it may just complete normally.
6975 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6976 A C statement to output to the stdio stream @var{stream} an assembler
6977 instruction to assemble a string constant containing the @var{len}
6978 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6979 @code{char *} and @var{len} a C expression of type @code{int}.
6981 If the assembler has a @code{.ascii} pseudo-op as found in the
6982 Berkeley Unix assembler, do not define the macro
6983 @code{ASM_OUTPUT_ASCII}.
6986 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6987 A C statement to output word @var{n} of a function descriptor for
6988 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6989 is defined, and is otherwise unused.
6992 @defmac CONSTANT_POOL_BEFORE_FUNCTION
6993 You may define this macro as a C expression. You should define the
6994 expression to have a nonzero value if GCC should output the constant
6995 pool for a function before the code for the function, or a zero value if
6996 GCC should output the constant pool after the function. If you do
6997 not define this macro, the usual case, GCC will output the constant
6998 pool before the function.
7001 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
7002 A C statement to output assembler commands to define the start of the
7003 constant pool for a function. @var{funname} is a string giving
7004 the name of the function. Should the return type of the function
7005 be required, it can be obtained via @var{fundecl}. @var{size}
7006 is the size, in bytes, of the constant pool that will be written
7007 immediately after this call.
7009 If no constant-pool prefix is required, the usual case, this macro need
7013 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
7014 A C statement (with or without semicolon) to output a constant in the
7015 constant pool, if it needs special treatment. (This macro need not do
7016 anything for RTL expressions that can be output normally.)
7018 The argument @var{file} is the standard I/O stream to output the
7019 assembler code on. @var{x} is the RTL expression for the constant to
7020 output, and @var{mode} is the machine mode (in case @var{x} is a
7021 @samp{const_int}). @var{align} is the required alignment for the value
7022 @var{x}; you should output an assembler directive to force this much
7025 The argument @var{labelno} is a number to use in an internal label for
7026 the address of this pool entry. The definition of this macro is
7027 responsible for outputting the label definition at the proper place.
7028 Here is how to do this:
7031 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
7034 When you output a pool entry specially, you should end with a
7035 @code{goto} to the label @var{jumpto}. This will prevent the same pool
7036 entry from being output a second time in the usual manner.
7038 You need not define this macro if it would do nothing.
7041 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
7042 A C statement to output assembler commands to at the end of the constant
7043 pool for a function. @var{funname} is a string giving the name of the
7044 function. Should the return type of the function be required, you can
7045 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
7046 constant pool that GCC wrote immediately before this call.
7048 If no constant-pool epilogue is required, the usual case, you need not
7052 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
7053 Define this macro as a C expression which is nonzero if @var{C} is
7054 used as a logical line separator by the assembler. @var{STR} points
7055 to the position in the string where @var{C} was found; this can be used if
7056 a line separator uses multiple characters.
7058 If you do not define this macro, the default is that only
7059 the character @samp{;} is treated as a logical line separator.
7062 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
7063 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
7064 These target hooks are C string constants, describing the syntax in the
7065 assembler for grouping arithmetic expressions. If not overridden, they
7066 default to normal parentheses, which is correct for most assemblers.
7069 These macros are provided by @file{real.h} for writing the definitions
7070 of @code{ASM_OUTPUT_DOUBLE} and the like:
7072 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
7073 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
7074 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
7075 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
7076 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
7077 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
7078 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
7079 target's floating point representation, and store its bit pattern in
7080 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
7081 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
7082 simple @code{long int}. For the others, it should be an array of
7083 @code{long int}. The number of elements in this array is determined
7084 by the size of the desired target floating point data type: 32 bits of
7085 it go in each @code{long int} array element. Each array element holds
7086 32 bits of the result, even if @code{long int} is wider than 32 bits
7087 on the host machine.
7089 The array element values are designed so that you can print them out
7090 using @code{fprintf} in the order they should appear in the target
7094 @node Uninitialized Data
7095 @subsection Output of Uninitialized Variables
7097 Each of the macros in this section is used to do the whole job of
7098 outputting a single uninitialized variable.
7100 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
7101 A C statement (sans semicolon) to output to the stdio stream
7102 @var{stream} the assembler definition of a common-label named
7103 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7104 is the size rounded up to whatever alignment the caller wants.
7106 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7107 output the name itself; before and after that, output the additional
7108 assembler syntax for defining the name, and a newline.
7110 This macro controls how the assembler definitions of uninitialized
7111 common global variables are output.
7114 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
7115 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
7116 separate, explicit argument. If you define this macro, it is used in
7117 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
7118 handling the required alignment of the variable. The alignment is specified
7119 as the number of bits.
7122 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7123 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
7124 variable to be output, if there is one, or @code{NULL_TREE} if there
7125 is no corresponding variable. If you define this macro, GCC will use it
7126 in place of both @code{ASM_OUTPUT_COMMON} and
7127 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
7128 the variable's decl in order to chose what to output.
7131 @defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
7132 A C statement (sans semicolon) to output to the stdio stream
7133 @var{stream} the assembler definition of uninitialized global @var{decl} named
7134 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7135 is the size rounded up to whatever alignment the caller wants.
7137 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
7138 defining this macro. If unable, use the expression
7139 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
7140 before and after that, output the additional assembler syntax for defining
7141 the name, and a newline.
7143 There are two ways of handling global BSS. One is to define either
7144 this macro or its aligned counterpart, @code{ASM_OUTPUT_ALIGNED_BSS}.
7145 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
7146 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
7147 You do not need to do both.
7149 Some languages do not have @code{common} data, and require a
7150 non-common form of global BSS in order to handle uninitialized globals
7151 efficiently. C++ is one example of this. However, if the target does
7152 not support global BSS, the front end may choose to make globals
7153 common in order to save space in the object file.
7156 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7157 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
7158 separate, explicit argument. If you define this macro, it is used in
7159 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
7160 handling the required alignment of the variable. The alignment is specified
7161 as the number of bits.
7163 Try to use function @code{asm_output_aligned_bss} defined in file
7164 @file{varasm.c} when defining this macro.
7167 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
7168 A C statement (sans semicolon) to output to the stdio stream
7169 @var{stream} the assembler definition of a local-common-label named
7170 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7171 is the size rounded up to whatever alignment the caller wants.
7173 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7174 output the name itself; before and after that, output the additional
7175 assembler syntax for defining the name, and a newline.
7177 This macro controls how the assembler definitions of uninitialized
7178 static variables are output.
7181 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
7182 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
7183 separate, explicit argument. If you define this macro, it is used in
7184 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
7185 handling the required alignment of the variable. The alignment is specified
7186 as the number of bits.
7189 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7190 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
7191 variable to be output, if there is one, or @code{NULL_TREE} if there
7192 is no corresponding variable. If you define this macro, GCC will use it
7193 in place of both @code{ASM_OUTPUT_DECL} and
7194 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
7195 the variable's decl in order to chose what to output.
7199 @subsection Output and Generation of Labels
7201 @c prevent bad page break with this line
7202 This is about outputting labels.
7204 @findex assemble_name
7205 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
7206 A C statement (sans semicolon) to output to the stdio stream
7207 @var{stream} the assembler definition of a label named @var{name}.
7208 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7209 output the name itself; before and after that, output the additional
7210 assembler syntax for defining the name, and a newline. A default
7211 definition of this macro is provided which is correct for most systems.
7214 @findex assemble_name_raw
7215 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
7216 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
7217 to refer to a compiler-generated label. The default definition uses
7218 @code{assemble_name_raw}, which is like @code{assemble_name} except
7219 that it is more efficient.
7223 A C string containing the appropriate assembler directive to specify the
7224 size of a symbol, without any arguments. On systems that use ELF, the
7225 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
7226 systems, the default is not to define this macro.
7228 Define this macro only if it is correct to use the default definitions
7229 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
7230 for your system. If you need your own custom definitions of those
7231 macros, or if you do not need explicit symbol sizes at all, do not
7235 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
7236 A C statement (sans semicolon) to output to the stdio stream
7237 @var{stream} a directive telling the assembler that the size of the
7238 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
7239 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7243 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
7244 A C statement (sans semicolon) to output to the stdio stream
7245 @var{stream} a directive telling the assembler to calculate the size of
7246 the symbol @var{name} by subtracting its address from the current
7249 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7250 provided. The default assumes that the assembler recognizes a special
7251 @samp{.} symbol as referring to the current address, and can calculate
7252 the difference between this and another symbol. If your assembler does
7253 not recognize @samp{.} or cannot do calculations with it, you will need
7254 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
7258 A C string containing the appropriate assembler directive to specify the
7259 type of a symbol, without any arguments. On systems that use ELF, the
7260 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
7261 systems, the default is not to define this macro.
7263 Define this macro only if it is correct to use the default definition of
7264 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7265 custom definition of this macro, or if you do not need explicit symbol
7266 types at all, do not define this macro.
7269 @defmac TYPE_OPERAND_FMT
7270 A C string which specifies (using @code{printf} syntax) the format of
7271 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
7272 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
7273 the default is not to define this macro.
7275 Define this macro only if it is correct to use the default definition of
7276 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7277 custom definition of this macro, or if you do not need explicit symbol
7278 types at all, do not define this macro.
7281 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
7282 A C statement (sans semicolon) to output to the stdio stream
7283 @var{stream} a directive telling the assembler that the type of the
7284 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
7285 that string is always either @samp{"function"} or @samp{"object"}, but
7286 you should not count on this.
7288 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
7289 definition of this macro is provided.
7292 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
7293 A C statement (sans semicolon) to output to the stdio stream
7294 @var{stream} any text necessary for declaring the name @var{name} of a
7295 function which is being defined. This macro is responsible for
7296 outputting the label definition (perhaps using
7297 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
7298 @code{FUNCTION_DECL} tree node representing the function.
7300 If this macro is not defined, then the function name is defined in the
7301 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7303 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7307 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
7308 A C statement (sans semicolon) to output to the stdio stream
7309 @var{stream} any text necessary for declaring the size of a function
7310 which is being defined. The argument @var{name} is the name of the
7311 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
7312 representing the function.
7314 If this macro is not defined, then the function size is not defined.
7316 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
7320 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
7321 A C statement (sans semicolon) to output to the stdio stream
7322 @var{stream} any text necessary for declaring the name @var{name} of an
7323 initialized variable which is being defined. This macro must output the
7324 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
7325 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
7327 If this macro is not defined, then the variable name is defined in the
7328 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7330 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
7331 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
7334 @defmac ASM_DECLARE_CONSTANT_NAME (@var{stream}, @var{name}, @var{exp}, @var{size})
7335 A C statement (sans semicolon) to output to the stdio stream
7336 @var{stream} any text necessary for declaring the name @var{name} of a
7337 constant which is being defined. This macro is responsible for
7338 outputting the label definition (perhaps using
7339 @code{ASM_OUTPUT_LABEL}). The argument @var{exp} is the
7340 value of the constant, and @var{size} is the size of the constant
7341 in bytes. @var{name} will be an internal label.
7343 If this macro is not defined, then the @var{name} is defined in the
7344 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7346 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7350 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
7351 A C statement (sans semicolon) to output to the stdio stream
7352 @var{stream} any text necessary for claiming a register @var{regno}
7353 for a global variable @var{decl} with name @var{name}.
7355 If you don't define this macro, that is equivalent to defining it to do
7359 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
7360 A C statement (sans semicolon) to finish up declaring a variable name
7361 once the compiler has processed its initializer fully and thus has had a
7362 chance to determine the size of an array when controlled by an
7363 initializer. This is used on systems where it's necessary to declare
7364 something about the size of the object.
7366 If you don't define this macro, that is equivalent to defining it to do
7369 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
7370 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
7373 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
7374 This target hook is a function to output to the stdio stream
7375 @var{stream} some commands that will make the label @var{name} global;
7376 that is, available for reference from other files.
7378 The default implementation relies on a proper definition of
7379 @code{GLOBAL_ASM_OP}.
7382 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *@var{stream}, tree @var{decl})
7383 This target hook is a function to output to the stdio stream
7384 @var{stream} some commands that will make the name associated with @var{decl}
7385 global; that is, available for reference from other files.
7387 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
7390 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
7391 A C statement (sans semicolon) to output to the stdio stream
7392 @var{stream} some commands that will make the label @var{name} weak;
7393 that is, available for reference from other files but only used if
7394 no other definition is available. Use the expression
7395 @code{assemble_name (@var{stream}, @var{name})} to output the name
7396 itself; before and after that, output the additional assembler syntax
7397 for making that name weak, and a newline.
7399 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
7400 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
7404 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
7405 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
7406 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
7407 or variable decl. If @var{value} is not @code{NULL}, this C statement
7408 should output to the stdio stream @var{stream} assembler code which
7409 defines (equates) the weak symbol @var{name} to have the value
7410 @var{value}. If @var{value} is @code{NULL}, it should output commands
7411 to make @var{name} weak.
7414 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
7415 Outputs a directive that enables @var{name} to be used to refer to
7416 symbol @var{value} with weak-symbol semantics. @code{decl} is the
7417 declaration of @code{name}.
7420 @defmac SUPPORTS_WEAK
7421 A C expression which evaluates to true if the target supports weak symbols.
7423 If you don't define this macro, @file{defaults.h} provides a default
7424 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
7425 is defined, the default definition is @samp{1}; otherwise, it is
7426 @samp{0}. Define this macro if you want to control weak symbol support
7427 with a compiler flag such as @option{-melf}.
7430 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
7431 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
7432 public symbol such that extra copies in multiple translation units will
7433 be discarded by the linker. Define this macro if your object file
7434 format provides support for this concept, such as the @samp{COMDAT}
7435 section flags in the Microsoft Windows PE/COFF format, and this support
7436 requires changes to @var{decl}, such as putting it in a separate section.
7439 @defmac SUPPORTS_ONE_ONLY
7440 A C expression which evaluates to true if the target supports one-only
7443 If you don't define this macro, @file{varasm.c} provides a default
7444 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
7445 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
7446 you want to control one-only symbol support with a compiler flag, or if
7447 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
7448 be emitted as one-only.
