1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,
2 @c 2002, 2003, 2004, 2005 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
7 @chapter Target Description Macros and Functions
8 @cindex machine description macros
9 @cindex target description macros
10 @cindex macros, target description
11 @cindex @file{tm.h} macros
13 In addition to the file @file{@var{machine}.md}, a machine description
14 includes a C header file conventionally given the name
15 @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}.
16 The header file defines numerous macros that convey the information
17 about the target machine that does not fit into the scheme of the
18 @file{.md} file. The file @file{tm.h} should be a link to
19 @file{@var{machine}.h}. The header file @file{config.h} includes
20 @file{tm.h} and most compiler source files include @file{config.h}. The
21 source file defines a variable @code{targetm}, which is a structure
22 containing pointers to functions and data relating to the target
23 machine. @file{@var{machine}.c} should also contain their definitions,
24 if they are not defined elsewhere in GCC, and other functions called
25 through the macros defined in the @file{.h} file.
28 * Target Structure:: The @code{targetm} variable.
29 * Driver:: Controlling how the driver runs the compilation passes.
30 * Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}.
31 * Per-Function Data:: Defining data structures for per-function information.
32 * Storage Layout:: Defining sizes and alignments of data.
33 * Type Layout:: Defining sizes and properties of basic user data types.
34 * Registers:: Naming and describing the hardware registers.
35 * Register Classes:: Defining the classes of hardware registers.
36 * 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}, if any.
621 The directories specified by the environment variable @code{COMPILER_PATH}.
624 The macro @code{STANDARD_EXEC_PREFIX}.
627 @file{/usr/lib/gcc/}.
630 The macro @code{MD_EXEC_PREFIX}, if any.
633 Here is the order of prefixes tried for startfiles:
637 Any prefixes specified by the user with @option{-B}.
640 The environment variable @code{GCC_EXEC_PREFIX}, if any.
643 The directories specified by the environment variable @code{LIBRARY_PATH}
644 (or port-specific name; native only, cross compilers do not use this).
647 The macro @code{STANDARD_EXEC_PREFIX}.
650 @file{/usr/lib/gcc/}.
653 The macro @code{MD_EXEC_PREFIX}, if any.
656 The macro @code{MD_STARTFILE_PREFIX}, if any.
659 The macro @code{STANDARD_STARTFILE_PREFIX}.
668 @node Run-time Target
669 @section Run-time Target Specification
670 @cindex run-time target specification
671 @cindex predefined macros
672 @cindex target specifications
674 @c prevent bad page break with this line
675 Here are run-time target specifications.
677 @defmac TARGET_CPU_CPP_BUILTINS ()
678 This function-like macro expands to a block of code that defines
679 built-in preprocessor macros and assertions for the target cpu, using
680 the functions @code{builtin_define}, @code{builtin_define_std} and
681 @code{builtin_assert}. When the front end
682 calls this macro it provides a trailing semicolon, and since it has
683 finished command line option processing your code can use those
686 @code{builtin_assert} takes a string in the form you pass to the
687 command-line option @option{-A}, such as @code{cpu=mips}, and creates
688 the assertion. @code{builtin_define} takes a string in the form
689 accepted by option @option{-D} and unconditionally defines the macro.
691 @code{builtin_define_std} takes a string representing the name of an
692 object-like macro. If it doesn't lie in the user's namespace,
693 @code{builtin_define_std} defines it unconditionally. Otherwise, it
694 defines a version with two leading underscores, and another version
695 with two leading and trailing underscores, and defines the original
696 only if an ISO standard was not requested on the command line. For
697 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
698 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
699 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
700 defines only @code{_ABI64}.
702 You can also test for the C dialect being compiled. The variable
703 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
704 or @code{clk_objective_c}. Note that if we are preprocessing
705 assembler, this variable will be @code{clk_c} but the function-like
706 macro @code{preprocessing_asm_p()} will return true, so you might want
707 to check for that first. If you need to check for strict ANSI, the
708 variable @code{flag_iso} can be used. The function-like macro
709 @code{preprocessing_trad_p()} can be used to check for traditional
713 @defmac TARGET_OS_CPP_BUILTINS ()
714 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
715 and is used for the target operating system instead.
718 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
719 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
720 and is used for the target object format. @file{elfos.h} uses this
721 macro to define @code{__ELF__}, so you probably do not need to define
725 @deftypevar {extern int} target_flags
726 This variable is declared in @file{options.h}, which is included before
727 any target-specific headers.
730 @deftypevar {Target Hook} int TARGET_DEFAULT_TARGET_FLAGS
731 This variable specifies the initial value of @code{target_flags}.
732 Its default setting is 0.
735 @cindex optional hardware or system features
736 @cindex features, optional, in system conventions
738 @deftypefn {Target Hook} bool TARGET_HANDLE_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
739 This hook is called whenever the user specifies one of the
740 target-specific options described by the @file{.opt} definition files
741 (@pxref{Options}). It has the opportunity to do some option-specific
742 processing and should return true if the option is valid. The default
743 definition does nothing but return true.
745 @var{code} specifies the @code{OPT_@var{name}} enumeration value
746 associated with the selected option; @var{name} is just a rendering of
747 the option name in which non-alphanumeric characters are replaced by
748 underscores. @var{arg} specifies the string argument and is null if
749 no argument was given. If the option is flagged as a @code{UInteger}
750 (@pxref{Option properties}), @var{value} is the numeric value of the
751 argument. Otherwise @var{value} is 1 if the positive form of the
752 option was used and 0 if the ``no-'' form was.
755 @defmac TARGET_VERSION
756 This macro is a C statement to print on @code{stderr} a string
757 describing the particular machine description choice. Every machine
758 description should define @code{TARGET_VERSION}. For example:
762 #define TARGET_VERSION \
763 fprintf (stderr, " (68k, Motorola syntax)");
765 #define TARGET_VERSION \
766 fprintf (stderr, " (68k, MIT syntax)");
771 @defmac OVERRIDE_OPTIONS
772 Sometimes certain combinations of command options do not make sense on
773 a particular target machine. You can define a macro
774 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
775 defined, is executed once just after all the command options have been
778 Don't use this macro to turn on various extra optimizations for
779 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
782 @defmac C_COMMON_OVERRIDE_OPTIONS
783 This is similar to @code{OVERRIDE_OPTIONS} but is only used in the C
784 language frontends (C, Objective-C, C++, Objective-C++) and so can be
785 used to alter option flag variables which only exist in those
789 @defmac OPTIMIZATION_OPTIONS (@var{level}, @var{size})
790 Some machines may desire to change what optimizations are performed for
791 various optimization levels. This macro, if defined, is executed once
792 just after the optimization level is determined and before the remainder
793 of the command options have been parsed. Values set in this macro are
794 used as the default values for the other command line options.
796 @var{level} is the optimization level specified; 2 if @option{-O2} is
797 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
799 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
801 You should not use this macro to change options that are not
802 machine-specific. These should uniformly selected by the same
803 optimization level on all supported machines. Use this macro to enable
804 machine-specific optimizations.
806 @strong{Do not examine @code{write_symbols} in
807 this macro!} The debugging options are not supposed to alter the
811 @defmac CAN_DEBUG_WITHOUT_FP
812 Define this macro if debugging can be performed even without a frame
813 pointer. If this macro is defined, GCC will turn on the
814 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
817 @node Per-Function Data
818 @section Defining data structures for per-function information.
819 @cindex per-function data
820 @cindex data structures
822 If the target needs to store information on a per-function basis, GCC
823 provides a macro and a couple of variables to allow this. Note, just
824 using statics to store the information is a bad idea, since GCC supports
825 nested functions, so you can be halfway through encoding one function
826 when another one comes along.
828 GCC defines a data structure called @code{struct function} which
829 contains all of the data specific to an individual function. This
830 structure contains a field called @code{machine} whose type is
831 @code{struct machine_function *}, which can be used by targets to point
832 to their own specific data.
834 If a target needs per-function specific data it should define the type
835 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
836 This macro should be used to initialize the function pointer
837 @code{init_machine_status}. This pointer is explained below.
839 One typical use of per-function, target specific data is to create an
840 RTX to hold the register containing the function's return address. This
841 RTX can then be used to implement the @code{__builtin_return_address}
842 function, for level 0.
844 Note---earlier implementations of GCC used a single data area to hold
845 all of the per-function information. Thus when processing of a nested
846 function began the old per-function data had to be pushed onto a
847 stack, and when the processing was finished, it had to be popped off the
848 stack. GCC used to provide function pointers called
849 @code{save_machine_status} and @code{restore_machine_status} to handle
850 the saving and restoring of the target specific information. Since the
851 single data area approach is no longer used, these pointers are no
854 @defmac INIT_EXPANDERS
855 Macro called to initialize any target specific information. This macro
856 is called once per function, before generation of any RTL has begun.
857 The intention of this macro is to allow the initialization of the
858 function pointer @code{init_machine_status}.
861 @deftypevar {void (*)(struct function *)} init_machine_status
862 If this function pointer is non-@code{NULL} it will be called once per
863 function, before function compilation starts, in order to allow the
864 target to perform any target specific initialization of the
865 @code{struct function} structure. It is intended that this would be
866 used to initialize the @code{machine} of that structure.
868 @code{struct machine_function} structures are expected to be freed by GC@.
869 Generally, any memory that they reference must be allocated by using
870 @code{ggc_alloc}, including the structure itself.
874 @section Storage Layout
875 @cindex storage layout
877 Note that the definitions of the macros in this table which are sizes or
878 alignments measured in bits do not need to be constant. They can be C
879 expressions that refer to static variables, such as the @code{target_flags}.
880 @xref{Run-time Target}.
882 @defmac BITS_BIG_ENDIAN
883 Define this macro to have the value 1 if the most significant bit in a
884 byte has the lowest number; otherwise define it to have the value zero.
885 This means that bit-field instructions count from the most significant
886 bit. If the machine has no bit-field instructions, then this must still
887 be defined, but it doesn't matter which value it is defined to. This
888 macro need not be a constant.
890 This macro does not affect the way structure fields are packed into
891 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
894 @defmac BYTES_BIG_ENDIAN
895 Define this macro to have the value 1 if the most significant byte in a
896 word has the lowest number. This macro need not be a constant.
899 @defmac WORDS_BIG_ENDIAN
900 Define this macro to have the value 1 if, in a multiword object, the
901 most significant word has the lowest number. This applies to both
902 memory locations and registers; GCC fundamentally assumes that the
903 order of words in memory is the same as the order in registers. This
904 macro need not be a constant.
907 @defmac LIBGCC2_WORDS_BIG_ENDIAN
908 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
909 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
910 used only when compiling @file{libgcc2.c}. Typically the value will be set
911 based on preprocessor defines.
914 @defmac FLOAT_WORDS_BIG_ENDIAN
915 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
916 @code{TFmode} floating point numbers are stored in memory with the word
917 containing the sign bit at the lowest address; otherwise define it to
918 have the value 0. This macro need not be a constant.
920 You need not define this macro if the ordering is the same as for
924 @defmac BITS_PER_UNIT
925 Define this macro to be the number of bits in an addressable storage
926 unit (byte). If you do not define this macro the default is 8.
929 @defmac BITS_PER_WORD
930 Number of bits in a word. If you do not define this macro, the default
931 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
934 @defmac MAX_BITS_PER_WORD
935 Maximum number of bits in a word. If this is undefined, the default is
936 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
937 largest value that @code{BITS_PER_WORD} can have at run-time.
940 @defmac UNITS_PER_WORD
941 Number of storage units in a word; normally the size of a general-purpose
942 register, a power of two from 1 or 8.
945 @defmac MIN_UNITS_PER_WORD
946 Minimum number of units in a word. If this is undefined, the default is
947 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
948 smallest value that @code{UNITS_PER_WORD} can have at run-time.
951 @defmac UNITS_PER_SIMD_WORD
952 Number of units in the vectors that the vectorizer can produce.
953 The default is equal to @code{UNITS_PER_WORD}, because the vectorizer
954 can do some transformations even in absence of specialized @acronym{SIMD}
959 Width of a pointer, in bits. You must specify a value no wider than the
960 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
961 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
962 a value the default is @code{BITS_PER_WORD}.
965 @defmac POINTERS_EXTEND_UNSIGNED
966 A C expression whose value is greater than zero if pointers that need to be
967 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
968 be zero-extended and zero if they are to be sign-extended. If the value
969 is less then zero then there must be an "ptr_extend" instruction that
970 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
972 You need not define this macro if the @code{POINTER_SIZE} is equal
973 to the width of @code{Pmode}.
976 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
977 A macro to update @var{m} and @var{unsignedp} when an object whose type
978 is @var{type} and which has the specified mode and signedness is to be
979 stored in a register. This macro is only called when @var{type} is a
982 On most RISC machines, which only have operations that operate on a full
983 register, define this macro to set @var{m} to @code{word_mode} if
984 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
985 cases, only integer modes should be widened because wider-precision
986 floating-point operations are usually more expensive than their narrower
989 For most machines, the macro definition does not change @var{unsignedp}.
990 However, some machines, have instructions that preferentially handle
991 either signed or unsigned quantities of certain modes. For example, on
992 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
993 sign-extend the result to 64 bits. On such machines, set
994 @var{unsignedp} according to which kind of extension is more efficient.
996 Do not define this macro if it would never modify @var{m}.
999 @defmac PROMOTE_FUNCTION_MODE
1000 Like @code{PROMOTE_MODE}, but is applied to outgoing function arguments or
1001 function return values, as specified by @code{TARGET_PROMOTE_FUNCTION_ARGS}
1002 and @code{TARGET_PROMOTE_FUNCTION_RETURN}, respectively.
1004 The default is @code{PROMOTE_MODE}.
1007 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_ARGS (tree @var{fntype})
1008 This target hook should return @code{true} if the promotion described by
1009 @code{PROMOTE_FUNCTION_MODE} should be done for outgoing function
1013 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_RETURN (tree @var{fntype})
1014 This target hook should return @code{true} if the promotion described by
1015 @code{PROMOTE_FUNCTION_MODE} should be done for the return value of
1018 If this target hook returns @code{true}, @code{FUNCTION_VALUE} must
1019 perform the same promotions done by @code{PROMOTE_FUNCTION_MODE}.
1022 @defmac PARM_BOUNDARY
1023 Normal alignment required for function parameters on the stack, in
1024 bits. All stack parameters receive at least this much alignment
1025 regardless of data type. On most machines, this is the same as the
1029 @defmac STACK_BOUNDARY
1030 Define this macro to the minimum alignment enforced by hardware for the
1031 stack pointer on this machine. The definition is a C expression for the
1032 desired alignment (measured in bits). This value is used as a default
1033 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1034 this should be the same as @code{PARM_BOUNDARY}.
1037 @defmac PREFERRED_STACK_BOUNDARY
1038 Define this macro if you wish to preserve a certain alignment for the
1039 stack pointer, greater than what the hardware enforces. The definition
1040 is a C expression for the desired alignment (measured in bits). This
1041 macro must evaluate to a value equal to or larger than
1042 @code{STACK_BOUNDARY}.
1045 @defmac FUNCTION_BOUNDARY
1046 Alignment required for a function entry point, in bits.
1049 @defmac BIGGEST_ALIGNMENT
1050 Biggest alignment that any data type can require on this machine, in bits.
1053 @defmac MINIMUM_ATOMIC_ALIGNMENT
1054 If defined, the smallest alignment, in bits, that can be given to an
1055 object that can be referenced in one operation, without disturbing any
1056 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1057 on machines that don't have byte or half-word store operations.
1060 @defmac BIGGEST_FIELD_ALIGNMENT
1061 Biggest alignment that any structure or union field can require on this
1062 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1063 structure and union fields only, unless the field alignment has been set
1064 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1067 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1068 An expression for the alignment of a structure field @var{field} if the
1069 alignment computed in the usual way (including applying of
1070 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1071 alignment) is @var{computed}. It overrides alignment only if the
1072 field alignment has not been set by the
1073 @code{__attribute__ ((aligned (@var{n})))} construct.
1076 @defmac MAX_OFILE_ALIGNMENT
1077 Biggest alignment supported by the object file format of this machine.
1078 Use this macro to limit the alignment which can be specified using the
1079 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1080 the default value is @code{BIGGEST_ALIGNMENT}.
1083 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1084 If defined, a C expression to compute the alignment for a variable in
1085 the static store. @var{type} is the data type, and @var{basic-align} is
1086 the alignment that the object would ordinarily have. The value of this
1087 macro is used instead of that alignment to align the object.
1089 If this macro is not defined, then @var{basic-align} is used.
1092 One use of this macro is to increase alignment of medium-size data to
1093 make it all fit in fewer cache lines. Another is to cause character
1094 arrays to be word-aligned so that @code{strcpy} calls that copy
1095 constants to character arrays can be done inline.
1098 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1099 If defined, a C expression to compute the alignment given to a constant
1100 that is being placed in memory. @var{constant} is the constant and
1101 @var{basic-align} is the alignment that the object would ordinarily
1102 have. The value of this macro is used instead of that alignment to
1105 If this macro is not defined, then @var{basic-align} is used.
1107 The typical use of this macro is to increase alignment for string
1108 constants to be word aligned so that @code{strcpy} calls that copy
1109 constants can be done inline.
1112 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1113 If defined, a C expression to compute the alignment for a variable in
1114 the local store. @var{type} is the data type, and @var{basic-align} is
1115 the alignment that the object would ordinarily have. The value of this
1116 macro is used instead of that alignment to align the object.
1118 If this macro is not defined, then @var{basic-align} is used.
1120 One use of this macro is to increase alignment of medium-size data to
1121 make it all fit in fewer cache lines.
1124 @defmac EMPTY_FIELD_BOUNDARY
1125 Alignment in bits to be given to a structure bit-field that follows an
1126 empty field such as @code{int : 0;}.
1128 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1131 @defmac STRUCTURE_SIZE_BOUNDARY
1132 Number of bits which any structure or union's size must be a multiple of.
1133 Each structure or union's size is rounded up to a multiple of this.
1135 If you do not define this macro, the default is the same as
1136 @code{BITS_PER_UNIT}.
1139 @defmac STRICT_ALIGNMENT
1140 Define this macro to be the value 1 if instructions will fail to work
1141 if given data not on the nominal alignment. If instructions will merely
1142 go slower in that case, define this macro as 0.
1145 @defmac PCC_BITFIELD_TYPE_MATTERS
1146 Define this if you wish to imitate the way many other C compilers handle
1147 alignment of bit-fields and the structures that contain them.
1149 The behavior is that the type written for a named bit-field (@code{int},
1150 @code{short}, or other integer type) imposes an alignment for the entire
1151 structure, as if the structure really did contain an ordinary field of
1152 that type. In addition, the bit-field is placed within the structure so
1153 that it would fit within such a field, not crossing a boundary for it.
1155 Thus, on most machines, a named bit-field whose type is written as
1156 @code{int} would not cross a four-byte boundary, and would force
1157 four-byte alignment for the whole structure. (The alignment used may
1158 not be four bytes; it is controlled by the other alignment parameters.)
1160 An unnamed bit-field will not affect the alignment of the containing
1163 If the macro is defined, its definition should be a C expression;
1164 a nonzero value for the expression enables this behavior.
1166 Note that if this macro is not defined, or its value is zero, some
1167 bit-fields may cross more than one alignment boundary. The compiler can
1168 support such references if there are @samp{insv}, @samp{extv}, and
1169 @samp{extzv} insns that can directly reference memory.
1171 The other known way of making bit-fields work is to define
1172 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1173 Then every structure can be accessed with fullwords.
1175 Unless the machine has bit-field instructions or you define
1176 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1177 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1179 If your aim is to make GCC use the same conventions for laying out
1180 bit-fields as are used by another compiler, here is how to investigate
1181 what the other compiler does. Compile and run this program:
1200 printf ("Size of foo1 is %d\n",
1201 sizeof (struct foo1));
1202 printf ("Size of foo2 is %d\n",
1203 sizeof (struct foo2));
1208 If this prints 2 and 5, then the compiler's behavior is what you would
1209 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1212 @defmac BITFIELD_NBYTES_LIMITED
1213 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1214 to aligning a bit-field within the structure.
1217 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELDS (void)
1218 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1219 whether unnamed bitfields affect the alignment of the containing
1220 structure. The hook should return true if the structure should inherit
1221 the alignment requirements of an unnamed bitfield's type.
1224 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1225 Return 1 if a structure or array containing @var{field} should be accessed using
1228 If @var{field} is the only field in the structure, @var{mode} is its
1229 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1230 case where structures of one field would require the structure's mode to
1231 retain the field's mode.
1233 Normally, this is not needed. See the file @file{c4x.h} for an example
1234 of how to use this macro to prevent a structure having a floating point
1235 field from being accessed in an integer mode.
1238 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1239 Define this macro as an expression for the alignment of a type (given
1240 by @var{type} as a tree node) if the alignment computed in the usual
1241 way is @var{computed} and the alignment explicitly specified was
1244 The default is to use @var{specified} if it is larger; otherwise, use
1245 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1248 @defmac MAX_FIXED_MODE_SIZE
1249 An integer expression for the size in bits of the largest integer
1250 machine mode that should actually be used. All integer machine modes of
1251 this size or smaller can be used for structures and unions with the
1252 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1253 (DImode)} is assumed.
1256 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1257 If defined, an expression of type @code{enum machine_mode} that
1258 specifies the mode of the save area operand of a
1259 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1260 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1261 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1262 having its mode specified.
1264 You need not define this macro if it always returns @code{Pmode}. You
1265 would most commonly define this macro if the
1266 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1270 @defmac STACK_SIZE_MODE
1271 If defined, an expression of type @code{enum machine_mode} that
1272 specifies the mode of the size increment operand of an
1273 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1275 You need not define this macro if it always returns @code{word_mode}.
1276 You would most commonly define this macro if the @code{allocate_stack}
1277 pattern needs to support both a 32- and a 64-bit mode.
1280 @defmac TARGET_FLOAT_FORMAT
1281 A code distinguishing the floating point format of the target machine.
1282 There are four defined values:
1285 @item IEEE_FLOAT_FORMAT
1286 This code indicates IEEE floating point. It is the default; there is no
1287 need to define @code{TARGET_FLOAT_FORMAT} when the format is IEEE@.
1289 @item VAX_FLOAT_FORMAT
1290 This code indicates the ``F float'' (for @code{float}) and ``D float''
1291 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
1293 @item IBM_FLOAT_FORMAT
1294 This code indicates the format used on the IBM System/370.
1296 @item C4X_FLOAT_FORMAT
1297 This code indicates the format used on the TMS320C3x/C4x.
1300 If your target uses a floating point format other than these, you must
1301 define a new @var{name}_FLOAT_FORMAT code for it, and add support for
1302 it to @file{real.c}.
1304 The ordering of the component words of floating point values stored in
1305 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1308 @defmac MODE_HAS_NANS (@var{mode})
1309 When defined, this macro should be true if @var{mode} has a NaN
1310 representation. The compiler assumes that NaNs are not equal to
1311 anything (including themselves) and that addition, subtraction,
1312 multiplication and division all return NaNs when one operand is
1315 By default, this macro is true if @var{mode} is a floating-point
1316 mode and the target floating-point format is IEEE@.
1319 @defmac MODE_HAS_INFINITIES (@var{mode})
1320 This macro should be true if @var{mode} can represent infinity. At
1321 present, the compiler uses this macro to decide whether @samp{x - x}
1322 is always defined. By default, the macro is true when @var{mode}
1323 is a floating-point mode and the target format is IEEE@.
1326 @defmac MODE_HAS_SIGNED_ZEROS (@var{mode})
1327 True if @var{mode} distinguishes between positive and negative zero.
1328 The rules are expected to follow the IEEE standard:
1332 @samp{x + x} has the same sign as @samp{x}.
1335 If the sum of two values with opposite sign is zero, the result is
1336 positive for all rounding modes expect towards @minus{}infinity, for
1337 which it is negative.
1340 The sign of a product or quotient is negative when exactly one
1341 of the operands is negative.
1344 The default definition is true if @var{mode} is a floating-point
1345 mode and the target format is IEEE@.
1348 @defmac MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1349 If defined, this macro should be true for @var{mode} if it has at
1350 least one rounding mode in which @samp{x} and @samp{-x} can be
1351 rounded to numbers of different magnitude. Two such modes are
1352 towards @minus{}infinity and towards +infinity.
1354 The default definition of this macro is true if @var{mode} is
1355 a floating-point mode and the target format is IEEE@.
1358 @defmac ROUND_TOWARDS_ZERO
1359 If defined, this macro should be true if the prevailing rounding
1360 mode is towards zero. A true value has the following effects:
1364 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1367 @file{libgcc.a}'s floating-point emulator will round towards zero
1368 rather than towards nearest.
1371 The compiler's floating-point emulator will round towards zero after
1372 doing arithmetic, and when converting from the internal float format to
1376 The macro does not affect the parsing of string literals. When the
1377 primary rounding mode is towards zero, library functions like
1378 @code{strtod} might still round towards nearest, and the compiler's
1379 parser should behave like the target's @code{strtod} where possible.
1381 Not defining this macro is equivalent to returning zero.
1384 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1385 This macro should return true if floats with @var{size}
1386 bits do not have a NaN or infinity representation, but use the largest
1387 exponent for normal numbers instead.
1389 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1390 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1391 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1392 floating-point arithmetic.
1394 The default definition of this macro returns false for all sizes.
1397 @deftypefn {Target Hook} bool TARGET_VECTOR_OPAQUE_P (tree @var{type})
1398 This target hook should return @code{true} a vector is opaque. That
1399 is, if no cast is needed when copying a vector value of type
1400 @var{type} into another vector lvalue of the same size. Vector opaque
1401 types cannot be initialized. The default is that there are no such
1405 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1406 This target hook returns @code{true} if bit-fields in the given
1407 @var{record_type} are to be laid out following the rules of Microsoft
1408 Visual C/C++, namely: (i) a bit-field won't share the same storage
1409 unit with the previous bit-field if their underlying types have
1410 different sizes, and the bit-field will be aligned to the highest
1411 alignment of the underlying types of itself and of the previous
1412 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1413 the whole enclosing structure, even if it is unnamed; except that
1414 (iii) a zero-sized bit-field will be disregarded unless it follows
1415 another bit-field of nonzero size. If this hook returns @code{true},
1416 other macros that control bit-field layout are ignored.
1418 When a bit-field is inserted into a packed record, the whole size
1419 of the underlying type is used by one or more same-size adjacent
1420 bit-fields (that is, if its long:3, 32 bits is used in the record,
1421 and any additional adjacent long bit-fields are packed into the same
1422 chunk of 32 bits. However, if the size changes, a new field of that
1423 size is allocated). In an unpacked record, this is the same as using
1424 alignment, but not equivalent when packing.
1426 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1427 the latter will take precedence. If @samp{__attribute__((packed))} is
1428 used on a single field when MS bit-fields are in use, it will take
1429 precedence for that field, but the alignment of the rest of the structure
1430 may affect its placement.
1433 @deftypefn {Target Hook} {bool} TARGET_DECIMAL_FLOAT_SUPPORTED_P (void)
1434 Returns true if the target supports decimal floating point.
1437 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_FUNDAMENTAL_TYPE (tree @var{type})
1438 If your target defines any fundamental types, define this hook to
1439 return the appropriate encoding for these types as part of a C++
1440 mangled name. The @var{type} argument is the tree structure
1441 representing the type to be mangled. The hook may be applied to trees
1442 which are not target-specific fundamental types; it should return
1443 @code{NULL} for all such types, as well as arguments it does not
1444 recognize. If the return value is not @code{NULL}, it must point to
1445 a statically-allocated string constant.
1447 Target-specific fundamental types might be new fundamental types or
1448 qualified versions of ordinary fundamental types. Encode new
1449 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1450 is the name used for the type in source code, and @var{n} is the
1451 length of @var{name} in decimal. Encode qualified versions of
1452 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1453 @var{name} is the name used for the type qualifier in source code,
1454 @var{n} is the length of @var{name} as above, and @var{code} is the
1455 code used to represent the unqualified version of this type. (See
1456 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1457 codes.) In both cases the spaces are for clarity; do not include any
1458 spaces in your string.
1460 The default version of this hook always returns @code{NULL}, which is
1461 appropriate for a target that does not define any new fundamental
1466 @section Layout of Source Language Data Types
1468 These macros define the sizes and other characteristics of the standard
1469 basic data types used in programs being compiled. Unlike the macros in
1470 the previous section, these apply to specific features of C and related
1471 languages, rather than to fundamental aspects of storage layout.
1473 @defmac INT_TYPE_SIZE
1474 A C expression for the size in bits of the type @code{int} on the
1475 target machine. If you don't define this, the default is one word.
1478 @defmac SHORT_TYPE_SIZE
1479 A C expression for the size in bits of the type @code{short} on the
1480 target machine. If you don't define this, the default is half a word.
1481 (If this would be less than one storage unit, it is rounded up to one
1485 @defmac LONG_TYPE_SIZE
1486 A C expression for the size in bits of the type @code{long} on the
1487 target machine. If you don't define this, the default is one word.
1490 @defmac ADA_LONG_TYPE_SIZE
1491 On some machines, the size used for the Ada equivalent of the type
1492 @code{long} by a native Ada compiler differs from that used by C@. In
1493 that situation, define this macro to be a C expression to be used for
1494 the size of that type. If you don't define this, the default is the
1495 value of @code{LONG_TYPE_SIZE}.
1498 @defmac LONG_LONG_TYPE_SIZE
1499 A C expression for the size in bits of the type @code{long long} on the
1500 target machine. If you don't define this, the default is two
1501 words. If you want to support GNU Ada on your machine, the value of this
1502 macro must be at least 64.
1505 @defmac CHAR_TYPE_SIZE
1506 A C expression for the size in bits of the type @code{char} on the
1507 target machine. If you don't define this, the default is
1508 @code{BITS_PER_UNIT}.
1511 @defmac BOOL_TYPE_SIZE
1512 A C expression for the size in bits of the C++ type @code{bool} and
1513 C99 type @code{_Bool} on the target machine. If you don't define
1514 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1517 @defmac FLOAT_TYPE_SIZE
1518 A C expression for the size in bits of the type @code{float} on the
1519 target machine. If you don't define this, the default is one word.
1522 @defmac DOUBLE_TYPE_SIZE
1523 A C expression for the size in bits of the type @code{double} on the
1524 target machine. If you don't define this, the default is two
1528 @defmac LONG_DOUBLE_TYPE_SIZE
1529 A C expression for the size in bits of the type @code{long double} on
1530 the target machine. If you don't define this, the default is two
1534 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1535 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1536 if you want routines in @file{libgcc2.a} for a size other than
1537 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1538 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1541 @defmac LIBGCC2_HAS_DF_MODE
1542 Define this macro if neither @code{LIBGCC2_DOUBLE_TYPE_SIZE} nor
1543 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1544 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1545 anyway. If you don't define this and either @code{LIBGCC2_DOUBLE_TYPE_SIZE}
1546 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1550 @defmac LIBGCC2_HAS_XF_MODE
1551 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1552 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1553 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1554 is 80 then the default is 1, otherwise it is 0.
1557 @defmac LIBGCC2_HAS_TF_MODE
1558 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1559 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1560 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1561 is 128 then the default is 1, otherwise it is 0.
1568 Define these macros to be the size in bits of the mantissa of
1569 @code{SFmode}, @code{DFmode}, @code{XFmode} and @code{TFmode} values,
1570 if the defaults in @file{libgcc2.h} are inappropriate. By default,
1571 @code{FLT_MANT_DIG} is used for @code{SF_SIZE}, @code{LDBL_MANT_DIG}
1572 for @code{XF_SIZE} and @code{TF_SIZE}, and @code{DBL_MANT_DIG} or
1573 @code{LDBL_MANT_DIG} for @code{DF_SIZE} according to whether
1574 @code{LIBGCC2_DOUBLE_TYPE_SIZE} or
1575 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64.
1578 @defmac TARGET_FLT_EVAL_METHOD
1579 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1580 assuming, if applicable, that the floating-point control word is in its
1581 default state. If you do not define this macro the value of
1582 @code{FLT_EVAL_METHOD} will be zero.
1585 @defmac WIDEST_HARDWARE_FP_SIZE
1586 A C expression for the size in bits of the widest floating-point format
1587 supported by the hardware. If you define this macro, you must specify a
1588 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1589 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1593 @defmac DEFAULT_SIGNED_CHAR
1594 An expression whose value is 1 or 0, according to whether the type
1595 @code{char} should be signed or unsigned by default. The user can
1596 always override this default with the options @option{-fsigned-char}
1597 and @option{-funsigned-char}.
1600 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1601 This target hook should return true if the compiler should give an
1602 @code{enum} type only as many bytes as it takes to represent the range
1603 of possible values of that type. It should return false if all
1604 @code{enum} types should be allocated like @code{int}.
1606 The default is to return false.
1610 A C expression for a string describing the name of the data type to use
1611 for size values. The typedef name @code{size_t} is defined using the
1612 contents of the string.
1614 The string can contain more than one keyword. If so, separate them with
1615 spaces, and write first any length keyword, then @code{unsigned} if
1616 appropriate, and finally @code{int}. The string must exactly match one
1617 of the data type names defined in the function
1618 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1619 omit @code{int} or change the order---that would cause the compiler to
1622 If you don't define this macro, the default is @code{"long unsigned
1626 @defmac PTRDIFF_TYPE
1627 A C expression for a string describing the name of the data type to use
1628 for the result of subtracting two pointers. The typedef name
1629 @code{ptrdiff_t} is defined using the contents of the string. See
1630 @code{SIZE_TYPE} above for more information.
1632 If you don't define this macro, the default is @code{"long int"}.
1636 A C expression for a string describing the name of the data type to use
1637 for wide characters. The typedef name @code{wchar_t} is defined using
1638 the contents of the string. See @code{SIZE_TYPE} above for more
1641 If you don't define this macro, the default is @code{"int"}.
1644 @defmac WCHAR_TYPE_SIZE
1645 A C expression for the size in bits of the data type for wide
1646 characters. This is used in @code{cpp}, which cannot make use of
1651 A C expression for a string describing the name of the data type to
1652 use for wide characters passed to @code{printf} and returned from
1653 @code{getwc}. The typedef name @code{wint_t} is defined using the
1654 contents of the string. See @code{SIZE_TYPE} above for more
1657 If you don't define this macro, the default is @code{"unsigned int"}.
1661 A C expression for a string describing the name of the data type that
1662 can represent any value of any standard or extended signed integer type.
1663 The typedef name @code{intmax_t} is defined using the contents of the
1664 string. See @code{SIZE_TYPE} above for more information.
1666 If you don't define this macro, the default is the first of
1667 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1668 much precision as @code{long long int}.
1671 @defmac UINTMAX_TYPE
1672 A C expression for a string describing the name of the data type that
1673 can represent any value of any standard or extended unsigned integer
1674 type. The typedef name @code{uintmax_t} is defined using the contents
1675 of the string. See @code{SIZE_TYPE} above for more information.
1677 If you don't define this macro, the default is the first of
1678 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1679 unsigned int"} that has as much precision as @code{long long unsigned
1683 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1684 The C++ compiler represents a pointer-to-member-function with a struct
1691 ptrdiff_t vtable_index;
1698 The C++ compiler must use one bit to indicate whether the function that
1699 will be called through a pointer-to-member-function is virtual.
1700 Normally, we assume that the low-order bit of a function pointer must
1701 always be zero. Then, by ensuring that the vtable_index is odd, we can
1702 distinguish which variant of the union is in use. But, on some
1703 platforms function pointers can be odd, and so this doesn't work. In
1704 that case, we use the low-order bit of the @code{delta} field, and shift
1705 the remainder of the @code{delta} field to the left.
1707 GCC will automatically make the right selection about where to store
1708 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1709 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1710 set such that functions always start at even addresses, but the lowest
1711 bit of pointers to functions indicate whether the function at that
1712 address is in ARM or Thumb mode. If this is the case of your
1713 architecture, you should define this macro to
1714 @code{ptrmemfunc_vbit_in_delta}.
1716 In general, you should not have to define this macro. On architectures
1717 in which function addresses are always even, according to
1718 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1719 @code{ptrmemfunc_vbit_in_pfn}.
1722 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1723 Normally, the C++ compiler uses function pointers in vtables. This
1724 macro allows the target to change to use ``function descriptors''
1725 instead. Function descriptors are found on targets for whom a
1726 function pointer is actually a small data structure. Normally the
1727 data structure consists of the actual code address plus a data
1728 pointer to which the function's data is relative.
1730 If vtables are used, the value of this macro should be the number
1731 of words that the function descriptor occupies.
1734 @defmac TARGET_VTABLE_ENTRY_ALIGN
1735 By default, the vtable entries are void pointers, the so the alignment
1736 is the same as pointer alignment. The value of this macro specifies
1737 the alignment of the vtable entry in bits. It should be defined only
1738 when special alignment is necessary. */
1741 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1742 There are a few non-descriptor entries in the vtable at offsets below
1743 zero. If these entries must be padded (say, to preserve the alignment
1744 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1745 of words in each data entry.
1749 @section Register Usage
1750 @cindex register usage
1752 This section explains how to describe what registers the target machine
1753 has, and how (in general) they can be used.
1755 The description of which registers a specific instruction can use is
1756 done with register classes; see @ref{Register Classes}. For information
1757 on using registers to access a stack frame, see @ref{Frame Registers}.
1758 For passing values in registers, see @ref{Register Arguments}.
1759 For returning values in registers, see @ref{Scalar Return}.
1762 * Register Basics:: Number and kinds of registers.
1763 * Allocation Order:: Order in which registers are allocated.
1764 * Values in Registers:: What kinds of values each reg can hold.
1765 * Leaf Functions:: Renumbering registers for leaf functions.
1766 * Stack Registers:: Handling a register stack such as 80387.
1769 @node Register Basics
1770 @subsection Basic Characteristics of Registers
1772 @c prevent bad page break with this line
1773 Registers have various characteristics.
1775 @defmac FIRST_PSEUDO_REGISTER
1776 Number of hardware registers known to the compiler. They receive
1777 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1778 pseudo register's number really is assigned the number
1779 @code{FIRST_PSEUDO_REGISTER}.
