1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,
2 @c 2002, 2003, 2004, 2005 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
7 @chapter Target Description Macros and Functions
8 @cindex machine description macros
9 @cindex target description macros
10 @cindex macros, target description
11 @cindex @file{tm.h} macros
13 In addition to the file @file{@var{machine}.md}, a machine description
14 includes a C header file conventionally given the name
15 @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}.
16 The header file defines numerous macros that convey the information
17 about the target machine that does not fit into the scheme of the
18 @file{.md} file. The file @file{tm.h} should be a link to
19 @file{@var{machine}.h}. The header file @file{config.h} includes
20 @file{tm.h} and most compiler source files include @file{config.h}. The
21 source file defines a variable @code{targetm}, which is a structure
22 containing pointers to functions and data relating to the target
23 machine. @file{@var{machine}.c} should also contain their definitions,
24 if they are not defined elsewhere in GCC, and other functions called
25 through the macros defined in the @file{.h} file.
28 * Target Structure:: The @code{targetm} variable.
29 * Driver:: Controlling how the driver runs the compilation passes.
30 * Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}.
31 * Per-Function Data:: Defining data structures for per-function information.
32 * Storage Layout:: Defining sizes and alignments of data.
33 * Type Layout:: Defining sizes and properties of basic user data types.
34 * Registers:: Naming and describing the hardware registers.
35 * Register Classes:: Defining the classes of hardware registers.
36 * Stack and Calling:: Defining which way the stack grows and by how much.
37 * Varargs:: Defining the varargs macros.
38 * Trampolines:: Code set up at run time to enter a nested function.
39 * Library Calls:: Controlling how library routines are implicitly called.
40 * Addressing Modes:: Defining addressing modes valid for memory operands.
41 * Condition Code:: Defining how insns update the condition code.
42 * Costs:: Defining relative costs of different operations.
43 * Scheduling:: Adjusting the behavior of the instruction scheduler.
44 * Sections:: Dividing storage into text, data, and other sections.
45 * PIC:: Macros for position independent code.
46 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
47 * Debugging Info:: Defining the format of debugging output.
48 * Floating Point:: Handling floating point for cross-compilers.
49 * Mode Switching:: Insertion of mode-switching instructions.
50 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
51 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
52 * PCH Target:: Validity checking for precompiled headers.
53 * C++ ABI:: Controlling C++ ABI changes.
54 * Misc:: Everything else.
57 @node Target Structure
58 @section The Global @code{targetm} Variable
60 @cindex target functions
62 @deftypevar {struct gcc_target} targetm
63 The target @file{.c} file must define the global @code{targetm} variable
64 which contains pointers to functions and data relating to the target
65 machine. The variable is declared in @file{target.h};
66 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
67 used to initialize the variable, and macros for the default initializers
68 for elements of the structure. The @file{.c} file should override those
69 macros for which the default definition is inappropriate. For example:
72 #include "target-def.h"
74 /* @r{Initialize the GCC target structure.} */
76 #undef TARGET_COMP_TYPE_ATTRIBUTES
77 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
79 struct gcc_target targetm = TARGET_INITIALIZER;
83 Where a macro should be defined in the @file{.c} file in this manner to
84 form part of the @code{targetm} structure, it is documented below as a
85 ``Target Hook'' with a prototype. Many macros will change in future
86 from being defined in the @file{.h} file to being part of the
87 @code{targetm} structure.
90 @section Controlling the Compilation Driver, @file{gcc}
92 @cindex controlling the compilation driver
94 @c prevent bad page break with this line
95 You can control the compilation driver.
97 @defmac SWITCH_TAKES_ARG (@var{char})
98 A C expression which determines whether the option @option{-@var{char}}
99 takes arguments. The value should be the number of arguments that
100 option takes--zero, for many options.
102 By default, this macro is defined as
103 @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options
104 properly. You need not define @code{SWITCH_TAKES_ARG} unless you
105 wish to add additional options which take arguments. Any redefinition
106 should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for
110 @defmac WORD_SWITCH_TAKES_ARG (@var{name})
111 A C expression which determines whether the option @option{-@var{name}}
112 takes arguments. The value should be the number of arguments that
113 option takes--zero, for many options. This macro rather than
114 @code{SWITCH_TAKES_ARG} is used for multi-character option names.
116 By default, this macro is defined as
117 @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
118 properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
119 wish to add additional options which take arguments. Any redefinition
120 should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
124 @defmac SWITCH_CURTAILS_COMPILATION (@var{char})
125 A C expression which determines whether the option @option{-@var{char}}
126 stops compilation before the generation of an executable. The value is
127 boolean, nonzero if the option does stop an executable from being
128 generated, zero otherwise.
130 By default, this macro is defined as
131 @code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard
132 options properly. You need not define
133 @code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional
134 options which affect the generation of an executable. Any redefinition
135 should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check
136 for additional options.
139 @defmac SWITCHES_NEED_SPACES
140 A string-valued C expression which enumerates the options for which
141 the linker needs a space between the option and its argument.
143 If this macro is not defined, the default value is @code{""}.
146 @defmac TARGET_OPTION_TRANSLATE_TABLE
147 If defined, a list of pairs of strings, the first of which is a
148 potential command line target to the @file{gcc} driver program, and the
149 second of which is a space-separated (tabs and other whitespace are not
150 supported) list of options with which to replace the first option. The
151 target defining this list is responsible for assuring that the results
152 are valid. Replacement options may not be the @code{--opt} style, they
153 must be the @code{-opt} style. It is the intention of this macro to
154 provide a mechanism for substitution that affects the multilibs chosen,
155 such as one option that enables many options, some of which select
156 multilibs. Example nonsensical definition, where @option{-malt-abi},
157 @option{-EB}, and @option{-mspoo} cause different multilibs to be chosen:
160 #define TARGET_OPTION_TRANSLATE_TABLE \
161 @{ "-fast", "-march=fast-foo -malt-abi -I/usr/fast-foo" @}, \
162 @{ "-compat", "-EB -malign=4 -mspoo" @}
166 @defmac DRIVER_SELF_SPECS
167 A list of specs for the driver itself. It should be a suitable
168 initializer for an array of strings, with no surrounding braces.
170 The driver applies these specs to its own command line between loading
171 default @file{specs} files (but not command-line specified ones) and
172 choosing the multilib directory or running any subcommands. It
173 applies them in the order given, so each spec can depend on the
174 options added by earlier ones. It is also possible to remove options
175 using @samp{%<@var{option}} in the usual way.
177 This macro can be useful when a port has several interdependent target
178 options. It provides a way of standardizing the command line so
179 that the other specs are easier to write.
181 Do not define this macro if it does not need to do anything.
184 @defmac OPTION_DEFAULT_SPECS
185 A list of specs used to support configure-time default options (i.e.@:
186 @option{--with} options) in the driver. It should be a suitable initializer
187 for an array of structures, each containing two strings, without the
188 outermost pair of surrounding braces.
190 The first item in the pair is the name of the default. This must match
191 the code in @file{config.gcc} for the target. The second item is a spec
192 to apply if a default with this name was specified. The string
193 @samp{%(VALUE)} in the spec will be replaced by the value of the default
194 everywhere it occurs.
196 The driver will apply these specs to its own command line between loading
197 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
198 the same mechanism as @code{DRIVER_SELF_SPECS}.
200 Do not define this macro if it does not need to do anything.
204 A C string constant that tells the GCC driver program options to
205 pass to CPP@. It can also specify how to translate options you
206 give to GCC into options for GCC to pass to the CPP@.
208 Do not define this macro if it does not need to do anything.
211 @defmac CPLUSPLUS_CPP_SPEC
212 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
213 than C@. If you do not define this macro, then the value of
214 @code{CPP_SPEC} (if any) will be used instead.
218 A C string constant that tells the GCC driver program options to
219 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
221 It can also specify how to translate options you give to GCC into options
222 for GCC to pass to front ends.
224 Do not define this macro if it does not need to do anything.
228 A C string constant that tells the GCC driver program options to
229 pass to @code{cc1plus}. It can also specify how to translate options you
230 give to GCC into options for GCC to pass to the @code{cc1plus}.
232 Do not define this macro if it does not need to do anything.
233 Note that everything defined in CC1_SPEC is already passed to
234 @code{cc1plus} so there is no need to duplicate the contents of
235 CC1_SPEC in CC1PLUS_SPEC@.
239 A C string constant that tells the GCC driver program options to
240 pass to the assembler. It can also specify how to translate options
241 you give to GCC into options for GCC to pass to the assembler.
242 See the file @file{sun3.h} for an example of this.
244 Do not define this macro if it does not need to do anything.
247 @defmac ASM_FINAL_SPEC
248 A C string constant that tells the GCC driver program how to
249 run any programs which cleanup after the normal assembler.
250 Normally, this is not needed. See the file @file{mips.h} for
253 Do not define this macro if it does not need to do anything.
256 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
257 Define this macro, with no value, if the driver should give the assembler
258 an argument consisting of a single dash, @option{-}, to instruct it to
259 read from its standard input (which will be a pipe connected to the
260 output of the compiler proper). This argument is given after any
261 @option{-o} option specifying the name of the output file.
263 If you do not define this macro, the assembler is assumed to read its
264 standard input if given no non-option arguments. If your assembler
265 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
266 see @file{mips.h} for instance.
270 A C string constant that tells the GCC driver program options to
271 pass to the linker. It can also specify how to translate options you
272 give to GCC into options for GCC to pass to the linker.
274 Do not define this macro if it does not need to do anything.
278 Another C string constant used much like @code{LINK_SPEC}. The difference
279 between the two is that @code{LIB_SPEC} is used at the end of the
280 command given to the linker.
282 If this macro is not defined, a default is provided that
283 loads the standard C library from the usual place. See @file{gcc.c}.
287 Another C string constant that tells the GCC driver program
288 how and when to place a reference to @file{libgcc.a} into the
289 linker command line. This constant is placed both before and after
290 the value of @code{LIB_SPEC}.
292 If this macro is not defined, the GCC driver provides a default that
293 passes the string @option{-lgcc} to the linker.
296 @defmac REAL_LIBGCC_SPEC
297 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
298 @code{LIBGCC_SPEC} is not directly used by the driver program but is
299 instead modified to refer to different versions of @file{libgcc.a}
300 depending on the values of the command line flags @option{-static},
301 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
302 targets where these modifications are inappropriate, define
303 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
304 driver how to place a reference to @file{libgcc} on the link command
305 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
308 @defmac USE_LD_AS_NEEDED
309 A macro that controls the modifications to @code{LIBGCC_SPEC}
310 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
311 generated that uses --as-needed and the shared libgcc in place of the
312 static exception handler library, when linking without any of
313 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
317 If defined, this C string constant is added to @code{LINK_SPEC}.
318 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
319 the modifications to @code{LIBGCC_SPEC} mentioned in
320 @code{REAL_LIBGCC_SPEC}.
323 @defmac STARTFILE_SPEC
324 Another C string constant used much like @code{LINK_SPEC}. The
325 difference between the two is that @code{STARTFILE_SPEC} is used at
326 the very beginning of the command given to the linker.
328 If this macro is not defined, a default is provided that loads the
329 standard C startup file from the usual place. See @file{gcc.c}.
333 Another C string constant used much like @code{LINK_SPEC}. The
334 difference between the two is that @code{ENDFILE_SPEC} is used at
335 the very end of the command given to the linker.
337 Do not define this macro if it does not need to do anything.
340 @defmac THREAD_MODEL_SPEC
341 GCC @code{-v} will print the thread model GCC was configured to use.
342 However, this doesn't work on platforms that are multilibbed on thread
343 models, such as AIX 4.3. On such platforms, define
344 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
345 blanks that names one of the recognized thread models. @code{%*}, the
346 default value of this macro, will expand to the value of
347 @code{thread_file} set in @file{config.gcc}.
350 @defmac SYSROOT_SUFFIX_SPEC
351 Define this macro to add a suffix to the target sysroot when GCC is
352 configured with a sysroot. This will cause GCC to search for usr/lib,
353 et al, within sysroot+suffix.
356 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
357 Define this macro to add a headers_suffix to the target sysroot when
358 GCC is configured with a sysroot. This will cause GCC to pass the
359 updated sysroot+headers_suffix to CPP, causing it to search for
360 usr/include, et al, within sysroot+headers_suffix.
364 Define this macro to provide additional specifications to put in the
365 @file{specs} file that can be used in various specifications like
368 The definition should be an initializer for an array of structures,
369 containing a string constant, that defines the specification name, and a
370 string constant that provides the specification.
372 Do not define this macro if it does not need to do anything.
374 @code{EXTRA_SPECS} is useful when an architecture contains several
375 related targets, which have various @code{@dots{}_SPECS} which are similar
376 to each other, and the maintainer would like one central place to keep
379 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
380 define either @code{_CALL_SYSV} when the System V calling sequence is
381 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
384 The @file{config/rs6000/rs6000.h} target file defines:
387 #define EXTRA_SPECS \
388 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
390 #define CPP_SYS_DEFAULT ""
393 The @file{config/rs6000/sysv.h} target file defines:
397 "%@{posix: -D_POSIX_SOURCE @} \
398 %@{mcall-sysv: -D_CALL_SYSV @} \
399 %@{!mcall-sysv: %(cpp_sysv_default) @} \
400 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
402 #undef CPP_SYSV_DEFAULT
403 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
406 while the @file{config/rs6000/eabiaix.h} target file defines
407 @code{CPP_SYSV_DEFAULT} as:
410 #undef CPP_SYSV_DEFAULT
411 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
415 @defmac LINK_LIBGCC_SPECIAL
416 Define this macro if the driver program should find the library
417 @file{libgcc.a} itself and should not pass @option{-L} options to the
418 linker. If you do not define this macro, the driver program will pass
419 the argument @option{-lgcc} to tell the linker to do the search and will
420 pass @option{-L} options to it.
423 @defmac LINK_LIBGCC_SPECIAL_1
424 Define this macro if the driver program should find the library
425 @file{libgcc.a}. If you do not define this macro, the driver program will pass
426 the argument @option{-lgcc} to tell the linker to do the search.
427 This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does
428 not affect @option{-L} options.
431 @defmac LINK_GCC_C_SEQUENCE_SPEC
432 The sequence in which libgcc and libc are specified to the linker.
433 By default this is @code{%G %L %G}.
436 @defmac LINK_COMMAND_SPEC
437 A C string constant giving the complete command line need to execute the
438 linker. When you do this, you will need to update your port each time a
439 change is made to the link command line within @file{gcc.c}. Therefore,
440 define this macro only if you need to completely redefine the command
441 line for invoking the linker and there is no other way to accomplish
442 the effect you need. Overriding this macro may be avoidable by overriding
443 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
446 @defmac LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
447 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
448 directories from linking commands. Do not give it a nonzero value if
449 removing duplicate search directories changes the linker's semantics.
452 @defmac MULTILIB_DEFAULTS
453 Define this macro as a C expression for the initializer of an array of
454 string to tell the driver program which options are defaults for this
455 target and thus do not need to be handled specially when using
456 @code{MULTILIB_OPTIONS}.
458 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
459 the target makefile fragment or if none of the options listed in
460 @code{MULTILIB_OPTIONS} are set by default.
461 @xref{Target Fragment}.
464 @defmac RELATIVE_PREFIX_NOT_LINKDIR
465 Define this macro to tell @command{gcc} that it should only translate
466 a @option{-B} prefix into a @option{-L} linker option if the prefix
467 indicates an absolute file name.
470 @defmac MD_EXEC_PREFIX
471 If defined, this macro is an additional prefix to try after
472 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
473 when the @option{-b} option is used, or the compiler is built as a cross
474 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
475 to the list of directories used to find the assembler in @file{configure.in}.
478 @defmac STANDARD_STARTFILE_PREFIX
479 Define this macro as a C string constant if you wish to override the
480 standard choice of @code{libdir} as the default prefix to
481 try when searching for startup files such as @file{crt0.o}.
482 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
483 is built as a cross compiler.
486 @defmac STANDARD_STARTFILE_PREFIX_1
487 Define this macro as a C string constant if you wish to override the
488 standard choice of @code{/lib} as a prefix to try after the default prefix
489 when searching for startup files such as @file{crt0.o}.
490 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
491 is built as a cross compiler.
494 @defmac STANDARD_STARTFILE_PREFIX_2
495 Define this macro as a C string constant if you wish to override the
496 standard choice of @code{/lib} as yet another prefix to try after the
497 default prefix when searching for startup files such as @file{crt0.o}.
498 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
499 is built as a cross compiler.
502 @defmac MD_STARTFILE_PREFIX
503 If defined, this macro supplies an additional prefix to try after the
504 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
505 @option{-b} option is used, or when the compiler is built as a cross
509 @defmac MD_STARTFILE_PREFIX_1
510 If defined, this macro supplies yet another prefix to try after the
511 standard prefixes. It is not searched when the @option{-b} option is
512 used, or when the compiler is built as a cross compiler.
515 @defmac INIT_ENVIRONMENT
516 Define this macro as a C string constant if you wish to set environment
517 variables for programs called by the driver, such as the assembler and
518 loader. The driver passes the value of this macro to @code{putenv} to
519 initialize the necessary environment variables.
522 @defmac LOCAL_INCLUDE_DIR
523 Define this macro as a C string constant if you wish to override the
524 standard choice of @file{/usr/local/include} as the default prefix to
525 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
526 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
528 Cross compilers do not search either @file{/usr/local/include} or its
532 @defmac MODIFY_TARGET_NAME
533 Define this macro if you wish to define command-line switches that
534 modify the default target name.
536 For each switch, you can include a string to be appended to the first
537 part of the configuration name or a string to be deleted from the
538 configuration name, if present. The definition should be an initializer
539 for an array of structures. Each array element should have three
540 elements: the switch name (a string constant, including the initial
541 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
542 indicate whether the string should be inserted or deleted, and the string
543 to be inserted or deleted (a string constant).
545 For example, on a machine where @samp{64} at the end of the
546 configuration name denotes a 64-bit target and you want the @option{-32}
547 and @option{-64} switches to select between 32- and 64-bit targets, you would
551 #define MODIFY_TARGET_NAME \
552 @{ @{ "-32", DELETE, "64"@}, \
553 @{"-64", ADD, "64"@}@}
557 @defmac SYSTEM_INCLUDE_DIR
558 Define this macro as a C string constant if you wish to specify a
559 system-specific directory to search for header files before the standard
560 directory. @code{SYSTEM_INCLUDE_DIR} comes before
561 @code{STANDARD_INCLUDE_DIR} in the search order.
563 Cross compilers do not use this macro and do not search the directory
567 @defmac STANDARD_INCLUDE_DIR
568 Define this macro as a C string constant if you wish to override the
569 standard choice of @file{/usr/include} as the default prefix to
570 try when searching for header files.
572 Cross compilers ignore this macro and do not search either
573 @file{/usr/include} or its replacement.
576 @defmac STANDARD_INCLUDE_COMPONENT
577 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
578 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
579 If you do not define this macro, no component is used.
582 @defmac INCLUDE_DEFAULTS
583 Define this macro if you wish to override the entire default search path
584 for include files. For a native compiler, the default search path
585 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
586 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
587 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
588 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
589 and specify private search areas for GCC@. The directory
590 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
592 The definition should be an initializer for an array of structures.
593 Each array element should have four elements: the directory name (a
594 string constant), the component name (also a string constant), a flag
595 for C++-only directories,
596 and a flag showing that the includes in the directory don't need to be
597 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
598 the array with a null element.
600 The component name denotes what GNU package the include file is part of,
601 if any, in all uppercase letters. For example, it might be @samp{GCC}
602 or @samp{BINUTILS}. If the package is part of a vendor-supplied
603 operating system, code the component name as @samp{0}.
605 For example, here is the definition used for VAX/VMS:
608 #define INCLUDE_DEFAULTS \
610 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
611 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
612 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
619 Here is the order of prefixes tried for exec files:
623 Any prefixes specified by the user with @option{-B}.
626 The environment variable @code{GCC_EXEC_PREFIX}, if any.
629 The directories specified by the environment variable @code{COMPILER_PATH}.
632 The macro @code{STANDARD_EXEC_PREFIX}.
635 @file{/usr/lib/gcc/}.
638 The macro @code{MD_EXEC_PREFIX}, if any.
641 Here is the order of prefixes tried for startfiles:
645 Any prefixes specified by the user with @option{-B}.
648 The environment variable @code{GCC_EXEC_PREFIX}, if any.
651 The directories specified by the environment variable @code{LIBRARY_PATH}
652 (or port-specific name; native only, cross compilers do not use this).
655 The macro @code{STANDARD_EXEC_PREFIX}.
658 @file{/usr/lib/gcc/}.
661 The macro @code{MD_EXEC_PREFIX}, if any.
664 The macro @code{MD_STARTFILE_PREFIX}, if any.
667 The macro @code{STANDARD_STARTFILE_PREFIX}.
676 @node Run-time Target
677 @section Run-time Target Specification
678 @cindex run-time target specification
679 @cindex predefined macros
680 @cindex target specifications
682 @c prevent bad page break with this line
683 Here are run-time target specifications.
685 @defmac TARGET_CPU_CPP_BUILTINS ()
686 This function-like macro expands to a block of code that defines
687 built-in preprocessor macros and assertions for the target cpu, using
688 the functions @code{builtin_define}, @code{builtin_define_std} and
689 @code{builtin_assert}. When the front end
690 calls this macro it provides a trailing semicolon, and since it has
691 finished command line option processing your code can use those
694 @code{builtin_assert} takes a string in the form you pass to the
695 command-line option @option{-A}, such as @code{cpu=mips}, and creates
696 the assertion. @code{builtin_define} takes a string in the form
697 accepted by option @option{-D} and unconditionally defines the macro.
699 @code{builtin_define_std} takes a string representing the name of an
700 object-like macro. If it doesn't lie in the user's namespace,
701 @code{builtin_define_std} defines it unconditionally. Otherwise, it
702 defines a version with two leading underscores, and another version
703 with two leading and trailing underscores, and defines the original
704 only if an ISO standard was not requested on the command line. For
705 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
706 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
707 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
708 defines only @code{_ABI64}.
710 You can also test for the C dialect being compiled. The variable
711 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
712 or @code{clk_objective_c}. Note that if we are preprocessing
713 assembler, this variable will be @code{clk_c} but the function-like
714 macro @code{preprocessing_asm_p()} will return true, so you might want
715 to check for that first. If you need to check for strict ANSI, the
716 variable @code{flag_iso} can be used. The function-like macro
717 @code{preprocessing_trad_p()} can be used to check for traditional
721 @defmac TARGET_OS_CPP_BUILTINS ()
722 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
723 and is used for the target operating system instead.
726 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
727 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
728 and is used for the target object format. @file{elfos.h} uses this
729 macro to define @code{__ELF__}, so you probably do not need to define
733 @deftypevar {extern int} target_flags
734 This declaration should be present.
737 @cindex optional hardware or system features
738 @cindex features, optional, in system conventions
740 @defmac TARGET_@var{featurename}
741 This series of macros is to allow compiler command arguments to
742 enable or disable the use of optional features of the target machine.
743 For example, one machine description serves both the 68000 and
744 the 68020; a command argument tells the compiler whether it should
745 use 68020-only instructions or not. This command argument works
746 by means of a macro @code{TARGET_68020} that tests a bit in
749 Define a macro @code{TARGET_@var{featurename}} for each such option.
750 Its definition should test a bit in @code{target_flags}. It is
751 recommended that a helper macro @code{MASK_@var{featurename}}
752 is defined for each bit-value to test, and used in
753 @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For
757 #define TARGET_MASK_68020 1
758 #define TARGET_68020 (target_flags & MASK_68020)
761 One place where these macros are used is in the condition-expressions
762 of instruction patterns. Note how @code{TARGET_68020} appears
763 frequently in the 68000 machine description file, @file{m68k.md}.
764 Another place they are used is in the definitions of the other
765 macros in the @file{@var{machine}.h} file.
768 @defmac TARGET_SWITCHES
769 This macro defines names of command options to set and clear
770 bits in @code{target_flags}. Its definition is an initializer
771 with a subgrouping for each command option.
773 Each subgrouping contains a string constant, that defines the option
774 name, a number, which contains the bits to set in
775 @code{target_flags}, and a second string which is the description
776 displayed by @option{--help}. If the number is negative then the bits specified
777 by the number are cleared instead of being set. If the description
778 string is present but empty, then no help information will be displayed
779 for that option, but it will not count as an undocumented option. The
780 actual option name is made by appending @samp{-m} to the specified name.
781 Non-empty description strings should be marked with @code{N_(@dots{})} for
782 @command{xgettext}. Please do not mark empty strings because the empty
783 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
784 of the message catalog with meta information, not the empty string.
786 In addition to the description for @option{--help},
787 more detailed documentation for each option should be added to
790 One of the subgroupings should have a null string. The number in
791 this grouping is the default value for @code{target_flags}. Any
792 target options act starting with that value.
794 Here is an example which defines @option{-m68000} and @option{-m68020}
795 with opposite meanings, and picks the latter as the default:
798 #define TARGET_SWITCHES \
799 @{ @{ "68020", MASK_68020, "" @}, \
800 @{ "68000", -MASK_68020, \
801 N_("Compile for the 68000") @}, \
802 @{ "", MASK_68020, "" @}, \
807 @defmac TARGET_OPTIONS
808 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
809 options that have values. Its definition is an initializer with a
810 subgrouping for each command option.
812 Each subgrouping contains a string constant, that defines the option
813 name, the address of a variable, a description string, and a value.
814 Non-empty description strings should be marked with @code{N_(@dots{})}
815 for @command{xgettext}. Please do not mark empty strings because the
816 empty string is reserved by GNU gettext. @code{gettext("")} returns the
817 header entry of the message catalog with meta information, not the empty
820 If the value listed in the table is @code{NULL}, then the variable, type
821 @code{char *}, is set to the variable part of the given option if the
822 fixed part matches. In other words, if the first part of the option
823 matches what's in the table, the variable will be set to point to the
824 rest of the option. This allows the user to specify a value for that
825 option. The actual option name is made by appending @samp{-m} to the
826 specified name. Again, each option should also be documented in
829 If the value listed in the table is non-@code{NULL}, then the option
830 must match the option in the table exactly (with @samp{-m}), and the
831 variable is set to point to the value listed in the table.
833 Here is an example which defines @option{-mshort-data-@var{number}}. If the
834 given option is @option{-mshort-data-512}, the variable @code{m88k_short_data}
835 will be set to the string @code{"512"}.
838 extern char *m88k_short_data;
839 #define TARGET_OPTIONS \
840 @{ @{ "short-data-", &m88k_short_data, \
841 N_("Specify the size of the short data section"), 0 @} @}
844 Here is a variant of the above that allows the user to also specify
845 just @option{-mshort-data} where a default of @code{"64"} is used.
848 extern char *m88k_short_data;
849 #define TARGET_OPTIONS \
850 @{ @{ "short-data-", &m88k_short_data, \
851 N_("Specify the size of the short data section"), 0 @} \
852 @{ "short-data", &m88k_short_data, "", "64" @},
856 Here is an example which defines @option{-mno-alu}, @option{-malu1}, and
857 @option{-malu2} as a three-state switch, along with suitable macros for
858 checking the state of the option (documentation is elided for brevity).
862 char *chip_alu = ""; /* Specify default here. */
865 extern char *chip_alu;
866 #define TARGET_OPTIONS \
867 @{ @{ "no-alu", &chip_alu, "", "" @}, \
868 @{ "alu1", &chip_alu, "", "1" @}, \
869 @{ "alu2", &chip_alu, "", "2" @}, @}
870 #define TARGET_ALU (chip_alu[0] != '\0')
871 #define TARGET_ALU1 (chip_alu[0] == '1')
872 #define TARGET_ALU2 (chip_alu[0] == '2')
876 @defmac TARGET_VERSION
877 This macro is a C statement to print on @code{stderr} a string
878 describing the particular machine description choice. Every machine
879 description should define @code{TARGET_VERSION}. For example:
883 #define TARGET_VERSION \
884 fprintf (stderr, " (68k, Motorola syntax)");
886 #define TARGET_VERSION \
887 fprintf (stderr, " (68k, MIT syntax)");
892 @defmac OVERRIDE_OPTIONS
893 Sometimes certain combinations of command options do not make sense on
894 a particular target machine. You can define a macro
895 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
896 defined, is executed once just after all the command options have been
899 Don't use this macro to turn on various extra optimizations for
900 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
903 @defmac OPTIMIZATION_OPTIONS (@var{level}, @var{size})
904 Some machines may desire to change what optimizations are performed for
905 various optimization levels. This macro, if defined, is executed once
906 just after the optimization level is determined and before the remainder
907 of the command options have been parsed. Values set in this macro are
908 used as the default values for the other command line options.
910 @var{level} is the optimization level specified; 2 if @option{-O2} is
911 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
913 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
915 You should not use this macro to change options that are not
916 machine-specific. These should uniformly selected by the same
917 optimization level on all supported machines. Use this macro to enable
918 machine-specific optimizations.
920 @strong{Do not examine @code{write_symbols} in
921 this macro!} The debugging options are not supposed to alter the
925 @defmac CAN_DEBUG_WITHOUT_FP
926 Define this macro if debugging can be performed even without a frame
927 pointer. If this macro is defined, GCC will turn on the
928 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
931 @node Per-Function Data
932 @section Defining data structures for per-function information.
933 @cindex per-function data
934 @cindex data structures
936 If the target needs to store information on a per-function basis, GCC
937 provides a macro and a couple of variables to allow this. Note, just
938 using statics to store the information is a bad idea, since GCC supports
939 nested functions, so you can be halfway through encoding one function
940 when another one comes along.
942 GCC defines a data structure called @code{struct function} which
943 contains all of the data specific to an individual function. This
944 structure contains a field called @code{machine} whose type is
945 @code{struct machine_function *}, which can be used by targets to point
946 to their own specific data.
948 If a target needs per-function specific data it should define the type
949 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
950 This macro should be used to initialize the function pointer
951 @code{init_machine_status}. This pointer is explained below.
953 One typical use of per-function, target specific data is to create an
954 RTX to hold the register containing the function's return address. This
955 RTX can then be used to implement the @code{__builtin_return_address}
956 function, for level 0.
958 Note---earlier implementations of GCC used a single data area to hold
959 all of the per-function information. Thus when processing of a nested
960 function began the old per-function data had to be pushed onto a
961 stack, and when the processing was finished, it had to be popped off the
962 stack. GCC used to provide function pointers called
963 @code{save_machine_status} and @code{restore_machine_status} to handle
964 the saving and restoring of the target specific information. Since the
965 single data area approach is no longer used, these pointers are no
968 @defmac INIT_EXPANDERS
969 Macro called to initialize any target specific information. This macro
970 is called once per function, before generation of any RTL has begun.
971 The intention of this macro is to allow the initialization of the
972 function pointer @code{init_machine_status}.
975 @deftypevar {void (*)(struct function *)} init_machine_status
976 If this function pointer is non-@code{NULL} it will be called once per
977 function, before function compilation starts, in order to allow the
978 target to perform any target specific initialization of the
979 @code{struct function} structure. It is intended that this would be
980 used to initialize the @code{machine} of that structure.
982 @code{struct machine_function} structures are expected to be freed by GC@.
983 Generally, any memory that they reference must be allocated by using
984 @code{ggc_alloc}, including the structure itself.
988 @section Storage Layout
989 @cindex storage layout
991 Note that the definitions of the macros in this table which are sizes or
992 alignments measured in bits do not need to be constant. They can be C
993 expressions that refer to static variables, such as the @code{target_flags}.
994 @xref{Run-time Target}.
996 @defmac BITS_BIG_ENDIAN
997 Define this macro to have the value 1 if the most significant bit in a
998 byte has the lowest number; otherwise define it to have the value zero.
999 This means that bit-field instructions count from the most significant
1000 bit. If the machine has no bit-field instructions, then this must still
1001 be defined, but it doesn't matter which value it is defined to. This
1002 macro need not be a constant.
1004 This macro does not affect the way structure fields are packed into
1005 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
1008 @defmac BYTES_BIG_ENDIAN
1009 Define this macro to have the value 1 if the most significant byte in a
1010 word has the lowest number. This macro need not be a constant.
1013 @defmac WORDS_BIG_ENDIAN
1014 Define this macro to have the value 1 if, in a multiword object, the
1015 most significant word has the lowest number. This applies to both
1016 memory locations and registers; GCC fundamentally assumes that the
1017 order of words in memory is the same as the order in registers. This
1018 macro need not be a constant.
1021 @defmac LIBGCC2_WORDS_BIG_ENDIAN
1022 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
1023 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
1024 used only when compiling @file{libgcc2.c}. Typically the value will be set
1025 based on preprocessor defines.
1028 @defmac FLOAT_WORDS_BIG_ENDIAN
1029 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
1030 @code{TFmode} floating point numbers are stored in memory with the word
1031 containing the sign bit at the lowest address; otherwise define it to
1032 have the value 0. This macro need not be a constant.
1034 You need not define this macro if the ordering is the same as for
1035 multi-word integers.
1038 @defmac BITS_PER_UNIT
1039 Define this macro to be the number of bits in an addressable storage
1040 unit (byte). If you do not define this macro the default is 8.
1043 @defmac BITS_PER_WORD
1044 Number of bits in a word. If you do not define this macro, the default
1045 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
1048 @defmac MAX_BITS_PER_WORD
1049 Maximum number of bits in a word. If this is undefined, the default is
1050 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
1051 largest value that @code{BITS_PER_WORD} can have at run-time.
1054 @defmac UNITS_PER_WORD
1055 Number of storage units in a word; normally 4.
1058 @defmac MIN_UNITS_PER_WORD
1059 Minimum number of units in a word. If this is undefined, the default is
1060 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
1061 smallest value that @code{UNITS_PER_WORD} can have at run-time.
1064 @defmac POINTER_SIZE
1065 Width of a pointer, in bits. You must specify a value no wider than the
1066 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
1067 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
1068 a value the default is @code{BITS_PER_WORD}.
1071 @defmac POINTERS_EXTEND_UNSIGNED
1072 A C expression whose value is greater than zero if pointers that need to be
1073 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
1074 be zero-extended and zero if they are to be sign-extended. If the value
1075 is less then zero then there must be an "ptr_extend" instruction that
1076 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
1078 You need not define this macro if the @code{POINTER_SIZE} is equal
1079 to the width of @code{Pmode}.
1082 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
1083 A macro to update @var{m} and @var{unsignedp} when an object whose type
1084 is @var{type} and which has the specified mode and signedness is to be
1085 stored in a register. This macro is only called when @var{type} is a
1088 On most RISC machines, which only have operations that operate on a full
1089 register, define this macro to set @var{m} to @code{word_mode} if
1090 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
1091 cases, only integer modes should be widened because wider-precision
1092 floating-point operations are usually more expensive than their narrower
1095 For most machines, the macro definition does not change @var{unsignedp}.
1096 However, some machines, have instructions that preferentially handle
1097 either signed or unsigned quantities of certain modes. For example, on
1098 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
1099 sign-extend the result to 64 bits. On such machines, set
1100 @var{unsignedp} according to which kind of extension is more efficient.
1102 Do not define this macro if it would never modify @var{m}.
1105 @defmac PROMOTE_FUNCTION_MODE
1106 Like @code{PROMOTE_MODE}, but is applied to outgoing function arguments or
1107 function return values, as specified by @code{TARGET_PROMOTE_FUNCTION_ARGS}
1108 and @code{TARGET_PROMOTE_FUNCTION_RETURN}, respectively.
1110 The default is @code{PROMOTE_MODE}.
1113 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_ARGS (tree @var{fntype})
1114 This target hook should return @code{true} if the promotion described by
1115 @code{PROMOTE_FUNCTION_MODE} should be done for outgoing function
1119 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_RETURN (tree @var{fntype})
1120 This target hook should return @code{true} if the promotion described by
1121 @code{PROMOTE_FUNCTION_MODE} should be done for the return value of
1124 If this target hook returns @code{true}, @code{FUNCTION_VALUE} must
1125 perform the same promotions done by @code{PROMOTE_FUNCTION_MODE}.
1128 @defmac PARM_BOUNDARY
1129 Normal alignment required for function parameters on the stack, in
1130 bits. All stack parameters receive at least this much alignment
1131 regardless of data type. On most machines, this is the same as the
1135 @defmac STACK_BOUNDARY
1136 Define this macro to the minimum alignment enforced by hardware for the
1137 stack pointer on this machine. The definition is a C expression for the
1138 desired alignment (measured in bits). This value is used as a default
1139 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1140 this should be the same as @code{PARM_BOUNDARY}.
1143 @defmac PREFERRED_STACK_BOUNDARY
1144 Define this macro if you wish to preserve a certain alignment for the
1145 stack pointer, greater than what the hardware enforces. The definition
1146 is a C expression for the desired alignment (measured in bits). This
1147 macro must evaluate to a value equal to or larger than
1148 @code{STACK_BOUNDARY}.