7451 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
7452 This target hook is a function to output to @var{asm_out_file} some
7453 commands that will make the symbol(s) associated with @var{decl} have
7454 hidden, protected or internal visibility as specified by @var{visibility}.
7457 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
7458 A C expression that evaluates to true if the target's linker expects
7459 that weak symbols do not appear in a static archive's table of contents.
7460 The default is @code{0}.
7462 Leaving weak symbols out of an archive's table of contents means that,
7463 if a symbol will only have a definition in one translation unit and
7464 will have undefined references from other translation units, that
7465 symbol should not be weak. Defining this macro to be nonzero will
7466 thus have the effect that certain symbols that would normally be weak
7467 (explicit template instantiations, and vtables for polymorphic classes
7468 with noninline key methods) will instead be nonweak.
7470 The C++ ABI requires this macro to be zero. Define this macro for
7471 targets where full C++ ABI compliance is impossible and where linker
7472 restrictions require weak symbols to be left out of a static archive's
7476 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
7477 A C statement (sans semicolon) to output to the stdio stream
7478 @var{stream} any text necessary for declaring the name of an external
7479 symbol named @var{name} which is referenced in this compilation but
7480 not defined. The value of @var{decl} is the tree node for the
7483 This macro need not be defined if it does not need to output anything.
7484 The GNU assembler and most Unix assemblers don't require anything.
7487 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
7488 This target hook is a function to output to @var{asm_out_file} an assembler
7489 pseudo-op to declare a library function name external. The name of the
7490 library function is given by @var{symref}, which is a @code{symbol_ref}.
7493 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (tree @var{decl})
7494 This target hook is a function to output to @var{asm_out_file} an assembler
7495 directive to annotate used symbol. Darwin target use .no_dead_code_strip
7499 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
7500 A C statement (sans semicolon) to output to the stdio stream
7501 @var{stream} a reference in assembler syntax to a label named
7502 @var{name}. This should add @samp{_} to the front of the name, if that
7503 is customary on your operating system, as it is in most Berkeley Unix
7504 systems. This macro is used in @code{assemble_name}.
7507 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
7508 A C statement (sans semicolon) to output a reference to
7509 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
7510 will be used to output the name of the symbol. This macro may be used
7511 to modify the way a symbol is referenced depending on information
7512 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
7515 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
7516 A C statement (sans semicolon) to output a reference to @var{buf}, the
7517 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
7518 @code{assemble_name} will be used to output the name of the symbol.
7519 This macro is not used by @code{output_asm_label}, or the @code{%l}
7520 specifier that calls it; the intention is that this macro should be set
7521 when it is necessary to output a label differently when its address is
7525 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
7526 A function to output to the stdio stream @var{stream} a label whose
7527 name is made from the string @var{prefix} and the number @var{labelno}.
7529 It is absolutely essential that these labels be distinct from the labels
7530 used for user-level functions and variables. Otherwise, certain programs
7531 will have name conflicts with internal labels.
7533 It is desirable to exclude internal labels from the symbol table of the
7534 object file. Most assemblers have a naming convention for labels that
7535 should be excluded; on many systems, the letter @samp{L} at the
7536 beginning of a label has this effect. You should find out what
7537 convention your system uses, and follow it.
7539 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
7542 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
7543 A C statement to output to the stdio stream @var{stream} a debug info
7544 label whose name is made from the string @var{prefix} and the number
7545 @var{num}. This is useful for VLIW targets, where debug info labels
7546 may need to be treated differently than branch target labels. On some
7547 systems, branch target labels must be at the beginning of instruction
7548 bundles, but debug info labels can occur in the middle of instruction
7551 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
7555 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
7556 A C statement to store into the string @var{string} a label whose name
7557 is made from the string @var{prefix} and the number @var{num}.
7559 This string, when output subsequently by @code{assemble_name}, should
7560 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
7561 with the same @var{prefix} and @var{num}.
7563 If the string begins with @samp{*}, then @code{assemble_name} will
7564 output the rest of the string unchanged. It is often convenient for
7565 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
7566 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
7567 to output the string, and may change it. (Of course,
7568 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
7569 you should know what it does on your machine.)
7572 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
7573 A C expression to assign to @var{outvar} (which is a variable of type
7574 @code{char *}) a newly allocated string made from the string
7575 @var{name} and the number @var{number}, with some suitable punctuation
7576 added. Use @code{alloca} to get space for the string.
7578 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
7579 produce an assembler label for an internal static variable whose name is
7580 @var{name}. Therefore, the string must be such as to result in valid
7581 assembler code. The argument @var{number} is different each time this
7582 macro is executed; it prevents conflicts between similarly-named
7583 internal static variables in different scopes.
7585 Ideally this string should not be a valid C identifier, to prevent any
7586 conflict with the user's own symbols. Most assemblers allow periods
7587 or percent signs in assembler symbols; putting at least one of these
7588 between the name and the number will suffice.
7590 If this macro is not defined, a default definition will be provided
7591 which is correct for most systems.
7594 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
7595 A C statement to output to the stdio stream @var{stream} assembler code
7596 which defines (equates) the symbol @var{name} to have the value @var{value}.
7599 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7600 correct for most systems.
7603 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
7604 A C statement to output to the stdio stream @var{stream} assembler code
7605 which defines (equates) the symbol whose tree node is @var{decl_of_name}
7606 to have the value of the tree node @var{decl_of_value}. This macro will
7607 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
7608 the tree nodes are available.
7611 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7612 correct for most systems.
7615 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
7616 A C statement that evaluates to true if the assembler code which defines
7617 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
7618 of the tree node @var{decl_of_value} should be emitted near the end of the
7619 current compilation unit. The default is to not defer output of defines.
7620 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
7621 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
7624 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
7625 A C statement to output to the stdio stream @var{stream} assembler code
7626 which defines (equates) the weak symbol @var{name} to have the value
7627 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
7628 an undefined weak symbol.
7630 Define this macro if the target only supports weak aliases; define
7631 @code{ASM_OUTPUT_DEF} instead if possible.
7634 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
7635 Define this macro to override the default assembler names used for
7636 Objective-C methods.
7638 The default name is a unique method number followed by the name of the
7639 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
7640 the category is also included in the assembler name (e.g.@:
7643 These names are safe on most systems, but make debugging difficult since
7644 the method's selector is not present in the name. Therefore, particular
7645 systems define other ways of computing names.
7647 @var{buf} is an expression of type @code{char *} which gives you a
7648 buffer in which to store the name; its length is as long as
7649 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
7650 50 characters extra.
7652 The argument @var{is_inst} specifies whether the method is an instance
7653 method or a class method; @var{class_name} is the name of the class;
7654 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
7655 in a category); and @var{sel_name} is the name of the selector.
7657 On systems where the assembler can handle quoted names, you can use this
7658 macro to provide more human-readable names.
7661 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
7662 A C statement (sans semicolon) to output to the stdio stream
7663 @var{stream} commands to declare that the label @var{name} is an
7664 Objective-C class reference. This is only needed for targets whose
7665 linkers have special support for NeXT-style runtimes.
7668 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
7669 A C statement (sans semicolon) to output to the stdio stream
7670 @var{stream} commands to declare that the label @var{name} is an
7671 unresolved Objective-C class reference. This is only needed for targets
7672 whose linkers have special support for NeXT-style runtimes.
7675 @node Initialization
7676 @subsection How Initialization Functions Are Handled
7677 @cindex initialization routines
7678 @cindex termination routines
7679 @cindex constructors, output of
7680 @cindex destructors, output of
7682 The compiled code for certain languages includes @dfn{constructors}
7683 (also called @dfn{initialization routines})---functions to initialize
7684 data in the program when the program is started. These functions need
7685 to be called before the program is ``started''---that is to say, before
7686 @code{main} is called.
7688 Compiling some languages generates @dfn{destructors} (also called
7689 @dfn{termination routines}) that should be called when the program
7692 To make the initialization and termination functions work, the compiler
7693 must output something in the assembler code to cause those functions to
7694 be called at the appropriate time. When you port the compiler to a new
7695 system, you need to specify how to do this.
7697 There are two major ways that GCC currently supports the execution of
7698 initialization and termination functions. Each way has two variants.
7699 Much of the structure is common to all four variations.
7701 @findex __CTOR_LIST__
7702 @findex __DTOR_LIST__
7703 The linker must build two lists of these functions---a list of
7704 initialization functions, called @code{__CTOR_LIST__}, and a list of
7705 termination functions, called @code{__DTOR_LIST__}.
7707 Each list always begins with an ignored function pointer (which may hold
7708 0, @minus{}1, or a count of the function pointers after it, depending on
7709 the environment). This is followed by a series of zero or more function
7710 pointers to constructors (or destructors), followed by a function
7711 pointer containing zero.
7713 Depending on the operating system and its executable file format, either
7714 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
7715 time and exit time. Constructors are called in reverse order of the
7716 list; destructors in forward order.
7718 The best way to handle static constructors works only for object file
7719 formats which provide arbitrarily-named sections. A section is set
7720 aside for a list of constructors, and another for a list of destructors.
7721 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
7722 object file that defines an initialization function also puts a word in
7723 the constructor section to point to that function. The linker
7724 accumulates all these words into one contiguous @samp{.ctors} section.
7725 Termination functions are handled similarly.
7727 This method will be chosen as the default by @file{target-def.h} if
7728 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
7729 support arbitrary sections, but does support special designated
7730 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
7731 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
7733 When arbitrary sections are available, there are two variants, depending
7734 upon how the code in @file{crtstuff.c} is called. On systems that
7735 support a @dfn{.init} section which is executed at program startup,
7736 parts of @file{crtstuff.c} are compiled into that section. The
7737 program is linked by the @command{gcc} driver like this:
7740 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
7743 The prologue of a function (@code{__init}) appears in the @code{.init}
7744 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
7745 for the function @code{__fini} in the @dfn{.fini} section. Normally these
7746 files are provided by the operating system or by the GNU C library, but
7747 are provided by GCC for a few targets.
7749 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
7750 compiled from @file{crtstuff.c}. They contain, among other things, code
7751 fragments within the @code{.init} and @code{.fini} sections that branch
7752 to routines in the @code{.text} section. The linker will pull all parts
7753 of a section together, which results in a complete @code{__init} function
7754 that invokes the routines we need at startup.
7756 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
7759 If no init section is available, when GCC compiles any function called
7760 @code{main} (or more accurately, any function designated as a program
7761 entry point by the language front end calling @code{expand_main_function}),
7762 it inserts a procedure call to @code{__main} as the first executable code
7763 after the function prologue. The @code{__main} function is defined
7764 in @file{libgcc2.c} and runs the global constructors.
7766 In file formats that don't support arbitrary sections, there are again
7767 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
7768 and an `a.out' format must be used. In this case,
7769 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
7770 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
7771 and with the address of the void function containing the initialization
7772 code as its value. The GNU linker recognizes this as a request to add
7773 the value to a @dfn{set}; the values are accumulated, and are eventually
7774 placed in the executable as a vector in the format described above, with
7775 a leading (ignored) count and a trailing zero element.
7776 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
7777 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
7778 the compilation of @code{main} to call @code{__main} as above, starting
7779 the initialization process.
7781 The last variant uses neither arbitrary sections nor the GNU linker.
7782 This is preferable when you want to do dynamic linking and when using
7783 file formats which the GNU linker does not support, such as `ECOFF'@. In
7784 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
7785 termination functions are recognized simply by their names. This requires
7786 an extra program in the linkage step, called @command{collect2}. This program
7787 pretends to be the linker, for use with GCC; it does its job by running
7788 the ordinary linker, but also arranges to include the vectors of
7789 initialization and termination functions. These functions are called
7790 via @code{__main} as described above. In order to use this method,
7791 @code{use_collect2} must be defined in the target in @file{config.gcc}.
7794 The following section describes the specific macros that control and
7795 customize the handling of initialization and termination functions.
7798 @node Macros for Initialization
7799 @subsection Macros Controlling Initialization Routines
7801 Here are the macros that control how the compiler handles initialization
7802 and termination functions:
7804 @defmac INIT_SECTION_ASM_OP
7805 If defined, a C string constant, including spacing, for the assembler
7806 operation to identify the following data as initialization code. If not
7807 defined, GCC will assume such a section does not exist. When you are
7808 using special sections for initialization and termination functions, this
7809 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7810 run the initialization functions.
7813 @defmac HAS_INIT_SECTION
7814 If defined, @code{main} will not call @code{__main} as described above.
7815 This macro should be defined for systems that control start-up code
7816 on a symbol-by-symbol basis, such as OSF/1, and should not
7817 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7820 @defmac LD_INIT_SWITCH
7821 If defined, a C string constant for a switch that tells the linker that
7822 the following symbol is an initialization routine.
7825 @defmac LD_FINI_SWITCH
7826 If defined, a C string constant for a switch that tells the linker that
7827 the following symbol is a finalization routine.
7830 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7831 If defined, a C statement that will write a function that can be
7832 automatically called when a shared library is loaded. The function
7833 should call @var{func}, which takes no arguments. If not defined, and
7834 the object format requires an explicit initialization function, then a
7835 function called @code{_GLOBAL__DI} will be generated.
7837 This function and the following one are used by collect2 when linking a
7838 shared library that needs constructors or destructors, or has DWARF2
7839 exception tables embedded in the code.
7842 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7843 If defined, a C statement that will write a function that can be
7844 automatically called when a shared library is unloaded. The function
7845 should call @var{func}, which takes no arguments. If not defined, and
7846 the object format requires an explicit finalization function, then a
7847 function called @code{_GLOBAL__DD} will be generated.
7850 @defmac INVOKE__main
7851 If defined, @code{main} will call @code{__main} despite the presence of
7852 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7853 where the init section is not actually run automatically, but is still
7854 useful for collecting the lists of constructors and destructors.