1782 @defmac FIXED_REGISTERS
1783 @cindex fixed register
1784 An initializer that says which registers are used for fixed purposes
1785 all throughout the compiled code and are therefore not available for
1786 general allocation. These would include the stack pointer, the frame
1787 pointer (except on machines where that can be used as a general
1788 register when no frame pointer is needed), the program counter on
1789 machines where that is considered one of the addressable registers,
1790 and any other numbered register with a standard use.
1792 This information is expressed as a sequence of numbers, separated by
1793 commas and surrounded by braces. The @var{n}th number is 1 if
1794 register @var{n} is fixed, 0 otherwise.
1796 The table initialized from this macro, and the table initialized by
1797 the following one, may be overridden at run time either automatically,
1798 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1799 the user with the command options @option{-ffixed-@var{reg}},
1800 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1803 @defmac CALL_USED_REGISTERS
1804 @cindex call-used register
1805 @cindex call-clobbered register
1806 @cindex call-saved register
1807 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1808 clobbered (in general) by function calls as well as for fixed
1809 registers. This macro therefore identifies the registers that are not
1810 available for general allocation of values that must live across
1813 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1814 automatically saves it on function entry and restores it on function
1815 exit, if the register is used within the function.
1818 @defmac CALL_REALLY_USED_REGISTERS
1819 @cindex call-used register
1820 @cindex call-clobbered register
1821 @cindex call-saved register
1822 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1823 that the entire set of @code{FIXED_REGISTERS} be included.
1824 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1825 This macro is optional. If not specified, it defaults to the value
1826 of @code{CALL_USED_REGISTERS}.
1829 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1830 @cindex call-used register
1831 @cindex call-clobbered register
1832 @cindex call-saved register
1833 A C expression that is nonzero if it is not permissible to store a
1834 value of mode @var{mode} in hard register number @var{regno} across a
1835 call without some part of it being clobbered. For most machines this
1836 macro need not be defined. It is only required for machines that do not
1837 preserve the entire contents of a register across a call.
1841 @findex call_used_regs
1844 @findex reg_class_contents
1845 @defmac CONDITIONAL_REGISTER_USAGE
1846 Zero or more C statements that may conditionally modify five variables
1847 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1848 @code{reg_names}, and @code{reg_class_contents}, to take into account
1849 any dependence of these register sets on target flags. The first three
1850 of these are of type @code{char []} (interpreted as Boolean vectors).
1851 @code{global_regs} is a @code{const char *[]}, and
1852 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1853 called, @code{fixed_regs}, @code{call_used_regs},
1854 @code{reg_class_contents}, and @code{reg_names} have been initialized
1855 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1856 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1857 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1858 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1859 command options have been applied.
1861 You need not define this macro if it has no work to do.
1863 @cindex disabling certain registers
1864 @cindex controlling register usage
1865 If the usage of an entire class of registers depends on the target
1866 flags, you may indicate this to GCC by using this macro to modify
1867 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1868 registers in the classes which should not be used by GCC@. Also define
1869 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1870 to return @code{NO_REGS} if it
1871 is called with a letter for a class that shouldn't be used.
1873 (However, if this class is not included in @code{GENERAL_REGS} and all
1874 of the insn patterns whose constraints permit this class are
1875 controlled by target switches, then GCC will automatically avoid using
1876 these registers when the target switches are opposed to them.)
1879 @defmac INCOMING_REGNO (@var{out})
1880 Define this macro if the target machine has register windows. This C
1881 expression returns the register number as seen by the called function
1882 corresponding to the register number @var{out} as seen by the calling
1883 function. Return @var{out} if register number @var{out} is not an
1887 @defmac OUTGOING_REGNO (@var{in})
1888 Define this macro if the target machine has register windows. This C
1889 expression returns the register number as seen by the calling function
1890 corresponding to the register number @var{in} as seen by the called
1891 function. Return @var{in} if register number @var{in} is not an inbound
1895 @defmac LOCAL_REGNO (@var{regno})
1896 Define this macro if the target machine has register windows. This C
1897 expression returns true if the register is call-saved but is in the
1898 register window. Unlike most call-saved registers, such registers
1899 need not be explicitly restored on function exit or during non-local
1904 If the program counter has a register number, define this as that
1905 register number. Otherwise, do not define it.
1908 @node Allocation Order
1909 @subsection Order of Allocation of Registers
1910 @cindex order of register allocation
1911 @cindex register allocation order
1913 @c prevent bad page break with this line
1914 Registers are allocated in order.
1916 @defmac REG_ALLOC_ORDER
1917 If defined, an initializer for a vector of integers, containing the
1918 numbers of hard registers in the order in which GCC should prefer
1919 to use them (from most preferred to least).
1921 If this macro is not defined, registers are used lowest numbered first
1922 (all else being equal).
1924 One use of this macro is on machines where the highest numbered
1925 registers must always be saved and the save-multiple-registers
1926 instruction supports only sequences of consecutive registers. On such
1927 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1928 the highest numbered allocable register first.
1931 @defmac ORDER_REGS_FOR_LOCAL_ALLOC
1932 A C statement (sans semicolon) to choose the order in which to allocate
1933 hard registers for pseudo-registers local to a basic block.
1935 Store the desired register order in the array @code{reg_alloc_order}.
1936 Element 0 should be the register to allocate first; element 1, the next
1937 register; and so on.
1939 The macro body should not assume anything about the contents of
1940 @code{reg_alloc_order} before execution of the macro.
1942 On most machines, it is not necessary to define this macro.
1945 @node Values in Registers
1946 @subsection How Values Fit in Registers
1948 This section discusses the macros that describe which kinds of values
1949 (specifically, which machine modes) each register can hold, and how many
1950 consecutive registers are needed for a given mode.
1952 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
1953 A C expression for the number of consecutive hard registers, starting
1954 at register number @var{regno}, required to hold a value of mode
1957 On a machine where all registers are exactly one word, a suitable
1958 definition of this macro is
1961 #define HARD_REGNO_NREGS(REGNO, MODE) \
1962 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
1967 @defmac REGMODE_NATURAL_SIZE (@var{mode})
1968 Define this macro if the natural size of registers that hold values
1969 of mode @var{mode} is not the word size. It is a C expression that
1970 should give the natural size in bytes for the specified mode. It is
1971 used by the register allocator to try to optimize its results. This
1972 happens for example on SPARC 64-bit where the natural size of
1973 floating-point registers is still 32-bit.
1976 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
1977 A C expression that is nonzero if it is permissible to store a value
1978 of mode @var{mode} in hard register number @var{regno} (or in several
1979 registers starting with that one). For a machine where all registers
1980 are equivalent, a suitable definition is
1983 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
1986 You need not include code to check for the numbers of fixed registers,
1987 because the allocation mechanism considers them to be always occupied.
1989 @cindex register pairs
1990 On some machines, double-precision values must be kept in even/odd
1991 register pairs. You can implement that by defining this macro to reject
1992 odd register numbers for such modes.
1994 The minimum requirement for a mode to be OK in a register is that the
1995 @samp{mov@var{mode}} instruction pattern support moves between the
1996 register and other hard register in the same class and that moving a
1997 value into the register and back out not alter it.
1999 Since the same instruction used to move @code{word_mode} will work for
2000 all narrower integer modes, it is not necessary on any machine for
2001 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2002 you define patterns @samp{movhi}, etc., to take advantage of this. This
2003 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2004 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2007 Many machines have special registers for floating point arithmetic.
2008 Often people assume that floating point machine modes are allowed only
2009 in floating point registers. This is not true. Any registers that
2010 can hold integers can safely @emph{hold} a floating point machine
2011 mode, whether or not floating arithmetic can be done on it in those
2012 registers. Integer move instructions can be used to move the values.
2014 On some machines, though, the converse is true: fixed-point machine
2015 modes may not go in floating registers. This is true if the floating
2016 registers normalize any value stored in them, because storing a
2017 non-floating value there would garble it. In this case,
2018 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2019 floating registers. But if the floating registers do not automatically
2020 normalize, if you can store any bit pattern in one and retrieve it
2021 unchanged without a trap, then any machine mode may go in a floating
2022 register, so you can define this macro to say so.
2024 The primary significance of special floating registers is rather that
2025 they are the registers acceptable in floating point arithmetic
2026 instructions. However, this is of no concern to
2027 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2028 constraints for those instructions.
2030 On some machines, the floating registers are especially slow to access,
2031 so that it is better to store a value in a stack frame than in such a
2032 register if floating point arithmetic is not being done. As long as the
2033 floating registers are not in class @code{GENERAL_REGS}, they will not
2034 be used unless some pattern's constraint asks for one.
2037 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2038 A C expression that is nonzero if it is OK to rename a hard register
2039 @var{from} to another hard register @var{to}.
2041 One common use of this macro is to prevent renaming of a register to
2042 another register that is not saved by a prologue in an interrupt
2045 The default is always nonzero.
2048 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2049 A C expression that is nonzero if a value of mode
2050 @var{mode1} is accessible in mode @var{mode2} without copying.
2052 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2053 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2054 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2055 should be nonzero. If they differ for any @var{r}, you should define
2056 this macro to return zero unless some other mechanism ensures the
2057 accessibility of the value in a narrower mode.
2059 You should define this macro to return nonzero in as many cases as
2060 possible since doing so will allow GCC to perform better register
2064 @defmac AVOID_CCMODE_COPIES
2065 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2066 registers. You should only define this macro if support for copying to/from
2067 @code{CCmode} is incomplete.
2070 @node Leaf Functions
2071 @subsection Handling Leaf Functions
2073 @cindex leaf functions
2074 @cindex functions, leaf
2075 On some machines, a leaf function (i.e., one which makes no calls) can run
2076 more efficiently if it does not make its own register window. Often this
2077 means it is required to receive its arguments in the registers where they
2078 are passed by the caller, instead of the registers where they would
2081 The special treatment for leaf functions generally applies only when
2082 other conditions are met; for example, often they may use only those
2083 registers for its own variables and temporaries. We use the term ``leaf
2084 function'' to mean a function that is suitable for this special
2085 handling, so that functions with no calls are not necessarily ``leaf
2088 GCC assigns register numbers before it knows whether the function is
2089 suitable for leaf function treatment. So it needs to renumber the
2090 registers in order to output a leaf function. The following macros
2093 @defmac LEAF_REGISTERS
2094 Name of a char vector, indexed by hard register number, which
2095 contains 1 for a register that is allowable in a candidate for leaf
2098 If leaf function treatment involves renumbering the registers, then the
2099 registers marked here should be the ones before renumbering---those that
2100 GCC would ordinarily allocate. The registers which will actually be
2101 used in the assembler code, after renumbering, should not be marked with 1
2104 Define this macro only if the target machine offers a way to optimize
2105 the treatment of leaf functions.
2108 @defmac LEAF_REG_REMAP (@var{regno})
2109 A C expression whose value is the register number to which @var{regno}
2110 should be renumbered, when a function is treated as a leaf function.
2112 If @var{regno} is a register number which should not appear in a leaf
2113 function before renumbering, then the expression should yield @minus{}1, which
2114 will cause the compiler to abort.
2116 Define this macro only if the target machine offers a way to optimize the
2117 treatment of leaf functions, and registers need to be renumbered to do
2121 @findex current_function_is_leaf
2122 @findex current_function_uses_only_leaf_regs
2123 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2124 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2125 specially. They can test the C variable @code{current_function_is_leaf}
2126 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2127 set prior to local register allocation and is valid for the remaining
2128 compiler passes. They can also test the C variable
2129 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2130 functions which only use leaf registers.
2131 @code{current_function_uses_only_leaf_regs} is valid after all passes
2132 that modify the instructions have been run and is only useful if
2133 @code{LEAF_REGISTERS} is defined.
2134 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2135 @c of the next paragraph?! --mew 2feb93
2137 @node Stack Registers
2138 @subsection Registers That Form a Stack
2140 There are special features to handle computers where some of the
2141 ``registers'' form a stack. Stack registers are normally written by
2142 pushing onto the stack, and are numbered relative to the top of the
2145 Currently, GCC can only handle one group of stack-like registers, and
2146 they must be consecutively numbered. Furthermore, the existing
2147 support for stack-like registers is specific to the 80387 floating
2148 point coprocessor. If you have a new architecture that uses
2149 stack-like registers, you will need to do substantial work on
2150 @file{reg-stack.c} and write your machine description to cooperate
2151 with it, as well as defining these macros.
2154 Define this if the machine has any stack-like registers.
2157 @defmac FIRST_STACK_REG
2158 The number of the first stack-like register. This one is the top
2162 @defmac LAST_STACK_REG
2163 The number of the last stack-like register. This one is the bottom of
2167 @node Register Classes
2168 @section Register Classes
2169 @cindex register class definitions
2170 @cindex class definitions, register
2172 On many machines, the numbered registers are not all equivalent.
2173 For example, certain registers may not be allowed for indexed addressing;
2174 certain registers may not be allowed in some instructions. These machine
2175 restrictions are described to the compiler using @dfn{register classes}.
2177 You define a number of register classes, giving each one a name and saying
2178 which of the registers belong to it. Then you can specify register classes
2179 that are allowed as operands to particular instruction patterns.
2183 In general, each register will belong to several classes. In fact, one
2184 class must be named @code{ALL_REGS} and contain all the registers. Another
2185 class must be named @code{NO_REGS} and contain no registers. Often the
2186 union of two classes will be another class; however, this is not required.
2188 @findex GENERAL_REGS
2189 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2190 terribly special about the name, but the operand constraint letters
2191 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2192 the same as @code{ALL_REGS}, just define it as a macro which expands
2195 Order the classes so that if class @var{x} is contained in class @var{y}
2196 then @var{x} has a lower class number than @var{y}.
2198 The way classes other than @code{GENERAL_REGS} are specified in operand
2199 constraints is through machine-dependent operand constraint letters.
2200 You can define such letters to correspond to various classes, then use
2201 them in operand constraints.
2203 You should define a class for the union of two classes whenever some
2204 instruction allows both classes. For example, if an instruction allows
2205 either a floating point (coprocessor) register or a general register for a
2206 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2207 which includes both of them. Otherwise you will get suboptimal code.
2209 You must also specify certain redundant information about the register
2210 classes: for each class, which classes contain it and which ones are
2211 contained in it; for each pair of classes, the largest class contained
2214 When a value occupying several consecutive registers is expected in a
2215 certain class, all the registers used must belong to that class.
2216 Therefore, register classes cannot be used to enforce a requirement for
2217 a register pair to start with an even-numbered register. The way to
2218 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2220 Register classes used for input-operands of bitwise-and or shift
2221 instructions have a special requirement: each such class must have, for
2222 each fixed-point machine mode, a subclass whose registers can transfer that
2223 mode to or from memory. For example, on some machines, the operations for
2224 single-byte values (@code{QImode}) are limited to certain registers. When
2225 this is so, each register class that is used in a bitwise-and or shift
2226 instruction must have a subclass consisting of registers from which
2227 single-byte values can be loaded or stored. This is so that
2228 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2230 @deftp {Data type} {enum reg_class}
2231 An enumerated type that must be defined with all the register class names
2232 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2233 must be the last register class, followed by one more enumerated value,
2234 @code{LIM_REG_CLASSES}, which is not a register class but rather
2235 tells how many classes there are.
2237 Each register class has a number, which is the value of casting
2238 the class name to type @code{int}. The number serves as an index
2239 in many of the tables described below.
2242 @defmac N_REG_CLASSES
2243 The number of distinct register classes, defined as follows:
2246 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2250 @defmac REG_CLASS_NAMES
2251 An initializer containing the names of the register classes as C string
2252 constants. These names are used in writing some of the debugging dumps.
2255 @defmac REG_CLASS_CONTENTS
2256 An initializer containing the contents of the register classes, as integers
2257 which are bit masks. The @var{n}th integer specifies the contents of class
2258 @var{n}. The way the integer @var{mask} is interpreted is that
2259 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2261 When the machine has more than 32 registers, an integer does not suffice.
2262 Then the integers are replaced by sub-initializers, braced groupings containing
2263 several integers. Each sub-initializer must be suitable as an initializer
2264 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2265 In this situation, the first integer in each sub-initializer corresponds to
2266 registers 0 through 31, the second integer to registers 32 through 63, and
2270 @defmac REGNO_REG_CLASS (@var{regno})
2271 A C expression whose value is a register class containing hard register
2272 @var{regno}. In general there is more than one such class; choose a class
2273 which is @dfn{minimal}, meaning that no smaller class also contains the
2277 @defmac BASE_REG_CLASS
2278 A macro whose definition is the name of the class to which a valid
2279 base register must belong. A base register is one used in an address
2280 which is the register value plus a displacement.
2283 @defmac MODE_BASE_REG_CLASS (@var{mode})
2284 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2285 the selection of a base register in a mode dependent manner. If
2286 @var{mode} is VOIDmode then it should return the same value as
2287 @code{BASE_REG_CLASS}.
2290 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2291 A C expression whose value is the register class to which a valid
2292 base register must belong in order to be used in a base plus index
2293 register address. You should define this macro if base plus index
2294 addresses have different requirements than other base register uses.
2297 @defmac INDEX_REG_CLASS
2298 A macro whose definition is the name of the class to which a valid
2299 index register must belong. An index register is one used in an
2300 address where its value is either multiplied by a scale factor or
2301 added to another register (as well as added to a displacement).
2304 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2305 A C expression which is nonzero if register number @var{num} is
2306 suitable for use as a base register in operand addresses. It may be
2307 either a suitable hard register or a pseudo register that has been
2308 allocated such a hard register.
2311 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2312 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2313 that expression may examine the mode of the memory reference in
2314 @var{mode}. You should define this macro if the mode of the memory
2315 reference affects whether a register may be used as a base register. If
2316 you define this macro, the compiler will use it instead of
2317 @code{REGNO_OK_FOR_BASE_P}.
2320 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2321 A C expression which is nonzero if register number @var{num} is suitable for
2322 use as a base register in base plus index operand addresses, accessing
2323 memory in mode @var{mode}. It may be either a suitable hard register or a
2324 pseudo register that has been allocated such a hard register. You should
2325 define this macro if base plus index addresses have different requirements
2326 than other base register uses.
2329 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2330 A C expression which is nonzero if register number @var{num} is
2331 suitable for use as an index register in operand addresses. It may be
2332 either a suitable hard register or a pseudo register that has been
2333 allocated such a hard register.
2335 The difference between an index register and a base register is that
2336 the index register may be scaled. If an address involves the sum of
2337 two registers, neither one of them scaled, then either one may be
2338 labeled the ``base'' and the other the ``index''; but whichever
2339 labeling is used must fit the machine's constraints of which registers
2340 may serve in each capacity. The compiler will try both labelings,
2341 looking for one that is valid, and will reload one or both registers
2342 only if neither labeling works.
2345 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2346 A C expression that places additional restrictions on the register class
2347 to use when it is necessary to copy value @var{x} into a register in class
2348 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2349 another, smaller class. On many machines, the following definition is
2353 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2356 Sometimes returning a more restrictive class makes better code. For
2357 example, on the 68000, when @var{x} is an integer constant that is in range
2358 for a @samp{moveq} instruction, the value of this macro is always
2359 @code{DATA_REGS} as long as @var{class} includes the data registers.
2360 Requiring a data register guarantees that a @samp{moveq} will be used.
2362 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2363 @var{class} is if @var{x} is a legitimate constant which cannot be
2364 loaded into some register class. By returning @code{NO_REGS} you can
2365 force @var{x} into a memory location. For example, rs6000 can load
2366 immediate values into general-purpose registers, but does not have an
2367 instruction for loading an immediate value into a floating-point
2368 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2369 @var{x} is a floating-point constant. If the constant can't be loaded
2370 into any kind of register, code generation will be better if
2371 @code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2372 of using @code{PREFERRED_RELOAD_CLASS}.
2375 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2376 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2377 input reloads. If you don't define this macro, the default is to use
2378 @var{class}, unchanged.
2381 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2382 A C expression that places additional restrictions on the register class
2383 to use when it is necessary to be able to hold a value of mode
2384 @var{mode} in a reload register for which class @var{class} would
2387 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2388 there are certain modes that simply can't go in certain reload classes.
2390 The value is a register class; perhaps @var{class}, or perhaps another,
2393 Don't define this macro unless the target machine has limitations which
2394 require the macro to do something nontrivial.
2397 @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})
2398 Many machines have some registers that cannot be copied directly to or
2399 from memory or even from other types of registers. An example is the
2400 @samp{MQ} register, which on most machines, can only be copied to or
2401 from general registers, but not memory. Below, we shall be using the
2402 term 'intermediate register' when a move operation cannot be performed
2403 directly, but has to be done by copying the source into the intermediate
2404 register first, and then copying the intermediate register to the
2405 destination. An intermediate register always has the same mode as
2406 source and destination. Since it holds the actual value being copied,
2407 reload might apply optimizations to re-use an intermediate register
2408 and eliding the copy from the source when it can determine that the
2409 intermediate register still holds the required value.
2411 Another kind of secondary reload is required on some machines which
2412 allow copying all registers to and from memory, but require a scratch
2413 register for stores to some memory locations (e.g., those with symbolic
2414 address on the RT, and those with certain symbolic address on the SPARC
2415 when compiling PIC)@. Scratch registers need not have the same mode
2416 as the value being copied, and usually hold a different value that
2417 that being copied. Special patterns in the md file are needed to
2418 describe how the copy is performed with the help of the scratch register;
2419 these patterns also describe the number, register class(es) and mode(s)
2420 of the scratch register(s).
2422 In some cases, both an intermediate and a scratch register are required.
2424 For input reloads, this target hook is called with nonzero @var{in_p},
2425 and @var{x} is an rtx that needs to be copied to a register in of class
2426 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2427 hook is called with zero @var{in_p}, and a register of class @var{reload_mode}
2428 needs to be copied to rtx @var{x} in @var{reload_mode}.
2430 If copying a register of @var{reload_class} from/to @var{x} requires
2431 an intermediate register, the hook @code{secondary_reload} should
2432 return the register class required for this intermediate register.
2433 If no intermediate register is required, it should return NO_REGS.
2434 If more than one intermediate register is required, describe the one
2435 that is closest in the copy chain to the reload register.
2437 If scratch registers are needed, you also have to describe how to
2438 perform the copy from/to the reload register to/from this
2439 closest intermediate register. Or if no intermediate register is
2440 required, but still a scratch register is needed, describe the
2441 copy from/to the reload register to/from the reload operand @var{x}.
2443 You do this by setting @code{sri->icode} to the instruction code of a pattern
2444 in the md file which performs the move. Operands 0 and 1 are the output
2445 and input of this copy, respectively. Operands from operand 2 onward are
2446 for scratch operands. These scratch operands must have a mode, and a
2447 single-register-class
2448 @c [later: or memory]
2451 When an intermediate register is used, the @code{secondary_reload}
2452 hook will be called again to determine how to copy the intermediate
2453 register to/from the reload operand @var{x}, so your hook must also
2454 have code to handle the register class of the intermediate operand.
2456 @c [For later: maybe we'll allow multi-alternative reload patterns -
2457 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2458 @c and match the constraints of input and output to determine the required
2459 @c alternative. A restriction would be that constraints used to match
2460 @c against reloads registers would have to be written as register class
2461 @c constraints, or we need a new target macro / hook that tells us if an
2462 @c arbitrary constraint can match an unknown register of a given class.
2463 @c Such a macro / hook would also be useful in other places.]
2466 @var{x} might be a pseudo-register or a @code{subreg} of a
2467 pseudo-register, which could either be in a hard register or in memory.
2468 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2469 in memory and the hard register number if it is in a register.
2471 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2472 currently not supported. For the time being, you will have to continue
2473 to use @code{SECONDARY_MEMORY_NEEDED} for that purpose.
2475 @code{copy_cost} also uses this target hook to find out how values are
2476 copied. If you want it to include some extra cost for the need to allocate
2477 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2478 Or if two dependent moves are supposed to have a lower cost than the sum
2479 of the individual moves due to expected fortuitous scheduling and/or special
2480 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2483 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2484 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2485 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2486 These macros are obsolete, new ports should use the target hook
2487 @code{TARGET_SECONDARY_RELOAD} instead.
2489 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2490 target hook. Older ports still define these macros to indicate to the
2491 reload phase that it may
2492 need to allocate at least one register for a reload in addition to the
2493 register to contain the data. Specifically, if copying @var{x} to a
2494 register @var{class} in @var{mode} requires an intermediate register,
2495 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2496 largest register class all of whose registers can be used as
2497 intermediate registers or scratch registers.
2499 If copying a register @var{class} in @var{mode} to @var{x} requires an
2500 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2501 was supposed to be defined be defined to return the largest register
2502 class required. If the
2503 requirements for input and output reloads were the same, the macro
2504 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2507 The values returned by these macros are often @code{GENERAL_REGS}.
2508 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2509 can be directly copied to or from a register of @var{class} in
2510 @var{mode} without requiring a scratch register. Do not define this
2511 macro if it would always return @code{NO_REGS}.
2513 If a scratch register is required (either with or without an
2514 intermediate register), you were supposed to define patterns for
2515 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2516 (@pxref{Standard Names}. These patterns, which were normally
2517 implemented with a @code{define_expand}, should be similar to the
2518 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2521 These patterns need constraints for the reload register and scratch
2523 contain a single register class. If the original reload register (whose
2524 class is @var{class}) can meet the constraint given in the pattern, the
2525 value returned by these macros is used for the class of the scratch
2526 register. Otherwise, two additional reload registers are required.
2527 Their classes are obtained from the constraints in the insn pattern.
2529 @var{x} might be a pseudo-register or a @code{subreg} of a
2530 pseudo-register, which could either be in a hard register or in memory.
2531 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2532 in memory and the hard register number if it is in a register.
2534 These macros should not be used in the case where a particular class of
2535 registers can only be copied to memory and not to another class of
2536 registers. In that case, secondary reload registers are not needed and
2537 would not be helpful. Instead, a stack location must be used to perform
2538 the copy and the @code{mov@var{m}} pattern should use memory as an
2539 intermediate storage. This case often occurs between floating-point and
2543 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2544 Certain machines have the property that some registers cannot be copied
2545 to some other registers without using memory. Define this macro on
2546 those machines to be a C expression that is nonzero if objects of mode
2547 @var{m} in registers of @var{class1} can only be copied to registers of
2548 class @var{class2} by storing a register of @var{class1} into memory
2549 and loading that memory location into a register of @var{class2}.
2551 Do not define this macro if its value would always be zero.
2554 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2555 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2556 allocates a stack slot for a memory location needed for register copies.
2557 If this macro is defined, the compiler instead uses the memory location
2558 defined by this macro.
2560 Do not define this macro if you do not define
2561 @code{SECONDARY_MEMORY_NEEDED}.
2564 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2565 When the compiler needs a secondary memory location to copy between two
2566 registers of mode @var{mode}, it normally allocates sufficient memory to
2567 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2568 load operations in a mode that many bits wide and whose class is the
2569 same as that of @var{mode}.
2571 This is right thing to do on most machines because it ensures that all
2572 bits of the register are copied and prevents accesses to the registers
2573 in a narrower mode, which some machines prohibit for floating-point
2576 However, this default behavior is not correct on some machines, such as
2577 the DEC Alpha, that store short integers in floating-point registers
2578 differently than in integer registers. On those machines, the default
2579 widening will not work correctly and you must define this macro to
2580 suppress that widening in some cases. See the file @file{alpha.h} for
2583 Do not define this macro if you do not define
2584 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2585 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2588 @defmac SMALL_REGISTER_CLASSES
2589 On some machines, it is risky to let hard registers live across arbitrary
2590 insns. Typically, these machines have instructions that require values
2591 to be in specific registers (like an accumulator), and reload will fail
2592 if the required hard register is used for another purpose across such an
2595 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2596 value on these machines. When this macro has a nonzero value, the
2597 compiler will try to minimize the lifetime of hard registers.
2599 It is always safe to define this macro with a nonzero value, but if you
2600 unnecessarily define it, you will reduce the amount of optimizations
2601 that can be performed in some cases. If you do not define this macro
2602 with a nonzero value when it is required, the compiler will run out of
2603 spill registers and print a fatal error message. For most machines, you
2604 should not define this macro at all.
2607 @defmac CLASS_LIKELY_SPILLED_P (@var{class})
2608 A C expression whose value is nonzero if pseudos that have been assigned
2609 to registers of class @var{class} would likely be spilled because
2610 registers of @var{class} are needed for spill registers.
2612 The default value of this macro returns 1 if @var{class} has exactly one
2613 register and zero otherwise. On most machines, this default should be
2614 used. Only define this macro to some other expression if pseudos
2615 allocated by @file{local-alloc.c} end up in memory because their hard
2616 registers were needed for spill registers. If this macro returns nonzero
2617 for those classes, those pseudos will only be allocated by
2618 @file{global.c}, which knows how to reallocate the pseudo to another
2619 register. If there would not be another register available for
2620 reallocation, you should not change the definition of this macro since
2621 the only effect of such a definition would be to slow down register
2625 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2626 A C expression for the maximum number of consecutive registers
2627 of class @var{class} needed to hold a value of mode @var{mode}.
2629 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2630 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2631 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2632 @var{mode})} for all @var{regno} values in the class @var{class}.
2634 This macro helps control the handling of multiple-word values
2638 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2639 If defined, a C expression that returns nonzero for a @var{class} for which
2640 a change from mode @var{from} to mode @var{to} is invalid.
2642 For the example, loading 32-bit integer or floating-point objects into
2643 floating-point registers on the Alpha extends them to 64 bits.
2644 Therefore loading a 64-bit object and then storing it as a 32-bit object
2645 does not store the low-order 32 bits, as would be the case for a normal
2646 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2650 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2651 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2652 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2656 @node Old Constraints
2657 @section Obsolete Macros for Defining Constraints
2658 @cindex defining constraints, obsolete method
2659 @cindex constraints, defining, obsolete method
2661 Machine-specific constraints can be defined with these macros instead
2662 of the machine description constructs described in @ref{Define
2663 Constraints}. This mechanism is obsolete. New ports should not use
2664 it; old ports should convert to the new mechanism.
2666 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2667 For the constraint at the start of @var{str}, which starts with the letter
2668 @var{c}, return the length. This allows you to have register class /
2669 constant / extra constraints that are longer than a single letter;
2670 you don't need to define this macro if you can do with single-letter
2671 constraints only. The definition of this macro should use
2672 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2673 to handle specially.
2674 There are some sanity checks in genoutput.c that check the constraint lengths
2675 for the md file, so you can also use this macro to help you while you are
2676 transitioning from a byzantine single-letter-constraint scheme: when you
2677 return a negative length for a constraint you want to re-use, genoutput
2678 will complain about every instance where it is used in the md file.
2681 @defmac REG_CLASS_FROM_LETTER (@var{char})
2682 A C expression which defines the machine-dependent operand constraint
2683 letters for register classes. If @var{char} is such a letter, the
2684 value should be the register class corresponding to it. Otherwise,
2685 the value should be @code{NO_REGS}. The register letter @samp{r},
2686 corresponding to class @code{GENERAL_REGS}, will not be passed
2687 to this macro; you do not need to handle it.
2690 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2691 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2692 passed in @var{str}, so that you can use suffixes to distinguish between
2696 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2697 A C expression that defines the machine-dependent operand constraint
2698 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2699 particular ranges of integer values. If @var{c} is one of those
2700 letters, the expression should check that @var{value}, an integer, is in
2701 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2702 not one of those letters, the value should be 0 regardless of
2706 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2707 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2708 string passed in @var{str}, so that you can use suffixes to distinguish
2709 between different variants.
2712 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2713 A C expression that defines the machine-dependent operand constraint
2714 letters that specify particular ranges of @code{const_double} values
2715 (@samp{G} or @samp{H}).
2717 If @var{c} is one of those letters, the expression should check that
2718 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2719 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2720 letters, the value should be 0 regardless of @var{value}.
2722 @code{const_double} is used for all floating-point constants and for
2723 @code{DImode} fixed-point constants. A given letter can accept either
2724 or both kinds of values. It can use @code{GET_MODE} to distinguish
2725 between these kinds.
2728 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2729 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2730 string passed in @var{str}, so that you can use suffixes to distinguish
2731 between different variants.
2734 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2735 A C expression that defines the optional machine-dependent constraint
2736 letters that can be used to segregate specific types of operands, usually
2737 memory references, for the target machine. Any letter that is not
2738 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2739 @code{REG_CLASS_FROM_CONSTRAINT}
2740 may be used. Normally this macro will not be defined.
2742 If it is required for a particular target machine, it should return 1
2743 if @var{value} corresponds to the operand type represented by the
2744 constraint letter @var{c}. If @var{c} is not defined as an extra
2745 constraint, the value returned should be 0 regardless of @var{value}.
2747 For example, on the ROMP, load instructions cannot have their output
2748 in r0 if the memory reference contains a symbolic address. Constraint
2749 letter @samp{Q} is defined as representing a memory address that does
2750 @emph{not} contain a symbolic address. An alternative is specified with
2751 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2752 alternative specifies @samp{m} on the input and a register class that
2753 does not include r0 on the output.
2756 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2757 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2758 in @var{str}, so that you can use suffixes to distinguish between different
2762 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2763 A C expression that defines the optional machine-dependent constraint
2764 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2765 be treated like memory constraints by the reload pass.
2767 It should return 1 if the operand type represented by the constraint
2768 at the start of @var{str}, the first letter of which is the letter @var{c},
2769 comprises a subset of all memory references including
2770 all those whose address is simply a base register. This allows the reload
2771 pass to reload an operand, if it does not directly correspond to the operand
2772 type of @var{c}, by copying its address into a base register.
2774 For example, on the S/390, some instructions do not accept arbitrary
2775 memory references, but only those that do not make use of an index
2776 register. The constraint letter @samp{Q} is defined via
2777 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2778 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2779 a @samp{Q} constraint can handle any memory operand, because the
2780 reload pass knows it can be reloaded by copying the memory address
2781 into a base register if required. This is analogous to the way
2782 a @samp{o} constraint can handle any memory operand.
2785 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2786 A C expression that defines the optional machine-dependent constraint
2787 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2788 @code{EXTRA_CONSTRAINT_STR}, that should
2789 be treated like address constraints by the reload pass.
2791 It should return 1 if the operand type represented by the constraint
2792 at the start of @var{str}, which starts with the letter @var{c}, comprises
2793 a subset of all memory addresses including
2794 all those that consist of just a base register. This allows the reload
2795 pass to reload an operand, if it does not directly correspond to the operand
2796 type of @var{str}, by copying it into a base register.
2798 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2799 be used with the @code{address_operand} predicate. It is treated
2800 analogously to the @samp{p} constraint.
2803 @node Stack and Calling
2804 @section Stack Layout and Calling Conventions
2805 @cindex calling conventions
2807 @c prevent bad page break with this line
2808 This describes the stack layout and calling conventions.
2812 * Exception Handling::
2817 * Register Arguments::
2819 * Aggregate Return::
2824 * Stack Smashing Protection::
2828 @subsection Basic Stack Layout
2829 @cindex stack frame layout
2830 @cindex frame layout
2832 @c prevent bad page break with this line
2833 Here is the basic stack layout.
2835 @defmac STACK_GROWS_DOWNWARD
2836 Define this macro if pushing a word onto the stack moves the stack
2837 pointer to a smaller address.
2839 When we say, ``define this macro if @dots{}'', it means that the
2840 compiler checks this macro only with @code{#ifdef} so the precise
2841 definition used does not matter.
2844 @defmac STACK_PUSH_CODE
2845 This macro defines the operation used when something is pushed
2846 on the stack. In RTL, a push operation will be
2847 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2849 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2850 and @code{POST_INC}. Which of these is correct depends on
2851 the stack direction and on whether the stack pointer points
2852 to the last item on the stack or whether it points to the
2853 space for the next item on the stack.
2855 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2856 defined, which is almost always right, and @code{PRE_INC} otherwise,
2857 which is often wrong.
2860 @defmac FRAME_GROWS_DOWNWARD
2861 Define this macro to nonzero value if the addresses of local variable slots
2862 are at negative offsets from the frame pointer.
2865 @defmac ARGS_GROW_DOWNWARD
2866 Define this macro if successive arguments to a function occupy decreasing
2867 addresses on the stack.
2870 @defmac STARTING_FRAME_OFFSET
2871 Offset from the frame pointer to the first local variable slot to be allocated.
2873 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2874 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2875 Otherwise, it is found by adding the length of the first slot to the
2876 value @code{STARTING_FRAME_OFFSET}.
2877 @c i'm not sure if the above is still correct.. had to change it to get
2878 @c rid of an overfull. --mew 2feb93
2881 @defmac STACK_ALIGNMENT_NEEDED
2882 Define to zero to disable final alignment of the stack during reload.
2883 The nonzero default for this macro is suitable for most ports.
2885 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
2886 is a register save block following the local block that doesn't require
2887 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2888 stack alignment and do it in the backend.
2891 @defmac STACK_POINTER_OFFSET
2892 Offset from the stack pointer register to the first location at which
2893 outgoing arguments are placed. If not specified, the default value of
2894 zero is used. This is the proper value for most machines.
2896 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2897 the first location at which outgoing arguments are placed.
2900 @defmac FIRST_PARM_OFFSET (@var{fundecl})
2901 Offset from the argument pointer register to the first argument's
2902 address. On some machines it may depend on the data type of the
2905 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2906 the first argument's address.
2909 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
2910 Offset from the stack pointer register to an item dynamically allocated
2911 on the stack, e.g., by @code{alloca}.
2913 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2914 length of the outgoing arguments. The default is correct for most
2915 machines. See @file{function.c} for details.
2918 @defmac INITIAL_FRAME_ADDRESS_RTX
2919 A C expression whose value is RTL representing the address of the initial
2920 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
2921 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
2922 default value will be used. Define this macro in order to make frame pointer
2923 elimination work in the presence of @code{__builtin_frame_address (count)} and
2924 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
2927 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2928 A C expression whose value is RTL representing the address in a stack
2929 frame where the pointer to the caller's frame is stored. Assume that
2930 @var{frameaddr} is an RTL expression for the address of the stack frame
2933 If you don't define this macro, the default is to return the value
2934 of @var{frameaddr}---that is, the stack frame address is also the
2935 address of the stack word that points to the previous frame.