1151 @defmac FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1152 A C expression that evaluates true if @code{PREFERRED_STACK_BOUNDARY} is
1153 not guaranteed by the runtime and we should emit code to align the stack
1154 at the beginning of @code{main}.
1156 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
1157 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
1158 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
1159 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
1160 be momentarily unaligned while pushing arguments.
1163 @defmac FUNCTION_BOUNDARY
1164 Alignment required for a function entry point, in bits.
1167 @defmac BIGGEST_ALIGNMENT
1168 Biggest alignment that any data type can require on this machine, in bits.
1171 @defmac MINIMUM_ATOMIC_ALIGNMENT
1172 If defined, the smallest alignment, in bits, that can be given to an
1173 object that can be referenced in one operation, without disturbing any
1174 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1175 on machines that don't have byte or half-word store operations.
1178 @defmac BIGGEST_FIELD_ALIGNMENT
1179 Biggest alignment that any structure or union field can require on this
1180 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1181 structure and union fields only, unless the field alignment has been set
1182 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1185 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1186 An expression for the alignment of a structure field @var{field} if the
1187 alignment computed in the usual way (including applying of
1188 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1189 alignment) is @var{computed}. It overrides alignment only if the
1190 field alignment has not been set by the
1191 @code{__attribute__ ((aligned (@var{n})))} construct.
1194 @defmac MAX_OFILE_ALIGNMENT
1195 Biggest alignment supported by the object file format of this machine.
1196 Use this macro to limit the alignment which can be specified using the
1197 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1198 the default value is @code{BIGGEST_ALIGNMENT}.
1201 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1202 If defined, a C expression to compute the alignment for a variable in
1203 the static store. @var{type} is the data type, and @var{basic-align} is
1204 the alignment that the object would ordinarily have. The value of this
1205 macro is used instead of that alignment to align the object.
1207 If this macro is not defined, then @var{basic-align} is used.
1210 One use of this macro is to increase alignment of medium-size data to
1211 make it all fit in fewer cache lines. Another is to cause character
1212 arrays to be word-aligned so that @code{strcpy} calls that copy
1213 constants to character arrays can be done inline.
1216 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1217 If defined, a C expression to compute the alignment given to a constant
1218 that is being placed in memory. @var{constant} is the constant and
1219 @var{basic-align} is the alignment that the object would ordinarily
1220 have. The value of this macro is used instead of that alignment to
1223 If this macro is not defined, then @var{basic-align} is used.
1225 The typical use of this macro is to increase alignment for string
1226 constants to be word aligned so that @code{strcpy} calls that copy
1227 constants can be done inline.
1230 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1231 If defined, a C expression to compute the alignment for a variable in
1232 the local store. @var{type} is the data type, and @var{basic-align} is
1233 the alignment that the object would ordinarily have. The value of this
1234 macro is used instead of that alignment to align the object.
1236 If this macro is not defined, then @var{basic-align} is used.
1238 One use of this macro is to increase alignment of medium-size data to
1239 make it all fit in fewer cache lines.
1242 @defmac EMPTY_FIELD_BOUNDARY
1243 Alignment in bits to be given to a structure bit-field that follows an
1244 empty field such as @code{int : 0;}.
1246 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1249 @defmac STRUCTURE_SIZE_BOUNDARY
1250 Number of bits which any structure or union's size must be a multiple of.
1251 Each structure or union's size is rounded up to a multiple of this.
1253 If you do not define this macro, the default is the same as
1254 @code{BITS_PER_UNIT}.
1257 @defmac STRICT_ALIGNMENT
1258 Define this macro to be the value 1 if instructions will fail to work
1259 if given data not on the nominal alignment. If instructions will merely
1260 go slower in that case, define this macro as 0.
1263 @defmac PCC_BITFIELD_TYPE_MATTERS
1264 Define this if you wish to imitate the way many other C compilers handle
1265 alignment of bit-fields and the structures that contain them.
1267 The behavior is that the type written for a named bit-field (@code{int},
1268 @code{short}, or other integer type) imposes an alignment for the entire
1269 structure, as if the structure really did contain an ordinary field of
1270 that type. In addition, the bit-field is placed within the structure so
1271 that it would fit within such a field, not crossing a boundary for it.
1273 Thus, on most machines, a named bit-field whose type is written as
1274 @code{int} would not cross a four-byte boundary, and would force
1275 four-byte alignment for the whole structure. (The alignment used may
1276 not be four bytes; it is controlled by the other alignment parameters.)
1278 An unnamed bit-field will not affect the alignment of the containing
1281 If the macro is defined, its definition should be a C expression;
1282 a nonzero value for the expression enables this behavior.
1284 Note that if this macro is not defined, or its value is zero, some
1285 bit-fields may cross more than one alignment boundary. The compiler can
1286 support such references if there are @samp{insv}, @samp{extv}, and
1287 @samp{extzv} insns that can directly reference memory.
1289 The other known way of making bit-fields work is to define
1290 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1291 Then every structure can be accessed with fullwords.
1293 Unless the machine has bit-field instructions or you define
1294 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1295 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1297 If your aim is to make GCC use the same conventions for laying out
1298 bit-fields as are used by another compiler, here is how to investigate
1299 what the other compiler does. Compile and run this program:
1318 printf ("Size of foo1 is %d\n",
1319 sizeof (struct foo1));
1320 printf ("Size of foo2 is %d\n",
1321 sizeof (struct foo2));
1326 If this prints 2 and 5, then the compiler's behavior is what you would
1327 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1330 @defmac BITFIELD_NBYTES_LIMITED
1331 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1332 to aligning a bit-field within the structure.
1335 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELDS (void)
1336 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1337 whether unnamed bitfields affect the alignment of the containing
1338 structure. The hook should return true if the structure should inherit
1339 the alignment requirements of an unnamed bitfield's type.
1342 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1343 Return 1 if a structure or array containing @var{field} should be accessed using
1346 If @var{field} is the only field in the structure, @var{mode} is its
1347 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1348 case where structures of one field would require the structure's mode to
1349 retain the field's mode.
1351 Normally, this is not needed. See the file @file{c4x.h} for an example
1352 of how to use this macro to prevent a structure having a floating point
1353 field from being accessed in an integer mode.
1356 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1357 Define this macro as an expression for the alignment of a type (given
1358 by @var{type} as a tree node) if the alignment computed in the usual
1359 way is @var{computed} and the alignment explicitly specified was
1362 The default is to use @var{specified} if it is larger; otherwise, use
1363 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1366 @defmac MAX_FIXED_MODE_SIZE
1367 An integer expression for the size in bits of the largest integer
1368 machine mode that should actually be used. All integer machine modes of
1369 this size or smaller can be used for structures and unions with the
1370 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1371 (DImode)} is assumed.
1374 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1375 If defined, an expression of type @code{enum machine_mode} that
1376 specifies the mode of the save area operand of a
1377 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1378 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1379 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1380 having its mode specified.
1382 You need not define this macro if it always returns @code{Pmode}. You
1383 would most commonly define this macro if the
1384 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1388 @defmac STACK_SIZE_MODE
1389 If defined, an expression of type @code{enum machine_mode} that
1390 specifies the mode of the size increment operand of an
1391 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1393 You need not define this macro if it always returns @code{word_mode}.
1394 You would most commonly define this macro if the @code{allocate_stack}
1395 pattern needs to support both a 32- and a 64-bit mode.
1398 @defmac TARGET_FLOAT_FORMAT
1399 A code distinguishing the floating point format of the target machine.
1400 There are four defined values:
1403 @item IEEE_FLOAT_FORMAT
1404 This code indicates IEEE floating point. It is the default; there is no
1405 need to define @code{TARGET_FLOAT_FORMAT} when the format is IEEE@.
1407 @item VAX_FLOAT_FORMAT
1408 This code indicates the ``F float'' (for @code{float}) and ``D float''
1409 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
1411 @item IBM_FLOAT_FORMAT
1412 This code indicates the format used on the IBM System/370.
1414 @item C4X_FLOAT_FORMAT
1415 This code indicates the format used on the TMS320C3x/C4x.
1418 If your target uses a floating point format other than these, you must
1419 define a new @var{name}_FLOAT_FORMAT code for it, and add support for
1420 it to @file{real.c}.
1422 The ordering of the component words of floating point values stored in
1423 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1426 @defmac MODE_HAS_NANS (@var{mode})
1427 When defined, this macro should be true if @var{mode} has a NaN
1428 representation. The compiler assumes that NaNs are not equal to
1429 anything (including themselves) and that addition, subtraction,
1430 multiplication and division all return NaNs when one operand is
1433 By default, this macro is true if @var{mode} is a floating-point
1434 mode and the target floating-point format is IEEE@.
1437 @defmac MODE_HAS_INFINITIES (@var{mode})
1438 This macro should be true if @var{mode} can represent infinity. At
1439 present, the compiler uses this macro to decide whether @samp{x - x}
1440 is always defined. By default, the macro is true when @var{mode}
1441 is a floating-point mode and the target format is IEEE@.
1444 @defmac MODE_HAS_SIGNED_ZEROS (@var{mode})
1445 True if @var{mode} distinguishes between positive and negative zero.
1446 The rules are expected to follow the IEEE standard:
1450 @samp{x + x} has the same sign as @samp{x}.
1453 If the sum of two values with opposite sign is zero, the result is
1454 positive for all rounding modes expect towards @minus{}infinity, for
1455 which it is negative.
1458 The sign of a product or quotient is negative when exactly one
1459 of the operands is negative.
1462 The default definition is true if @var{mode} is a floating-point
1463 mode and the target format is IEEE@.
1466 @defmac MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1467 If defined, this macro should be true for @var{mode} if it has at
1468 least one rounding mode in which @samp{x} and @samp{-x} can be
1469 rounded to numbers of different magnitude. Two such modes are
1470 towards @minus{}infinity and towards +infinity.
1472 The default definition of this macro is true if @var{mode} is
1473 a floating-point mode and the target format is IEEE@.
1476 @defmac ROUND_TOWARDS_ZERO
1477 If defined, this macro should be true if the prevailing rounding
1478 mode is towards zero. A true value has the following effects:
1482 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1485 @file{libgcc.a}'s floating-point emulator will round towards zero
1486 rather than towards nearest.
1489 The compiler's floating-point emulator will round towards zero after
1490 doing arithmetic, and when converting from the internal float format to
1494 The macro does not affect the parsing of string literals. When the
1495 primary rounding mode is towards zero, library functions like
1496 @code{strtod} might still round towards nearest, and the compiler's
1497 parser should behave like the target's @code{strtod} where possible.
1499 Not defining this macro is equivalent to returning zero.
1502 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1503 This macro should return true if floats with @var{size}
1504 bits do not have a NaN or infinity representation, but use the largest
1505 exponent for normal numbers instead.
1507 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1508 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1509 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1510 floating-point arithmetic.
1512 The default definition of this macro returns false for all sizes.
1515 @deftypefn {Target Hook} bool TARGET_VECTOR_OPAQUE_P (tree @var{type})
1516 This target hook should return @code{true} a vector is opaque. That
1517 is, if no cast is needed when copying a vector value of type
1518 @var{type} into another vector lvalue of the same size. Vector opaque
1519 types cannot be initialized. The default is that there are no such
1523 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1524 This target hook returns @code{true} if bit-fields in the given
1525 @var{record_type} are to be laid out following the rules of Microsoft
1526 Visual C/C++, namely: (i) a bit-field won't share the same storage
1527 unit with the previous bit-field if their underlying types have
1528 different sizes, and the bit-field will be aligned to the highest
1529 alignment of the underlying types of itself and of the previous
1530 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1531 the whole enclosing structure, even if it is unnamed; except that
1532 (iii) a zero-sized bit-field will be disregarded unless it follows
1533 another bit-field of nonzero size. If this hook returns @code{true},
1534 other macros that control bit-field layout are ignored.
1536 When a bit-field is inserted into a packed record, the whole size
1537 of the underlying type is used by one or more same-size adjacent
1538 bit-fields (that is, if its long:3, 32 bits is used in the record,
1539 and any additional adjacent long bit-fields are packed into the same
1540 chunk of 32 bits. However, if the size changes, a new field of that
1541 size is allocated). In an unpacked record, this is the same as using
1542 alignment, but not equivalent when packing.
1544 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1545 the latter will take precedence. If @samp{__attribute__((packed))} is
1546 used on a single field when MS bit-fields are in use, it will take
1547 precedence for that field, but the alignment of the rest of the structure
1548 may affect its placement.
1551 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_FUNDAMENTAL_TYPE (tree @var{type})
1552 If your target defines any fundamental types, define this hook to
1553 return the appropriate encoding for these types as part of a C++
1554 mangled name. The @var{type} argument is the tree structure
1555 representing the type to be mangled. The hook may be applied to trees
1556 which are not target-specific fundamental types; it should return
1557 @code{NULL} for all such types, as well as arguments it does not
1558 recognize. If the return value is not @code{NULL}, it must point to
1559 a statically-allocated string constant.
1561 Target-specific fundamental types might be new fundamental types or
1562 qualified versions of ordinary fundamental types. Encode new
1563 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1564 is the name used for the type in source code, and @var{n} is the
1565 length of @var{name} in decimal. Encode qualified versions of
1566 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1567 @var{name} is the name used for the type qualifier in source code,
1568 @var{n} is the length of @var{name} as above, and @var{code} is the
1569 code used to represent the unqualified version of this type. (See
1570 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1571 codes.) In both cases the spaces are for clarity; do not include any
1572 spaces in your string.
1574 The default version of this hook always returns @code{NULL}, which is
1575 appropriate for a target that does not define any new fundamental
1580 @section Layout of Source Language Data Types
1582 These macros define the sizes and other characteristics of the standard
1583 basic data types used in programs being compiled. Unlike the macros in
1584 the previous section, these apply to specific features of C and related
1585 languages, rather than to fundamental aspects of storage layout.
1587 @defmac INT_TYPE_SIZE
1588 A C expression for the size in bits of the type @code{int} on the
1589 target machine. If you don't define this, the default is one word.
1592 @defmac SHORT_TYPE_SIZE
1593 A C expression for the size in bits of the type @code{short} on the
1594 target machine. If you don't define this, the default is half a word.
1595 (If this would be less than one storage unit, it is rounded up to one
1599 @defmac LONG_TYPE_SIZE
1600 A C expression for the size in bits of the type @code{long} on the
1601 target machine. If you don't define this, the default is one word.
1604 @defmac ADA_LONG_TYPE_SIZE
1605 On some machines, the size used for the Ada equivalent of the type
1606 @code{long} by a native Ada compiler differs from that used by C@. In
1607 that situation, define this macro to be a C expression to be used for
1608 the size of that type. If you don't define this, the default is the
1609 value of @code{LONG_TYPE_SIZE}.
1612 @defmac LONG_LONG_TYPE_SIZE
1613 A C expression for the size in bits of the type @code{long long} on the
1614 target machine. If you don't define this, the default is two
1615 words. If you want to support GNU Ada on your machine, the value of this
1616 macro must be at least 64.
1619 @defmac CHAR_TYPE_SIZE
1620 A C expression for the size in bits of the type @code{char} on the
1621 target machine. If you don't define this, the default is
1622 @code{BITS_PER_UNIT}.
1625 @defmac BOOL_TYPE_SIZE
1626 A C expression for the size in bits of the C++ type @code{bool} and
1627 C99 type @code{_Bool} on the target machine. If you don't define
1628 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1631 @defmac FLOAT_TYPE_SIZE
1632 A C expression for the size in bits of the type @code{float} on the
1633 target machine. If you don't define this, the default is one word.
1636 @defmac DOUBLE_TYPE_SIZE
1637 A C expression for the size in bits of the type @code{double} on the
1638 target machine. If you don't define this, the default is two
1642 @defmac LONG_DOUBLE_TYPE_SIZE
1643 A C expression for the size in bits of the type @code{long double} on
1644 the target machine. If you don't define this, the default is two
1648 @defmac TARGET_FLT_EVAL_METHOD
1649 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1650 assuming, if applicable, that the floating-point control word is in its
1651 default state. If you do not define this macro the value of
1652 @code{FLT_EVAL_METHOD} will be zero.
1655 @defmac WIDEST_HARDWARE_FP_SIZE
1656 A C expression for the size in bits of the widest floating-point format
1657 supported by the hardware. If you define this macro, you must specify a
1658 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1659 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1663 @defmac DEFAULT_SIGNED_CHAR
1664 An expression whose value is 1 or 0, according to whether the type
1665 @code{char} should be signed or unsigned by default. The user can
1666 always override this default with the options @option{-fsigned-char}
1667 and @option{-funsigned-char}.
1670 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1671 This target hook should return true if the compiler should give an
1672 @code{enum} type only as many bytes as it takes to represent the range
1673 of possible values of that type. It should return false if all
1674 @code{enum} types should be allocated like @code{int}.
1676 The default is to return false.
1680 A C expression for a string describing the name of the data type to use
1681 for size values. The typedef name @code{size_t} is defined using the
1682 contents of the string.
1684 The string can contain more than one keyword. If so, separate them with
1685 spaces, and write first any length keyword, then @code{unsigned} if
1686 appropriate, and finally @code{int}. The string must exactly match one
1687 of the data type names defined in the function
1688 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1689 omit @code{int} or change the order---that would cause the compiler to
1692 If you don't define this macro, the default is @code{"long unsigned
1696 @defmac PTRDIFF_TYPE
1697 A C expression for a string describing the name of the data type to use
1698 for the result of subtracting two pointers. The typedef name
1699 @code{ptrdiff_t} is defined using the contents of the string. See
1700 @code{SIZE_TYPE} above for more information.
1702 If you don't define this macro, the default is @code{"long int"}.
1706 A C expression for a string describing the name of the data type to use
1707 for wide characters. The typedef name @code{wchar_t} is defined using
1708 the contents of the string. See @code{SIZE_TYPE} above for more
1711 If you don't define this macro, the default is @code{"int"}.
1714 @defmac WCHAR_TYPE_SIZE
1715 A C expression for the size in bits of the data type for wide
1716 characters. This is used in @code{cpp}, which cannot make use of
1721 A C expression for a string describing the name of the data type to
1722 use for wide characters passed to @code{printf} and returned from
1723 @code{getwc}. The typedef name @code{wint_t} is defined using the
1724 contents of the string. See @code{SIZE_TYPE} above for more
1727 If you don't define this macro, the default is @code{"unsigned int"}.
1731 A C expression for a string describing the name of the data type that
1732 can represent any value of any standard or extended signed integer type.
1733 The typedef name @code{intmax_t} is defined using the contents of the
1734 string. See @code{SIZE_TYPE} above for more information.
1736 If you don't define this macro, the default is the first of
1737 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1738 much precision as @code{long long int}.
1741 @defmac UINTMAX_TYPE
1742 A C expression for a string describing the name of the data type that
1743 can represent any value of any standard or extended unsigned integer
1744 type. The typedef name @code{uintmax_t} is defined using the contents
1745 of the string. See @code{SIZE_TYPE} above for more information.
1747 If you don't define this macro, the default is the first of
1748 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1749 unsigned int"} that has as much precision as @code{long long unsigned
1753 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1754 The C++ compiler represents a pointer-to-member-function with a struct
1761 ptrdiff_t vtable_index;
1768 The C++ compiler must use one bit to indicate whether the function that
1769 will be called through a pointer-to-member-function is virtual.
1770 Normally, we assume that the low-order bit of a function pointer must
1771 always be zero. Then, by ensuring that the vtable_index is odd, we can
1772 distinguish which variant of the union is in use. But, on some
1773 platforms function pointers can be odd, and so this doesn't work. In
1774 that case, we use the low-order bit of the @code{delta} field, and shift
1775 the remainder of the @code{delta} field to the left.
1777 GCC will automatically make the right selection about where to store
1778 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1779 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1780 set such that functions always start at even addresses, but the lowest
1781 bit of pointers to functions indicate whether the function at that
1782 address is in ARM or Thumb mode. If this is the case of your
1783 architecture, you should define this macro to
1784 @code{ptrmemfunc_vbit_in_delta}.
1786 In general, you should not have to define this macro. On architectures
1787 in which function addresses are always even, according to
1788 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1789 @code{ptrmemfunc_vbit_in_pfn}.
1792 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1793 Normally, the C++ compiler uses function pointers in vtables. This
1794 macro allows the target to change to use ``function descriptors''
1795 instead. Function descriptors are found on targets for whom a
1796 function pointer is actually a small data structure. Normally the
1797 data structure consists of the actual code address plus a data
1798 pointer to which the function's data is relative.
1800 If vtables are used, the value of this macro should be the number
1801 of words that the function descriptor occupies.
1804 @defmac TARGET_VTABLE_ENTRY_ALIGN
1805 By default, the vtable entries are void pointers, the so the alignment
1806 is the same as pointer alignment. The value of this macro specifies
1807 the alignment of the vtable entry in bits. It should be defined only
1808 when special alignment is necessary. */
1811 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1812 There are a few non-descriptor entries in the vtable at offsets below
1813 zero. If these entries must be padded (say, to preserve the alignment
1814 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1815 of words in each data entry.
1819 @section Register Usage
1820 @cindex register usage
1822 This section explains how to describe what registers the target machine
1823 has, and how (in general) they can be used.
1825 The description of which registers a specific instruction can use is
1826 done with register classes; see @ref{Register Classes}. For information
1827 on using registers to access a stack frame, see @ref{Frame Registers}.
1828 For passing values in registers, see @ref{Register Arguments}.
1829 For returning values in registers, see @ref{Scalar Return}.
1832 * Register Basics:: Number and kinds of registers.
1833 * Allocation Order:: Order in which registers are allocated.
1834 * Values in Registers:: What kinds of values each reg can hold.
1835 * Leaf Functions:: Renumbering registers for leaf functions.
1836 * Stack Registers:: Handling a register stack such as 80387.
1839 @node Register Basics
1840 @subsection Basic Characteristics of Registers
1842 @c prevent bad page break with this line
1843 Registers have various characteristics.
1845 @defmac FIRST_PSEUDO_REGISTER
1846 Number of hardware registers known to the compiler. They receive
1847 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1848 pseudo register's number really is assigned the number
1849 @code{FIRST_PSEUDO_REGISTER}.
1852 @defmac FIXED_REGISTERS
1853 @cindex fixed register
1854 An initializer that says which registers are used for fixed purposes
1855 all throughout the compiled code and are therefore not available for
1856 general allocation. These would include the stack pointer, the frame
1857 pointer (except on machines where that can be used as a general
1858 register when no frame pointer is needed), the program counter on
1859 machines where that is considered one of the addressable registers,
1860 and any other numbered register with a standard use.
1862 This information is expressed as a sequence of numbers, separated by
1863 commas and surrounded by braces. The @var{n}th number is 1 if
1864 register @var{n} is fixed, 0 otherwise.
1866 The table initialized from this macro, and the table initialized by
1867 the following one, may be overridden at run time either automatically,
1868 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1869 the user with the command options @option{-ffixed-@var{reg}},
1870 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1873 @defmac CALL_USED_REGISTERS
1874 @cindex call-used register
1875 @cindex call-clobbered register
1876 @cindex call-saved register
1877 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1878 clobbered (in general) by function calls as well as for fixed
1879 registers. This macro therefore identifies the registers that are not
1880 available for general allocation of values that must live across
1883 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1884 automatically saves it on function entry and restores it on function
1885 exit, if the register is used within the function.
1888 @defmac CALL_REALLY_USED_REGISTERS
1889 @cindex call-used register
1890 @cindex call-clobbered register
1891 @cindex call-saved register
1892 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1893 that the entire set of @code{FIXED_REGISTERS} be included.
1894 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1895 This macro is optional. If not specified, it defaults to the value
1896 of @code{CALL_USED_REGISTERS}.
1899 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1900 @cindex call-used register
1901 @cindex call-clobbered register
1902 @cindex call-saved register
1903 A C expression that is nonzero if it is not permissible to store a
1904 value of mode @var{mode} in hard register number @var{regno} across a
1905 call without some part of it being clobbered. For most machines this
1906 macro need not be defined. It is only required for machines that do not
1907 preserve the entire contents of a register across a call.
1911 @findex call_used_regs
1914 @findex reg_class_contents
1915 @defmac CONDITIONAL_REGISTER_USAGE
1916 Zero or more C statements that may conditionally modify five variables
1917 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1918 @code{reg_names}, and @code{reg_class_contents}, to take into account
1919 any dependence of these register sets on target flags. The first three
1920 of these are of type @code{char []} (interpreted as Boolean vectors).
1921 @code{global_regs} is a @code{const char *[]}, and
1922 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1923 called, @code{fixed_regs}, @code{call_used_regs},
1924 @code{reg_class_contents}, and @code{reg_names} have been initialized
1925 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1926 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1927 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1928 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1929 command options have been applied.
1931 You need not define this macro if it has no work to do.
1933 @cindex disabling certain registers
1934 @cindex controlling register usage
1935 If the usage of an entire class of registers depends on the target
1936 flags, you may indicate this to GCC by using this macro to modify
1937 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1938 registers in the classes which should not be used by GCC@. Also define
1939 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1940 to return @code{NO_REGS} if it
1941 is called with a letter for a class that shouldn't be used.
1943 (However, if this class is not included in @code{GENERAL_REGS} and all
1944 of the insn patterns whose constraints permit this class are
1945 controlled by target switches, then GCC will automatically avoid using
1946 these registers when the target switches are opposed to them.)
1949 @defmac INCOMING_REGNO (@var{out})
1950 Define this macro if the target machine has register windows. This C
1951 expression returns the register number as seen by the called function
1952 corresponding to the register number @var{out} as seen by the calling
1953 function. Return @var{out} if register number @var{out} is not an
1957 @defmac OUTGOING_REGNO (@var{in})
1958 Define this macro if the target machine has register windows. This C
1959 expression returns the register number as seen by the calling function
1960 corresponding to the register number @var{in} as seen by the called
1961 function. Return @var{in} if register number @var{in} is not an inbound
1965 @defmac LOCAL_REGNO (@var{regno})
1966 Define this macro if the target machine has register windows. This C
1967 expression returns true if the register is call-saved but is in the
1968 register window. Unlike most call-saved registers, such registers
1969 need not be explicitly restored on function exit or during non-local
1974 If the program counter has a register number, define this as that
1975 register number. Otherwise, do not define it.
1978 @node Allocation Order
1979 @subsection Order of Allocation of Registers
1980 @cindex order of register allocation
1981 @cindex register allocation order
1983 @c prevent bad page break with this line
1984 Registers are allocated in order.
1986 @defmac REG_ALLOC_ORDER
1987 If defined, an initializer for a vector of integers, containing the
1988 numbers of hard registers in the order in which GCC should prefer
1989 to use them (from most preferred to least).
1991 If this macro is not defined, registers are used lowest numbered first
1992 (all else being equal).
1994 One use of this macro is on machines where the highest numbered
1995 registers must always be saved and the save-multiple-registers
1996 instruction supports only sequences of consecutive registers. On such
1997 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1998 the highest numbered allocable register first.
2001 @defmac ORDER_REGS_FOR_LOCAL_ALLOC
2002 A C statement (sans semicolon) to choose the order in which to allocate
2003 hard registers for pseudo-registers local to a basic block.
2005 Store the desired register order in the array @code{reg_alloc_order}.
2006 Element 0 should be the register to allocate first; element 1, the next
2007 register; and so on.
2009 The macro body should not assume anything about the contents of
2010 @code{reg_alloc_order} before execution of the macro.
2012 On most machines, it is not necessary to define this macro.
2015 @node Values in Registers
2016 @subsection How Values Fit in Registers
2018 This section discusses the macros that describe which kinds of values
2019 (specifically, which machine modes) each register can hold, and how many
2020 consecutive registers are needed for a given mode.
2022 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2023 A C expression for the number of consecutive hard registers, starting
2024 at register number @var{regno}, required to hold a value of mode
2027 On a machine where all registers are exactly one word, a suitable
2028 definition of this macro is
2031 #define HARD_REGNO_NREGS(REGNO, MODE) \
2032 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2037 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2038 Define this macro if the natural size of registers that hold values
2039 of mode @var{mode} is not the word size. It is a C expression that
2040 should give the natural size in bytes for the specified mode. It is
2041 used by the register allocator to try to optimize its results. This
2042 happens for example on SPARC 64-bit where the natural size of
2043 floating-point registers is still 32-bit.
2046 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2047 A C expression that is nonzero if it is permissible to store a value
2048 of mode @var{mode} in hard register number @var{regno} (or in several
2049 registers starting with that one). For a machine where all registers
2050 are equivalent, a suitable definition is
2053 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2056 You need not include code to check for the numbers of fixed registers,
2057 because the allocation mechanism considers them to be always occupied.
2059 @cindex register pairs
2060 On some machines, double-precision values must be kept in even/odd
2061 register pairs. You can implement that by defining this macro to reject
2062 odd register numbers for such modes.
2064 The minimum requirement for a mode to be OK in a register is that the
2065 @samp{mov@var{mode}} instruction pattern support moves between the
2066 register and other hard register in the same class and that moving a
2067 value into the register and back out not alter it.
2069 Since the same instruction used to move @code{word_mode} will work for
2070 all narrower integer modes, it is not necessary on any machine for
2071 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2072 you define patterns @samp{movhi}, etc., to take advantage of this. This
2073 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2074 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2077 Many machines have special registers for floating point arithmetic.
2078 Often people assume that floating point machine modes are allowed only
2079 in floating point registers. This is not true. Any registers that
2080 can hold integers can safely @emph{hold} a floating point machine
2081 mode, whether or not floating arithmetic can be done on it in those
2082 registers. Integer move instructions can be used to move the values.
2084 On some machines, though, the converse is true: fixed-point machine
2085 modes may not go in floating registers. This is true if the floating
2086 registers normalize any value stored in them, because storing a
2087 non-floating value there would garble it. In this case,
2088 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2089 floating registers. But if the floating registers do not automatically
2090 normalize, if you can store any bit pattern in one and retrieve it
2091 unchanged without a trap, then any machine mode may go in a floating
2092 register, so you can define this macro to say so.
2094 The primary significance of special floating registers is rather that
2095 they are the registers acceptable in floating point arithmetic
2096 instructions. However, this is of no concern to
2097 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2098 constraints for those instructions.
2100 On some machines, the floating registers are especially slow to access,
2101 so that it is better to store a value in a stack frame than in such a
2102 register if floating point arithmetic is not being done. As long as the
2103 floating registers are not in class @code{GENERAL_REGS}, they will not
2104 be used unless some pattern's constraint asks for one.
2107 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2108 A C expression that is nonzero if it is OK to rename a hard register
2109 @var{from} to another hard register @var{to}.
2111 One common use of this macro is to prevent renaming of a register to
2112 another register that is not saved by a prologue in an interrupt
2115 The default is always nonzero.
2118 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2119 A C expression that is nonzero if a value of mode
2120 @var{mode1} is accessible in mode @var{mode2} without copying.
2122 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2123 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2124 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2125 should be nonzero. If they differ for any @var{r}, you should define
2126 this macro to return zero unless some other mechanism ensures the
2127 accessibility of the value in a narrower mode.
2129 You should define this macro to return nonzero in as many cases as
2130 possible since doing so will allow GCC to perform better register
2134 @defmac AVOID_CCMODE_COPIES
2135 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2136 registers. You should only define this macro if support for copying to/from
2137 @code{CCmode} is incomplete.
2140 @node Leaf Functions
2141 @subsection Handling Leaf Functions
2143 @cindex leaf functions
2144 @cindex functions, leaf
2145 On some machines, a leaf function (i.e., one which makes no calls) can run
2146 more efficiently if it does not make its own register window. Often this
2147 means it is required to receive its arguments in the registers where they
2148 are passed by the caller, instead of the registers where they would
2151 The special treatment for leaf functions generally applies only when
2152 other conditions are met; for example, often they may use only those
2153 registers for its own variables and temporaries. We use the term ``leaf
2154 function'' to mean a function that is suitable for this special
2155 handling, so that functions with no calls are not necessarily ``leaf
2158 GCC assigns register numbers before it knows whether the function is
2159 suitable for leaf function treatment. So it needs to renumber the
2160 registers in order to output a leaf function. The following macros
2163 @defmac LEAF_REGISTERS
2164 Name of a char vector, indexed by hard register number, which
2165 contains 1 for a register that is allowable in a candidate for leaf
2168 If leaf function treatment involves renumbering the registers, then the
2169 registers marked here should be the ones before renumbering---those that
2170 GCC would ordinarily allocate. The registers which will actually be
2171 used in the assembler code, after renumbering, should not be marked with 1
2174 Define this macro only if the target machine offers a way to optimize
2175 the treatment of leaf functions.
2178 @defmac LEAF_REG_REMAP (@var{regno})
2179 A C expression whose value is the register number to which @var{regno}
2180 should be renumbered, when a function is treated as a leaf function.
2182 If @var{regno} is a register number which should not appear in a leaf
2183 function before renumbering, then the expression should yield @minus{}1, which
2184 will cause the compiler to abort.
2186 Define this macro only if the target machine offers a way to optimize the
2187 treatment of leaf functions, and registers need to be renumbered to do
2191 @findex current_function_is_leaf
2192 @findex current_function_uses_only_leaf_regs
2193 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2194 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2195 specially. They can test the C variable @code{current_function_is_leaf}
2196 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2197 set prior to local register allocation and is valid for the remaining
2198 compiler passes. They can also test the C variable
2199 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2200 functions which only use leaf registers.
2201 @code{current_function_uses_only_leaf_regs} is valid after all passes
2202 that modify the instructions have been run and is only useful if
2203 @code{LEAF_REGISTERS} is defined.
2204 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2205 @c of the next paragraph?! --mew 2feb93
2207 @node Stack Registers
2208 @subsection Registers That Form a Stack
2210 There are special features to handle computers where some of the
2211 ``registers'' form a stack. Stack registers are normally written by
2212 pushing onto the stack, and are numbered relative to the top of the
2215 Currently, GCC can only handle one group of stack-like registers, and
2216 they must be consecutively numbered. Furthermore, the existing
2217 support for stack-like registers is specific to the 80387 floating
2218 point coprocessor. If you have a new architecture that uses
2219 stack-like registers, you will need to do substantial work on
2220 @file{reg-stack.c} and write your machine description to cooperate
2221 with it, as well as defining these macros.
2224 Define this if the machine has any stack-like registers.
2227 @defmac FIRST_STACK_REG
2228 The number of the first stack-like register. This one is the top
2232 @defmac LAST_STACK_REG
2233 The number of the last stack-like register. This one is the bottom of
2237 @node Register Classes
2238 @section Register Classes
2239 @cindex register class definitions
2240 @cindex class definitions, register
2242 On many machines, the numbered registers are not all equivalent.
2243 For example, certain registers may not be allowed for indexed addressing;
2244 certain registers may not be allowed in some instructions. These machine
2245 restrictions are described to the compiler using @dfn{register classes}.
2247 You define a number of register classes, giving each one a name and saying
2248 which of the registers belong to it. Then you can specify register classes
2249 that are allowed as operands to particular instruction patterns.
2253 In general, each register will belong to several classes. In fact, one
2254 class must be named @code{ALL_REGS} and contain all the registers. Another
2255 class must be named @code{NO_REGS} and contain no registers. Often the
2256 union of two classes will be another class; however, this is not required.
2258 @findex GENERAL_REGS
2259 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2260 terribly special about the name, but the operand constraint letters
2261 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2262 the same as @code{ALL_REGS}, just define it as a macro which expands
2265 Order the classes so that if class @var{x} is contained in class @var{y}
2266 then @var{x} has a lower class number than @var{y}.
2268 The way classes other than @code{GENERAL_REGS} are specified in operand
2269 constraints is through machine-dependent operand constraint letters.
2270 You can define such letters to correspond to various classes, then use
2271 them in operand constraints.
2273 You should define a class for the union of two classes whenever some
2274 instruction allows both classes. For example, if an instruction allows
2275 either a floating point (coprocessor) register or a general register for a
2276 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2277 which includes both of them. Otherwise you will get suboptimal code.
2279 You must also specify certain redundant information about the register
2280 classes: for each class, which classes contain it and which ones are
2281 contained in it; for each pair of classes, the largest class contained
2284 When a value occupying several consecutive registers is expected in a
2285 certain class, all the registers used must belong to that class.
2286 Therefore, register classes cannot be used to enforce a requirement for
2287 a register pair to start with an even-numbered register. The way to
2288 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2290 Register classes used for input-operands of bitwise-and or shift
2291 instructions have a special requirement: each such class must have, for
2292 each fixed-point machine mode, a subclass whose registers can transfer that
2293 mode to or from memory. For example, on some machines, the operations for
2294 single-byte values (@code{QImode}) are limited to certain registers. When
2295 this is so, each register class that is used in a bitwise-and or shift
2296 instruction must have a subclass consisting of registers from which
2297 single-byte values can be loaded or stored. This is so that
2298 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2300 @deftp {Data type} {enum reg_class}
2301 An enumerated type that must be defined with all the register class names
2302 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2303 must be the last register class, followed by one more enumerated value,
2304 @code{LIM_REG_CLASSES}, which is not a register class but rather
2305 tells how many classes there are.