7857 @defmac SUPPORTS_INIT_PRIORITY
7858 If nonzero, the C++ @code{init_priority} attribute is supported and the
7859 compiler should emit instructions to control the order of initialization
7860 of objects. If zero, the compiler will issue an error message upon
7861 encountering an @code{init_priority} attribute.
7864 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7865 This value is true if the target supports some ``native'' method of
7866 collecting constructors and destructors to be run at startup and exit.
7867 It is false if we must use @command{collect2}.
7870 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7871 If defined, a function that outputs assembler code to arrange to call
7872 the function referenced by @var{symbol} at initialization time.
7874 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7875 no arguments and with no return value. If the target supports initialization
7876 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7877 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7879 If this macro is not defined by the target, a suitable default will
7880 be chosen if (1) the target supports arbitrary section names, (2) the
7881 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7885 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7886 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7887 functions rather than initialization functions.
7890 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7891 generated for the generated object file will have static linkage.
7893 If your system uses @command{collect2} as the means of processing
7894 constructors, then that program normally uses @command{nm} to scan
7895 an object file for constructor functions to be called.
7897 On certain kinds of systems, you can define this macro to make
7898 @command{collect2} work faster (and, in some cases, make it work at all):
7900 @defmac OBJECT_FORMAT_COFF
7901 Define this macro if the system uses COFF (Common Object File Format)
7902 object files, so that @command{collect2} can assume this format and scan
7903 object files directly for dynamic constructor/destructor functions.
7905 This macro is effective only in a native compiler; @command{collect2} as
7906 part of a cross compiler always uses @command{nm} for the target machine.
7909 @defmac REAL_NM_FILE_NAME
7910 Define this macro as a C string constant containing the file name to use
7911 to execute @command{nm}. The default is to search the path normally for
7914 If your system supports shared libraries and has a program to list the
7915 dynamic dependencies of a given library or executable, you can define
7916 these macros to enable support for running initialization and
7917 termination functions in shared libraries:
7921 Define this macro to a C string constant containing the name of the program
7922 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7925 @defmac PARSE_LDD_OUTPUT (@var{ptr})
7926 Define this macro to be C code that extracts filenames from the output
7927 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7928 of type @code{char *} that points to the beginning of a line of output
7929 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7930 code must advance @var{ptr} to the beginning of the filename on that
7931 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7934 @node Instruction Output
7935 @subsection Output of Assembler Instructions
7937 @c prevent bad page break with this line
7938 This describes assembler instruction output.
7940 @defmac REGISTER_NAMES
7941 A C initializer containing the assembler's names for the machine
7942 registers, each one as a C string constant. This is what translates
7943 register numbers in the compiler into assembler language.
7946 @defmac ADDITIONAL_REGISTER_NAMES
7947 If defined, a C initializer for an array of structures containing a name
7948 and a register number. This macro defines additional names for hard
7949 registers, thus allowing the @code{asm} option in declarations to refer
7950 to registers using alternate names.
7953 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7954 Define this macro if you are using an unusual assembler that
7955 requires different names for the machine instructions.
7957 The definition is a C statement or statements which output an
7958 assembler instruction opcode to the stdio stream @var{stream}. The
7959 macro-operand @var{ptr} is a variable of type @code{char *} which
7960 points to the opcode name in its ``internal'' form---the form that is
7961 written in the machine description. The definition should output the
7962 opcode name to @var{stream}, performing any translation you desire, and
7963 increment the variable @var{ptr} to point at the end of the opcode
7964 so that it will not be output twice.
7966 In fact, your macro definition may process less than the entire opcode
7967 name, or more than the opcode name; but if you want to process text
7968 that includes @samp{%}-sequences to substitute operands, you must take
7969 care of the substitution yourself. Just be sure to increment
7970 @var{ptr} over whatever text should not be output normally.
7972 @findex recog_data.operand
7973 If you need to look at the operand values, they can be found as the
7974 elements of @code{recog_data.operand}.
7976 If the macro definition does nothing, the instruction is output
7980 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7981 If defined, a C statement to be executed just prior to the output of
7982 assembler code for @var{insn}, to modify the extracted operands so
7983 they will be output differently.
7985 Here the argument @var{opvec} is the vector containing the operands
7986 extracted from @var{insn}, and @var{noperands} is the number of
7987 elements of the vector which contain meaningful data for this insn.
7988 The contents of this vector are what will be used to convert the insn
7989 template into assembler code, so you can change the assembler output
7990 by changing the contents of the vector.
7992 This macro is useful when various assembler syntaxes share a single
7993 file of instruction patterns; by defining this macro differently, you
7994 can cause a large class of instructions to be output differently (such
7995 as with rearranged operands). Naturally, variations in assembler
7996 syntax affecting individual insn patterns ought to be handled by
7997 writing conditional output routines in those patterns.
7999 If this macro is not defined, it is equivalent to a null statement.
8002 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
8003 A C compound statement to output to stdio stream @var{stream} the
8004 assembler syntax for an instruction operand @var{x}. @var{x} is an
8007 @var{code} is a value that can be used to specify one of several ways
8008 of printing the operand. It is used when identical operands must be
8009 printed differently depending on the context. @var{code} comes from
8010 the @samp{%} specification that was used to request printing of the
8011 operand. If the specification was just @samp{%@var{digit}} then
8012 @var{code} is 0; if the specification was @samp{%@var{ltr}
8013 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
8016 If @var{x} is a register, this macro should print the register's name.
8017 The names can be found in an array @code{reg_names} whose type is
8018 @code{char *[]}. @code{reg_names} is initialized from
8019 @code{REGISTER_NAMES}.
8021 When the machine description has a specification @samp{%@var{punct}}
8022 (a @samp{%} followed by a punctuation character), this macro is called
8023 with a null pointer for @var{x} and the punctuation character for
8027 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
8028 A C expression which evaluates to true if @var{code} is a valid
8029 punctuation character for use in the @code{PRINT_OPERAND} macro. If
8030 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
8031 punctuation characters (except for the standard one, @samp{%}) are used
8035 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
8036 A C compound statement to output to stdio stream @var{stream} the
8037 assembler syntax for an instruction operand that is a memory reference
8038 whose address is @var{x}. @var{x} is an RTL expression.
8040 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
8041 On some machines, the syntax for a symbolic address depends on the
8042 section that the address refers to. On these machines, define the hook
8043 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
8044 @code{symbol_ref}, and then check for it here. @xref{Assembler
8048 @findex dbr_sequence_length
8049 @defmac DBR_OUTPUT_SEQEND (@var{file})
8050 A C statement, to be executed after all slot-filler instructions have
8051 been output. If necessary, call @code{dbr_sequence_length} to
8052 determine the number of slots filled in a sequence (zero if not
8053 currently outputting a sequence), to decide how many no-ops to output,
8056 Don't define this macro if it has nothing to do, but it is helpful in
8057 reading assembly output if the extent of the delay sequence is made
8058 explicit (e.g.@: with white space).
8061 @findex final_sequence
8062 Note that output routines for instructions with delay slots must be
8063 prepared to deal with not being output as part of a sequence
8064 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
8065 found.) The variable @code{final_sequence} is null when not
8066 processing a sequence, otherwise it contains the @code{sequence} rtx
8070 @defmac REGISTER_PREFIX
8071 @defmacx LOCAL_LABEL_PREFIX
8072 @defmacx USER_LABEL_PREFIX
8073 @defmacx IMMEDIATE_PREFIX
8074 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
8075 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
8076 @file{final.c}). These are useful when a single @file{md} file must
8077 support multiple assembler formats. In that case, the various @file{tm.h}
8078 files can define these macros differently.
8081 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
8082 If defined this macro should expand to a series of @code{case}
8083 statements which will be parsed inside the @code{switch} statement of
8084 the @code{asm_fprintf} function. This allows targets to define extra
8085 printf formats which may useful when generating their assembler
8086 statements. Note that uppercase letters are reserved for future
8087 generic extensions to asm_fprintf, and so are not available to target
8088 specific code. The output file is given by the parameter @var{file}.
8089 The varargs input pointer is @var{argptr} and the rest of the format
8090 string, starting the character after the one that is being switched
8091 upon, is pointed to by @var{format}.
8094 @defmac ASSEMBLER_DIALECT
8095 If your target supports multiple dialects of assembler language (such as
8096 different opcodes), define this macro as a C expression that gives the
8097 numeric index of the assembler language dialect to use, with zero as the
8100 If this macro is defined, you may use constructs of the form
8102 @samp{@{option0|option1|option2@dots{}@}}
8105 in the output templates of patterns (@pxref{Output Template}) or in the
8106 first argument of @code{asm_fprintf}. This construct outputs
8107 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
8108 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
8109 within these strings retain their usual meaning. If there are fewer
8110 alternatives within the braces than the value of
8111 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
8113 If you do not define this macro, the characters @samp{@{}, @samp{|} and
8114 @samp{@}} do not have any special meaning when used in templates or
8115 operands to @code{asm_fprintf}.
8117 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
8118 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
8119 the variations in assembler language syntax with that mechanism. Define
8120 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
8121 if the syntax variant are larger and involve such things as different
8122 opcodes or operand order.
8125 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
8126 A C expression to output to @var{stream} some assembler code
8127 which will push hard register number @var{regno} onto the stack.
8128 The code need not be optimal, since this macro is used only when
8132 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
8133 A C expression to output to @var{stream} some assembler code
8134 which will pop hard register number @var{regno} off of the stack.
8135 The code need not be optimal, since this macro is used only when
8139 @node Dispatch Tables
8140 @subsection Output of Dispatch Tables
8142 @c prevent bad page break with this line
8143 This concerns dispatch tables.
8145 @cindex dispatch table
8146 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
8147 A C statement to output to the stdio stream @var{stream} an assembler
8148 pseudo-instruction to generate a difference between two labels.
8149 @var{value} and @var{rel} are the numbers of two internal labels. The
8150 definitions of these labels are output using
8151 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
8152 way here. For example,
8155 fprintf (@var{stream}, "\t.word L%d-L%d\n",
8156 @var{value}, @var{rel})
8159 You must provide this macro on machines where the addresses in a
8160 dispatch table are relative to the table's own address. If defined, GCC
8161 will also use this macro on all machines when producing PIC@.
8162 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
8163 mode and flags can be read.
8166 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
8167 This macro should be provided on machines where the addresses
8168 in a dispatch table are absolute.
8170 The definition should be a C statement to output to the stdio stream
8171 @var{stream} an assembler pseudo-instruction to generate a reference to
8172 a label. @var{value} is the number of an internal label whose
8173 definition is output using @code{(*targetm.asm_out.internal_label)}.
8177 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
8181 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
8182 Define this if the label before a jump-table needs to be output
8183 specially. The first three arguments are the same as for
8184 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
8185 jump-table which follows (a @code{jump_insn} containing an
8186 @code{addr_vec} or @code{addr_diff_vec}).
8188 This feature is used on system V to output a @code{swbeg} statement
8191 If this macro is not defined, these labels are output with
8192 @code{(*targetm.asm_out.internal_label)}.
8195 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
8196 Define this if something special must be output at the end of a
8197 jump-table. The definition should be a C statement to be executed
8198 after the assembler code for the table is written. It should write
8199 the appropriate code to stdio stream @var{stream}. The argument
8200 @var{table} is the jump-table insn, and @var{num} is the label-number
8201 of the preceding label.
8203 If this macro is not defined, nothing special is output at the end of
8207 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (@var{stream}, @var{decl}, @var{for_eh}, @var{empty})
8208 This target hook emits a label at the beginning of each FDE@. It
8209 should be defined on targets where FDEs need special labels, and it
8210 should write the appropriate label, for the FDE associated with the
8211 function declaration @var{decl}, to the stdio stream @var{stream}.
8212 The third argument, @var{for_eh}, is a boolean: true if this is for an
8213 exception table. The fourth argument, @var{empty}, is a boolean:
8214 true if this is a placeholder label for an omitted FDE@.
8216 The default is that FDEs are not given nonlocal labels.
8219 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (@var{stream})
8220 This target hook emits a label at the beginning of the exception table.
8221 It should be defined on targets where it is desirable for the table
8222 to be broken up according to function.
8224 The default is that no label is emitted.
8227 @deftypefn {Target Hook} void TARGET_UNWIND_EMIT (FILE * @var{stream}, rtx @var{insn})
8228 This target hook emits and assembly directives required to unwind the
8229 given instruction. This is only used when TARGET_UNWIND_INFO is set.
8232 @node Exception Region Output
8233 @subsection Assembler Commands for Exception Regions
8235 @c prevent bad page break with this line
8237 This describes commands marking the start and the end of an exception
8240 @defmac EH_FRAME_SECTION_NAME
8241 If defined, a C string constant for the name of the section containing
8242 exception handling frame unwind information. If not defined, GCC will
8243 provide a default definition if the target supports named sections.
8244 @file{crtstuff.c} uses this macro to switch to the appropriate section.
8246 You should define this symbol if your target supports DWARF 2 frame
8247 unwind information and the default definition does not work.
8250 @defmac EH_FRAME_IN_DATA_SECTION
8251 If defined, DWARF 2 frame unwind information will be placed in the
8252 data section even though the target supports named sections. This
8253 might be necessary, for instance, if the system linker does garbage
8254 collection and sections cannot be marked as not to be collected.
8256 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
8260 @defmac EH_TABLES_CAN_BE_READ_ONLY
8261 Define this macro to 1 if your target is such that no frame unwind
8262 information encoding used with non-PIC code will ever require a
8263 runtime relocation, but the linker may not support merging read-only
8264 and read-write sections into a single read-write section.
8267 @defmac MASK_RETURN_ADDR
8268 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
8269 that it does not contain any extraneous set bits in it.
8272 @defmac DWARF2_UNWIND_INFO
8273 Define this macro to 0 if your target supports DWARF 2 frame unwind
8274 information, but it does not yet work with exception handling.
8275 Otherwise, if your target supports this information (if it defines
8276 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
8277 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of 1.