2938 @defmac SETUP_FRAME_ADDRESSES
2939 If defined, a C expression that produces the machine-specific code to
2940 setup the stack so that arbitrary frames can be accessed. For example,
2941 on the SPARC, we must flush all of the register windows to the stack
2942 before we can access arbitrary stack frames. You will seldom need to
2946 @deftypefn {Target Hook} bool TARGET_BUILTIN_SETJMP_FRAME_VALUE ()
2947 This target hook should return an rtx that is used to store
2948 the address of the current frame into the built in @code{setjmp} buffer.
2949 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2950 machines. One reason you may need to define this target hook is if
2951 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2954 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2955 A C expression whose value is RTL representing the value of the return
2956 address for the frame @var{count} steps up from the current frame, after
2957 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2958 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2959 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2961 The value of the expression must always be the correct address when
2962 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2963 determine the return address of other frames.
2966 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
2967 Define this if the return address of a particular stack frame is accessed
2968 from the frame pointer of the previous stack frame.
2971 @defmac INCOMING_RETURN_ADDR_RTX
2972 A C expression whose value is RTL representing the location of the
2973 incoming return address at the beginning of any function, before the
2974 prologue. This RTL is either a @code{REG}, indicating that the return
2975 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2978 You only need to define this macro if you want to support call frame
2979 debugging information like that provided by DWARF 2.
2981 If this RTL is a @code{REG}, you should also define
2982 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2985 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
2986 A C expression whose value is an integer giving a DWARF 2 column
2987 number that may be used as an alternate return column. This should
2988 be defined only if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
2989 general register, but an alternate column needs to be used for
2993 @defmac DWARF_ZERO_REG
2994 A C expression whose value is an integer giving a DWARF 2 register
2995 number that is considered to always have the value zero. This should
2996 only be defined if the target has an architected zero register, and
2997 someone decided it was a good idea to use that register number to
2998 terminate the stack backtrace. New ports should avoid this.
3001 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3002 This target hook allows the backend to emit frame-related insns that
3003 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3004 info engine will invoke it on insns of the form
3006 (set (reg) (unspec [...] UNSPEC_INDEX))
3010 (set (reg) (unspec_volatile [...] UNSPECV_INDEX)).
3012 to let the backend emit the call frame instructions. @var{label} is
3013 the CFI label attached to the insn, @var{pattern} is the pattern of
3014 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3017 @defmac INCOMING_FRAME_SP_OFFSET
3018 A C expression whose value is an integer giving the offset, in bytes,
3019 from the value of the stack pointer register to the top of the stack
3020 frame at the beginning of any function, before the prologue. The top of
3021 the frame is defined to be the value of the stack pointer in the
3022 previous frame, just before the call instruction.
3024 You only need to define this macro if you want to support call frame
3025 debugging information like that provided by DWARF 2.
3028 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3029 A C expression whose value is an integer giving the offset, in bytes,
3030 from the argument pointer to the canonical frame address (cfa). The
3031 final value should coincide with that calculated by
3032 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3033 during virtual register instantiation.
3035 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
3036 which is correct for most machines; in general, the arguments are found
3037 immediately before the stack frame. Note that this is not the case on
3038 some targets that save registers into the caller's frame, such as SPARC
3039 and rs6000, and so such targets need to define this macro.
3041 You only need to define this macro if the default is incorrect, and you
3042 want to support call frame debugging information like that provided by
3046 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3047 If defined, a C expression whose value is an integer giving the offset
3048 in bytes from the frame pointer to the canonical frame address (cfa).
3049 The final value should conincide with that calculated by
3050 @code{INCOMING_FRAME_SP_OFFSET}.
3052 Normally the CFA is calculated as an offset from the argument pointer,
3053 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3054 variable due to the ABI, this may not be possible. If this macro is
3055 defined, it implies that the virtual register instantiation should be
3056 based on the frame pointer instead of the argument pointer. Only one
3057 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3061 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3062 If defined, a C expression whose value is an integer giving the offset
3063 in bytes from the canonical frame address (cfa) to the frame base used
3064 in DWARF 2 debug information. The default is zero. A different value
3065 may reduce the size of debug information on some ports.
3068 @node Exception Handling
3069 @subsection Exception Handling Support
3070 @cindex exception handling
3072 @defmac EH_RETURN_DATA_REGNO (@var{N})
3073 A C expression whose value is the @var{N}th register number used for
3074 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3075 @var{N} registers are usable.
3077 The exception handling library routines communicate with the exception
3078 handlers via a set of agreed upon registers. Ideally these registers
3079 should be call-clobbered; it is possible to use call-saved registers,
3080 but may negatively impact code size. The target must support at least
3081 2 data registers, but should define 4 if there are enough free registers.
3083 You must define this macro if you want to support call frame exception
3084 handling like that provided by DWARF 2.
3087 @defmac EH_RETURN_STACKADJ_RTX
3088 A C expression whose value is RTL representing a location in which
3089 to store a stack adjustment to be applied before function return.
3090 This is used to unwind the stack to an exception handler's call frame.
3091 It will be assigned zero on code paths that return normally.
3093 Typically this is a call-clobbered hard register that is otherwise
3094 untouched by the epilogue, but could also be a stack slot.
3096 Do not define this macro if the stack pointer is saved and restored
3097 by the regular prolog and epilog code in the call frame itself; in
3098 this case, the exception handling library routines will update the
3099 stack location to be restored in place. Otherwise, you must define
3100 this macro if you want to support call frame exception handling like
3101 that provided by DWARF 2.
3104 @defmac EH_RETURN_HANDLER_RTX
3105 A C expression whose value is RTL representing a location in which
3106 to store the address of an exception handler to which we should
3107 return. It will not be assigned on code paths that return normally.
3109 Typically this is the location in the call frame at which the normal
3110 return address is stored. For targets that return by popping an
3111 address off the stack, this might be a memory address just below
3112 the @emph{target} call frame rather than inside the current call
3113 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3114 been assigned, so it may be used to calculate the location of the
3117 Some targets have more complex requirements than storing to an
3118 address calculable during initial code generation. In that case
3119 the @code{eh_return} instruction pattern should be used instead.
3121 If you want to support call frame exception handling, you must
3122 define either this macro or the @code{eh_return} instruction pattern.
3125 @defmac RETURN_ADDR_OFFSET
3126 If defined, an integer-valued C expression for which rtl will be generated
3127 to add it to the exception handler address before it is searched in the
3128 exception handling tables, and to subtract it again from the address before
3129 using it to return to the exception handler.
3132 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3133 This macro chooses the encoding of pointers embedded in the exception
3134 handling sections. If at all possible, this should be defined such
3135 that the exception handling section will not require dynamic relocations,
3136 and so may be read-only.
3138 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3139 @var{global} is true if the symbol may be affected by dynamic relocations.
3140 The macro should return a combination of the @code{DW_EH_PE_*} defines
3141 as found in @file{dwarf2.h}.
3143 If this macro is not defined, pointers will not be encoded but
3144 represented directly.
3147 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3148 This macro allows the target to emit whatever special magic is required
3149 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3150 Generic code takes care of pc-relative and indirect encodings; this must
3151 be defined if the target uses text-relative or data-relative encodings.
3153 This is a C statement that branches to @var{done} if the format was
3154 handled. @var{encoding} is the format chosen, @var{size} is the number
3155 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3159 @defmac MD_UNWIND_SUPPORT
3160 A string specifying a file to be #include'd in unwind-dw2.c. The file
3161 so included typically defines @code{MD_FALLBACK_FRAME_STATE_FOR}.
3164 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3165 This macro allows the target to add cpu and operating system specific
3166 code to the call-frame unwinder for use when there is no unwind data
3167 available. The most common reason to implement this macro is to unwind
3168 through signal frames.
3170 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
3171 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3172 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3173 for the address of the code being executed and @code{context->cfa} for
3174 the stack pointer value. If the frame can be decoded, the register save
3175 addresses should be updated in @var{fs} and the macro should evaluate to
3176 @code{_URC_NO_REASON}. If the frame cannot be decoded, the macro should
3177 evaluate to @code{_URC_END_OF_STACK}.
3179 For proper signal handling in Java this macro is accompanied by
3180 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3183 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3184 This macro allows the target to add operating system specific code to the
3185 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3186 usually used for signal or interrupt frames.
3188 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3189 @var{context} is an @code{_Unwind_Context};
3190 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3191 for the abi and context in the @code{.unwabi} directive. If the
3192 @code{.unwabi} directive can be handled, the register save addresses should
3193 be updated in @var{fs}.
3196 @defmac TARGET_USES_WEAK_UNWIND_INFO
3197 A C expression that evaluates to true if the target requires unwind
3198 info to be given comdat linkage. Define it to be @code{1} if comdat
3199 linkage is necessary. The default is @code{0}.
3202 @node Stack Checking
3203 @subsection Specifying How Stack Checking is Done
3205 GCC will check that stack references are within the boundaries of
3206 the stack, if the @option{-fstack-check} is specified, in one of three ways:
3210 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3211 will assume that you have arranged for stack checking to be done at
3212 appropriate places in the configuration files, e.g., in
3213 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
3217 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
3218 called @code{check_stack} in your @file{md} file, GCC will call that
3219 pattern with one argument which is the address to compare the stack
3220 value against. You must arrange for this pattern to report an error if
3221 the stack pointer is out of range.
3224 If neither of the above are true, GCC will generate code to periodically
3225 ``probe'' the stack pointer using the values of the macros defined below.
3228 Normally, you will use the default values of these macros, so GCC
3229 will use the third approach.
3231 @defmac STACK_CHECK_BUILTIN
3232 A nonzero value if stack checking is done by the configuration files in a
3233 machine-dependent manner. You should define this macro if stack checking
3234 is require by the ABI of your machine or if you would like to have to stack
3235 checking in some more efficient way than GCC's portable approach.
3236 The default value of this macro is zero.
3239 @defmac STACK_CHECK_PROBE_INTERVAL
3240 An integer representing the interval at which GCC must generate stack
3241 probe instructions. You will normally define this macro to be no larger
3242 than the size of the ``guard pages'' at the end of a stack area. The
3243 default value of 4096 is suitable for most systems.
3246 @defmac STACK_CHECK_PROBE_LOAD
3247 A integer which is nonzero if GCC should perform the stack probe
3248 as a load instruction and zero if GCC should use a store instruction.
3249 The default is zero, which is the most efficient choice on most systems.
3252 @defmac STACK_CHECK_PROTECT
3253 The number of bytes of stack needed to recover from a stack overflow,
3254 for languages where such a recovery is supported. The default value of
3255 75 words should be adequate for most machines.
3258 @defmac STACK_CHECK_MAX_FRAME_SIZE
3259 The maximum size of a stack frame, in bytes. GCC will generate probe
3260 instructions in non-leaf functions to ensure at least this many bytes of
3261 stack are available. If a stack frame is larger than this size, stack
3262 checking will not be reliable and GCC will issue a warning. The
3263 default is chosen so that GCC only generates one instruction on most
3264 systems. You should normally not change the default value of this macro.
3267 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3268 GCC uses this value to generate the above warning message. It
3269 represents the amount of fixed frame used by a function, not including
3270 space for any callee-saved registers, temporaries and user variables.
3271 You need only specify an upper bound for this amount and will normally
3272 use the default of four words.
3275 @defmac STACK_CHECK_MAX_VAR_SIZE
3276 The maximum size, in bytes, of an object that GCC will place in the
3277 fixed area of the stack frame when the user specifies
3278 @option{-fstack-check}.
3279 GCC computed the default from the values of the above macros and you will
3280 normally not need to override that default.
3284 @node Frame Registers
3285 @subsection Registers That Address the Stack Frame
3287 @c prevent bad page break with this line
3288 This discusses registers that address the stack frame.
3290 @defmac STACK_POINTER_REGNUM
3291 The register number of the stack pointer register, which must also be a
3292 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3293 the hardware determines which register this is.
3296 @defmac FRAME_POINTER_REGNUM
3297 The register number of the frame pointer register, which is used to
3298 access automatic variables in the stack frame. On some machines, the
3299 hardware determines which register this is. On other machines, you can
3300 choose any register you wish for this purpose.
3303 @defmac HARD_FRAME_POINTER_REGNUM
3304 On some machines the offset between the frame pointer and starting
3305 offset of the automatic variables is not known until after register
3306 allocation has been done (for example, because the saved registers are
3307 between these two locations). On those machines, define
3308 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3309 be used internally until the offset is known, and define
3310 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3311 used for the frame pointer.
3313 You should define this macro only in the very rare circumstances when it
3314 is not possible to calculate the offset between the frame pointer and
3315 the automatic variables until after register allocation has been
3316 completed. When this macro is defined, you must also indicate in your
3317 definition of @code{ELIMINABLE_REGS} how to eliminate
3318 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3319 or @code{STACK_POINTER_REGNUM}.
3321 Do not define this macro if it would be the same as
3322 @code{FRAME_POINTER_REGNUM}.
3325 @defmac ARG_POINTER_REGNUM
3326 The register number of the arg pointer register, which is used to access
3327 the function's argument list. On some machines, this is the same as the
3328 frame pointer register. On some machines, the hardware determines which
3329 register this is. On other machines, you can choose any register you
3330 wish for this purpose. If this is not the same register as the frame
3331 pointer register, then you must mark it as a fixed register according to
3332 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3333 (@pxref{Elimination}).
3336 @defmac RETURN_ADDRESS_POINTER_REGNUM
3337 The register number of the return address pointer register, which is used to
3338 access the current function's return address from the stack. On some
3339 machines, the return address is not at a fixed offset from the frame
3340 pointer or stack pointer or argument pointer. This register can be defined
3341 to point to the return address on the stack, and then be converted by
3342 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3344 Do not define this macro unless there is no other way to get the return
3345 address from the stack.
3348 @defmac STATIC_CHAIN_REGNUM
3349 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3350 Register numbers used for passing a function's static chain pointer. If
3351 register windows are used, the register number as seen by the called
3352 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3353 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3354 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3357 The static chain register need not be a fixed register.
3359 If the static chain is passed in memory, these macros should not be
3360 defined; instead, the next two macros should be defined.
3363 @defmac STATIC_CHAIN
3364 @defmacx STATIC_CHAIN_INCOMING
3365 If the static chain is passed in memory, these macros provide rtx giving
3366 @code{mem} expressions that denote where they are stored.
3367 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3368 as seen by the calling and called functions, respectively. Often the former
3369 will be at an offset from the stack pointer and the latter at an offset from
3372 @findex stack_pointer_rtx
3373 @findex frame_pointer_rtx
3374 @findex arg_pointer_rtx
3375 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3376 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3377 macros and should be used to refer to those items.
3379 If the static chain is passed in a register, the two previous macros should
3383 @defmac DWARF_FRAME_REGISTERS
3384 This macro specifies the maximum number of hard registers that can be
3385 saved in a call frame. This is used to size data structures used in
3386 DWARF2 exception handling.
3388 Prior to GCC 3.0, this macro was needed in order to establish a stable
3389 exception handling ABI in the face of adding new hard registers for ISA
3390 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3391 in the number of hard registers. Nevertheless, this macro can still be
3392 used to reduce the runtime memory requirements of the exception handling
3393 routines, which can be substantial if the ISA contains a lot of
3394 registers that are not call-saved.
3396 If this macro is not defined, it defaults to
3397 @code{FIRST_PSEUDO_REGISTER}.
3400 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3402 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3403 for backward compatibility in pre GCC 3.0 compiled code.
3405 If this macro is not defined, it defaults to
3406 @code{DWARF_FRAME_REGISTERS}.
3409 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3411 Define this macro if the target's representation for dwarf registers
3412 is different than the internal representation for unwind column.
3413 Given a dwarf register, this macro should return the internal unwind
3414 column number to use instead.
3416 See the PowerPC's SPE target for an example.
3419 @defmac DWARF_FRAME_REGNUM (@var{regno})
3421 Define this macro if the target's representation for dwarf registers
3422 used in .eh_frame or .debug_frame is different from that used in other
3423 debug info sections. Given a GCC hard register number, this macro
3424 should return the .eh_frame register number. The default is
3425 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3429 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3431 Define this macro to map register numbers held in the call frame info
3432 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3433 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3434 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3435 return @code{@var{regno}}.
3440 @subsection Eliminating Frame Pointer and Arg Pointer
3442 @c prevent bad page break with this line
3443 This is about eliminating the frame pointer and arg pointer.
3445 @defmac FRAME_POINTER_REQUIRED
3446 A C expression which is nonzero if a function must have and use a frame
3447 pointer. This expression is evaluated in the reload pass. If its value is
3448 nonzero the function will have a frame pointer.
3450 The expression can in principle examine the current function and decide
3451 according to the facts, but on most machines the constant 0 or the
3452 constant 1 suffices. Use 0 when the machine allows code to be generated
3453 with no frame pointer, and doing so saves some time or space. Use 1
3454 when there is no possible advantage to avoiding a frame pointer.
3456 In certain cases, the compiler does not know how to produce valid code
3457 without a frame pointer. The compiler recognizes those cases and
3458 automatically gives the function a frame pointer regardless of what
3459 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3462 In a function that does not require a frame pointer, the frame pointer
3463 register can be allocated for ordinary usage, unless you mark it as a
3464 fixed register. See @code{FIXED_REGISTERS} for more information.
3467 @findex get_frame_size
3468 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3469 A C statement to store in the variable @var{depth-var} the difference
3470 between the frame pointer and the stack pointer values immediately after
3471 the function prologue. The value would be computed from information
3472 such as the result of @code{get_frame_size ()} and the tables of
3473 registers @code{regs_ever_live} and @code{call_used_regs}.
3475 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3476 need not be defined. Otherwise, it must be defined even if
3477 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3478 case, you may set @var{depth-var} to anything.
3481 @defmac ELIMINABLE_REGS
3482 If defined, this macro specifies a table of register pairs used to
3483 eliminate unneeded registers that point into the stack frame. If it is not
3484 defined, the only elimination attempted by the compiler is to replace
3485 references to the frame pointer with references to the stack pointer.
3487 The definition of this macro is a list of structure initializations, each
3488 of which specifies an original and replacement register.
3490 On some machines, the position of the argument pointer is not known until
3491 the compilation is completed. In such a case, a separate hard register
3492 must be used for the argument pointer. This register can be eliminated by
3493 replacing it with either the frame pointer or the argument pointer,
3494 depending on whether or not the frame pointer has been eliminated.
3496 In this case, you might specify:
3498 #define ELIMINABLE_REGS \
3499 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3500 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3501 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3504 Note that the elimination of the argument pointer with the stack pointer is
3505 specified first since that is the preferred elimination.
3508 @defmac CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3509 A C expression that returns nonzero if the compiler is allowed to try
3510 to replace register number @var{from-reg} with register number
3511 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3512 is defined, and will usually be the constant 1, since most of the cases
3513 preventing register elimination are things that the compiler already
3517 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3518 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3519 specifies the initial difference between the specified pair of
3520 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3524 @node Stack Arguments
3525 @subsection Passing Function Arguments on the Stack
3526 @cindex arguments on stack
3527 @cindex stack arguments
3529 The macros in this section control how arguments are passed
3530 on the stack. See the following section for other macros that
3531 control passing certain arguments in registers.
3533 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (tree @var{fntype})
3534 This target hook returns @code{true} if an argument declared in a
3535 prototype as an integral type smaller than @code{int} should actually be
3536 passed as an @code{int}. In addition to avoiding errors in certain
3537 cases of mismatch, it also makes for better code on certain machines.
3538 The default is to not promote prototypes.
3542 A C expression. If nonzero, push insns will be used to pass
3544 If the target machine does not have a push instruction, set it to zero.
3545 That directs GCC to use an alternate strategy: to
3546 allocate the entire argument block and then store the arguments into
3547 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3550 @defmac PUSH_ARGS_REVERSED
3551 A C expression. If nonzero, function arguments will be evaluated from
3552 last to first, rather than from first to last. If this macro is not
3553 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3554 and args grow in opposite directions, and 0 otherwise.
3557 @defmac PUSH_ROUNDING (@var{npushed})
3558 A C expression that is the number of bytes actually pushed onto the
3559 stack when an instruction attempts to push @var{npushed} bytes.
3561 On some machines, the definition
3564 #define PUSH_ROUNDING(BYTES) (BYTES)
3568 will suffice. But on other machines, instructions that appear
3569 to push one byte actually push two bytes in an attempt to maintain
3570 alignment. Then the definition should be
3573 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3577 @findex current_function_outgoing_args_size
3578 @defmac ACCUMULATE_OUTGOING_ARGS
3579 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3580 will be computed and placed into the variable
3581 @code{current_function_outgoing_args_size}. No space will be pushed
3582 onto the stack for each call; instead, the function prologue should
3583 increase the stack frame size by this amount.
3585 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3589 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3590 Define this macro if functions should assume that stack space has been
3591 allocated for arguments even when their values are passed in
3594 The value of this macro is the size, in bytes, of the area reserved for
3595 arguments passed in registers for the function represented by @var{fndecl},
3596 which can be zero if GCC is calling a library function.
3598 This space can be allocated by the caller, or be a part of the
3599 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3602 @c above is overfull. not sure what to do. --mew 5feb93 did
3603 @c something, not sure if it looks good. --mew 10feb93
3605 @defmac OUTGOING_REG_PARM_STACK_SPACE
3606 Define this if it is the responsibility of the caller to allocate the area
3607 reserved for arguments passed in registers.
3609 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3610 whether the space for these arguments counts in the value of
3611 @code{current_function_outgoing_args_size}.
3614 @defmac STACK_PARMS_IN_REG_PARM_AREA
3615 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3616 stack parameters don't skip the area specified by it.
3617 @c i changed this, makes more sens and it should have taken care of the
3618 @c overfull.. not as specific, tho. --mew 5feb93
3620 Normally, when a parameter is not passed in registers, it is placed on the
3621 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3622 suppresses this behavior and causes the parameter to be passed on the
3623 stack in its natural location.
3626 @defmac RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3627 A C expression that should indicate the number of bytes of its own
3628 arguments that a function pops on returning, or 0 if the
3629 function pops no arguments and the caller must therefore pop them all
3630 after the function returns.
3632 @var{fundecl} is a C variable whose value is a tree node that describes
3633 the function in question. Normally it is a node of type
3634 @code{FUNCTION_DECL} that describes the declaration of the function.
3635 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3637 @var{funtype} is a C variable whose value is a tree node that
3638 describes the function in question. Normally it is a node of type
3639 @code{FUNCTION_TYPE} that describes the data type of the function.
3640 From this it is possible to obtain the data types of the value and
3641 arguments (if known).
3643 When a call to a library function is being considered, @var{fundecl}
3644 will contain an identifier node for the library function. Thus, if
3645 you need to distinguish among various library functions, you can do so
3646 by their names. Note that ``library function'' in this context means
3647 a function used to perform arithmetic, whose name is known specially
3648 in the compiler and was not mentioned in the C code being compiled.
3650 @var{stack-size} is the number of bytes of arguments passed on the
3651 stack. If a variable number of bytes is passed, it is zero, and
3652 argument popping will always be the responsibility of the calling function.
3654 On the VAX, all functions always pop their arguments, so the definition
3655 of this macro is @var{stack-size}. On the 68000, using the standard
3656 calling convention, no functions pop their arguments, so the value of
3657 the macro is always 0 in this case. But an alternative calling
3658 convention is available in which functions that take a fixed number of
3659 arguments pop them but other functions (such as @code{printf}) pop
3660 nothing (the caller pops all). When this convention is in use,
3661 @var{funtype} is examined to determine whether a function takes a fixed
3662 number of arguments.
3665 @defmac CALL_POPS_ARGS (@var{cum})
3666 A C expression that should indicate the number of bytes a call sequence
3667 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3668 when compiling a function call.
3670 @var{cum} is the variable in which all arguments to the called function
3671 have been accumulated.
3673 On certain architectures, such as the SH5, a call trampoline is used
3674 that pops certain registers off the stack, depending on the arguments
3675 that have been passed to the function. Since this is a property of the
3676 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3680 @node Register Arguments
3681 @subsection Passing Arguments in Registers
3682 @cindex arguments in registers
3683 @cindex registers arguments
3685 This section describes the macros which let you control how various
3686 types of arguments are passed in registers or how they are arranged in
3689 @defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3690 A C expression that controls whether a function argument is passed
3691 in a register, and which register.
3693 The arguments are @var{cum}, which summarizes all the previous
3694 arguments; @var{mode}, the machine mode of the argument; @var{type},
3695 the data type of the argument as a tree node or 0 if that is not known
3696 (which happens for C support library functions); and @var{named},
3697 which is 1 for an ordinary argument and 0 for nameless arguments that
3698 correspond to @samp{@dots{}} in the called function's prototype.
3699 @var{type} can be an incomplete type if a syntax error has previously
3702 The value of the expression is usually either a @code{reg} RTX for the
3703 hard register in which to pass the argument, or zero to pass the
3704 argument on the stack.
3706 For machines like the VAX and 68000, where normally all arguments are
3707 pushed, zero suffices as a definition.
3709 The value of the expression can also be a @code{parallel} RTX@. This is
3710 used when an argument is passed in multiple locations. The mode of the
3711 @code{parallel} should be the mode of the entire argument. The
3712 @code{parallel} holds any number of @code{expr_list} pairs; each one
3713 describes where part of the argument is passed. In each
3714 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3715 register in which to pass this part of the argument, and the mode of the
3716 register RTX indicates how large this part of the argument is. The
3717 second operand of the @code{expr_list} is a @code{const_int} which gives
3718 the offset in bytes into the entire argument of where this part starts.
3719 As a special exception the first @code{expr_list} in the @code{parallel}
3720 RTX may have a first operand of zero. This indicates that the entire
3721 argument is also stored on the stack.
3723 The last time this macro is called, it is called with @code{MODE ==
3724 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3725 pattern as operands 2 and 3 respectively.
3727 @cindex @file{stdarg.h} and register arguments
3728 The usual way to make the ISO library @file{stdarg.h} work on a machine
3729 where some arguments are usually passed in registers, is to cause
3730 nameless arguments to be passed on the stack instead. This is done
3731 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3733 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3734 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3735 You may use the hook @code{targetm.calls.must_pass_in_stack}
3736 in the definition of this macro to determine if this argument is of a
3737 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3738 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3739 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3740 defined, the argument will be computed in the stack and then loaded into
3744 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, tree @var{type})
3745 This target hook should return @code{true} if we should not pass @var{type}
3746 solely in registers. The file @file{expr.h} defines a
3747 definition that is usually appropriate, refer to @file{expr.h} for additional
3751 @defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3752 Define this macro if the target machine has ``register windows'', so
3753 that the register in which a function sees an arguments is not
3754 necessarily the same as the one in which the caller passed the
3757 For such machines, @code{FUNCTION_ARG} computes the register in which
3758 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3759 be defined in a similar fashion to tell the function being called
3760 where the arguments will arrive.
3762 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3763 serves both purposes.
3766 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3767 This target hook returns the number of bytes at the beginning of an
3768 argument that must be put in registers. The value must be zero for
3769 arguments that are passed entirely in registers or that are entirely
3770 pushed on the stack.
3772 On some machines, certain arguments must be passed partially in
3773 registers and partially in memory. On these machines, typically the
3774 first few words of arguments are passed in registers, and the rest
3775 on the stack. If a multi-word argument (a @code{double} or a
3776 structure) crosses that boundary, its first few words must be passed
3777 in registers and the rest must be pushed. This macro tells the
3778 compiler when this occurs, and how many bytes should go in registers.
3780 @code{FUNCTION_ARG} for these arguments should return the first
3781 register to be used by the caller for this argument; likewise
3782 @code{FUNCTION_INCOMING_ARG}, for the called function.
3785 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3786 This target hook should return @code{true} if an argument at the
3787 position indicated by @var{cum} should be passed by reference. This
3788 predicate is queried after target independent reasons for being
3789 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
3791 If the hook returns true, a copy of that argument is made in memory and a
3792 pointer to the argument is passed instead of the argument itself.
3793 The pointer is passed in whatever way is appropriate for passing a pointer
3797 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3798 The function argument described by the parameters to this hook is
3799 known to be passed by reference. The hook should return true if the
3800 function argument should be copied by the callee instead of copied
3803 For any argument for which the hook returns true, if it can be
3804 determined that the argument is not modified, then a copy need
3807 The default version of this hook always returns false.
3810 @defmac CUMULATIVE_ARGS
3811 A C type for declaring a variable that is used as the first argument of
3812 @code{FUNCTION_ARG} and other related values. For some target machines,
3813 the type @code{int} suffices and can hold the number of bytes of
3816 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3817 arguments that have been passed on the stack. The compiler has other
3818 variables to keep track of that. For target machines on which all
3819 arguments are passed on the stack, there is no need to store anything in
3820 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3821 should not be empty, so use @code{int}.
3824 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
3825 A C statement (sans semicolon) for initializing the variable
3826 @var{cum} for the state at the beginning of the argument list. The
3827 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
3828 is the tree node for the data type of the function which will receive
3829 the args, or 0 if the args are to a compiler support library function.
3830 For direct calls that are not libcalls, @var{fndecl} contain the
3831 declaration node of the function. @var{fndecl} is also set when
3832 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3833 being compiled. @var{n_named_args} is set to the number of named
3834 arguments, including a structure return address if it is passed as a
3835 parameter, when making a call. When processing incoming arguments,
3836 @var{n_named_args} is set to @minus{}1.
3838 When processing a call to a compiler support library function,
3839 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3840 contains the name of the function, as a string. @var{libname} is 0 when
3841 an ordinary C function call is being processed. Thus, each time this
3842 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3843 never both of them at once.
3846 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3847 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3848 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3849 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3850 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3851 0)} is used instead.
3854 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3855 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3856 finding the arguments for the function being compiled. If this macro is
3857 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3859 The value passed for @var{libname} is always 0, since library routines
3860 with special calling conventions are never compiled with GCC@. The
3861 argument @var{libname} exists for symmetry with
3862 @code{INIT_CUMULATIVE_ARGS}.
3863 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3864 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3867 @defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3868 A C statement (sans semicolon) to update the summarizer variable
3869 @var{cum} to advance past an argument in the argument list. The
3870 values @var{mode}, @var{type} and @var{named} describe that argument.
3871 Once this is done, the variable @var{cum} is suitable for analyzing
3872 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
3874 This macro need not do anything if the argument in question was passed
3875 on the stack. The compiler knows how to track the amount of stack space
3876 used for arguments without any special help.
3879 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3880 If defined, a C expression which determines whether, and in which direction,
3881 to pad out an argument with extra space. The value should be of type
3882 @code{enum direction}: either @code{upward} to pad above the argument,
3883 @code{downward} to pad below, or @code{none} to inhibit padding.
3885 The @emph{amount} of padding is always just enough to reach the next
3886 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3889 This macro has a default definition which is right for most systems.
3890 For little-endian machines, the default is to pad upward. For
3891 big-endian machines, the default is to pad downward for an argument of
3892 constant size shorter than an @code{int}, and upward otherwise.
3895 @defmac PAD_VARARGS_DOWN
3896 If defined, a C expression which determines whether the default
3897 implementation of va_arg will attempt to pad down before reading the
3898 next argument, if that argument is smaller than its aligned space as
3899 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3900 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3903 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
3904 Specify padding for the last element of a block move between registers and
3905 memory. @var{first} is nonzero if this is the only element. Defining this
3906 macro allows better control of register function parameters on big-endian
3907 machines, without using @code{PARALLEL} rtl. In particular,
3908 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
3909 registers, as there is no longer a "wrong" part of a register; For example,
3910 a three byte aggregate may be passed in the high part of a register if so
3914 @defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3915 If defined, a C expression that gives the alignment boundary, in bits,
3916 of an argument with the specified mode and type. If it is not defined,
3917 @code{PARM_BOUNDARY} is used for all arguments.
3920 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
3921 A C expression that is nonzero if @var{regno} is the number of a hard
3922 register in which function arguments are sometimes passed. This does
3923 @emph{not} include implicit arguments such as the static chain and
3924 the structure-value address. On many machines, no registers can be
3925 used for this purpose since all function arguments are pushed on the
3929 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (tree @var{type})
3930 This hook should return true if parameter of type @var{type} are passed
3931 as two scalar parameters. By default, GCC will attempt to pack complex
3932 arguments into the target's word size. Some ABIs require complex arguments
3933 to be split and treated as their individual components. For example, on
3934 AIX64, complex floats should be passed in a pair of floating point
3935 registers, even though a complex float would fit in one 64-bit floating
3938 The default value of this hook is @code{NULL}, which is treated as always
3942 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
3943 This hook returns a type node for @code{va_list} for the target.
3944 The default version of the hook returns @code{void*}.
3947 @deftypefn {Target Hook} tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree @var{valist}, tree @var{type}, tree *@var{pre_p}, tree *@var{post_p})
3948 This hook performs target-specific gimplification of
3949 @code{VA_ARG_EXPR}. The first two parameters correspond to the
3950 arguments to @code{va_arg}; the latter two are as in
3951 @code{gimplify.c:gimplify_expr}.
3954 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (enum machine_mode @var{mode})
3955 Define this to return nonzero if the port can handle pointers
3956 with machine mode @var{mode}. The default version of this
3957 hook returns true for both @code{ptr_mode} and @code{Pmode}.
3960 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
3961 Define this to return nonzero if the port is prepared to handle
3962 insns involving scalar mode @var{mode}. For a scalar mode to be
3963 considered supported, all the basic arithmetic and comparisons
3966 The default version of this hook returns true for any mode
3967 required to handle the basic C types (as defined by the port).
3968 Included here are the double-word arithmetic supported by the
3969 code in @file{optabs.c}.
3972 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
3973 Define this to return nonzero if the port is prepared to handle
3974 insns involving vector mode @var{mode}. At the very least, it
3975 must have move patterns for this mode.
3979 @subsection How Scalar Function Values Are Returned
3980 @cindex return values in registers
3981 @cindex values, returned by functions
3982 @cindex scalars, returned as values
3984 This section discusses the macros that control returning scalars as
3985 values---values that can fit in registers.
3987 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
3988 A C expression to create an RTX representing the place where a
3989 function returns a value of data type @var{valtype}. @var{valtype} is
3990 a tree node representing a data type. Write @code{TYPE_MODE
3991 (@var{valtype})} to get the machine mode used to represent that type.
3992 On many machines, only the mode is relevant. (Actually, on most
3993 machines, scalar values are returned in the same place regardless of
3996 The value of the expression is usually a @code{reg} RTX for the hard
3997 register where the return value is stored. The value can also be a
3998 @code{parallel} RTX, if the return value is in multiple places. See
3999 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
4001 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply the same
4002 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
4005 If the precise function being called is known, @var{func} is a tree
4006 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4007 pointer. This makes it possible to use a different value-returning
4008 convention for specific functions when all their calls are
4011 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
4012 types, because these are returned in another way. See
4013 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4016 @defmac FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
4017 Define this macro if the target machine has ``register windows''
4018 so that the register in which a function returns its value is not
4019 the same as the one in which the caller sees the value.
4021 For such machines, @code{FUNCTION_VALUE} computes the register in which
4022 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
4023 defined in a similar fashion to tell the function where to put the
4026 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
4027 @code{FUNCTION_VALUE} serves both purposes.
4029 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
4030 aggregate data types, because these are returned in another way. See
4031 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4034 @defmac LIBCALL_VALUE (@var{mode})
4035 A C expression to create an RTX representing the place where a library
4036 function returns a value of mode @var{mode}. If the precise function
4037 being called is known, @var{func} is a tree node
4038 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4039 pointer. This makes it possible to use a different value-returning
4040 convention for specific functions when all their calls are
4043 Note that ``library function'' in this context means a compiler
4044 support routine, used to perform arithmetic, whose name is known
4045 specially by the compiler and was not mentioned in the C code being
4048 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
4049 data types, because none of the library functions returns such types.
4052 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4053 A C expression that is nonzero if @var{regno} is the number of a hard
4054 register in which the values of called function may come back.
4056 A register whose use for returning values is limited to serving as the
4057 second of a pair (for a value of type @code{double}, say) need not be
4058 recognized by this macro. So for most machines, this definition
4062 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4065 If the machine has register windows, so that the caller and the called
4066 function use different registers for the return value, this macro
4067 should recognize only the caller's register numbers.
4070 @defmac APPLY_RESULT_SIZE
4071 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4072 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4073 saving and restoring an arbitrary return value.
4076 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (tree @var{type})
4077 This hook should return true if values of type @var{type} are returned
4078 at the most significant end of a register (in other words, if they are
4079 padded at the least significant end). You can assume that @var{type}
4080 is returned in a register; the caller is required to check this.
4082 Note that the register provided by @code{FUNCTION_VALUE} must be able
4083 to hold the complete return value. For example, if a 1-, 2- or 3-byte
4084 structure is returned at the most significant end of a 4-byte register,
4085 @code{FUNCTION_VALUE} should provide an @code{SImode} rtx.
4088 @node Aggregate Return
4089 @subsection How Large Values Are Returned
4090 @cindex aggregates as return values
4091 @cindex large return values
4092 @cindex returning aggregate values
4093 @cindex structure value address
4095 When a function value's mode is @code{BLKmode} (and in some other
4096 cases), the value is not returned according to @code{FUNCTION_VALUE}
4097 (@pxref{Scalar Return}). Instead, the caller passes the address of a
4098 block of memory in which the value should be stored. This address
4099 is called the @dfn{structure value address}.
4101 This section describes how to control returning structure values in
4104 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (tree @var{type}, tree @var{fntype})
4105 This target hook should return a nonzero value to say to return the
4106 function value in memory, just as large structures are always returned.
4107 Here @var{type} will be the data type of the value, and @var{fntype}
4108 will be the type of the function doing the returning, or @code{NULL} for
4111 Note that values of mode @code{BLKmode} must be explicitly handled
4112 by this function. Also, the option @option{-fpcc-struct-return}
4113 takes effect regardless of this macro. On most systems, it is
4114 possible to leave the hook undefined; this causes a default
4115 definition to be used, whose value is the constant 1 for @code{BLKmode}
4116 values, and 0 otherwise.