2307 Each register class has a number, which is the value of casting
2308 the class name to type @code{int}. The number serves as an index
2309 in many of the tables described below.
2312 @defmac N_REG_CLASSES
2313 The number of distinct register classes, defined as follows:
2316 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2320 @defmac REG_CLASS_NAMES
2321 An initializer containing the names of the register classes as C string
2322 constants. These names are used in writing some of the debugging dumps.
2325 @defmac REG_CLASS_CONTENTS
2326 An initializer containing the contents of the register classes, as integers
2327 which are bit masks. The @var{n}th integer specifies the contents of class
2328 @var{n}. The way the integer @var{mask} is interpreted is that
2329 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2331 When the machine has more than 32 registers, an integer does not suffice.
2332 Then the integers are replaced by sub-initializers, braced groupings containing
2333 several integers. Each sub-initializer must be suitable as an initializer
2334 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2335 In this situation, the first integer in each sub-initializer corresponds to
2336 registers 0 through 31, the second integer to registers 32 through 63, and
2340 @defmac REGNO_REG_CLASS (@var{regno})
2341 A C expression whose value is a register class containing hard register
2342 @var{regno}. In general there is more than one such class; choose a class
2343 which is @dfn{minimal}, meaning that no smaller class also contains the
2347 @defmac BASE_REG_CLASS
2348 A macro whose definition is the name of the class to which a valid
2349 base register must belong. A base register is one used in an address
2350 which is the register value plus a displacement.
2353 @defmac MODE_BASE_REG_CLASS (@var{mode})
2354 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2355 the selection of a base register in a mode dependent manner. If
2356 @var{mode} is VOIDmode then it should return the same value as
2357 @code{BASE_REG_CLASS}.
2360 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2361 A C expression whose value is the register class to which a valid
2362 base register must belong in order to be used in a base plus index
2363 register address. You should define this macro if base plus index
2364 addresses have different requirements than other base register uses.
2367 @defmac INDEX_REG_CLASS
2368 A macro whose definition is the name of the class to which a valid
2369 index register must belong. An index register is one used in an
2370 address where its value is either multiplied by a scale factor or
2371 added to another register (as well as added to a displacement).
2374 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2375 For the constraint at the start of @var{str}, which starts with the letter
2376 @var{c}, return the length. This allows you to have register class /
2377 constant / extra constraints that are longer than a single letter;
2378 you don't need to define this macro if you can do with single-letter
2379 constraints only. The definition of this macro should use
2380 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2381 to handle specially.
2382 There are some sanity checks in genoutput.c that check the constraint lengths
2383 for the md file, so you can also use this macro to help you while you are
2384 transitioning from a byzantine single-letter-constraint scheme: when you
2385 return a negative length for a constraint you want to re-use, genoutput
2386 will complain about every instance where it is used in the md file.
2389 @defmac REG_CLASS_FROM_LETTER (@var{char})
2390 A C expression which defines the machine-dependent operand constraint
2391 letters for register classes. If @var{char} is such a letter, the
2392 value should be the register class corresponding to it. Otherwise,
2393 the value should be @code{NO_REGS}. The register letter @samp{r},
2394 corresponding to class @code{GENERAL_REGS}, will not be passed
2395 to this macro; you do not need to handle it.
2398 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2399 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2400 passed in @var{str}, so that you can use suffixes to distinguish between
2404 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2405 A C expression which is nonzero if register number @var{num} is
2406 suitable for use as a base register in operand addresses. It may be
2407 either a suitable hard register or a pseudo register that has been
2408 allocated such a hard register.
2411 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2412 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2413 that expression may examine the mode of the memory reference in
2414 @var{mode}. You should define this macro if the mode of the memory
2415 reference affects whether a register may be used as a base register. If
2416 you define this macro, the compiler will use it instead of
2417 @code{REGNO_OK_FOR_BASE_P}.
2420 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2421 A C expression which is nonzero if register number @var{num} is suitable for
2422 use as a base register in base plus index operand addresses, accessing
2423 memory in mode @var{mode}. It may be either a suitable hard register or a
2424 pseudo register that has been allocated such a hard register. You should
2425 define this macro if base plus index addresses have different requirements
2426 than other base register uses.
2429 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2430 A C expression which is nonzero if register number @var{num} is
2431 suitable for use as an index register in operand addresses. It may be
2432 either a suitable hard register or a pseudo register that has been
2433 allocated such a hard register.
2435 The difference between an index register and a base register is that
2436 the index register may be scaled. If an address involves the sum of
2437 two registers, neither one of them scaled, then either one may be
2438 labeled the ``base'' and the other the ``index''; but whichever
2439 labeling is used must fit the machine's constraints of which registers
2440 may serve in each capacity. The compiler will try both labelings,
2441 looking for one that is valid, and will reload one or both registers
2442 only if neither labeling works.
2445 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2446 A C expression that places additional restrictions on the register class
2447 to use when it is necessary to copy value @var{x} into a register in class
2448 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2449 another, smaller class. On many machines, the following definition is
2453 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2456 Sometimes returning a more restrictive class makes better code. For
2457 example, on the 68000, when @var{x} is an integer constant that is in range
2458 for a @samp{moveq} instruction, the value of this macro is always
2459 @code{DATA_REGS} as long as @var{class} includes the data registers.
2460 Requiring a data register guarantees that a @samp{moveq} will be used.
2462 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2463 @var{class} is if @var{x} is a legitimate constant which cannot be
2464 loaded into some register class. By returning @code{NO_REGS} you can
2465 force @var{x} into a memory location. For example, rs6000 can load
2466 immediate values into general-purpose registers, but does not have an
2467 instruction for loading an immediate value into a floating-point
2468 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2469 @var{x} is a floating-point constant. If the constant can't be loaded
2470 into any kind of register, code generation will be better if
2471 @code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2472 of using @code{PREFERRED_RELOAD_CLASS}.
2475 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2476 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2477 input reloads. If you don't define this macro, the default is to use
2478 @var{class}, unchanged.
2481 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2482 A C expression that places additional restrictions on the register class
2483 to use when it is necessary to be able to hold a value of mode
2484 @var{mode} in a reload register for which class @var{class} would
2487 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2488 there are certain modes that simply can't go in certain reload classes.
2490 The value is a register class; perhaps @var{class}, or perhaps another,
2493 Don't define this macro unless the target machine has limitations which
2494 require the macro to do something nontrivial.
2497 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2498 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2499 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2500 Many machines have some registers that cannot be copied directly to or
2501 from memory or even from other types of registers. An example is the
2502 @samp{MQ} register, which on most machines, can only be copied to or
2503 from general registers, but not memory. Some machines allow copying all
2504 registers to and from memory, but require a scratch register for stores
2505 to some memory locations (e.g., those with symbolic address on the RT,
2506 and those with certain symbolic address on the SPARC when compiling
2507 PIC)@. In some cases, both an intermediate and a scratch register are
2510 You should define these macros to indicate to the reload phase that it may
2511 need to allocate at least one register for a reload in addition to the
2512 register to contain the data. Specifically, if copying @var{x} to a
2513 register @var{class} in @var{mode} requires an intermediate register,
2514 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2515 largest register class all of whose registers can be used as
2516 intermediate registers or scratch registers.
2518 If copying a register @var{class} in @var{mode} to @var{x} requires an
2519 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2520 should be defined to return the largest register class required. If the
2521 requirements for input and output reloads are the same, the macro
2522 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2525 The values returned by these macros are often @code{GENERAL_REGS}.
2526 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2527 can be directly copied to or from a register of @var{class} in
2528 @var{mode} without requiring a scratch register. Do not define this
2529 macro if it would always return @code{NO_REGS}.
2531 If a scratch register is required (either with or without an
2532 intermediate register), you should define patterns for
2533 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2534 (@pxref{Standard Names}. These patterns, which will normally be
2535 implemented with a @code{define_expand}, should be similar to the
2536 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2539 Define constraints for the reload register and scratch register that
2540 contain a single register class. If the original reload register (whose
2541 class is @var{class}) can meet the constraint given in the pattern, the
2542 value returned by these macros is used for the class of the scratch
2543 register. Otherwise, two additional reload registers are required.
2544 Their classes are obtained from the constraints in the insn pattern.
2546 @var{x} might be a pseudo-register or a @code{subreg} of a
2547 pseudo-register, which could either be in a hard register or in memory.
2548 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2549 in memory and the hard register number if it is in a register.
2551 These macros should not be used in the case where a particular class of
2552 registers can only be copied to memory and not to another class of
2553 registers. In that case, secondary reload registers are not needed and
2554 would not be helpful. Instead, a stack location must be used to perform
2555 the copy and the @code{mov@var{m}} pattern should use memory as an
2556 intermediate storage. This case often occurs between floating-point and
2560 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2561 Certain machines have the property that some registers cannot be copied
2562 to some other registers without using memory. Define this macro on
2563 those machines to be a C expression that is nonzero if objects of mode
2564 @var{m} in registers of @var{class1} can only be copied to registers of
2565 class @var{class2} by storing a register of @var{class1} into memory
2566 and loading that memory location into a register of @var{class2}.
2568 Do not define this macro if its value would always be zero.
2571 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2572 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2573 allocates a stack slot for a memory location needed for register copies.
2574 If this macro is defined, the compiler instead uses the memory location
2575 defined by this macro.
2577 Do not define this macro if you do not define
2578 @code{SECONDARY_MEMORY_NEEDED}.
2581 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2582 When the compiler needs a secondary memory location to copy between two
2583 registers of mode @var{mode}, it normally allocates sufficient memory to
2584 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2585 load operations in a mode that many bits wide and whose class is the
2586 same as that of @var{mode}.
2588 This is right thing to do on most machines because it ensures that all
2589 bits of the register are copied and prevents accesses to the registers
2590 in a narrower mode, which some machines prohibit for floating-point
2593 However, this default behavior is not correct on some machines, such as
2594 the DEC Alpha, that store short integers in floating-point registers
2595 differently than in integer registers. On those machines, the default
2596 widening will not work correctly and you must define this macro to
2597 suppress that widening in some cases. See the file @file{alpha.h} for
2600 Do not define this macro if you do not define
2601 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2602 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2605 @defmac SMALL_REGISTER_CLASSES
2606 On some machines, it is risky to let hard registers live across arbitrary
2607 insns. Typically, these machines have instructions that require values
2608 to be in specific registers (like an accumulator), and reload will fail
2609 if the required hard register is used for another purpose across such an
2612 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2613 value on these machines. When this macro has a nonzero value, the
2614 compiler will try to minimize the lifetime of hard registers.
2616 It is always safe to define this macro with a nonzero value, but if you
2617 unnecessarily define it, you will reduce the amount of optimizations
2618 that can be performed in some cases. If you do not define this macro
2619 with a nonzero value when it is required, the compiler will run out of
2620 spill registers and print a fatal error message. For most machines, you
2621 should not define this macro at all.
2624 @defmac CLASS_LIKELY_SPILLED_P (@var{class})
2625 A C expression whose value is nonzero if pseudos that have been assigned
2626 to registers of class @var{class} would likely be spilled because
2627 registers of @var{class} are needed for spill registers.
2629 The default value of this macro returns 1 if @var{class} has exactly one
2630 register and zero otherwise. On most machines, this default should be
2631 used. Only define this macro to some other expression if pseudos
2632 allocated by @file{local-alloc.c} end up in memory because their hard
2633 registers were needed for spill registers. If this macro returns nonzero
2634 for those classes, those pseudos will only be allocated by
2635 @file{global.c}, which knows how to reallocate the pseudo to another
2636 register. If there would not be another register available for
2637 reallocation, you should not change the definition of this macro since
2638 the only effect of such a definition would be to slow down register
2642 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2643 A C expression for the maximum number of consecutive registers
2644 of class @var{class} needed to hold a value of mode @var{mode}.
2646 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2647 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2648 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2649 @var{mode})} for all @var{regno} values in the class @var{class}.
2651 This macro helps control the handling of multiple-word values
2655 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2656 If defined, a C expression that returns nonzero for a @var{class} for which
2657 a change from mode @var{from} to mode @var{to} is invalid.
2659 For the example, loading 32-bit integer or floating-point objects into
2660 floating-point registers on the Alpha extends them to 64 bits.
2661 Therefore loading a 64-bit object and then storing it as a 32-bit object
2662 does not store the low-order 32 bits, as would be the case for a normal
2663 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2667 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2668 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2669 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2673 Three other special macros describe which operands fit which constraint
2676 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2677 A C expression that defines the machine-dependent operand constraint
2678 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2679 particular ranges of integer values. If @var{c} is one of those
2680 letters, the expression should check that @var{value}, an integer, is in
2681 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2682 not one of those letters, the value should be 0 regardless of
2686 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2687 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2688 string passed in @var{str}, so that you can use suffixes to distinguish
2689 between different variants.
2692 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2693 A C expression that defines the machine-dependent operand constraint
2694 letters that specify particular ranges of @code{const_double} values
2695 (@samp{G} or @samp{H}).
2697 If @var{c} is one of those letters, the expression should check that
2698 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2699 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2700 letters, the value should be 0 regardless of @var{value}.
2702 @code{const_double} is used for all floating-point constants and for
2703 @code{DImode} fixed-point constants. A given letter can accept either
2704 or both kinds of values. It can use @code{GET_MODE} to distinguish
2705 between these kinds.
2708 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2709 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2710 string passed in @var{str}, so that you can use suffixes to distinguish
2711 between different variants.
2714 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2715 A C expression that defines the optional machine-dependent constraint
2716 letters that can be used to segregate specific types of operands, usually
2717 memory references, for the target machine. Any letter that is not
2718 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2719 @code{REG_CLASS_FROM_CONSTRAINT}
2720 may be used. Normally this macro will not be defined.
2722 If it is required for a particular target machine, it should return 1
2723 if @var{value} corresponds to the operand type represented by the
2724 constraint letter @var{c}. If @var{c} is not defined as an extra
2725 constraint, the value returned should be 0 regardless of @var{value}.
2727 For example, on the ROMP, load instructions cannot have their output
2728 in r0 if the memory reference contains a symbolic address. Constraint
2729 letter @samp{Q} is defined as representing a memory address that does
2730 @emph{not} contain a symbolic address. An alternative is specified with
2731 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2732 alternative specifies @samp{m} on the input and a register class that
2733 does not include r0 on the output.
2736 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2737 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2738 in @var{str}, so that you can use suffixes to distinguish between different
2742 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2743 A C expression that defines the optional machine-dependent constraint
2744 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2745 be treated like memory constraints by the reload pass.
2747 It should return 1 if the operand type represented by the constraint
2748 at the start of @var{str}, the first letter of which is the letter @var{c},
2749 comprises a subset of all memory references including
2750 all those whose address is simply a base register. This allows the reload
2751 pass to reload an operand, if it does not directly correspond to the operand
2752 type of @var{c}, by copying its address into a base register.
2754 For example, on the S/390, some instructions do not accept arbitrary
2755 memory references, but only those that do not make use of an index
2756 register. The constraint letter @samp{Q} is defined via
2757 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2758 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2759 a @samp{Q} constraint can handle any memory operand, because the
2760 reload pass knows it can be reloaded by copying the memory address
2761 into a base register if required. This is analogous to the way
2762 a @samp{o} constraint can handle any memory operand.
2765 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2766 A C expression that defines the optional machine-dependent constraint
2767 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2768 @code{EXTRA_CONSTRAINT_STR}, that should
2769 be treated like address constraints by the reload pass.
2771 It should return 1 if the operand type represented by the constraint
2772 at the start of @var{str}, which starts with the letter @var{c}, comprises
2773 a subset of all memory addresses including
2774 all those that consist of just a base register. This allows the reload
2775 pass to reload an operand, if it does not directly correspond to the operand
2776 type of @var{str}, by copying it into a base register.
2778 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2779 be used with the @code{address_operand} predicate. It is treated
2780 analogously to the @samp{p} constraint.
2783 @node Stack and Calling
2784 @section Stack Layout and Calling Conventions
2785 @cindex calling conventions
2787 @c prevent bad page break with this line
2788 This describes the stack layout and calling conventions.
2792 * Exception Handling::
2797 * Register Arguments::
2799 * Aggregate Return::
2807 @subsection Basic Stack Layout
2808 @cindex stack frame layout
2809 @cindex frame layout
2811 @c prevent bad page break with this line
2812 Here is the basic stack layout.
2814 @defmac STACK_GROWS_DOWNWARD
2815 Define this macro if pushing a word onto the stack moves the stack
2816 pointer to a smaller address.
2818 When we say, ``define this macro if @dots{}'', it means that the
2819 compiler checks this macro only with @code{#ifdef} so the precise
2820 definition used does not matter.
2823 @defmac STACK_PUSH_CODE
2824 This macro defines the operation used when something is pushed
2825 on the stack. In RTL, a push operation will be
2826 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2828 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2829 and @code{POST_INC}. Which of these is correct depends on
2830 the stack direction and on whether the stack pointer points
2831 to the last item on the stack or whether it points to the
2832 space for the next item on the stack.
2834 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2835 defined, which is almost always right, and @code{PRE_INC} otherwise,
2836 which is often wrong.
2839 @defmac FRAME_GROWS_DOWNWARD
2840 Define this macro if the addresses of local variable slots are at negative
2841 offsets from the frame pointer.
2844 @defmac ARGS_GROW_DOWNWARD
2845 Define this macro if successive arguments to a function occupy decreasing
2846 addresses on the stack.
2849 @defmac STARTING_FRAME_OFFSET
2850 Offset from the frame pointer to the first local variable slot to be allocated.
2852 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2853 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2854 Otherwise, it is found by adding the length of the first slot to the
2855 value @code{STARTING_FRAME_OFFSET}.
2856 @c i'm not sure if the above is still correct.. had to change it to get
2857 @c rid of an overfull. --mew 2feb93
2860 @defmac STACK_ALIGNMENT_NEEDED
2861 Define to zero to disable final alignment of the stack during reload.
2862 The nonzero default for this macro is suitable for most ports.
2864 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
2865 is a register save block following the local block that doesn't require
2866 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2867 stack alignment and do it in the backend.
2870 @defmac STACK_POINTER_OFFSET
2871 Offset from the stack pointer register to the first location at which
2872 outgoing arguments are placed. If not specified, the default value of
2873 zero is used. This is the proper value for most machines.
2875 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2876 the first location at which outgoing arguments are placed.
2879 @defmac FIRST_PARM_OFFSET (@var{fundecl})
2880 Offset from the argument pointer register to the first argument's
2881 address. On some machines it may depend on the data type of the
2884 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2885 the first argument's address.
2888 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
2889 Offset from the stack pointer register to an item dynamically allocated
2890 on the stack, e.g., by @code{alloca}.
2892 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2893 length of the outgoing arguments. The default is correct for most
2894 machines. See @file{function.c} for details.
2897 @defmac INITIAL_FRAME_ADDRESS_RTX
2898 A C expression whose value is RTL representing the address of the initial
2899 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
2900 @code{DYNAMIC_CHAIN_ADDRESS}.
2901 If you don't define this macro, the default is to return
2902 @code{hard_frame_pointer_rtx}.
2903 This default is usually correct unless @code{-fomit-frame-pointer} is in
2905 Define this macro in order to make @code{__builtin_frame_address (0)} and
2906 @code{__builtin_return_address (0)} work even in absence of a hard frame pointer.
2909 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2910 A C expression whose value is RTL representing the address in a stack
2911 frame where the pointer to the caller's frame is stored. Assume that
2912 @var{frameaddr} is an RTL expression for the address of the stack frame
2915 If you don't define this macro, the default is to return the value
2916 of @var{frameaddr}---that is, the stack frame address is also the
2917 address of the stack word that points to the previous frame.
2920 @defmac SETUP_FRAME_ADDRESSES
2921 If defined, a C expression that produces the machine-specific code to
2922 setup the stack so that arbitrary frames can be accessed. For example,
2923 on the SPARC, we must flush all of the register windows to the stack
2924 before we can access arbitrary stack frames. You will seldom need to
2928 @deftypefn {Target Hook} bool TARGET_BUILTIN_SETJMP_FRAME_VALUE ()
2929 This target hook should return an rtx that is used to store
2930 the address of the current frame into the built in @code{setjmp} buffer.
2931 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2932 machines. One reason you may need to define this target hook is if
2933 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2936 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2937 A C expression whose value is RTL representing the value of the return
2938 address for the frame @var{count} steps up from the current frame, after
2939 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2940 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2941 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2943 The value of the expression must always be the correct address when
2944 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2945 determine the return address of other frames.
2948 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
2949 Define this if the return address of a particular stack frame is accessed
2950 from the frame pointer of the previous stack frame.
2953 @defmac INCOMING_RETURN_ADDR_RTX
2954 A C expression whose value is RTL representing the location of the
2955 incoming return address at the beginning of any function, before the
2956 prologue. This RTL is either a @code{REG}, indicating that the return
2957 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2960 You only need to define this macro if you want to support call frame
2961 debugging information like that provided by DWARF 2.
2963 If this RTL is a @code{REG}, you should also define
2964 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2967 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
2968 A C expression whose value is an integer giving a DWARF 2 column
2969 number that may be used as an alternate return column. This should
2970 be defined only if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
2971 general register, but an alternate column needs to be used for
2975 @defmac DWARF_ZERO_REG
2976 A C expression whose value is an integer giving a DWARF 2 register
2977 number that is considered to always have the value zero. This should
2978 only be defined if the target has an architected zero register, and
2979 someone decided it was a good idea to use that register number to
2980 terminate the stack backtrace. New ports should avoid this.
2983 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
2984 This target hook allows the backend to emit frame-related insns that
2985 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
2986 info engine will invoke it on insns of the form
2988 (set (reg) (unspec [...] UNSPEC_INDEX))
2992 (set (reg) (unspec_volatile [...] UNSPECV_INDEX)).
2994 to let the backend emit the call frame instructions. @var{label} is
2995 the CFI label attached to the insn, @var{pattern} is the pattern of
2996 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
2999 @defmac INCOMING_FRAME_SP_OFFSET
3000 A C expression whose value is an integer giving the offset, in bytes,
3001 from the value of the stack pointer register to the top of the stack
3002 frame at the beginning of any function, before the prologue. The top of
3003 the frame is defined to be the value of the stack pointer in the
3004 previous frame, just before the call instruction.
3006 You only need to define this macro if you want to support call frame
3007 debugging information like that provided by DWARF 2.
3010 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3011 A C expression whose value is an integer giving the offset, in bytes,
3012 from the argument pointer to the canonical frame address (cfa). The
3013 final value should coincide with that calculated by
3014 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3015 during virtual register instantiation.
3017 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
3018 which is correct for most machines; in general, the arguments are found
3019 immediately before the stack frame. Note that this is not the case on
3020 some targets that save registers into the caller's frame, such as SPARC
3021 and rs6000, and so such targets need to define this macro.
3023 You only need to define this macro if the default is incorrect, and you
3024 want to support call frame debugging information like that provided by
3028 @node Exception Handling
3029 @subsection Exception Handling Support
3030 @cindex exception handling
3032 @defmac EH_RETURN_DATA_REGNO (@var{N})
3033 A C expression whose value is the @var{N}th register number used for
3034 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3035 @var{N} registers are usable.
3037 The exception handling library routines communicate with the exception
3038 handlers via a set of agreed upon registers. Ideally these registers
3039 should be call-clobbered; it is possible to use call-saved registers,
3040 but may negatively impact code size. The target must support at least
3041 2 data registers, but should define 4 if there are enough free registers.
3043 You must define this macro if you want to support call frame exception
3044 handling like that provided by DWARF 2.
3047 @defmac EH_RETURN_STACKADJ_RTX
3048 A C expression whose value is RTL representing a location in which
3049 to store a stack adjustment to be applied before function return.
3050 This is used to unwind the stack to an exception handler's call frame.
3051 It will be assigned zero on code paths that return normally.
3053 Typically this is a call-clobbered hard register that is otherwise
3054 untouched by the epilogue, but could also be a stack slot.
3056 Do not define this macro if the stack pointer is saved and restored
3057 by the regular prolog and epilog code in the call frame itself; in
3058 this case, the exception handling library routines will update the
3059 stack location to be restored in place. Otherwise, you must define
3060 this macro if you want to support call frame exception handling like
3061 that provided by DWARF 2.
3064 @defmac EH_RETURN_HANDLER_RTX
3065 A C expression whose value is RTL representing a location in which
3066 to store the address of an exception handler to which we should
3067 return. It will not be assigned on code paths that return normally.
3069 Typically this is the location in the call frame at which the normal
3070 return address is stored. For targets that return by popping an
3071 address off the stack, this might be a memory address just below
3072 the @emph{target} call frame rather than inside the current call
3073 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3074 been assigned, so it may be used to calculate the location of the
3077 Some targets have more complex requirements than storing to an
3078 address calculable during initial code generation. In that case
3079 the @code{eh_return} instruction pattern should be used instead.
3081 If you want to support call frame exception handling, you must
3082 define either this macro or the @code{eh_return} instruction pattern.
3085 @defmac RETURN_ADDR_OFFSET
3086 If defined, an integer-valued C expression for which rtl will be generated
3087 to add it to the exception handler address before it is searched in the
3088 exception handling tables, and to subtract it again from the address before
3089 using it to return to the exception handler.
3092 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3093 This macro chooses the encoding of pointers embedded in the exception
3094 handling sections. If at all possible, this should be defined such
3095 that the exception handling section will not require dynamic relocations,
3096 and so may be read-only.
3098 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3099 @var{global} is true if the symbol may be affected by dynamic relocations.
3100 The macro should return a combination of the @code{DW_EH_PE_*} defines
3101 as found in @file{dwarf2.h}.
3103 If this macro is not defined, pointers will not be encoded but
3104 represented directly.
3107 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3108 This macro allows the target to emit whatever special magic is required
3109 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3110 Generic code takes care of pc-relative and indirect encodings; this must
3111 be defined if the target uses text-relative or data-relative encodings.
3113 This is a C statement that branches to @var{done} if the format was
3114 handled. @var{encoding} is the format chosen, @var{size} is the number
3115 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3119 @defmac MD_UNWIND_SUPPORT
3120 A string specifying a file to be #include'd in unwind-dw2.c. The file
3121 so included typically defines @code{MD_FALLBACK_FRAME_STATE_FOR}.
3124 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3125 This macro allows the target to add cpu and operating system specific
3126 code to the call-frame unwinder for use when there is no unwind data
3127 available. The most common reason to implement this macro is to unwind
3128 through signal frames.
3130 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
3131 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3132 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3133 for the address of the code being executed and @code{context->cfa} for
3134 the stack pointer value. If the frame can be decoded, the register save
3135 addresses should be updated in @var{fs} and the macro should evaluate to
3136 @code{_URC_NO_REASON}. If the frame cannot be decoded, the macro should
3137 evaluate to @code{_URC_END_OF_STACK}.
3139 For proper signal handling in Java this macro is accompanied by
3140 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3143 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3144 This macro allows the target to add operating system specific code to the
3145 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3146 usually used for signal or interrupt frames.
3148 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3149 @var{context} is an @code{_Unwind_Context};
3150 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3151 for the abi and context in the @code{.unwabi} directive. If the
3152 @code{.unwabi} directive can be handled, the register save addresses should
3153 be updated in @var{fs}.
3156 @defmac TARGET_USES_WEAK_UNWIND_INFO
3157 A C expression that evaluates to true if the target requires unwind
3158 info to be given comdat linkage. Define it to be @code{1} if comdat
3159 linkage is necessary. The default is @code{0}.
3162 @node Stack Checking
3163 @subsection Specifying How Stack Checking is Done
3165 GCC will check that stack references are within the boundaries of
3166 the stack, if the @option{-fstack-check} is specified, in one of three ways:
3170 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3171 will assume that you have arranged for stack checking to be done at
3172 appropriate places in the configuration files, e.g., in
3173 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
3177 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
3178 called @code{check_stack} in your @file{md} file, GCC will call that
3179 pattern with one argument which is the address to compare the stack
3180 value against. You must arrange for this pattern to report an error if
3181 the stack pointer is out of range.
3184 If neither of the above are true, GCC will generate code to periodically
3185 ``probe'' the stack pointer using the values of the macros defined below.
3188 Normally, you will use the default values of these macros, so GCC
3189 will use the third approach.
3191 @defmac STACK_CHECK_BUILTIN
3192 A nonzero value if stack checking is done by the configuration files in a
3193 machine-dependent manner. You should define this macro if stack checking
3194 is require by the ABI of your machine or if you would like to have to stack
3195 checking in some more efficient way than GCC's portable approach.
3196 The default value of this macro is zero.
3199 @defmac STACK_CHECK_PROBE_INTERVAL
3200 An integer representing the interval at which GCC must generate stack
3201 probe instructions. You will normally define this macro to be no larger
3202 than the size of the ``guard pages'' at the end of a stack area. The
3203 default value of 4096 is suitable for most systems.
3206 @defmac STACK_CHECK_PROBE_LOAD
3207 A integer which is nonzero if GCC should perform the stack probe
3208 as a load instruction and zero if GCC should use a store instruction.
3209 The default is zero, which is the most efficient choice on most systems.
3212 @defmac STACK_CHECK_PROTECT
3213 The number of bytes of stack needed to recover from a stack overflow,
3214 for languages where such a recovery is supported. The default value of
3215 75 words should be adequate for most machines.
3218 @defmac STACK_CHECK_MAX_FRAME_SIZE
3219 The maximum size of a stack frame, in bytes. GCC will generate probe
3220 instructions in non-leaf functions to ensure at least this many bytes of
3221 stack are available. If a stack frame is larger than this size, stack
3222 checking will not be reliable and GCC will issue a warning. The
3223 default is chosen so that GCC only generates one instruction on most
3224 systems. You should normally not change the default value of this macro.
3227 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3228 GCC uses this value to generate the above warning message. It
3229 represents the amount of fixed frame used by a function, not including
3230 space for any callee-saved registers, temporaries and user variables.
3231 You need only specify an upper bound for this amount and will normally
3232 use the default of four words.
3235 @defmac STACK_CHECK_MAX_VAR_SIZE
3236 The maximum size, in bytes, of an object that GCC will place in the
3237 fixed area of the stack frame when the user specifies
3238 @option{-fstack-check}.
3239 GCC computed the default from the values of the above macros and you will
3240 normally not need to override that default.
3244 @node Frame Registers
3245 @subsection Registers That Address the Stack Frame
3247 @c prevent bad page break with this line
3248 This discusses registers that address the stack frame.
3250 @defmac STACK_POINTER_REGNUM
3251 The register number of the stack pointer register, which must also be a
3252 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3253 the hardware determines which register this is.
3256 @defmac FRAME_POINTER_REGNUM
3257 The register number of the frame pointer register, which is used to
3258 access automatic variables in the stack frame. On some machines, the
3259 hardware determines which register this is. On other machines, you can
3260 choose any register you wish for this purpose.
3263 @defmac HARD_FRAME_POINTER_REGNUM
3264 On some machines the offset between the frame pointer and starting
3265 offset of the automatic variables is not known until after register
3266 allocation has been done (for example, because the saved registers are
3267 between these two locations). On those machines, define
3268 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3269 be used internally until the offset is known, and define
3270 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3271 used for the frame pointer.
3273 You should define this macro only in the very rare circumstances when it
3274 is not possible to calculate the offset between the frame pointer and
3275 the automatic variables until after register allocation has been
3276 completed. When this macro is defined, you must also indicate in your
3277 definition of @code{ELIMINABLE_REGS} how to eliminate
3278 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3279 or @code{STACK_POINTER_REGNUM}.
3281 Do not define this macro if it would be the same as
3282 @code{FRAME_POINTER_REGNUM}.
3285 @defmac ARG_POINTER_REGNUM
3286 The register number of the arg pointer register, which is used to access
3287 the function's argument list. On some machines, this is the same as the
3288 frame pointer register. On some machines, the hardware determines which
3289 register this is. On other machines, you can choose any register you
3290 wish for this purpose. If this is not the same register as the frame
3291 pointer register, then you must mark it as a fixed register according to
3292 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3293 (@pxref{Elimination}).
3296 @defmac RETURN_ADDRESS_POINTER_REGNUM
3297 The register number of the return address pointer register, which is used to
3298 access the current function's return address from the stack. On some
3299 machines, the return address is not at a fixed offset from the frame
3300 pointer or stack pointer or argument pointer. This register can be defined
3301 to point to the return address on the stack, and then be converted by
3302 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3304 Do not define this macro unless there is no other way to get the return
3305 address from the stack.
3308 @defmac STATIC_CHAIN_REGNUM
3309 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3310 Register numbers used for passing a function's static chain pointer. If
3311 register windows are used, the register number as seen by the called
3312 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3313 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3314 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3317 The static chain register need not be a fixed register.
3319 If the static chain is passed in memory, these macros should not be
3320 defined; instead, the next two macros should be defined.
3323 @defmac STATIC_CHAIN
3324 @defmacx STATIC_CHAIN_INCOMING
3325 If the static chain is passed in memory, these macros provide rtx giving
3326 @code{mem} expressions that denote where they are stored.
3327 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3328 as seen by the calling and called functions, respectively. Often the former
3329 will be at an offset from the stack pointer and the latter at an offset from
3332 @findex stack_pointer_rtx
3333 @findex frame_pointer_rtx
3334 @findex arg_pointer_rtx
3335 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3336 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3337 macros and should be used to refer to those items.
3339 If the static chain is passed in a register, the two previous macros should
3343 @defmac DWARF_FRAME_REGISTERS
3344 This macro specifies the maximum number of hard registers that can be
3345 saved in a call frame. This is used to size data structures used in
3346 DWARF2 exception handling.
3348 Prior to GCC 3.0, this macro was needed in order to establish a stable
3349 exception handling ABI in the face of adding new hard registers for ISA
3350 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3351 in the number of hard registers. Nevertheless, this macro can still be
3352 used to reduce the runtime memory requirements of the exception handling
3353 routines, which can be substantial if the ISA contains a lot of
3354 registers that are not call-saved.
3356 If this macro is not defined, it defaults to
3357 @code{FIRST_PSEUDO_REGISTER}.
3360 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3362 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3363 for backward compatibility in pre GCC 3.0 compiled code.
3365 If this macro is not defined, it defaults to
3366 @code{DWARF_FRAME_REGISTERS}.
3369 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3371 Define this macro if the target's representation for dwarf registers
3372 is different than the internal representation for unwind column.
3373 Given a dwarf register, this macro should return the internal unwind
3374 column number to use instead.
3376 See the PowerPC's SPE target for an example.
3379 @defmac DWARF_FRAME_REGNUM (@var{regno})
3381 Define this macro if the target's representation for dwarf registers
3382 used in .eh_frame or .debug_frame is different from that used in other
3383 debug info sections. Given a GCC hard register number, this macro
3384 should return the .eh_frame register number. The default is
3385 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3389 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3391 Define this macro to map register numbers held in the call frame info
3392 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3393 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3394 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3395 return @code{@var{regno}}.
3400 @subsection Eliminating Frame Pointer and Arg Pointer
3402 @c prevent bad page break with this line
3403 This is about eliminating the frame pointer and arg pointer.
3405 @defmac FRAME_POINTER_REQUIRED
3406 A C expression which is nonzero if a function must have and use a frame
3407 pointer. This expression is evaluated in the reload pass. If its value is
3408 nonzero the function will have a frame pointer.
3410 The expression can in principle examine the current function and decide
3411 according to the facts, but on most machines the constant 0 or the
3412 constant 1 suffices. Use 0 when the machine allows code to be generated
3413 with no frame pointer, and doing so saves some time or space. Use 1
3414 when there is no possible advantage to avoiding a frame pointer.
3416 In certain cases, the compiler does not know how to produce valid code
3417 without a frame pointer. The compiler recognizes those cases and
3418 automatically gives the function a frame pointer regardless of what
3419 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3422 In a function that does not require a frame pointer, the frame pointer
3423 register can be allocated for ordinary usage, unless you mark it as a
3424 fixed register. See @code{FIXED_REGISTERS} for more information.
3427 @findex get_frame_size
3428 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3429 A C statement to store in the variable @var{depth-var} the difference
3430 between the frame pointer and the stack pointer values immediately after
3431 the function prologue. The value would be computed from information
3432 such as the result of @code{get_frame_size ()} and the tables of
3433 registers @code{regs_ever_live} and @code{call_used_regs}.
3435 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3436 need not be defined. Otherwise, it must be defined even if
3437 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3438 case, you may set @var{depth-var} to anything.