8279 If @code{TARGET_UNWIND_INFO} is defined, the target specific unwinder
8280 will be used in all cases. Defining this macro will enable the generation
8281 of DWARF 2 frame debugging information.
8283 If @code{TARGET_UNWIND_INFO} is not defined, and this macro is defined to 1,
8284 the DWARF 2 unwinder will be the default exception handling mechanism;
8285 otherwise, the @code{setjmp}/@code{longjmp}-based scheme will be used by
8289 @defmac TARGET_UNWIND_INFO
8290 Define this macro if your target has ABI specified unwind tables. Usually
8291 these will be output by @code{TARGET_UNWIND_EMIT}.
8294 @deftypevar {Target Hook} bool TARGET_UNWIND_TABLES_DEFAULT
8295 This variable should be set to @code{true} if the target ABI requires unwinding
8296 tables even when exceptions are not used.
8299 @defmac MUST_USE_SJLJ_EXCEPTIONS
8300 This macro need only be defined if @code{DWARF2_UNWIND_INFO} is
8301 runtime-variable. In that case, @file{except.h} cannot correctly
8302 determine the corresponding definition of @code{MUST_USE_SJLJ_EXCEPTIONS},
8303 so the target must provide it directly.
8306 @defmac DONT_USE_BUILTIN_SETJMP
8307 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
8308 should use the @code{setjmp}/@code{longjmp} functions from the C library
8309 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
8312 @defmac DWARF_CIE_DATA_ALIGNMENT
8313 This macro need only be defined if the target might save registers in the
8314 function prologue at an offset to the stack pointer that is not aligned to
8315 @code{UNITS_PER_WORD}. The definition should be the negative minimum
8316 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
8317 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
8318 the target supports DWARF 2 frame unwind information.
8321 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
8322 Contains the value true if the target should add a zero word onto the
8323 end of a Dwarf-2 frame info section when used for exception handling.
8324 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
8328 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
8329 Given a register, this hook should return a parallel of registers to
8330 represent where to find the register pieces. Define this hook if the
8331 register and its mode are represented in Dwarf in non-contiguous
8332 locations, or if the register should be represented in more than one
8333 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
8334 If not defined, the default is to return @code{NULL_RTX}.
8337 @deftypefn {Target Hook} void TARGET_INIT_DWARF_REG_SIZES_EXTRA (tree @var{address})
8338 If some registers are represented in Dwarf-2 unwind information in
8339 multiple pieces, define this hook to fill in information about the
8340 sizes of those pieces in the table used by the unwinder at runtime.
8341 It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after
8342 filling in a single size corresponding to each hard register;
8343 @var{address} is the address of the table.
8346 @deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym})
8347 This hook is used to output a reference from a frame unwinding table to
8348 the type_info object identified by @var{sym}. It should return @code{true}
8349 if the reference was output. Returning @code{false} will cause the
8350 reference to be output using the normal Dwarf2 routines.
8353 @deftypefn {Target Hook} bool TARGET_ARM_EABI_UNWINDER
8354 This hook should be set to @code{true} on targets that use an ARM EABI
8355 based unwinding library, and @code{false} on other targets. This effects
8356 the format of unwinding tables, and how the unwinder in entered after
8357 running a cleanup. The default is @code{false}.
8360 @node Alignment Output
8361 @subsection Assembler Commands for Alignment
8363 @c prevent bad page break with this line
8364 This describes commands for alignment.
8366 @defmac JUMP_ALIGN (@var{label})
8367 The alignment (log base 2) to put in front of @var{label}, which is
8368 a common destination of jumps and has no fallthru incoming edge.
8370 This macro need not be defined if you don't want any special alignment
8371 to be done at such a time. Most machine descriptions do not currently
8374 Unless it's necessary to inspect the @var{label} parameter, it is better
8375 to set the variable @var{align_jumps} in the target's
8376 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8377 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
8380 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
8381 The alignment (log base 2) to put in front of @var{label}, which follows
8384 This macro need not be defined if you don't want any special alignment
8385 to be done at such a time. Most machine descriptions do not currently
8389 @defmac LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
8390 The maximum number of bytes to skip when applying
8391 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
8392 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8395 @defmac LOOP_ALIGN (@var{label})
8396 The alignment (log base 2) to put in front of @var{label}, which follows
8397 a @code{NOTE_INSN_LOOP_BEG} note.
8399 This macro need not be defined if you don't want any special alignment
8400 to be done at such a time. Most machine descriptions do not currently
8403 Unless it's necessary to inspect the @var{label} parameter, it is better
8404 to set the variable @code{align_loops} in the target's
8405 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8406 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
8409 @defmac LOOP_ALIGN_MAX_SKIP
8410 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
8411 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8414 @defmac LABEL_ALIGN (@var{label})
8415 The alignment (log base 2) to put in front of @var{label}.
8416 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
8417 the maximum of the specified values is used.
8419 Unless it's necessary to inspect the @var{label} parameter, it is better
8420 to set the variable @code{align_labels} in the target's
8421 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8422 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
8425 @defmac LABEL_ALIGN_MAX_SKIP
8426 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
8427 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8430 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
8431 A C statement to output to the stdio stream @var{stream} an assembler
8432 instruction to advance the location counter by @var{nbytes} bytes.
8433 Those bytes should be zero when loaded. @var{nbytes} will be a C
8434 expression of type @code{unsigned HOST_WIDE_INT}.
8437 @defmac ASM_NO_SKIP_IN_TEXT
8438 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
8439 text section because it fails to put zeros in the bytes that are skipped.
8440 This is true on many Unix systems, where the pseudo--op to skip bytes
8441 produces no-op instructions rather than zeros when used in the text
8445 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
8446 A C statement to output to the stdio stream @var{stream} an assembler
8447 command to advance the location counter to a multiple of 2 to the
8448 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
8451 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
8452 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
8453 for padding, if necessary.
8456 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
8457 A C statement to output to the stdio stream @var{stream} an assembler
8458 command to advance the location counter to a multiple of 2 to the
8459 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
8460 satisfy the alignment request. @var{power} and @var{max_skip} will be
8461 a C expression of type @code{int}.
8465 @node Debugging Info
8466 @section Controlling Debugging Information Format
8468 @c prevent bad page break with this line
8469 This describes how to specify debugging information.
8472 * All Debuggers:: Macros that affect all debugging formats uniformly.
8473 * DBX Options:: Macros enabling specific options in DBX format.
8474 * DBX Hooks:: Hook macros for varying DBX format.
8475 * File Names and DBX:: Macros controlling output of file names in DBX format.
8476 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
8477 * VMS Debug:: Macros for VMS debug format.
8481 @subsection Macros Affecting All Debugging Formats
8483 @c prevent bad page break with this line
8484 These macros affect all debugging formats.
8486 @defmac DBX_REGISTER_NUMBER (@var{regno})
8487 A C expression that returns the DBX register number for the compiler
8488 register number @var{regno}. In the default macro provided, the value
8489 of this expression will be @var{regno} itself. But sometimes there are
8490 some registers that the compiler knows about and DBX does not, or vice
8491 versa. In such cases, some register may need to have one number in the
8492 compiler and another for DBX@.
8494 If two registers have consecutive numbers inside GCC, and they can be
8495 used as a pair to hold a multiword value, then they @emph{must} have
8496 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
8497 Otherwise, debuggers will be unable to access such a pair, because they
8498 expect register pairs to be consecutive in their own numbering scheme.
8500 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
8501 does not preserve register pairs, then what you must do instead is
8502 redefine the actual register numbering scheme.
8505 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
8506 A C expression that returns the integer offset value for an automatic
8507 variable having address @var{x} (an RTL expression). The default
8508 computation assumes that @var{x} is based on the frame-pointer and
8509 gives the offset from the frame-pointer. This is required for targets
8510 that produce debugging output for DBX or COFF-style debugging output
8511 for SDB and allow the frame-pointer to be eliminated when the
8512 @option{-g} options is used.
8515 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
8516 A C expression that returns the integer offset value for an argument
8517 having address @var{x} (an RTL expression). The nominal offset is
8521 @defmac PREFERRED_DEBUGGING_TYPE
8522 A C expression that returns the type of debugging output GCC should
8523 produce when the user specifies just @option{-g}. Define
8524 this if you have arranged for GCC to support more than one format of
8525 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
8526 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
8527 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
8529 When the user specifies @option{-ggdb}, GCC normally also uses the
8530 value of this macro to select the debugging output format, but with two
8531 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
8532 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
8533 defined, GCC uses @code{DBX_DEBUG}.
8535 The value of this macro only affects the default debugging output; the
8536 user can always get a specific type of output by using @option{-gstabs},
8537 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
8541 @subsection Specific Options for DBX Output
8543 @c prevent bad page break with this line
8544 These are specific options for DBX output.
8546 @defmac DBX_DEBUGGING_INFO
8547 Define this macro if GCC should produce debugging output for DBX
8548 in response to the @option{-g} option.
8551 @defmac XCOFF_DEBUGGING_INFO
8552 Define this macro if GCC should produce XCOFF format debugging output
8553 in response to the @option{-g} option. This is a variant of DBX format.
8556 @defmac DEFAULT_GDB_EXTENSIONS
8557 Define this macro to control whether GCC should by default generate
8558 GDB's extended version of DBX debugging information (assuming DBX-format
8559 debugging information is enabled at all). If you don't define the
8560 macro, the default is 1: always generate the extended information
8561 if there is any occasion to.
8564 @defmac DEBUG_SYMS_TEXT
8565 Define this macro if all @code{.stabs} commands should be output while
8566 in the text section.
8569 @defmac ASM_STABS_OP
8570 A C string constant, including spacing, naming the assembler pseudo op to
8571 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
8572 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
8573 applies only to DBX debugging information format.
8576 @defmac ASM_STABD_OP
8577 A C string constant, including spacing, naming the assembler pseudo op to
8578 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
8579 value is the current location. If you don't define this macro,
8580 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
8584 @defmac ASM_STABN_OP
8585 A C string constant, including spacing, naming the assembler pseudo op to
8586 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
8587 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
8588 macro applies only to DBX debugging information format.
8591 @defmac DBX_NO_XREFS
8592 Define this macro if DBX on your system does not support the construct
8593 @samp{xs@var{tagname}}. On some systems, this construct is used to
8594 describe a forward reference to a structure named @var{tagname}.
8595 On other systems, this construct is not supported at all.
8598 @defmac DBX_CONTIN_LENGTH
8599 A symbol name in DBX-format debugging information is normally
8600 continued (split into two separate @code{.stabs} directives) when it
8601 exceeds a certain length (by default, 80 characters). On some
8602 operating systems, DBX requires this splitting; on others, splitting
8603 must not be done. You can inhibit splitting by defining this macro
8604 with the value zero. You can override the default splitting-length by
8605 defining this macro as an expression for the length you desire.
8608 @defmac DBX_CONTIN_CHAR
8609 Normally continuation is indicated by adding a @samp{\} character to
8610 the end of a @code{.stabs} string when a continuation follows. To use
8611 a different character instead, define this macro as a character
8612 constant for the character you want to use. Do not define this macro
8613 if backslash is correct for your system.
8616 @defmac DBX_STATIC_STAB_DATA_SECTION
8617 Define this macro if it is necessary to go to the data section before
8618 outputting the @samp{.stabs} pseudo-op for a non-global static
8622 @defmac DBX_TYPE_DECL_STABS_CODE
8623 The value to use in the ``code'' field of the @code{.stabs} directive
8624 for a typedef. The default is @code{N_LSYM}.
8627 @defmac DBX_STATIC_CONST_VAR_CODE
8628 The value to use in the ``code'' field of the @code{.stabs} directive
8629 for a static variable located in the text section. DBX format does not
8630 provide any ``right'' way to do this. The default is @code{N_FUN}.
8633 @defmac DBX_REGPARM_STABS_CODE
8634 The value to use in the ``code'' field of the @code{.stabs} directive
8635 for a parameter passed in registers. DBX format does not provide any
8636 ``right'' way to do this. The default is @code{N_RSYM}.
8639 @defmac DBX_REGPARM_STABS_LETTER
8640 The letter to use in DBX symbol data to identify a symbol as a parameter
8641 passed in registers. DBX format does not customarily provide any way to
8642 do this. The default is @code{'P'}.
8645 @defmac DBX_FUNCTION_FIRST
8646 Define this macro if the DBX information for a function and its
8647 arguments should precede the assembler code for the function. Normally,
8648 in DBX format, the debugging information entirely follows the assembler
8652 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
8653 Define this macro, with value 1, if the value of a symbol describing
8654 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
8655 relative to the start of the enclosing function. Normally, GCC uses
8656 an absolute address.
8659 @defmac DBX_LINES_FUNCTION_RELATIVE
8660 Define this macro, with value 1, if the value of a symbol indicating
8661 the current line number (@code{N_SLINE}) should be relative to the
8662 start of the enclosing function. Normally, GCC uses an absolute address.
8665 @defmac DBX_USE_BINCL
8666 Define this macro if GCC should generate @code{N_BINCL} and
8667 @code{N_EINCL} stabs for included header files, as on Sun systems. This
8668 macro also directs GCC to output a type number as a pair of a file
8669 number and a type number within the file. Normally, GCC does not
8670 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
8671 number for a type number.
8675 @subsection Open-Ended Hooks for DBX Format
8677 @c prevent bad page break with this line
8678 These are hooks for DBX format.
8680 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
8681 Define this macro to say how to output to @var{stream} the debugging
8682 information for the start of a scope level for variable names. The
8683 argument @var{name} is the name of an assembler symbol (for use with
8684 @code{assemble_name}) whose value is the address where the scope begins.
8687 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
8688 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
8691 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
8692 Define this macro if the target machine requires special handling to
8693 output an @code{N_FUN} entry for the function @var{decl}.
8696 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
8697 A C statement to output DBX debugging information before code for line
8698 number @var{line} of the current source file to the stdio stream
8699 @var{stream}. @var{counter} is the number of time the macro was
8700 invoked, including the current invocation; it is intended to generate
8701 unique labels in the assembly output.