4118 Do not use this hook to indicate that structures and unions should always
4119 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4123 @defmac DEFAULT_PCC_STRUCT_RETURN
4124 Define this macro to be 1 if all structure and union return values must be
4125 in memory. Since this results in slower code, this should be defined
4126 only if needed for compatibility with other compilers or with an ABI@.
4127 If you define this macro to be 0, then the conventions used for structure
4128 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4131 If not defined, this defaults to the value 1.
4134 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4135 This target hook should return the location of the structure value
4136 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4137 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4138 be @code{NULL}, for libcalls. You do not need to define this target
4139 hook if the address is always passed as an ``invisible'' first
4142 On some architectures the place where the structure value address
4143 is found by the called function is not the same place that the
4144 caller put it. This can be due to register windows, or it could
4145 be because the function prologue moves it to a different place.
4146 @var{incoming} is @code{true} when the location is needed in
4147 the context of the called function, and @code{false} in the context of
4150 If @var{incoming} is @code{true} and the address is to be found on the
4151 stack, return a @code{mem} which refers to the frame pointer.
4154 @defmac PCC_STATIC_STRUCT_RETURN
4155 Define this macro if the usual system convention on the target machine
4156 for returning structures and unions is for the called function to return
4157 the address of a static variable containing the value.
4159 Do not define this if the usual system convention is for the caller to
4160 pass an address to the subroutine.
4162 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4163 nothing when you use @option{-freg-struct-return} mode.
4167 @subsection Caller-Saves Register Allocation
4169 If you enable it, GCC can save registers around function calls. This
4170 makes it possible to use call-clobbered registers to hold variables that
4171 must live across calls.
4173 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4174 A C expression to determine whether it is worthwhile to consider placing
4175 a pseudo-register in a call-clobbered hard register and saving and
4176 restoring it around each function call. The expression should be 1 when
4177 this is worth doing, and 0 otherwise.
4179 If you don't define this macro, a default is used which is good on most
4180 machines: @code{4 * @var{calls} < @var{refs}}.
4183 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4184 A C expression specifying which mode is required for saving @var{nregs}
4185 of a pseudo-register in call-clobbered hard register @var{regno}. If
4186 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4187 returned. For most machines this macro need not be defined since GCC
4188 will select the smallest suitable mode.
4191 @node Function Entry
4192 @subsection Function Entry and Exit
4193 @cindex function entry and exit
4197 This section describes the macros that output function entry
4198 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4200 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4201 If defined, a function that outputs the assembler code for entry to a
4202 function. The prologue is responsible for setting up the stack frame,
4203 initializing the frame pointer register, saving registers that must be
4204 saved, and allocating @var{size} additional bytes of storage for the
4205 local variables. @var{size} is an integer. @var{file} is a stdio
4206 stream to which the assembler code should be output.
4208 The label for the beginning of the function need not be output by this
4209 macro. That has already been done when the macro is run.
4211 @findex regs_ever_live
4212 To determine which registers to save, the macro can refer to the array
4213 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4214 @var{r} is used anywhere within the function. This implies the function
4215 prologue should save register @var{r}, provided it is not one of the
4216 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4217 @code{regs_ever_live}.)
4219 On machines that have ``register windows'', the function entry code does
4220 not save on the stack the registers that are in the windows, even if
4221 they are supposed to be preserved by function calls; instead it takes
4222 appropriate steps to ``push'' the register stack, if any non-call-used
4223 registers are used in the function.
4225 @findex frame_pointer_needed
4226 On machines where functions may or may not have frame-pointers, the
4227 function entry code must vary accordingly; it must set up the frame
4228 pointer if one is wanted, and not otherwise. To determine whether a
4229 frame pointer is in wanted, the macro can refer to the variable
4230 @code{frame_pointer_needed}. The variable's value will be 1 at run
4231 time in a function that needs a frame pointer. @xref{Elimination}.
4233 The function entry code is responsible for allocating any stack space
4234 required for the function. This stack space consists of the regions
4235 listed below. In most cases, these regions are allocated in the
4236 order listed, with the last listed region closest to the top of the
4237 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4238 the highest address if it is not defined). You can use a different order
4239 for a machine if doing so is more convenient or required for
4240 compatibility reasons. Except in cases where required by standard
4241 or by a debugger, there is no reason why the stack layout used by GCC
4242 need agree with that used by other compilers for a machine.
4245 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4246 If defined, a function that outputs assembler code at the end of a
4247 prologue. This should be used when the function prologue is being
4248 emitted as RTL, and you have some extra assembler that needs to be
4249 emitted. @xref{prologue instruction pattern}.
4252 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4253 If defined, a function that outputs assembler code at the start of an
4254 epilogue. This should be used when the function epilogue is being
4255 emitted as RTL, and you have some extra assembler that needs to be
4256 emitted. @xref{epilogue instruction pattern}.
4259 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4260 If defined, a function that outputs the assembler code for exit from a
4261 function. The epilogue is responsible for restoring the saved
4262 registers and stack pointer to their values when the function was
4263 called, and returning control to the caller. This macro takes the
4264 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4265 registers to restore are determined from @code{regs_ever_live} and
4266 @code{CALL_USED_REGISTERS} in the same way.
4268 On some machines, there is a single instruction that does all the work
4269 of returning from the function. On these machines, give that
4270 instruction the name @samp{return} and do not define the macro
4271 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4273 Do not define a pattern named @samp{return} if you want the
4274 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4275 switches to control whether return instructions or epilogues are used,
4276 define a @samp{return} pattern with a validity condition that tests the
4277 target switches appropriately. If the @samp{return} pattern's validity
4278 condition is false, epilogues will be used.
4280 On machines where functions may or may not have frame-pointers, the
4281 function exit code must vary accordingly. Sometimes the code for these
4282 two cases is completely different. To determine whether a frame pointer
4283 is wanted, the macro can refer to the variable
4284 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4285 a function that needs a frame pointer.
4287 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4288 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4289 The C variable @code{current_function_is_leaf} is nonzero for such a
4290 function. @xref{Leaf Functions}.
4292 On some machines, some functions pop their arguments on exit while
4293 others leave that for the caller to do. For example, the 68020 when
4294 given @option{-mrtd} pops arguments in functions that take a fixed
4295 number of arguments.
4297 @findex current_function_pops_args
4298 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4299 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4300 needs to know what was decided. The variable that is called
4301 @code{current_function_pops_args} is the number of bytes of its
4302 arguments that a function should pop. @xref{Scalar Return}.
4303 @c what is the "its arguments" in the above sentence referring to, pray
4304 @c tell? --mew 5feb93
4309 @findex current_function_pretend_args_size
4310 A region of @code{current_function_pretend_args_size} bytes of
4311 uninitialized space just underneath the first argument arriving on the
4312 stack. (This may not be at the very start of the allocated stack region
4313 if the calling sequence has pushed anything else since pushing the stack
4314 arguments. But usually, on such machines, nothing else has been pushed
4315 yet, because the function prologue itself does all the pushing.) This
4316 region is used on machines where an argument may be passed partly in
4317 registers and partly in memory, and, in some cases to support the
4318 features in @code{<stdarg.h>}.
4321 An area of memory used to save certain registers used by the function.
4322 The size of this area, which may also include space for such things as
4323 the return address and pointers to previous stack frames, is
4324 machine-specific and usually depends on which registers have been used
4325 in the function. Machines with register windows often do not require
4329 A region of at least @var{size} bytes, possibly rounded up to an allocation
4330 boundary, to contain the local variables of the function. On some machines,
4331 this region and the save area may occur in the opposite order, with the
4332 save area closer to the top of the stack.
4335 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4336 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4337 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4338 argument lists of the function. @xref{Stack Arguments}.
4341 @defmac EXIT_IGNORE_STACK
4342 Define this macro as a C expression that is nonzero if the return
4343 instruction or the function epilogue ignores the value of the stack
4344 pointer; in other words, if it is safe to delete an instruction to
4345 adjust the stack pointer before a return from the function. The
4348 Note that this macro's value is relevant only for functions for which
4349 frame pointers are maintained. It is never safe to delete a final
4350 stack adjustment in a function that has no frame pointer, and the
4351 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4354 @defmac EPILOGUE_USES (@var{regno})
4355 Define this macro as a C expression that is nonzero for registers that are
4356 used by the epilogue or the @samp{return} pattern. The stack and frame
4357 pointer registers are already be assumed to be used as needed.
4360 @defmac EH_USES (@var{regno})
4361 Define this macro as a C expression that is nonzero for registers that are
4362 used by the exception handling mechanism, and so should be considered live
4363 on entry to an exception edge.
4366 @defmac DELAY_SLOTS_FOR_EPILOGUE
4367 Define this macro if the function epilogue contains delay slots to which
4368 instructions from the rest of the function can be ``moved''. The
4369 definition should be a C expression whose value is an integer
4370 representing the number of delay slots there.
4373 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4374 A C expression that returns 1 if @var{insn} can be placed in delay
4375 slot number @var{n} of the epilogue.
4377 The argument @var{n} is an integer which identifies the delay slot now
4378 being considered (since different slots may have different rules of
4379 eligibility). It is never negative and is always less than the number
4380 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4381 If you reject a particular insn for a given delay slot, in principle, it
4382 may be reconsidered for a subsequent delay slot. Also, other insns may
4383 (at least in principle) be considered for the so far unfilled delay
4386 @findex current_function_epilogue_delay_list
4387 @findex final_scan_insn
4388 The insns accepted to fill the epilogue delay slots are put in an RTL
4389 list made with @code{insn_list} objects, stored in the variable
4390 @code{current_function_epilogue_delay_list}. The insn for the first
4391 delay slot comes first in the list. Your definition of the macro
4392 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4393 outputting the insns in this list, usually by calling
4394 @code{final_scan_insn}.
4396 You need not define this macro if you did not define
4397 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4400 @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})
4401 A function that outputs the assembler code for a thunk
4402 function, used to implement C++ virtual function calls with multiple
4403 inheritance. The thunk acts as a wrapper around a virtual function,
4404 adjusting the implicit object parameter before handing control off to
4407 First, emit code to add the integer @var{delta} to the location that
4408 contains the incoming first argument. Assume that this argument
4409 contains a pointer, and is the one used to pass the @code{this} pointer
4410 in C++. This is the incoming argument @emph{before} the function prologue,
4411 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4412 all other incoming arguments.
4414 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4415 made after adding @code{delta}. In particular, if @var{p} is the
4416 adjusted pointer, the following adjustment should be made:
4419 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4422 After the additions, emit code to jump to @var{function}, which is a
4423 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4424 not touch the return address. Hence returning from @var{FUNCTION} will
4425 return to whoever called the current @samp{thunk}.
4427 The effect must be as if @var{function} had been called directly with
4428 the adjusted first argument. This macro is responsible for emitting all
4429 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4430 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4432 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4433 have already been extracted from it.) It might possibly be useful on
4434 some targets, but probably not.
4436 If you do not define this macro, the target-independent code in the C++
4437 front end will generate a less efficient heavyweight thunk that calls
4438 @var{function} instead of jumping to it. The generic approach does
4439 not support varargs.
4442 @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})
4443 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4444 to output the assembler code for the thunk function specified by the
4445 arguments it is passed, and false otherwise. In the latter case, the
4446 generic approach will be used by the C++ front end, with the limitations
4451 @subsection Generating Code for Profiling
4452 @cindex profiling, code generation
4454 These macros will help you generate code for profiling.
4456 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4457 A C statement or compound statement to output to @var{file} some
4458 assembler code to call the profiling subroutine @code{mcount}.
4461 The details of how @code{mcount} expects to be called are determined by
4462 your operating system environment, not by GCC@. To figure them out,
4463 compile a small program for profiling using the system's installed C
4464 compiler and look at the assembler code that results.
4466 Older implementations of @code{mcount} expect the address of a counter
4467 variable to be loaded into some register. The name of this variable is
4468 @samp{LP} followed by the number @var{labelno}, so you would generate
4469 the name using @samp{LP%d} in a @code{fprintf}.
4472 @defmac PROFILE_HOOK
4473 A C statement or compound statement to output to @var{file} some assembly
4474 code to call the profiling subroutine @code{mcount} even the target does
4475 not support profiling.
4478 @defmac NO_PROFILE_COUNTERS
4479 Define this macro if the @code{mcount} subroutine on your system does
4480 not need a counter variable allocated for each function. This is true
4481 for almost all modern implementations. If you define this macro, you
4482 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
4485 @defmac PROFILE_BEFORE_PROLOGUE
4486 Define this macro if the code for function profiling should come before
4487 the function prologue. Normally, the profiling code comes after.
4491 @subsection Permitting tail calls
4494 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4495 True if it is ok to do sibling call optimization for the specified
4496 call expression @var{exp}. @var{decl} will be the called function,
4497 or @code{NULL} if this is an indirect call.
4499 It is not uncommon for limitations of calling conventions to prevent
4500 tail calls to functions outside the current unit of translation, or
4501 during PIC compilation. The hook is used to enforce these restrictions,
4502 as the @code{sibcall} md pattern can not fail, or fall over to a
4503 ``normal'' call. The criteria for successful sibling call optimization
4504 may vary greatly between different architectures.
4507 @deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap *@var{regs})
4508 Add any hard registers to @var{regs} that are live on entry to the
4509 function. This hook only needs to be defined to provide registers that
4510 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4511 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4512 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4513 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4516 @node Stack Smashing Protection
4517 @subsection Stack smashing protection
4518 @cindex stack smashing protection
4520 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void)
4521 This hook returns a @code{DECL} node for the external variable to use
4522 for the stack protection guard. This variable is initialized by the
4523 runtime to some random value and is used to initialize the guard value
4524 that is placed at the top of the local stack frame. The type of this
4525 variable must be @code{ptr_type_node}.
4527 The default version of this hook creates a variable called
4528 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
4531 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void)
4532 This hook returns a tree expression that alerts the runtime that the
4533 stack protect guard variable has been modified. This expression should
4534 involve a call to a @code{noreturn} function.
4536 The default version of this hook invokes a function called
4537 @samp{__stack_chk_fail}, taking no arguments. This function is
4538 normally defined in @file{libgcc2.c}.
4542 @section Implementing the Varargs Macros
4543 @cindex varargs implementation
4545 GCC comes with an implementation of @code{<varargs.h>} and
4546 @code{<stdarg.h>} that work without change on machines that pass arguments
4547 on the stack. Other machines require their own implementations of
4548 varargs, and the two machine independent header files must have
4549 conditionals to include it.
4551 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4552 the calling convention for @code{va_start}. The traditional
4553 implementation takes just one argument, which is the variable in which
4554 to store the argument pointer. The ISO implementation of
4555 @code{va_start} takes an additional second argument. The user is
4556 supposed to write the last named argument of the function here.
4558 However, @code{va_start} should not use this argument. The way to find
4559 the end of the named arguments is with the built-in functions described
4562 @defmac __builtin_saveregs ()
4563 Use this built-in function to save the argument registers in memory so
4564 that the varargs mechanism can access them. Both ISO and traditional
4565 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4566 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4568 On some machines, @code{__builtin_saveregs} is open-coded under the
4569 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4570 other machines, it calls a routine written in assembler language,
4571 found in @file{libgcc2.c}.
4573 Code generated for the call to @code{__builtin_saveregs} appears at the
4574 beginning of the function, as opposed to where the call to
4575 @code{__builtin_saveregs} is written, regardless of what the code is.
4576 This is because the registers must be saved before the function starts
4577 to use them for its own purposes.
4578 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4582 @defmac __builtin_args_info (@var{category})
4583 Use this built-in function to find the first anonymous arguments in
4586 In general, a machine may have several categories of registers used for
4587 arguments, each for a particular category of data types. (For example,
4588 on some machines, floating-point registers are used for floating-point
4589 arguments while other arguments are passed in the general registers.)
4590 To make non-varargs functions use the proper calling convention, you
4591 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4592 registers in each category have been used so far
4594 @code{__builtin_args_info} accesses the same data structure of type
4595 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4596 with it, with @var{category} specifying which word to access. Thus, the
4597 value indicates the first unused register in a given category.
4599 Normally, you would use @code{__builtin_args_info} in the implementation
4600 of @code{va_start}, accessing each category just once and storing the
4601 value in the @code{va_list} object. This is because @code{va_list} will
4602 have to update the values, and there is no way to alter the
4603 values accessed by @code{__builtin_args_info}.
4606 @defmac __builtin_next_arg (@var{lastarg})
4607 This is the equivalent of @code{__builtin_args_info}, for stack
4608 arguments. It returns the address of the first anonymous stack
4609 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4610 returns the address of the location above the first anonymous stack
4611 argument. Use it in @code{va_start} to initialize the pointer for
4612 fetching arguments from the stack. Also use it in @code{va_start} to
4613 verify that the second parameter @var{lastarg} is the last named argument
4614 of the current function.
4617 @defmac __builtin_classify_type (@var{object})
4618 Since each machine has its own conventions for which data types are
4619 passed in which kind of register, your implementation of @code{va_arg}
4620 has to embody these conventions. The easiest way to categorize the
4621 specified data type is to use @code{__builtin_classify_type} together
4622 with @code{sizeof} and @code{__alignof__}.
4624 @code{__builtin_classify_type} ignores the value of @var{object},
4625 considering only its data type. It returns an integer describing what
4626 kind of type that is---integer, floating, pointer, structure, and so on.
4628 The file @file{typeclass.h} defines an enumeration that you can use to
4629 interpret the values of @code{__builtin_classify_type}.
4632 These machine description macros help implement varargs:
4634 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
4635 If defined, this hook produces the machine-specific code for a call to
4636 @code{__builtin_saveregs}. This code will be moved to the very
4637 beginning of the function, before any parameter access are made. The
4638 return value of this function should be an RTX that contains the value
4639 to use as the return of @code{__builtin_saveregs}.
4642 @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})
4643 This target hook offers an alternative to using
4644 @code{__builtin_saveregs} and defining the hook
4645 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
4646 register arguments into the stack so that all the arguments appear to
4647 have been passed consecutively on the stack. Once this is done, you can
4648 use the standard implementation of varargs that works for machines that
4649 pass all their arguments on the stack.
4651 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
4652 structure, containing the values that are obtained after processing the
4653 named arguments. The arguments @var{mode} and @var{type} describe the
4654 last named argument---its machine mode and its data type as a tree node.
4656 The target hook should do two things: first, push onto the stack all the
4657 argument registers @emph{not} used for the named arguments, and second,
4658 store the size of the data thus pushed into the @code{int}-valued
4659 variable pointed to by @var{pretend_args_size}. The value that you
4660 store here will serve as additional offset for setting up the stack
4663 Because you must generate code to push the anonymous arguments at
4664 compile time without knowing their data types,
4665 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
4666 have just a single category of argument register and use it uniformly
4669 If the argument @var{second_time} is nonzero, it means that the
4670 arguments of the function are being analyzed for the second time. This
4671 happens for an inline function, which is not actually compiled until the
4672 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
4673 not generate any instructions in this case.
4676 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca})
4677 Define this hook to return @code{true} if the location where a function
4678 argument is passed depends on whether or not it is a named argument.
4680 This hook controls how the @var{named} argument to @code{FUNCTION_ARG}
4681 is set for varargs and stdarg functions. If this hook returns
4682 @code{true}, the @var{named} argument is always true for named
4683 arguments, and false for unnamed arguments. If it returns @code{false},
4684 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
4685 then all arguments are treated as named. Otherwise, all named arguments
4686 except the last are treated as named.
4688 You need not define this hook if it always returns zero.
4691 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED
4692 If you need to conditionally change ABIs so that one works with
4693 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
4694 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
4695 defined, then define this hook to return @code{true} if
4696 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
4697 Otherwise, you should not define this hook.
4701 @section Trampolines for Nested Functions
4702 @cindex trampolines for nested functions
4703 @cindex nested functions, trampolines for
4705 A @dfn{trampoline} is a small piece of code that is created at run time
4706 when the address of a nested function is taken. It normally resides on
4707 the stack, in the stack frame of the containing function. These macros
4708 tell GCC how to generate code to allocate and initialize a
4711 The instructions in the trampoline must do two things: load a constant
4712 address into the static chain register, and jump to the real address of
4713 the nested function. On CISC machines such as the m68k, this requires
4714 two instructions, a move immediate and a jump. Then the two addresses
4715 exist in the trampoline as word-long immediate operands. On RISC
4716 machines, it is often necessary to load each address into a register in
4717 two parts. Then pieces of each address form separate immediate
4720 The code generated to initialize the trampoline must store the variable
4721 parts---the static chain value and the function address---into the
4722 immediate operands of the instructions. On a CISC machine, this is
4723 simply a matter of copying each address to a memory reference at the
4724 proper offset from the start of the trampoline. On a RISC machine, it
4725 may be necessary to take out pieces of the address and store them
4728 @defmac TRAMPOLINE_TEMPLATE (@var{file})
4729 A C statement to output, on the stream @var{file}, assembler code for a
4730 block of data that contains the constant parts of a trampoline. This
4731 code should not include a label---the label is taken care of
4734 If you do not define this macro, it means no template is needed
4735 for the target. Do not define this macro on systems where the block move
4736 code to copy the trampoline into place would be larger than the code
4737 to generate it on the spot.
4740 @defmac TRAMPOLINE_SECTION
4741 Return the section into which the trampoline template is to be placed
4742 (@pxref{Sections}). The default value is @code{readonly_data_section}.
4745 @defmac TRAMPOLINE_SIZE
4746 A C expression for the size in bytes of the trampoline, as an integer.
4749 @defmac TRAMPOLINE_ALIGNMENT
4750 Alignment required for trampolines, in bits.
4752 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4753 is used for aligning trampolines.
4756 @defmac INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4757 A C statement to initialize the variable parts of a trampoline.
4758 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4759 an RTX for the address of the nested function; @var{static_chain} is an
4760 RTX for the static chain value that should be passed to the function
4764 @defmac TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4765 A C statement that should perform any machine-specific adjustment in
4766 the address of the trampoline. Its argument contains the address that
4767 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4768 used for a function call should be different from the address in which
4769 the template was stored, the different address should be assigned to
4770 @var{addr}. If this macro is not defined, @var{addr} will be used for
4773 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4774 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4775 If this macro is not defined, by default the trampoline is allocated as
4776 a stack slot. This default is right for most machines. The exceptions
4777 are machines where it is impossible to execute instructions in the stack
4778 area. On such machines, you may have to implement a separate stack,
4779 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4780 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4782 @var{fp} points to a data structure, a @code{struct function}, which
4783 describes the compilation status of the immediate containing function of
4784 the function which the trampoline is for. The stack slot for the
4785 trampoline is in the stack frame of this containing function. Other
4786 allocation strategies probably must do something analogous with this
4790 Implementing trampolines is difficult on many machines because they have
4791 separate instruction and data caches. Writing into a stack location
4792 fails to clear the memory in the instruction cache, so when the program
4793 jumps to that location, it executes the old contents.
4795 Here are two possible solutions. One is to clear the relevant parts of
4796 the instruction cache whenever a trampoline is set up. The other is to
4797 make all trampolines identical, by having them jump to a standard
4798 subroutine. The former technique makes trampoline execution faster; the
4799 latter makes initialization faster.
4801 To clear the instruction cache when a trampoline is initialized, define
4802 the following macro.
4804 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
4805 If defined, expands to a C expression clearing the @emph{instruction
4806 cache} in the specified interval. The definition of this macro would
4807 typically be a series of @code{asm} statements. Both @var{beg} and
4808 @var{end} are both pointer expressions.
4811 The operating system may also require the stack to be made executable
4812 before calling the trampoline. To implement this requirement, define
4813 the following macro.
4815 @defmac ENABLE_EXECUTE_STACK
4816 Define this macro if certain operations must be performed before executing
4817 code located on the stack. The macro should expand to a series of C
4818 file-scope constructs (e.g.@: functions) and provide a unique entry point
4819 named @code{__enable_execute_stack}. The target is responsible for
4820 emitting calls to the entry point in the code, for example from the
4821 @code{INITIALIZE_TRAMPOLINE} macro.
4824 To use a standard subroutine, define the following macro. In addition,
4825 you must make sure that the instructions in a trampoline fill an entire
4826 cache line with identical instructions, or else ensure that the
4827 beginning of the trampoline code is always aligned at the same point in
4828 its cache line. Look in @file{m68k.h} as a guide.
4830 @defmac TRANSFER_FROM_TRAMPOLINE
4831 Define this macro if trampolines need a special subroutine to do their
4832 work. The macro should expand to a series of @code{asm} statements
4833 which will be compiled with GCC@. They go in a library function named
4834 @code{__transfer_from_trampoline}.
4836 If you need to avoid executing the ordinary prologue code of a compiled
4837 C function when you jump to the subroutine, you can do so by placing a
4838 special label of your own in the assembler code. Use one @code{asm}
4839 statement to generate an assembler label, and another to make the label
4840 global. Then trampolines can use that label to jump directly to your
4841 special assembler code.
4845 @section Implicit Calls to Library Routines
4846 @cindex library subroutine names
4847 @cindex @file{libgcc.a}
4849 @c prevent bad page break with this line
4850 Here is an explanation of implicit calls to library routines.
4852 @defmac DECLARE_LIBRARY_RENAMES
4853 This macro, if defined, should expand to a piece of C code that will get
4854 expanded when compiling functions for libgcc.a. It can be used to
4855 provide alternate names for GCC's internal library functions if there
4856 are ABI-mandated names that the compiler should provide.
4859 @findex init_one_libfunc
4860 @findex set_optab_libfunc
4861 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
4862 This hook should declare additional library routines or rename
4863 existing ones, using the functions @code{set_optab_libfunc} and
4864 @code{init_one_libfunc} defined in @file{optabs.c}.
4865 @code{init_optabs} calls this macro after initializing all the normal
4868 The default is to do nothing. Most ports don't need to define this hook.
4871 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4872 This macro should return @code{true} if the library routine that
4873 implements the floating point comparison operator @var{comparison} in
4874 mode @var{mode} will return a boolean, and @var{false} if it will
4877 GCC's own floating point libraries return tristates from the
4878 comparison operators, so the default returns false always. Most ports
4879 don't need to define this macro.
4882 @defmac TARGET_LIB_INT_CMP_BIASED
4883 This macro should evaluate to @code{true} if the integer comparison
4884 functions (like @code{__cmpdi2}) return 0 to indicate that the first
4885 operand is smaller than the second, 1 to indicate that they are equal,
4886 and 2 to indicate that the first operand is greater than the second.
4887 If this macro evaluates to @code{false} the comparison functions return
4888 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
4889 in @file{libgcc.a}, you do not need to define this macro.
4892 @cindex US Software GOFAST, floating point emulation library
4893 @cindex floating point emulation library, US Software GOFAST
4894 @cindex GOFAST, floating point emulation library
4895 @findex gofast_maybe_init_libfuncs
4896 @defmac US_SOFTWARE_GOFAST
4897 Define this macro if your system C library uses the US Software GOFAST
4898 library to provide floating point emulation.
4900 In addition to defining this macro, your architecture must set
4901 @code{TARGET_INIT_LIBFUNCS} to @code{gofast_maybe_init_libfuncs}, or
4902 else call that function from its version of that hook. It is defined
4903 in @file{config/gofast.h}, which must be included by your
4904 architecture's @file{@var{cpu}.c} file. See @file{sparc/sparc.c} for
4907 If this macro is defined, the
4908 @code{TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL} target hook must return
4909 false for @code{SFmode} and @code{DFmode} comparisons.
4912 @cindex @code{EDOM}, implicit usage
4915 The value of @code{EDOM} on the target machine, as a C integer constant
4916 expression. If you don't define this macro, GCC does not attempt to
4917 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4918 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4921 If you do not define @code{TARGET_EDOM}, then compiled code reports
4922 domain errors by calling the library function and letting it report the
4923 error. If mathematical functions on your system use @code{matherr} when
4924 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4925 that @code{matherr} is used normally.
4928 @cindex @code{errno}, implicit usage
4929 @defmac GEN_ERRNO_RTX
4930 Define this macro as a C expression to create an rtl expression that
4931 refers to the global ``variable'' @code{errno}. (On certain systems,
4932 @code{errno} may not actually be a variable.) If you don't define this
4933 macro, a reasonable default is used.
4936 @cindex C99 math functions, implicit usage
4937 @defmac TARGET_C99_FUNCTIONS
4938 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
4939 @code{sinf} and similarly for other functions defined by C99 standard. The
4940 default is nonzero that should be proper value for most modern systems, however
4941 number of existing systems lacks support for these functions in the runtime so
4942 they needs this macro to be redefined to 0.
4945 @defmac NEXT_OBJC_RUNTIME
4946 Define this macro to generate code for Objective-C message sending using
4947 the calling convention of the NeXT system. This calling convention
4948 involves passing the object, the selector and the method arguments all
4949 at once to the method-lookup library function.
4951 The default calling convention passes just the object and the selector
4952 to the lookup function, which returns a pointer to the method.
4955 @node Addressing Modes
4956 @section Addressing Modes
4957 @cindex addressing modes
4959 @c prevent bad page break with this line
4960 This is about addressing modes.
4962 @defmac HAVE_PRE_INCREMENT
4963 @defmacx HAVE_PRE_DECREMENT
4964 @defmacx HAVE_POST_INCREMENT
4965 @defmacx HAVE_POST_DECREMENT
4966 A C expression that is nonzero if the machine supports pre-increment,
4967 pre-decrement, post-increment, or post-decrement addressing respectively.
4970 @defmac HAVE_PRE_MODIFY_DISP
4971 @defmacx HAVE_POST_MODIFY_DISP
4972 A C expression that is nonzero if the machine supports pre- or
4973 post-address side-effect generation involving constants other than
4974 the size of the memory operand.
4977 @defmac HAVE_PRE_MODIFY_REG
4978 @defmacx HAVE_POST_MODIFY_REG
4979 A C expression that is nonzero if the machine supports pre- or
4980 post-address side-effect generation involving a register displacement.
4983 @defmac CONSTANT_ADDRESS_P (@var{x})
4984 A C expression that is 1 if the RTX @var{x} is a constant which
4985 is a valid address. On most machines, this can be defined as
4986 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4987 in which constant addresses are supported.
4990 @defmac CONSTANT_P (@var{x})
4991 @code{CONSTANT_P}, which is defined by target-independent code,
4992 accepts integer-values expressions whose values are not explicitly
4993 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
4994 expressions and @code{const} arithmetic expressions, in addition to
4995 @code{const_int} and @code{const_double} expressions.
4998 @defmac MAX_REGS_PER_ADDRESS
4999 A number, the maximum number of registers that can appear in a valid
5000 memory address. Note that it is up to you to specify a value equal to
5001 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
5005 @defmac GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5006 A C compound statement with a conditional @code{goto @var{label};}
5007 executed if @var{x} (an RTX) is a legitimate memory address on the
5008 target machine for a memory operand of mode @var{mode}.
5010 It usually pays to define several simpler macros to serve as
5011 subroutines for this one. Otherwise it may be too complicated to
5014 This macro must exist in two variants: a strict variant and a
5015 non-strict one. The strict variant is used in the reload pass. It
5016 must be defined so that any pseudo-register that has not been
5017 allocated a hard register is considered a memory reference. In
5018 contexts where some kind of register is required, a pseudo-register
5019 with no hard register must be rejected.
5021 The non-strict variant is used in other passes. It must be defined to
5022 accept all pseudo-registers in every context where some kind of
5023 register is required.
5025 @findex REG_OK_STRICT
5026 Compiler source files that want to use the strict variant of this
5027 macro define the macro @code{REG_OK_STRICT}. You should use an
5028 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
5029 in that case and the non-strict variant otherwise.
5031 Subroutines to check for acceptable registers for various purposes (one
5032 for base registers, one for index registers, and so on) are typically
5033 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
5034 Then only these subroutine macros need have two variants; the higher
5035 levels of macros may be the same whether strict or not.
5037 Normally, constant addresses which are the sum of a @code{symbol_ref}
5038 and an integer are stored inside a @code{const} RTX to mark them as
5039 constant. Therefore, there is no need to recognize such sums
5040 specifically as legitimate addresses. Normally you would simply
5041 recognize any @code{const} as legitimate.
5043 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5044 sums that are not marked with @code{const}. It assumes that a naked
5045 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5046 naked constant sums as illegitimate addresses, so that none of them will
5047 be given to @code{PRINT_OPERAND_ADDRESS}.
5049 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5050 On some machines, whether a symbolic address is legitimate depends on
5051 the section that the address refers to. On these machines, define the
5052 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5053 into the @code{symbol_ref}, and then check for it here. When you see a
5054 @code{const}, you will have to look inside it to find the
5055 @code{symbol_ref} in order to determine the section. @xref{Assembler
5059 @defmac REG_OK_FOR_BASE_P (@var{x})
5060 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
5061 RTX) is valid for use as a base register. For hard registers, it
5062 should always accept those which the hardware permits and reject the
5063 others. Whether the macro accepts or rejects pseudo registers must be
5064 controlled by @code{REG_OK_STRICT} as described above. This usually
5065 requires two variant definitions, of which @code{REG_OK_STRICT}
5066 controls the one actually used.
5069 @defmac REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
5070 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
5071 that expression may examine the mode of the memory reference in
5072 @var{mode}. You should define this macro if the mode of the memory
5073 reference affects whether a register may be used as a base register. If
5074 you define this macro, the compiler will use it instead of
5075 @code{REG_OK_FOR_BASE_P}.
5078 @defmac REG_MODE_OK_FOR_REG_BASE_P (@var{x}, @var{mode})
5079 A C expression which is nonzero if @var{x} (assumed to be a @code{reg} RTX)
5080 is suitable for use as a base register in base plus index operand addresses,
5081 accessing memory in mode @var{mode}. It may be either a suitable hard
5082 register or a pseudo register that has been allocated such a hard register.
5083 You should define this macro if base plus index addresses have different
5084 requirements than other base register uses.
5087 @defmac REG_OK_FOR_INDEX_P (@var{x})
5088 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
5089 RTX) is valid for use as an index register.
5091 The difference between an index register and a base register is that
5092 the index register may be scaled. If an address involves the sum of
5093 two registers, neither one of them scaled, then either one may be
5094 labeled the ``base'' and the other the ``index''; but whichever
5095 labeling is used must fit the machine's constraints of which registers
5096 may serve in each capacity. The compiler will try both labelings,
5097 looking for one that is valid, and will reload one or both registers
5098 only if neither labeling works.
5101 @defmac FIND_BASE_TERM (@var{x})
5102 A C expression to determine the base term of address @var{x}.
5103 This macro is used in only one place: `find_base_term' in alias.c.
5105 It is always safe for this macro to not be defined. It exists so
5106 that alias analysis can understand machine-dependent addresses.
5108 The typical use of this macro is to handle addresses containing
5109 a label_ref or symbol_ref within an UNSPEC@.
5112 @defmac LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
5113 A C compound statement that attempts to replace @var{x} with a valid
5114 memory address for an operand of mode @var{mode}. @var{win} will be a
5115 C statement label elsewhere in the code; the macro definition may use
5118 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
5122 to avoid further processing if the address has become legitimate.
5124 @findex break_out_memory_refs
5125 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5126 and @var{oldx} will be the operand that was given to that function to produce
5129 The code generated by this macro should not alter the substructure of
5130 @var{x}. If it transforms @var{x} into a more legitimate form, it
5131 should assign @var{x} (which will always be a C variable) a new value.
5133 It is not necessary for this macro to come up with a legitimate
5134 address. The compiler has standard ways of doing so in all cases. In
5135 fact, it is safe to omit this macro. But often a
5136 machine-dependent strategy can generate better code.
5139 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5140 A C compound statement that attempts to replace @var{x}, which is an address
5141 that needs reloading, with a valid memory address for an operand of mode
5142 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5143 It is not necessary to define this macro, but it might be useful for
5144 performance reasons.
5146 For example, on the i386, it is sometimes possible to use a single
5147 reload register instead of two by reloading a sum of two pseudo
5148 registers into a register. On the other hand, for number of RISC
5149 processors offsets are limited so that often an intermediate address
5150 needs to be generated in order to address a stack slot. By defining
5151 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5152 generated for adjacent some stack slots can be made identical, and thus
5155 @emph{Note}: This macro should be used with caution. It is necessary
5156 to know something of how reload works in order to effectively use this,
5157 and it is quite easy to produce macros that build in too much knowledge
5158 of reload internals.
5160 @emph{Note}: This macro must be able to reload an address created by a
5161 previous invocation of this macro. If it fails to handle such addresses
5162 then the compiler may generate incorrect code or abort.
5165 The macro definition should use @code{push_reload} to indicate parts that
5166 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5167 suitable to be passed unaltered to @code{push_reload}.
5169 The code generated by this macro must not alter the substructure of
5170 @var{x}. If it transforms @var{x} into a more legitimate form, it
5171 should assign @var{x} (which will always be a C variable) a new value.
5172 This also applies to parts that you change indirectly by calling
5175 @findex strict_memory_address_p
5176 The macro definition may use @code{strict_memory_address_p} to test if
5177 the address has become legitimate.
5180 If you want to change only a part of @var{x}, one standard way of doing
5181 this is to use @code{copy_rtx}. Note, however, that is unshares only a
5182 single level of rtl. Thus, if the part to be changed is not at the
5183 top level, you'll need to replace first the top level.
5184 It is not necessary for this macro to come up with a legitimate
5185 address; but often a machine-dependent strategy can generate better code.
5188 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5189 A C statement or compound statement with a conditional @code{goto
5190 @var{label};} executed if memory address @var{x} (an RTX) can have
5191 different meanings depending on the machine mode of the memory
5192 reference it is used for or if the address is valid for some modes
5195 Autoincrement and autodecrement addresses typically have mode-dependent
5196 effects because the amount of the increment or decrement is the size
5197 of the operand being addressed. Some machines have other mode-dependent
5198 addresses. Many RISC machines have no mode-dependent addresses.
5200 You may assume that @var{addr} is a valid address for the machine.
5203 @defmac LEGITIMATE_CONSTANT_P (@var{x})
5204 A C expression that is nonzero if @var{x} is a legitimate constant for
5205 an immediate operand on the target machine. You can assume that
5206 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5207 @samp{1} is a suitable definition for this macro on machines where
5208 anything @code{CONSTANT_P} is valid.