3441 @defmac ELIMINABLE_REGS
3442 If defined, this macro specifies a table of register pairs used to
3443 eliminate unneeded registers that point into the stack frame. If it is not
3444 defined, the only elimination attempted by the compiler is to replace
3445 references to the frame pointer with references to the stack pointer.
3447 The definition of this macro is a list of structure initializations, each
3448 of which specifies an original and replacement register.
3450 On some machines, the position of the argument pointer is not known until
3451 the compilation is completed. In such a case, a separate hard register
3452 must be used for the argument pointer. This register can be eliminated by
3453 replacing it with either the frame pointer or the argument pointer,
3454 depending on whether or not the frame pointer has been eliminated.
3456 In this case, you might specify:
3458 #define ELIMINABLE_REGS \
3459 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3460 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3461 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3464 Note that the elimination of the argument pointer with the stack pointer is
3465 specified first since that is the preferred elimination.
3468 @defmac CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3469 A C expression that returns nonzero if the compiler is allowed to try
3470 to replace register number @var{from-reg} with register number
3471 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3472 is defined, and will usually be the constant 1, since most of the cases
3473 preventing register elimination are things that the compiler already
3477 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3478 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3479 specifies the initial difference between the specified pair of
3480 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3484 @node Stack Arguments
3485 @subsection Passing Function Arguments on the Stack
3486 @cindex arguments on stack
3487 @cindex stack arguments
3489 The macros in this section control how arguments are passed
3490 on the stack. See the following section for other macros that
3491 control passing certain arguments in registers.
3493 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (tree @var{fntype})
3494 This target hook returns @code{true} if an argument declared in a
3495 prototype as an integral type smaller than @code{int} should actually be
3496 passed as an @code{int}. In addition to avoiding errors in certain
3497 cases of mismatch, it also makes for better code on certain machines.
3498 The default is to not promote prototypes.
3502 A C expression. If nonzero, push insns will be used to pass
3504 If the target machine does not have a push instruction, set it to zero.
3505 That directs GCC to use an alternate strategy: to
3506 allocate the entire argument block and then store the arguments into
3507 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3510 @defmac PUSH_ARGS_REVERSED
3511 A C expression. If nonzero, function arguments will be evaluated from
3512 last to first, rather than from first to last. If this macro is not
3513 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3514 and args grow in opposite directions, and 0 otherwise.
3517 @defmac PUSH_ROUNDING (@var{npushed})
3518 A C expression that is the number of bytes actually pushed onto the
3519 stack when an instruction attempts to push @var{npushed} bytes.
3521 On some machines, the definition
3524 #define PUSH_ROUNDING(BYTES) (BYTES)
3528 will suffice. But on other machines, instructions that appear
3529 to push one byte actually push two bytes in an attempt to maintain
3530 alignment. Then the definition should be
3533 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3537 @findex current_function_outgoing_args_size
3538 @defmac ACCUMULATE_OUTGOING_ARGS
3539 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3540 will be computed and placed into the variable
3541 @code{current_function_outgoing_args_size}. No space will be pushed
3542 onto the stack for each call; instead, the function prologue should
3543 increase the stack frame size by this amount.
3545 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3549 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3550 Define this macro if functions should assume that stack space has been
3551 allocated for arguments even when their values are passed in
3554 The value of this macro is the size, in bytes, of the area reserved for
3555 arguments passed in registers for the function represented by @var{fndecl},
3556 which can be zero if GCC is calling a library function.
3558 This space can be allocated by the caller, or be a part of the
3559 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3562 @c above is overfull. not sure what to do. --mew 5feb93 did
3563 @c something, not sure if it looks good. --mew 10feb93
3565 @defmac OUTGOING_REG_PARM_STACK_SPACE
3566 Define this if it is the responsibility of the caller to allocate the area
3567 reserved for arguments passed in registers.
3569 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3570 whether the space for these arguments counts in the value of
3571 @code{current_function_outgoing_args_size}.
3574 @defmac STACK_PARMS_IN_REG_PARM_AREA
3575 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3576 stack parameters don't skip the area specified by it.
3577 @c i changed this, makes more sens and it should have taken care of the
3578 @c overfull.. not as specific, tho. --mew 5feb93
3580 Normally, when a parameter is not passed in registers, it is placed on the
3581 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3582 suppresses this behavior and causes the parameter to be passed on the
3583 stack in its natural location.
3586 @defmac RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3587 A C expression that should indicate the number of bytes of its own
3588 arguments that a function pops on returning, or 0 if the
3589 function pops no arguments and the caller must therefore pop them all
3590 after the function returns.
3592 @var{fundecl} is a C variable whose value is a tree node that describes
3593 the function in question. Normally it is a node of type
3594 @code{FUNCTION_DECL} that describes the declaration of the function.
3595 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3597 @var{funtype} is a C variable whose value is a tree node that
3598 describes the function in question. Normally it is a node of type
3599 @code{FUNCTION_TYPE} that describes the data type of the function.
3600 From this it is possible to obtain the data types of the value and
3601 arguments (if known).
3603 When a call to a library function is being considered, @var{fundecl}
3604 will contain an identifier node for the library function. Thus, if
3605 you need to distinguish among various library functions, you can do so
3606 by their names. Note that ``library function'' in this context means
3607 a function used to perform arithmetic, whose name is known specially
3608 in the compiler and was not mentioned in the C code being compiled.
3610 @var{stack-size} is the number of bytes of arguments passed on the
3611 stack. If a variable number of bytes is passed, it is zero, and
3612 argument popping will always be the responsibility of the calling function.
3614 On the VAX, all functions always pop their arguments, so the definition
3615 of this macro is @var{stack-size}. On the 68000, using the standard
3616 calling convention, no functions pop their arguments, so the value of
3617 the macro is always 0 in this case. But an alternative calling
3618 convention is available in which functions that take a fixed number of
3619 arguments pop them but other functions (such as @code{printf}) pop
3620 nothing (the caller pops all). When this convention is in use,
3621 @var{funtype} is examined to determine whether a function takes a fixed
3622 number of arguments.
3625 @defmac CALL_POPS_ARGS (@var{cum})
3626 A C expression that should indicate the number of bytes a call sequence
3627 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3628 when compiling a function call.
3630 @var{cum} is the variable in which all arguments to the called function
3631 have been accumulated.
3633 On certain architectures, such as the SH5, a call trampoline is used
3634 that pops certain registers off the stack, depending on the arguments
3635 that have been passed to the function. Since this is a property of the
3636 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3640 @node Register Arguments
3641 @subsection Passing Arguments in Registers
3642 @cindex arguments in registers
3643 @cindex registers arguments
3645 This section describes the macros which let you control how various
3646 types of arguments are passed in registers or how they are arranged in
3649 @defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3650 A C expression that controls whether a function argument is passed
3651 in a register, and which register.
3653 The arguments are @var{cum}, which summarizes all the previous
3654 arguments; @var{mode}, the machine mode of the argument; @var{type},
3655 the data type of the argument as a tree node or 0 if that is not known
3656 (which happens for C support library functions); and @var{named},
3657 which is 1 for an ordinary argument and 0 for nameless arguments that
3658 correspond to @samp{@dots{}} in the called function's prototype.
3659 @var{type} can be an incomplete type if a syntax error has previously
3662 The value of the expression is usually either a @code{reg} RTX for the
3663 hard register in which to pass the argument, or zero to pass the
3664 argument on the stack.
3666 For machines like the VAX and 68000, where normally all arguments are
3667 pushed, zero suffices as a definition.
3669 The value of the expression can also be a @code{parallel} RTX@. This is
3670 used when an argument is passed in multiple locations. The mode of the
3671 @code{parallel} should be the mode of the entire argument. The
3672 @code{parallel} holds any number of @code{expr_list} pairs; each one
3673 describes where part of the argument is passed. In each
3674 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3675 register in which to pass this part of the argument, and the mode of the
3676 register RTX indicates how large this part of the argument is. The
3677 second operand of the @code{expr_list} is a @code{const_int} which gives
3678 the offset in bytes into the entire argument of where this part starts.
3679 As a special exception the first @code{expr_list} in the @code{parallel}
3680 RTX may have a first operand of zero. This indicates that the entire
3681 argument is also stored on the stack.
3683 The last time this macro is called, it is called with @code{MODE ==
3684 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3685 pattern as operands 2 and 3 respectively.
3687 @cindex @file{stdarg.h} and register arguments
3688 The usual way to make the ISO library @file{stdarg.h} work on a machine
3689 where some arguments are usually passed in registers, is to cause
3690 nameless arguments to be passed on the stack instead. This is done
3691 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3693 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3694 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3695 You may use the hook @code{targetm.calls.must_pass_in_stack}
3696 in the definition of this macro to determine if this argument is of a
3697 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3698 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3699 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3700 defined, the argument will be computed in the stack and then loaded into
3704 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, tree @var{type})
3705 This target hook should return @code{true} if we should not pass @var{type}
3706 solely in registers. The file @file{expr.h} defines a
3707 definition that is usually appropriate, refer to @file{expr.h} for additional
3711 @defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3712 Define this macro if the target machine has ``register windows'', so
3713 that the register in which a function sees an arguments is not
3714 necessarily the same as the one in which the caller passed the
3717 For such machines, @code{FUNCTION_ARG} computes the register in which
3718 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3719 be defined in a similar fashion to tell the function being called
3720 where the arguments will arrive.
3722 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3723 serves both purposes.
3726 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3727 This target hook returns the number of bytes at the beginning of an
3728 argument that must be put in registers. The value must be zero for
3729 arguments that are passed entirely in registers or that are entirely
3730 pushed on the stack.
3732 On some machines, certain arguments must be passed partially in
3733 registers and partially in memory. On these machines, typically the
3734 first few words of arguments are passed in registers, and the rest
3735 on the stack. If a multi-word argument (a @code{double} or a
3736 structure) crosses that boundary, its first few words must be passed
3737 in registers and the rest must be pushed. This macro tells the
3738 compiler when this occurs, and how many bytes should go in registers.
3740 @code{FUNCTION_ARG} for these arguments should return the first
3741 register to be used by the caller for this argument; likewise
3742 @code{FUNCTION_INCOMING_ARG}, for the called function.
3745 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3746 This target hook should return @code{true} if an argument at the
3747 position indicated by @var{cum} should be passed by reference. This
3748 predicate is queried after target independent reasons for being
3749 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
3751 If the hook returns true, a copy of that argument is made in memory and a
3752 pointer to the argument is passed instead of the argument itself.
3753 The pointer is passed in whatever way is appropriate for passing a pointer
3757 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3758 The function argument described by the parameters to this hook is
3759 known to be passed by reference. The hook should return true if the
3760 function argument should be copied by the callee instead of copied
3763 For any argument for which the hook returns true, if it can be
3764 determined that the argument is not modified, then a copy need
3767 The default version of this hook always returns false.
3770 @defmac CUMULATIVE_ARGS
3771 A C type for declaring a variable that is used as the first argument of
3772 @code{FUNCTION_ARG} and other related values. For some target machines,
3773 the type @code{int} suffices and can hold the number of bytes of
3776 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3777 arguments that have been passed on the stack. The compiler has other
3778 variables to keep track of that. For target machines on which all
3779 arguments are passed on the stack, there is no need to store anything in
3780 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3781 should not be empty, so use @code{int}.
3784 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
3785 A C statement (sans semicolon) for initializing the variable
3786 @var{cum} for the state at the beginning of the argument list. The
3787 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
3788 is the tree node for the data type of the function which will receive
3789 the args, or 0 if the args are to a compiler support library function.
3790 For direct calls that are not libcalls, @var{fndecl} contain the
3791 declaration node of the function. @var{fndecl} is also set when
3792 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3793 being compiled. @var{n_named_args} is set to the number of named
3794 arguments, including a structure return address if it is passed as a
3795 parameter, when making a call. When processing incoming arguments,
3796 @var{n_named_args} is set to @minus{}1.
3798 When processing a call to a compiler support library function,
3799 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3800 contains the name of the function, as a string. @var{libname} is 0 when
3801 an ordinary C function call is being processed. Thus, each time this
3802 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3803 never both of them at once.
3806 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3807 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3808 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3809 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3810 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3811 0)} is used instead.
3814 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3815 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3816 finding the arguments for the function being compiled. If this macro is
3817 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3819 The value passed for @var{libname} is always 0, since library routines
3820 with special calling conventions are never compiled with GCC@. The
3821 argument @var{libname} exists for symmetry with
3822 @code{INIT_CUMULATIVE_ARGS}.
3823 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3824 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3827 @defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3828 A C statement (sans semicolon) to update the summarizer variable
3829 @var{cum} to advance past an argument in the argument list. The
3830 values @var{mode}, @var{type} and @var{named} describe that argument.
3831 Once this is done, the variable @var{cum} is suitable for analyzing
3832 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
3834 This macro need not do anything if the argument in question was passed
3835 on the stack. The compiler knows how to track the amount of stack space
3836 used for arguments without any special help.
3839 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3840 If defined, a C expression which determines whether, and in which direction,
3841 to pad out an argument with extra space. The value should be of type
3842 @code{enum direction}: either @code{upward} to pad above the argument,
3843 @code{downward} to pad below, or @code{none} to inhibit padding.
3845 The @emph{amount} of padding is always just enough to reach the next
3846 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3849 This macro has a default definition which is right for most systems.
3850 For little-endian machines, the default is to pad upward. For
3851 big-endian machines, the default is to pad downward for an argument of
3852 constant size shorter than an @code{int}, and upward otherwise.
3855 @defmac PAD_VARARGS_DOWN
3856 If defined, a C expression which determines whether the default
3857 implementation of va_arg will attempt to pad down before reading the
3858 next argument, if that argument is smaller than its aligned space as
3859 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3860 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3863 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
3864 Specify padding for the last element of a block move between registers and
3865 memory. @var{first} is nonzero if this is the only element. Defining this
3866 macro allows better control of register function parameters on big-endian
3867 machines, without using @code{PARALLEL} rtl. In particular,
3868 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
3869 registers, as there is no longer a "wrong" part of a register; For example,
3870 a three byte aggregate may be passed in the high part of a register if so
3874 @defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3875 If defined, a C expression that gives the alignment boundary, in bits,
3876 of an argument with the specified mode and type. If it is not defined,
3877 @code{PARM_BOUNDARY} is used for all arguments.
3880 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
3881 A C expression that is nonzero if @var{regno} is the number of a hard
3882 register in which function arguments are sometimes passed. This does
3883 @emph{not} include implicit arguments such as the static chain and
3884 the structure-value address. On many machines, no registers can be
3885 used for this purpose since all function arguments are pushed on the
3889 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (tree @var{type})
3890 This hook should return true if parameter of type @var{type} are passed
3891 as two scalar parameters. By default, GCC will attempt to pack complex
3892 arguments into the target's word size. Some ABIs require complex arguments
3893 to be split and treated as their individual components. For example, on
3894 AIX64, complex floats should be passed in a pair of floating point
3895 registers, even though a complex float would fit in one 64-bit floating
3898 The default value of this hook is @code{NULL}, which is treated as always
3902 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
3903 This hook returns a type node for @code{va_list} for the target.
3904 The default version of the hook returns @code{void*}.
3907 @deftypefn {Target Hook} tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree @var{valist}, tree @var{type}, tree *@var{pre_p}, tree *@var{post_p})
3908 This hook performs target-specific gimplification of
3909 @code{VA_ARG_EXPR}. The first two parameters correspond to the
3910 arguments to @code{va_arg}; the latter two are as in
3911 @code{gimplify.c:gimplify_expr}.
3914 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (enum machine_mode @var{mode})
3915 Define this to return nonzero if the port can handle pointers
3916 with machine mode @var{mode}. The default version of this
3917 hook returns true for both @code{ptr_mode} and @code{Pmode}.
3920 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
3921 Define this to return nonzero if the port is prepared to handle
3922 insns involving scalar mode @var{mode}. For a scalar mode to be
3923 considered supported, all the basic arithmetic and comparisons
3926 The default version of this hook returns true for any mode
3927 required to handle the basic C types (as defined by the port).
3928 Included here are the double-word arithmetic supported by the
3929 code in @file{optabs.c}.
3932 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
3933 Define this to return nonzero if the port is prepared to handle
3934 insns involving vector mode @var{mode}. At the very least, it
3935 must have move patterns for this mode.
3939 @subsection How Scalar Function Values Are Returned
3940 @cindex return values in registers
3941 @cindex values, returned by functions
3942 @cindex scalars, returned as values
3944 This section discusses the macros that control returning scalars as
3945 values---values that can fit in registers.
3947 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
3948 A C expression to create an RTX representing the place where a
3949 function returns a value of data type @var{valtype}. @var{valtype} is
3950 a tree node representing a data type. Write @code{TYPE_MODE
3951 (@var{valtype})} to get the machine mode used to represent that type.
3952 On many machines, only the mode is relevant. (Actually, on most
3953 machines, scalar values are returned in the same place regardless of
3956 The value of the expression is usually a @code{reg} RTX for the hard
3957 register where the return value is stored. The value can also be a
3958 @code{parallel} RTX, if the return value is in multiple places. See
3959 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3961 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply the same
3962 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3965 If the precise function being called is known, @var{func} is a tree
3966 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3967 pointer. This makes it possible to use a different value-returning
3968 convention for specific functions when all their calls are
3971 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3972 types, because these are returned in another way. See
3973 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
3976 @defmac FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3977 Define this macro if the target machine has ``register windows''
3978 so that the register in which a function returns its value is not
3979 the same as the one in which the caller sees the value.
3981 For such machines, @code{FUNCTION_VALUE} computes the register in which
3982 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3983 defined in a similar fashion to tell the function where to put the
3986 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3987 @code{FUNCTION_VALUE} serves both purposes.
3989 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3990 aggregate data types, because these are returned in another way. See
3991 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
3994 @defmac LIBCALL_VALUE (@var{mode})
3995 A C expression to create an RTX representing the place where a library
3996 function returns a value of mode @var{mode}. If the precise function
3997 being called is known, @var{func} is a tree node
3998 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3999 pointer. This makes it possible to use a different value-returning
4000 convention for specific functions when all their calls are
4003 Note that ``library function'' in this context means a compiler
4004 support routine, used to perform arithmetic, whose name is known
4005 specially by the compiler and was not mentioned in the C code being
4008 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
4009 data types, because none of the library functions returns such types.
4012 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4013 A C expression that is nonzero if @var{regno} is the number of a hard
4014 register in which the values of called function may come back.
4016 A register whose use for returning values is limited to serving as the
4017 second of a pair (for a value of type @code{double}, say) need not be
4018 recognized by this macro. So for most machines, this definition
4022 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4025 If the machine has register windows, so that the caller and the called
4026 function use different registers for the return value, this macro
4027 should recognize only the caller's register numbers.
4030 @defmac APPLY_RESULT_SIZE
4031 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4032 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4033 saving and restoring an arbitrary return value.
4036 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (tree @var{type})
4037 This hook should return true if values of type @var{type} are returned
4038 at the most significant end of a register (in other words, if they are
4039 padded at the least significant end). You can assume that @var{type}
4040 is returned in a register; the caller is required to check this.
4042 Note that the register provided by @code{FUNCTION_VALUE} must be able
4043 to hold the complete return value. For example, if a 1-, 2- or 3-byte
4044 structure is returned at the most significant end of a 4-byte register,
4045 @code{FUNCTION_VALUE} should provide an @code{SImode} rtx.
4048 @node Aggregate Return
4049 @subsection How Large Values Are Returned
4050 @cindex aggregates as return values
4051 @cindex large return values
4052 @cindex returning aggregate values
4053 @cindex structure value address
4055 When a function value's mode is @code{BLKmode} (and in some other
4056 cases), the value is not returned according to @code{FUNCTION_VALUE}
4057 (@pxref{Scalar Return}). Instead, the caller passes the address of a
4058 block of memory in which the value should be stored. This address
4059 is called the @dfn{structure value address}.
4061 This section describes how to control returning structure values in
4064 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (tree @var{type}, tree @var{fntype})
4065 This target hook should return a nonzero value to say to return the
4066 function value in memory, just as large structures are always returned.
4067 Here @var{type} will be the data type of the value, and @var{fntype}
4068 will be the type of the function doing the returning, or @code{NULL} for
4071 Note that values of mode @code{BLKmode} must be explicitly handled
4072 by this function. Also, the option @option{-fpcc-struct-return}
4073 takes effect regardless of this macro. On most systems, it is
4074 possible to leave the hook undefined; this causes a default
4075 definition to be used, whose value is the constant 1 for @code{BLKmode}
4076 values, and 0 otherwise.
4078 Do not use this hook to indicate that structures and unions should always
4079 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4083 @defmac DEFAULT_PCC_STRUCT_RETURN
4084 Define this macro to be 1 if all structure and union return values must be
4085 in memory. Since this results in slower code, this should be defined
4086 only if needed for compatibility with other compilers or with an ABI@.
4087 If you define this macro to be 0, then the conventions used for structure
4088 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4091 If not defined, this defaults to the value 1.
4094 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4095 This target hook should return the location of the structure value
4096 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4097 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4098 be @code{NULL}, for libcalls. You do not need to define this target
4099 hook if the address is always passed as an ``invisible'' first
4102 On some architectures the place where the structure value address
4103 is found by the called function is not the same place that the
4104 caller put it. This can be due to register windows, or it could
4105 be because the function prologue moves it to a different place.
4106 @var{incoming} is @code{true} when the location is needed in
4107 the context of the called function, and @code{false} in the context of
4110 If @var{incoming} is @code{true} and the address is to be found on the
4111 stack, return a @code{mem} which refers to the frame pointer.
4114 @defmac PCC_STATIC_STRUCT_RETURN
4115 Define this macro if the usual system convention on the target machine
4116 for returning structures and unions is for the called function to return
4117 the address of a static variable containing the value.
4119 Do not define this if the usual system convention is for the caller to
4120 pass an address to the subroutine.
4122 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4123 nothing when you use @option{-freg-struct-return} mode.
4127 @subsection Caller-Saves Register Allocation
4129 If you enable it, GCC can save registers around function calls. This
4130 makes it possible to use call-clobbered registers to hold variables that
4131 must live across calls.
4133 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4134 A C expression to determine whether it is worthwhile to consider placing
4135 a pseudo-register in a call-clobbered hard register and saving and
4136 restoring it around each function call. The expression should be 1 when
4137 this is worth doing, and 0 otherwise.
4139 If you don't define this macro, a default is used which is good on most
4140 machines: @code{4 * @var{calls} < @var{refs}}.
4143 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4144 A C expression specifying which mode is required for saving @var{nregs}
4145 of a pseudo-register in call-clobbered hard register @var{regno}. If
4146 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4147 returned. For most machines this macro need not be defined since GCC
4148 will select the smallest suitable mode.
4151 @node Function Entry
4152 @subsection Function Entry and Exit
4153 @cindex function entry and exit
4157 This section describes the macros that output function entry
4158 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4160 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4161 If defined, a function that outputs the assembler code for entry to a
4162 function. The prologue is responsible for setting up the stack frame,
4163 initializing the frame pointer register, saving registers that must be
4164 saved, and allocating @var{size} additional bytes of storage for the
4165 local variables. @var{size} is an integer. @var{file} is a stdio
4166 stream to which the assembler code should be output.
4168 The label for the beginning of the function need not be output by this
4169 macro. That has already been done when the macro is run.
4171 @findex regs_ever_live
4172 To determine which registers to save, the macro can refer to the array
4173 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4174 @var{r} is used anywhere within the function. This implies the function
4175 prologue should save register @var{r}, provided it is not one of the
4176 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4177 @code{regs_ever_live}.)
4179 On machines that have ``register windows'', the function entry code does
4180 not save on the stack the registers that are in the windows, even if
4181 they are supposed to be preserved by function calls; instead it takes
4182 appropriate steps to ``push'' the register stack, if any non-call-used
4183 registers are used in the function.
4185 @findex frame_pointer_needed
4186 On machines where functions may or may not have frame-pointers, the
4187 function entry code must vary accordingly; it must set up the frame
4188 pointer if one is wanted, and not otherwise. To determine whether a
4189 frame pointer is in wanted, the macro can refer to the variable
4190 @code{frame_pointer_needed}. The variable's value will be 1 at run
4191 time in a function that needs a frame pointer. @xref{Elimination}.
4193 The function entry code is responsible for allocating any stack space
4194 required for the function. This stack space consists of the regions
4195 listed below. In most cases, these regions are allocated in the
4196 order listed, with the last listed region closest to the top of the
4197 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4198 the highest address if it is not defined). You can use a different order
4199 for a machine if doing so is more convenient or required for
4200 compatibility reasons. Except in cases where required by standard
4201 or by a debugger, there is no reason why the stack layout used by GCC
4202 need agree with that used by other compilers for a machine.
4205 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4206 If defined, a function that outputs assembler code at the end of a
4207 prologue. This should be used when the function prologue is being
4208 emitted as RTL, and you have some extra assembler that needs to be
4209 emitted. @xref{prologue instruction pattern}.
4212 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4213 If defined, a function that outputs assembler code at the start of an
4214 epilogue. This should be used when the function epilogue is being
4215 emitted as RTL, and you have some extra assembler that needs to be
4216 emitted. @xref{epilogue instruction pattern}.
4219 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4220 If defined, a function that outputs the assembler code for exit from a
4221 function. The epilogue is responsible for restoring the saved
4222 registers and stack pointer to their values when the function was
4223 called, and returning control to the caller. This macro takes the
4224 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4225 registers to restore are determined from @code{regs_ever_live} and
4226 @code{CALL_USED_REGISTERS} in the same way.
4228 On some machines, there is a single instruction that does all the work
4229 of returning from the function. On these machines, give that
4230 instruction the name @samp{return} and do not define the macro
4231 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4233 Do not define a pattern named @samp{return} if you want the
4234 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4235 switches to control whether return instructions or epilogues are used,
4236 define a @samp{return} pattern with a validity condition that tests the
4237 target switches appropriately. If the @samp{return} pattern's validity
4238 condition is false, epilogues will be used.
4240 On machines where functions may or may not have frame-pointers, the
4241 function exit code must vary accordingly. Sometimes the code for these
4242 two cases is completely different. To determine whether a frame pointer
4243 is wanted, the macro can refer to the variable
4244 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4245 a function that needs a frame pointer.
4247 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4248 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4249 The C variable @code{current_function_is_leaf} is nonzero for such a
4250 function. @xref{Leaf Functions}.
4252 On some machines, some functions pop their arguments on exit while
4253 others leave that for the caller to do. For example, the 68020 when
4254 given @option{-mrtd} pops arguments in functions that take a fixed
4255 number of arguments.
4257 @findex current_function_pops_args
4258 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4259 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4260 needs to know what was decided. The variable that is called
4261 @code{current_function_pops_args} is the number of bytes of its
4262 arguments that a function should pop. @xref{Scalar Return}.
4263 @c what is the "its arguments" in the above sentence referring to, pray
4264 @c tell? --mew 5feb93
4269 @findex current_function_pretend_args_size
4270 A region of @code{current_function_pretend_args_size} bytes of
4271 uninitialized space just underneath the first argument arriving on the
4272 stack. (This may not be at the very start of the allocated stack region
4273 if the calling sequence has pushed anything else since pushing the stack
4274 arguments. But usually, on such machines, nothing else has been pushed
4275 yet, because the function prologue itself does all the pushing.) This
4276 region is used on machines where an argument may be passed partly in
4277 registers and partly in memory, and, in some cases to support the
4278 features in @code{<stdarg.h>}.
4281 An area of memory used to save certain registers used by the function.
4282 The size of this area, which may also include space for such things as
4283 the return address and pointers to previous stack frames, is
4284 machine-specific and usually depends on which registers have been used
4285 in the function. Machines with register windows often do not require
4289 A region of at least @var{size} bytes, possibly rounded up to an allocation
4290 boundary, to contain the local variables of the function. On some machines,
4291 this region and the save area may occur in the opposite order, with the
4292 save area closer to the top of the stack.
4295 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4296 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4297 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4298 argument lists of the function. @xref{Stack Arguments}.
4301 @defmac EXIT_IGNORE_STACK
4302 Define this macro as a C expression that is nonzero if the return
4303 instruction or the function epilogue ignores the value of the stack
4304 pointer; in other words, if it is safe to delete an instruction to
4305 adjust the stack pointer before a return from the function. The
4308 Note that this macro's value is relevant only for functions for which
4309 frame pointers are maintained. It is never safe to delete a final
4310 stack adjustment in a function that has no frame pointer, and the
4311 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4314 @defmac EPILOGUE_USES (@var{regno})
4315 Define this macro as a C expression that is nonzero for registers that are
4316 used by the epilogue or the @samp{return} pattern. The stack and frame
4317 pointer registers are already be assumed to be used as needed.
4320 @defmac EH_USES (@var{regno})
4321 Define this macro as a C expression that is nonzero for registers that are
4322 used by the exception handling mechanism, and so should be considered live
4323 on entry to an exception edge.
4326 @defmac DELAY_SLOTS_FOR_EPILOGUE
4327 Define this macro if the function epilogue contains delay slots to which
4328 instructions from the rest of the function can be ``moved''. The
4329 definition should be a C expression whose value is an integer
4330 representing the number of delay slots there.
4333 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4334 A C expression that returns 1 if @var{insn} can be placed in delay
4335 slot number @var{n} of the epilogue.
4337 The argument @var{n} is an integer which identifies the delay slot now
4338 being considered (since different slots may have different rules of
4339 eligibility). It is never negative and is always less than the number
4340 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4341 If you reject a particular insn for a given delay slot, in principle, it
4342 may be reconsidered for a subsequent delay slot. Also, other insns may
4343 (at least in principle) be considered for the so far unfilled delay
4346 @findex current_function_epilogue_delay_list
4347 @findex final_scan_insn
4348 The insns accepted to fill the epilogue delay slots are put in an RTL
4349 list made with @code{insn_list} objects, stored in the variable
4350 @code{current_function_epilogue_delay_list}. The insn for the first
4351 delay slot comes first in the list. Your definition of the macro
4352 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4353 outputting the insns in this list, usually by calling
4354 @code{final_scan_insn}.
4356 You need not define this macro if you did not define
4357 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4360 @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})
4361 A function that outputs the assembler code for a thunk
4362 function, used to implement C++ virtual function calls with multiple
4363 inheritance. The thunk acts as a wrapper around a virtual function,
4364 adjusting the implicit object parameter before handing control off to
4367 First, emit code to add the integer @var{delta} to the location that
4368 contains the incoming first argument. Assume that this argument
4369 contains a pointer, and is the one used to pass the @code{this} pointer
4370 in C++. This is the incoming argument @emph{before} the function prologue,
4371 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4372 all other incoming arguments.
4374 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4375 made after adding @code{delta}. In particular, if @var{p} is the
4376 adjusted pointer, the following adjustment should be made:
4379 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4382 After the additions, emit code to jump to @var{function}, which is a
4383 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4384 not touch the return address. Hence returning from @var{FUNCTION} will
4385 return to whoever called the current @samp{thunk}.
4387 The effect must be as if @var{function} had been called directly with
4388 the adjusted first argument. This macro is responsible for emitting all
4389 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4390 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4392 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4393 have already been extracted from it.) It might possibly be useful on
4394 some targets, but probably not.
4396 If you do not define this macro, the target-independent code in the C++
4397 front end will generate a less efficient heavyweight thunk that calls
4398 @var{function} instead of jumping to it. The generic approach does
4399 not support varargs.
4402 @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})
4403 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4404 to output the assembler code for the thunk function specified by the
4405 arguments it is passed, and false otherwise. In the latter case, the
4406 generic approach will be used by the C++ front end, with the limitations
4411 @subsection Generating Code for Profiling
4412 @cindex profiling, code generation
4414 These macros will help you generate code for profiling.
4416 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4417 A C statement or compound statement to output to @var{file} some
4418 assembler code to call the profiling subroutine @code{mcount}.
4421 The details of how @code{mcount} expects to be called are determined by
4422 your operating system environment, not by GCC@. To figure them out,
4423 compile a small program for profiling using the system's installed C
4424 compiler and look at the assembler code that results.
4426 Older implementations of @code{mcount} expect the address of a counter
4427 variable to be loaded into some register. The name of this variable is
4428 @samp{LP} followed by the number @var{labelno}, so you would generate
4429 the name using @samp{LP%d} in a @code{fprintf}.
4432 @defmac PROFILE_HOOK
4433 A C statement or compound statement to output to @var{file} some assembly
4434 code to call the profiling subroutine @code{mcount} even the target does
4435 not support profiling.
4438 @defmac NO_PROFILE_COUNTERS
4439 Define this macro if the @code{mcount} subroutine on your system does
4440 not need a counter variable allocated for each function. This is true
4441 for almost all modern implementations. If you define this macro, you
4442 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
4445 @defmac PROFILE_BEFORE_PROLOGUE
4446 Define this macro if the code for function profiling should come before
4447 the function prologue. Normally, the profiling code comes after.
4451 @subsection Permitting tail calls
4454 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4455 True if it is ok to do sibling call optimization for the specified
4456 call expression @var{exp}. @var{decl} will be the called function,
4457 or @code{NULL} if this is an indirect call.
4459 It is not uncommon for limitations of calling conventions to prevent
4460 tail calls to functions outside the current unit of translation, or
4461 during PIC compilation. The hook is used to enforce these restrictions,
4462 as the @code{sibcall} md pattern can not fail, or fall over to a
4463 ``normal'' call. The criteria for successful sibling call optimization
4464 may vary greatly between different architectures.
4468 @section Implementing the Varargs Macros
4469 @cindex varargs implementation
4471 GCC comes with an implementation of @code{<varargs.h>} and
4472 @code{<stdarg.h>} that work without change on machines that pass arguments
4473 on the stack. Other machines require their own implementations of
4474 varargs, and the two machine independent header files must have
4475 conditionals to include it.
4477 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4478 the calling convention for @code{va_start}. The traditional
4479 implementation takes just one argument, which is the variable in which
4480 to store the argument pointer. The ISO implementation of
4481 @code{va_start} takes an additional second argument. The user is
4482 supposed to write the last named argument of the function here.
4484 However, @code{va_start} should not use this argument. The way to find
4485 the end of the named arguments is with the built-in functions described
4488 @defmac __builtin_saveregs ()
4489 Use this built-in function to save the argument registers in memory so
4490 that the varargs mechanism can access them. Both ISO and traditional
4491 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4492 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4494 On some machines, @code{__builtin_saveregs} is open-coded under the
4495 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4496 other machines, it calls a routine written in assembler language,
4497 found in @file{libgcc2.c}.
4499 Code generated for the call to @code{__builtin_saveregs} appears at the
4500 beginning of the function, as opposed to where the call to
4501 @code{__builtin_saveregs} is written, regardless of what the code is.
4502 This is because the registers must be saved before the function starts
4503 to use them for its own purposes.
4504 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4508 @defmac __builtin_args_info (@var{category})
4509 Use this built-in function to find the first anonymous arguments in
4512 In general, a machine may have several categories of registers used for
4513 arguments, each for a particular category of data types. (For example,
4514 on some machines, floating-point registers are used for floating-point
4515 arguments while other arguments are passed in the general registers.)
4516 To make non-varargs functions use the proper calling convention, you
4517 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4518 registers in each category have been used so far
4520 @code{__builtin_args_info} accesses the same data structure of type
4521 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4522 with it, with @var{category} specifying which word to access. Thus, the
4523 value indicates the first unused register in a given category.
4525 Normally, you would use @code{__builtin_args_info} in the implementation
4526 of @code{va_start}, accessing each category just once and storing the
4527 value in the @code{va_list} object. This is because @code{va_list} will
4528 have to update the values, and there is no way to alter the
4529 values accessed by @code{__builtin_args_info}.
4532 @defmac __builtin_next_arg (@var{lastarg})
4533 This is the equivalent of @code{__builtin_args_info}, for stack
4534 arguments. It returns the address of the first anonymous stack
4535 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4536 returns the address of the location above the first anonymous stack
4537 argument. Use it in @code{va_start} to initialize the pointer for
4538 fetching arguments from the stack. Also use it in @code{va_start} to
4539 verify that the second parameter @var{lastarg} is the last named argument
4540 of the current function.
4543 @defmac __builtin_classify_type (@var{object})
4544 Since each machine has its own conventions for which data types are
4545 passed in which kind of register, your implementation of @code{va_arg}
4546 has to embody these conventions. The easiest way to categorize the
4547 specified data type is to use @code{__builtin_classify_type} together
4548 with @code{sizeof} and @code{__alignof__}.
4550 @code{__builtin_classify_type} ignores the value of @var{object},
4551 considering only its data type. It returns an integer describing what
4552 kind of type that is---integer, floating, pointer, structure, and so on.
4554 The file @file{typeclass.h} defines an enumeration that you can use to
4555 interpret the values of @code{__builtin_classify_type}.