8703 This macro should not be defined if the default output is correct, or
8704 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
8707 @defmac NO_DBX_FUNCTION_END
8708 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
8709 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
8710 On those machines, define this macro to turn this feature off without
8711 disturbing the rest of the gdb extensions.
8714 @defmac NO_DBX_BNSYM_ENSYM
8715 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
8716 extension construct. On those machines, define this macro to turn this
8717 feature off without disturbing the rest of the gdb extensions.
8720 @node File Names and DBX
8721 @subsection File Names in DBX Format
8723 @c prevent bad page break with this line
8724 This describes file names in DBX format.
8726 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
8727 A C statement to output DBX debugging information to the stdio stream
8728 @var{stream}, which indicates that file @var{name} is the main source
8729 file---the file specified as the input file for compilation.
8730 This macro is called only once, at the beginning of compilation.
8732 This macro need not be defined if the standard form of output
8733 for DBX debugging information is appropriate.
8735 It may be necessary to refer to a label equal to the beginning of the
8736 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
8737 to do so. If you do this, you must also set the variable
8738 @var{used_ltext_label_name} to @code{true}.
8741 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
8742 Define this macro, with value 1, if GCC should not emit an indication
8743 of the current directory for compilation and current source language at
8744 the beginning of the file.
8747 @defmac NO_DBX_GCC_MARKER
8748 Define this macro, with value 1, if GCC should not emit an indication
8749 that this object file was compiled by GCC@. The default is to emit
8750 an @code{N_OPT} stab at the beginning of every source file, with
8751 @samp{gcc2_compiled.} for the string and value 0.
8754 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
8755 A C statement to output DBX debugging information at the end of
8756 compilation of the main source file @var{name}. Output should be
8757 written to the stdio stream @var{stream}.
8759 If you don't define this macro, nothing special is output at the end
8760 of compilation, which is correct for most machines.
8763 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
8764 Define this macro @emph{instead of} defining
8765 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
8766 the end of compilation is a @code{N_SO} stab with an empty string,
8767 whose value is the highest absolute text address in the file.
8772 @subsection Macros for SDB and DWARF Output
8774 @c prevent bad page break with this line
8775 Here are macros for SDB and DWARF output.
8777 @defmac SDB_DEBUGGING_INFO
8778 Define this macro if GCC should produce COFF-style debugging output
8779 for SDB in response to the @option{-g} option.
8782 @defmac DWARF2_DEBUGGING_INFO
8783 Define this macro if GCC should produce dwarf version 2 format
8784 debugging output in response to the @option{-g} option.
8786 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (tree @var{function})
8787 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
8788 be emitted for each function. Instead of an integer return the enum
8789 value for the @code{DW_CC_} tag.
8792 To support optional call frame debugging information, you must also
8793 define @code{INCOMING_RETURN_ADDR_RTX} and either set
8794 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
8795 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
8796 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
8799 @defmac DWARF2_FRAME_INFO
8800 Define this macro to a nonzero value if GCC should always output
8801 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
8802 (@pxref{Exception Region Output} is nonzero, GCC will output this
8803 information not matter how you define @code{DWARF2_FRAME_INFO}.
8806 @defmac DWARF2_ASM_LINE_DEBUG_INFO
8807 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8808 line debug info sections. This will result in much more compact line number
8809 tables, and hence is desirable if it works.
8812 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8813 A C statement to issue assembly directives that create a difference
8814 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
8817 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
8818 A C statement to issue assembly directives that create a
8819 section-relative reference to the given @var{label}, using an integer of the
8820 given @var{size}. The label is known to be defined in the given @var{section}.
8823 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8824 A C statement to issue assembly directives that create a self-relative
8825 reference to the given @var{label}, using an integer of the given @var{size}.
8828 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{FILE}, int @var{size}, rtx @var{x})
8829 If defined, this target hook is a function which outputs a DTP-relative
8830 reference to the given TLS symbol of the specified size.
8833 @defmac PUT_SDB_@dots{}
8834 Define these macros to override the assembler syntax for the special
8835 SDB assembler directives. See @file{sdbout.c} for a list of these
8836 macros and their arguments. If the standard syntax is used, you need
8837 not define them yourself.
8841 Some assemblers do not support a semicolon as a delimiter, even between
8842 SDB assembler directives. In that case, define this macro to be the
8843 delimiter to use (usually @samp{\n}). It is not necessary to define
8844 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8848 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
8849 Define this macro to allow references to unknown structure,
8850 union, or enumeration tags to be emitted. Standard COFF does not
8851 allow handling of unknown references, MIPS ECOFF has support for
8855 @defmac SDB_ALLOW_FORWARD_REFERENCES
8856 Define this macro to allow references to structure, union, or
8857 enumeration tags that have not yet been seen to be handled. Some
8858 assemblers choke if forward tags are used, while some require it.
8861 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
8862 A C statement to output SDB debugging information before code for line
8863 number @var{line} of the current source file to the stdio stream
8864 @var{stream}. The default is to emit an @code{.ln} directive.
8869 @subsection Macros for VMS Debug Format
8871 @c prevent bad page break with this line
8872 Here are macros for VMS debug format.
8874 @defmac VMS_DEBUGGING_INFO
8875 Define this macro if GCC should produce debugging output for VMS
8876 in response to the @option{-g} option. The default behavior for VMS
8877 is to generate minimal debug info for a traceback in the absence of
8878 @option{-g} unless explicitly overridden with @option{-g0}. This
8879 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8880 @code{OVERRIDE_OPTIONS}.
8883 @node Floating Point
8884 @section Cross Compilation and Floating Point
8885 @cindex cross compilation and floating point
8886 @cindex floating point and cross compilation
8888 While all modern machines use twos-complement representation for integers,
8889 there are a variety of representations for floating point numbers. This
8890 means that in a cross-compiler the representation of floating point numbers
8891 in the compiled program may be different from that used in the machine
8892 doing the compilation.
8894 Because different representation systems may offer different amounts of
8895 range and precision, all floating point constants must be represented in
8896 the target machine's format. Therefore, the cross compiler cannot
8897 safely use the host machine's floating point arithmetic; it must emulate
8898 the target's arithmetic. To ensure consistency, GCC always uses
8899 emulation to work with floating point values, even when the host and
8900 target floating point formats are identical.
8902 The following macros are provided by @file{real.h} for the compiler to
8903 use. All parts of the compiler which generate or optimize
8904 floating-point calculations must use these macros. They may evaluate
8905 their operands more than once, so operands must not have side effects.
8907 @defmac REAL_VALUE_TYPE
8908 The C data type to be used to hold a floating point value in the target
8909 machine's format. Typically this is a @code{struct} containing an
8910 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8914 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8915 Compares for equality the two values, @var{x} and @var{y}. If the target
8916 floating point format supports negative zeroes and/or NaNs,
8917 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8918 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8921 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8922 Tests whether @var{x} is less than @var{y}.
8925 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8926 Truncates @var{x} to a signed integer, rounding toward zero.
8929 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8930 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8931 @var{x} is negative, returns zero.
8934 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8935 Converts @var{string} into a floating point number in the target machine's
8936 representation for mode @var{mode}. This routine can handle both
8937 decimal and hexadecimal floating point constants, using the syntax
8938 defined by the C language for both.
8941 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8942 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8945 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8946 Determines whether @var{x} represents infinity (positive or negative).
8949 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8950 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8953 @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})
8954 Calculates an arithmetic operation on the two floating point values
8955 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8958 The operation to be performed is specified by @var{code}. Only the
8959 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8960 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8962 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8963 target's floating point format cannot represent infinity, it will call
8964 @code{abort}. Callers should check for this situation first, using
8965 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8968 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8969 Returns the negative of the floating point value @var{x}.
8972 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8973 Returns the absolute value of @var{x}.
8976 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8977 Truncates the floating point value @var{x} to fit in @var{mode}. The
8978 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8979 appropriate bit pattern to be output as a floating constant whose
8980 precision accords with mode @var{mode}.
8983 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8984 Converts a floating point value @var{x} into a double-precision integer
8985 which is then stored into @var{low} and @var{high}. If the value is not
8986 integral, it is truncated.
8989 @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})
8990 Converts a double-precision integer found in @var{low} and @var{high},
8991 into a floating point value which is then stored into @var{x}. The
8992 value is truncated to fit in mode @var{mode}.
8995 @node Mode Switching
8996 @section Mode Switching Instructions
8997 @cindex mode switching
8998 The following macros control mode switching optimizations:
9000 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
9001 Define this macro if the port needs extra instructions inserted for mode
9002 switching in an optimizing compilation.
9004 For an example, the SH4 can perform both single and double precision
9005 floating point operations, but to perform a single precision operation,
9006 the FPSCR PR bit has to be cleared, while for a double precision
9007 operation, this bit has to be set. Changing the PR bit requires a general
9008 purpose register as a scratch register, hence these FPSCR sets have to
9009 be inserted before reload, i.e.@: you can't put this into instruction emitting
9010 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
9012 You can have multiple entities that are mode-switched, and select at run time
9013 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
9014 return nonzero for any @var{entity} that needs mode-switching.
9015 If you define this macro, you also have to define
9016 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
9017 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
9018 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
9022 @defmac NUM_MODES_FOR_MODE_SWITCHING
9023 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
9024 initializer for an array of integers. Each initializer element
9025 N refers to an entity that needs mode switching, and specifies the number
9026 of different modes that might need to be set for this entity.
9027 The position of the initializer in the initializer---starting counting at
9028 zero---determines the integer that is used to refer to the mode-switched
9030 In macros that take mode arguments / yield a mode result, modes are
9031 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
9032 switch is needed / supplied.
9035 @defmac MODE_NEEDED (@var{entity}, @var{insn})
9036 @var{entity} is an integer specifying a mode-switched entity. If
9037 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
9038 return an integer value not larger than the corresponding element in
9039 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
9040 be switched into prior to the execution of @var{insn}.
9043 @defmac MODE_AFTER (@var{mode}, @var{insn})
9044 If this macro is defined, it is evaluated for every @var{insn} during
9045 mode switching. It determines the mode that an insn results in (if
9046 different from the incoming mode).
9049 @defmac MODE_ENTRY (@var{entity})
9050 If this macro is defined, it is evaluated for every @var{entity} that needs
9051 mode switching. It should evaluate to an integer, which is a mode that
9052 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
9053 is defined then @code{MODE_EXIT} must be defined.
9056 @defmac MODE_EXIT (@var{entity})
9057 If this macro is defined, it is evaluated for every @var{entity} that needs
9058 mode switching. It should evaluate to an integer, which is a mode that
9059 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
9060 is defined then @code{MODE_ENTRY} must be defined.
9063 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
9064 This macro specifies the order in which modes for @var{entity} are processed.
9065 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
9066 lowest. The value of the macro should be an integer designating a mode
9067 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
9068 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
9069 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
9072 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
9073 Generate one or more insns to set @var{entity} to @var{mode}.
9074 @var{hard_reg_live} is the set of hard registers live at the point where
9075 the insn(s) are to be inserted.
9078 @node Target Attributes
9079 @section Defining target-specific uses of @code{__attribute__}
9080 @cindex target attributes
9081 @cindex machine attributes
9082 @cindex attributes, target-specific
9084 Target-specific attributes may be defined for functions, data and types.
9085 These are described using the following target hooks; they also need to
9086 be documented in @file{extend.texi}.
9088 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
9089 If defined, this target hook points to an array of @samp{struct
9090 attribute_spec} (defined in @file{tree.h}) specifying the machine
9091 specific attributes for this target and some of the restrictions on the
9092 entities to which these attributes are applied and the arguments they
9096 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
9097 If defined, this target hook is a function which returns zero if the attributes on
9098 @var{type1} and @var{type2} are incompatible, one if they are compatible,
9099 and two if they are nearly compatible (which causes a warning to be
9100 generated). If this is not defined, machine-specific attributes are
9101 supposed always to be compatible.
9104 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
9105 If defined, this target hook is a function which assigns default attributes to
9106 newly defined @var{type}.
9109 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
9110 Define this target hook if the merging of type attributes needs special
9111 handling. If defined, the result is a list of the combined
9112 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
9113 that @code{comptypes} has already been called and returned 1. This
9114 function may call @code{merge_attributes} to handle machine-independent
9118 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
9119 Define this target hook if the merging of decl attributes needs special
9120 handling. If defined, the result is a list of the combined
9121 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
9122 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
9123 when this is needed are when one attribute overrides another, or when an
9124 attribute is nullified by a subsequent definition. This function may
9125 call @code{merge_attributes} to handle machine-independent merging.
9127 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
9128 If the only target-specific handling you require is @samp{dllimport}
9129 for Microsoft Windows targets, you should define the macro
9130 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
9131 will then define a function called
9132 @code{merge_dllimport_decl_attributes} which can then be defined as
9133 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
9134 add @code{handle_dll_attribute} in the attribute table for your port
9135 to perform initial processing of the @samp{dllimport} and
9136 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
9137 @file{i386/i386.c}, for example.
9140 @deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (tree @var{decl})
9141 @var{decl} is a variable or function with @code{__attribute__((dllimport))}
9142 specified. Use this hook if the target needs to add extra validation
9143 checks to @code{handle_dll_attribute}.
9146 @defmac TARGET_DECLSPEC
9147 Define this macro to a nonzero value if you want to treat
9148 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
9149 default, this behavior is enabled only for targets that define
9150 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
9151 of @code{__declspec} is via a built-in macro, but you should not rely
9152 on this implementation detail.
9155 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
9156 Define this target hook if you want to be able to add attributes to a decl
9157 when it is being created. This is normally useful for back ends which
9158 wish to implement a pragma by using the attributes which correspond to
9159 the pragma's effect. The @var{node} argument is the decl which is being
9160 created. The @var{attr_ptr} argument is a pointer to the attribute list
9161 for this decl. The list itself should not be modified, since it may be
9162 shared with other decls, but attributes may be chained on the head of
9163 the list and @code{*@var{attr_ptr}} modified to point to the new
9164 attributes, or a copy of the list may be made if further changes are
9168 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
9170 This target hook returns @code{true} if it is ok to inline @var{fndecl}
9171 into the current function, despite its having target-specific
9172 attributes, @code{false} otherwise. By default, if a function has a
9173 target specific attribute attached to it, it will not be inlined.