5211 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5212 This hook is used to undo the possibly obfuscating effects of the
5213 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5214 macros. Some backend implementations of these macros wrap symbol
5215 references inside an @code{UNSPEC} rtx to represent PIC or similar
5216 addressing modes. This target hook allows GCC's optimizers to understand
5217 the semantics of these opaque @code{UNSPEC}s by converting them back
5218 into their original form.
5221 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (rtx @var{x})
5222 This hook should return true if @var{x} is of a form that cannot (or
5223 should not) be spilled to the constant pool. The default version of
5224 this hook returns false.
5226 The primary reason to define this hook is to prevent reload from
5227 deciding that a non-legitimate constant would be better reloaded
5228 from the constant pool instead of spilling and reloading a register
5229 holding the constant. This restriction is often true of addresses
5230 of TLS symbols for various targets.
5233 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (enum machine_mode @var{mode}, rtx @var{x})
5234 This hook should return true if pool entries for constant @var{x} can
5235 be placed in an @code{object_block} structure. @var{mode} is the mode
5238 The default version returns false for all constants.
5241 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5242 This hook should return the DECL of a function @var{f} that given an
5243 address @var{addr} as an argument returns a mask @var{m} that can be
5244 used to extract from two vectors the relevant data that resides in
5245 @var{addr} in case @var{addr} is not properly aligned.
5247 The autovectrizer, when vectorizing a load operation from an address
5248 @var{addr} that may be unaligned, will generate two vector loads from
5249 the two aligned addresses around @var{addr}. It then generates a
5250 @code{REALIGN_LOAD} operation to extract the relevant data from the
5251 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5252 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5253 the third argument, @var{OFF}, defines how the data will be extracted
5254 from these two vectors: if @var{OFF} is 0, then the returned vector is
5255 @var{v2}; otherwise, the returned vector is composed from the last
5256 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5257 @var{OFF} elements of @var{v2}.
5259 If this hook is defined, the autovectorizer will generate a call
5260 to @var{f} (using the DECL tree that this hook returns) and will
5261 use the return value of @var{f} as the argument @var{OFF} to
5262 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5263 should comply with the semantics expected by @code{REALIGN_LOAD}
5265 If this hook is not defined, then @var{addr} will be used as
5266 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5267 log2(@var{VS})-1 bits of @var{addr} will be considered.
5270 @node Anchored Addresses
5271 @section Anchored Addresses
5272 @cindex anchored addresses
5273 @cindex @option{-fsection-anchors}
5275 GCC usually addresses every static object as a separate entity.
5276 For example, if we have:
5280 int foo (void) @{ return a + b + c; @}
5283 the code for @code{foo} will usually calculate three separate symbolic
5284 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
5285 it would be better to calculate just one symbolic address and access
5286 the three variables relative to it. The equivalent pseudocode would
5292 register int *xr = &x;
5293 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5297 (which isn't valid C). We refer to shared addresses like @code{x} as
5298 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
5300 The hooks below describe the target properties that GCC needs to know
5301 in order to make effective use of section anchors. It won't use
5302 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
5303 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
5305 @deftypevar {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET
5306 The minimum offset that should be applied to a section anchor.
5307 On most targets, it should be the smallest offset that can be
5308 applied to a base register while still giving a legitimate address
5309 for every mode. The default value is 0.
5312 @deftypevar {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET
5313 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
5314 offset that should be applied to section anchors. The default
5318 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x})
5319 Write the assembly code to define section anchor @var{x}, which is a
5320 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
5321 The hook is called with the assembly output position set to the beginning
5322 of @code{SYMBOL_REF_BLOCK (@var{x})}.
5324 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
5325 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
5326 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
5327 is @code{NULL}, which disables the use of section anchors altogether.
5330 @deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (rtx @var{x})
5331 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
5332 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
5333 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
5335 The default version is correct for most targets, but you might need to
5336 intercept this hook to handle things like target-specific attributes
5337 or target-specific sections.
5340 @node Condition Code
5341 @section Condition Code Status
5342 @cindex condition code status
5344 @c prevent bad page break with this line
5345 This describes the condition code status.
5348 The file @file{conditions.h} defines a variable @code{cc_status} to
5349 describe how the condition code was computed (in case the interpretation of
5350 the condition code depends on the instruction that it was set by). This
5351 variable contains the RTL expressions on which the condition code is
5352 currently based, and several standard flags.
5354 Sometimes additional machine-specific flags must be defined in the machine
5355 description header file. It can also add additional machine-specific
5356 information by defining @code{CC_STATUS_MDEP}.
5358 @defmac CC_STATUS_MDEP
5359 C code for a data type which is used for declaring the @code{mdep}
5360 component of @code{cc_status}. It defaults to @code{int}.
5362 This macro is not used on machines that do not use @code{cc0}.
5365 @defmac CC_STATUS_MDEP_INIT
5366 A C expression to initialize the @code{mdep} field to ``empty''.
5367 The default definition does nothing, since most machines don't use
5368 the field anyway. If you want to use the field, you should probably
5369 define this macro to initialize it.
5371 This macro is not used on machines that do not use @code{cc0}.
5374 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5375 A C compound statement to set the components of @code{cc_status}
5376 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5377 this macro's responsibility to recognize insns that set the condition
5378 code as a byproduct of other activity as well as those that explicitly
5381 This macro is not used on machines that do not use @code{cc0}.
5383 If there are insns that do not set the condition code but do alter
5384 other machine registers, this macro must check to see whether they
5385 invalidate the expressions that the condition code is recorded as
5386 reflecting. For example, on the 68000, insns that store in address
5387 registers do not set the condition code, which means that usually
5388 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5389 insns. But suppose that the previous insn set the condition code
5390 based on location @samp{a4@@(102)} and the current insn stores a new
5391 value in @samp{a4}. Although the condition code is not changed by
5392 this, it will no longer be true that it reflects the contents of
5393 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5394 @code{cc_status} in this case to say that nothing is known about the
5395 condition code value.
5397 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5398 with the results of peephole optimization: insns whose patterns are
5399 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5400 constants which are just the operands. The RTL structure of these
5401 insns is not sufficient to indicate what the insns actually do. What
5402 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5403 @code{CC_STATUS_INIT}.
5405 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5406 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5407 @samp{cc}. This avoids having detailed information about patterns in
5408 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5411 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5412 Returns a mode from class @code{MODE_CC} to be used when comparison
5413 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5414 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5415 @pxref{Jump Patterns} for a description of the reason for this
5419 #define SELECT_CC_MODE(OP,X,Y) \
5420 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5421 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5422 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5423 || GET_CODE (X) == NEG) \
5424 ? CC_NOOVmode : CCmode))
5427 You should define this macro if and only if you define extra CC modes
5428 in @file{@var{machine}-modes.def}.
5431 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5432 On some machines not all possible comparisons are defined, but you can
5433 convert an invalid comparison into a valid one. For example, the Alpha
5434 does not have a @code{GT} comparison, but you can use an @code{LT}
5435 comparison instead and swap the order of the operands.
5437 On such machines, define this macro to be a C statement to do any
5438 required conversions. @var{code} is the initial comparison code
5439 and @var{op0} and @var{op1} are the left and right operands of the
5440 comparison, respectively. You should modify @var{code}, @var{op0}, and
5441 @var{op1} as required.
5443 GCC will not assume that the comparison resulting from this macro is
5444 valid but will see if the resulting insn matches a pattern in the
5447 You need not define this macro if it would never change the comparison
5451 @defmac REVERSIBLE_CC_MODE (@var{mode})
5452 A C expression whose value is one if it is always safe to reverse a
5453 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5454 can ever return @var{mode} for a floating-point inequality comparison,
5455 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5457 You need not define this macro if it would always returns zero or if the
5458 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5459 For example, here is the definition used on the SPARC, where floating-point
5460 inequality comparisons are always given @code{CCFPEmode}:
5463 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5467 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
5468 A C expression whose value is reversed condition code of the @var{code} for
5469 comparison done in CC_MODE @var{mode}. The macro is used only in case
5470 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5471 machine has some non-standard way how to reverse certain conditionals. For
5472 instance in case all floating point conditions are non-trapping, compiler may
5473 freely convert unordered compares to ordered one. Then definition may look
5477 #define REVERSE_CONDITION(CODE, MODE) \
5478 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5479 : reverse_condition_maybe_unordered (CODE))
5483 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
5484 A C expression that returns true if the conditional execution predicate
5485 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
5486 versa. Define this to return 0 if the target has conditional execution
5487 predicates that cannot be reversed safely. There is no need to validate
5488 that the arguments of op1 and op2 are the same, this is done separately.
5489 If no expansion is specified, this macro is defined as follows:
5492 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5493 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
5497 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *, unsigned int *)
5498 On targets which do not use @code{(cc0)}, and which use a hard
5499 register rather than a pseudo-register to hold condition codes, the
5500 regular CSE passes are often not able to identify cases in which the
5501 hard register is set to a common value. Use this hook to enable a
5502 small pass which optimizes such cases. This hook should return true
5503 to enable this pass, and it should set the integers to which its
5504 arguments point to the hard register numbers used for condition codes.
5505 When there is only one such register, as is true on most systems, the
5506 integer pointed to by the second argument should be set to
5507 @code{INVALID_REGNUM}.
5509 The default version of this hook returns false.
5512 @deftypefn {Target Hook} enum machine_mode TARGET_CC_MODES_COMPATIBLE (enum machine_mode, enum machine_mode)
5513 On targets which use multiple condition code modes in class
5514 @code{MODE_CC}, it is sometimes the case that a comparison can be
5515 validly done in more than one mode. On such a system, define this
5516 target hook to take two mode arguments and to return a mode in which
5517 both comparisons may be validly done. If there is no such mode,
5518 return @code{VOIDmode}.
5520 The default version of this hook checks whether the modes are the
5521 same. If they are, it returns that mode. If they are different, it
5522 returns @code{VOIDmode}.
5526 @section Describing Relative Costs of Operations
5527 @cindex costs of instructions
5528 @cindex relative costs
5529 @cindex speed of instructions
5531 These macros let you describe the relative speed of various operations
5532 on the target machine.
5534 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5535 A C expression for the cost of moving data of mode @var{mode} from a
5536 register in class @var{from} to one in class @var{to}. The classes are
5537 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5538 value of 2 is the default; other values are interpreted relative to
5541 It is not required that the cost always equal 2 when @var{from} is the
5542 same as @var{to}; on some machines it is expensive to move between
5543 registers if they are not general registers.
5545 If reload sees an insn consisting of a single @code{set} between two
5546 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5547 classes returns a value of 2, reload does not check to ensure that the
5548 constraints of the insn are met. Setting a cost of other than 2 will
5549 allow reload to verify that the constraints are met. You should do this
5550 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5553 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5554 A C expression for the cost of moving data of mode @var{mode} between a
5555 register of class @var{class} and memory; @var{in} is zero if the value
5556 is to be written to memory, nonzero if it is to be read in. This cost
5557 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5558 registers and memory is more expensive than between two registers, you
5559 should define this macro to express the relative cost.
5561 If you do not define this macro, GCC uses a default cost of 4 plus
5562 the cost of copying via a secondary reload register, if one is
5563 needed. If your machine requires a secondary reload register to copy
5564 between memory and a register of @var{class} but the reload mechanism is
5565 more complex than copying via an intermediate, define this macro to
5566 reflect the actual cost of the move.
5568 GCC defines the function @code{memory_move_secondary_cost} if
5569 secondary reloads are needed. It computes the costs due to copying via
5570 a secondary register. If your machine copies from memory using a
5571 secondary register in the conventional way but the default base value of
5572 4 is not correct for your machine, define this macro to add some other
5573 value to the result of that function. The arguments to that function
5574 are the same as to this macro.
5578 A C expression for the cost of a branch instruction. A value of 1 is
5579 the default; other values are interpreted relative to that.
5582 Here are additional macros which do not specify precise relative costs,
5583 but only that certain actions are more expensive than GCC would
5586 @defmac SLOW_BYTE_ACCESS
5587 Define this macro as a C expression which is nonzero if accessing less
5588 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5589 faster than accessing a word of memory, i.e., if such access
5590 require more than one instruction or if there is no difference in cost
5591 between byte and (aligned) word loads.
5593 When this macro is not defined, the compiler will access a field by
5594 finding the smallest containing object; when it is defined, a fullword
5595 load will be used if alignment permits. Unless bytes accesses are
5596 faster than word accesses, using word accesses is preferable since it
5597 may eliminate subsequent memory access if subsequent accesses occur to
5598 other fields in the same word of the structure, but to different bytes.
5601 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5602 Define this macro to be the value 1 if memory accesses described by the
5603 @var{mode} and @var{alignment} parameters have a cost many times greater
5604 than aligned accesses, for example if they are emulated in a trap
5607 When this macro is nonzero, the compiler will act as if
5608 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5609 moves. This can cause significantly more instructions to be produced.
5610 Therefore, do not set this macro nonzero if unaligned accesses only add a
5611 cycle or two to the time for a memory access.
5613 If the value of this macro is always zero, it need not be defined. If
5614 this macro is defined, it should produce a nonzero value when
5615 @code{STRICT_ALIGNMENT} is nonzero.
5619 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5620 which a sequence of insns should be generated instead of a
5621 string move insn or a library call. Increasing the value will always
5622 make code faster, but eventually incurs high cost in increased code size.
5624 Note that on machines where the corresponding move insn is a
5625 @code{define_expand} that emits a sequence of insns, this macro counts
5626 the number of such sequences.
5628 If you don't define this, a reasonable default is used.
5631 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5632 A C expression used to determine whether @code{move_by_pieces} will be used to
5633 copy a chunk of memory, or whether some other block move mechanism
5634 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5635 than @code{MOVE_RATIO}.
5638 @defmac MOVE_MAX_PIECES
5639 A C expression used by @code{move_by_pieces} to determine the largest unit
5640 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5644 The threshold of number of scalar move insns, @emph{below} which a sequence
5645 of insns should be generated to clear memory instead of a string clear insn
5646 or a library call. Increasing the value will always make code faster, but
5647 eventually incurs high cost in increased code size.
5649 If you don't define this, a reasonable default is used.
5652 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5653 A C expression used to determine whether @code{clear_by_pieces} will be used
5654 to clear a chunk of memory, or whether some other block clear mechanism
5655 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5656 than @code{CLEAR_RATIO}.
5659 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
5660 A C expression used to determine whether @code{store_by_pieces} will be
5661 used to set a chunk of memory to a constant value, or whether some other
5662 mechanism will be used. Used by @code{__builtin_memset} when storing
5663 values other than constant zero and by @code{__builtin_strcpy} when
5664 when called with a constant source string.
5665 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5666 than @code{MOVE_RATIO}.
5669 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
5670 A C expression used to determine whether a load postincrement is a good
5671 thing to use for a given mode. Defaults to the value of
5672 @code{HAVE_POST_INCREMENT}.
5675 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
5676 A C expression used to determine whether a load postdecrement is a good
5677 thing to use for a given mode. Defaults to the value of
5678 @code{HAVE_POST_DECREMENT}.
5681 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
5682 A C expression used to determine whether a load preincrement is a good
5683 thing to use for a given mode. Defaults to the value of
5684 @code{HAVE_PRE_INCREMENT}.
5687 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
5688 A C expression used to determine whether a load predecrement is a good
5689 thing to use for a given mode. Defaults to the value of
5690 @code{HAVE_PRE_DECREMENT}.
5693 @defmac USE_STORE_POST_INCREMENT (@var{mode})
5694 A C expression used to determine whether a store postincrement is a good
5695 thing to use for a given mode. Defaults to the value of
5696 @code{HAVE_POST_INCREMENT}.
5699 @defmac USE_STORE_POST_DECREMENT (@var{mode})
5700 A C expression used to determine whether a store postdecrement is a good
5701 thing to use for a given mode. Defaults to the value of
5702 @code{HAVE_POST_DECREMENT}.
5705 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
5706 This macro is used to determine whether a store preincrement is a good
5707 thing to use for a given mode. Defaults to the value of
5708 @code{HAVE_PRE_INCREMENT}.
5711 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
5712 This macro is used to determine whether a store predecrement is a good
5713 thing to use for a given mode. Defaults to the value of
5714 @code{HAVE_PRE_DECREMENT}.
5717 @defmac NO_FUNCTION_CSE
5718 Define this macro if it is as good or better to call a constant
5719 function address than to call an address kept in a register.
5722 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
5723 Define this macro if a non-short-circuit operation produced by
5724 @samp{fold_range_test ()} is optimal. This macro defaults to true if
5725 @code{BRANCH_COST} is greater than or equal to the value 2.
5728 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total})
5729 This target hook describes the relative costs of RTL expressions.
5731 The cost may depend on the precise form of the expression, which is
5732 available for examination in @var{x}, and the rtx code of the expression
5733 in which it is contained, found in @var{outer_code}. @var{code} is the
5734 expression code---redundant, since it can be obtained with
5735 @code{GET_CODE (@var{x})}.
5737 In implementing this hook, you can use the construct
5738 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
5741 On entry to the hook, @code{*@var{total}} contains a default estimate
5742 for the cost of the expression. The hook should modify this value as
5743 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
5744 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
5745 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
5747 When optimizing for code size, i.e.@: when @code{optimize_size} is
5748 nonzero, this target hook should be used to estimate the relative
5749 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
5751 The hook returns true when all subexpressions of @var{x} have been
5752 processed, and false when @code{rtx_cost} should recurse.
5755 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address})
5756 This hook computes the cost of an addressing mode that contains
5757 @var{address}. If not defined, the cost is computed from
5758 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
5760 For most CISC machines, the default cost is a good approximation of the
5761 true cost of the addressing mode. However, on RISC machines, all
5762 instructions normally have the same length and execution time. Hence
5763 all addresses will have equal costs.
5765 In cases where more than one form of an address is known, the form with
5766 the lowest cost will be used. If multiple forms have the same, lowest,
5767 cost, the one that is the most complex will be used.
5769 For example, suppose an address that is equal to the sum of a register
5770 and a constant is used twice in the same basic block. When this macro
5771 is not defined, the address will be computed in a register and memory
5772 references will be indirect through that register. On machines where
5773 the cost of the addressing mode containing the sum is no higher than
5774 that of a simple indirect reference, this will produce an additional
5775 instruction and possibly require an additional register. Proper
5776 specification of this macro eliminates this overhead for such machines.
5778 This hook is never called with an invalid address.
5780 On machines where an address involving more than one register is as
5781 cheap as an address computation involving only one register, defining
5782 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
5783 be live over a region of code where only one would have been if
5784 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
5785 should be considered in the definition of this macro. Equivalent costs
5786 should probably only be given to addresses with different numbers of
5787 registers on machines with lots of registers.
5791 @section Adjusting the Instruction Scheduler
5793 The instruction scheduler may need a fair amount of machine-specific
5794 adjustment in order to produce good code. GCC provides several target
5795 hooks for this purpose. It is usually enough to define just a few of
5796 them: try the first ones in this list first.
5798 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
5799 This hook returns the maximum number of instructions that can ever
5800 issue at the same time on the target machine. The default is one.
5801 Although the insn scheduler can define itself the possibility of issue
5802 an insn on the same cycle, the value can serve as an additional
5803 constraint to issue insns on the same simulated processor cycle (see
5804 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
5805 This value must be constant over the entire compilation. If you need
5806 it to vary depending on what the instructions are, you must use
5807 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
5810 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
5811 This hook is executed by the scheduler after it has scheduled an insn
5812 from the ready list. It should return the number of insns which can
5813 still be issued in the current cycle. The default is
5814 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
5815 @code{USE}, which normally are not counted against the issue rate.
5816 You should define this hook if some insns take more machine resources
5817 than others, so that fewer insns can follow them in the same cycle.
5818 @var{file} is either a null pointer, or a stdio stream to write any
5819 debug output to. @var{verbose} is the verbose level provided by
5820 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
5824 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
5825 This function corrects the value of @var{cost} based on the
5826 relationship between @var{insn} and @var{dep_insn} through the
5827 dependence @var{link}. It should return the new value. The default
5828 is to make no adjustment to @var{cost}. This can be used for example
5829 to specify to the scheduler using the traditional pipeline description
5830 that an output- or anti-dependence does not incur the same cost as a
5831 data-dependence. If the scheduler using the automaton based pipeline
5832 description, the cost of anti-dependence is zero and the cost of
5833 output-dependence is maximum of one and the difference of latency
5834 times of the first and the second insns. If these values are not
5835 acceptable, you could use the hook to modify them too. See also
5836 @pxref{Processor pipeline description}.
5839 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
5840 This hook adjusts the integer scheduling priority @var{priority} of
5841 @var{insn}. It should return the new priority. Increase the priority to
5842 execute @var{insn} earlier, reduce the priority to execute @var{insn}
5843 later. Do not define this hook if you do not need to adjust the
5844 scheduling priorities of insns.
5847 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
5848 This hook is executed by the scheduler after it has scheduled the ready
5849 list, to allow the machine description to reorder it (for example to
5850 combine two small instructions together on @samp{VLIW} machines).
5851 @var{file} is either a null pointer, or a stdio stream to write any
5852 debug output to. @var{verbose} is the verbose level provided by
5853 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
5854 list of instructions that are ready to be scheduled. @var{n_readyp} is
5855 a pointer to the number of elements in the ready list. The scheduler
5856 reads the ready list in reverse order, starting with
5857 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
5858 is the timer tick of the scheduler. You may modify the ready list and
5859 the number of ready insns. The return value is the number of insns that
5860 can issue this cycle; normally this is just @code{issue_rate}. See also
5861 @samp{TARGET_SCHED_REORDER2}.
5864 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
5865 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
5866 function is called whenever the scheduler starts a new cycle. This one
5867 is called once per iteration over a cycle, immediately after
5868 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
5869 return the number of insns to be scheduled in the same cycle. Defining
5870 this hook can be useful if there are frequent situations where
5871 scheduling one insn causes other insns to become ready in the same
5872 cycle. These other insns can then be taken into account properly.
5875 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
5876 This hook is called after evaluation forward dependencies of insns in
5877 chain given by two parameter values (@var{head} and @var{tail}
5878 correspondingly) but before insns scheduling of the insn chain. For
5879 example, it can be used for better insn classification if it requires
5880 analysis of dependencies. This hook can use backward and forward
5881 dependencies of the insn scheduler because they are already
5885 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
5886 This hook is executed by the scheduler at the beginning of each block of
5887 instructions that are to be scheduled. @var{file} is either a null
5888 pointer, or a stdio stream to write any debug output to. @var{verbose}
5889 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5890 @var{max_ready} is the maximum number of insns in the current scheduling
5891 region that can be live at the same time. This can be used to allocate
5892 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
5895 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
5896 This hook is executed by the scheduler at the end of each block of
5897 instructions that are to be scheduled. It can be used to perform
5898 cleanup of any actions done by the other scheduling hooks. @var{file}
5899 is either a null pointer, or a stdio stream to write any debug output
5900 to. @var{verbose} is the verbose level provided by
5901 @option{-fsched-verbose-@var{n}}.
5904 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
5905 This hook is executed by the scheduler after function level initializations.
5906 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
5907 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5908 @var{old_max_uid} is the maximum insn uid when scheduling begins.
5911 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
5912 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
5913 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
5914 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5917 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
5918 The hook returns an RTL insn. The automaton state used in the
5919 pipeline hazard recognizer is changed as if the insn were scheduled
5920 when the new simulated processor cycle starts. Usage of the hook may
5921 simplify the automaton pipeline description for some @acronym{VLIW}
5922 processors. If the hook is defined, it is used only for the automaton
5923 based pipeline description. The default is not to change the state
5924 when the new simulated processor cycle starts.
5927 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
5928 The hook can be used to initialize data used by the previous hook.
5931 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
5932 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
5933 to changed the state as if the insn were scheduled when the new
5934 simulated processor cycle finishes.
5937 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
5938 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
5939 used to initialize data used by the previous hook.
5942 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
5943 This hook controls better choosing an insn from the ready insn queue
5944 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
5945 chooses the first insn from the queue. If the hook returns a positive
5946 value, an additional scheduler code tries all permutations of
5947 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
5948 subsequent ready insns to choose an insn whose issue will result in
5949 maximal number of issued insns on the same cycle. For the
5950 @acronym{VLIW} processor, the code could actually solve the problem of
5951 packing simple insns into the @acronym{VLIW} insn. Of course, if the
5952 rules of @acronym{VLIW} packing are described in the automaton.
5954 This code also could be used for superscalar @acronym{RISC}
5955 processors. Let us consider a superscalar @acronym{RISC} processor
5956 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
5957 @var{B}, some insns can be executed only in pipelines @var{B} or
5958 @var{C}, and one insn can be executed in pipeline @var{B}. The
5959 processor may issue the 1st insn into @var{A} and the 2nd one into
5960 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
5961 until the next cycle. If the scheduler issues the 3rd insn the first,
5962 the processor could issue all 3 insns per cycle.
5964 Actually this code demonstrates advantages of the automaton based
5965 pipeline hazard recognizer. We try quickly and easy many insn
5966 schedules to choose the best one.
5968 The default is no multipass scheduling.
5971 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
5973 This hook controls what insns from the ready insn queue will be
5974 considered for the multipass insn scheduling. If the hook returns
5975 zero for insn passed as the parameter, the insn will be not chosen to
5978 The default is that any ready insns can be chosen to be issued.
5981 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
5983 This hook is called by the insn scheduler before issuing insn passed
5984 as the third parameter on given cycle. If the hook returns nonzero,
5985 the insn is not issued on given processors cycle. Instead of that,
5986 the processor cycle is advanced. If the value passed through the last
5987 parameter is zero, the insn ready queue is not sorted on the new cycle
5988 start as usually. The first parameter passes file for debugging
5989 output. The second one passes the scheduler verbose level of the
5990 debugging output. The forth and the fifth parameter values are
5991 correspondingly processor cycle on which the previous insn has been
5992 issued and the current processor cycle.
5995 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (rtx @var{insn1}, rtx @var{insn2}, rtx @var{dep_link}, int @var{dep_cost}, int @var{distance})
5996 This hook is used to define which dependences are considered costly by
5997 the target, so costly that it is not advisable to schedule the insns that
5998 are involved in the dependence too close to one another. The parameters
5999 to this hook are as follows: The second parameter @var{insn2} is dependent
6000 upon the first parameter @var{insn1}. The dependence between @var{insn1}
6001 and @var{insn2} is represented by the third parameter @var{dep_link}. The
6002 fourth parameter @var{cost} is the cost of the dependence, and the fifth
6003 parameter @var{distance} is the distance in cycles between the two insns.
6004 The hook returns @code{true} if considering the distance between the two
6005 insns the dependence between them is considered costly by the target,
6006 and @code{false} otherwise.
6008 Defining this hook can be useful in multiple-issue out-of-order machines,
6009 where (a) it's practically hopeless to predict the actual data/resource
6010 delays, however: (b) there's a better chance to predict the actual grouping
6011 that will be formed, and (c) correctly emulating the grouping can be very
6012 important. In such targets one may want to allow issuing dependent insns
6013 closer to one another---i.e., closer than the dependence distance; however,
6014 not in cases of "costly dependences", which this hooks allows to define.
6017 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST_2 (rtx @var{insn}, int @var{dep_type}, rtx @var{dep_insn}, int @var{cost})
6018 This hook is a modified version of @samp{TARGET_SCHED_ADJUST_COST}. Instead
6019 of passing dependence as a second parameter, it passes a type of that
6020 dependence. This is useful to calculate cost of dependence between insns
6021 not having the corresponding link. If @samp{TARGET_SCHED_ADJUST_COST_2} is
6022 definded it is used instead of @samp{TARGET_SCHED_ADJUST_COST}.
6025 @deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void)
6026 This hook is called by the insn scheduler after emitting a new instruction to
6027 the instruction stream. The hook notifies a target backend to extend its
6028 per instruction data structures.
6031 @deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx @var{insn}, int @var{request}, rtx *@var{new_pat})
6032 This hook is called by the insn scheduler when @var{insn} has only
6033 speculative dependencies and therefore can be scheduled speculatively.
6034 The hook is used to check if the pattern of @var{insn} has a speculative
6035 version and, in case of successful check, to generate that speculative
6036 pattern. The hook should return 1, if the instruction has a speculative form,
6037 or -1, if it doesn't. @var{request} describes the type of requested
6038 speculation. If the return value equals 1 then @var{new_pat} is assigned
6039 the generated speculative pattern.
6042 @deftypefn {Target Hook} int TARGET_SCHED_NEEDS_BLOCK_P (rtx @var{insn})
6043 This hook is called by the insn scheduler during generation of recovery code
6044 for @var{insn}. It should return non-zero, if the corresponding check
6045 instruction should branch to recovery code, or zero otherwise.
6048 @deftypefn {Target Hook} rtx TARGET_SCHED_GEN_CHECK (rtx @var{insn}, rtx @var{label}, int @var{mutate_p})
6049 This hook is called by the insn scheduler to generate a pattern for recovery
6050 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
6051 speculative instruction for which the check should be generated.
6052 @var{label} is either a label of a basic block, where recovery code should
6053 be emitted, or a null pointer, when requested check doesn't branch to
6054 recovery code (a simple check). If @var{mutate_p} is non-zero, then
6055 a pattern for a branchy check corresponding to a simple check denoted by
6056 @var{insn} should be generated. In this case @var{label} can't be null.
6059 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC (rtx @var{insn})
6060 This hook is used as a workaround for
6061 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD} not being
6062 called on the first instruction of the ready list. The hook is used to
6063 discard speculative instruction that stand first in the ready list from
6064 being scheduled on the current cycle. For non-speculative instructions,
6065 the hook should always return non-zero. For example, in the ia64 backend
6066 the hook is used to cancel data speculative insns when the ALAT table
6070 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (unsigned int *@var{flags}, spec_info_t @var{spec_info})
6071 This hook is used by the insn scheduler to find out what features should be
6072 enabled/used. @var{flags} initially may have either the SCHED_RGN or SCHED_EBB
6073 bit set. This denotes the scheduler pass for which the data should be
6074 provided. The target backend should modify @var{flags} by modifying
6075 the bits correponding to the following features: USE_DEPS_LIST, USE_GLAT,
6076 DETACH_LIFE_INFO, and DO_SPECULATION. For the DO_SPECULATION feature
6077 an additional structure @var{spec_info} should be filled by the target.
6078 The structure describes speculation types that can be used in the scheduler.
6082 @section Dividing the Output into Sections (Texts, Data, @dots{})
6083 @c the above section title is WAY too long. maybe cut the part between
6084 @c the (...)? --mew 10feb93
6086 An object file is divided into sections containing different types of
6087 data. In the most common case, there are three sections: the @dfn{text
6088 section}, which holds instructions and read-only data; the @dfn{data
6089 section}, which holds initialized writable data; and the @dfn{bss
6090 section}, which holds uninitialized data. Some systems have other kinds
6093 @file{varasm.c} provides several well-known sections, such as
6094 @code{text_section}, @code{data_section} and @code{bss_section}.
6095 The normal way of controlling a @code{@var{foo}_section} variable
6096 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
6097 as described below. The macros are only read once, when @file{varasm.c}
6098 initializes itself, so their values must be run-time constants.
6099 They may however depend on command-line flags.
6101 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
6102 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
6103 to be string literals.
6105 Some assemblers require a different string to be written every time a
6106 section is selected. If your assembler falls into this category, you
6107 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
6108 @code{get_unnamed_section} to set up the sections.
6110 You must always create a @code{text_section}, either by defining
6111 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
6112 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
6113 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
6114 create a distinct @code{readonly_data_section}, the default is to
6115 reuse @code{text_section}.
6117 All the other @file{varasm.c} sections are optional, and are null
6118 if the target does not provide them.
6120 @defmac TEXT_SECTION_ASM_OP
6121 A C expression whose value is a string, including spacing, containing the
6122 assembler operation that should precede instructions and read-only data.
6123 Normally @code{"\t.text"} is right.
6126 @defmac HOT_TEXT_SECTION_NAME
6127 If defined, a C string constant for the name of the section containing most
6128 frequently executed functions of the program. If not defined, GCC will provide
6129 a default definition if the target supports named sections.
6132 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
6133 If defined, a C string constant for the name of the section containing unlikely
6134 executed functions in the program.
6137 @defmac DATA_SECTION_ASM_OP
6138 A C expression whose value is a string, including spacing, containing the
6139 assembler operation to identify the following data as writable initialized
6140 data. Normally @code{"\t.data"} is right.
6143 @defmac SDATA_SECTION_ASM_OP
6144 If defined, a C expression whose value is a string, including spacing,
6145 containing the assembler operation to identify the following data as
6146 initialized, writable small data.
6149 @defmac READONLY_DATA_SECTION_ASM_OP
6150 A C expression whose value is a string, including spacing, containing the
6151 assembler operation to identify the following data as read-only initialized
6155 @defmac BSS_SECTION_ASM_OP
6156 If defined, a C expression whose value is a string, including spacing,
6157 containing the assembler operation to identify the following data as
6158 uninitialized global data. If not defined, and neither
6159 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
6160 uninitialized global data will be output in the data section if
6161 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
6165 @defmac SBSS_SECTION_ASM_OP
6166 If defined, a C expression whose value is a string, including spacing,
6167 containing the assembler operation to identify the following data as
6168 uninitialized, writable small data.
6171 @defmac INIT_SECTION_ASM_OP
6172 If defined, a C expression whose value is a string, including spacing,
6173 containing the assembler operation to identify the following data as
6174 initialization code. If not defined, GCC will assume such a section does
6175 not exist. This section has no corresponding @code{init_section}
6176 variable; it is used entirely in runtime code.
6179 @defmac FINI_SECTION_ASM_OP
6180 If defined, a C expression whose value is a string, including spacing,
6181 containing the assembler operation to identify the following data as
6182 finalization code. If not defined, GCC will assume such a section does
6183 not exist. This section has no corresponding @code{fini_section}
6184 variable; it is used entirely in runtime code.
6187 @defmac INIT_ARRAY_SECTION_ASM_OP
6188 If defined, a C expression whose value is a string, including spacing,
6189 containing the assembler operation to identify the following data as
6190 part of the @code{.init_array} (or equivalent) section. If not
6191 defined, GCC will assume such a section does not exist. Do not define
6192 both this macro and @code{INIT_SECTION_ASM_OP}.
6195 @defmac FINI_ARRAY_SECTION_ASM_OP
6196 If defined, a C expression whose value is a string, including spacing,
6197 containing the assembler operation to identify the following data as
6198 part of the @code{.fini_array} (or equivalent) section. If not
6199 defined, GCC will assume such a section does not exist. Do not define
6200 both this macro and @code{FINI_SECTION_ASM_OP}.
6203 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
6204 If defined, an ASM statement that switches to a different section
6205 via @var{section_op}, calls @var{function}, and switches back to
6206 the text section. This is used in @file{crtstuff.c} if
6207 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
6208 to initialization and finalization functions from the init and fini
6209 sections. By default, this macro uses a simple function call. Some
6210 ports need hand-crafted assembly code to avoid dependencies on
6211 registers initialized in the function prologue or to ensure that
6212 constant pools don't end up too far way in the text section.
6215 @defmac FORCE_CODE_SECTION_ALIGN
6216 If defined, an ASM statement that aligns a code section to some
6217 arbitrary boundary. This is used to force all fragments of the
6218 @code{.init} and @code{.fini} sections to have to same alignment
6219 and thus prevent the linker from having to add any padding.
6222 @defmac JUMP_TABLES_IN_TEXT_SECTION
6223 Define this macro to be an expression with a nonzero value if jump
6224 tables (for @code{tablejump} insns) should be output in the text
6225 section, along with the assembler instructions. Otherwise, the
6226 readonly data section is used.
6228 This macro is irrelevant if there is no separate readonly data section.
6231 @deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void)
6232 Define this hook if you need to do something special to set up the
6233 @file{varasm.c} sections, or if your target has some special sections
6234 of its own that you need to create.
6236 GCC calls this hook after processing the command line, but before writing
6237 any assembly code, and before calling any of the section-returning hooks
6241 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
6242 Return the section into which @var{exp} should be placed. You can
6243 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
6244 some sort. @var{reloc} indicates whether the initial value of @var{exp}
6245 requires link-time relocations. Bit 0 is set when variable contains
6246 local relocations only, while bit 1 is set for global relocations.
6247 @var{align} is the constant alignment in bits.
6249 The default version of this function takes care of putting read-only
6250 variables in @code{readonly_data_section}.
6252 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
6255 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
6256 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
6257 for @code{FUNCTION_DECL}s as well as for variables and constants.
6259 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
6260 function has been determined to be likely to be called, and nonzero if
6261 it is unlikely to be called.
6264 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
6265 Build up a unique section name, expressed as a @code{STRING_CST} node,
6266 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
6267 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
6268 the initial value of @var{exp} requires link-time relocations.
6270 The default version of this function appends the symbol name to the
6271 ELF section name that would normally be used for the symbol. For
6272 example, the function @code{foo} would be placed in @code{.text.foo}.
6273 Whatever the actual target object format, this is often good enough.
6276 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
6277 Return the readonly data section associated with
6278 @samp{DECL_SECTION_NAME (@var{decl})}.
6279 The default version of this function selects @code{.gnu.linkonce.r.name} if
6280 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
6281 if function is in @code{.text.name}, and the normal readonly-data section
6285 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
6286 Return the section into which a constant @var{x}, of mode @var{mode},
6287 should be placed. You can assume that @var{x} is some kind of
6288 constant in RTL@. The argument @var{mode} is redundant except in the
6289 case of a @code{const_int} rtx. @var{align} is the constant alignment
6292 The default version of this function takes care of putting symbolic
6293 constants in @code{flag_pic} mode in @code{data_section} and everything
6294 else in @code{readonly_data_section}.
6297 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
6298 Define this hook if references to a symbol or a constant must be
6299 treated differently depending on something about the variable or
6300 function named by the symbol (such as what section it is in).
6302 The hook is executed immediately after rtl has been created for
6303 @var{decl}, which may be a variable or function declaration or
6304 an entry in the constant pool. In either case, @var{rtl} is the
6305 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
6306 in this hook; that field may not have been initialized yet.