4558 These machine description macros help implement varargs:
4560 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
4561 If defined, this hook produces the machine-specific code for a call to
4562 @code{__builtin_saveregs}. This code will be moved to the very
4563 beginning of the function, before any parameter access are made. The
4564 return value of this function should be an RTX that contains the value
4565 to use as the return of @code{__builtin_saveregs}.
4568 @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})
4569 This target hook offers an alternative to using
4570 @code{__builtin_saveregs} and defining the hook
4571 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
4572 register arguments into the stack so that all the arguments appear to
4573 have been passed consecutively on the stack. Once this is done, you can
4574 use the standard implementation of varargs that works for machines that
4575 pass all their arguments on the stack.
4577 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
4578 structure, containing the values that are obtained after processing the
4579 named arguments. The arguments @var{mode} and @var{type} describe the
4580 last named argument---its machine mode and its data type as a tree node.
4582 The target hook should do two things: first, push onto the stack all the
4583 argument registers @emph{not} used for the named arguments, and second,
4584 store the size of the data thus pushed into the @code{int}-valued
4585 variable pointed to by @var{pretend_args_size}. The value that you
4586 store here will serve as additional offset for setting up the stack
4589 Because you must generate code to push the anonymous arguments at
4590 compile time without knowing their data types,
4591 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
4592 have just a single category of argument register and use it uniformly
4595 If the argument @var{second_time} is nonzero, it means that the
4596 arguments of the function are being analyzed for the second time. This
4597 happens for an inline function, which is not actually compiled until the
4598 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
4599 not generate any instructions in this case.
4602 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca})
4603 Define this hook to return @code{true} if the location where a function
4604 argument is passed depends on whether or not it is a named argument.
4606 This hook controls how the @var{named} argument to @code{FUNCTION_ARG}
4607 is set for varargs and stdarg functions. If this hook returns
4608 @code{true}, the @var{named} argument is always true for named
4609 arguments, and false for unnamed arguments. If it returns @code{false},
4610 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
4611 then all arguments are treated as named. Otherwise, all named arguments
4612 except the last are treated as named.
4614 You need not define this hook if it always returns zero.
4617 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED
4618 If you need to conditionally change ABIs so that one works with
4619 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
4620 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
4621 defined, then define this hook to return @code{true} if
4622 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
4623 Otherwise, you should not define this hook.
4627 @section Trampolines for Nested Functions
4628 @cindex trampolines for nested functions
4629 @cindex nested functions, trampolines for
4631 A @dfn{trampoline} is a small piece of code that is created at run time
4632 when the address of a nested function is taken. It normally resides on
4633 the stack, in the stack frame of the containing function. These macros
4634 tell GCC how to generate code to allocate and initialize a
4637 The instructions in the trampoline must do two things: load a constant
4638 address into the static chain register, and jump to the real address of
4639 the nested function. On CISC machines such as the m68k, this requires
4640 two instructions, a move immediate and a jump. Then the two addresses
4641 exist in the trampoline as word-long immediate operands. On RISC
4642 machines, it is often necessary to load each address into a register in
4643 two parts. Then pieces of each address form separate immediate
4646 The code generated to initialize the trampoline must store the variable
4647 parts---the static chain value and the function address---into the
4648 immediate operands of the instructions. On a CISC machine, this is
4649 simply a matter of copying each address to a memory reference at the
4650 proper offset from the start of the trampoline. On a RISC machine, it
4651 may be necessary to take out pieces of the address and store them
4654 @defmac TRAMPOLINE_TEMPLATE (@var{file})
4655 A C statement to output, on the stream @var{file}, assembler code for a
4656 block of data that contains the constant parts of a trampoline. This
4657 code should not include a label---the label is taken care of
4660 If you do not define this macro, it means no template is needed
4661 for the target. Do not define this macro on systems where the block move
4662 code to copy the trampoline into place would be larger than the code
4663 to generate it on the spot.
4666 @defmac TRAMPOLINE_SECTION
4667 The name of a subroutine to switch to the section in which the
4668 trampoline template is to be placed (@pxref{Sections}). The default is
4669 a value of @samp{readonly_data_section}, which places the trampoline in
4670 the section containing read-only data.
4673 @defmac TRAMPOLINE_SIZE
4674 A C expression for the size in bytes of the trampoline, as an integer.
4677 @defmac TRAMPOLINE_ALIGNMENT
4678 Alignment required for trampolines, in bits.
4680 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4681 is used for aligning trampolines.
4684 @defmac INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4685 A C statement to initialize the variable parts of a trampoline.
4686 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4687 an RTX for the address of the nested function; @var{static_chain} is an
4688 RTX for the static chain value that should be passed to the function
4692 @defmac TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4693 A C statement that should perform any machine-specific adjustment in
4694 the address of the trampoline. Its argument contains the address that
4695 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4696 used for a function call should be different from the address in which
4697 the template was stored, the different address should be assigned to
4698 @var{addr}. If this macro is not defined, @var{addr} will be used for
4701 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4702 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4703 If this macro is not defined, by default the trampoline is allocated as
4704 a stack slot. This default is right for most machines. The exceptions
4705 are machines where it is impossible to execute instructions in the stack
4706 area. On such machines, you may have to implement a separate stack,
4707 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4708 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4710 @var{fp} points to a data structure, a @code{struct function}, which
4711 describes the compilation status of the immediate containing function of
4712 the function which the trampoline is for. The stack slot for the
4713 trampoline is in the stack frame of this containing function. Other
4714 allocation strategies probably must do something analogous with this
4718 Implementing trampolines is difficult on many machines because they have
4719 separate instruction and data caches. Writing into a stack location
4720 fails to clear the memory in the instruction cache, so when the program
4721 jumps to that location, it executes the old contents.
4723 Here are two possible solutions. One is to clear the relevant parts of
4724 the instruction cache whenever a trampoline is set up. The other is to
4725 make all trampolines identical, by having them jump to a standard
4726 subroutine. The former technique makes trampoline execution faster; the
4727 latter makes initialization faster.
4729 To clear the instruction cache when a trampoline is initialized, define
4730 the following macro.
4732 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
4733 If defined, expands to a C expression clearing the @emph{instruction
4734 cache} in the specified interval. The definition of this macro would
4735 typically be a series of @code{asm} statements. Both @var{beg} and
4736 @var{end} are both pointer expressions.
4739 The operating system may also require the stack to be made executable
4740 before calling the trampoline. To implement this requirement, define
4741 the following macro.
4743 @defmac ENABLE_EXECUTE_STACK
4744 Define this macro if certain operations must be performed before executing
4745 code located on the stack. The macro should expand to a series of C
4746 file-scope constructs (e.g.@: functions) and provide a unique entry point
4747 named @code{__enable_execute_stack}. The target is responsible for
4748 emitting calls to the entry point in the code, for example from the
4749 @code{INITIALIZE_TRAMPOLINE} macro.
4752 To use a standard subroutine, define the following macro. In addition,
4753 you must make sure that the instructions in a trampoline fill an entire
4754 cache line with identical instructions, or else ensure that the
4755 beginning of the trampoline code is always aligned at the same point in
4756 its cache line. Look in @file{m68k.h} as a guide.
4758 @defmac TRANSFER_FROM_TRAMPOLINE
4759 Define this macro if trampolines need a special subroutine to do their
4760 work. The macro should expand to a series of @code{asm} statements
4761 which will be compiled with GCC@. They go in a library function named
4762 @code{__transfer_from_trampoline}.
4764 If you need to avoid executing the ordinary prologue code of a compiled
4765 C function when you jump to the subroutine, you can do so by placing a
4766 special label of your own in the assembler code. Use one @code{asm}
4767 statement to generate an assembler label, and another to make the label
4768 global. Then trampolines can use that label to jump directly to your
4769 special assembler code.
4773 @section Implicit Calls to Library Routines
4774 @cindex library subroutine names
4775 @cindex @file{libgcc.a}
4777 @c prevent bad page break with this line
4778 Here is an explanation of implicit calls to library routines.
4780 @defmac DECLARE_LIBRARY_RENAMES
4781 This macro, if defined, should expand to a piece of C code that will get
4782 expanded when compiling functions for libgcc.a. It can be used to
4783 provide alternate names for GCC's internal library functions if there
4784 are ABI-mandated names that the compiler should provide.
4787 @findex init_one_libfunc
4788 @findex set_optab_libfunc
4789 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
4790 This hook should declare additional library routines or rename
4791 existing ones, using the functions @code{set_optab_libfunc} and
4792 @code{init_one_libfunc} defined in @file{optabs.c}.
4793 @code{init_optabs} calls this macro after initializing all the normal
4796 The default is to do nothing. Most ports don't need to define this hook.
4799 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4800 This macro should return @code{true} if the library routine that
4801 implements the floating point comparison operator @var{comparison} in
4802 mode @var{mode} will return a boolean, and @var{false} if it will
4805 GCC's own floating point libraries return tristates from the
4806 comparison operators, so the default returns false always. Most ports
4807 don't need to define this macro.
4810 @defmac TARGET_LIB_INT_CMP_BIASED
4811 This macro should evaluate to @code{true} if the integer comparison
4812 functions (like @code{__cmpdi2}) return 0 to indicate that the first
4813 operand is smaller than the second, 1 to indicate that they are equal,
4814 and 2 to indicate that the first operand is greater than the second.
4815 If this macro evalutes to @code{false} the comparison functions return
4816 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
4817 in @file{libgcc.a}, you do not need to define this macro.
4820 @cindex US Software GOFAST, floating point emulation library
4821 @cindex floating point emulation library, US Software GOFAST
4822 @cindex GOFAST, floating point emulation library
4823 @findex gofast_maybe_init_libfuncs
4824 @defmac US_SOFTWARE_GOFAST
4825 Define this macro if your system C library uses the US Software GOFAST
4826 library to provide floating point emulation.
4828 In addition to defining this macro, your architecture must set
4829 @code{TARGET_INIT_LIBFUNCS} to @code{gofast_maybe_init_libfuncs}, or
4830 else call that function from its version of that hook. It is defined
4831 in @file{config/gofast.h}, which must be included by your
4832 architecture's @file{@var{cpu}.c} file. See @file{sparc/sparc.c} for
4835 If this macro is defined, the
4836 @code{TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL} target hook must return
4837 false for @code{SFmode} and @code{DFmode} comparisons.
4840 @cindex @code{EDOM}, implicit usage
4843 The value of @code{EDOM} on the target machine, as a C integer constant
4844 expression. If you don't define this macro, GCC does not attempt to
4845 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4846 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4849 If you do not define @code{TARGET_EDOM}, then compiled code reports
4850 domain errors by calling the library function and letting it report the
4851 error. If mathematical functions on your system use @code{matherr} when
4852 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4853 that @code{matherr} is used normally.
4856 @cindex @code{errno}, implicit usage
4857 @defmac GEN_ERRNO_RTX
4858 Define this macro as a C expression to create an rtl expression that
4859 refers to the global ``variable'' @code{errno}. (On certain systems,
4860 @code{errno} may not actually be a variable.) If you don't define this
4861 macro, a reasonable default is used.
4864 @cindex C99 math functions, implicit usage
4865 @defmac TARGET_C99_FUNCTIONS
4866 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
4867 @code{sinf} and similarly for other functions defined by C99 standard. The
4868 default is nonzero that should be proper value for most modern systems, however
4869 number of existing systems lacks support for these functions in the runtime so
4870 they needs this macro to be redefined to 0.
4873 @defmac NEXT_OBJC_RUNTIME
4874 Define this macro to generate code for Objective-C message sending using
4875 the calling convention of the NeXT system. This calling convention
4876 involves passing the object, the selector and the method arguments all
4877 at once to the method-lookup library function.
4879 The default calling convention passes just the object and the selector
4880 to the lookup function, which returns a pointer to the method.
4883 @node Addressing Modes
4884 @section Addressing Modes
4885 @cindex addressing modes
4887 @c prevent bad page break with this line
4888 This is about addressing modes.
4890 @defmac HAVE_PRE_INCREMENT
4891 @defmacx HAVE_PRE_DECREMENT
4892 @defmacx HAVE_POST_INCREMENT
4893 @defmacx HAVE_POST_DECREMENT
4894 A C expression that is nonzero if the machine supports pre-increment,
4895 pre-decrement, post-increment, or post-decrement addressing respectively.
4898 @defmac HAVE_PRE_MODIFY_DISP
4899 @defmacx HAVE_POST_MODIFY_DISP
4900 A C expression that is nonzero if the machine supports pre- or
4901 post-address side-effect generation involving constants other than
4902 the size of the memory operand.
4905 @defmac HAVE_PRE_MODIFY_REG
4906 @defmacx HAVE_POST_MODIFY_REG
4907 A C expression that is nonzero if the machine supports pre- or
4908 post-address side-effect generation involving a register displacement.
4911 @defmac CONSTANT_ADDRESS_P (@var{x})
4912 A C expression that is 1 if the RTX @var{x} is a constant which
4913 is a valid address. On most machines, this can be defined as
4914 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4915 in which constant addresses are supported.
4918 @defmac CONSTANT_P (@var{x})
4919 @code{CONSTANT_P}, which is defined by target-independent code,
4920 accepts integer-values expressions whose values are not explicitly
4921 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
4922 expressions and @code{const} arithmetic expressions, in addition to
4923 @code{const_int} and @code{const_double} expressions.
4926 @defmac MAX_REGS_PER_ADDRESS
4927 A number, the maximum number of registers that can appear in a valid
4928 memory address. Note that it is up to you to specify a value equal to
4929 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4933 @defmac GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4934 A C compound statement with a conditional @code{goto @var{label};}
4935 executed if @var{x} (an RTX) is a legitimate memory address on the
4936 target machine for a memory operand of mode @var{mode}.
4938 It usually pays to define several simpler macros to serve as
4939 subroutines for this one. Otherwise it may be too complicated to
4942 This macro must exist in two variants: a strict variant and a
4943 non-strict one. The strict variant is used in the reload pass. It
4944 must be defined so that any pseudo-register that has not been
4945 allocated a hard register is considered a memory reference. In
4946 contexts where some kind of register is required, a pseudo-register
4947 with no hard register must be rejected.
4949 The non-strict variant is used in other passes. It must be defined to
4950 accept all pseudo-registers in every context where some kind of
4951 register is required.
4953 @findex REG_OK_STRICT
4954 Compiler source files that want to use the strict variant of this
4955 macro define the macro @code{REG_OK_STRICT}. You should use an
4956 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4957 in that case and the non-strict variant otherwise.
4959 Subroutines to check for acceptable registers for various purposes (one
4960 for base registers, one for index registers, and so on) are typically
4961 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4962 Then only these subroutine macros need have two variants; the higher
4963 levels of macros may be the same whether strict or not.
4965 Normally, constant addresses which are the sum of a @code{symbol_ref}
4966 and an integer are stored inside a @code{const} RTX to mark them as
4967 constant. Therefore, there is no need to recognize such sums
4968 specifically as legitimate addresses. Normally you would simply
4969 recognize any @code{const} as legitimate.
4971 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4972 sums that are not marked with @code{const}. It assumes that a naked
4973 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4974 naked constant sums as illegitimate addresses, so that none of them will
4975 be given to @code{PRINT_OPERAND_ADDRESS}.
4977 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
4978 On some machines, whether a symbolic address is legitimate depends on
4979 the section that the address refers to. On these machines, define the
4980 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
4981 into the @code{symbol_ref}, and then check for it here. When you see a
4982 @code{const}, you will have to look inside it to find the
4983 @code{symbol_ref} in order to determine the section. @xref{Assembler
4987 @defmac REG_OK_FOR_BASE_P (@var{x})
4988 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4989 RTX) is valid for use as a base register. For hard registers, it
4990 should always accept those which the hardware permits and reject the
4991 others. Whether the macro accepts or rejects pseudo registers must be
4992 controlled by @code{REG_OK_STRICT} as described above. This usually
4993 requires two variant definitions, of which @code{REG_OK_STRICT}
4994 controls the one actually used.
4997 @defmac REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4998 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4999 that expression may examine the mode of the memory reference in
5000 @var{mode}. You should define this macro if the mode of the memory
5001 reference affects whether a register may be used as a base register. If
5002 you define this macro, the compiler will use it instead of
5003 @code{REG_OK_FOR_BASE_P}.
5006 @defmac REG_MODE_OK_FOR_REG_BASE_P (@var{x}, @var{mode})
5007 A C expression which is nonzero if @var{x} (assumed to be a @code{reg} RTX)
5008 is suitable for use as a base register in base plus index operand addresses,
5009 accessing memory in mode @var{mode}. It may be either a suitable hard
5010 register or a pseudo register that has been allocated such a hard register.
5011 You should define this macro if base plus index addresses have different
5012 requirements than other base register uses.
5015 @defmac REG_OK_FOR_INDEX_P (@var{x})
5016 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
5017 RTX) is valid for use as an index register.
5019 The difference between an index register and a base register is that
5020 the index register may be scaled. If an address involves the sum of
5021 two registers, neither one of them scaled, then either one may be
5022 labeled the ``base'' and the other the ``index''; but whichever
5023 labeling is used must fit the machine's constraints of which registers
5024 may serve in each capacity. The compiler will try both labelings,
5025 looking for one that is valid, and will reload one or both registers
5026 only if neither labeling works.
5029 @defmac FIND_BASE_TERM (@var{x})
5030 A C expression to determine the base term of address @var{x}.
5031 This macro is used in only one place: `find_base_term' in alias.c.
5033 It is always safe for this macro to not be defined. It exists so
5034 that alias analysis can understand machine-dependent addresses.
5036 The typical use of this macro is to handle addresses containing
5037 a label_ref or symbol_ref within an UNSPEC@.
5040 @defmac LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
5041 A C compound statement that attempts to replace @var{x} with a valid
5042 memory address for an operand of mode @var{mode}. @var{win} will be a
5043 C statement label elsewhere in the code; the macro definition may use
5046 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
5050 to avoid further processing if the address has become legitimate.
5052 @findex break_out_memory_refs
5053 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5054 and @var{oldx} will be the operand that was given to that function to produce
5057 The code generated by this macro should not alter the substructure of
5058 @var{x}. If it transforms @var{x} into a more legitimate form, it
5059 should assign @var{x} (which will always be a C variable) a new value.
5061 It is not necessary for this macro to come up with a legitimate
5062 address. The compiler has standard ways of doing so in all cases. In
5063 fact, it is safe to omit this macro. But often a
5064 machine-dependent strategy can generate better code.
5067 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5068 A C compound statement that attempts to replace @var{x}, which is an address
5069 that needs reloading, with a valid memory address for an operand of mode
5070 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5071 It is not necessary to define this macro, but it might be useful for
5072 performance reasons.
5074 For example, on the i386, it is sometimes possible to use a single
5075 reload register instead of two by reloading a sum of two pseudo
5076 registers into a register. On the other hand, for number of RISC
5077 processors offsets are limited so that often an intermediate address
5078 needs to be generated in order to address a stack slot. By defining
5079 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5080 generated for adjacent some stack slots can be made identical, and thus
5083 @emph{Note}: This macro should be used with caution. It is necessary
5084 to know something of how reload works in order to effectively use this,
5085 and it is quite easy to produce macros that build in too much knowledge
5086 of reload internals.
5088 @emph{Note}: This macro must be able to reload an address created by a
5089 previous invocation of this macro. If it fails to handle such addresses
5090 then the compiler may generate incorrect code or abort.
5093 The macro definition should use @code{push_reload} to indicate parts that
5094 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5095 suitable to be passed unaltered to @code{push_reload}.
5097 The code generated by this macro must not alter the substructure of
5098 @var{x}. If it transforms @var{x} into a more legitimate form, it
5099 should assign @var{x} (which will always be a C variable) a new value.
5100 This also applies to parts that you change indirectly by calling
5103 @findex strict_memory_address_p
5104 The macro definition may use @code{strict_memory_address_p} to test if
5105 the address has become legitimate.
5108 If you want to change only a part of @var{x}, one standard way of doing
5109 this is to use @code{copy_rtx}. Note, however, that is unshares only a
5110 single level of rtl. Thus, if the part to be changed is not at the
5111 top level, you'll need to replace first the top level.
5112 It is not necessary for this macro to come up with a legitimate
5113 address; but often a machine-dependent strategy can generate better code.
5116 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5117 A C statement or compound statement with a conditional @code{goto
5118 @var{label};} executed if memory address @var{x} (an RTX) can have
5119 different meanings depending on the machine mode of the memory
5120 reference it is used for or if the address is valid for some modes
5123 Autoincrement and autodecrement addresses typically have mode-dependent
5124 effects because the amount of the increment or decrement is the size
5125 of the operand being addressed. Some machines have other mode-dependent
5126 addresses. Many RISC machines have no mode-dependent addresses.
5128 You may assume that @var{addr} is a valid address for the machine.
5131 @defmac LEGITIMATE_CONSTANT_P (@var{x})
5132 A C expression that is nonzero if @var{x} is a legitimate constant for
5133 an immediate operand on the target machine. You can assume that
5134 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5135 @samp{1} is a suitable definition for this macro on machines where
5136 anything @code{CONSTANT_P} is valid.
5139 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5140 This hook is used to undo the possibly obfuscating effects of the
5141 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5142 macros. Some backend implementations of these macros wrap symbol
5143 references inside an @code{UNSPEC} rtx to represent PIC or similar
5144 addressing modes. This target hook allows GCC's optimizers to understand
5145 the semantics of these opaque @code{UNSPEC}s by converting them back
5146 into their original form.
5149 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (rtx @var{x})
5150 This hook should return true if @var{x} is of a form that cannot (or
5151 should not) be spilled to the constant pool. The default version of
5152 this hook returns false.
5154 The primary reason to define this hook is to prevent reload from
5155 deciding that a non-legitimate constant would be better reloaded
5156 from the constant pool instead of spilling and reloading a register
5157 holding the constant. This restriction is often true of addresses
5158 of TLS symbols for various targets.
5161 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5162 This hook should return the DECL of a function @var{f} that given an
5163 address @var{addr} as an argument returns a mask @var{m} that can be
5164 used to extract from two vectors the relevant data that resides in
5165 @var{addr} in case @var{addr} is not properly aligned.
5167 The autovectrizer, when vectorizing a load operation from an address
5168 @var{addr} that may be unaligned, will generate two vector loads from
5169 the two aligned addresses around @var{addr}. It then generates a
5170 @code{REALIGN_LOAD} operation to extract the relevant data from the
5171 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5172 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5173 the third argument, @var{OFF}, defines how the data will be extracted
5174 from these two vectors: if @var{OFF} is 0, then the returned vector is
5175 @var{v2}; otherwise, the returned vector is composed from the last
5176 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5177 @var{OFF} elements of @var{v2}.
5179 If this hook is defined, the autovectorizer will generate a call
5180 to @var{f} (using the DECL tree that this hook returns) and will
5181 use the return value of @var{f} as the argument @var{OFF} to
5182 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5183 should comply with the semantics expected by @code{REALIGN_LOAD}
5185 If this hook is not defined, then @var{addr} will be used as
5186 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5187 log2(@var{VS})-1 bits of @var{addr} will be considered.
5190 @node Condition Code
5191 @section Condition Code Status
5192 @cindex condition code status
5194 @c prevent bad page break with this line
5195 This describes the condition code status.
5198 The file @file{conditions.h} defines a variable @code{cc_status} to
5199 describe how the condition code was computed (in case the interpretation of
5200 the condition code depends on the instruction that it was set by). This
5201 variable contains the RTL expressions on which the condition code is
5202 currently based, and several standard flags.
5204 Sometimes additional machine-specific flags must be defined in the machine
5205 description header file. It can also add additional machine-specific
5206 information by defining @code{CC_STATUS_MDEP}.
5208 @defmac CC_STATUS_MDEP
5209 C code for a data type which is used for declaring the @code{mdep}
5210 component of @code{cc_status}. It defaults to @code{int}.
5212 This macro is not used on machines that do not use @code{cc0}.
5215 @defmac CC_STATUS_MDEP_INIT
5216 A C expression to initialize the @code{mdep} field to ``empty''.
5217 The default definition does nothing, since most machines don't use
5218 the field anyway. If you want to use the field, you should probably
5219 define this macro to initialize it.
5221 This macro is not used on machines that do not use @code{cc0}.
5224 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5225 A C compound statement to set the components of @code{cc_status}
5226 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5227 this macro's responsibility to recognize insns that set the condition
5228 code as a byproduct of other activity as well as those that explicitly
5231 This macro is not used on machines that do not use @code{cc0}.
5233 If there are insns that do not set the condition code but do alter
5234 other machine registers, this macro must check to see whether they
5235 invalidate the expressions that the condition code is recorded as
5236 reflecting. For example, on the 68000, insns that store in address
5237 registers do not set the condition code, which means that usually
5238 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5239 insns. But suppose that the previous insn set the condition code
5240 based on location @samp{a4@@(102)} and the current insn stores a new
5241 value in @samp{a4}. Although the condition code is not changed by
5242 this, it will no longer be true that it reflects the contents of
5243 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5244 @code{cc_status} in this case to say that nothing is known about the
5245 condition code value.
5247 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5248 with the results of peephole optimization: insns whose patterns are
5249 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5250 constants which are just the operands. The RTL structure of these
5251 insns is not sufficient to indicate what the insns actually do. What
5252 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5253 @code{CC_STATUS_INIT}.
5255 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5256 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5257 @samp{cc}. This avoids having detailed information about patterns in
5258 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5261 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5262 Returns a mode from class @code{MODE_CC} to be used when comparison
5263 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5264 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5265 @pxref{Jump Patterns} for a description of the reason for this
5269 #define SELECT_CC_MODE(OP,X,Y) \
5270 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5271 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5272 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5273 || GET_CODE (X) == NEG) \
5274 ? CC_NOOVmode : CCmode))
5277 You should define this macro if and only if you define extra CC modes
5278 in @file{@var{machine}-modes.def}.
5281 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5282 On some machines not all possible comparisons are defined, but you can
5283 convert an invalid comparison into a valid one. For example, the Alpha
5284 does not have a @code{GT} comparison, but you can use an @code{LT}
5285 comparison instead and swap the order of the operands.
5287 On such machines, define this macro to be a C statement to do any
5288 required conversions. @var{code} is the initial comparison code
5289 and @var{op0} and @var{op1} are the left and right operands of the
5290 comparison, respectively. You should modify @var{code}, @var{op0}, and
5291 @var{op1} as required.
5293 GCC will not assume that the comparison resulting from this macro is
5294 valid but will see if the resulting insn matches a pattern in the
5297 You need not define this macro if it would never change the comparison
5301 @defmac REVERSIBLE_CC_MODE (@var{mode})
5302 A C expression whose value is one if it is always safe to reverse a
5303 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5304 can ever return @var{mode} for a floating-point inequality comparison,
5305 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5307 You need not define this macro if it would always returns zero or if the
5308 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5309 For example, here is the definition used on the SPARC, where floating-point
5310 inequality comparisons are always given @code{CCFPEmode}:
5313 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5317 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
5318 A C expression whose value is reversed condition code of the @var{code} for
5319 comparison done in CC_MODE @var{mode}. The macro is used only in case
5320 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5321 machine has some non-standard way how to reverse certain conditionals. For
5322 instance in case all floating point conditions are non-trapping, compiler may
5323 freely convert unordered compares to ordered one. Then definition may look
5327 #define REVERSE_CONDITION(CODE, MODE) \
5328 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5329 : reverse_condition_maybe_unordered (CODE))
5333 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
5334 A C expression that returns true if the conditional execution predicate
5335 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
5336 versa. Define this to return 0 if the target has conditional execution
5337 predicates that cannot be reversed safely. There is no need to validate
5338 that the arguments of op1 and op2 are the same, this is done separately.
5339 If no expansion is specified, this macro is defined as follows:
5342 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5343 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
5347 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *, unsigned int *)
5348 On targets which do not use @code{(cc0)}, and which use a hard
5349 register rather than a pseudo-register to hold condition codes, the
5350 regular CSE passes are often not able to identify cases in which the
5351 hard register is set to a common value. Use this hook to enable a
5352 small pass which optimizes such cases. This hook should return true
5353 to enable this pass, and it should set the integers to which its
5354 arguments point to the hard register numbers used for condition codes.
5355 When there is only one such register, as is true on most systems, the
5356 integer pointed to by the second argument should be set to
5357 @code{INVALID_REGNUM}.
5359 The default version of this hook returns false.
5362 @deftypefn {Target Hook} enum machine_mode TARGET_CC_MODES_COMPATIBLE (enum machine_mode, enum machine_mode)
5363 On targets which use multiple condition code modes in class
5364 @code{MODE_CC}, it is sometimes the case that a comparison can be
5365 validly done in more than one mode. On such a system, define this
5366 target hook to take two mode arguments and to return a mode in which
5367 both comparisons may be validly done. If there is no such mode,
5368 return @code{VOIDmode}.
5370 The default version of this hook checks whether the modes are the
5371 same. If they are, it returns that mode. If they are different, it
5372 returns @code{VOIDmode}.
5376 @section Describing Relative Costs of Operations
5377 @cindex costs of instructions
5378 @cindex relative costs
5379 @cindex speed of instructions
5381 These macros let you describe the relative speed of various operations
5382 on the target machine.
5384 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5385 A C expression for the cost of moving data of mode @var{mode} from a
5386 register in class @var{from} to one in class @var{to}. The classes are
5387 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5388 value of 2 is the default; other values are interpreted relative to
5391 It is not required that the cost always equal 2 when @var{from} is the
5392 same as @var{to}; on some machines it is expensive to move between
5393 registers if they are not general registers.
5395 If reload sees an insn consisting of a single @code{set} between two
5396 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5397 classes returns a value of 2, reload does not check to ensure that the
5398 constraints of the insn are met. Setting a cost of other than 2 will
5399 allow reload to verify that the constraints are met. You should do this
5400 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5403 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5404 A C expression for the cost of moving data of mode @var{mode} between a
5405 register of class @var{class} and memory; @var{in} is zero if the value
5406 is to be written to memory, nonzero if it is to be read in. This cost
5407 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5408 registers and memory is more expensive than between two registers, you
5409 should define this macro to express the relative cost.
5411 If you do not define this macro, GCC uses a default cost of 4 plus
5412 the cost of copying via a secondary reload register, if one is
5413 needed. If your machine requires a secondary reload register to copy
5414 between memory and a register of @var{class} but the reload mechanism is
5415 more complex than copying via an intermediate, define this macro to
5416 reflect the actual cost of the move.
5418 GCC defines the function @code{memory_move_secondary_cost} if
5419 secondary reloads are needed. It computes the costs due to copying via
5420 a secondary register. If your machine copies from memory using a
5421 secondary register in the conventional way but the default base value of
5422 4 is not correct for your machine, define this macro to add some other
5423 value to the result of that function. The arguments to that function
5424 are the same as to this macro.
5428 A C expression for the cost of a branch instruction. A value of 1 is
5429 the default; other values are interpreted relative to that.
5432 Here are additional macros which do not specify precise relative costs,
5433 but only that certain actions are more expensive than GCC would
5436 @defmac SLOW_BYTE_ACCESS
5437 Define this macro as a C expression which is nonzero if accessing less
5438 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5439 faster than accessing a word of memory, i.e., if such access
5440 require more than one instruction or if there is no difference in cost
5441 between byte and (aligned) word loads.
5443 When this macro is not defined, the compiler will access a field by
5444 finding the smallest containing object; when it is defined, a fullword
5445 load will be used if alignment permits. Unless bytes accesses are
5446 faster than word accesses, using word accesses is preferable since it
5447 may eliminate subsequent memory access if subsequent accesses occur to
5448 other fields in the same word of the structure, but to different bytes.
5451 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5452 Define this macro to be the value 1 if memory accesses described by the
5453 @var{mode} and @var{alignment} parameters have a cost many times greater
5454 than aligned accesses, for example if they are emulated in a trap
5457 When this macro is nonzero, the compiler will act as if
5458 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5459 moves. This can cause significantly more instructions to be produced.
5460 Therefore, do not set this macro nonzero if unaligned accesses only add a
5461 cycle or two to the time for a memory access.
5463 If the value of this macro is always zero, it need not be defined. If
5464 this macro is defined, it should produce a nonzero value when
5465 @code{STRICT_ALIGNMENT} is nonzero.
5469 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5470 which a sequence of insns should be generated instead of a
5471 string move insn or a library call. Increasing the value will always
5472 make code faster, but eventually incurs high cost in increased code size.
5474 Note that on machines where the corresponding move insn is a
5475 @code{define_expand} that emits a sequence of insns, this macro counts
5476 the number of such sequences.
5478 If you don't define this, a reasonable default is used.
5481 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5482 A C expression used to determine whether @code{move_by_pieces} will be used to
5483 copy a chunk of memory, or whether some other block move mechanism
5484 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5485 than @code{MOVE_RATIO}.
5488 @defmac MOVE_MAX_PIECES
5489 A C expression used by @code{move_by_pieces} to determine the largest unit
5490 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5494 The threshold of number of scalar move insns, @emph{below} which a sequence
5495 of insns should be generated to clear memory instead of a string clear insn
5496 or a library call. Increasing the value will always make code faster, but
5497 eventually incurs high cost in increased code size.
5499 If you don't define this, a reasonable default is used.
5502 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5503 A C expression used to determine whether @code{clear_by_pieces} will be used
5504 to clear a chunk of memory, or whether some other block clear mechanism
5505 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5506 than @code{CLEAR_RATIO}.
5509 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
5510 A C expression used to determine whether @code{store_by_pieces} will be
5511 used to set a chunk of memory to a constant value, or whether some other
5512 mechanism will be used. Used by @code{__builtin_memset} when storing
5513 values other than constant zero and by @code{__builtin_strcpy} when
5514 when called with a constant source string.
5515 Defaults to to 1 if @code{move_by_pieces_ninsns} returns less
5516 than @code{MOVE_RATIO}.
5519 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
5520 A C expression used to determine whether a load postincrement is a good
5521 thing to use for a given mode. Defaults to the value of
5522 @code{HAVE_POST_INCREMENT}.
5525 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
5526 A C expression used to determine whether a load postdecrement is a good
5527 thing to use for a given mode. Defaults to the value of
5528 @code{HAVE_POST_DECREMENT}.
5531 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
5532 A C expression used to determine whether a load preincrement is a good
5533 thing to use for a given mode. Defaults to the value of
5534 @code{HAVE_PRE_INCREMENT}.
5537 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
5538 A C expression used to determine whether a load predecrement is a good
5539 thing to use for a given mode. Defaults to the value of
5540 @code{HAVE_PRE_DECREMENT}.
5543 @defmac USE_STORE_POST_INCREMENT (@var{mode})
5544 A C expression used to determine whether a store postincrement is a good
5545 thing to use for a given mode. Defaults to the value of
5546 @code{HAVE_POST_INCREMENT}.
5549 @defmac USE_STORE_POST_DECREMENT (@var{mode})
5550 A C expression used to determine whether a store postdecrement is a good
5551 thing to use for a given mode. Defaults to the value of
5552 @code{HAVE_POST_DECREMENT}.
5555 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
5556 This macro is used to determine whether a store preincrement is a good
5557 thing to use for a given mode. Defaults to the value of
5558 @code{HAVE_PRE_INCREMENT}.
5561 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
5562 This macro is used to determine whether a store predecrement is a good
5563 thing to use for a given mode. Defaults to the value of
5564 @code{HAVE_PRE_DECREMENT}.
5567 @defmac NO_FUNCTION_CSE
5568 Define this macro if it is as good or better to call a constant
5569 function address than to call an address kept in a register.
5572 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
5573 Define this macro if a non-short-circuit operation produced by
5574 @samp{fold_range_test ()} is optimal. This macro defaults to true if
5575 @code{BRANCH_COST} is greater than or equal to the value 2.
5578 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total})
5579 This target hook describes the relative costs of RTL expressions.
5581 The cost may depend on the precise form of the expression, which is
5582 available for examination in @var{x}, and the rtx code of the expression
5583 in which it is contained, found in @var{outer_code}. @var{code} is the
5584 expression code---redundant, since it can be obtained with
5585 @code{GET_CODE (@var{x})}.
5587 In implementing this hook, you can use the construct
5588 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
5591 On entry to the hook, @code{*@var{total}} contains a default estimate
5592 for the cost of the expression. The hook should modify this value as
5593 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
5594 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
5595 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
5597 When optimizing for code size, i.e.@: when @code{optimize_size} is
5598 nonzero, this target hook should be used to estimate the relative
5599 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
5601 The hook returns true when all subexpressions of @var{x} have been
5602 processed, and false when @code{rtx_cost} should recurse.
5605 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address})
5606 This hook computes the cost of an addressing mode that contains
5607 @var{address}. If not defined, the cost is computed from
5608 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
5610 For most CISC machines, the default cost is a good approximation of the
5611 true cost of the addressing mode. However, on RISC machines, all
5612 instructions normally have the same length and execution time. Hence
5613 all addresses will have equal costs.