9176 @node MIPS Coprocessors
9177 @section Defining coprocessor specifics for MIPS targets.
9178 @cindex MIPS coprocessor-definition macros
9180 The MIPS specification allows MIPS implementations to have as many as 4
9181 coprocessors, each with as many as 32 private registers. GCC supports
9182 accessing these registers and transferring values between the registers
9183 and memory using asm-ized variables. For example:
9186 register unsigned int cp0count asm ("c0r1");
9192 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
9193 names may be added as described below, or the default names may be
9194 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
9196 Coprocessor registers are assumed to be epilogue-used; sets to them will
9197 be preserved even if it does not appear that the register is used again
9198 later in the function.
9200 Another note: according to the MIPS spec, coprocessor 1 (if present) is
9201 the FPU@. One accesses COP1 registers through standard mips
9202 floating-point support; they are not included in this mechanism.
9204 There is one macro used in defining the MIPS coprocessor interface which
9205 you may want to override in subtargets; it is described below.
9207 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
9208 A comma-separated list (with leading comma) of pairs describing the
9209 alternate names of coprocessor registers. The format of each entry should be
9211 @{ @var{alternatename}, @var{register_number}@}
9217 @section Parameters for Precompiled Header Validity Checking
9218 @cindex parameters, precompiled headers
9220 @deftypefn {Target Hook} void *TARGET_GET_PCH_VALIDITY (size_t *@var{sz})
9221 This hook returns the data needed by @code{TARGET_PCH_VALID_P} and sets
9222 @samp{*@var{sz}} to the size of the data in bytes.
9225 @deftypefn {Target Hook} const char *TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz})
9226 This hook checks whether the options used to create a PCH file are
9227 compatible with the current settings. It returns @code{NULL}
9228 if so and a suitable error message if not. Error messages will
9229 be presented to the user and must be localized using @samp{_(@var{msg})}.
9231 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
9232 when the PCH file was created and @var{sz} is the size of that data in bytes.
9233 It's safe to assume that the data was created by the same version of the
9234 compiler, so no format checking is needed.
9236 The default definition of @code{default_pch_valid_p} should be
9237 suitable for most targets.
9240 @deftypefn {Target Hook} const char *TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags})
9241 If this hook is nonnull, the default implementation of
9242 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
9243 of @code{target_flags}. @var{pch_flags} specifies the value that
9244 @code{target_flags} had when the PCH file was created. The return
9245 value is the same as for @code{TARGET_PCH_VALID_P}.
9249 @section C++ ABI parameters
9250 @cindex parameters, c++ abi
9252 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
9253 Define this hook to override the integer type used for guard variables.
9254 These are used to implement one-time construction of static objects. The
9255 default is long_long_integer_type_node.
9258 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
9259 This hook determines how guard variables are used. It should return
9260 @code{false} (the default) if first byte should be used. A return value of
9261 @code{true} indicates the least significant bit should be used.
9264 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
9265 This hook returns the size of the cookie to use when allocating an array
9266 whose elements have the indicated @var{type}. Assumes that it is already
9267 known that a cookie is needed. The default is
9268 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
9269 IA64/Generic C++ ABI@.
9272 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
9273 This hook should return @code{true} if the element size should be stored in
9274 array cookies. The default is to return @code{false}.
9277 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
9278 If defined by a backend this hook allows the decision made to export
9279 class @var{type} to be overruled. Upon entry @var{import_export}
9280 will contain 1 if the class is going to be exported, @minus{}1 if it is going
9281 to be imported and 0 otherwise. This function should return the
9282 modified value and perform any other actions necessary to support the
9283 backend's targeted operating system.
9286 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
9287 This hook should return @code{true} if constructors and destructors return
9288 the address of the object created/destroyed. The default is to return
9292 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
9293 This hook returns true if the key method for a class (i.e., the method
9294 which, if defined in the current translation unit, causes the virtual
9295 table to be emitted) may be an inline function. Under the standard
9296 Itanium C++ ABI the key method may be an inline function so long as
9297 the function is not declared inline in the class definition. Under
9298 some variants of the ABI, an inline function can never be the key
9299 method. The default is to return @code{true}.
9302 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
9303 @var{decl} is a virtual table, virtual table table, typeinfo object,
9304 or other similar implicit class data object that will be emitted with
9305 external linkage in this translation unit. No ELF visibility has been
9306 explicitly specified. If the target needs to specify a visibility
9307 other than that of the containing class, use this hook to set
9308 @code{DECL_VISIBILITY} and @code{DECL_VISIBILITY_SPECIFIED}.
9311 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
9312 This hook returns true (the default) if virtual tables and other
9313 similar implicit class data objects are always COMDAT if they have
9314 external linkage. If this hook returns false, then class data for
9315 classes whose virtual table will be emitted in only one translation
9316 unit will not be COMDAT.
9319 @deftypefn {Target Hook} bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void)
9320 This hook returns true (the default) if the RTTI information for
9321 the basic types which is defined in the C++ runtime should always
9322 be COMDAT, false if it should not be COMDAT.
9325 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
9326 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
9327 should be used to register static destructors when @option{-fuse-cxa-atexit}
9328 is in effect. The default is to return false to use @code{__cxa_atexit}.
9331 @deftypefn {Target Hook} bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void)
9332 This hook returns true if the target @code{atexit} function can be used
9333 in the same manner as @code{__cxa_atexit} to register C++ static
9334 destructors. This requires that @code{atexit}-registered functions in
9335 shared libraries are run in the correct order when the libraries are
9336 unloaded. The default is to return false.
9339 @deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type})
9340 @var{type} is a C++ class (i.e., RECORD_TYPE or UNION_TYPE) that has just been
9341 defined. Use this hook to make adjustments to the class (eg, tweak
9342 visibility or perform any other required target modifications).
9346 @section Miscellaneous Parameters
9347 @cindex parameters, miscellaneous
9349 @c prevent bad page break with this line
9350 Here are several miscellaneous parameters.
9352 @defmac HAS_LONG_COND_BRANCH
9353 Define this boolean macro to indicate whether or not your architecture
9354 has conditional branches that can span all of memory. It is used in
9355 conjunction with an optimization that partitions hot and cold basic
9356 blocks into separate sections of the executable. If this macro is
9357 set to false, gcc will convert any conditional branches that attempt
9358 to cross between sections into unconditional branches or indirect jumps.
9361 @defmac HAS_LONG_UNCOND_BRANCH
9362 Define this boolean macro to indicate whether or not your architecture
9363 has unconditional branches that can span all of memory. It is used in
9364 conjunction with an optimization that partitions hot and cold basic
9365 blocks into separate sections of the executable. If this macro is
9366 set to false, gcc will convert any unconditional branches that attempt
9367 to cross between sections into indirect jumps.
9370 @defmac CASE_VECTOR_MODE
9371 An alias for a machine mode name. This is the machine mode that
9372 elements of a jump-table should have.
9375 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
9376 Optional: return the preferred mode for an @code{addr_diff_vec}
9377 when the minimum and maximum offset are known. If you define this,
9378 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
9379 To make this work, you also have to define @code{INSN_ALIGN} and
9380 make the alignment for @code{addr_diff_vec} explicit.
9381 The @var{body} argument is provided so that the offset_unsigned and scale
9382 flags can be updated.
9385 @defmac CASE_VECTOR_PC_RELATIVE
9386 Define this macro to be a C expression to indicate when jump-tables
9387 should contain relative addresses. You need not define this macro if
9388 jump-tables never contain relative addresses, or jump-tables should
9389 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
9393 @defmac CASE_VALUES_THRESHOLD
9394 Define this to be the smallest number of different values for which it
9395 is best to use a jump-table instead of a tree of conditional branches.
9396 The default is four for machines with a @code{casesi} instruction and
9397 five otherwise. This is best for most machines.
9400 @defmac CASE_USE_BIT_TESTS
9401 Define this macro to be a C expression to indicate whether C switch
9402 statements may be implemented by a sequence of bit tests. This is
9403 advantageous on processors that can efficiently implement left shift
9404 of 1 by the number of bits held in a register, but inappropriate on
9405 targets that would require a loop. By default, this macro returns
9406 @code{true} if the target defines an @code{ashlsi3} pattern, and
9407 @code{false} otherwise.
9410 @defmac WORD_REGISTER_OPERATIONS
9411 Define this macro if operations between registers with integral mode
9412 smaller than a word are always performed on the entire register.
9413 Most RISC machines have this property and most CISC machines do not.
9416 @defmac LOAD_EXTEND_OP (@var{mem_mode})
9417 Define this macro to be a C expression indicating when insns that read
9418 memory in @var{mem_mode}, an integral mode narrower than a word, set the
9419 bits outside of @var{mem_mode} to be either the sign-extension or the
9420 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
9421 of @var{mem_mode} for which the
9422 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
9423 @code{UNKNOWN} for other modes.
9425 This macro is not called with @var{mem_mode} non-integral or with a width
9426 greater than or equal to @code{BITS_PER_WORD}, so you may return any
9427 value in this case. Do not define this macro if it would always return
9428 @code{UNKNOWN}. On machines where this macro is defined, you will normally
9429 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
9431 You may return a non-@code{UNKNOWN} value even if for some hard registers
9432 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
9433 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
9434 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
9435 integral mode larger than this but not larger than @code{word_mode}.
9437 You must return @code{UNKNOWN} if for some hard registers that allow this
9438 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
9439 @code{word_mode}, but that they can change to another integral mode that
9440 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
9443 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
9444 Define this macro if loading short immediate values into registers sign
9448 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
9449 Define this macro if the same instructions that convert a floating
9450 point number to a signed fixed point number also convert validly to an
9454 @deftypefn {Target Hook} int TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (enum machine_mode @var{mode})
9455 When @option{-ffast-math} is in effect, GCC tries to optimize
9456 divisions by the same divisor, by turning them into multiplications by
9457 the reciprocal. This target hook specifies the minimum number of divisions
9458 that should be there for GCC to perform the optimization for a variable
9459 of mode @var{mode}. The default implementation returns 3 if the machine
9460 has an instruction for the division, and 2 if it does not.
9464 The maximum number of bytes that a single instruction can move quickly
9465 between memory and registers or between two memory locations.
9468 @defmac MAX_MOVE_MAX
9469 The maximum number of bytes that a single instruction can move quickly
9470 between memory and registers or between two memory locations. If this
9471 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
9472 constant value that is the largest value that @code{MOVE_MAX} can have
9476 @defmac SHIFT_COUNT_TRUNCATED
9477 A C expression that is nonzero if on this machine the number of bits
9478 actually used for the count of a shift operation is equal to the number
9479 of bits needed to represent the size of the object being shifted. When
9480 this macro is nonzero, the compiler will assume that it is safe to omit
9481 a sign-extend, zero-extend, and certain bitwise `and' instructions that
9482 truncates the count of a shift operation. On machines that have
9483 instructions that act on bit-fields at variable positions, which may
9484 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
9485 also enables deletion of truncations of the values that serve as
9486 arguments to bit-field instructions.
9488 If both types of instructions truncate the count (for shifts) and
9489 position (for bit-field operations), or if no variable-position bit-field
9490 instructions exist, you should define this macro.
9492 However, on some machines, such as the 80386 and the 680x0, truncation
9493 only applies to shift operations and not the (real or pretended)
9494 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
9495 such machines. Instead, add patterns to the @file{md} file that include
9496 the implied truncation of the shift instructions.
9498 You need not define this macro if it would always have the value of zero.
9501 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
9502 @deftypefn {Target Hook} int TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode})
9503 This function describes how the standard shift patterns for @var{mode}
9504 deal with shifts by negative amounts or by more than the width of the mode.
9505 @xref{shift patterns}.
9507 On many machines, the shift patterns will apply a mask @var{m} to the
9508 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
9509 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
9510 this is true for mode @var{mode}, the function should return @var{m},
9511 otherwise it should return 0. A return value of 0 indicates that no
9512 particular behavior is guaranteed.
9514 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
9515 @emph{not} apply to general shift rtxes; it applies only to instructions
9516 that are generated by the named shift patterns.
9518 The default implementation of this function returns
9519 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
9520 and 0 otherwise. This definition is always safe, but if
9521 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
9522 nevertheless truncate the shift count, you may get better code
9526 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
9527 A C expression which is nonzero if on this machine it is safe to
9528 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
9529 bits (where @var{outprec} is smaller than @var{inprec}) by merely
9530 operating on it as if it had only @var{outprec} bits.
9532 On many machines, this expression can be 1.
9534 @c rearranged this, removed the phrase "it is reported that". this was
9535 @c to fix an overfull hbox. --mew 10feb93
9536 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
9537 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
9538 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
9539 such cases may improve things.
9542 @deftypefn {Target Hook} int TARGET_MODE_REP_EXTENDED (enum machine_mode @var{mode}, enum machine_mode @var{rep_mode})
9543 The representation of an integral mode can be such that the values
9544 are always extended to a wider integral mode. Return
9545 @code{SIGN_EXTEND} if values of @var{mode} are represented in
9546 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
9547 otherwise. (Currently, none of the targets use zero-extended
9548 representation this way so unlike @code{LOAD_EXTEND_OP},
9549 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
9550 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
9551 @var{mode} to @var{mode_rep} so that @var{mode_rep} is not the next
9552 widest integral mode and currently we take advantage of this fact.)
9554 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
9555 value even if the extension is not performed on certain hard registers
9556 as long as for the @code{REGNO_REG_CLASS} of these hard registers
9557 @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero.