6308 In the case of a constant, it is safe to assume that the rtl is
6309 a @code{mem} whose address is a @code{symbol_ref}. Most decls
6310 will also have this form, but that is not guaranteed. Global
6311 register variables, for instance, will have a @code{reg} for their
6312 rtl. (Normally the right thing to do with such unusual rtl is
6315 The @var{new_decl_p} argument will be true if this is the first time
6316 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
6317 be false for subsequent invocations, which will happen for duplicate
6318 declarations. Whether or not anything must be done for the duplicate
6319 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
6320 @var{new_decl_p} is always true when the hook is called for a constant.
6322 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
6323 The usual thing for this hook to do is to record flags in the
6324 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
6325 Historically, the name string was modified if it was necessary to
6326 encode more than one bit of information, but this practice is now
6327 discouraged; use @code{SYMBOL_REF_FLAGS}.
6329 The default definition of this hook, @code{default_encode_section_info}
6330 in @file{varasm.c}, sets a number of commonly-useful bits in
6331 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
6332 before overriding it.
6335 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
6336 Decode @var{name} and return the real name part, sans
6337 the characters that @code{TARGET_ENCODE_SECTION_INFO}
6341 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
6342 Returns true if @var{exp} should be placed into a ``small data'' section.
6343 The default version of this hook always returns false.
6346 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
6347 Contains the value true if the target places read-only
6348 ``small data'' into a separate section. The default value is false.
6351 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
6352 Returns true if @var{exp} names an object for which name resolution
6353 rules must resolve to the current ``module'' (dynamic shared library
6354 or executable image).
6356 The default version of this hook implements the name resolution rules
6357 for ELF, which has a looser model of global name binding than other
6358 currently supported object file formats.
6361 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
6362 Contains the value true if the target supports thread-local storage.
6363 The default value is false.
6368 @section Position Independent Code
6369 @cindex position independent code
6372 This section describes macros that help implement generation of position
6373 independent code. Simply defining these macros is not enough to
6374 generate valid PIC; you must also add support to the macros
6375 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
6376 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
6377 @samp{movsi} to do something appropriate when the source operand
6378 contains a symbolic address. You may also need to alter the handling of
6379 switch statements so that they use relative addresses.
6380 @c i rearranged the order of the macros above to try to force one of
6381 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
6383 @defmac PIC_OFFSET_TABLE_REGNUM
6384 The register number of the register used to address a table of static
6385 data addresses in memory. In some cases this register is defined by a
6386 processor's ``application binary interface'' (ABI)@. When this macro
6387 is defined, RTL is generated for this register once, as with the stack
6388 pointer and frame pointer registers. If this macro is not defined, it
6389 is up to the machine-dependent files to allocate such a register (if
6390 necessary). Note that this register must be fixed when in use (e.g.@:
6391 when @code{flag_pic} is true).
6394 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6395 Define this macro if the register defined by
6396 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
6397 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
6400 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
6401 A C expression that is nonzero if @var{x} is a legitimate immediate
6402 operand on the target machine when generating position independent code.
6403 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6404 check this. You can also assume @var{flag_pic} is true, so you need not
6405 check it either. You need not define this macro if all constants
6406 (including @code{SYMBOL_REF}) can be immediate operands when generating
6407 position independent code.
6410 @node Assembler Format
6411 @section Defining the Output Assembler Language
6413 This section describes macros whose principal purpose is to describe how
6414 to write instructions in assembler language---rather than what the
6418 * File Framework:: Structural information for the assembler file.
6419 * Data Output:: Output of constants (numbers, strings, addresses).
6420 * Uninitialized Data:: Output of uninitialized variables.
6421 * Label Output:: Output and generation of labels.
6422 * Initialization:: General principles of initialization
6423 and termination routines.
6424 * Macros for Initialization::
6425 Specific macros that control the handling of
6426 initialization and termination routines.
6427 * Instruction Output:: Output of actual instructions.
6428 * Dispatch Tables:: Output of jump tables.
6429 * Exception Region Output:: Output of exception region code.
6430 * Alignment Output:: Pseudo ops for alignment and skipping data.
6433 @node File Framework
6434 @subsection The Overall Framework of an Assembler File
6435 @cindex assembler format
6436 @cindex output of assembler code
6438 @c prevent bad page break with this line
6439 This describes the overall framework of an assembly file.
6441 @deftypefn {Target Hook} void TARGET_ASM_FILE_START ()
6442 @findex default_file_start
6443 Output to @code{asm_out_file} any text which the assembler expects to
6444 find at the beginning of a file. The default behavior is controlled
6445 by two flags, documented below. Unless your target's assembler is
6446 quite unusual, if you override the default, you should call
6447 @code{default_file_start} at some point in your target hook. This
6448 lets other target files rely on these variables.
6451 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
6452 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
6453 printed as the very first line in the assembly file, unless
6454 @option{-fverbose-asm} is in effect. (If that macro has been defined
6455 to the empty string, this variable has no effect.) With the normal
6456 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
6457 assembler that it need not bother stripping comments or extra
6458 whitespace from its input. This allows it to work a bit faster.
6460 The default is false. You should not set it to true unless you have
6461 verified that your port does not generate any extra whitespace or
6462 comments that will cause GAS to issue errors in NO_APP mode.
6465 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
6466 If this flag is true, @code{output_file_directive} will be called
6467 for the primary source file, immediately after printing
6468 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
6469 this to be done. The default is false.
6472 @deftypefn {Target Hook} void TARGET_ASM_FILE_END ()
6473 Output to @code{asm_out_file} any text which the assembler expects
6474 to find at the end of a file. The default is to output nothing.
6477 @deftypefun void file_end_indicate_exec_stack ()
6478 Some systems use a common convention, the @samp{.note.GNU-stack}
6479 special section, to indicate whether or not an object file relies on
6480 the stack being executable. If your system uses this convention, you
6481 should define @code{TARGET_ASM_FILE_END} to this function. If you
6482 need to do other things in that hook, have your hook function call
6486 @defmac ASM_COMMENT_START
6487 A C string constant describing how to begin a comment in the target
6488 assembler language. The compiler assumes that the comment will end at
6489 the end of the line.
6493 A C string constant for text to be output before each @code{asm}
6494 statement or group of consecutive ones. Normally this is
6495 @code{"#APP"}, which is a comment that has no effect on most
6496 assemblers but tells the GNU assembler that it must check the lines
6497 that follow for all valid assembler constructs.
6501 A C string constant for text to be output after each @code{asm}
6502 statement or group of consecutive ones. Normally this is
6503 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6504 time-saving assumptions that are valid for ordinary compiler output.
6507 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6508 A C statement to output COFF information or DWARF debugging information
6509 which indicates that filename @var{name} is the current source file to
6510 the stdio stream @var{stream}.
6512 This macro need not be defined if the standard form of output
6513 for the file format in use is appropriate.
6516 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6517 A C statement to output the string @var{string} to the stdio stream
6518 @var{stream}. If you do not call the function @code{output_quoted_string}
6519 in your config files, GCC will only call it to output filenames to
6520 the assembler source. So you can use it to canonicalize the format
6521 of the filename using this macro.
6524 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6525 A C statement to output something to the assembler file to handle a
6526 @samp{#ident} directive containing the text @var{string}. If this
6527 macro is not defined, nothing is output for a @samp{#ident} directive.
6530 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6531 Output assembly directives to switch to section @var{name}. The section
6532 should have attributes as specified by @var{flags}, which is a bit mask
6533 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6534 is nonzero, it contains an alignment in bytes to be used for the section,
6535 otherwise some target default should be used. Only targets that must
6536 specify an alignment within the section directive need pay attention to
6537 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6540 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6541 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6544 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
6545 @deftypefn {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
6546 This flag is true if we can create zeroed data by switching to a BSS
6547 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
6548 This is true on most ELF targets.
6551 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6552 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6553 based on a variable or function decl, a section name, and whether or not the
6554 declaration's initializer may contain runtime relocations. @var{decl} may be
6555 null, in which case read-write data should be assumed.
6557 The default version if this function handles choosing code vs data,
6558 read-only vs read-write data, and @code{flag_pic}. You should only
6559 need to override this if your target has special flags that might be
6560 set via @code{__attribute__}.
6565 @subsection Output of Data
6568 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6569 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6570 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6571 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6572 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6573 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6574 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6575 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6576 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6577 These hooks specify assembly directives for creating certain kinds
6578 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6579 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6580 aligned two-byte object, and so on. Any of the hooks may be
6581 @code{NULL}, indicating that no suitable directive is available.
6583 The compiler will print these strings at the start of a new line,
6584 followed immediately by the object's initial value. In most cases,
6585 the string should contain a tab, a pseudo-op, and then another tab.
6588 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6589 The @code{assemble_integer} function uses this hook to output an
6590 integer object. @var{x} is the object's value, @var{size} is its size
6591 in bytes and @var{aligned_p} indicates whether it is aligned. The
6592 function should return @code{true} if it was able to output the
6593 object. If it returns false, @code{assemble_integer} will try to
6594 split the object into smaller parts.
6596 The default implementation of this hook will use the
6597 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6598 when the relevant string is @code{NULL}.
6601 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6602 A C statement to recognize @var{rtx} patterns that
6603 @code{output_addr_const} can't deal with, and output assembly code to
6604 @var{stream} corresponding to the pattern @var{x}. This may be used to
6605 allow machine-dependent @code{UNSPEC}s to appear within constants.
6607 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6608 @code{goto fail}, so that a standard error message is printed. If it
6609 prints an error message itself, by calling, for example,
6610 @code{output_operand_lossage}, it may just complete normally.
6613 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6614 A C statement to output to the stdio stream @var{stream} an assembler
6615 instruction to assemble a string constant containing the @var{len}
6616 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6617 @code{char *} and @var{len} a C expression of type @code{int}.
6619 If the assembler has a @code{.ascii} pseudo-op as found in the
6620 Berkeley Unix assembler, do not define the macro
6621 @code{ASM_OUTPUT_ASCII}.
6624 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6625 A C statement to output word @var{n} of a function descriptor for
6626 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6627 is defined, and is otherwise unused.
6630 @defmac CONSTANT_POOL_BEFORE_FUNCTION
6631 You may define this macro as a C expression. You should define the
6632 expression to have a nonzero value if GCC should output the constant
6633 pool for a function before the code for the function, or a zero value if
6634 GCC should output the constant pool after the function. If you do
6635 not define this macro, the usual case, GCC will output the constant
6636 pool before the function.
6639 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
6640 A C statement to output assembler commands to define the start of the
6641 constant pool for a function. @var{funname} is a string giving
6642 the name of the function. Should the return type of the function
6643 be required, it can be obtained via @var{fundecl}. @var{size}
6644 is the size, in bytes, of the constant pool that will be written
6645 immediately after this call.
6647 If no constant-pool prefix is required, the usual case, this macro need
6651 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
6652 A C statement (with or without semicolon) to output a constant in the
6653 constant pool, if it needs special treatment. (This macro need not do
6654 anything for RTL expressions that can be output normally.)
6656 The argument @var{file} is the standard I/O stream to output the
6657 assembler code on. @var{x} is the RTL expression for the constant to
6658 output, and @var{mode} is the machine mode (in case @var{x} is a
6659 @samp{const_int}). @var{align} is the required alignment for the value
6660 @var{x}; you should output an assembler directive to force this much
6663 The argument @var{labelno} is a number to use in an internal label for
6664 the address of this pool entry. The definition of this macro is
6665 responsible for outputting the label definition at the proper place.
6666 Here is how to do this:
6669 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
6672 When you output a pool entry specially, you should end with a
6673 @code{goto} to the label @var{jumpto}. This will prevent the same pool
6674 entry from being output a second time in the usual manner.
6676 You need not define this macro if it would do nothing.
6679 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
6680 A C statement to output assembler commands to at the end of the constant
6681 pool for a function. @var{funname} is a string giving the name of the
6682 function. Should the return type of the function be required, you can
6683 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
6684 constant pool that GCC wrote immediately before this call.
6686 If no constant-pool epilogue is required, the usual case, you need not
6690 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
6691 Define this macro as a C expression which is nonzero if @var{C} is
6692 used as a logical line separator by the assembler.
6694 If you do not define this macro, the default is that only
6695 the character @samp{;} is treated as a logical line separator.
6698 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
6699 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
6700 These target hooks are C string constants, describing the syntax in the
6701 assembler for grouping arithmetic expressions. If not overridden, they
6702 default to normal parentheses, which is correct for most assemblers.
6705 These macros are provided by @file{real.h} for writing the definitions
6706 of @code{ASM_OUTPUT_DOUBLE} and the like:
6708 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
6709 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
6710 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
6711 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
6712 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
6713 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
6714 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
6715 target's floating point representation, and store its bit pattern in
6716 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
6717 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
6718 simple @code{long int}. For the others, it should be an array of
6719 @code{long int}. The number of elements in this array is determined
6720 by the size of the desired target floating point data type: 32 bits of
6721 it go in each @code{long int} array element. Each array element holds
6722 32 bits of the result, even if @code{long int} is wider than 32 bits
6723 on the host machine.
6725 The array element values are designed so that you can print them out
6726 using @code{fprintf} in the order they should appear in the target
6730 @node Uninitialized Data
6731 @subsection Output of Uninitialized Variables
6733 Each of the macros in this section is used to do the whole job of
6734 outputting a single uninitialized variable.
6736 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6737 A C statement (sans semicolon) to output to the stdio stream
6738 @var{stream} the assembler definition of a common-label named
6739 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6740 is the size rounded up to whatever alignment the caller wants.
6742 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6743 output the name itself; before and after that, output the additional
6744 assembler syntax for defining the name, and a newline.
6746 This macro controls how the assembler definitions of uninitialized
6747 common global variables are output.
6750 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
6751 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
6752 separate, explicit argument. If you define this macro, it is used in
6753 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
6754 handling the required alignment of the variable. The alignment is specified
6755 as the number of bits.
6758 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6759 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
6760 variable to be output, if there is one, or @code{NULL_TREE} if there
6761 is no corresponding variable. If you define this macro, GCC will use it
6762 in place of both @code{ASM_OUTPUT_COMMON} and
6763 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
6764 the variable's decl in order to chose what to output.
6767 @defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6768 A C statement (sans semicolon) to output to the stdio stream
6769 @var{stream} the assembler definition of uninitialized global @var{decl} named
6770 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6771 is the size rounded up to whatever alignment the caller wants.
6773 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
6774 defining this macro. If unable, use the expression
6775 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
6776 before and after that, output the additional assembler syntax for defining
6777 the name, and a newline.
6779 There are two ways of handling global BSS. One is to define either
6780 this macro or its aligned counterpart, @code{ASM_OUTPUT_ALIGNED_BSS}.
6781 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
6782 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
6783 You do not need to do both.
6785 Some languages do not have @code{common} data, and require a
6786 non-common form of global BSS in order to handle uninitialized globals
6787 efficiently. C++ is one example of this. However, if the target does
6788 not support global BSS, the front end may choose to make globals
6789 common in order to save space in the object file.
6792 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6793 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
6794 separate, explicit argument. If you define this macro, it is used in
6795 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
6796 handling the required alignment of the variable. The alignment is specified
6797 as the number of bits.
6799 Try to use function @code{asm_output_aligned_bss} defined in file
6800 @file{varasm.c} when defining this macro.
6803 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6804 A C statement (sans semicolon) to output to the stdio stream
6805 @var{stream} the assembler definition of a local-common-label named
6806 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6807 is the size rounded up to whatever alignment the caller wants.
6809 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6810 output the name itself; before and after that, output the additional
6811 assembler syntax for defining the name, and a newline.
6813 This macro controls how the assembler definitions of uninitialized
6814 static variables are output.
6817 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
6818 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
6819 separate, explicit argument. If you define this macro, it is used in
6820 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
6821 handling the required alignment of the variable. The alignment is specified
6822 as the number of bits.
6825 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6826 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
6827 variable to be output, if there is one, or @code{NULL_TREE} if there
6828 is no corresponding variable. If you define this macro, GCC will use it
6829 in place of both @code{ASM_OUTPUT_DECL} and
6830 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
6831 the variable's decl in order to chose what to output.
6835 @subsection Output and Generation of Labels
6837 @c prevent bad page break with this line
6838 This is about outputting labels.
6840 @findex assemble_name
6841 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
6842 A C statement (sans semicolon) to output to the stdio stream
6843 @var{stream} the assembler definition of a label named @var{name}.
6844 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6845 output the name itself; before and after that, output the additional
6846 assembler syntax for defining the name, and a newline. A default
6847 definition of this macro is provided which is correct for most systems.
6850 @findex assemble_name_raw
6851 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
6852 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
6853 to refer to a compiler-generated label. The default definition uses
6854 @code{assemble_name_raw}, which is like @code{assemble_name} except
6855 that it is more efficient.
6859 A C string containing the appropriate assembler directive to specify the
6860 size of a symbol, without any arguments. On systems that use ELF, the
6861 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
6862 systems, the default is not to define this macro.
6864 Define this macro only if it is correct to use the default definitions
6865 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
6866 for your system. If you need your own custom definitions of those
6867 macros, or if you do not need explicit symbol sizes at all, do not
6871 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
6872 A C statement (sans semicolon) to output to the stdio stream
6873 @var{stream} a directive telling the assembler that the size of the
6874 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
6875 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6879 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
6880 A C statement (sans semicolon) to output to the stdio stream
6881 @var{stream} a directive telling the assembler to calculate the size of
6882 the symbol @var{name} by subtracting its address from the current
6885 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6886 provided. The default assumes that the assembler recognizes a special
6887 @samp{.} symbol as referring to the current address, and can calculate
6888 the difference between this and another symbol. If your assembler does
6889 not recognize @samp{.} or cannot do calculations with it, you will need
6890 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
6894 A C string containing the appropriate assembler directive to specify the
6895 type of a symbol, without any arguments. On systems that use ELF, the
6896 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
6897 systems, the default is not to define this macro.
6899 Define this macro only if it is correct to use the default definition of
6900 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6901 custom definition of this macro, or if you do not need explicit symbol
6902 types at all, do not define this macro.
6905 @defmac TYPE_OPERAND_FMT
6906 A C string which specifies (using @code{printf} syntax) the format of
6907 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
6908 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
6909 the default is not to define this macro.
6911 Define this macro only if it is correct to use the default definition of
6912 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6913 custom definition of this macro, or if you do not need explicit symbol
6914 types at all, do not define this macro.
6917 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
6918 A C statement (sans semicolon) to output to the stdio stream
6919 @var{stream} a directive telling the assembler that the type of the
6920 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
6921 that string is always either @samp{"function"} or @samp{"object"}, but
6922 you should not count on this.
6924 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
6925 definition of this macro is provided.
6928 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
6929 A C statement (sans semicolon) to output to the stdio stream
6930 @var{stream} any text necessary for declaring the name @var{name} of a
6931 function which is being defined. This macro is responsible for
6932 outputting the label definition (perhaps using
6933 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
6934 @code{FUNCTION_DECL} tree node representing the function.
6936 If this macro is not defined, then the function name is defined in the
6937 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6939 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6943 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
6944 A C statement (sans semicolon) to output to the stdio stream
6945 @var{stream} any text necessary for declaring the size of a function
6946 which is being defined. The argument @var{name} is the name of the
6947 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
6948 representing the function.
6950 If this macro is not defined, then the function size is not defined.
6952 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
6956 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6957 A C statement (sans semicolon) to output to the stdio stream
6958 @var{stream} any text necessary for declaring the name @var{name} of an
6959 initialized variable which is being defined. This macro must output the
6960 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6961 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6963 If this macro is not defined, then the variable name is defined in the
6964 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6966 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
6967 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
6970 @defmac ASM_DECLARE_CONSTANT_NAME (@var{stream}, @var{name}, @var{exp}, @var{size})
6971 A C statement (sans semicolon) to output to the stdio stream
6972 @var{stream} any text necessary for declaring the name @var{name} of a
6973 constant which is being defined. This macro is responsible for
6974 outputting the label definition (perhaps using
6975 @code{ASM_OUTPUT_LABEL}). The argument @var{exp} is the
6976 value of the constant, and @var{size} is the size of the constant
6977 in bytes. @var{name} will be an internal label.
6979 If this macro is not defined, then the @var{name} is defined in the
6980 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6982 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6986 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6987 A C statement (sans semicolon) to output to the stdio stream
6988 @var{stream} any text necessary for claiming a register @var{regno}
6989 for a global variable @var{decl} with name @var{name}.
6991 If you don't define this macro, that is equivalent to defining it to do
6995 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6996 A C statement (sans semicolon) to finish up declaring a variable name
6997 once the compiler has processed its initializer fully and thus has had a
6998 chance to determine the size of an array when controlled by an
6999 initializer. This is used on systems where it's necessary to declare
7000 something about the size of the object.
7002 If you don't define this macro, that is equivalent to defining it to do
7005 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
7006 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
7009 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
7010 This target hook is a function to output to the stdio stream
7011 @var{stream} some commands that will make the label @var{name} global;
7012 that is, available for reference from other files.
7014 The default implementation relies on a proper definition of
7015 @code{GLOBAL_ASM_OP}.
7018 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
7019 A C statement (sans semicolon) to output to the stdio stream
7020 @var{stream} some commands that will make the label @var{name} weak;
7021 that is, available for reference from other files but only used if
7022 no other definition is available. Use the expression
7023 @code{assemble_name (@var{stream}, @var{name})} to output the name
7024 itself; before and after that, output the additional assembler syntax
7025 for making that name weak, and a newline.
7027 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
7028 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
7032 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
7033 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
7034 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
7035 or variable decl. If @var{value} is not @code{NULL}, this C statement
7036 should output to the stdio stream @var{stream} assembler code which
7037 defines (equates) the weak symbol @var{name} to have the value
7038 @var{value}. If @var{value} is @code{NULL}, it should output commands
7039 to make @var{name} weak.
7042 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
7043 Outputs a directive that enables @var{name} to be used to refer to
7044 symbol @var{value} with weak-symbol semantics. @code{decl} is the
7045 declaration of @code{name}.
7048 @defmac SUPPORTS_WEAK
7049 A C expression which evaluates to true if the target supports weak symbols.
7051 If you don't define this macro, @file{defaults.h} provides a default
7052 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
7053 is defined, the default definition is @samp{1}; otherwise, it is
7054 @samp{0}. Define this macro if you want to control weak symbol support
7055 with a compiler flag such as @option{-melf}.
7058 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
7059 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
7060 public symbol such that extra copies in multiple translation units will
7061 be discarded by the linker. Define this macro if your object file
7062 format provides support for this concept, such as the @samp{COMDAT}
7063 section flags in the Microsoft Windows PE/COFF format, and this support
7064 requires changes to @var{decl}, such as putting it in a separate section.
7067 @defmac SUPPORTS_ONE_ONLY
7068 A C expression which evaluates to true if the target supports one-only
7071 If you don't define this macro, @file{varasm.c} provides a default
7072 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
7073 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
7074 you want to control one-only symbol support with a compiler flag, or if
7075 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
7076 be emitted as one-only.
7079 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
7080 This target hook is a function to output to @var{asm_out_file} some
7081 commands that will make the symbol(s) associated with @var{decl} have
7082 hidden, protected or internal visibility as specified by @var{visibility}.
7085 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
7086 A C expression that evaluates to true if the target's linker expects
7087 that weak symbols do not appear in a static archive's table of contents.
7088 The default is @code{0}.
7090 Leaving weak symbols out of an archive's table of contents means that,
7091 if a symbol will only have a definition in one translation unit and
7092 will have undefined references from other translation units, that
7093 symbol should not be weak. Defining this macro to be nonzero will
7094 thus have the effect that certain symbols that would normally be weak
7095 (explicit template instantiations, and vtables for polymorphic classes
7096 with noninline key methods) will instead be nonweak.
7098 The C++ ABI requires this macro to be zero. Define this macro for
7099 targets where full C++ ABI compliance is impossible and where linker
7100 restrictions require weak symbols to be left out of a static archive's
7104 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
7105 A C statement (sans semicolon) to output to the stdio stream
7106 @var{stream} any text necessary for declaring the name of an external
7107 symbol named @var{name} which is referenced in this compilation but
7108 not defined. The value of @var{decl} is the tree node for the
7111 This macro need not be defined if it does not need to output anything.
7112 The GNU assembler and most Unix assemblers don't require anything.
7115 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
7116 This target hook is a function to output to @var{asm_out_file} an assembler
7117 pseudo-op to declare a library function name external. The name of the
7118 library function is given by @var{symref}, which is a @code{symbol_ref}.
7121 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (tree @var{decl})
7122 This target hook is a function to output to @var{asm_out_file} an assembler
7123 directive to annotate used symbol. Darwin target use .no_dead_code_strip
7127 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
7128 A C statement (sans semicolon) to output to the stdio stream
7129 @var{stream} a reference in assembler syntax to a label named
7130 @var{name}. This should add @samp{_} to the front of the name, if that
7131 is customary on your operating system, as it is in most Berkeley Unix
7132 systems. This macro is used in @code{assemble_name}.
7135 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
7136 A C statement (sans semicolon) to output a reference to
7137 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
7138 will be used to output the name of the symbol. This macro may be used
7139 to modify the way a symbol is referenced depending on information
7140 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
7143 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
7144 A C statement (sans semicolon) to output a reference to @var{buf}, the
7145 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
7146 @code{assemble_name} will be used to output the name of the symbol.
7147 This macro is not used by @code{output_asm_label}, or the @code{%l}
7148 specifier that calls it; the intention is that this macro should be set
7149 when it is necessary to output a label differently when its address is
7153 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
7154 A function to output to the stdio stream @var{stream} a label whose
7155 name is made from the string @var{prefix} and the number @var{labelno}.
7157 It is absolutely essential that these labels be distinct from the labels
7158 used for user-level functions and variables. Otherwise, certain programs
7159 will have name conflicts with internal labels.
7161 It is desirable to exclude internal labels from the symbol table of the
7162 object file. Most assemblers have a naming convention for labels that
7163 should be excluded; on many systems, the letter @samp{L} at the
7164 beginning of a label has this effect. You should find out what
7165 convention your system uses, and follow it.
7167 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
7170 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
7171 A C statement to output to the stdio stream @var{stream} a debug info
7172 label whose name is made from the string @var{prefix} and the number
7173 @var{num}. This is useful for VLIW targets, where debug info labels
7174 may need to be treated differently than branch target labels. On some
7175 systems, branch target labels must be at the beginning of instruction
7176 bundles, but debug info labels can occur in the middle of instruction
7179 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
7183 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
7184 A C statement to store into the string @var{string} a label whose name
7185 is made from the string @var{prefix} and the number @var{num}.
7187 This string, when output subsequently by @code{assemble_name}, should
7188 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
7189 with the same @var{prefix} and @var{num}.
7191 If the string begins with @samp{*}, then @code{assemble_name} will
7192 output the rest of the string unchanged. It is often convenient for
7193 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
7194 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
7195 to output the string, and may change it. (Of course,
7196 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
7197 you should know what it does on your machine.)
7200 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
7201 A C expression to assign to @var{outvar} (which is a variable of type
7202 @code{char *}) a newly allocated string made from the string
7203 @var{name} and the number @var{number}, with some suitable punctuation
7204 added. Use @code{alloca} to get space for the string.
7206 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
7207 produce an assembler label for an internal static variable whose name is
7208 @var{name}. Therefore, the string must be such as to result in valid
7209 assembler code. The argument @var{number} is different each time this
7210 macro is executed; it prevents conflicts between similarly-named
7211 internal static variables in different scopes.
7213 Ideally this string should not be a valid C identifier, to prevent any
7214 conflict with the user's own symbols. Most assemblers allow periods
7215 or percent signs in assembler symbols; putting at least one of these
7216 between the name and the number will suffice.
7218 If this macro is not defined, a default definition will be provided
7219 which is correct for most systems.
7222 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
7223 A C statement to output to the stdio stream @var{stream} assembler code
7224 which defines (equates) the symbol @var{name} to have the value @var{value}.
7227 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7228 correct for most systems.
7231 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
7232 A C statement to output to the stdio stream @var{stream} assembler code
7233 which defines (equates) the symbol whose tree node is @var{decl_of_name}
7234 to have the value of the tree node @var{decl_of_value}. This macro will
7235 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
7236 the tree nodes are available.
7239 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7240 correct for most systems.
7243 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
7244 A C statement that evaluates to true if the assembler code which defines
7245 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
7246 of the tree node @var{decl_of_value} should be emitted near the end of the
7247 current compilation unit. The default is to not defer output of defines.
7248 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
7249 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
7252 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
7253 A C statement to output to the stdio stream @var{stream} assembler code
7254 which defines (equates) the weak symbol @var{name} to have the value
7255 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
7256 an undefined weak symbol.
7258 Define this macro if the target only supports weak aliases; define
7259 @code{ASM_OUTPUT_DEF} instead if possible.
7262 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
7263 Define this macro to override the default assembler names used for
7264 Objective-C methods.
7266 The default name is a unique method number followed by the name of the
7267 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
7268 the category is also included in the assembler name (e.g.@:
7271 These names are safe on most systems, but make debugging difficult since
7272 the method's selector is not present in the name. Therefore, particular
7273 systems define other ways of computing names.
7275 @var{buf} is an expression of type @code{char *} which gives you a
7276 buffer in which to store the name; its length is as long as
7277 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
7278 50 characters extra.
7280 The argument @var{is_inst} specifies whether the method is an instance
7281 method or a class method; @var{class_name} is the name of the class;
7282 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
7283 in a category); and @var{sel_name} is the name of the selector.
7285 On systems where the assembler can handle quoted names, you can use this
7286 macro to provide more human-readable names.
7289 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
7290 A C statement (sans semicolon) to output to the stdio stream
7291 @var{stream} commands to declare that the label @var{name} is an
7292 Objective-C class reference. This is only needed for targets whose
7293 linkers have special support for NeXT-style runtimes.
7296 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
7297 A C statement (sans semicolon) to output to the stdio stream
7298 @var{stream} commands to declare that the label @var{name} is an
7299 unresolved Objective-C class reference. This is only needed for targets
7300 whose linkers have special support for NeXT-style runtimes.
7303 @node Initialization
7304 @subsection How Initialization Functions Are Handled
7305 @cindex initialization routines
7306 @cindex termination routines
7307 @cindex constructors, output of
7308 @cindex destructors, output of
7310 The compiled code for certain languages includes @dfn{constructors}
7311 (also called @dfn{initialization routines})---functions to initialize
7312 data in the program when the program is started. These functions need
7313 to be called before the program is ``started''---that is to say, before
7314 @code{main} is called.
7316 Compiling some languages generates @dfn{destructors} (also called
7317 @dfn{termination routines}) that should be called when the program
7320 To make the initialization and termination functions work, the compiler
7321 must output something in the assembler code to cause those functions to
7322 be called at the appropriate time. When you port the compiler to a new
7323 system, you need to specify how to do this.
7325 There are two major ways that GCC currently supports the execution of
7326 initialization and termination functions. Each way has two variants.
7327 Much of the structure is common to all four variations.
7329 @findex __CTOR_LIST__
7330 @findex __DTOR_LIST__
7331 The linker must build two lists of these functions---a list of
7332 initialization functions, called @code{__CTOR_LIST__}, and a list of
7333 termination functions, called @code{__DTOR_LIST__}.
7335 Each list always begins with an ignored function pointer (which may hold
7336 0, @minus{}1, or a count of the function pointers after it, depending on
7337 the environment). This is followed by a series of zero or more function
7338 pointers to constructors (or destructors), followed by a function
7339 pointer containing zero.
7341 Depending on the operating system and its executable file format, either
7342 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
7343 time and exit time. Constructors are called in reverse order of the
7344 list; destructors in forward order.
7346 The best way to handle static constructors works only for object file
7347 formats which provide arbitrarily-named sections. A section is set
7348 aside for a list of constructors, and another for a list of destructors.
7349 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
7350 object file that defines an initialization function also puts a word in
7351 the constructor section to point to that function. The linker
7352 accumulates all these words into one contiguous @samp{.ctors} section.
7353 Termination functions are handled similarly.
7355 This method will be chosen as the default by @file{target-def.h} if
7356 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
7357 support arbitrary sections, but does support special designated
7358 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
7359 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
7361 When arbitrary sections are available, there are two variants, depending
7362 upon how the code in @file{crtstuff.c} is called. On systems that
7363 support a @dfn{.init} section which is executed at program startup,
7364 parts of @file{crtstuff.c} are compiled into that section. The
7365 program is linked by the @command{gcc} driver like this:
7368 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
7371 The prologue of a function (@code{__init}) appears in the @code{.init}
7372 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
7373 for the function @code{__fini} in the @dfn{.fini} section. Normally these
7374 files are provided by the operating system or by the GNU C library, but
7375 are provided by GCC for a few targets.
7377 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
7378 compiled from @file{crtstuff.c}. They contain, among other things, code
7379 fragments within the @code{.init} and @code{.fini} sections that branch
7380 to routines in the @code{.text} section. The linker will pull all parts
7381 of a section together, which results in a complete @code{__init} function
7382 that invokes the routines we need at startup.
7384 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
7387 If no init section is available, when GCC compiles any function called
7388 @code{main} (or more accurately, any function designated as a program
7389 entry point by the language front end calling @code{expand_main_function}),
7390 it inserts a procedure call to @code{__main} as the first executable code
7391 after the function prologue. The @code{__main} function is defined
7392 in @file{libgcc2.c} and runs the global constructors.
7394 In file formats that don't support arbitrary sections, there are again
7395 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
7396 and an `a.out' format must be used. In this case,
7397 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
7398 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
7399 and with the address of the void function containing the initialization
7400 code as its value. The GNU linker recognizes this as a request to add
7401 the value to a @dfn{set}; the values are accumulated, and are eventually
7402 placed in the executable as a vector in the format described above, with
7403 a leading (ignored) count and a trailing zero element.
7404 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
7405 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
7406 the compilation of @code{main} to call @code{__main} as above, starting
7407 the initialization process.
7409 The last variant uses neither arbitrary sections nor the GNU linker.
7410 This is preferable when you want to do dynamic linking and when using
7411 file formats which the GNU linker does not support, such as `ECOFF'@. In
7412 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
7413 termination functions are recognized simply by their names. This requires
7414 an extra program in the linkage step, called @command{collect2}. This program
7415 pretends to be the linker, for use with GCC; it does its job by running
7416 the ordinary linker, but also arranges to include the vectors of
7417 initialization and termination functions. These functions are called
7418 via @code{__main} as described above. In order to use this method,
7419 @code{use_collect2} must be defined in the target in @file{config.gcc}.
7422 The following section describes the specific macros that control and
7423 customize the handling of initialization and termination functions.
7426 @node Macros for Initialization
7427 @subsection Macros Controlling Initialization Routines
7429 Here are the macros that control how the compiler handles initialization
7430 and termination functions:
7432 @defmac INIT_SECTION_ASM_OP
7433 If defined, a C string constant, including spacing, for the assembler
7434 operation to identify the following data as initialization code. If not
7435 defined, GCC will assume such a section does not exist. When you are
7436 using special sections for initialization and termination functions, this
7437 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7438 run the initialization functions.
7441 @defmac HAS_INIT_SECTION
7442 If defined, @code{main} will not call @code{__main} as described above.
7443 This macro should be defined for systems that control start-up code
7444 on a symbol-by-symbol basis, such as OSF/1, and should not
7445 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7448 @defmac LD_INIT_SWITCH
7449 If defined, a C string constant for a switch that tells the linker that
7450 the following symbol is an initialization routine.
7453 @defmac LD_FINI_SWITCH
7454 If defined, a C string constant for a switch that tells the linker that
7455 the following symbol is a finalization routine.
7458 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7459 If defined, a C statement that will write a function that can be
7460 automatically called when a shared library is loaded. The function
7461 should call @var{func}, which takes no arguments. If not defined, and
7462 the object format requires an explicit initialization function, then a
7463 function called @code{_GLOBAL__DI} will be generated.
7465 This function and the following one are used by collect2 when linking a
7466 shared library that needs constructors or destructors, or has DWARF2
7467 exception tables embedded in the code.
7470 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7471 If defined, a C statement that will write a function that can be
7472 automatically called when a shared library is unloaded. The function
7473 should call @var{func}, which takes no arguments. If not defined, and
7474 the object format requires an explicit finalization function, then a
7475 function called @code{_GLOBAL__DD} will be generated.
7478 @defmac INVOKE__main
7479 If defined, @code{main} will call @code{__main} despite the presence of
7480 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7481 where the init section is not actually run automatically, but is still
7482 useful for collecting the lists of constructors and destructors.
7485 @defmac SUPPORTS_INIT_PRIORITY
7486 If nonzero, the C++ @code{init_priority} attribute is supported and the
7487 compiler should emit instructions to control the order of initialization
7488 of objects. If zero, the compiler will issue an error message upon
7489 encountering an @code{init_priority} attribute.
7492 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7493 This value is true if the target supports some ``native'' method of
7494 collecting constructors and destructors to be run at startup and exit.
7495 It is false if we must use @command{collect2}.
7498 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7499 If defined, a function that outputs assembler code to arrange to call
7500 the function referenced by @var{symbol} at initialization time.
7502 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7503 no arguments and with no return value. If the target supports initialization
7504 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7505 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7507 If this macro is not defined by the target, a suitable default will
7508 be chosen if (1) the target supports arbitrary section names, (2) the
7509 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7513 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7514 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7515 functions rather than initialization functions.
7518 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7519 generated for the generated object file will have static linkage.
7521 If your system uses @command{collect2} as the means of processing
7522 constructors, then that program normally uses @command{nm} to scan
7523 an object file for constructor functions to be called.
7525 On certain kinds of systems, you can define this macro to make
7526 @command{collect2} work faster (and, in some cases, make it work at all):
7528 @defmac OBJECT_FORMAT_COFF
7529 Define this macro if the system uses COFF (Common Object File Format)
7530 object files, so that @command{collect2} can assume this format and scan
7531 object files directly for dynamic constructor/destructor functions.
7533 This macro is effective only in a native compiler; @command{collect2} as
7534 part of a cross compiler always uses @command{nm} for the target machine.