5615 In cases where more than one form of an address is known, the form with
5616 the lowest cost will be used. If multiple forms have the same, lowest,
5617 cost, the one that is the most complex will be used.
5619 For example, suppose an address that is equal to the sum of a register
5620 and a constant is used twice in the same basic block. When this macro
5621 is not defined, the address will be computed in a register and memory
5622 references will be indirect through that register. On machines where
5623 the cost of the addressing mode containing the sum is no higher than
5624 that of a simple indirect reference, this will produce an additional
5625 instruction and possibly require an additional register. Proper
5626 specification of this macro eliminates this overhead for such machines.
5628 This hook is never called with an invalid address.
5630 On machines where an address involving more than one register is as
5631 cheap as an address computation involving only one register, defining
5632 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
5633 be live over a region of code where only one would have been if
5634 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
5635 should be considered in the definition of this macro. Equivalent costs
5636 should probably only be given to addresses with different numbers of
5637 registers on machines with lots of registers.
5641 @section Adjusting the Instruction Scheduler
5643 The instruction scheduler may need a fair amount of machine-specific
5644 adjustment in order to produce good code. GCC provides several target
5645 hooks for this purpose. It is usually enough to define just a few of
5646 them: try the first ones in this list first.
5648 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
5649 This hook returns the maximum number of instructions that can ever
5650 issue at the same time on the target machine. The default is one.
5651 Although the insn scheduler can define itself the possibility of issue
5652 an insn on the same cycle, the value can serve as an additional
5653 constraint to issue insns on the same simulated processor cycle (see
5654 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
5655 This value must be constant over the entire compilation. If you need
5656 it to vary depending on what the instructions are, you must use
5657 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
5659 You could define this hook to return the value of the macro
5660 @code{MAX_DFA_ISSUE_RATE}.
5663 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
5664 This hook is executed by the scheduler after it has scheduled an insn
5665 from the ready list. It should return the number of insns which can
5666 still be issued in the current cycle. The default is
5667 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
5668 @code{USE}, which normally are not counted against the issue rate.
5669 You should define this hook if some insns take more machine resources
5670 than others, so that fewer insns can follow them in the same cycle.
5671 @var{file} is either a null pointer, or a stdio stream to write any
5672 debug output to. @var{verbose} is the verbose level provided by
5673 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
5677 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
5678 This function corrects the value of @var{cost} based on the
5679 relationship between @var{insn} and @var{dep_insn} through the
5680 dependence @var{link}. It should return the new value. The default
5681 is to make no adjustment to @var{cost}. This can be used for example
5682 to specify to the scheduler using the traditional pipeline description
5683 that an output- or anti-dependence does not incur the same cost as a
5684 data-dependence. If the scheduler using the automaton based pipeline
5685 description, the cost of anti-dependence is zero and the cost of
5686 output-dependence is maximum of one and the difference of latency
5687 times of the first and the second insns. If these values are not
5688 acceptable, you could use the hook to modify them too. See also
5689 @pxref{Processor pipeline description}.
5692 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
5693 This hook adjusts the integer scheduling priority @var{priority} of
5694 @var{insn}. It should return the new priority. Reduce the priority to
5695 execute @var{insn} earlier, increase the priority to execute @var{insn}
5696 later. Do not define this hook if you do not need to adjust the
5697 scheduling priorities of insns.
5700 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
5701 This hook is executed by the scheduler after it has scheduled the ready
5702 list, to allow the machine description to reorder it (for example to
5703 combine two small instructions together on @samp{VLIW} machines).
5704 @var{file} is either a null pointer, or a stdio stream to write any
5705 debug output to. @var{verbose} is the verbose level provided by
5706 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
5707 list of instructions that are ready to be scheduled. @var{n_readyp} is
5708 a pointer to the number of elements in the ready list. The scheduler
5709 reads the ready list in reverse order, starting with
5710 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
5711 is the timer tick of the scheduler. You may modify the ready list and
5712 the number of ready insns. The return value is the number of insns that
5713 can issue this cycle; normally this is just @code{issue_rate}. See also
5714 @samp{TARGET_SCHED_REORDER2}.
5717 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
5718 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
5719 function is called whenever the scheduler starts a new cycle. This one
5720 is called once per iteration over a cycle, immediately after
5721 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
5722 return the number of insns to be scheduled in the same cycle. Defining
5723 this hook can be useful if there are frequent situations where
5724 scheduling one insn causes other insns to become ready in the same
5725 cycle. These other insns can then be taken into account properly.
5728 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
5729 This hook is called after evaluation forward dependencies of insns in
5730 chain given by two parameter values (@var{head} and @var{tail}
5731 correspondingly) but before insns scheduling of the insn chain. For
5732 example, it can be used for better insn classification if it requires
5733 analysis of dependencies. This hook can use backward and forward
5734 dependencies of the insn scheduler because they are already
5738 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
5739 This hook is executed by the scheduler at the beginning of each block of
5740 instructions that are to be scheduled. @var{file} is either a null
5741 pointer, or a stdio stream to write any debug output to. @var{verbose}
5742 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5743 @var{max_ready} is the maximum number of insns in the current scheduling
5744 region that can be live at the same time. This can be used to allocate
5745 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
5748 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
5749 This hook is executed by the scheduler at the end of each block of
5750 instructions that are to be scheduled. It can be used to perform
5751 cleanup of any actions done by the other scheduling hooks. @var{file}
5752 is either a null pointer, or a stdio stream to write any debug output
5753 to. @var{verbose} is the verbose level provided by
5754 @option{-fsched-verbose-@var{n}}.
5757 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
5758 This hook is executed by the scheduler after function level initializations.
5759 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
5760 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5761 @var{old_max_uid} is the maximum insn uid when scheduling begins.
5764 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
5765 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
5766 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
5767 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5770 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
5771 The hook returns an RTL insn. The automaton state used in the
5772 pipeline hazard recognizer is changed as if the insn were scheduled
5773 when the new simulated processor cycle starts. Usage of the hook may
5774 simplify the automaton pipeline description for some @acronym{VLIW}
5775 processors. If the hook is defined, it is used only for the automaton
5776 based pipeline description. The default is not to change the state
5777 when the new simulated processor cycle starts.
5780 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
5781 The hook can be used to initialize data used by the previous hook.
5784 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
5785 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
5786 to changed the state as if the insn were scheduled when the new
5787 simulated processor cycle finishes.
5790 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
5791 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
5792 used to initialize data used by the previous hook.
5795 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
5796 This hook controls better choosing an insn from the ready insn queue
5797 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
5798 chooses the first insn from the queue. If the hook returns a positive
5799 value, an additional scheduler code tries all permutations of
5800 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
5801 subsequent ready insns to choose an insn whose issue will result in
5802 maximal number of issued insns on the same cycle. For the
5803 @acronym{VLIW} processor, the code could actually solve the problem of
5804 packing simple insns into the @acronym{VLIW} insn. Of course, if the
5805 rules of @acronym{VLIW} packing are described in the automaton.
5807 This code also could be used for superscalar @acronym{RISC}
5808 processors. Let us consider a superscalar @acronym{RISC} processor
5809 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
5810 @var{B}, some insns can be executed only in pipelines @var{B} or
5811 @var{C}, and one insn can be executed in pipeline @var{B}. The
5812 processor may issue the 1st insn into @var{A} and the 2nd one into
5813 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
5814 until the next cycle. If the scheduler issues the 3rd insn the first,
5815 the processor could issue all 3 insns per cycle.
5817 Actually this code demonstrates advantages of the automaton based
5818 pipeline hazard recognizer. We try quickly and easy many insn
5819 schedules to choose the best one.
5821 The default is no multipass scheduling.
5824 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
5826 This hook controls what insns from the ready insn queue will be
5827 considered for the multipass insn scheduling. If the hook returns
5828 zero for insn passed as the parameter, the insn will be not chosen to
5831 The default is that any ready insns can be chosen to be issued.
5834 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
5836 This hook is called by the insn scheduler before issuing insn passed
5837 as the third parameter on given cycle. If the hook returns nonzero,
5838 the insn is not issued on given processors cycle. Instead of that,
5839 the processor cycle is advanced. If the value passed through the last
5840 parameter is zero, the insn ready queue is not sorted on the new cycle
5841 start as usually. The first parameter passes file for debugging
5842 output. The second one passes the scheduler verbose level of the
5843 debugging output. The forth and the fifth parameter values are
5844 correspondingly processor cycle on which the previous insn has been
5845 issued and the current processor cycle.
5848 @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})
5849 This hook is used to define which dependences are considered costly by
5850 the target, so costly that it is not advisable to schedule the insns that
5851 are involved in the dependence too close to one another. The parameters
5852 to this hook are as follows: The second parameter @var{insn2} is dependent
5853 upon the first parameter @var{insn1}. The dependence between @var{insn1}
5854 and @var{insn2} is represented by the third parameter @var{dep_link}. The
5855 fourth parameter @var{cost} is the cost of the dependence, and the fifth
5856 parameter @var{distance} is the distance in cycles between the two insns.
5857 The hook returns @code{true} if considering the distance between the two
5858 insns the dependence between them is considered costly by the target,
5859 and @code{false} otherwise.
5861 Defining this hook can be useful in multiple-issue out-of-order machines,
5862 where (a) it's practically hopeless to predict the actual data/resource
5863 delays, however: (b) there's a better chance to predict the actual grouping
5864 that will be formed, and (c) correctly emulating the grouping can be very
5865 important. In such targets one may want to allow issuing dependent insns
5866 closer to one another---i.e., closer than the dependence distance; however,
5867 not in cases of "costly dependences", which this hooks allows to define.
5870 Macros in the following table are generated by the program
5871 @file{genattr} and can be useful for writing the hooks.
5873 @defmac MAX_DFA_ISSUE_RATE
5874 The macro definition is generated in the automaton based pipeline
5875 description interface. Its value is calculated from the automaton
5876 based pipeline description and is equal to maximal number of all insns
5877 described in constructions @samp{define_insn_reservation} which can be
5878 issued on the same processor cycle.
5882 @section Dividing the Output into Sections (Texts, Data, @dots{})
5883 @c the above section title is WAY too long. maybe cut the part between
5884 @c the (...)? --mew 10feb93
5886 An object file is divided into sections containing different types of
5887 data. In the most common case, there are three sections: the @dfn{text
5888 section}, which holds instructions and read-only data; the @dfn{data
5889 section}, which holds initialized writable data; and the @dfn{bss
5890 section}, which holds uninitialized data. Some systems have other kinds
5893 The compiler must tell the assembler when to switch sections. These
5894 macros control what commands to output to tell the assembler this. You
5895 can also define additional sections.
5897 @defmac TEXT_SECTION_ASM_OP
5898 A C expression whose value is a string, including spacing, containing the
5899 assembler operation that should precede instructions and read-only data.
5900 Normally @code{"\t.text"} is right.
5903 @defmac HOT_TEXT_SECTION_NAME
5904 If defined, a C string constant for the name of the section containing most
5905 frequently executed functions of the program. If not defined, GCC will provide
5906 a default definition if the target supports named sections.
5909 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5910 If defined, a C string constant for the name of the section containing unlikely
5911 executed functions in the program.
5914 @defmac DATA_SECTION_ASM_OP
5915 A C expression whose value is a string, including spacing, containing the
5916 assembler operation to identify the following data as writable initialized
5917 data. Normally @code{"\t.data"} is right.
5920 @defmac READONLY_DATA_SECTION_ASM_OP
5921 A C expression whose value is a string, including spacing, containing the
5922 assembler operation to identify the following data as read-only initialized
5926 @defmac READONLY_DATA_SECTION
5927 A macro naming a function to call to switch to the proper section for
5928 read-only data. The default is to use @code{READONLY_DATA_SECTION_ASM_OP}
5929 if defined, else fall back to @code{text_section}.
5931 The most common definition will be @code{data_section}, if the target
5932 does not have a special read-only data section, and does not put data
5933 in the text section.
5936 @defmac BSS_SECTION_ASM_OP
5937 If defined, a C expression whose value is a string, including spacing,
5938 containing the assembler operation to identify the following data as
5939 uninitialized global data. If not defined, and neither
5940 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5941 uninitialized global data will be output in the data section if
5942 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5946 @defmac INIT_SECTION_ASM_OP
5947 If defined, a C expression whose value is a string, including spacing,
5948 containing the assembler operation to identify the following data as
5949 initialization code. If not defined, GCC will assume such a section does
5953 @defmac FINI_SECTION_ASM_OP
5954 If defined, a C expression whose value is a string, including spacing,
5955 containing the assembler operation to identify the following data as
5956 finalization code. If not defined, GCC will assume such a section does
5960 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
5961 If defined, an ASM statement that switches to a different section
5962 via @var{section_op}, calls @var{function}, and switches back to
5963 the text section. This is used in @file{crtstuff.c} if
5964 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
5965 to initialization and finalization functions from the init and fini
5966 sections. By default, this macro uses a simple function call. Some
5967 ports need hand-crafted assembly code to avoid dependencies on
5968 registers initialized in the function prologue or to ensure that
5969 constant pools don't end up too far way in the text section.
5972 @defmac FORCE_CODE_SECTION_ALIGN
5973 If defined, an ASM statement that aligns a code section to some
5974 arbitrary boundary. This is used to force all fragments of the
5975 @code{.init} and @code{.fini} sections to have to same alignment
5976 and thus prevent the linker from having to add any padding.
5981 @defmac EXTRA_SECTIONS
5982 A list of names for sections other than the standard two, which are
5983 @code{in_text} and @code{in_data}. You need not define this macro
5984 on a system with no other sections (that GCC needs to use).
5987 @findex text_section
5988 @findex data_section
5989 @defmac EXTRA_SECTION_FUNCTIONS
5990 One or more functions to be defined in @file{varasm.c}. These
5991 functions should do jobs analogous to those of @code{text_section} and
5992 @code{data_section}, for your additional sections. Do not define this
5993 macro if you do not define @code{EXTRA_SECTIONS}.
5996 @defmac JUMP_TABLES_IN_TEXT_SECTION
5997 Define this macro to be an expression with a nonzero value if jump
5998 tables (for @code{tablejump} insns) should be output in the text
5999 section, along with the assembler instructions. Otherwise, the
6000 readonly data section is used.
6002 This macro is irrelevant if there is no separate readonly data section.
6005 @deftypefn {Target Hook} void TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
6006 Switches to the appropriate section for output of @var{exp}. You can
6007 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
6008 some sort. @var{reloc} indicates whether the initial value of @var{exp}
6009 requires link-time relocations. Bit 0 is set when variable contains
6010 local relocations only, while bit 1 is set for global relocations.
6011 Select the section by calling @code{data_section} or one of the
6012 alternatives for other sections. @var{align} is the constant alignment
6015 The default version of this function takes care of putting read-only
6016 variables in @code{readonly_data_section}.
6018 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
6021 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
6022 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
6023 for @code{FUNCTION_DECL}s as well as for variables and constants.
6025 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
6026 function has been determined to be likely to be called, and nonzero if
6027 it is unlikely to be called.
6030 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
6031 Build up a unique section name, expressed as a @code{STRING_CST} node,
6032 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
6033 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
6034 the initial value of @var{exp} requires link-time relocations.
6036 The default version of this function appends the symbol name to the
6037 ELF section name that would normally be used for the symbol. For
6038 example, the function @code{foo} would be placed in @code{.text.foo}.
6039 Whatever the actual target object format, this is often good enough.
6042 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
6043 Switches to a readonly data section associated with
6044 @samp{DECL_SECTION_NAME (@var{decl})}.
6045 The default version of this function switches to @code{.gnu.linkonce.r.name}
6046 section if function's section is @code{.gnu.linkonce.t.name}, to
6047 @code{.rodata.name} if function is in @code{.text.name} section
6048 and otherwise switches to the normal readonly data section.
6051 @deftypefn {Target Hook} void TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
6052 Switches to the appropriate section for output of constant pool entry
6053 @var{x} in @var{mode}. You can assume that @var{x} is some kind of
6054 constant in RTL@. The argument @var{mode} is redundant except in the
6055 case of a @code{const_int} rtx. Select the section by calling
6056 @code{readonly_data_section} or one of the alternatives for other
6057 sections. @var{align} is the constant alignment in bits.
6059 The default version of this function takes care of putting symbolic
6060 constants in @code{flag_pic} mode in @code{data_section} and everything
6061 else in @code{readonly_data_section}.
6064 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
6065 Define this hook if references to a symbol or a constant must be
6066 treated differently depending on something about the variable or
6067 function named by the symbol (such as what section it is in).
6069 The hook is executed immediately after rtl has been created for
6070 @var{decl}, which may be a variable or function declaration or
6071 an entry in the constant pool. In either case, @var{rtl} is the
6072 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
6073 in this hook; that field may not have been initialized yet.
6075 In the case of a constant, it is safe to assume that the rtl is
6076 a @code{mem} whose address is a @code{symbol_ref}. Most decls
6077 will also have this form, but that is not guaranteed. Global
6078 register variables, for instance, will have a @code{reg} for their
6079 rtl. (Normally the right thing to do with such unusual rtl is
6082 The @var{new_decl_p} argument will be true if this is the first time
6083 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
6084 be false for subsequent invocations, which will happen for duplicate
6085 declarations. Whether or not anything must be done for the duplicate
6086 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
6087 @var{new_decl_p} is always true when the hook is called for a constant.
6089 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
6090 The usual thing for this hook to do is to record flags in the
6091 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
6092 Historically, the name string was modified if it was necessary to
6093 encode more than one bit of information, but this practice is now
6094 discouraged; use @code{SYMBOL_REF_FLAGS}.
6096 The default definition of this hook, @code{default_encode_section_info}
6097 in @file{varasm.c}, sets a number of commonly-useful bits in
6098 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
6099 before overriding it.
6102 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
6103 Decode @var{name} and return the real name part, sans
6104 the characters that @code{TARGET_ENCODE_SECTION_INFO}
6108 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
6109 Returns true if @var{exp} should be placed into a ``small data'' section.
6110 The default version of this hook always returns false.
6113 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
6114 Contains the value true if the target places read-only
6115 ``small data'' into a separate section. The default value is false.
6118 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
6119 Returns true if @var{exp} names an object for which name resolution
6120 rules must resolve to the current ``module'' (dynamic shared library
6121 or executable image).
6123 The default version of this hook implements the name resolution rules
6124 for ELF, which has a looser model of global name binding than other
6125 currently supported object file formats.
6128 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
6129 Contains the value true if the target supports thread-local storage.
6130 The default value is false.
6135 @section Position Independent Code
6136 @cindex position independent code
6139 This section describes macros that help implement generation of position
6140 independent code. Simply defining these macros is not enough to
6141 generate valid PIC; you must also add support to the macros
6142 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
6143 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
6144 @samp{movsi} to do something appropriate when the source operand
6145 contains a symbolic address. You may also need to alter the handling of
6146 switch statements so that they use relative addresses.
6147 @c i rearranged the order of the macros above to try to force one of
6148 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
6150 @defmac PIC_OFFSET_TABLE_REGNUM
6151 The register number of the register used to address a table of static
6152 data addresses in memory. In some cases this register is defined by a
6153 processor's ``application binary interface'' (ABI)@. When this macro
6154 is defined, RTL is generated for this register once, as with the stack
6155 pointer and frame pointer registers. If this macro is not defined, it
6156 is up to the machine-dependent files to allocate such a register (if
6157 necessary). Note that this register must be fixed when in use (e.g.@:
6158 when @code{flag_pic} is true).
6161 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6162 Define this macro if the register defined by
6163 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
6164 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
6167 @defmac FINALIZE_PIC
6168 By generating position-independent code, when two different programs (A
6169 and B) share a common library (libC.a), the text of the library can be
6170 shared whether or not the library is linked at the same address for both
6171 programs. In some of these environments, position-independent code
6172 requires not only the use of different addressing modes, but also
6173 special code to enable the use of these addressing modes.
6175 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
6176 codes once the function is being compiled into assembly code, but not
6177 before. (It is not done before, because in the case of compiling an
6178 inline function, it would lead to multiple PIC prologues being
6179 included in functions which used inline functions and were compiled to
6183 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
6184 A C expression that is nonzero if @var{x} is a legitimate immediate
6185 operand on the target machine when generating position independent code.
6186 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6187 check this. You can also assume @var{flag_pic} is true, so you need not
6188 check it either. You need not define this macro if all constants
6189 (including @code{SYMBOL_REF}) can be immediate operands when generating
6190 position independent code.
6193 @node Assembler Format
6194 @section Defining the Output Assembler Language
6196 This section describes macros whose principal purpose is to describe how
6197 to write instructions in assembler language---rather than what the
6201 * File Framework:: Structural information for the assembler file.
6202 * Data Output:: Output of constants (numbers, strings, addresses).
6203 * Uninitialized Data:: Output of uninitialized variables.
6204 * Label Output:: Output and generation of labels.
6205 * Initialization:: General principles of initialization
6206 and termination routines.
6207 * Macros for Initialization::
6208 Specific macros that control the handling of
6209 initialization and termination routines.
6210 * Instruction Output:: Output of actual instructions.
6211 * Dispatch Tables:: Output of jump tables.
6212 * Exception Region Output:: Output of exception region code.
6213 * Alignment Output:: Pseudo ops for alignment and skipping data.
6216 @node File Framework
6217 @subsection The Overall Framework of an Assembler File
6218 @cindex assembler format
6219 @cindex output of assembler code
6221 @c prevent bad page break with this line
6222 This describes the overall framework of an assembly file.
6224 @deftypefn {Target Hook} void TARGET_ASM_FILE_START ()
6225 @findex default_file_start
6226 Output to @code{asm_out_file} any text which the assembler expects to
6227 find at the beginning of a file. The default behavior is controlled
6228 by two flags, documented below. Unless your target's assembler is
6229 quite unusual, if you override the default, you should call
6230 @code{default_file_start} at some point in your target hook. This
6231 lets other target files rely on these variables.
6234 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
6235 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
6236 printed as the very first line in the assembly file, unless
6237 @option{-fverbose-asm} is in effect. (If that macro has been defined
6238 to the empty string, this variable has no effect.) With the normal
6239 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
6240 assembler that it need not bother stripping comments or extra
6241 whitespace from its input. This allows it to work a bit faster.
6243 The default is false. You should not set it to true unless you have
6244 verified that your port does not generate any extra whitespace or
6245 comments that will cause GAS to issue errors in NO_APP mode.
6248 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
6249 If this flag is true, @code{output_file_directive} will be called
6250 for the primary source file, immediately after printing
6251 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
6252 this to be done. The default is false.
6255 @deftypefn {Target Hook} void TARGET_ASM_FILE_END ()
6256 Output to @code{asm_out_file} any text which the assembler expects
6257 to find at the end of a file. The default is to output nothing.
6260 @deftypefun void file_end_indicate_exec_stack ()
6261 Some systems use a common convention, the @samp{.note.GNU-stack}
6262 special section, to indicate whether or not an object file relies on
6263 the stack being executable. If your system uses this convention, you
6264 should define @code{TARGET_ASM_FILE_END} to this function. If you
6265 need to do other things in that hook, have your hook function call
6269 @defmac ASM_COMMENT_START
6270 A C string constant describing how to begin a comment in the target
6271 assembler language. The compiler assumes that the comment will end at
6272 the end of the line.
6276 A C string constant for text to be output before each @code{asm}
6277 statement or group of consecutive ones. Normally this is
6278 @code{"#APP"}, which is a comment that has no effect on most
6279 assemblers but tells the GNU assembler that it must check the lines
6280 that follow for all valid assembler constructs.
6284 A C string constant for text to be output after each @code{asm}
6285 statement or group of consecutive ones. Normally this is
6286 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6287 time-saving assumptions that are valid for ordinary compiler output.
6290 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6291 A C statement to output COFF information or DWARF debugging information
6292 which indicates that filename @var{name} is the current source file to
6293 the stdio stream @var{stream}.
6295 This macro need not be defined if the standard form of output
6296 for the file format in use is appropriate.
6299 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6300 A C statement to output the string @var{string} to the stdio stream
6301 @var{stream}. If you do not call the function @code{output_quoted_string}
6302 in your config files, GCC will only call it to output filenames to
6303 the assembler source. So you can use it to canonicalize the format
6304 of the filename using this macro.
6307 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6308 A C statement to output something to the assembler file to handle a
6309 @samp{#ident} directive containing the text @var{string}. If this
6310 macro is not defined, nothing is output for a @samp{#ident} directive.
6313 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6314 Output assembly directives to switch to section @var{name}. The section
6315 should have attributes as specified by @var{flags}, which is a bit mask
6316 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6317 is nonzero, it contains an alignment in bytes to be used for the section,
6318 otherwise some target default should be used. Only targets that must
6319 specify an alignment within the section directive need pay attention to
6320 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6323 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6324 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6327 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6328 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6329 based on a variable or function decl, a section name, and whether or not the
6330 declaration's initializer may contain runtime relocations. @var{decl} may be
6331 null, in which case read-write data should be assumed.
6333 The default version if this function handles choosing code vs data,
6334 read-only vs read-write data, and @code{flag_pic}. You should only
6335 need to override this if your target has special flags that might be
6336 set via @code{__attribute__}.
6341 @subsection Output of Data
6344 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6345 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6346 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6347 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6348 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6349 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6350 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6351 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6352 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6353 These hooks specify assembly directives for creating certain kinds
6354 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6355 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6356 aligned two-byte object, and so on. Any of the hooks may be
6357 @code{NULL}, indicating that no suitable directive is available.
6359 The compiler will print these strings at the start of a new line,
6360 followed immediately by the object's initial value. In most cases,
6361 the string should contain a tab, a pseudo-op, and then another tab.
6364 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6365 The @code{assemble_integer} function uses this hook to output an
6366 integer object. @var{x} is the object's value, @var{size} is its size
6367 in bytes and @var{aligned_p} indicates whether it is aligned. The
6368 function should return @code{true} if it was able to output the
6369 object. If it returns false, @code{assemble_integer} will try to
6370 split the object into smaller parts.
6372 The default implementation of this hook will use the
6373 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6374 when the relevant string is @code{NULL}.
6377 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6378 A C statement to recognize @var{rtx} patterns that
6379 @code{output_addr_const} can't deal with, and output assembly code to
6380 @var{stream} corresponding to the pattern @var{x}. This may be used to
6381 allow machine-dependent @code{UNSPEC}s to appear within constants.
6383 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6384 @code{goto fail}, so that a standard error message is printed. If it
6385 prints an error message itself, by calling, for example,
6386 @code{output_operand_lossage}, it may just complete normally.
6389 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6390 A C statement to output to the stdio stream @var{stream} an assembler
6391 instruction to assemble a string constant containing the @var{len}
6392 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6393 @code{char *} and @var{len} a C expression of type @code{int}.
6395 If the assembler has a @code{.ascii} pseudo-op as found in the
6396 Berkeley Unix assembler, do not define the macro
6397 @code{ASM_OUTPUT_ASCII}.
6400 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6401 A C statement to output word @var{n} of a function descriptor for
6402 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6403 is defined, and is otherwise unused.
6406 @defmac CONSTANT_POOL_BEFORE_FUNCTION
6407 You may define this macro as a C expression. You should define the
6408 expression to have a nonzero value if GCC should output the constant
6409 pool for a function before the code for the function, or a zero value if
6410 GCC should output the constant pool after the function. If you do
6411 not define this macro, the usual case, GCC will output the constant
6412 pool before the function.
6415 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
6416 A C statement to output assembler commands to define the start of the
6417 constant pool for a function. @var{funname} is a string giving
6418 the name of the function. Should the return type of the function
6419 be required, it can be obtained via @var{fundecl}. @var{size}
6420 is the size, in bytes, of the constant pool that will be written
6421 immediately after this call.
6423 If no constant-pool prefix is required, the usual case, this macro need
6427 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
6428 A C statement (with or without semicolon) to output a constant in the
6429 constant pool, if it needs special treatment. (This macro need not do
6430 anything for RTL expressions that can be output normally.)
6432 The argument @var{file} is the standard I/O stream to output the
6433 assembler code on. @var{x} is the RTL expression for the constant to
6434 output, and @var{mode} is the machine mode (in case @var{x} is a
6435 @samp{const_int}). @var{align} is the required alignment for the value
6436 @var{x}; you should output an assembler directive to force this much
6439 The argument @var{labelno} is a number to use in an internal label for
6440 the address of this pool entry. The definition of this macro is
6441 responsible for outputting the label definition at the proper place.
6442 Here is how to do this:
6445 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
6448 When you output a pool entry specially, you should end with a
6449 @code{goto} to the label @var{jumpto}. This will prevent the same pool
6450 entry from being output a second time in the usual manner.
6452 You need not define this macro if it would do nothing.
6455 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
6456 A C statement to output assembler commands to at the end of the constant
6457 pool for a function. @var{funname} is a string giving the name of the
6458 function. Should the return type of the function be required, you can
6459 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
6460 constant pool that GCC wrote immediately before this call.
6462 If no constant-pool epilogue is required, the usual case, you need not
6466 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
6467 Define this macro as a C expression which is nonzero if @var{C} is
6468 used as a logical line separator by the assembler.
6470 If you do not define this macro, the default is that only
6471 the character @samp{;} is treated as a logical line separator.
6474 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
6475 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
6476 These target hooks are C string constants, describing the syntax in the
6477 assembler for grouping arithmetic expressions. If not overridden, they
6478 default to normal parentheses, which is correct for most assemblers.
6481 These macros are provided by @file{real.h} for writing the definitions
6482 of @code{ASM_OUTPUT_DOUBLE} and the like:
6484 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
6485 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
6486 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
6487 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
6488 floating point representation, and store its bit pattern in the variable
6489 @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE}, this variable should
6490 be a simple @code{long int}. For the others, it should be an array of
6491 @code{long int}. The number of elements in this array is determined by
6492 the size of the desired target floating point data type: 32 bits of it
6493 go in each @code{long int} array element. Each array element holds 32
6494 bits of the result, even if @code{long int} is wider than 32 bits on the
6497 The array element values are designed so that you can print them out
6498 using @code{fprintf} in the order they should appear in the target
6502 @node Uninitialized Data
6503 @subsection Output of Uninitialized Variables
6505 Each of the macros in this section is used to do the whole job of
6506 outputting a single uninitialized variable.
6508 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6509 A C statement (sans semicolon) to output to the stdio stream
6510 @var{stream} the assembler definition of a common-label named
6511 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6512 is the size rounded up to whatever alignment the caller wants.
6514 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6515 output the name itself; before and after that, output the additional
6516 assembler syntax for defining the name, and a newline.
6518 This macro controls how the assembler definitions of uninitialized
6519 common global variables are output.
6522 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
6523 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
6524 separate, explicit argument. If you define this macro, it is used in
6525 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
6526 handling the required alignment of the variable. The alignment is specified
6527 as the number of bits.
6530 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6531 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
6532 variable to be output, if there is one, or @code{NULL_TREE} if there
6533 is no corresponding variable. If you define this macro, GCC will use it
6534 in place of both @code{ASM_OUTPUT_COMMON} and
6535 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
6536 the variable's decl in order to chose what to output.
6539 @defmac ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6540 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
6541 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
6545 @defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6546 A C statement (sans semicolon) to output to the stdio stream
6547 @var{stream} the assembler definition of uninitialized global @var{decl} named
6548 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6549 is the size rounded up to whatever alignment the caller wants.
6551 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
6552 defining this macro. If unable, use the expression
6553 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
6554 before and after that, output the additional assembler syntax for defining
6555 the name, and a newline.
6557 This macro controls how the assembler definitions of uninitialized global
6558 variables are output. This macro exists to properly support languages like
6559 C++ which do not have @code{common} data. However, this macro currently
6560 is not defined for all targets. If this macro and
6561 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
6562 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
6563 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
6566 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6567 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
6568 separate, explicit argument. If you define this macro, it is used in
6569 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
6570 handling the required alignment of the variable. The alignment is specified
6571 as the number of bits.
6573 Try to use function @code{asm_output_aligned_bss} defined in file
6574 @file{varasm.c} when defining this macro.
6577 @defmac ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6578 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
6579 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
6583 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6584 A C statement (sans semicolon) to output to the stdio stream
6585 @var{stream} the assembler definition of a local-common-label named
6586 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6587 is the size rounded up to whatever alignment the caller wants.
6589 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6590 output the name itself; before and after that, output the additional
6591 assembler syntax for defining the name, and a newline.
6593 This macro controls how the assembler definitions of uninitialized
6594 static variables are output.
6597 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
6598 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
6599 separate, explicit argument. If you define this macro, it is used in
6600 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
6601 handling the required alignment of the variable. The alignment is specified
6602 as the number of bits.
6605 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6606 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
6607 variable to be output, if there is one, or @code{NULL_TREE} if there
6608 is no corresponding variable. If you define this macro, GCC will use it
6609 in place of both @code{ASM_OUTPUT_DECL} and
6610 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
6611 the variable's decl in order to chose what to output.
6614 @defmac ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6615 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
6616 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
6621 @subsection Output and Generation of Labels
6623 @c prevent bad page break with this line
6624 This is about outputting labels.
6626 @findex assemble_name
6627 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
6628 A C statement (sans semicolon) to output to the stdio stream
6629 @var{stream} the assembler definition of a label named @var{name}.
6630 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6631 output the name itself; before and after that, output the additional
6632 assembler syntax for defining the name, and a newline. A default
6633 definition of this macro is provided which is correct for most systems.
6636 @findex assemble_name_raw
6637 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
6638 Identical to @code{ASM_OUTPUT_lABEL}, except that @var{name} is known
6639 to refer to a compiler-generated label. The default definition uses
6640 @code{assemble_name_raw}, which is like @code{assemble_name} except
6641 that it is more efficient.
6645 A C string containing the appropriate assembler directive to specify the
6646 size of a symbol, without any arguments. On systems that use ELF, the
6647 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
6648 systems, the default is not to define this macro.
6650 Define this macro only if it is correct to use the default definitions
6651 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
6652 for your system. If you need your own custom definitions of those
6653 macros, or if you do not need explicit symbol sizes at all, do not
6657 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
6658 A C statement (sans semicolon) to output to the stdio stream
6659 @var{stream} a directive telling the assembler that the size of the
6660 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
6661 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6665 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
6666 A C statement (sans semicolon) to output to the stdio stream
6667 @var{stream} a directive telling the assembler to calculate the size of
6668 the symbol @var{name} by subtracting its address from the current
6671 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6672 provided. The default assumes that the assembler recognizes a special
6673 @samp{.} symbol as referring to the current address, and can calculate
6674 the difference between this and another symbol. If your assembler does
6675 not recognize @samp{.} or cannot do calculations with it, you will need
6676 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
6680 A C string containing the appropriate assembler directive to specify the
6681 type of a symbol, without any arguments. On systems that use ELF, the
6682 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
6683 systems, the default is not to define this macro.
6685 Define this macro only if it is correct to use the default definition of
6686 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6687 custom definition of this macro, or if you do not need explicit symbol
6688 types at all, do not define this macro.
6691 @defmac TYPE_OPERAND_FMT
6692 A C string which specifies (using @code{printf} syntax) the format of
6693 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
6694 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
6695 the default is not to define this macro.
6697 Define this macro only if it is correct to use the default definition of
6698 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6699 custom definition of this macro, or if you do not need explicit symbol
6700 types at all, do not define this macro.
6703 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
6704 A C statement (sans semicolon) to output to the stdio stream
6705 @var{stream} a directive telling the assembler that the type of the
6706 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
6707 that string is always either @samp{"function"} or @samp{"object"}, but
6708 you should not count on this.
6710 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
6711 definition of this macro is provided.
6714 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
6715 A C statement (sans semicolon) to output to the stdio stream
6716 @var{stream} any text necessary for declaring the name @var{name} of a
6717 function which is being defined. This macro is responsible for
6718 outputting the label definition (perhaps using
6719 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
6720 @code{FUNCTION_DECL} tree node representing the function.
6722 If this macro is not defined, then the function name is defined in the
6723 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6725 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6729 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
6730 A C statement (sans semicolon) to output to the stdio stream
6731 @var{stream} any text necessary for declaring the size of a function
6732 which is being defined. The argument @var{name} is the name of the
6733 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
6734 representing the function.
6736 If this macro is not defined, then the function size is not defined.
6738 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
6742 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6743 A C statement (sans semicolon) to output to the stdio stream
6744 @var{stream} any text necessary for declaring the name @var{name} of an
6745 initialized variable which is being defined. This macro must output the
6746 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6747 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6749 If this macro is not defined, then the variable name is defined in the
6750 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6752 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
6753 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
6756 @defmac ASM_DECLARE_CONSTANT_NAME (@var{stream}, @var{name}, @var{exp}, @var{size})
6757 A C statement (sans semicolon) to output to the stdio stream
6758 @var{stream} any text necessary for declaring the name @var{name} of a
6759 constant which is being defined. This macro is responsible for
6760 outputting the label definition (perhaps using
6761 @code{ASM_OUTPUT_LABEL}). The argument @var{exp} is the
6762 value of the constant, and @var{size} is the size of the constant
6763 in bytes. @var{name} will be an internal label.