9559 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
9560 describe two related properties. If you define
9561 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
9562 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
9565 In order to enforce the representation of @code{mode},
9566 @code{TRULY_NOOP_TRUNCATION} should return false when truncating to
9570 @defmac STORE_FLAG_VALUE
9571 A C expression describing the value returned by a comparison operator
9572 with an integral mode and stored by a store-flag instruction
9573 (@samp{s@var{cond}}) when the condition is true. This description must
9574 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
9575 comparison operators whose results have a @code{MODE_INT} mode.
9577 A value of 1 or @minus{}1 means that the instruction implementing the
9578 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
9579 and 0 when the comparison is false. Otherwise, the value indicates
9580 which bits of the result are guaranteed to be 1 when the comparison is
9581 true. This value is interpreted in the mode of the comparison
9582 operation, which is given by the mode of the first operand in the
9583 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
9584 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
9587 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
9588 generate code that depends only on the specified bits. It can also
9589 replace comparison operators with equivalent operations if they cause
9590 the required bits to be set, even if the remaining bits are undefined.
9591 For example, on a machine whose comparison operators return an
9592 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
9593 @samp{0x80000000}, saying that just the sign bit is relevant, the
9597 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
9604 (ashift:SI @var{x} (const_int @var{n}))
9608 where @var{n} is the appropriate shift count to move the bit being
9609 tested into the sign bit.
9611 There is no way to describe a machine that always sets the low-order bit
9612 for a true value, but does not guarantee the value of any other bits,
9613 but we do not know of any machine that has such an instruction. If you
9614 are trying to port GCC to such a machine, include an instruction to
9615 perform a logical-and of the result with 1 in the pattern for the
9616 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
9618 Often, a machine will have multiple instructions that obtain a value
9619 from a comparison (or the condition codes). Here are rules to guide the
9620 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
9625 Use the shortest sequence that yields a valid definition for
9626 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
9627 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
9628 comparison operators to do so because there may be opportunities to
9629 combine the normalization with other operations.
9632 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
9633 slightly preferred on machines with expensive jumps and 1 preferred on
9637 As a second choice, choose a value of @samp{0x80000001} if instructions
9638 exist that set both the sign and low-order bits but do not define the
9642 Otherwise, use a value of @samp{0x80000000}.
9645 Many machines can produce both the value chosen for
9646 @code{STORE_FLAG_VALUE} and its negation in the same number of
9647 instructions. On those machines, you should also define a pattern for
9648 those cases, e.g., one matching
9651 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
9654 Some machines can also perform @code{and} or @code{plus} operations on
9655 condition code values with less instructions than the corresponding
9656 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
9657 machines, define the appropriate patterns. Use the names @code{incscc}
9658 and @code{decscc}, respectively, for the patterns which perform
9659 @code{plus} or @code{minus} operations on condition code values. See
9660 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
9661 find such instruction sequences on other machines.
9663 If this macro is not defined, the default value, 1, is used. You need
9664 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
9665 instructions, or if the value generated by these instructions is 1.
9668 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
9669 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
9670 returned when comparison operators with floating-point results are true.
9671 Define this macro on machines that have comparison operations that return
9672 floating-point values. If there are no such operations, do not define
9676 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
9677 A C expression that gives a rtx representing the nonzero true element
9678 for vector comparisons. The returned rtx should be valid for the inner
9679 mode of @var{mode} which is guaranteed to be a vector mode. Define
9680 this macro on machines that have vector comparison operations that
9681 return a vector result. If there are no such operations, do not define
9682 this macro. Typically, this macro is defined as @code{const1_rtx} or
9683 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
9684 the compiler optimizing such vector comparison operations for the
9688 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
9689 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
9690 A C expression that indicates whether the architecture defines a value
9691 for @code{clz} or @code{ctz} with a zero operand.
9692 A result of @code{0} indicates the value is undefined.
9693 If the value is defined for only the RTL expression, the macro should
9694 evaluate to @code{1}; if the value applies also to the corresponding optab
9695 entry (which is normally the case if it expands directly into
9696 the corresponding RTL), then the macro should evaluate to @code{2}.
9697 In the cases where the value is defined, @var{value} should be set to
9700 If this macro is not defined, the value of @code{clz} or
9701 @code{ctz} at zero is assumed to be undefined.
9703 This macro must be defined if the target's expansion for @code{ffs}
9704 relies on a particular value to get correct results. Otherwise it
9705 is not necessary, though it may be used to optimize some corner cases, and
9706 to provide a default expansion for the @code{ffs} optab.
9708 Note that regardless of this macro the ``definedness'' of @code{clz}
9709 and @code{ctz} at zero do @emph{not} extend to the builtin functions
9710 visible to the user. Thus one may be free to adjust the value at will
9711 to match the target expansion of these operations without fear of
9716 An alias for the machine mode for pointers. On most machines, define
9717 this to be the integer mode corresponding to the width of a hardware
9718 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
9719 On some machines you must define this to be one of the partial integer
9720 modes, such as @code{PSImode}.
9722 The width of @code{Pmode} must be at least as large as the value of
9723 @code{POINTER_SIZE}. If it is not equal, you must define the macro
9724 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
9728 @defmac FUNCTION_MODE
9729 An alias for the machine mode used for memory references to functions
9730 being called, in @code{call} RTL expressions. On most CISC machines,
9731 where an instruction can begin at any byte address, this should be
9732 @code{QImode}. On most RISC machines, where all instructions have fixed
9733 size and alignment, this should be a mode with the same size and alignment
9734 as the machine instruction words - typically @code{SImode} or @code{HImode}.
9737 @defmac STDC_0_IN_SYSTEM_HEADERS
9738 In normal operation, the preprocessor expands @code{__STDC__} to the
9739 constant 1, to signify that GCC conforms to ISO Standard C@. On some
9740 hosts, like Solaris, the system compiler uses a different convention,
9741 where @code{__STDC__} is normally 0, but is 1 if the user specifies
9742 strict conformance to the C Standard.
9744 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
9745 convention when processing system header files, but when processing user
9746 files @code{__STDC__} will always expand to 1.
9749 @defmac NO_IMPLICIT_EXTERN_C
9750 Define this macro if the system header files support C++ as well as C@.
9751 This macro inhibits the usual method of using system header files in
9752 C++, which is to pretend that the file's contents are enclosed in
9753 @samp{extern "C" @{@dots{}@}}.
9758 @defmac REGISTER_TARGET_PRAGMAS ()
9759 Define this macro if you want to implement any target-specific pragmas.
9760 If defined, it is a C expression which makes a series of calls to
9761 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
9762 for each pragma. The macro may also do any
9763 setup required for the pragmas.
9765 The primary reason to define this macro is to provide compatibility with
9766 other compilers for the same target. In general, we discourage
9767 definition of target-specific pragmas for GCC@.
9769 If the pragma can be implemented by attributes then you should consider
9770 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
9772 Preprocessor macros that appear on pragma lines are not expanded. All
9773 @samp{#pragma} directives that do not match any registered pragma are
9774 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
9777 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9778 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9780 Each call to @code{c_register_pragma} or
9781 @code{c_register_pragma_with_expansion} establishes one pragma. The
9782 @var{callback} routine will be called when the preprocessor encounters a
9786 #pragma [@var{space}] @var{name} @dots{}
9789 @var{space} is the case-sensitive namespace of the pragma, or
9790 @code{NULL} to put the pragma in the global namespace. The callback
9791 routine receives @var{pfile} as its first argument, which can be passed
9792 on to cpplib's functions if necessary. You can lex tokens after the
9793 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
9794 callback will be silently ignored. The end of the line is indicated by
9795 a token of type @code{CPP_EOF}. Macro expansion occurs on the
9796 arguments of pragmas registered with
9797 @code{c_register_pragma_with_expansion} but not on the arguments of
9798 pragmas registered with @code{c_register_pragma}.
9800 For an example use of this routine, see @file{c4x.h} and the callback
9801 routines defined in @file{c4x-c.c}.
9803 Note that the use of @code{pragma_lex} is specific to the C and C++
9804 compilers. It will not work in the Java or Fortran compilers, or any
9805 other language compilers for that matter. Thus if @code{pragma_lex} is going
9806 to be called from target-specific code, it must only be done so when
9807 building the C and C++ compilers. This can be done by defining the
9808 variables @code{c_target_objs} and @code{cxx_target_objs} in the
9809 target entry in the @file{config.gcc} file. These variables should name
9810 the target-specific, language-specific object file which contains the
9811 code that uses @code{pragma_lex}. Note it will also be necessary to add a
9812 rule to the makefile fragment pointed to by @code{tmake_file} that shows
9813 how to build this object file.
9818 @defmac HANDLE_SYSV_PRAGMA
9819 Define this macro (to a value of 1) if you want the System V style
9820 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
9821 [=<value>]} to be supported by gcc.
9823 The pack pragma specifies the maximum alignment (in bytes) of fields
9824 within a structure, in much the same way as the @samp{__aligned__} and
9825 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
9826 the behavior to the default.
9828 A subtlety for Microsoft Visual C/C++ style bit-field packing
9829 (e.g.@: -mms-bitfields) for targets that support it:
9830 When a bit-field is inserted into a packed record, the whole size
9831 of the underlying type is used by one or more same-size adjacent
9832 bit-fields (that is, if its long:3, 32 bits is used in the record,
9833 and any additional adjacent long bit-fields are packed into the same
9834 chunk of 32 bits. However, if the size changes, a new field of that
9837 If both MS bit-fields and @samp{__attribute__((packed))} are used,
9838 the latter will take precedence. If @samp{__attribute__((packed))} is
9839 used on a single field when MS bit-fields are in use, it will take
9840 precedence for that field, but the alignment of the rest of the structure
9841 may affect its placement.
9843 The weak pragma only works if @code{SUPPORTS_WEAK} and
9844 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
9845 of specifically named weak labels, optionally with a value.
9850 @defmac HANDLE_PRAGMA_PACK_PUSH_POP
9851 Define this macro (to a value of 1) if you want to support the Win32
9852 style pragmas @samp{#pragma pack(push[,@var{n}])} and @samp{#pragma
9853 pack(pop)}. The @samp{pack(push,[@var{n}])} pragma specifies the maximum
9854 alignment (in bytes) of fields within a structure, in much the same way as
9855 the @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
9856 pack value of zero resets the behavior to the default. Successive
9857 invocations of this pragma cause the previous values to be stacked, so
9858 that invocations of @samp{#pragma pack(pop)} will return to the previous
9862 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
9863 Define this macro, as well as
9864 @code{HANDLE_SYSV_PRAGMA}, if macros should be expanded in the
9865 arguments of @samp{#pragma pack}.
9868 @defmac TARGET_DEFAULT_PACK_STRUCT
9869 If your target requires a structure packing default other than 0 (meaning
9870 the machine default), define this macro to the necessary value (in bytes).
9871 This must be a value that would also be valid to use with
9872 @samp{#pragma pack()} (that is, a small power of two).
9877 @defmac HANDLE_PRAGMA_PUSH_POP_MACRO
9878 Define this macro if you want to support the Win32 style pragmas
9879 @samp{#pragma push_macro(macro-name-as-string)} and @samp{#pragma
9880 pop_macro(macro-name-as-string)}. The @samp{#pragma push_macro(
9881 macro-name-as-string)} pragma saves the named macro and via
9882 @samp{#pragma pop_macro(macro-name-as-string)} it will return to the
9887 @defmac DOLLARS_IN_IDENTIFIERS
9888 Define this macro to control use of the character @samp{$} in
9889 identifier names for the C family of languages. 0 means @samp{$} is
9890 not allowed by default; 1 means it is allowed. 1 is the default;
9891 there is no need to define this macro in that case.
9894 @defmac NO_DOLLAR_IN_LABEL
9895 Define this macro if the assembler does not accept the character
9896 @samp{$} in label names. By default constructors and destructors in
9897 G++ have @samp{$} in the identifiers. If this macro is defined,
9898 @samp{.} is used instead.
9901 @defmac NO_DOT_IN_LABEL
9902 Define this macro if the assembler does not accept the character
9903 @samp{.} in label names. By default constructors and destructors in G++
9904 have names that use @samp{.}. If this macro is defined, these names
9905 are rewritten to avoid @samp{.}.
9908 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
9909 Define this macro as a C expression that is nonzero if it is safe for the
9910 delay slot scheduler to place instructions in the delay slot of @var{insn},
9911 even if they appear to use a resource set or clobbered in @var{insn}.
9912 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
9913 every @code{call_insn} has this behavior. On machines where some @code{insn}
9914 or @code{jump_insn} is really a function call and hence has this behavior,
9915 you should define this macro.
9917 You need not define this macro if it would always return zero.
9920 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
9921 Define this macro as a C expression that is nonzero if it is safe for the
9922 delay slot scheduler to place instructions in the delay slot of @var{insn},
9923 even if they appear to set or clobber a resource referenced in @var{insn}.
9924 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
9925 some @code{insn} or @code{jump_insn} is really a function call and its operands
9926 are registers whose use is actually in the subroutine it calls, you should
9927 define this macro. Doing so allows the delay slot scheduler to move
9928 instructions which copy arguments into the argument registers into the delay
9931 You need not define this macro if it would always return zero.
9934 @defmac MULTIPLE_SYMBOL_SPACES
9935 Define this macro as a C expression that is nonzero if, in some cases,
9936 global symbols from one translation unit may not be bound to undefined
9937 symbols in another translation unit without user intervention. For
9938 instance, under Microsoft Windows symbols must be explicitly imported
9939 from shared libraries (DLLs).
9941 You need not define this macro if it would always evaluate to zero.
9944 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{outputs}, tree @var{inputs}, tree @var{clobbers})
9945 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
9946 any hard regs the port wishes to automatically clobber for an asm.
9947 It should return the result of the last @code{tree_cons} used to add a
9948 clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the
9949 corresponding parameters to the asm and may be inspected to avoid
9950 clobbering a register that is an input or output of the asm. You can use
9951 @code{tree_overlaps_hard_reg_set}, declared in @file{tree.h}, to test
9952 for overlap with regards to asm-declared registers.