7537 @defmac REAL_NM_FILE_NAME
7538 Define this macro as a C string constant containing the file name to use
7539 to execute @command{nm}. The default is to search the path normally for
7542 If your system supports shared libraries and has a program to list the
7543 dynamic dependencies of a given library or executable, you can define
7544 these macros to enable support for running initialization and
7545 termination functions in shared libraries:
7549 Define this macro to a C string constant containing the name of the program
7550 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7553 @defmac PARSE_LDD_OUTPUT (@var{ptr})
7554 Define this macro to be C code that extracts filenames from the output
7555 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7556 of type @code{char *} that points to the beginning of a line of output
7557 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7558 code must advance @var{ptr} to the beginning of the filename on that
7559 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7562 @node Instruction Output
7563 @subsection Output of Assembler Instructions
7565 @c prevent bad page break with this line
7566 This describes assembler instruction output.
7568 @defmac REGISTER_NAMES
7569 A C initializer containing the assembler's names for the machine
7570 registers, each one as a C string constant. This is what translates
7571 register numbers in the compiler into assembler language.
7574 @defmac ADDITIONAL_REGISTER_NAMES
7575 If defined, a C initializer for an array of structures containing a name
7576 and a register number. This macro defines additional names for hard
7577 registers, thus allowing the @code{asm} option in declarations to refer
7578 to registers using alternate names.
7581 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7582 Define this macro if you are using an unusual assembler that
7583 requires different names for the machine instructions.
7585 The definition is a C statement or statements which output an
7586 assembler instruction opcode to the stdio stream @var{stream}. The
7587 macro-operand @var{ptr} is a variable of type @code{char *} which
7588 points to the opcode name in its ``internal'' form---the form that is
7589 written in the machine description. The definition should output the
7590 opcode name to @var{stream}, performing any translation you desire, and
7591 increment the variable @var{ptr} to point at the end of the opcode
7592 so that it will not be output twice.
7594 In fact, your macro definition may process less than the entire opcode
7595 name, or more than the opcode name; but if you want to process text
7596 that includes @samp{%}-sequences to substitute operands, you must take
7597 care of the substitution yourself. Just be sure to increment
7598 @var{ptr} over whatever text should not be output normally.
7600 @findex recog_data.operand
7601 If you need to look at the operand values, they can be found as the
7602 elements of @code{recog_data.operand}.
7604 If the macro definition does nothing, the instruction is output
7608 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7609 If defined, a C statement to be executed just prior to the output of
7610 assembler code for @var{insn}, to modify the extracted operands so
7611 they will be output differently.
7613 Here the argument @var{opvec} is the vector containing the operands
7614 extracted from @var{insn}, and @var{noperands} is the number of
7615 elements of the vector which contain meaningful data for this insn.
7616 The contents of this vector are what will be used to convert the insn
7617 template into assembler code, so you can change the assembler output
7618 by changing the contents of the vector.
7620 This macro is useful when various assembler syntaxes share a single
7621 file of instruction patterns; by defining this macro differently, you
7622 can cause a large class of instructions to be output differently (such
7623 as with rearranged operands). Naturally, variations in assembler
7624 syntax affecting individual insn patterns ought to be handled by
7625 writing conditional output routines in those patterns.
7627 If this macro is not defined, it is equivalent to a null statement.
7630 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
7631 A C compound statement to output to stdio stream @var{stream} the
7632 assembler syntax for an instruction operand @var{x}. @var{x} is an
7635 @var{code} is a value that can be used to specify one of several ways
7636 of printing the operand. It is used when identical operands must be
7637 printed differently depending on the context. @var{code} comes from
7638 the @samp{%} specification that was used to request printing of the
7639 operand. If the specification was just @samp{%@var{digit}} then
7640 @var{code} is 0; if the specification was @samp{%@var{ltr}
7641 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
7644 If @var{x} is a register, this macro should print the register's name.
7645 The names can be found in an array @code{reg_names} whose type is
7646 @code{char *[]}. @code{reg_names} is initialized from
7647 @code{REGISTER_NAMES}.
7649 When the machine description has a specification @samp{%@var{punct}}
7650 (a @samp{%} followed by a punctuation character), this macro is called
7651 with a null pointer for @var{x} and the punctuation character for
7655 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
7656 A C expression which evaluates to true if @var{code} is a valid
7657 punctuation character for use in the @code{PRINT_OPERAND} macro. If
7658 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
7659 punctuation characters (except for the standard one, @samp{%}) are used
7663 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
7664 A C compound statement to output to stdio stream @var{stream} the
7665 assembler syntax for an instruction operand that is a memory reference
7666 whose address is @var{x}. @var{x} is an RTL expression.
7668 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
7669 On some machines, the syntax for a symbolic address depends on the
7670 section that the address refers to. On these machines, define the hook
7671 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
7672 @code{symbol_ref}, and then check for it here. @xref{Assembler
7676 @findex dbr_sequence_length
7677 @defmac DBR_OUTPUT_SEQEND (@var{file})
7678 A C statement, to be executed after all slot-filler instructions have
7679 been output. If necessary, call @code{dbr_sequence_length} to
7680 determine the number of slots filled in a sequence (zero if not
7681 currently outputting a sequence), to decide how many no-ops to output,
7684 Don't define this macro if it has nothing to do, but it is helpful in
7685 reading assembly output if the extent of the delay sequence is made
7686 explicit (e.g.@: with white space).
7689 @findex final_sequence
7690 Note that output routines for instructions with delay slots must be
7691 prepared to deal with not being output as part of a sequence
7692 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
7693 found.) The variable @code{final_sequence} is null when not
7694 processing a sequence, otherwise it contains the @code{sequence} rtx
7698 @defmac REGISTER_PREFIX
7699 @defmacx LOCAL_LABEL_PREFIX
7700 @defmacx USER_LABEL_PREFIX
7701 @defmacx IMMEDIATE_PREFIX
7702 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
7703 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
7704 @file{final.c}). These are useful when a single @file{md} file must
7705 support multiple assembler formats. In that case, the various @file{tm.h}
7706 files can define these macros differently.
7709 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
7710 If defined this macro should expand to a series of @code{case}
7711 statements which will be parsed inside the @code{switch} statement of
7712 the @code{asm_fprintf} function. This allows targets to define extra
7713 printf formats which may useful when generating their assembler
7714 statements. Note that uppercase letters are reserved for future
7715 generic extensions to asm_fprintf, and so are not available to target
7716 specific code. The output file is given by the parameter @var{file}.
7717 The varargs input pointer is @var{argptr} and the rest of the format
7718 string, starting the character after the one that is being switched
7719 upon, is pointed to by @var{format}.
7722 @defmac ASSEMBLER_DIALECT
7723 If your target supports multiple dialects of assembler language (such as
7724 different opcodes), define this macro as a C expression that gives the
7725 numeric index of the assembler language dialect to use, with zero as the
7728 If this macro is defined, you may use constructs of the form
7730 @samp{@{option0|option1|option2@dots{}@}}
7733 in the output templates of patterns (@pxref{Output Template}) or in the
7734 first argument of @code{asm_fprintf}. This construct outputs
7735 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
7736 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
7737 within these strings retain their usual meaning. If there are fewer
7738 alternatives within the braces than the value of
7739 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
7741 If you do not define this macro, the characters @samp{@{}, @samp{|} and
7742 @samp{@}} do not have any special meaning when used in templates or
7743 operands to @code{asm_fprintf}.
7745 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
7746 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
7747 the variations in assembler language syntax with that mechanism. Define
7748 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
7749 if the syntax variant are larger and involve such things as different
7750 opcodes or operand order.
7753 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
7754 A C expression to output to @var{stream} some assembler code
7755 which will push hard register number @var{regno} onto the stack.
7756 The code need not be optimal, since this macro is used only when
7760 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
7761 A C expression to output to @var{stream} some assembler code
7762 which will pop hard register number @var{regno} off of the stack.
7763 The code need not be optimal, since this macro is used only when
7767 @node Dispatch Tables
7768 @subsection Output of Dispatch Tables
7770 @c prevent bad page break with this line
7771 This concerns dispatch tables.
7773 @cindex dispatch table
7774 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
7775 A C statement to output to the stdio stream @var{stream} an assembler
7776 pseudo-instruction to generate a difference between two labels.
7777 @var{value} and @var{rel} are the numbers of two internal labels. The
7778 definitions of these labels are output using
7779 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
7780 way here. For example,
7783 fprintf (@var{stream}, "\t.word L%d-L%d\n",
7784 @var{value}, @var{rel})
7787 You must provide this macro on machines where the addresses in a
7788 dispatch table are relative to the table's own address. If defined, GCC
7789 will also use this macro on all machines when producing PIC@.
7790 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
7791 mode and flags can be read.
7794 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
7795 This macro should be provided on machines where the addresses
7796 in a dispatch table are absolute.
7798 The definition should be a C statement to output to the stdio stream
7799 @var{stream} an assembler pseudo-instruction to generate a reference to
7800 a label. @var{value} is the number of an internal label whose
7801 definition is output using @code{(*targetm.asm_out.internal_label)}.
7805 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
7809 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
7810 Define this if the label before a jump-table needs to be output
7811 specially. The first three arguments are the same as for
7812 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
7813 jump-table which follows (a @code{jump_insn} containing an
7814 @code{addr_vec} or @code{addr_diff_vec}).
7816 This feature is used on system V to output a @code{swbeg} statement
7819 If this macro is not defined, these labels are output with
7820 @code{(*targetm.asm_out.internal_label)}.
7823 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
7824 Define this if something special must be output at the end of a
7825 jump-table. The definition should be a C statement to be executed
7826 after the assembler code for the table is written. It should write
7827 the appropriate code to stdio stream @var{stream}. The argument
7828 @var{table} is the jump-table insn, and @var{num} is the label-number
7829 of the preceding label.
7831 If this macro is not defined, nothing special is output at the end of
7835 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (@var{stream}, @var{decl}, @var{for_eh}, @var{empty})
7836 This target hook emits a label at the beginning of each FDE@. It
7837 should be defined on targets where FDEs need special labels, and it
7838 should write the appropriate label, for the FDE associated with the
7839 function declaration @var{decl}, to the stdio stream @var{stream}.
7840 The third argument, @var{for_eh}, is a boolean: true if this is for an
7841 exception table. The fourth argument, @var{empty}, is a boolean:
7842 true if this is a placeholder label for an omitted FDE@.
7844 The default is that FDEs are not given nonlocal labels.
7847 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (@var{stream})
7848 This target hook emits a label at the beginning of the exception table.
7849 It should be defined on targets where it is desirable for the table
7850 to be broken up according to function.
7852 The default is that no label is emitted.
7855 @deftypefn {Target Hook} void TARGET_UNWIND_EMIT (FILE * @var{stream}, rtx @var{insn})
7856 This target hook emits and assembly directives required to unwind the
7857 given instruction. This is only used when TARGET_UNWIND_INFO is set.
7860 @node Exception Region Output
7861 @subsection Assembler Commands for Exception Regions
7863 @c prevent bad page break with this line
7865 This describes commands marking the start and the end of an exception
7868 @defmac EH_FRAME_SECTION_NAME
7869 If defined, a C string constant for the name of the section containing
7870 exception handling frame unwind information. If not defined, GCC will
7871 provide a default definition if the target supports named sections.
7872 @file{crtstuff.c} uses this macro to switch to the appropriate section.
7874 You should define this symbol if your target supports DWARF 2 frame
7875 unwind information and the default definition does not work.
7878 @defmac EH_FRAME_IN_DATA_SECTION
7879 If defined, DWARF 2 frame unwind information will be placed in the
7880 data section even though the target supports named sections. This
7881 might be necessary, for instance, if the system linker does garbage
7882 collection and sections cannot be marked as not to be collected.
7884 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
7888 @defmac EH_TABLES_CAN_BE_READ_ONLY
7889 Define this macro to 1 if your target is such that no frame unwind
7890 information encoding used with non-PIC code will ever require a
7891 runtime relocation, but the linker may not support merging read-only
7892 and read-write sections into a single read-write section.
7895 @defmac MASK_RETURN_ADDR
7896 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
7897 that it does not contain any extraneous set bits in it.
7900 @defmac DWARF2_UNWIND_INFO
7901 Define this macro to 0 if your target supports DWARF 2 frame unwind
7902 information, but it does not yet work with exception handling.
7903 Otherwise, if your target supports this information (if it defines
7904 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
7905 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
7908 If @code{TARGET_UNWIND_INFO} is defined, the target specific unwinder
7909 will be used in all cases. Defining this macro will enable the generation
7910 of DWARF 2 frame debugging information.
7912 If @code{TARGET_UNWIND_INFO} is not defined, and this macro is defined to 1,
7913 the DWARF 2 unwinder will be the default exception handling mechanism;
7914 otherwise, @code{setjmp}/@code{longjmp} will be used by default.
7917 @defmac TARGET_UNWIND_INFO
7918 Define this macro if your target has ABI specified unwind tables. Usually
7919 these will be output by @code{TARGET_UNWIND_EMIT}.
7922 @deftypevar {Target Hook} bool TARGET_UNWID_TABLES_DEFAULT
7923 This variable should be set to @code{true} if the target ABI requires unwinding
7924 tables even when exceptions are not used.
7927 @defmac MUST_USE_SJLJ_EXCEPTIONS
7928 This macro need only be defined if @code{DWARF2_UNWIND_INFO} is
7929 runtime-variable. In that case, @file{except.h} cannot correctly
7930 determine the corresponding definition of
7931 @code{MUST_USE_SJLJ_EXCEPTIONS}, so the target must provide it directly.
7934 @defmac DWARF_CIE_DATA_ALIGNMENT
7935 This macro need only be defined if the target might save registers in the
7936 function prologue at an offset to the stack pointer that is not aligned to
7937 @code{UNITS_PER_WORD}. The definition should be the negative minimum
7938 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
7939 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
7940 the target supports DWARF 2 frame unwind information.
7943 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
7944 Contains the value true if the target should add a zero word onto the
7945 end of a Dwarf-2 frame info section when used for exception handling.
7946 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
7950 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
7951 Given a register, this hook should return a parallel of registers to
7952 represent where to find the register pieces. Define this hook if the
7953 register and its mode are represented in Dwarf in non-contiguous
7954 locations, or if the register should be represented in more than one
7955 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
7956 If not defined, the default is to return @code{NULL_RTX}.
7959 @deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym})
7960 This hook is used to output a reference from a frame unwinding table to
7961 the type_info object identified by @var{sym}. It should return @code{true}
7962 if the reference was output. Returning @code{false} will cause the
7963 reference to be output using the normal Dwarf2 routines.
7966 @deftypefn {Target Hook} bool TARGET_ARM_EABI_UNWINDER
7967 This hook should be set to @code{true} on targets that use an ARM EABI
7968 based unwinding library, and @code{false} on other targets. This effects
7969 the format of unwinding tables, and how the unwinder in entered after
7970 running a cleanup. The default is @code{false}.
7973 @node Alignment Output
7974 @subsection Assembler Commands for Alignment
7976 @c prevent bad page break with this line
7977 This describes commands for alignment.
7979 @defmac JUMP_ALIGN (@var{label})
7980 The alignment (log base 2) to put in front of @var{label}, which is
7981 a common destination of jumps and has no fallthru incoming edge.
7983 This macro need not be defined if you don't want any special alignment
7984 to be done at such a time. Most machine descriptions do not currently
7987 Unless it's necessary to inspect the @var{label} parameter, it is better
7988 to set the variable @var{align_jumps} in the target's
7989 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7990 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
7993 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
7994 The alignment (log base 2) to put in front of @var{label}, which follows
7997 This macro need not be defined if you don't want any special alignment
7998 to be done at such a time. Most machine descriptions do not currently
8002 @defmac LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
8003 The maximum number of bytes to skip when applying
8004 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
8005 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8008 @defmac LOOP_ALIGN (@var{label})
8009 The alignment (log base 2) to put in front of @var{label}, which follows
8010 a @code{NOTE_INSN_LOOP_BEG} note.
8012 This macro need not be defined if you don't want any special alignment
8013 to be done at such a time. Most machine descriptions do not currently
8016 Unless it's necessary to inspect the @var{label} parameter, it is better
8017 to set the variable @code{align_loops} in the target's
8018 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8019 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
8022 @defmac LOOP_ALIGN_MAX_SKIP
8023 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
8024 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8027 @defmac LABEL_ALIGN (@var{label})
8028 The alignment (log base 2) to put in front of @var{label}.
8029 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
8030 the maximum of the specified values is used.
8032 Unless it's necessary to inspect the @var{label} parameter, it is better
8033 to set the variable @code{align_labels} in the target's
8034 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8035 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
8038 @defmac LABEL_ALIGN_MAX_SKIP
8039 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
8040 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8043 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
8044 A C statement to output to the stdio stream @var{stream} an assembler
8045 instruction to advance the location counter by @var{nbytes} bytes.
8046 Those bytes should be zero when loaded. @var{nbytes} will be a C
8047 expression of type @code{int}.
8050 @defmac ASM_NO_SKIP_IN_TEXT
8051 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
8052 text section because it fails to put zeros in the bytes that are skipped.
8053 This is true on many Unix systems, where the pseudo--op to skip bytes
8054 produces no-op instructions rather than zeros when used in the text
8058 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
8059 A C statement to output to the stdio stream @var{stream} an assembler
8060 command to advance the location counter to a multiple of 2 to the
8061 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
8064 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
8065 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
8066 for padding, if necessary.
8069 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
8070 A C statement to output to the stdio stream @var{stream} an assembler
8071 command to advance the location counter to a multiple of 2 to the
8072 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
8073 satisfy the alignment request. @var{power} and @var{max_skip} will be
8074 a C expression of type @code{int}.
8078 @node Debugging Info
8079 @section Controlling Debugging Information Format
8081 @c prevent bad page break with this line
8082 This describes how to specify debugging information.
8085 * All Debuggers:: Macros that affect all debugging formats uniformly.
8086 * DBX Options:: Macros enabling specific options in DBX format.
8087 * DBX Hooks:: Hook macros for varying DBX format.
8088 * File Names and DBX:: Macros controlling output of file names in DBX format.
8089 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
8090 * VMS Debug:: Macros for VMS debug format.
8094 @subsection Macros Affecting All Debugging Formats
8096 @c prevent bad page break with this line
8097 These macros affect all debugging formats.
8099 @defmac DBX_REGISTER_NUMBER (@var{regno})
8100 A C expression that returns the DBX register number for the compiler
8101 register number @var{regno}. In the default macro provided, the value
8102 of this expression will be @var{regno} itself. But sometimes there are
8103 some registers that the compiler knows about and DBX does not, or vice
8104 versa. In such cases, some register may need to have one number in the
8105 compiler and another for DBX@.
8107 If two registers have consecutive numbers inside GCC, and they can be
8108 used as a pair to hold a multiword value, then they @emph{must} have
8109 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
8110 Otherwise, debuggers will be unable to access such a pair, because they
8111 expect register pairs to be consecutive in their own numbering scheme.
8113 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
8114 does not preserve register pairs, then what you must do instead is
8115 redefine the actual register numbering scheme.
8118 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
8119 A C expression that returns the integer offset value for an automatic
8120 variable having address @var{x} (an RTL expression). The default
8121 computation assumes that @var{x} is based on the frame-pointer and
8122 gives the offset from the frame-pointer. This is required for targets
8123 that produce debugging output for DBX or COFF-style debugging output
8124 for SDB and allow the frame-pointer to be eliminated when the
8125 @option{-g} options is used.
8128 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
8129 A C expression that returns the integer offset value for an argument
8130 having address @var{x} (an RTL expression). The nominal offset is
8134 @defmac PREFERRED_DEBUGGING_TYPE
8135 A C expression that returns the type of debugging output GCC should
8136 produce when the user specifies just @option{-g}. Define
8137 this if you have arranged for GCC to support more than one format of
8138 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
8139 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
8140 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
8142 When the user specifies @option{-ggdb}, GCC normally also uses the
8143 value of this macro to select the debugging output format, but with two
8144 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
8145 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
8146 defined, GCC uses @code{DBX_DEBUG}.
8148 The value of this macro only affects the default debugging output; the
8149 user can always get a specific type of output by using @option{-gstabs},
8150 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
8154 @subsection Specific Options for DBX Output
8156 @c prevent bad page break with this line
8157 These are specific options for DBX output.
8159 @defmac DBX_DEBUGGING_INFO
8160 Define this macro if GCC should produce debugging output for DBX
8161 in response to the @option{-g} option.
8164 @defmac XCOFF_DEBUGGING_INFO
8165 Define this macro if GCC should produce XCOFF format debugging output
8166 in response to the @option{-g} option. This is a variant of DBX format.
8169 @defmac DEFAULT_GDB_EXTENSIONS
8170 Define this macro to control whether GCC should by default generate
8171 GDB's extended version of DBX debugging information (assuming DBX-format
8172 debugging information is enabled at all). If you don't define the
8173 macro, the default is 1: always generate the extended information
8174 if there is any occasion to.
8177 @defmac DEBUG_SYMS_TEXT
8178 Define this macro if all @code{.stabs} commands should be output while
8179 in the text section.
8182 @defmac ASM_STABS_OP
8183 A C string constant, including spacing, naming the assembler pseudo op to
8184 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
8185 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
8186 applies only to DBX debugging information format.
8189 @defmac ASM_STABD_OP
8190 A C string constant, including spacing, naming the assembler pseudo op to
8191 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
8192 value is the current location. If you don't define this macro,
8193 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
8197 @defmac ASM_STABN_OP
8198 A C string constant, including spacing, naming the assembler pseudo op to
8199 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
8200 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
8201 macro applies only to DBX debugging information format.
8204 @defmac DBX_NO_XREFS
8205 Define this macro if DBX on your system does not support the construct
8206 @samp{xs@var{tagname}}. On some systems, this construct is used to
8207 describe a forward reference to a structure named @var{tagname}.
8208 On other systems, this construct is not supported at all.
8211 @defmac DBX_CONTIN_LENGTH
8212 A symbol name in DBX-format debugging information is normally
8213 continued (split into two separate @code{.stabs} directives) when it
8214 exceeds a certain length (by default, 80 characters). On some
8215 operating systems, DBX requires this splitting; on others, splitting
8216 must not be done. You can inhibit splitting by defining this macro
8217 with the value zero. You can override the default splitting-length by
8218 defining this macro as an expression for the length you desire.
8221 @defmac DBX_CONTIN_CHAR
8222 Normally continuation is indicated by adding a @samp{\} character to
8223 the end of a @code{.stabs} string when a continuation follows. To use
8224 a different character instead, define this macro as a character
8225 constant for the character you want to use. Do not define this macro
8226 if backslash is correct for your system.
8229 @defmac DBX_STATIC_STAB_DATA_SECTION
8230 Define this macro if it is necessary to go to the data section before
8231 outputting the @samp{.stabs} pseudo-op for a non-global static
8235 @defmac DBX_TYPE_DECL_STABS_CODE
8236 The value to use in the ``code'' field of the @code{.stabs} directive
8237 for a typedef. The default is @code{N_LSYM}.
8240 @defmac DBX_STATIC_CONST_VAR_CODE
8241 The value to use in the ``code'' field of the @code{.stabs} directive
8242 for a static variable located in the text section. DBX format does not
8243 provide any ``right'' way to do this. The default is @code{N_FUN}.
8246 @defmac DBX_REGPARM_STABS_CODE
8247 The value to use in the ``code'' field of the @code{.stabs} directive
8248 for a parameter passed in registers. DBX format does not provide any
8249 ``right'' way to do this. The default is @code{N_RSYM}.
8252 @defmac DBX_REGPARM_STABS_LETTER
8253 The letter to use in DBX symbol data to identify a symbol as a parameter
8254 passed in registers. DBX format does not customarily provide any way to
8255 do this. The default is @code{'P'}.
8258 @defmac DBX_FUNCTION_FIRST
8259 Define this macro if the DBX information for a function and its
8260 arguments should precede the assembler code for the function. Normally,
8261 in DBX format, the debugging information entirely follows the assembler
8265 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
8266 Define this macro, with value 1, if the value of a symbol describing
8267 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
8268 relative to the start of the enclosing function. Normally, GCC uses
8269 an absolute address.
8272 @defmac DBX_LINES_FUNCTION_RELATIVE
8273 Define this macro, with value 1, if the value of a symbol indicating
8274 the current line number (@code{N_SLINE}) should be relative to the
8275 start of the enclosing function. Normally, GCC uses an absolute address.
8278 @defmac DBX_USE_BINCL
8279 Define this macro if GCC should generate @code{N_BINCL} and
8280 @code{N_EINCL} stabs for included header files, as on Sun systems. This
8281 macro also directs GCC to output a type number as a pair of a file
8282 number and a type number within the file. Normally, GCC does not
8283 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
8284 number for a type number.
8288 @subsection Open-Ended Hooks for DBX Format
8290 @c prevent bad page break with this line
8291 These are hooks for DBX format.
8293 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
8294 Define this macro to say how to output to @var{stream} the debugging
8295 information for the start of a scope level for variable names. The
8296 argument @var{name} is the name of an assembler symbol (for use with
8297 @code{assemble_name}) whose value is the address where the scope begins.
8300 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
8301 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
8304 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
8305 Define this macro if the target machine requires special handling to
8306 output an @code{N_FUN} entry for the function @var{decl}.
8309 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
8310 A C statement to output DBX debugging information before code for line
8311 number @var{line} of the current source file to the stdio stream
8312 @var{stream}. @var{counter} is the number of time the macro was
8313 invoked, including the current invocation; it is intended to generate
8314 unique labels in the assembly output.
8316 This macro should not be defined if the default output is correct, or
8317 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
8320 @defmac NO_DBX_FUNCTION_END
8321 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
8322 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
8323 On those machines, define this macro to turn this feature off without
8324 disturbing the rest of the gdb extensions.
8327 @defmac NO_DBX_BNSYM_ENSYM
8328 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
8329 extension construct. On those machines, define this macro to turn this
8330 feature off without disturbing the rest of the gdb extensions.
8333 @node File Names and DBX
8334 @subsection File Names in DBX Format
8336 @c prevent bad page break with this line
8337 This describes file names in DBX format.
8339 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
8340 A C statement to output DBX debugging information to the stdio stream
8341 @var{stream}, which indicates that file @var{name} is the main source
8342 file---the file specified as the input file for compilation.
8343 This macro is called only once, at the beginning of compilation.
8345 This macro need not be defined if the standard form of output
8346 for DBX debugging information is appropriate.
8348 It may be necessary to refer to a label equal to the beginning of the
8349 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
8350 to do so. If you do this, you must also set the variable
8351 @var{used_ltext_label_name} to @code{true}.
8354 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
8355 Define this macro, with value 1, if GCC should not emit an indication
8356 of the current directory for compilation and current source language at
8357 the beginning of the file.
8360 @defmac NO_DBX_GCC_MARKER
8361 Define this macro, with value 1, if GCC should not emit an indication
8362 that this object file was compiled by GCC@. The default is to emit
8363 an @code{N_OPT} stab at the beginning of every source file, with
8364 @samp{gcc2_compiled.} for the string and value 0.
8367 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
8368 A C statement to output DBX debugging information at the end of
8369 compilation of the main source file @var{name}. Output should be
8370 written to the stdio stream @var{stream}.
8372 If you don't define this macro, nothing special is output at the end
8373 of compilation, which is correct for most machines.
8376 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
8377 Define this macro @emph{instead of} defining
8378 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
8379 the end of compilation is a @code{N_SO} stab with an empty string,
8380 whose value is the highest absolute text address in the file.
8385 @subsection Macros for SDB and DWARF Output
8387 @c prevent bad page break with this line
8388 Here are macros for SDB and DWARF output.
8390 @defmac SDB_DEBUGGING_INFO
8391 Define this macro if GCC should produce COFF-style debugging output
8392 for SDB in response to the @option{-g} option.
8395 @defmac DWARF2_DEBUGGING_INFO
8396 Define this macro if GCC should produce dwarf version 2 format
8397 debugging output in response to the @option{-g} option.
8399 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (tree @var{function})
8400 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
8401 be emitted for each function. Instead of an integer return the enum
8402 value for the @code{DW_CC_} tag.
8405 To support optional call frame debugging information, you must also
8406 define @code{INCOMING_RETURN_ADDR_RTX} and either set
8407 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
8408 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
8409 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
8412 @defmac DWARF2_FRAME_INFO
8413 Define this macro to a nonzero value if GCC should always output
8414 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
8415 (@pxref{Exception Region Output} is nonzero, GCC will output this
8416 information not matter how you define @code{DWARF2_FRAME_INFO}.
8419 @defmac DWARF2_ASM_LINE_DEBUG_INFO
8420 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8421 line debug info sections. This will result in much more compact line number
8422 tables, and hence is desirable if it works.
8425 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8426 A C statement to issue assembly directives that create a difference
8427 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
8430 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
8431 A C statement to issue assembly directives that create a
8432 section-relative reference to the given @var{label}, using an integer of the
8433 given @var{size}. The label is known to be defined in the given @var{section}.
8436 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8437 A C statement to issue assembly directives that create a self-relative
8438 reference to the given @var{label}, using an integer of the given @var{size}.
8441 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{FILE}, int @var{size}, rtx @var{x})
8442 If defined, this target hook is a function which outputs a DTP-relative
8443 reference to the given TLS symbol of the specified size.
8446 @defmac PUT_SDB_@dots{}
8447 Define these macros to override the assembler syntax for the special
8448 SDB assembler directives. See @file{sdbout.c} for a list of these
8449 macros and their arguments. If the standard syntax is used, you need
8450 not define them yourself.
8454 Some assemblers do not support a semicolon as a delimiter, even between
8455 SDB assembler directives. In that case, define this macro to be the
8456 delimiter to use (usually @samp{\n}). It is not necessary to define
8457 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8461 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
8462 Define this macro to allow references to unknown structure,
8463 union, or enumeration tags to be emitted. Standard COFF does not
8464 allow handling of unknown references, MIPS ECOFF has support for
8468 @defmac SDB_ALLOW_FORWARD_REFERENCES
8469 Define this macro to allow references to structure, union, or
8470 enumeration tags that have not yet been seen to be handled. Some
8471 assemblers choke if forward tags are used, while some require it.
8474 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
8475 A C statement to output SDB debugging information before code for line
8476 number @var{line} of the current source file to the stdio stream
8477 @var{stream}. The default is to emit an @code{.ln} directive.
8482 @subsection Macros for VMS Debug Format
8484 @c prevent bad page break with this line
8485 Here are macros for VMS debug format.
8487 @defmac VMS_DEBUGGING_INFO
8488 Define this macro if GCC should produce debugging output for VMS
8489 in response to the @option{-g} option. The default behavior for VMS
8490 is to generate minimal debug info for a traceback in the absence of
8491 @option{-g} unless explicitly overridden with @option{-g0}. This
8492 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8493 @code{OVERRIDE_OPTIONS}.
8496 @node Floating Point
8497 @section Cross Compilation and Floating Point
8498 @cindex cross compilation and floating point
8499 @cindex floating point and cross compilation
8501 While all modern machines use twos-complement representation for integers,
8502 there are a variety of representations for floating point numbers. This
8503 means that in a cross-compiler the representation of floating point numbers
8504 in the compiled program may be different from that used in the machine
8505 doing the compilation.
8507 Because different representation systems may offer different amounts of
8508 range and precision, all floating point constants must be represented in
8509 the target machine's format. Therefore, the cross compiler cannot
8510 safely use the host machine's floating point arithmetic; it must emulate
8511 the target's arithmetic. To ensure consistency, GCC always uses
8512 emulation to work with floating point values, even when the host and
8513 target floating point formats are identical.
8515 The following macros are provided by @file{real.h} for the compiler to
8516 use. All parts of the compiler which generate or optimize
8517 floating-point calculations must use these macros. They may evaluate
8518 their operands more than once, so operands must not have side effects.
8520 @defmac REAL_VALUE_TYPE
8521 The C data type to be used to hold a floating point value in the target
8522 machine's format. Typically this is a @code{struct} containing an
8523 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8527 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8528 Compares for equality the two values, @var{x} and @var{y}. If the target
8529 floating point format supports negative zeroes and/or NaNs,
8530 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8531 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8534 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8535 Tests whether @var{x} is less than @var{y}.
8538 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8539 Truncates @var{x} to a signed integer, rounding toward zero.
8542 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8543 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8544 @var{x} is negative, returns zero.
8547 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8548 Converts @var{string} into a floating point number in the target machine's
8549 representation for mode @var{mode}. This routine can handle both
8550 decimal and hexadecimal floating point constants, using the syntax
8551 defined by the C language for both.
8554 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8555 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8558 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8559 Determines whether @var{x} represents infinity (positive or negative).
8562 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8563 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8566 @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})
8567 Calculates an arithmetic operation on the two floating point values
8568 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8571 The operation to be performed is specified by @var{code}. Only the
8572 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8573 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8575 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8576 target's floating point format cannot represent infinity, it will call
8577 @code{abort}. Callers should check for this situation first, using
8578 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8581 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8582 Returns the negative of the floating point value @var{x}.
8585 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8586 Returns the absolute value of @var{x}.
8589 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8590 Truncates the floating point value @var{x} to fit in @var{mode}. The
8591 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8592 appropriate bit pattern to be output asa floating constant whose
8593 precision accords with mode @var{mode}.
8596 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8597 Converts a floating point value @var{x} into a double-precision integer
8598 which is then stored into @var{low} and @var{high}. If the value is not
8599 integral, it is truncated.
8602 @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})
8603 Converts a double-precision integer found in @var{low} and @var{high},
8604 into a floating point value which is then stored into @var{x}. The
8605 value is truncated to fit in mode @var{mode}.
8608 @node Mode Switching
8609 @section Mode Switching Instructions
8610 @cindex mode switching
8611 The following macros control mode switching optimizations:
8613 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
8614 Define this macro if the port needs extra instructions inserted for mode
8615 switching in an optimizing compilation.
8617 For an example, the SH4 can perform both single and double precision
8618 floating point operations, but to perform a single precision operation,
8619 the FPSCR PR bit has to be cleared, while for a double precision
8620 operation, this bit has to be set. Changing the PR bit requires a general
8621 purpose register as a scratch register, hence these FPSCR sets have to
8622 be inserted before reload, i.e.@: you can't put this into instruction emitting
8623 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
8625 You can have multiple entities that are mode-switched, and select at run time
8626 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
8627 return nonzero for any @var{entity} that needs mode-switching.
8628 If you define this macro, you also have to define
8629 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
8630 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
8631 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
8635 @defmac NUM_MODES_FOR_MODE_SWITCHING
8636 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
8637 initializer for an array of integers. Each initializer element
8638 N refers to an entity that needs mode switching, and specifies the number
8639 of different modes that might need to be set for this entity.
8640 The position of the initializer in the initializer---starting counting at
8641 zero---determines the integer that is used to refer to the mode-switched
8643 In macros that take mode arguments / yield a mode result, modes are
8644 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
8645 switch is needed / supplied.
8648 @defmac MODE_NEEDED (@var{entity}, @var{insn})
8649 @var{entity} is an integer specifying a mode-switched entity. If
8650 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
8651 return an integer value not larger than the corresponding element in
8652 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
8653 be switched into prior to the execution of @var{insn}.
8656 @defmac MODE_AFTER (@var{mode}, @var{insn})
8657 If this macro is defined, it is evaluated for every @var{insn} during
8658 mode switching. It determines the mode that an insn results in (if
8659 different from the incoming mode).
8662 @defmac MODE_ENTRY (@var{entity})
8663 If this macro is defined, it is evaluated for every @var{entity} that needs
8664 mode switching. It should evaluate to an integer, which is a mode that
8665 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
8666 is defined then @code{MODE_EXIT} must be defined.
8669 @defmac MODE_EXIT (@var{entity})
8670 If this macro is defined, it is evaluated for every @var{entity} that needs
8671 mode switching. It should evaluate to an integer, which is a mode that
8672 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
8673 is defined then @code{MODE_ENTRY} must be defined.
8676 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
8677 This macro specifies the order in which modes for @var{entity} are processed.
8678 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
8679 lowest. The value of the macro should be an integer designating a mode
8680 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
8681 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
8682 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
8685 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
8686 Generate one or more insns to set @var{entity} to @var{mode}.
8687 @var{hard_reg_live} is the set of hard registers live at the point where
8688 the insn(s) are to be inserted.
8691 @node Target Attributes
8692 @section Defining target-specific uses of @code{__attribute__}
8693 @cindex target attributes
8694 @cindex machine attributes
8695 @cindex attributes, target-specific
8697 Target-specific attributes may be defined for functions, data and types.
8698 These are described using the following target hooks; they also need to
8699 be documented in @file{extend.texi}.
8701 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
8702 If defined, this target hook points to an array of @samp{struct
8703 attribute_spec} (defined in @file{tree.h}) specifying the machine
8704 specific attributes for this target and some of the restrictions on the
8705 entities to which these attributes are applied and the arguments they
8709 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8710 If defined, this target hook is a function which returns zero if the attributes on
8711 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8712 and two if they are nearly compatible (which causes a warning to be
8713 generated). If this is not defined, machine-specific attributes are
8714 supposed always to be compatible.
8717 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
8718 If defined, this target hook is a function which assigns default attributes to
8719 newly defined @var{type}.
8722 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8723 Define this target hook if the merging of type attributes needs special
8724 handling. If defined, the result is a list of the combined
8725 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
8726 that @code{comptypes} has already been called and returned 1. This
8727 function may call @code{merge_attributes} to handle machine-independent
8731 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
8732 Define this target hook if the merging of decl attributes needs special
8733 handling. If defined, the result is a list of the combined
8734 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
8735 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
8736 when this is needed are when one attribute overrides another, or when an
8737 attribute is nullified by a subsequent definition. This function may
8738 call @code{merge_attributes} to handle machine-independent merging.
8740 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
8741 If the only target-specific handling you require is @samp{dllimport}
8742 for Microsoft Windows targets, you should define the macro
8743 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
8744 will then define a function called
8745 @code{merge_dllimport_decl_attributes} which can then be defined as
8746 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
8747 add @code{handle_dll_attribute} in the attribute table for your port
8748 to perform initial processing of the @samp{dllimport} and
8749 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
8750 @file{i386/i386.c}, for example.