6765 If this macro is not defined, then the @var{name} is defined in the
6766 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6768 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6772 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6773 A C statement (sans semicolon) to output to the stdio stream
6774 @var{stream} any text necessary for claiming a register @var{regno}
6775 for a global variable @var{decl} with name @var{name}.
6777 If you don't define this macro, that is equivalent to defining it to do
6781 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6782 A C statement (sans semicolon) to finish up declaring a variable name
6783 once the compiler has processed its initializer fully and thus has had a
6784 chance to determine the size of an array when controlled by an
6785 initializer. This is used on systems where it's necessary to declare
6786 something about the size of the object.
6788 If you don't define this macro, that is equivalent to defining it to do
6791 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
6792 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
6795 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
6796 This target hook is a function to output to the stdio stream
6797 @var{stream} some commands that will make the label @var{name} global;
6798 that is, available for reference from other files.
6800 The default implementation relies on a proper definition of
6801 @code{GLOBAL_ASM_OP}.
6804 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
6805 A C statement (sans semicolon) to output to the stdio stream
6806 @var{stream} some commands that will make the label @var{name} weak;
6807 that is, available for reference from other files but only used if
6808 no other definition is available. Use the expression
6809 @code{assemble_name (@var{stream}, @var{name})} to output the name
6810 itself; before and after that, output the additional assembler syntax
6811 for making that name weak, and a newline.
6813 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
6814 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
6818 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
6819 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
6820 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
6821 or variable decl. If @var{value} is not @code{NULL}, this C statement
6822 should output to the stdio stream @var{stream} assembler code which
6823 defines (equates) the weak symbol @var{name} to have the value
6824 @var{value}. If @var{value} is @code{NULL}, it should output commands
6825 to make @var{name} weak.
6828 @defmac SUPPORTS_WEAK
6829 A C expression which evaluates to true if the target supports weak symbols.
6831 If you don't define this macro, @file{defaults.h} provides a default
6832 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
6833 is defined, the default definition is @samp{1}; otherwise, it is
6834 @samp{0}. Define this macro if you want to control weak symbol support
6835 with a compiler flag such as @option{-melf}.
6838 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
6839 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
6840 public symbol such that extra copies in multiple translation units will
6841 be discarded by the linker. Define this macro if your object file
6842 format provides support for this concept, such as the @samp{COMDAT}
6843 section flags in the Microsoft Windows PE/COFF format, and this support
6844 requires changes to @var{decl}, such as putting it in a separate section.
6847 @defmac SUPPORTS_ONE_ONLY
6848 A C expression which evaluates to true if the target supports one-only
6851 If you don't define this macro, @file{varasm.c} provides a default
6852 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6853 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6854 you want to control one-only symbol support with a compiler flag, or if
6855 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6856 be emitted as one-only.
6859 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
6860 This target hook is a function to output to @var{asm_out_file} some
6861 commands that will make the symbol(s) associated with @var{decl} have
6862 hidden, protected or internal visibility as specified by @var{visibility}.
6865 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
6866 A C expression that evaluates to true if the target's linker expects
6867 that weak symbols do not appear in a static archive's table of contents.
6868 The default is @code{0}.
6870 Leaving weak symbols out of an archive's table of contents means that,
6871 if a symbol will only have a definition in one translation unit and
6872 will have undefined references from other translation units, that
6873 symbol should not be weak. Defining this macro to be nonzero will
6874 thus have the effect that certain symbols that would normally be weak
6875 (explicit template instantiations, and vtables for polymorphic classes
6876 with noninline key methods) will instead be nonweak.
6878 The C++ ABI requires this macro to be zero. Define this macro for
6879 targets where full C++ ABI compliance is impossible and where linker
6880 restrictions require weak symbols to be left out of a static archive's
6884 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6885 A C statement (sans semicolon) to output to the stdio stream
6886 @var{stream} any text necessary for declaring the name of an external
6887 symbol named @var{name} which is referenced in this compilation but
6888 not defined. The value of @var{decl} is the tree node for the
6891 This macro need not be defined if it does not need to output anything.
6892 The GNU assembler and most Unix assemblers don't require anything.
6895 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
6896 This target hook is a function to output to @var{asm_out_file} an assembler
6897 pseudo-op to declare a library function name external. The name of the
6898 library function is given by @var{symref}, which is a @code{symbol_ref}.
6901 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (tree @var{decl})
6902 This target hook is a function to output to @var{asm_out_file} an assembler
6903 directive to annotate used symbol. Darwin target use .no_dead_code_strip
6907 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6908 A C statement (sans semicolon) to output to the stdio stream
6909 @var{stream} a reference in assembler syntax to a label named
6910 @var{name}. This should add @samp{_} to the front of the name, if that
6911 is customary on your operating system, as it is in most Berkeley Unix
6912 systems. This macro is used in @code{assemble_name}.
6915 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6916 A C statement (sans semicolon) to output a reference to
6917 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
6918 will be used to output the name of the symbol. This macro may be used
6919 to modify the way a symbol is referenced depending on information
6920 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
6923 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
6924 A C statement (sans semicolon) to output a reference to @var{buf}, the
6925 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
6926 @code{assemble_name} will be used to output the name of the symbol.
6927 This macro is not used by @code{output_asm_label}, or the @code{%l}
6928 specifier that calls it; the intention is that this macro should be set
6929 when it is necessary to output a label differently when its address is
6933 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
6934 A function to output to the stdio stream @var{stream} a label whose
6935 name is made from the string @var{prefix} and the number @var{labelno}.
6937 It is absolutely essential that these labels be distinct from the labels
6938 used for user-level functions and variables. Otherwise, certain programs
6939 will have name conflicts with internal labels.
6941 It is desirable to exclude internal labels from the symbol table of the
6942 object file. Most assemblers have a naming convention for labels that
6943 should be excluded; on many systems, the letter @samp{L} at the
6944 beginning of a label has this effect. You should find out what
6945 convention your system uses, and follow it.
6947 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
6950 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
6951 A C statement to output to the stdio stream @var{stream} a debug info
6952 label whose name is made from the string @var{prefix} and the number
6953 @var{num}. This is useful for VLIW targets, where debug info labels
6954 may need to be treated differently than branch target labels. On some
6955 systems, branch target labels must be at the beginning of instruction
6956 bundles, but debug info labels can occur in the middle of instruction
6959 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
6963 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
6964 A C statement to store into the string @var{string} a label whose name
6965 is made from the string @var{prefix} and the number @var{num}.
6967 This string, when output subsequently by @code{assemble_name}, should
6968 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
6969 with the same @var{prefix} and @var{num}.
6971 If the string begins with @samp{*}, then @code{assemble_name} will
6972 output the rest of the string unchanged. It is often convenient for
6973 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
6974 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
6975 to output the string, and may change it. (Of course,
6976 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
6977 you should know what it does on your machine.)
6980 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
6981 A C expression to assign to @var{outvar} (which is a variable of type
6982 @code{char *}) a newly allocated string made from the string
6983 @var{name} and the number @var{number}, with some suitable punctuation
6984 added. Use @code{alloca} to get space for the string.
6986 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
6987 produce an assembler label for an internal static variable whose name is
6988 @var{name}. Therefore, the string must be such as to result in valid
6989 assembler code. The argument @var{number} is different each time this
6990 macro is executed; it prevents conflicts between similarly-named
6991 internal static variables in different scopes.
6993 Ideally this string should not be a valid C identifier, to prevent any
6994 conflict with the user's own symbols. Most assemblers allow periods
6995 or percent signs in assembler symbols; putting at least one of these
6996 between the name and the number will suffice.
6998 If this macro is not defined, a default definition will be provided
6999 which is correct for most systems.
7002 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
7003 A C statement to output to the stdio stream @var{stream} assembler code
7004 which defines (equates) the symbol @var{name} to have the value @var{value}.
7007 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7008 correct for most systems.
7011 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
7012 A C statement to output to the stdio stream @var{stream} assembler code
7013 which defines (equates) the symbol whose tree node is @var{decl_of_name}
7014 to have the value of the tree node @var{decl_of_value}. This macro will
7015 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
7016 the tree nodes are available.
7019 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7020 correct for most systems.
7023 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
7024 A C statement that evaluates to true if the assembler code which defines
7025 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
7026 of the tree node @var{decl_of_value} should be emitted near the end of the
7027 current compilation unit. The default is to not defer output of defines.
7028 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
7029 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
7032 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
7033 A C statement to output to the stdio stream @var{stream} assembler code
7034 which defines (equates) the weak symbol @var{name} to have the value
7035 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
7036 an undefined weak symbol.
7038 Define this macro if the target only supports weak aliases; define
7039 @code{ASM_OUTPUT_DEF} instead if possible.
7042 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
7043 Define this macro to override the default assembler names used for
7044 Objective-C methods.
7046 The default name is a unique method number followed by the name of the
7047 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
7048 the category is also included in the assembler name (e.g.@:
7051 These names are safe on most systems, but make debugging difficult since
7052 the method's selector is not present in the name. Therefore, particular
7053 systems define other ways of computing names.
7055 @var{buf} is an expression of type @code{char *} which gives you a
7056 buffer in which to store the name; its length is as long as
7057 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
7058 50 characters extra.
7060 The argument @var{is_inst} specifies whether the method is an instance
7061 method or a class method; @var{class_name} is the name of the class;
7062 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
7063 in a category); and @var{sel_name} is the name of the selector.
7065 On systems where the assembler can handle quoted names, you can use this
7066 macro to provide more human-readable names.
7069 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
7070 A C statement (sans semicolon) to output to the stdio stream
7071 @var{stream} commands to declare that the label @var{name} is an
7072 Objective-C class reference. This is only needed for targets whose
7073 linkers have special support for NeXT-style runtimes.
7076 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
7077 A C statement (sans semicolon) to output to the stdio stream
7078 @var{stream} commands to declare that the label @var{name} is an
7079 unresolved Objective-C class reference. This is only needed for targets
7080 whose linkers have special support for NeXT-style runtimes.
7083 @node Initialization
7084 @subsection How Initialization Functions Are Handled
7085 @cindex initialization routines
7086 @cindex termination routines
7087 @cindex constructors, output of
7088 @cindex destructors, output of
7090 The compiled code for certain languages includes @dfn{constructors}
7091 (also called @dfn{initialization routines})---functions to initialize
7092 data in the program when the program is started. These functions need
7093 to be called before the program is ``started''---that is to say, before
7094 @code{main} is called.
7096 Compiling some languages generates @dfn{destructors} (also called
7097 @dfn{termination routines}) that should be called when the program
7100 To make the initialization and termination functions work, the compiler
7101 must output something in the assembler code to cause those functions to
7102 be called at the appropriate time. When you port the compiler to a new
7103 system, you need to specify how to do this.
7105 There are two major ways that GCC currently supports the execution of
7106 initialization and termination functions. Each way has two variants.
7107 Much of the structure is common to all four variations.
7109 @findex __CTOR_LIST__
7110 @findex __DTOR_LIST__
7111 The linker must build two lists of these functions---a list of
7112 initialization functions, called @code{__CTOR_LIST__}, and a list of
7113 termination functions, called @code{__DTOR_LIST__}.
7115 Each list always begins with an ignored function pointer (which may hold
7116 0, @minus{}1, or a count of the function pointers after it, depending on
7117 the environment). This is followed by a series of zero or more function
7118 pointers to constructors (or destructors), followed by a function
7119 pointer containing zero.
7121 Depending on the operating system and its executable file format, either
7122 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
7123 time and exit time. Constructors are called in reverse order of the
7124 list; destructors in forward order.
7126 The best way to handle static constructors works only for object file
7127 formats which provide arbitrarily-named sections. A section is set
7128 aside for a list of constructors, and another for a list of destructors.
7129 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
7130 object file that defines an initialization function also puts a word in
7131 the constructor section to point to that function. The linker
7132 accumulates all these words into one contiguous @samp{.ctors} section.
7133 Termination functions are handled similarly.
7135 This method will be chosen as the default by @file{target-def.h} if
7136 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
7137 support arbitrary sections, but does support special designated
7138 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
7139 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
7141 When arbitrary sections are available, there are two variants, depending
7142 upon how the code in @file{crtstuff.c} is called. On systems that
7143 support a @dfn{.init} section which is executed at program startup,
7144 parts of @file{crtstuff.c} are compiled into that section. The
7145 program is linked by the @command{gcc} driver like this:
7148 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
7151 The prologue of a function (@code{__init}) appears in the @code{.init}
7152 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
7153 for the function @code{__fini} in the @dfn{.fini} section. Normally these
7154 files are provided by the operating system or by the GNU C library, but
7155 are provided by GCC for a few targets.
7157 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
7158 compiled from @file{crtstuff.c}. They contain, among other things, code
7159 fragments within the @code{.init} and @code{.fini} sections that branch
7160 to routines in the @code{.text} section. The linker will pull all parts
7161 of a section together, which results in a complete @code{__init} function
7162 that invokes the routines we need at startup.
7164 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
7167 If no init section is available, when GCC compiles any function called
7168 @code{main} (or more accurately, any function designated as a program
7169 entry point by the language front end calling @code{expand_main_function}),
7170 it inserts a procedure call to @code{__main} as the first executable code
7171 after the function prologue. The @code{__main} function is defined
7172 in @file{libgcc2.c} and runs the global constructors.
7174 In file formats that don't support arbitrary sections, there are again
7175 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
7176 and an `a.out' format must be used. In this case,
7177 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
7178 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
7179 and with the address of the void function containing the initialization
7180 code as its value. The GNU linker recognizes this as a request to add
7181 the value to a @dfn{set}; the values are accumulated, and are eventually
7182 placed in the executable as a vector in the format described above, with
7183 a leading (ignored) count and a trailing zero element.
7184 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
7185 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
7186 the compilation of @code{main} to call @code{__main} as above, starting
7187 the initialization process.
7189 The last variant uses neither arbitrary sections nor the GNU linker.
7190 This is preferable when you want to do dynamic linking and when using
7191 file formats which the GNU linker does not support, such as `ECOFF'@. In
7192 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
7193 termination functions are recognized simply by their names. This requires
7194 an extra program in the linkage step, called @command{collect2}. This program
7195 pretends to be the linker, for use with GCC; it does its job by running
7196 the ordinary linker, but also arranges to include the vectors of
7197 initialization and termination functions. These functions are called
7198 via @code{__main} as described above. In order to use this method,
7199 @code{use_collect2} must be defined in the target in @file{config.gcc}.
7202 The following section describes the specific macros that control and
7203 customize the handling of initialization and termination functions.
7206 @node Macros for Initialization
7207 @subsection Macros Controlling Initialization Routines
7209 Here are the macros that control how the compiler handles initialization
7210 and termination functions:
7212 @defmac INIT_SECTION_ASM_OP
7213 If defined, a C string constant, including spacing, for the assembler
7214 operation to identify the following data as initialization code. If not
7215 defined, GCC will assume such a section does not exist. When you are
7216 using special sections for initialization and termination functions, this
7217 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7218 run the initialization functions.
7221 @defmac HAS_INIT_SECTION
7222 If defined, @code{main} will not call @code{__main} as described above.
7223 This macro should be defined for systems that control start-up code
7224 on a symbol-by-symbol basis, such as OSF/1, and should not
7225 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7228 @defmac LD_INIT_SWITCH
7229 If defined, a C string constant for a switch that tells the linker that
7230 the following symbol is an initialization routine.
7233 @defmac LD_FINI_SWITCH
7234 If defined, a C string constant for a switch that tells the linker that
7235 the following symbol is a finalization routine.
7238 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7239 If defined, a C statement that will write a function that can be
7240 automatically called when a shared library is loaded. The function
7241 should call @var{func}, which takes no arguments. If not defined, and
7242 the object format requires an explicit initialization function, then a
7243 function called @code{_GLOBAL__DI} will be generated.
7245 This function and the following one are used by collect2 when linking a
7246 shared library that needs constructors or destructors, or has DWARF2
7247 exception tables embedded in the code.
7250 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7251 If defined, a C statement that will write a function that can be
7252 automatically called when a shared library is unloaded. The function
7253 should call @var{func}, which takes no arguments. If not defined, and
7254 the object format requires an explicit finalization function, then a
7255 function called @code{_GLOBAL__DD} will be generated.
7258 @defmac INVOKE__main
7259 If defined, @code{main} will call @code{__main} despite the presence of
7260 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7261 where the init section is not actually run automatically, but is still
7262 useful for collecting the lists of constructors and destructors.
7265 @defmac SUPPORTS_INIT_PRIORITY
7266 If nonzero, the C++ @code{init_priority} attribute is supported and the
7267 compiler should emit instructions to control the order of initialization
7268 of objects. If zero, the compiler will issue an error message upon
7269 encountering an @code{init_priority} attribute.
7272 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7273 This value is true if the target supports some ``native'' method of
7274 collecting constructors and destructors to be run at startup and exit.
7275 It is false if we must use @command{collect2}.
7278 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7279 If defined, a function that outputs assembler code to arrange to call
7280 the function referenced by @var{symbol} at initialization time.
7282 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7283 no arguments and with no return value. If the target supports initialization
7284 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7285 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7287 If this macro is not defined by the target, a suitable default will
7288 be chosen if (1) the target supports arbitrary section names, (2) the
7289 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7293 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7294 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7295 functions rather than initialization functions.
7298 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7299 generated for the generated object file will have static linkage.
7301 If your system uses @command{collect2} as the means of processing
7302 constructors, then that program normally uses @command{nm} to scan
7303 an object file for constructor functions to be called.
7305 On certain kinds of systems, you can define this macro to make
7306 @command{collect2} work faster (and, in some cases, make it work at all):
7308 @defmac OBJECT_FORMAT_COFF
7309 Define this macro if the system uses COFF (Common Object File Format)
7310 object files, so that @command{collect2} can assume this format and scan
7311 object files directly for dynamic constructor/destructor functions.
7313 This macro is effective only in a native compiler; @command{collect2} as
7314 part of a cross compiler always uses @command{nm} for the target machine.
7317 @defmac REAL_NM_FILE_NAME
7318 Define this macro as a C string constant containing the file name to use
7319 to execute @command{nm}. The default is to search the path normally for
7322 If your system supports shared libraries and has a program to list the
7323 dynamic dependencies of a given library or executable, you can define
7324 these macros to enable support for running initialization and
7325 termination functions in shared libraries:
7329 Define this macro to a C string constant containing the name of the program
7330 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7333 @defmac PARSE_LDD_OUTPUT (@var{ptr})
7334 Define this macro to be C code that extracts filenames from the output
7335 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7336 of type @code{char *} that points to the beginning of a line of output
7337 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7338 code must advance @var{ptr} to the beginning of the filename on that
7339 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7342 @node Instruction Output
7343 @subsection Output of Assembler Instructions
7345 @c prevent bad page break with this line
7346 This describes assembler instruction output.
7348 @defmac REGISTER_NAMES
7349 A C initializer containing the assembler's names for the machine
7350 registers, each one as a C string constant. This is what translates
7351 register numbers in the compiler into assembler language.
7354 @defmac ADDITIONAL_REGISTER_NAMES
7355 If defined, a C initializer for an array of structures containing a name
7356 and a register number. This macro defines additional names for hard
7357 registers, thus allowing the @code{asm} option in declarations to refer
7358 to registers using alternate names.
7361 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7362 Define this macro if you are using an unusual assembler that
7363 requires different names for the machine instructions.
7365 The definition is a C statement or statements which output an
7366 assembler instruction opcode to the stdio stream @var{stream}. The
7367 macro-operand @var{ptr} is a variable of type @code{char *} which
7368 points to the opcode name in its ``internal'' form---the form that is
7369 written in the machine description. The definition should output the
7370 opcode name to @var{stream}, performing any translation you desire, and
7371 increment the variable @var{ptr} to point at the end of the opcode
7372 so that it will not be output twice.
7374 In fact, your macro definition may process less than the entire opcode
7375 name, or more than the opcode name; but if you want to process text
7376 that includes @samp{%}-sequences to substitute operands, you must take
7377 care of the substitution yourself. Just be sure to increment
7378 @var{ptr} over whatever text should not be output normally.
7380 @findex recog_data.operand
7381 If you need to look at the operand values, they can be found as the
7382 elements of @code{recog_data.operand}.
7384 If the macro definition does nothing, the instruction is output
7388 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7389 If defined, a C statement to be executed just prior to the output of
7390 assembler code for @var{insn}, to modify the extracted operands so
7391 they will be output differently.
7393 Here the argument @var{opvec} is the vector containing the operands
7394 extracted from @var{insn}, and @var{noperands} is the number of
7395 elements of the vector which contain meaningful data for this insn.
7396 The contents of this vector are what will be used to convert the insn
7397 template into assembler code, so you can change the assembler output
7398 by changing the contents of the vector.
7400 This macro is useful when various assembler syntaxes share a single
7401 file of instruction patterns; by defining this macro differently, you
7402 can cause a large class of instructions to be output differently (such
7403 as with rearranged operands). Naturally, variations in assembler
7404 syntax affecting individual insn patterns ought to be handled by
7405 writing conditional output routines in those patterns.
7407 If this macro is not defined, it is equivalent to a null statement.
7410 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
7411 A C compound statement to output to stdio stream @var{stream} the
7412 assembler syntax for an instruction operand @var{x}. @var{x} is an
7415 @var{code} is a value that can be used to specify one of several ways
7416 of printing the operand. It is used when identical operands must be
7417 printed differently depending on the context. @var{code} comes from
7418 the @samp{%} specification that was used to request printing of the
7419 operand. If the specification was just @samp{%@var{digit}} then
7420 @var{code} is 0; if the specification was @samp{%@var{ltr}
7421 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
7424 If @var{x} is a register, this macro should print the register's name.
7425 The names can be found in an array @code{reg_names} whose type is
7426 @code{char *[]}. @code{reg_names} is initialized from
7427 @code{REGISTER_NAMES}.
7429 When the machine description has a specification @samp{%@var{punct}}
7430 (a @samp{%} followed by a punctuation character), this macro is called
7431 with a null pointer for @var{x} and the punctuation character for
7435 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
7436 A C expression which evaluates to true if @var{code} is a valid
7437 punctuation character for use in the @code{PRINT_OPERAND} macro. If
7438 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
7439 punctuation characters (except for the standard one, @samp{%}) are used
7443 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
7444 A C compound statement to output to stdio stream @var{stream} the
7445 assembler syntax for an instruction operand that is a memory reference
7446 whose address is @var{x}. @var{x} is an RTL expression.
7448 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
7449 On some machines, the syntax for a symbolic address depends on the
7450 section that the address refers to. On these machines, define the hook
7451 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
7452 @code{symbol_ref}, and then check for it here. @xref{Assembler
7456 @findex dbr_sequence_length
7457 @defmac DBR_OUTPUT_SEQEND (@var{file})
7458 A C statement, to be executed after all slot-filler instructions have
7459 been output. If necessary, call @code{dbr_sequence_length} to
7460 determine the number of slots filled in a sequence (zero if not
7461 currently outputting a sequence), to decide how many no-ops to output,
7464 Don't define this macro if it has nothing to do, but it is helpful in
7465 reading assembly output if the extent of the delay sequence is made
7466 explicit (e.g.@: with white space).
7469 @findex final_sequence
7470 Note that output routines for instructions with delay slots must be
7471 prepared to deal with not being output as part of a sequence
7472 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
7473 found.) The variable @code{final_sequence} is null when not
7474 processing a sequence, otherwise it contains the @code{sequence} rtx
7478 @defmac REGISTER_PREFIX
7479 @defmacx LOCAL_LABEL_PREFIX
7480 @defmacx USER_LABEL_PREFIX
7481 @defmacx IMMEDIATE_PREFIX
7482 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
7483 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
7484 @file{final.c}). These are useful when a single @file{md} file must
7485 support multiple assembler formats. In that case, the various @file{tm.h}
7486 files can define these macros differently.
7489 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
7490 If defined this macro should expand to a series of @code{case}
7491 statements which will be parsed inside the @code{switch} statement of
7492 the @code{asm_fprintf} function. This allows targets to define extra
7493 printf formats which may useful when generating their assembler
7494 statements. Note that uppercase letters are reserved for future
7495 generic extensions to asm_fprintf, and so are not available to target
7496 specific code. The output file is given by the parameter @var{file}.
7497 The varargs input pointer is @var{argptr} and the rest of the format
7498 string, starting the character after the one that is being switched
7499 upon, is pointed to by @var{format}.
7502 @defmac ASSEMBLER_DIALECT
7503 If your target supports multiple dialects of assembler language (such as
7504 different opcodes), define this macro as a C expression that gives the
7505 numeric index of the assembler language dialect to use, with zero as the
7508 If this macro is defined, you may use constructs of the form
7510 @samp{@{option0|option1|option2@dots{}@}}
7513 in the output templates of patterns (@pxref{Output Template}) or in the
7514 first argument of @code{asm_fprintf}. This construct outputs
7515 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
7516 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
7517 within these strings retain their usual meaning. If there are fewer
7518 alternatives within the braces than the value of
7519 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
7521 If you do not define this macro, the characters @samp{@{}, @samp{|} and
7522 @samp{@}} do not have any special meaning when used in templates or
7523 operands to @code{asm_fprintf}.
7525 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
7526 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
7527 the variations in assembler language syntax with that mechanism. Define
7528 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
7529 if the syntax variant are larger and involve such things as different
7530 opcodes or operand order.
7533 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
7534 A C expression to output to @var{stream} some assembler code
7535 which will push hard register number @var{regno} onto the stack.
7536 The code need not be optimal, since this macro is used only when
7540 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
7541 A C expression to output to @var{stream} some assembler code
7542 which will pop hard register number @var{regno} off of the stack.
7543 The code need not be optimal, since this macro is used only when
7547 @node Dispatch Tables
7548 @subsection Output of Dispatch Tables
7550 @c prevent bad page break with this line
7551 This concerns dispatch tables.
7553 @cindex dispatch table
7554 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
7555 A C statement to output to the stdio stream @var{stream} an assembler
7556 pseudo-instruction to generate a difference between two labels.
7557 @var{value} and @var{rel} are the numbers of two internal labels. The
7558 definitions of these labels are output using
7559 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
7560 way here. For example,
7563 fprintf (@var{stream}, "\t.word L%d-L%d\n",
7564 @var{value}, @var{rel})
7567 You must provide this macro on machines where the addresses in a
7568 dispatch table are relative to the table's own address. If defined, GCC
7569 will also use this macro on all machines when producing PIC@.
7570 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
7571 mode and flags can be read.
7574 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
7575 This macro should be provided on machines where the addresses
7576 in a dispatch table are absolute.
7578 The definition should be a C statement to output to the stdio stream
7579 @var{stream} an assembler pseudo-instruction to generate a reference to
7580 a label. @var{value} is the number of an internal label whose
7581 definition is output using @code{(*targetm.asm_out.internal_label)}.
7585 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
7589 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
7590 Define this if the label before a jump-table needs to be output
7591 specially. The first three arguments are the same as for
7592 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
7593 jump-table which follows (a @code{jump_insn} containing an
7594 @code{addr_vec} or @code{addr_diff_vec}).
7596 This feature is used on system V to output a @code{swbeg} statement
7599 If this macro is not defined, these labels are output with
7600 @code{(*targetm.asm_out.internal_label)}.
7603 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
7604 Define this if something special must be output at the end of a
7605 jump-table. The definition should be a C statement to be executed
7606 after the assembler code for the table is written. It should write
7607 the appropriate code to stdio stream @var{stream}. The argument
7608 @var{table} is the jump-table insn, and @var{num} is the label-number
7609 of the preceding label.
7611 If this macro is not defined, nothing special is output at the end of
7615 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (@var{stream}, @var{decl}, @var{for_eh}, @var{empty})
7616 This target hook emits a label at the beginning of each FDE@. It
7617 should be defined on targets where FDEs need special labels, and it
7618 should write the appropriate label, for the FDE associated with the
7619 function declaration @var{decl}, to the stdio stream @var{stream}.
7620 The third argument, @var{for_eh}, is a boolean: true if this is for an
7621 exception table. The fourth argument, @var{empty}, is a boolean:
7622 true if this is a placeholder label for an omitted FDE@.
7624 The default is that FDEs are not given nonlocal labels.
7627 @deftypefn {Taget Hook} void TARGET_UNWIND_EMIT (FILE * @var{stream}, rtx @var{insn})
7628 This target hook emits and assembly directives required to unwind the
7629 given instruction. This is only used when TARGET_UNWIND_INFO is set.
7632 @node Exception Region Output
7633 @subsection Assembler Commands for Exception Regions
7635 @c prevent bad page break with this line
7637 This describes commands marking the start and the end of an exception
7640 @defmac EH_FRAME_SECTION_NAME
7641 If defined, a C string constant for the name of the section containing
7642 exception handling frame unwind information. If not defined, GCC will
7643 provide a default definition if the target supports named sections.
7644 @file{crtstuff.c} uses this macro to switch to the appropriate section.
7646 You should define this symbol if your target supports DWARF 2 frame
7647 unwind information and the default definition does not work.
7650 @defmac EH_FRAME_IN_DATA_SECTION
7651 If defined, DWARF 2 frame unwind information will be placed in the
7652 data section even though the target supports named sections. This
7653 might be necessary, for instance, if the system linker does garbage
7654 collection and sections cannot be marked as not to be collected.
7656 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
7660 @defmac EH_TABLES_CAN_BE_READ_ONLY
7661 Define this macro to 1 if your target is such that no frame unwind
7662 information encoding used with non-PIC code will ever require a
7663 runtime relocation, but the linker may not support merging read-only
7664 and read-write sections into a single read-write section.
7667 @defmac MASK_RETURN_ADDR
7668 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
7669 that it does not contain any extraneous set bits in it.
7672 @defmac DWARF2_UNWIND_INFO
7673 Define this macro to 0 if your target supports DWARF 2 frame unwind
7674 information, but it does not yet work with exception handling.
7675 Otherwise, if your target supports this information (if it defines
7676 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
7677 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
7680 If @code{TARGET_UNWIND_INFO} is defined, the target specific unwinder
7681 will be used in all cases. Defining this macro will enable the generation
7682 of DWARF 2 frame debugging information.
7684 If @code{TARGET_UNWIND_INFO} is not defined, and this macro is defined to 1,
7685 the DWARF 2 unwinder will be the default exception handling mechanism;
7686 otherwise, @code{setjmp}/@code{longjmp} will be used by default.
7689 @defmac TARGET_UNWIND_INFO
7690 Define this macro if your target has ABI specified unwind tables. Usually
7691 these will be output by @code{TARGET_UNWIND_EMIT}.
7694 @defmac MUST_USE_SJLJ_EXCEPTIONS
7695 This macro need only be defined if @code{DWARF2_UNWIND_INFO} is
7696 runtime-variable. In that case, @file{except.h} cannot correctly
7697 determine the corresponding definition of
7698 @code{MUST_USE_SJLJ_EXCEPTIONS}, so the target must provide it directly.
7701 @defmac DWARF_CIE_DATA_ALIGNMENT
7702 This macro need only be defined if the target might save registers in the
7703 function prologue at an offset to the stack pointer that is not aligned to
7704 @code{UNITS_PER_WORD}. The definition should be the negative minimum
7705 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
7706 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
7707 the target supports DWARF 2 frame unwind information.
7710 @deftypefn {Target Hook} void TARGET_ASM_EXCEPTION_SECTION ()
7711 If defined, a function that switches to the section in which the main
7712 exception table is to be placed (@pxref{Sections}). The default is a
7713 function that switches to a section named @code{.gcc_except_table} on
7714 machines that support named sections via
7715 @code{TARGET_ASM_NAMED_SECTION}, otherwise if @option{-fpic} or
7716 @option{-fPIC} is in effect, the @code{data_section}, otherwise the
7717 @code{readonly_data_section}.
7720 @deftypefn {Target Hook} void TARGET_ASM_EH_FRAME_SECTION ()
7721 If defined, a function that switches to the section in which the DWARF 2
7722 frame unwind information to be placed (@pxref{Sections}). The default
7723 is a function that outputs a standard GAS section directive, if
7724 @code{EH_FRAME_SECTION_NAME} is defined, or else a data section
7725 directive followed by a synthetic label.
7728 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
7729 Contains the value true if the target should add a zero word onto the
7730 end of a Dwarf-2 frame info section when used for exception handling.
7731 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
7735 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
7736 Given a register, this hook should return a parallel of registers to
7737 represent where to find the register pieces. Define this hook if the
7738 register and its mode are represented in Dwarf in non-contiguous
7739 locations, or if the register should be represented in more than one
7740 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
7741 If not defined, the default is to return @code{NULL_RTX}.
7744 @node Alignment Output
7745 @subsection Assembler Commands for Alignment
7747 @c prevent bad page break with this line
7748 This describes commands for alignment.
7750 @defmac JUMP_ALIGN (@var{label})
7751 The alignment (log base 2) to put in front of @var{label}, which is
7752 a common destination of jumps and has no fallthru incoming edge.
7754 This macro need not be defined if you don't want any special alignment
7755 to be done at such a time. Most machine descriptions do not currently
7758 Unless it's necessary to inspect the @var{label} parameter, it is better
7759 to set the variable @var{align_jumps} in the target's
7760 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7761 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
7764 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
7765 The alignment (log base 2) to put in front of @var{label}, which follows
7768 This macro need not be defined if you don't want any special alignment
7769 to be done at such a time. Most machine descriptions do not currently
7773 @defmac LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7774 The maximum number of bytes to skip when applying
7775 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
7776 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7779 @defmac LOOP_ALIGN (@var{label})
7780 The alignment (log base 2) to put in front of @var{label}, which follows
7781 a @code{NOTE_INSN_LOOP_BEG} note.
7783 This macro need not be defined if you don't want any special alignment
7784 to be done at such a time. Most machine descriptions do not currently
7787 Unless it's necessary to inspect the @var{label} parameter, it is better
7788 to set the variable @code{align_loops} in the target's
7789 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7790 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
7793 @defmac LOOP_ALIGN_MAX_SKIP
7794 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
7795 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7798 @defmac LABEL_ALIGN (@var{label})
7799 The alignment (log base 2) to put in front of @var{label}.
7800 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
7801 the maximum of the specified values is used.
7803 Unless it's necessary to inspect the @var{label} parameter, it is better
7804 to set the variable @code{align_labels} in the target's
7805 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7806 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
7809 @defmac LABEL_ALIGN_MAX_SKIP
7810 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
7811 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7814 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
7815 A C statement to output to the stdio stream @var{stream} an assembler
7816 instruction to advance the location counter by @var{nbytes} bytes.
7817 Those bytes should be zero when loaded. @var{nbytes} will be a C
7818 expression of type @code{int}.
7821 @defmac ASM_NO_SKIP_IN_TEXT
7822 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
7823 text section because it fails to put zeros in the bytes that are skipped.
7824 This is true on many Unix systems, where the pseudo--op to skip bytes
7825 produces no-op instructions rather than zeros when used in the text
7829 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
7830 A C statement to output to the stdio stream @var{stream} an assembler
7831 command to advance the location counter to a multiple of 2 to the
7832 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
7835 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
7836 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
7837 for padding, if necessary.
7840 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
7841 A C statement to output to the stdio stream @var{stream} an assembler
7842 command to advance the location counter to a multiple of 2 to the
7843 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
7844 satisfy the alignment request. @var{power} and @var{max_skip} will be
7845 a C expression of type @code{int}.
7849 @node Debugging Info
7850 @section Controlling Debugging Information Format
7852 @c prevent bad page break with this line
7853 This describes how to specify debugging information.
7856 * All Debuggers:: Macros that affect all debugging formats uniformly.
7857 * DBX Options:: Macros enabling specific options in DBX format.
7858 * DBX Hooks:: Hook macros for varying DBX format.
7859 * File Names and DBX:: Macros controlling output of file names in DBX format.
7860 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
7861 * VMS Debug:: Macros for VMS debug format.
7865 @subsection Macros Affecting All Debugging Formats
7867 @c prevent bad page break with this line
7868 These macros affect all debugging formats.
7870 @defmac DBX_REGISTER_NUMBER (@var{regno})
7871 A C expression that returns the DBX register number for the compiler
7872 register number @var{regno}. In the default macro provided, the value
7873 of this expression will be @var{regno} itself. But sometimes there are
7874 some registers that the compiler knows about and DBX does not, or vice
7875 versa. In such cases, some register may need to have one number in the
7876 compiler and another for DBX@.
7878 If two registers have consecutive numbers inside GCC, and they can be
7879 used as a pair to hold a multiword value, then they @emph{must} have
7880 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
7881 Otherwise, debuggers will be unable to access such a pair, because they
7882 expect register pairs to be consecutive in their own numbering scheme.
7884 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
7885 does not preserve register pairs, then what you must do instead is
7886 redefine the actual register numbering scheme.