9955 @defmac MATH_LIBRARY
9956 Define this macro as a C string constant for the linker argument to link
9957 in the system math library, or @samp{""} if the target does not have a
9958 separate math library.
9960 You need only define this macro if the default of @samp{"-lm"} is wrong.
9963 @defmac LIBRARY_PATH_ENV
9964 Define this macro as a C string constant for the environment variable that
9965 specifies where the linker should look for libraries.
9967 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
9971 @defmac TARGET_POSIX_IO
9972 Define this macro if the target supports the following POSIX@ file
9973 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
9974 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
9975 to use file locking when exiting a program, which avoids race conditions
9976 if the program has forked. It will also create directories at run-time
9977 for cross-profiling.
9980 @defmac MAX_CONDITIONAL_EXECUTE
9982 A C expression for the maximum number of instructions to execute via
9983 conditional execution instructions instead of a branch. A value of
9984 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
9985 1 if it does use cc0.
9988 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
9989 Used if the target needs to perform machine-dependent modifications on the
9990 conditionals used for turning basic blocks into conditionally executed code.
9991 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
9992 contains information about the currently processed blocks. @var{true_expr}
9993 and @var{false_expr} are the tests that are used for converting the
9994 then-block and the else-block, respectively. Set either @var{true_expr} or
9995 @var{false_expr} to a null pointer if the tests cannot be converted.
9998 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
9999 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
10000 if-statements into conditions combined by @code{and} and @code{or} operations.
10001 @var{bb} contains the basic block that contains the test that is currently
10002 being processed and about to be turned into a condition.
10005 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
10006 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
10007 be converted to conditional execution format. @var{ce_info} points to
10008 a data structure, @code{struct ce_if_block}, which contains information
10009 about the currently processed blocks.
10012 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
10013 A C expression to perform any final machine dependent modifications in
10014 converting code to conditional execution. The involved basic blocks
10015 can be found in the @code{struct ce_if_block} structure that is pointed
10016 to by @var{ce_info}.
10019 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
10020 A C expression to cancel any machine dependent modifications in
10021 converting code to conditional execution. The involved basic blocks
10022 can be found in the @code{struct ce_if_block} structure that is pointed
10023 to by @var{ce_info}.
10026 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
10027 A C expression to initialize any extra fields in a @code{struct ce_if_block}
10028 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
10031 @defmac IFCVT_EXTRA_FIELDS
10032 If defined, it should expand to a set of field declarations that will be
10033 added to the @code{struct ce_if_block} structure. These should be initialized
10034 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
10037 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG ()
10038 If non-null, this hook performs a target-specific pass over the
10039 instruction stream. The compiler will run it at all optimization levels,
10040 just before the point at which it normally does delayed-branch scheduling.
10042 The exact purpose of the hook varies from target to target. Some use
10043 it to do transformations that are necessary for correctness, such as
10044 laying out in-function constant pools or avoiding hardware hazards.
10045 Others use it as an opportunity to do some machine-dependent optimizations.
10047 You need not implement the hook if it has nothing to do. The default
10048 definition is null.
10051 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
10052 Define this hook if you have any machine-specific built-in functions
10053 that need to be defined. It should be a function that performs the
10056 Machine specific built-in functions can be useful to expand special machine
10057 instructions that would otherwise not normally be generated because
10058 they have no equivalent in the source language (for example, SIMD vector
10059 instructions or prefetch instructions).
10061 To create a built-in function, call the function
10062 @code{lang_hooks.builtin_function}
10063 which is defined by the language front end. You can use any type nodes set
10064 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
10065 only language front ends that use those two functions will call
10066 @samp{TARGET_INIT_BUILTINS}.
10069 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
10071 Expand a call to a machine specific built-in function that was set up by
10072 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
10073 function call; the result should go to @var{target} if that is
10074 convenient, and have mode @var{mode} if that is convenient.
10075 @var{subtarget} may be used as the target for computing one of
10076 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
10077 ignored. This function should return the result of the call to the
10081 @deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (tree @var{fndecl}, tree @var{arglist})
10083 Select a replacement for a machine specific built-in function that
10084 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
10085 @emph{before} regular type checking, and so allows the target to
10086 implement a crude form of function overloading. @var{fndecl} is the
10087 declaration of the built-in function. @var{arglist} is the list of
10088 arguments passed to the built-in function. The result is a
10089 complete expression that implements the operation, usually
10090 another @code{CALL_EXPR}.
10093 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, tree @var{arglist}, bool @var{ignore})
10095 Fold a call to a machine specific built-in function that was set up by
10096 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
10097 built-in function. @var{arglist} is the list of arguments passed to
10098 the built-in function. The result is another tree containing a
10099 simplified expression for the call's result. If @var{ignore} is true
10100 the value will be ignored.
10103 @deftypefn {Target Hook} const char * TARGET_INVALID_WITHIN_DOLOOP (rtx @var{insn})
10105 Take an instruction in @var{insn} and return NULL if it is valid within a
10106 low-overhead loop, otherwise return a string why doloop could not be applied.
10108 Many targets use special registers for low-overhead looping. For any
10109 instruction that clobbers these this function should return a string indicating
10110 the reason why the doloop could not be applied.
10111 By default, the RTL loop optimizer does not use a present doloop pattern for
10112 loops containing function calls or branch on table instructions.
10115 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
10117 Take a branch insn in @var{branch1} and another in @var{branch2}.
10118 Return true if redirecting @var{branch1} to the destination of
10119 @var{branch2} is possible.
10121 On some targets, branches may have a limited range. Optimizing the
10122 filling of delay slots can result in branches being redirected, and this
10123 may in turn cause a branch offset to overflow.
10126 @deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (rtx @var{x}, @var{outer_code})
10127 This target hook returns @code{true} if @var{x} is considered to be commutative.
10128 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
10129 PLUS to be commutative inside a MEM. @var{outer_code} is the rtx code
10130 of the enclosing rtl, if known, otherwise it is UNKNOWN.
10133 @deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg})
10135 When the initial value of a hard register has been copied in a pseudo
10136 register, it is often not necessary to actually allocate another register
10137 to this pseudo register, because the original hard register or a stack slot
10138 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
10139 is called at the start of register allocation once for each hard register
10140 that had its initial value copied by using
10141 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
10142 Possible values are @code{NULL_RTX}, if you don't want
10143 to do any special allocation, a @code{REG} rtx---that would typically be
10144 the hard register itself, if it is known not to be clobbered---or a
10146 If you are returning a @code{MEM}, this is only a hint for the allocator;
10147 it might decide to use another register anyways.
10148 You may use @code{current_function_leaf_function} in the hook, functions
10149 that use @code{REG_N_SETS}, to determine if the hard
10150 register in question will not be clobbered.
10151 The default value of this hook is @code{NULL}, which disables any special
10155 @deftypefn {Target Hook} int TARGET_UNSPEC_MAY_TRAP_P (const_rtx @var{x}, unsigned @var{flags})
10156 This target hook returns nonzero if @var{x}, an @code{unspec} or
10157 @code{unspec_volatile} operation, might cause a trap. Targets can use
10158 this hook to enhance precision of analysis for @code{unspec} and
10159 @code{unspec_volatile} operations. You may call @code{may_trap_p_1}
10160 to analyze inner elements of @var{x} in which case @var{flags} should be
10164 @deftypefn {Target Hook} void TARGET_SET_CURRENT_FUNCTION (tree @var{decl})
10165 The compiler invokes this hook whenever it changes its current function
10166 context (@code{cfun}). You can define this function if
10167 the back end needs to perform any initialization or reset actions on a
10168 per-function basis. For example, it may be used to implement function
10169 attributes that affect register usage or code generation patterns.
10170 The argument @var{decl} is the declaration for the new function context,
10171 and may be null to indicate that the compiler has left a function context
10172 and is returning to processing at the top level.
10173 The default hook function does nothing.
10175 GCC sets @code{cfun} to a dummy function context during initialization of
10176 some parts of the back end. The hook function is not invoked in this
10177 situation; you need not worry about the hook being invoked recursively,
10178 or when the back end is in a partially-initialized state.
10181 @defmac TARGET_OBJECT_SUFFIX
10182 Define this macro to be a C string representing the suffix for object
10183 files on your target machine. If you do not define this macro, GCC will
10184 use @samp{.o} as the suffix for object files.
10187 @defmac TARGET_EXECUTABLE_SUFFIX
10188 Define this macro to be a C string representing the suffix to be
10189 automatically added to executable files on your target machine. If you
10190 do not define this macro, GCC will use the null string as the suffix for
10194 @defmac COLLECT_EXPORT_LIST
10195 If defined, @code{collect2} will scan the individual object files
10196 specified on its command line and create an export list for the linker.
10197 Define this macro for systems like AIX, where the linker discards
10198 object files that are not referenced from @code{main} and uses export
10202 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
10203 Define this macro to a C expression representing a variant of the
10204 method call @var{mdecl}, if Java Native Interface (JNI) methods
10205 must be invoked differently from other methods on your target.
10206 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
10207 the @code{stdcall} calling convention and this macro is then
10208 defined as this expression:
10211 build_type_attribute_variant (@var{mdecl},
10213 (get_identifier ("stdcall"),
10218 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
10219 This target hook returns @code{true} past the point in which new jump
10220 instructions could be created. On machines that require a register for
10221 every jump such as the SHmedia ISA of SH5, this point would typically be
10222 reload, so this target hook should be defined to a function such as:
10226 cannot_modify_jumps_past_reload_p ()
10228 return (reload_completed || reload_in_progress);
10233 @deftypefn {Target Hook} int TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
10234 This target hook returns a register class for which branch target register
10235 optimizations should be applied. All registers in this class should be
10236 usable interchangeably. After reload, registers in this class will be
10237 re-allocated and loads will be hoisted out of loops and be subjected
10238 to inter-block scheduling.
10241 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
10242 Branch target register optimization will by default exclude callee-saved
10244 that are not already live during the current function; if this target hook
10245 returns true, they will be included. The target code must than make sure
10246 that all target registers in the class returned by
10247 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
10248 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
10249 epilogues have already been generated. Note, even if you only return
10250 true when @var{after_prologue_epilogue_gen} is false, you still are likely
10251 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
10252 to reserve space for caller-saved target registers.
10255 @defmac POWI_MAX_MULTS
10256 If defined, this macro is interpreted as a signed integer C expression
10257 that specifies the maximum number of floating point multiplications
10258 that should be emitted when expanding exponentiation by an integer
10259 constant inline. When this value is defined, exponentiation requiring
10260 more than this number of multiplications is implemented by calling the
10261 system library's @code{pow}, @code{powf} or @code{powl} routines.
10262 The default value places no upper bound on the multiplication count.
10265 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
10266 This target hook should register any extra include files for the
10267 target. The parameter @var{stdinc} indicates if normal include files
10268 are present. The parameter @var{sysroot} is the system root directory.
10269 The parameter @var{iprefix} is the prefix for the gcc directory.
10272 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
10273 This target hook should register any extra include files for the
10274 target before any standard headers. The parameter @var{stdinc}
10275 indicates if normal include files are present. The parameter
10276 @var{sysroot} is the system root directory. The parameter
10277 @var{iprefix} is the prefix for the gcc directory.
10280 @deftypefn Macro void TARGET_OPTF (char *@var{path})
10281 This target hook should register special include paths for the target.
10282 The parameter @var{path} is the include to register. On Darwin
10283 systems, this is used for Framework includes, which have semantics
10284 that are different from @option{-I}.
10287 @deftypefn {Target Hook} bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
10288 This target hook returns @code{true} if it is safe to use a local alias
10289 for a virtual function @var{fndecl} when constructing thunks,
10290 @code{false} otherwise. By default, the hook returns @code{true} for all
10291 functions, if a target supports aliases (i.e.@: defines
10292 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
10295 @defmac TARGET_FORMAT_TYPES
10296 If defined, this macro is the name of a global variable containing
10297 target-specific format checking information for the @option{-Wformat}
10298 option. The default is to have no target-specific format checks.
10301 @defmac TARGET_N_FORMAT_TYPES
10302 If defined, this macro is the number of entries in
10303 @code{TARGET_FORMAT_TYPES}.
10306 @deftypefn {Target Hook} bool TARGET_RELAXED_ORDERING
10307 If set to @code{true}, means that the target's memory model does not
10308 guarantee that loads which do not depend on one another will access
10309 main memory in the order of the instruction stream; if ordering is
10310 important, an explicit memory barrier must be used. This is true of
10311 many recent processors which implement a policy of ``relaxed,''
10312 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
10313 and ia64. The default is @code{false}.
10316 @deftypefn {Target Hook} const char *TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (tree @var{typelist}, tree @var{funcdecl}, tree @var{val})
10317 If defined, this macro returns the diagnostic message when it is
10318 illegal to pass argument @var{val} to function @var{funcdecl}
10319 with prototype @var{typelist}.
10322 @deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (tree @var{fromtype}, tree @var{totype})
10323 If defined, this macro returns the diagnostic message when it is
10324 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
10325 if validity should be determined by the front end.
10328 @deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, tree @var{type})
10329 If defined, this macro returns the diagnostic message when it is
10330 invalid to apply operation @var{op} (where unary plus is denoted by
10331 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
10332 if validity should be determined by the front end.
10335 @deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, tree @var{type1}, tree @var{type2})
10336 If defined, this macro returns the diagnostic message when it is
10337 invalid to apply operation @var{op} to operands of types @var{type1}
10338 and @var{type2}, or @code{NULL} if validity should be determined by
10342 @defmac TARGET_USE_JCR_SECTION
10343 This macro determines whether to use the JCR section to register Java
10344 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
10345 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.
10349 This macro determines the size of the objective C jump buffer for the
10350 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
10353 @defmac LIBGCC2_UNWIND_ATTRIBUTE
10354 Define this macro if any target-specific attributes need to be attached
10355 to the functions in @file{libgcc} that provide low-level support for
10356 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
10357 and the associated definitions of those functions.