8753 @deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (tree @var{decl})
8754 @var{decl} is a variable or function with @code{__attribute__((dllimport))}
8755 specified. Use this hook if the target needs to add extra validation
8756 checks to @code{handle_dll_attribute}.
8759 @defmac TARGET_DECLSPEC
8760 Define this macro to a nonzero value if you want to treat
8761 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
8762 default, this behavior is enabled only for targets that define
8763 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
8764 of @code{__declspec} is via a built-in macro, but you should not rely
8765 on this implementation detail.
8768 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
8769 Define this target hook if you want to be able to add attributes to a decl
8770 when it is being created. This is normally useful for back ends which
8771 wish to implement a pragma by using the attributes which correspond to
8772 the pragma's effect. The @var{node} argument is the decl which is being
8773 created. The @var{attr_ptr} argument is a pointer to the attribute list
8774 for this decl. The list itself should not be modified, since it may be
8775 shared with other decls, but attributes may be chained on the head of
8776 the list and @code{*@var{attr_ptr}} modified to point to the new
8777 attributes, or a copy of the list may be made if further changes are
8781 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
8783 This target hook returns @code{true} if it is ok to inline @var{fndecl}
8784 into the current function, despite its having target-specific
8785 attributes, @code{false} otherwise. By default, if a function has a
8786 target specific attribute attached to it, it will not be inlined.
8789 @node MIPS Coprocessors
8790 @section Defining coprocessor specifics for MIPS targets.
8791 @cindex MIPS coprocessor-definition macros
8793 The MIPS specification allows MIPS implementations to have as many as 4
8794 coprocessors, each with as many as 32 private registers. GCC supports
8795 accessing these registers and transferring values between the registers
8796 and memory using asm-ized variables. For example:
8799 register unsigned int cp0count asm ("c0r1");
8805 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
8806 names may be added as described below, or the default names may be
8807 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
8809 Coprocessor registers are assumed to be epilogue-used; sets to them will
8810 be preserved even if it does not appear that the register is used again
8811 later in the function.
8813 Another note: according to the MIPS spec, coprocessor 1 (if present) is
8814 the FPU@. One accesses COP1 registers through standard mips
8815 floating-point support; they are not included in this mechanism.
8817 There is one macro used in defining the MIPS coprocessor interface which
8818 you may want to override in subtargets; it is described below.
8820 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
8821 A comma-separated list (with leading comma) of pairs describing the
8822 alternate names of coprocessor registers. The format of each entry should be
8824 @{ @var{alternatename}, @var{register_number}@}
8830 @section Parameters for Precompiled Header Validity Checking
8831 @cindex parameters, precompiled headers
8833 @deftypefn {Target Hook} void *TARGET_GET_PCH_VALIDITY (size_t *@var{sz})
8834 This hook returns the data needed by @code{TARGET_PCH_VALID_P} and sets
8835 @samp{*@var{sz}} to the size of the data in bytes.
8838 @deftypefn {Target Hook} const char *TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz})
8839 This hook checks whether the options used to create a PCH file are
8840 compatible with the current settings. It returns @code{NULL}
8841 if so and a suitable error message if not. Error messages will
8842 be presented to the user and must be localized using @samp{_(@var{msg})}.
8844 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
8845 when the PCH file was created and @var{sz} is the size of that data in bytes.
8846 It's safe to assume that the data was created by the same version of the
8847 compiler, so no format checking is needed.
8849 The default definition of @code{default_pch_valid_p} should be
8850 suitable for most targets.
8853 @deftypefn {Target Hook} const char *TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags})
8854 If this hook is nonnull, the default implementation of
8855 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
8856 of @code{target_flags}. @var{pch_flags} specifies the value that
8857 @code{target_flags} had when the PCH file was created. The return
8858 value is the same as for @code{TARGET_PCH_VALID_P}.
8862 @section C++ ABI parameters
8863 @cindex parameters, c++ abi
8865 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
8866 Define this hook to override the integer type used for guard variables.
8867 These are used to implement one-time construction of static objects. The
8868 default is long_long_integer_type_node.
8871 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
8872 This hook determines how guard variables are used. It should return
8873 @code{false} (the default) if first byte should be used. A return value of
8874 @code{true} indicates the least significant bit should be used.
8877 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
8878 This hook returns the size of the cookie to use when allocating an array
8879 whose elements have the indicated @var{type}. Assumes that it is already
8880 known that a cookie is needed. The default is
8881 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
8882 IA64/Generic C++ ABI@.
8885 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
8886 This hook should return @code{true} if the element size should be stored in
8887 array cookies. The default is to return @code{false}.
8890 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
8891 If defined by a backend this hook allows the decision made to export
8892 class @var{type} to be overruled. Upon entry @var{import_export}
8893 will contain 1 if the class is going to be exported, @minus{}1 if it is going
8894 to be imported and 0 otherwise. This function should return the
8895 modified value and perform any other actions necessary to support the
8896 backend's targeted operating system.
8899 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
8900 This hook should return @code{true} if constructors and destructors return
8901 the address of the object created/destroyed. The default is to return
8905 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
8906 This hook returns true if the key method for a class (i.e., the method
8907 which, if defined in the current translation unit, causes the virtual
8908 table to be emitted) may be an inline function. Under the standard
8909 Itanium C++ ABI the key method may be an inline function so long as
8910 the function is not declared inline in the class definition. Under
8911 some variants of the ABI, an inline function can never be the key
8912 method. The default is to return @code{true}.
8915 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
8916 @var{decl} is a virtual table, virtual table table, typeinfo object,
8917 or other similar implicit class data object that will be emitted with
8918 external linkage in this translation unit. No ELF visibility has been
8919 explicitly specified. If the target needs to specify a visibility
8920 other than that of the containing class, use this hook to set
8921 @code{DECL_VISIBILITY} and @code{DECL_VISIBILITY_SPECIFIED}.
8924 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
8925 This hook returns true (the default) if virtual tables and other
8926 similar implicit class data objects are always COMDAT if they have
8927 external linkage. If this hook returns false, then class data for
8928 classes whose virtual table will be emitted in only one translation
8929 unit will not be COMDAT.
8932 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
8933 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
8934 should be used to register static destructors when @option{-fuse-cxa-atexit}
8935 is in effect. The default is to return false to use @code{__cxa_atexit}.
8938 @deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type})
8939 @var{type} is a C++ class (i.e., RECORD_TYPE or UNION_TYPE) that has just been
8940 defined. Use this hook to make adjustments to the class (eg, tweak
8941 visibility or perform any other required target modifications).
8945 @section Miscellaneous Parameters
8946 @cindex parameters, miscellaneous
8948 @c prevent bad page break with this line
8949 Here are several miscellaneous parameters.
8951 @defmac HAS_LONG_COND_BRANCH
8952 Define this boolean macro to indicate whether or not your architecture
8953 has conditional branches that can span all of memory. It is used in
8954 conjunction with an optimization that partitions hot and cold basic
8955 blocks into separate sections of the executable. If this macro is
8956 set to false, gcc will convert any conditional branches that attempt
8957 to cross between sections into unconditional branches or indirect jumps.
8960 @defmac HAS_LONG_UNCOND_BRANCH
8961 Define this boolean macro to indicate whether or not your architecture
8962 has unconditional branches that can span all of memory. It is used in
8963 conjunction with an optimization that partitions hot and cold basic
8964 blocks into separate sections of the executable. If this macro is
8965 set to false, gcc will convert any unconditional branches that attempt
8966 to cross between sections into indirect jumps.
8969 @defmac CASE_VECTOR_MODE
8970 An alias for a machine mode name. This is the machine mode that
8971 elements of a jump-table should have.
8974 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
8975 Optional: return the preferred mode for an @code{addr_diff_vec}
8976 when the minimum and maximum offset are known. If you define this,
8977 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
8978 To make this work, you also have to define @code{INSN_ALIGN} and
8979 make the alignment for @code{addr_diff_vec} explicit.
8980 The @var{body} argument is provided so that the offset_unsigned and scale
8981 flags can be updated.
8984 @defmac CASE_VECTOR_PC_RELATIVE
8985 Define this macro to be a C expression to indicate when jump-tables
8986 should contain relative addresses. You need not define this macro if
8987 jump-tables never contain relative addresses, or jump-tables should
8988 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
8992 @defmac CASE_VALUES_THRESHOLD
8993 Define this to be the smallest number of different values for which it
8994 is best to use a jump-table instead of a tree of conditional branches.
8995 The default is four for machines with a @code{casesi} instruction and
8996 five otherwise. This is best for most machines.
8999 @defmac CASE_USE_BIT_TESTS
9000 Define this macro to be a C expression to indicate whether C switch
9001 statements may be implemented by a sequence of bit tests. This is
9002 advantageous on processors that can efficiently implement left shift
9003 of 1 by the number of bits held in a register, but inappropriate on
9004 targets that would require a loop. By default, this macro returns
9005 @code{true} if the target defines an @code{ashlsi3} pattern, and
9006 @code{false} otherwise.
9009 @defmac WORD_REGISTER_OPERATIONS
9010 Define this macro if operations between registers with integral mode
9011 smaller than a word are always performed on the entire register.
9012 Most RISC machines have this property and most CISC machines do not.
9015 @defmac LOAD_EXTEND_OP (@var{mem_mode})
9016 Define this macro to be a C expression indicating when insns that read
9017 memory in @var{mem_mode}, an integral mode narrower than a word, set the
9018 bits outside of @var{mem_mode} to be either the sign-extension or the
9019 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
9020 of @var{mem_mode} for which the
9021 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
9022 @code{UNKNOWN} for other modes.
9024 This macro is not called with @var{mem_mode} non-integral or with a width
9025 greater than or equal to @code{BITS_PER_WORD}, so you may return any
9026 value in this case. Do not define this macro if it would always return
9027 @code{UNKNOWN}. On machines where this macro is defined, you will normally
9028 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
9030 You may return a non-@code{UNKNOWN} value even if for some hard registers
9031 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
9032 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
9033 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
9034 integral mode larger than this but not larger than @code{word_mode}.
9036 You must return @code{UNKNOWN} if for some hard registers that allow this
9037 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
9038 @code{word_mode}, but that they can change to another integral mode that
9039 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
9042 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
9043 Define this macro if loading short immediate values into registers sign
9047 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
9048 Define this macro if the same instructions that convert a floating
9049 point number to a signed fixed point number also convert validly to an
9053 @deftypefn {Target Hook} int TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (enum machine_mode @var{mode})
9054 When @option{-ffast-math} is in effect, GCC tries to optimize
9055 divisions by the same divisor, by turning them into multiplications by
9056 the reciprocal. This target hook specifies the minimum number of divisions
9057 that should be there for GCC to perform the optimization for a variable
9058 of mode @var{mode}. The default implementation returns 3 if the machine
9059 has an instruction for the division, and 2 if it does not.
9063 The maximum number of bytes that a single instruction can move quickly
9064 between memory and registers or between two memory locations.
9067 @defmac MAX_MOVE_MAX
9068 The maximum number of bytes that a single instruction can move quickly
9069 between memory and registers or between two memory locations. If this
9070 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
9071 constant value that is the largest value that @code{MOVE_MAX} can have
9075 @defmac SHIFT_COUNT_TRUNCATED
9076 A C expression that is nonzero if on this machine the number of bits
9077 actually used for the count of a shift operation is equal to the number
9078 of bits needed to represent the size of the object being shifted. When
9079 this macro is nonzero, the compiler will assume that it is safe to omit
9080 a sign-extend, zero-extend, and certain bitwise `and' instructions that
9081 truncates the count of a shift operation. On machines that have
9082 instructions that act on bit-fields at variable positions, which may
9083 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
9084 also enables deletion of truncations of the values that serve as
9085 arguments to bit-field instructions.
9087 If both types of instructions truncate the count (for shifts) and
9088 position (for bit-field operations), or if no variable-position bit-field
9089 instructions exist, you should define this macro.
9091 However, on some machines, such as the 80386 and the 680x0, truncation
9092 only applies to shift operations and not the (real or pretended)
9093 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
9094 such machines. Instead, add patterns to the @file{md} file that include
9095 the implied truncation of the shift instructions.
9097 You need not define this macro if it would always have the value of zero.
9100 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
9101 @deftypefn {Target Hook} int TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode})
9102 This function describes how the standard shift patterns for @var{mode}
9103 deal with shifts by negative amounts or by more than the width of the mode.
9104 @xref{shift patterns}.
9106 On many machines, the shift patterns will apply a mask @var{m} to the
9107 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
9108 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
9109 this is true for mode @var{mode}, the function should return @var{m},
9110 otherwise it should return 0. A return value of 0 indicates that no
9111 particular behavior is guaranteed.
9113 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
9114 @emph{not} apply to general shift rtxes; it applies only to instructions
9115 that are generated by the named shift patterns.
9117 The default implementation of this function returns
9118 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
9119 and 0 otherwise. This definition is always safe, but if
9120 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
9121 nevertheless truncate the shift count, you may get better code
9125 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
9126 A C expression which is nonzero if on this machine it is safe to
9127 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
9128 bits (where @var{outprec} is smaller than @var{inprec}) by merely
9129 operating on it as if it had only @var{outprec} bits.
9131 On many machines, this expression can be 1.
9133 @c rearranged this, removed the phrase "it is reported that". this was
9134 @c to fix an overfull hbox. --mew 10feb93
9135 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
9136 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
9137 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
9138 such cases may improve things.
9141 @defmac STORE_FLAG_VALUE
9142 A C expression describing the value returned by a comparison operator
9143 with an integral mode and stored by a store-flag instruction
9144 (@samp{s@var{cond}}) when the condition is true. This description must
9145 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
9146 comparison operators whose results have a @code{MODE_INT} mode.
9148 A value of 1 or @minus{}1 means that the instruction implementing the
9149 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
9150 and 0 when the comparison is false. Otherwise, the value indicates
9151 which bits of the result are guaranteed to be 1 when the comparison is
9152 true. This value is interpreted in the mode of the comparison
9153 operation, which is given by the mode of the first operand in the
9154 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
9155 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
9158 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
9159 generate code that depends only on the specified bits. It can also
9160 replace comparison operators with equivalent operations if they cause
9161 the required bits to be set, even if the remaining bits are undefined.
9162 For example, on a machine whose comparison operators return an
9163 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
9164 @samp{0x80000000}, saying that just the sign bit is relevant, the
9168 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
9175 (ashift:SI @var{x} (const_int @var{n}))
9179 where @var{n} is the appropriate shift count to move the bit being
9180 tested into the sign bit.
9182 There is no way to describe a machine that always sets the low-order bit
9183 for a true value, but does not guarantee the value of any other bits,
9184 but we do not know of any machine that has such an instruction. If you
9185 are trying to port GCC to such a machine, include an instruction to
9186 perform a logical-and of the result with 1 in the pattern for the
9187 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
9189 Often, a machine will have multiple instructions that obtain a value
9190 from a comparison (or the condition codes). Here are rules to guide the
9191 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
9196 Use the shortest sequence that yields a valid definition for
9197 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
9198 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
9199 comparison operators to do so because there may be opportunities to
9200 combine the normalization with other operations.
9203 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
9204 slightly preferred on machines with expensive jumps and 1 preferred on
9208 As a second choice, choose a value of @samp{0x80000001} if instructions
9209 exist that set both the sign and low-order bits but do not define the
9213 Otherwise, use a value of @samp{0x80000000}.
9216 Many machines can produce both the value chosen for
9217 @code{STORE_FLAG_VALUE} and its negation in the same number of
9218 instructions. On those machines, you should also define a pattern for
9219 those cases, e.g., one matching
9222 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
9225 Some machines can also perform @code{and} or @code{plus} operations on
9226 condition code values with less instructions than the corresponding
9227 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
9228 machines, define the appropriate patterns. Use the names @code{incscc}
9229 and @code{decscc}, respectively, for the patterns which perform
9230 @code{plus} or @code{minus} operations on condition code values. See
9231 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
9232 find such instruction sequences on other machines.
9234 If this macro is not defined, the default value, 1, is used. You need
9235 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
9236 instructions, or if the value generated by these instructions is 1.
9239 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
9240 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
9241 returned when comparison operators with floating-point results are true.
9242 Define this macro on machines that have comparison operations that return
9243 floating-point values. If there are no such operations, do not define
9247 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
9248 A C expression that gives a rtx representing the nonzero true element
9249 for vector comparisons. The returned rtx should be valid for the inner
9250 mode of @var{mode} which is guaranteed to be a vector mode. Define
9251 this macro on machines that have vector comparison operations that
9252 return a vector result. If there are no such operations, do not define
9253 this macro. Typically, this macro is defined as @code{const1_rtx} or
9254 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
9255 the compiler optimizing such vector comparison operations for the
9259 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
9260 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
9261 A C expression that evaluates to true if the architecture defines a value
9262 for @code{clz} or @code{ctz} with a zero operand. If so, @var{value}
9263 should be set to this value. If this macro is not defined, the value of
9264 @code{clz} or @code{ctz} is assumed to be undefined.
9266 This macro must be defined if the target's expansion for @code{ffs}
9267 relies on a particular value to get correct results. Otherwise it
9268 is not necessary, though it may be used to optimize some corner cases.
9270 Note that regardless of this macro the ``definedness'' of @code{clz}
9271 and @code{ctz} at zero do @emph{not} extend to the builtin functions
9272 visible to the user. Thus one may be free to adjust the value at will
9273 to match the target expansion of these operations without fear of
9278 An alias for the machine mode for pointers. On most machines, define
9279 this to be the integer mode corresponding to the width of a hardware
9280 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
9281 On some machines you must define this to be one of the partial integer
9282 modes, such as @code{PSImode}.
9284 The width of @code{Pmode} must be at least as large as the value of
9285 @code{POINTER_SIZE}. If it is not equal, you must define the macro
9286 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
9290 @defmac FUNCTION_MODE
9291 An alias for the machine mode used for memory references to functions
9292 being called, in @code{call} RTL expressions. On most machines this
9293 should be @code{QImode}.
9296 @defmac STDC_0_IN_SYSTEM_HEADERS
9297 In normal operation, the preprocessor expands @code{__STDC__} to the
9298 constant 1, to signify that GCC conforms to ISO Standard C@. On some
9299 hosts, like Solaris, the system compiler uses a different convention,
9300 where @code{__STDC__} is normally 0, but is 1 if the user specifies
9301 strict conformance to the C Standard.
9303 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
9304 convention when processing system header files, but when processing user
9305 files @code{__STDC__} will always expand to 1.
9308 @defmac NO_IMPLICIT_EXTERN_C
9309 Define this macro if the system header files support C++ as well as C@.
9310 This macro inhibits the usual method of using system header files in
9311 C++, which is to pretend that the file's contents are enclosed in
9312 @samp{extern "C" @{@dots{}@}}.
9317 @defmac REGISTER_TARGET_PRAGMAS ()
9318 Define this macro if you want to implement any target-specific pragmas.
9319 If defined, it is a C expression which makes a series of calls to
9320 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
9321 for each pragma. The macro may also do any
9322 setup required for the pragmas.
9324 The primary reason to define this macro is to provide compatibility with
9325 other compilers for the same target. In general, we discourage
9326 definition of target-specific pragmas for GCC@.
9328 If the pragma can be implemented by attributes then you should consider
9329 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
9331 Preprocessor macros that appear on pragma lines are not expanded. All
9332 @samp{#pragma} directives that do not match any registered pragma are
9333 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
9336 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9337 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9339 Each call to @code{c_register_pragma} or
9340 @code{c_register_pragma_with_expansion} establishes one pragma. The
9341 @var{callback} routine will be called when the preprocessor encounters a
9345 #pragma [@var{space}] @var{name} @dots{}
9348 @var{space} is the case-sensitive namespace of the pragma, or
9349 @code{NULL} to put the pragma in the global namespace. The callback
9350 routine receives @var{pfile} as its first argument, which can be passed
9351 on to cpplib's functions if necessary. You can lex tokens after the
9352 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
9353 callback will be silently ignored. The end of the line is indicated by
9354 a token of type @code{CPP_EOF}. Macro expansion occurs on the
9355 arguments of pragmas registered with
9356 @code{c_register_pragma_with_expansion} but not on the arguments of
9357 pragmas registered with @code{c_register_pragma}.
9359 For an example use of this routine, see @file{c4x.h} and the callback
9360 routines defined in @file{c4x-c.c}.
9362 Note that the use of @code{pragma_lex} is specific to the C and C++
9363 compilers. It will not work in the Java or Fortran compilers, or any
9364 other language compilers for that matter. Thus if @code{pragma_lex} is going
9365 to be called from target-specific code, it must only be done so when
9366 building the C and C++ compilers. This can be done by defining the
9367 variables @code{c_target_objs} and @code{cxx_target_objs} in the
9368 target entry in the @file{config.gcc} file. These variables should name
9369 the target-specific, language-specific object file which contains the
9370 code that uses @code{pragma_lex}. Note it will also be necessary to add a
9371 rule to the makefile fragment pointed to by @code{tmake_file} that shows
9372 how to build this object file.
9377 @defmac HANDLE_SYSV_PRAGMA
9378 Define this macro (to a value of 1) if you want the System V style
9379 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
9380 [=<value>]} to be supported by gcc.
9382 The pack pragma specifies the maximum alignment (in bytes) of fields
9383 within a structure, in much the same way as the @samp{__aligned__} and
9384 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
9385 the behavior to the default.
9387 A subtlety for Microsoft Visual C/C++ style bit-field packing
9388 (e.g.@: -mms-bitfields) for targets that support it:
9389 When a bit-field is inserted into a packed record, the whole size
9390 of the underlying type is used by one or more same-size adjacent
9391 bit-fields (that is, if its long:3, 32 bits is used in the record,
9392 and any additional adjacent long bit-fields are packed into the same
9393 chunk of 32 bits. However, if the size changes, a new field of that
9396 If both MS bit-fields and @samp{__attribute__((packed))} are used,
9397 the latter will take precedence. If @samp{__attribute__((packed))} is
9398 used on a single field when MS bit-fields are in use, it will take
9399 precedence for that field, but the alignment of the rest of the structure
9400 may affect its placement.
9402 The weak pragma only works if @code{SUPPORTS_WEAK} and
9403 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
9404 of specifically named weak labels, optionally with a value.
9409 @defmac HANDLE_PRAGMA_PACK_PUSH_POP
9410 Define this macro (to a value of 1) if you want to support the Win32
9411 style pragmas @samp{#pragma pack(push[,@var{n}])} and @samp{#pragma
9412 pack(pop)}. The @samp{pack(push,[@var{n}])} pragma specifies the maximum
9413 alignment (in bytes) of fields within a structure, in much the same way as
9414 the @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
9415 pack value of zero resets the behavior to the default. Successive
9416 invocations of this pragma cause the previous values to be stacked, so
9417 that invocations of @samp{#pragma pack(pop)} will return to the previous
9421 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
9422 Define this macro, as well as
9423 @code{HANDLE_SYSV_PRAGMA}, if macros should be expanded in the
9424 arguments of @samp{#pragma pack}.
9427 @defmac TARGET_DEFAULT_PACK_STRUCT
9428 If your target requires a structure packing default other than 0 (meaning
9429 the machine default), define this macro to the necessary value (in bytes).
9430 This must be a value that would also valid to be used with
9431 @samp{#pragma pack()} (that is, a small power of two).
9434 @defmac DOLLARS_IN_IDENTIFIERS
9435 Define this macro to control use of the character @samp{$} in
9436 identifier names for the C family of languages. 0 means @samp{$} is
9437 not allowed by default; 1 means it is allowed. 1 is the default;
9438 there is no need to define this macro in that case.
9441 @defmac NO_DOLLAR_IN_LABEL
9442 Define this macro if the assembler does not accept the character
9443 @samp{$} in label names. By default constructors and destructors in
9444 G++ have @samp{$} in the identifiers. If this macro is defined,
9445 @samp{.} is used instead.
9448 @defmac NO_DOT_IN_LABEL
9449 Define this macro if the assembler does not accept the character
9450 @samp{.} in label names. By default constructors and destructors in G++
9451 have names that use @samp{.}. If this macro is defined, these names
9452 are rewritten to avoid @samp{.}.
9455 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
9456 Define this macro as a C expression that is nonzero if it is safe for the
9457 delay slot scheduler to place instructions in the delay slot of @var{insn},
9458 even if they appear to use a resource set or clobbered in @var{insn}.
9459 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
9460 every @code{call_insn} has this behavior. On machines where some @code{insn}
9461 or @code{jump_insn} is really a function call and hence has this behavior,
9462 you should define this macro.
9464 You need not define this macro if it would always return zero.
9467 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
9468 Define this macro as a C expression that is nonzero if it is safe for the
9469 delay slot scheduler to place instructions in the delay slot of @var{insn},
9470 even if they appear to set or clobber a resource referenced in @var{insn}.
9471 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
9472 some @code{insn} or @code{jump_insn} is really a function call and its operands
9473 are registers whose use is actually in the subroutine it calls, you should
9474 define this macro. Doing so allows the delay slot scheduler to move
9475 instructions which copy arguments into the argument registers into the delay
9478 You need not define this macro if it would always return zero.
9481 @defmac MULTIPLE_SYMBOL_SPACES
9482 Define this macro as a C expression that is nonzero if, in some cases,
9483 global symbols from one translation unit may not be bound to undefined
9484 symbols in another translation unit without user intervention. For
9485 instance, under Microsoft Windows symbols must be explicitly imported
9486 from shared libraries (DLLs).
9488 You need not define this macro if it would always evaluate to zero.
9491 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{outputs}, tree @var{inputs}, tree @var{clobbers})
9492 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
9493 any hard regs the port wishes to automatically clobber for an asm.
9494 It should return the result of the last @code{tree_cons} used to add a
9495 clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the
9496 corresponding parameters to the asm and may be inspected to avoid
9497 clobbering a register that is an input or output of the asm. You can use
9498 @code{tree_overlaps_hard_reg_set}, declared in @file{tree.h}, to test
9499 for overlap with regards to asm-declared registers.
9502 @defmac MATH_LIBRARY
9503 Define this macro as a C string constant for the linker argument to link
9504 in the system math library, or @samp{""} if the target does not have a
9505 separate math library.
9507 You need only define this macro if the default of @samp{"-lm"} is wrong.
9510 @defmac LIBRARY_PATH_ENV
9511 Define this macro as a C string constant for the environment variable that
9512 specifies where the linker should look for libraries.
9514 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
9518 @defmac TARGET_POSIX_IO
9519 Define this macro if the target supports the following POSIX@ file
9520 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
9521 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
9522 to use file locking when exiting a program, which avoids race conditions
9523 if the program has forked. It will also create directories at run-time
9524 for cross-profiling.
9527 @defmac MAX_CONDITIONAL_EXECUTE
9529 A C expression for the maximum number of instructions to execute via
9530 conditional execution instructions instead of a branch. A value of
9531 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
9532 1 if it does use cc0.
9535 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
9536 Used if the target needs to perform machine-dependent modifications on the
9537 conditionals used for turning basic blocks into conditionally executed code.
9538 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
9539 contains information about the currently processed blocks. @var{true_expr}
9540 and @var{false_expr} are the tests that are used for converting the
9541 then-block and the else-block, respectively. Set either @var{true_expr} or
9542 @var{false_expr} to a null pointer if the tests cannot be converted.
9545 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
9546 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
9547 if-statements into conditions combined by @code{and} and @code{or} operations.
9548 @var{bb} contains the basic block that contains the test that is currently
9549 being processed and about to be turned into a condition.
9552 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
9553 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
9554 be converted to conditional execution format. @var{ce_info} points to
9555 a data structure, @code{struct ce_if_block}, which contains information
9556 about the currently processed blocks.
9559 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
9560 A C expression to perform any final machine dependent modifications in
9561 converting code to conditional execution. The involved basic blocks
9562 can be found in the @code{struct ce_if_block} structure that is pointed
9563 to by @var{ce_info}.
9566 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
9567 A C expression to cancel any machine dependent modifications in
9568 converting code to conditional execution. The involved basic blocks
9569 can be found in the @code{struct ce_if_block} structure that is pointed
9570 to by @var{ce_info}.
9573 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
9574 A C expression to initialize any extra fields in a @code{struct ce_if_block}
9575 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
9578 @defmac IFCVT_EXTRA_FIELDS
9579 If defined, it should expand to a set of field declarations that will be
9580 added to the @code{struct ce_if_block} structure. These should be initialized
9581 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
9584 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG ()
9585 If non-null, this hook performs a target-specific pass over the
9586 instruction stream. The compiler will run it at all optimization levels,
9587 just before the point at which it normally does delayed-branch scheduling.
9589 The exact purpose of the hook varies from target to target. Some use
9590 it to do transformations that are necessary for correctness, such as
9591 laying out in-function constant pools or avoiding hardware hazards.
9592 Others use it as an opportunity to do some machine-dependent optimizations.
9594 You need not implement the hook if it has nothing to do. The default
9598 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
9599 Define this hook if you have any machine-specific built-in functions
9600 that need to be defined. It should be a function that performs the
9603 Machine specific built-in functions can be useful to expand special machine
9604 instructions that would otherwise not normally be generated because
9605 they have no equivalent in the source language (for example, SIMD vector
9606 instructions or prefetch instructions).
9608 To create a built-in function, call the function
9609 @code{lang_hooks.builtin_function}
9610 which is defined by the language front end. You can use any type nodes set
9611 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
9612 only language front ends that use those two functions will call
9613 @samp{TARGET_INIT_BUILTINS}.
9616 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
9618 Expand a call to a machine specific built-in function that was set up by
9619 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
9620 function call; the result should go to @var{target} if that is
9621 convenient, and have mode @var{mode} if that is convenient.
9622 @var{subtarget} may be used as the target for computing one of
9623 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
9624 ignored. This function should return the result of the call to the
9628 @deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (tree @var{fndecl}, tree @var{arglist})
9630 Select a replacement for a machine specific built-in function that
9631 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
9632 @emph{before} regular type checking, and so allows the target to
9633 implement a crude form of function overloading. @var{fndecl} is the
9634 declaration of the built-in function. @var{arglist} is the list of
9635 arguments passed to the built-in function. The result is a
9636 complete expression that implements the operation, usually
9637 another @code{CALL_EXPR}.
9640 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, tree @var{arglist}, bool @var{ignore})
9642 Fold a call to a machine specific built-in function that was set up by
9643 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
9644 built-in function. @var{arglist} is the list of arguments passed to
9645 the built-in function. The result is another tree containing a
9646 simplified expression for the call's result. If @var{ignore} is true
9647 the value will be ignored.
9650 @deftypefn {Target Hook} const char * TARGET_INVALID_WITHIN_DOLOOP (rtx @var{insn})
9652 Take an instruction in @var{insn} and return NULL if it is valid within a
9653 low-overhead loop, otherwise return a string why doloop could not be applied.
9655 Many targets use special registers for low-overhead looping. For any
9656 instruction that clobbers these this function should return a string indicating
9657 the reason why the doloop could not be applied.
9658 By default, the RTL loop optimizer does not use a present doloop pattern for
9659 loops containing function calls or branch on table instructions.
9662 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
9664 Take a branch insn in @var{branch1} and another in @var{branch2}.
9665 Return true if redirecting @var{branch1} to the destination of
9666 @var{branch2} is possible.
9668 On some targets, branches may have a limited range. Optimizing the
9669 filling of delay slots can result in branches being redirected, and this
9670 may in turn cause a branch offset to overflow.
9673 @deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (rtx @var{x}, @var{outer_code})
9674 This target hook returns @code{true} if @var{x} is considered to be commutative.
9675 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
9676 PLUS to be commutative inside a MEM. @var{outer_code} is the rtx code
9677 of the enclosing rtl, if known, otherwise it is UNKNOWN.
9680 @deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg})
9682 When the initial value of a hard register has been copied in a pseudo
9683 register, it is often not necessary to actually allocate another register
9684 to this pseudo register, because the original hard register or a stack slot
9685 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
9686 is called at the start of register allocation once for each hard register
9687 that had its initial value copied by using
9688 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
9689 Possible values are @code{NULL_RTX}, if you don't want
9690 to do any special allocation, a @code{REG} rtx---that would typically be
9691 the hard register itself, if it is known not to be clobbered---or a
9693 If you are returning a @code{MEM}, this is only a hint for the allocator;
9694 it might decide to use another register anyways.
9695 You may use @code{current_function_leaf_function} in the hook, functions
9696 that use @code{REG_N_SETS}, to determine if the hard
9697 register in question will not be clobbered.
9698 The default value of this hook is @code{NULL}, which disables any special
9702 @defmac TARGET_OBJECT_SUFFIX
9703 Define this macro to be a C string representing the suffix for object
9704 files on your target machine. If you do not define this macro, GCC will
9705 use @samp{.o} as the suffix for object files.
9708 @defmac TARGET_EXECUTABLE_SUFFIX
9709 Define this macro to be a C string representing the suffix to be
9710 automatically added to executable files on your target machine. If you
9711 do not define this macro, GCC will use the null string as the suffix for
9715 @defmac COLLECT_EXPORT_LIST
9716 If defined, @code{collect2} will scan the individual object files
9717 specified on its command line and create an export list for the linker.
9718 Define this macro for systems like AIX, where the linker discards
9719 object files that are not referenced from @code{main} and uses export
9723 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
9724 Define this macro to a C expression representing a variant of the
9725 method call @var{mdecl}, if Java Native Interface (JNI) methods
9726 must be invoked differently from other methods on your target.
9727 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
9728 the @code{stdcall} calling convention and this macro is then
9729 defined as this expression:
9732 build_type_attribute_variant (@var{mdecl},
9734 (get_identifier ("stdcall"),
9739 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
9740 This target hook returns @code{true} past the point in which new jump
9741 instructions could be created. On machines that require a register for
9742 every jump such as the SHmedia ISA of SH5, this point would typically be
9743 reload, so this target hook should be defined to a function such as:
9747 cannot_modify_jumps_past_reload_p ()
9749 return (reload_completed || reload_in_progress);
9754 @deftypefn {Target Hook} int TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
9755 This target hook returns a register class for which branch target register
9756 optimizations should be applied. All registers in this class should be
9757 usable interchangeably. After reload, registers in this class will be
9758 re-allocated and loads will be hoisted out of loops and be subjected
9759 to inter-block scheduling.
9762 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
9763 Branch target register optimization will by default exclude callee-saved
9765 that are not already live during the current function; if this target hook
9766 returns true, they will be included. The target code must than make sure
9767 that all target registers in the class returned by
9768 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
9769 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
9770 epilogues have already been generated. Note, even if you only return
9771 true when @var{after_prologue_epilogue_gen} is false, you still are likely
9772 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
9773 to reserve space for caller-saved target registers.
9776 @defmac POWI_MAX_MULTS
9777 If defined, this macro is interpreted as a signed integer C expression
9778 that specifies the maximum number of floating point multiplications
9779 that should be emitted when expanding exponentiation by an integer
9780 constant inline. When this value is defined, exponentiation requiring
9781 more than this number of multiplications is implemented by calling the
9782 system library's @code{pow}, @code{powf} or @code{powl} routines.
9783 The default value places no upper bound on the multiplication count.
9786 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
9787 This target hook should register any extra include files for the
9788 target. The parameter @var{stdinc} indicates if normal include files
9789 are present. The parameter @var{sysroot} is the system root directory.
9790 The parameter @var{iprefix} is the prefix for the gcc directory.
9793 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
9794 This target hook should register any extra include files for the
9795 target before any standard headers. The parameter @var{stdinc}
9796 indicates if normal include files are present. The parameter
9797 @var{sysroot} is the system root directory. The parameter
9798 @var{iprefix} is the prefix for the gcc directory.
9801 @deftypefn Macro void TARGET_OPTF (char *@var{path})
9802 This target hook should register special include paths for the target.
9803 The parameter @var{path} is the include to register. On Darwin
9804 systems, this is used for Framework includes, which have semantics
9805 that are different from @option{-I}.
9808 @deftypefn {Target Hook} bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
9809 This target hook returns @code{true} if it is safe to use a local alias
9810 for a virtual function @var{fndecl} when constructing thunks,
9811 @code{false} otherwise. By default, the hook returns @code{true} for all
9812 functions, if a target supports aliases (i.e.@: defines
9813 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
9816 @defmac TARGET_FORMAT_TYPES
9817 If defined, this macro is the name of a global variable containing
9818 target-specific format checking information for the @option{-Wformat}
9819 option. The default is to have no target-specific format checks.
9822 @defmac TARGET_N_FORMAT_TYPES
9823 If defined, this macro is the number of entries in
9824 @code{TARGET_FORMAT_TYPES}.
9827 @deftypefn {Target Hook} bool TARGET_RELAXED_ORDERING
9828 If set to @code{true}, means that the target's memory model does not
9829 guarantee that loads which do not depend on one another will access
9830 main memory in the order of the instruction stream; if ordering is
9831 important, an explicit memory barrier must be used. This is true of
9832 many recent processors which implement a policy of ``relaxed,''
9833 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
9834 and ia64. The default is @code{false}.
9837 @deftypefn {Target Hook} const char *TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (tree @var{typelist}, tree @var{funcdecl}, tree @var{val})
9838 If defined, this macro returns the diagnostic message when it is
9839 illegal to pass argument @var{val} to function @var{funcdecl}
9840 with prototype @var{typelist}.
9843 @deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (tree @var{fromtype}, tree @var{totype})
9844 If defined, this macro returns the diagnostic message when it is
9845 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
9846 if validity should be determined by the front end.
9849 @deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, tree @var{type})
9850 If defined, this macro returns the diagnostic message when it is
9851 invalid to apply operation @var{op} (where unary plus is denoted by
9852 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
9853 if validity should be determined by the front end.
9856 @deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, tree @var{type1}, tree @var{type2})
9857 If defined, this macro returns the diagnostic message when it is
9858 invalid to apply operation @var{op} to operands of types @var{type1}
9859 and @var{type2}, or @code{NULL} if validity should be determined by
9863 @defmac TARGET_USE_JCR_SECTION
9864 This macro determines whether to use the JCR section to register Java
9865 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
9866 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.