7889 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
7890 A C expression that returns the integer offset value for an automatic
7891 variable having address @var{x} (an RTL expression). The default
7892 computation assumes that @var{x} is based on the frame-pointer and
7893 gives the offset from the frame-pointer. This is required for targets
7894 that produce debugging output for DBX or COFF-style debugging output
7895 for SDB and allow the frame-pointer to be eliminated when the
7896 @option{-g} options is used.
7899 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
7900 A C expression that returns the integer offset value for an argument
7901 having address @var{x} (an RTL expression). The nominal offset is
7905 @defmac PREFERRED_DEBUGGING_TYPE
7906 A C expression that returns the type of debugging output GCC should
7907 produce when the user specifies just @option{-g}. Define
7908 this if you have arranged for GCC to support more than one format of
7909 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7910 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
7911 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
7913 When the user specifies @option{-ggdb}, GCC normally also uses the
7914 value of this macro to select the debugging output format, but with two
7915 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
7916 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7917 defined, GCC uses @code{DBX_DEBUG}.
7919 The value of this macro only affects the default debugging output; the
7920 user can always get a specific type of output by using @option{-gstabs},
7921 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
7925 @subsection Specific Options for DBX Output
7927 @c prevent bad page break with this line
7928 These are specific options for DBX output.
7930 @defmac DBX_DEBUGGING_INFO
7931 Define this macro if GCC should produce debugging output for DBX
7932 in response to the @option{-g} option.
7935 @defmac XCOFF_DEBUGGING_INFO
7936 Define this macro if GCC should produce XCOFF format debugging output
7937 in response to the @option{-g} option. This is a variant of DBX format.
7940 @defmac DEFAULT_GDB_EXTENSIONS
7941 Define this macro to control whether GCC should by default generate
7942 GDB's extended version of DBX debugging information (assuming DBX-format
7943 debugging information is enabled at all). If you don't define the
7944 macro, the default is 1: always generate the extended information
7945 if there is any occasion to.
7948 @defmac DEBUG_SYMS_TEXT
7949 Define this macro if all @code{.stabs} commands should be output while
7950 in the text section.
7953 @defmac ASM_STABS_OP
7954 A C string constant, including spacing, naming the assembler pseudo op to
7955 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
7956 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
7957 applies only to DBX debugging information format.
7960 @defmac ASM_STABD_OP
7961 A C string constant, including spacing, naming the assembler pseudo op to
7962 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
7963 value is the current location. If you don't define this macro,
7964 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
7968 @defmac ASM_STABN_OP
7969 A C string constant, including spacing, naming the assembler pseudo op to
7970 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
7971 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
7972 macro applies only to DBX debugging information format.
7975 @defmac DBX_NO_XREFS
7976 Define this macro if DBX on your system does not support the construct
7977 @samp{xs@var{tagname}}. On some systems, this construct is used to
7978 describe a forward reference to a structure named @var{tagname}.
7979 On other systems, this construct is not supported at all.
7982 @defmac DBX_CONTIN_LENGTH
7983 A symbol name in DBX-format debugging information is normally
7984 continued (split into two separate @code{.stabs} directives) when it
7985 exceeds a certain length (by default, 80 characters). On some
7986 operating systems, DBX requires this splitting; on others, splitting
7987 must not be done. You can inhibit splitting by defining this macro
7988 with the value zero. You can override the default splitting-length by
7989 defining this macro as an expression for the length you desire.
7992 @defmac DBX_CONTIN_CHAR
7993 Normally continuation is indicated by adding a @samp{\} character to
7994 the end of a @code{.stabs} string when a continuation follows. To use
7995 a different character instead, define this macro as a character
7996 constant for the character you want to use. Do not define this macro
7997 if backslash is correct for your system.
8000 @defmac DBX_STATIC_STAB_DATA_SECTION
8001 Define this macro if it is necessary to go to the data section before
8002 outputting the @samp{.stabs} pseudo-op for a non-global static
8006 @defmac DBX_TYPE_DECL_STABS_CODE
8007 The value to use in the ``code'' field of the @code{.stabs} directive
8008 for a typedef. The default is @code{N_LSYM}.
8011 @defmac DBX_STATIC_CONST_VAR_CODE
8012 The value to use in the ``code'' field of the @code{.stabs} directive
8013 for a static variable located in the text section. DBX format does not
8014 provide any ``right'' way to do this. The default is @code{N_FUN}.
8017 @defmac DBX_REGPARM_STABS_CODE
8018 The value to use in the ``code'' field of the @code{.stabs} directive
8019 for a parameter passed in registers. DBX format does not provide any
8020 ``right'' way to do this. The default is @code{N_RSYM}.
8023 @defmac DBX_REGPARM_STABS_LETTER
8024 The letter to use in DBX symbol data to identify a symbol as a parameter
8025 passed in registers. DBX format does not customarily provide any way to
8026 do this. The default is @code{'P'}.
8029 @defmac DBX_FUNCTION_FIRST
8030 Define this macro if the DBX information for a function and its
8031 arguments should precede the assembler code for the function. Normally,
8032 in DBX format, the debugging information entirely follows the assembler
8036 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
8037 Define this macro, with value 1, if the value of a symbol describing
8038 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
8039 relative to the start of the enclosing function. Normally, GCC uses
8040 an absolute address.
8043 @defmac DBX_LINES_FUNCTION_RELATIVE
8044 Define this macro, with value 1, if the value of a symbol indicating
8045 the current line number (@code{N_SLINE}) should be relative to the
8046 start of the enclosing function. Normally, GCC uses an absolute address.
8049 @defmac DBX_USE_BINCL
8050 Define this macro if GCC should generate @code{N_BINCL} and
8051 @code{N_EINCL} stabs for included header files, as on Sun systems. This
8052 macro also directs GCC to output a type number as a pair of a file
8053 number and a type number within the file. Normally, GCC does not
8054 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
8055 number for a type number.
8059 @subsection Open-Ended Hooks for DBX Format
8061 @c prevent bad page break with this line
8062 These are hooks for DBX format.
8064 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
8065 Define this macro to say how to output to @var{stream} the debugging
8066 information for the start of a scope level for variable names. The
8067 argument @var{name} is the name of an assembler symbol (for use with
8068 @code{assemble_name}) whose value is the address where the scope begins.
8071 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
8072 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
8075 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
8076 Define this macro if the target machine requires special handling to
8077 output an @code{N_FUN} entry for the function @var{decl}.
8080 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
8081 A C statement to output DBX debugging information before code for line
8082 number @var{line} of the current source file to the stdio stream
8083 @var{stream}. @var{counter} is the number of time the macro was
8084 invoked, including the current invocation; it is intended to generate
8085 unique labels in the assembly output.
8087 This macro should not be defined if the default output is correct, or
8088 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
8091 @defmac NO_DBX_FUNCTION_END
8092 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
8093 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
8094 On those machines, define this macro to turn this feature off without
8095 disturbing the rest of the gdb extensions.
8098 @defmac NO_DBX_BNSYM_ENSYM
8099 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
8100 extension construct. On those machines, define this macro to turn this
8101 feature off without disturbing the rest of the gdb extensions.
8104 @node File Names and DBX
8105 @subsection File Names in DBX Format
8107 @c prevent bad page break with this line
8108 This describes file names in DBX format.
8110 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
8111 A C statement to output DBX debugging information to the stdio stream
8112 @var{stream}, which indicates that file @var{name} is the main source
8113 file---the file specified as the input file for compilation.
8114 This macro is called only once, at the beginning of compilation.
8116 This macro need not be defined if the standard form of output
8117 for DBX debugging information is appropriate.
8119 It may be necessary to refer to a label equal to the beginning of the
8120 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
8121 to do so. If you do this, you must also set the variable
8122 @var{used_ltext_label_name} to @code{true}.
8125 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
8126 Define this macro, with value 1, if GCC should not emit an indication
8127 of the current directory for compilation and current source language at
8128 the beginning of the file.
8131 @defmac NO_DBX_GCC_MARKER
8132 Define this macro, with value 1, if GCC should not emit an indication
8133 that this object file was compiled by GCC@. The default is to emit
8134 an @code{N_OPT} stab at the beginning of every source file, with
8135 @samp{gcc2_compiled.} for the string and value 0.
8138 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
8139 A C statement to output DBX debugging information at the end of
8140 compilation of the main source file @var{name}. Output should be
8141 written to the stdio stream @var{stream}.
8143 If you don't define this macro, nothing special is output at the end
8144 of compilation, which is correct for most machines.
8147 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
8148 Define this macro @emph{instead of} defining
8149 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
8150 the end of compilation is a @code{N_SO} stab with an empty string,
8151 whose value is the highest absolute text address in the file.
8156 @subsection Macros for SDB and DWARF Output
8158 @c prevent bad page break with this line
8159 Here are macros for SDB and DWARF output.
8161 @defmac SDB_DEBUGGING_INFO
8162 Define this macro if GCC should produce COFF-style debugging output
8163 for SDB in response to the @option{-g} option.
8166 @defmac DWARF2_DEBUGGING_INFO
8167 Define this macro if GCC should produce dwarf version 2 format
8168 debugging output in response to the @option{-g} option.
8170 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (tree @var{function})
8171 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
8172 be emitted for each function. Instead of an integer return the enum
8173 value for the @code{DW_CC_} tag.
8176 To support optional call frame debugging information, you must also
8177 define @code{INCOMING_RETURN_ADDR_RTX} and either set
8178 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
8179 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
8180 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
8183 @defmac DWARF2_FRAME_INFO
8184 Define this macro to a nonzero value if GCC should always output
8185 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
8186 (@pxref{Exception Region Output} is nonzero, GCC will output this
8187 information not matter how you define @code{DWARF2_FRAME_INFO}.
8190 @defmac DWARF2_ASM_LINE_DEBUG_INFO
8191 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8192 line debug info sections. This will result in much more compact line number
8193 tables, and hence is desirable if it works.
8196 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8197 A C statement to issue assembly directives that create a difference
8198 between the two given labels, using an integer of the given size.
8201 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label})
8202 A C statement to issue assembly directives that create a
8203 section-relative reference to the given label, using an integer of the
8207 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8208 A C statement to issue assembly directives that create a self-relative
8209 reference to the given label, using an integer of the given size.
8212 @defmac PUT_SDB_@dots{}
8213 Define these macros to override the assembler syntax for the special
8214 SDB assembler directives. See @file{sdbout.c} for a list of these
8215 macros and their arguments. If the standard syntax is used, you need
8216 not define them yourself.
8220 Some assemblers do not support a semicolon as a delimiter, even between
8221 SDB assembler directives. In that case, define this macro to be the
8222 delimiter to use (usually @samp{\n}). It is not necessary to define
8223 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8227 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
8228 Define this macro to allow references to unknown structure,
8229 union, or enumeration tags to be emitted. Standard COFF does not
8230 allow handling of unknown references, MIPS ECOFF has support for
8234 @defmac SDB_ALLOW_FORWARD_REFERENCES
8235 Define this macro to allow references to structure, union, or
8236 enumeration tags that have not yet been seen to be handled. Some
8237 assemblers choke if forward tags are used, while some require it.
8240 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
8241 A C statement to output SDB debugging information before code for line
8242 number @var{line} of the current source file to the stdio stream
8243 @var{stream}. The default is to emit an @code{.ln} directive.
8248 @subsection Macros for VMS Debug Format
8250 @c prevent bad page break with this line
8251 Here are macros for VMS debug format.
8253 @defmac VMS_DEBUGGING_INFO
8254 Define this macro if GCC should produce debugging output for VMS
8255 in response to the @option{-g} option. The default behavior for VMS
8256 is to generate minimal debug info for a traceback in the absence of
8257 @option{-g} unless explicitly overridden with @option{-g0}. This
8258 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8259 @code{OVERRIDE_OPTIONS}.
8262 @node Floating Point
8263 @section Cross Compilation and Floating Point
8264 @cindex cross compilation and floating point
8265 @cindex floating point and cross compilation
8267 While all modern machines use twos-complement representation for integers,
8268 there are a variety of representations for floating point numbers. This
8269 means that in a cross-compiler the representation of floating point numbers
8270 in the compiled program may be different from that used in the machine
8271 doing the compilation.
8273 Because different representation systems may offer different amounts of
8274 range and precision, all floating point constants must be represented in
8275 the target machine's format. Therefore, the cross compiler cannot
8276 safely use the host machine's floating point arithmetic; it must emulate
8277 the target's arithmetic. To ensure consistency, GCC always uses
8278 emulation to work with floating point values, even when the host and
8279 target floating point formats are identical.
8281 The following macros are provided by @file{real.h} for the compiler to
8282 use. All parts of the compiler which generate or optimize
8283 floating-point calculations must use these macros. They may evaluate
8284 their operands more than once, so operands must not have side effects.
8286 @defmac REAL_VALUE_TYPE
8287 The C data type to be used to hold a floating point value in the target
8288 machine's format. Typically this is a @code{struct} containing an
8289 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8293 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8294 Compares for equality the two values, @var{x} and @var{y}. If the target
8295 floating point format supports negative zeroes and/or NaNs,
8296 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8297 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8300 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8301 Tests whether @var{x} is less than @var{y}.
8304 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8305 Truncates @var{x} to a signed integer, rounding toward zero.
8308 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8309 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8310 @var{x} is negative, returns zero.
8313 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8314 Converts @var{string} into a floating point number in the target machine's
8315 representation for mode @var{mode}. This routine can handle both
8316 decimal and hexadecimal floating point constants, using the syntax
8317 defined by the C language for both.
8320 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8321 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8324 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8325 Determines whether @var{x} represents infinity (positive or negative).
8328 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8329 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8332 @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})
8333 Calculates an arithmetic operation on the two floating point values
8334 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8337 The operation to be performed is specified by @var{code}. Only the
8338 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8339 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8341 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8342 target's floating point format cannot represent infinity, it will call
8343 @code{abort}. Callers should check for this situation first, using
8344 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8347 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8348 Returns the negative of the floating point value @var{x}.
8351 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8352 Returns the absolute value of @var{x}.
8355 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8356 Truncates the floating point value @var{x} to fit in @var{mode}. The
8357 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8358 appropriate bit pattern to be output asa floating constant whose
8359 precision accords with mode @var{mode}.
8362 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8363 Converts a floating point value @var{x} into a double-precision integer
8364 which is then stored into @var{low} and @var{high}. If the value is not
8365 integral, it is truncated.
8368 @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})
8369 Converts a double-precision integer found in @var{low} and @var{high},
8370 into a floating point value which is then stored into @var{x}. The
8371 value is truncated to fit in mode @var{mode}.
8374 @node Mode Switching
8375 @section Mode Switching Instructions
8376 @cindex mode switching
8377 The following macros control mode switching optimizations:
8379 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
8380 Define this macro if the port needs extra instructions inserted for mode
8381 switching in an optimizing compilation.
8383 For an example, the SH4 can perform both single and double precision
8384 floating point operations, but to perform a single precision operation,
8385 the FPSCR PR bit has to be cleared, while for a double precision
8386 operation, this bit has to be set. Changing the PR bit requires a general
8387 purpose register as a scratch register, hence these FPSCR sets have to
8388 be inserted before reload, i.e.@: you can't put this into instruction emitting
8389 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
8391 You can have multiple entities that are mode-switched, and select at run time
8392 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
8393 return nonzero for any @var{entity} that needs mode-switching.
8394 If you define this macro, you also have to define
8395 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
8396 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
8397 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
8401 @defmac NUM_MODES_FOR_MODE_SWITCHING
8402 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
8403 initializer for an array of integers. Each initializer element
8404 N refers to an entity that needs mode switching, and specifies the number
8405 of different modes that might need to be set for this entity.
8406 The position of the initializer in the initializer---starting counting at
8407 zero---determines the integer that is used to refer to the mode-switched
8409 In macros that take mode arguments / yield a mode result, modes are
8410 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
8411 switch is needed / supplied.
8414 @defmac MODE_NEEDED (@var{entity}, @var{insn})
8415 @var{entity} is an integer specifying a mode-switched entity. If
8416 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
8417 return an integer value not larger than the corresponding element in
8418 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
8419 be switched into prior to the execution of @var{insn}.
8422 @defmac MODE_AFTER (@var{mode}, @var{insn})
8423 If this macro is defined, it is evaluated for every @var{insn} during
8424 mode switching. It determines the mode that an insn results in (if
8425 different from the incoming mode).
8428 @defmac MODE_ENTRY (@var{entity})
8429 If this macro is defined, it is evaluated for every @var{entity} that needs
8430 mode switching. It should evaluate to an integer, which is a mode that
8431 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
8432 is defined then @code{MODE_EXIT} must be defined.
8435 @defmac MODE_EXIT (@var{entity})
8436 If this macro is defined, it is evaluated for every @var{entity} that needs
8437 mode switching. It should evaluate to an integer, which is a mode that
8438 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
8439 is defined then @code{MODE_ENTRY} must be defined.
8442 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
8443 This macro specifies the order in which modes for @var{entity} are processed.
8444 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
8445 lowest. The value of the macro should be an integer designating a mode
8446 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
8447 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
8448 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
8451 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
8452 Generate one or more insns to set @var{entity} to @var{mode}.
8453 @var{hard_reg_live} is the set of hard registers live at the point where
8454 the insn(s) are to be inserted.
8457 @node Target Attributes
8458 @section Defining target-specific uses of @code{__attribute__}
8459 @cindex target attributes
8460 @cindex machine attributes
8461 @cindex attributes, target-specific
8463 Target-specific attributes may be defined for functions, data and types.
8464 These are described using the following target hooks; they also need to
8465 be documented in @file{extend.texi}.
8467 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
8468 If defined, this target hook points to an array of @samp{struct
8469 attribute_spec} (defined in @file{tree.h}) specifying the machine
8470 specific attributes for this target and some of the restrictions on the
8471 entities to which these attributes are applied and the arguments they
8475 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8476 If defined, this target hook is a function which returns zero if the attributes on
8477 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8478 and two if they are nearly compatible (which causes a warning to be
8479 generated). If this is not defined, machine-specific attributes are
8480 supposed always to be compatible.
8483 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
8484 If defined, this target hook is a function which assigns default attributes to
8485 newly defined @var{type}.
8488 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8489 Define this target hook if the merging of type attributes needs special
8490 handling. If defined, the result is a list of the combined
8491 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
8492 that @code{comptypes} has already been called and returned 1. This
8493 function may call @code{merge_attributes} to handle machine-independent
8497 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
8498 Define this target hook if the merging of decl attributes needs special
8499 handling. If defined, the result is a list of the combined
8500 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
8501 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
8502 when this is needed are when one attribute overrides another, or when an
8503 attribute is nullified by a subsequent definition. This function may
8504 call @code{merge_attributes} to handle machine-independent merging.
8506 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
8507 If the only target-specific handling you require is @samp{dllimport}
8508 for Microsoft Windows targets, you should define the macro
8509 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
8510 will then define a function called
8511 @code{merge_dllimport_decl_attributes} which can then be defined as
8512 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
8513 add @code{handle_dll_attribute} in the attribute table for your port
8514 to perform initial processing of the @samp{dllimport} and
8515 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
8516 @file{i386/i386.c}, for example.
8519 @defmac TARGET_DECLSPEC
8520 Define this macro to a nonzero value if you want to treat
8521 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
8522 default, this behavior is enabled only for targets that define
8523 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
8524 of @code{__declspec} is via a built-in macro, but you should not rely
8525 on this implementation detail.
8528 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
8529 Define this target hook if you want to be able to add attributes to a decl
8530 when it is being created. This is normally useful for back ends which
8531 wish to implement a pragma by using the attributes which correspond to
8532 the pragma's effect. The @var{node} argument is the decl which is being
8533 created. The @var{attr_ptr} argument is a pointer to the attribute list
8534 for this decl. The list itself should not be modified, since it may be
8535 shared with other decls, but attributes may be chained on the head of
8536 the list and @code{*@var{attr_ptr}} modified to point to the new
8537 attributes, or a copy of the list may be made if further changes are
8541 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
8543 This target hook returns @code{true} if it is ok to inline @var{fndecl}
8544 into the current function, despite its having target-specific
8545 attributes, @code{false} otherwise. By default, if a function has a
8546 target specific attribute attached to it, it will not be inlined.
8549 @node MIPS Coprocessors
8550 @section Defining coprocessor specifics for MIPS targets.
8551 @cindex MIPS coprocessor-definition macros
8553 The MIPS specification allows MIPS implementations to have as many as 4
8554 coprocessors, each with as many as 32 private registers. GCC supports
8555 accessing these registers and transferring values between the registers
8556 and memory using asm-ized variables. For example:
8559 register unsigned int cp0count asm ("c0r1");
8565 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
8566 names may be added as described below, or the default names may be
8567 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
8569 Coprocessor registers are assumed to be epilogue-used; sets to them will
8570 be preserved even if it does not appear that the register is used again
8571 later in the function.
8573 Another note: according to the MIPS spec, coprocessor 1 (if present) is
8574 the FPU@. One accesses COP1 registers through standard mips
8575 floating-point support; they are not included in this mechanism.
8577 There is one macro used in defining the MIPS coprocessor interface which
8578 you may want to override in subtargets; it is described below.
8580 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
8581 A comma-separated list (with leading comma) of pairs describing the
8582 alternate names of coprocessor registers. The format of each entry should be
8584 @{ @var{alternatename}, @var{register_number}@}
8590 @section Parameters for Precompiled Header Validity Checking
8591 @cindex parameters, precompiled headers
8593 @deftypefn {Target Hook} void * TARGET_GET_PCH_VALIDITY (size_t * @var{sz})
8594 Define this hook if your target needs to check a different collection
8595 of flags than the default, which is every flag defined by
8596 @code{TARGET_SWITCHES} and @code{TARGET_OPTIONS}. It should return
8597 some data which will be saved in the PCH file and presented to
8598 @code{TARGET_PCH_VALID_P} later; it should set @code{SZ} to the size
8602 @deftypefn {Target Hook} const char * TARGET_PCH_VALID_P (const void * @var{data}, size_t @var{sz})
8603 Define this hook if your target needs to check a different collection of
8604 flags than the default, which is every flag defined by @code{TARGET_SWITCHES}
8605 and @code{TARGET_OPTIONS}. It is given data which came from
8606 @code{TARGET_GET_PCH_VALIDITY} (in this version of this compiler, so there
8607 is no need for extensive validity checking). It returns @code{NULL} if
8608 it is safe to load a PCH file with this data, or a suitable error message
8609 if not. The error message will be presented to the user, so it should
8614 @section C++ ABI parameters
8615 @cindex parameters, c++ abi
8617 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
8618 Define this hook to override the integer type used for guard variables.
8619 These are used to implement one-time construction of static objects. The
8620 default is long_long_integer_type_node.
8623 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
8624 This hook determines how guard variables are used. It should return
8625 @code{false} (the default) if first byte should be used. A return value of
8626 @code{true} indicates the least significant bit should be used.
8629 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
8630 This hook returns the size of the cookie to use when allocating an array
8631 whose elements have the indicated @var{type}. Assumes that it is already
8632 known that a cookie is needed. The default is
8633 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
8634 IA64/Generic C++ ABI@.
8637 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
8638 This hook should return @code{true} if the element size should be stored in
8639 array cookies. The default is to return @code{false}.
8642 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
8643 If defined by a backend this hook allows the decision made to export
8644 class @var{type} to be overruled. Upon entry @var{import_export}
8645 will contain 1 if the class is going to be exported, @minus{}1 if it is going
8646 to be imported and 0 otherwise. This function should return the
8647 modified value and perform any other actions necessary to support the
8648 backend's targeted operating system.
8651 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
8652 This hook should return @code{true} if constructors and destructors return
8653 the address of the object created/destroyed. The default is to return
8657 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
8658 This hook returns true if the key method for a class (i.e., the method
8659 which, if defined in the current translation unit, causes the virtual
8660 table to be emitted) may be an inline function. Under the standard
8661 Itanium C++ ABI the key method may be an inline function so long as
8662 the function is not declared inline in the class definition. Under
8663 some variants of the ABI, an inline function can never be the key
8664 method. The default is to return @code{true}.
8667 @deftypefn {Target Hook} bool TARGET_CXX_EXPORT_CLASS_DATA (void)
8668 If this hook returns false (the default), then virtual tables and RTTI
8669 data structures will have the ELF visibility of their containing
8670 class. If this hook returns true, then these data structures will
8671 have ELF ``default'' visibility, independently of the visibility of
8672 the containing class.
8676 @section Miscellaneous Parameters
8677 @cindex parameters, miscellaneous
8679 @c prevent bad page break with this line
8680 Here are several miscellaneous parameters.
8682 @defmac PREDICATE_CODES
8683 Define this if you have defined special-purpose predicates in the file
8684 @file{@var{machine}.c}. This macro is called within an initializer of an
8685 array of structures. The first field in the structure is the name of a
8686 predicate and the second field is an array of rtl codes. For each
8687 predicate, list all rtl codes that can be in expressions matched by the
8688 predicate. The list should have a trailing comma. Here is an example
8689 of two entries in the list for a typical RISC machine:
8692 #define PREDICATE_CODES \
8693 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
8694 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
8697 Defining this macro does not affect the generated code (however,
8698 incorrect definitions that omit an rtl code that may be matched by the
8699 predicate can cause the compiler to malfunction). Instead, it allows
8700 the table built by @file{genrecog} to be more compact and efficient,
8701 thus speeding up the compiler. The most important predicates to include
8702 in the list specified by this macro are those used in the most insn
8705 For each predicate function named in @code{PREDICATE_CODES}, a
8706 declaration will be generated in @file{insn-codes.h}.
8708 Use of this macro is deprecated; use @code{define_predicate} instead.
8709 @xref{Defining Predicates}.
8712 @defmac SPECIAL_MODE_PREDICATES
8713 Define this if you have special predicates that know special things
8714 about modes. Genrecog will warn about certain forms of
8715 @code{match_operand} without a mode; if the operand predicate is
8716 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
8719 Here is an example from the IA-32 port (@code{ext_register_operand}
8720 specially checks for @code{HImode} or @code{SImode} in preparation
8721 for a byte extraction from @code{%ah} etc.).
8724 #define SPECIAL_MODE_PREDICATES \
8725 "ext_register_operand",
8728 Use of this macro is deprecated; use @code{define_special_predicate}
8729 instead. @xref{Defining Predicates}.
8732 @defmac HAS_LONG_COND_BRANCH
8733 Define this boolean macro to indicate whether or not your architecture
8734 has conditional branches that can span all of memory. It is used in
8735 conjunction with an optimization that partitions hot and cold basic
8736 blocks into separate sections of the executable. If this macro is
8737 set to false, gcc will convert any conditional branches that attempt
8738 to cross between sections into unconditional branches or indirect jumps.
8741 @defmac HAS_LONG_UNCOND_BRANCH
8742 Define this boolean macro to indicate whether or not your architecture
8743 has unconditional branches that can span all of memory. It is used in
8744 conjunction with an optimization that partitions hot and cold basic
8745 blocks into separate sections of the executable. If this macro is
8746 set to false, gcc will convert any unconditional branches that attempt
8747 to cross between sections into indirect jumps.
8750 @defmac CASE_VECTOR_MODE
8751 An alias for a machine mode name. This is the machine mode that
8752 elements of a jump-table should have.
8755 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
8756 Optional: return the preferred mode for an @code{addr_diff_vec}
8757 when the minimum and maximum offset are known. If you define this,
8758 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
8759 To make this work, you also have to define @code{INSN_ALIGN} and
8760 make the alignment for @code{addr_diff_vec} explicit.
8761 The @var{body} argument is provided so that the offset_unsigned and scale
8762 flags can be updated.
8765 @defmac CASE_VECTOR_PC_RELATIVE
8766 Define this macro to be a C expression to indicate when jump-tables
8767 should contain relative addresses. You need not define this macro if
8768 jump-tables never contain relative addresses, or jump-tables should
8769 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
8773 @defmac CASE_VALUES_THRESHOLD
8774 Define this to be the smallest number of different values for which it
8775 is best to use a jump-table instead of a tree of conditional branches.
8776 The default is four for machines with a @code{casesi} instruction and
8777 five otherwise. This is best for most machines.
8780 @defmac CASE_USE_BIT_TESTS
8781 Define this macro to be a C expression to indicate whether C switch
8782 statements may be implemented by a sequence of bit tests. This is
8783 advantageous on processors that can efficiently implement left shift
8784 of 1 by the number of bits held in a register, but inappropriate on
8785 targets that would require a loop. By default, this macro returns
8786 @code{true} if the target defines an @code{ashlsi3} pattern, and
8787 @code{false} otherwise.
8790 @defmac WORD_REGISTER_OPERATIONS
8791 Define this macro if operations between registers with integral mode
8792 smaller than a word are always performed on the entire register.
8793 Most RISC machines have this property and most CISC machines do not.
8796 @defmac LOAD_EXTEND_OP (@var{mem_mode})
8797 Define this macro to be a C expression indicating when insns that read
8798 memory in @var{mem_mode}, an integral mode narrower than a word, set the
8799 bits outside of @var{mem_mode} to be either the sign-extension or the
8800 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
8801 of @var{mem_mode} for which the
8802 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
8803 @code{UNKNOWN} for other modes.
8805 This macro is not called with @var{mem_mode} non-integral or with a width
8806 greater than or equal to @code{BITS_PER_WORD}, so you may return any
8807 value in this case. Do not define this macro if it would always return
8808 @code{UNKNOWN}. On machines where this macro is defined, you will normally
8809 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
8811 You may return a non-@code{UNKNOWN} value even if for some hard registers
8812 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
8813 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
8814 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
8815 integral mode larger than this but not larger than @code{word_mode}.
8817 You must return @code{UNKNOWN} if for some hard registers that allow this
8818 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
8819 @code{word_mode}, but that they can change to another integral mode that
8820 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
8823 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
8824 Define this macro if loading short immediate values into registers sign
8828 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
8829 Define this macro if the same instructions that convert a floating
8830 point number to a signed fixed point number also convert validly to an
8835 The maximum number of bytes that a single instruction can move quickly
8836 between memory and registers or between two memory locations.
8839 @defmac MAX_MOVE_MAX
8840 The maximum number of bytes that a single instruction can move quickly
8841 between memory and registers or between two memory locations. If this
8842 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
8843 constant value that is the largest value that @code{MOVE_MAX} can have
8847 @defmac SHIFT_COUNT_TRUNCATED
8848 A C expression that is nonzero if on this machine the number of bits
8849 actually used for the count of a shift operation is equal to the number
8850 of bits needed to represent the size of the object being shifted. When
8851 this macro is nonzero, the compiler will assume that it is safe to omit
8852 a sign-extend, zero-extend, and certain bitwise `and' instructions that
8853 truncates the count of a shift operation. On machines that have
8854 instructions that act on bit-fields at variable positions, which may
8855 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
8856 also enables deletion of truncations of the values that serve as
8857 arguments to bit-field instructions.
8859 If both types of instructions truncate the count (for shifts) and
8860 position (for bit-field operations), or if no variable-position bit-field
8861 instructions exist, you should define this macro.
8863 However, on some machines, such as the 80386 and the 680x0, truncation
8864 only applies to shift operations and not the (real or pretended)
8865 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
8866 such machines. Instead, add patterns to the @file{md} file that include
8867 the implied truncation of the shift instructions.
8869 You need not define this macro if it would always have the value of zero.
8872 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
8873 @deftypefn {Target Hook} int TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode})
8874 This function describes how the standard shift patterns for @var{mode}
8875 deal with shifts by negative amounts or by more than the width of the mode.
8876 @xref{shift patterns}.
8878 On many machines, the shift patterns will apply a mask @var{m} to the
8879 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
8880 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
8881 this is true for mode @var{mode}, the function should return @var{m},
8882 otherwise it should return 0. A return value of 0 indicates that no
8883 particular behavior is guaranteed.
8885 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
8886 @emph{not} apply to general shift rtxes; it applies only to instructions
8887 that are generated by the named shift patterns.
8889 The default implementation of this function returns
8890 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
8891 and 0 otherwise. This definition is always safe, but if
8892 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
8893 nevertheless truncate the shift count, you may get better code
8897 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
8898 A C expression which is nonzero if on this machine it is safe to
8899 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
8900 bits (where @var{outprec} is smaller than @var{inprec}) by merely
8901 operating on it as if it had only @var{outprec} bits.
8903 On many machines, this expression can be 1.
8905 @c rearranged this, removed the phrase "it is reported that". this was
8906 @c to fix an overfull hbox. --mew 10feb93
8907 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
8908 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
8909 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
8910 such cases may improve things.
8913 @defmac STORE_FLAG_VALUE
8914 A C expression describing the value returned by a comparison operator
8915 with an integral mode and stored by a store-flag instruction
8916 (@samp{s@var{cond}}) when the condition is true. This description must
8917 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
8918 comparison operators whose results have a @code{MODE_INT} mode.
8920 A value of 1 or @minus{}1 means that the instruction implementing the
8921 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
8922 and 0 when the comparison is false. Otherwise, the value indicates
8923 which bits of the result are guaranteed to be 1 when the comparison is
8924 true. This value is interpreted in the mode of the comparison
8925 operation, which is given by the mode of the first operand in the
8926 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
8927 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
8930 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
8931 generate code that depends only on the specified bits. It can also
8932 replace comparison operators with equivalent operations if they cause
8933 the required bits to be set, even if the remaining bits are undefined.
8934 For example, on a machine whose comparison operators return an
8935 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
8936 @samp{0x80000000}, saying that just the sign bit is relevant, the
8940 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
8947 (ashift:SI @var{x} (const_int @var{n}))
8951 where @var{n} is the appropriate shift count to move the bit being
8952 tested into the sign bit.
8954 There is no way to describe a machine that always sets the low-order bit
8955 for a true value, but does not guarantee the value of any other bits,
8956 but we do not know of any machine that has such an instruction. If you
8957 are trying to port GCC to such a machine, include an instruction to
8958 perform a logical-and of the result with 1 in the pattern for the
8959 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
8961 Often, a machine will have multiple instructions that obtain a value
8962 from a comparison (or the condition codes). Here are rules to guide the
8963 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
8968 Use the shortest sequence that yields a valid definition for
8969 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
8970 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
8971 comparison operators to do so because there may be opportunities to
8972 combine the normalization with other operations.
8975 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
8976 slightly preferred on machines with expensive jumps and 1 preferred on
8980 As a second choice, choose a value of @samp{0x80000001} if instructions
8981 exist that set both the sign and low-order bits but do not define the
8985 Otherwise, use a value of @samp{0x80000000}.
8988 Many machines can produce both the value chosen for
8989 @code{STORE_FLAG_VALUE} and its negation in the same number of
8990 instructions. On those machines, you should also define a pattern for
8991 those cases, e.g., one matching
8994 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
8997 Some machines can also perform @code{and} or @code{plus} operations on
8998 condition code values with less instructions than the corresponding
8999 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
9000 machines, define the appropriate patterns. Use the names @code{incscc}
9001 and @code{decscc}, respectively, for the patterns which perform
9002 @code{plus} or @code{minus} operations on condition code values. See
9003 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
9004 find such instruction sequences on other machines.
9006 If this macro is not defined, the default value, 1, is used. You need
9007 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
9008 instructions, or if the value generated by these instructions is 1.
9011 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
9012 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
9013 returned when comparison operators with floating-point results are true.
9014 Define this macro on machines that have comparison operations that return
9015 floating-point values. If there are no such operations, do not define
9019 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
9020 A C expression that gives a rtx representing the non-zero true element
9021 for vector comparisons. The returned rtx should be valid for the inner
9022 mode of @var{mode} which is guaranteed to be a vector mode. Define
9023 this macro on machines that have vector comparison operations that
9024 return a vector result. If there are no such operations, do not define
9025 this macro. Typically, this macro is defined as @code{const1_rtx} or
9026 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
9027 the compiler optimizing such vector comparison operations for the
9031 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
9032 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
9033 A C expression that evaluates to true if the architecture defines a value
9034 for @code{clz} or @code{ctz} with a zero operand. If so, @var{value}
9035 should be set to this value. If this macro is not defined, the value of
9036 @code{clz} or @code{ctz} is assumed to be undefined.
9038 This macro must be defined if the target's expansion for @code{ffs}
9039 relies on a particular value to get correct results. Otherwise it
9040 is not necessary, though it may be used to optimize some corner cases.
9042 Note that regardless of this macro the ``definedness'' of @code{clz}
9043 and @code{ctz} at zero do @emph{not} extend to the builtin functions
9044 visible to the user. Thus one may be free to adjust the value at will
9045 to match the target expansion of these operations without fear of
9050 An alias for the machine mode for pointers. On most machines, define
9051 this to be the integer mode corresponding to the width of a hardware
9052 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
9053 On some machines you must define this to be one of the partial integer
9054 modes, such as @code{PSImode}.
9056 The width of @code{Pmode} must be at least as large as the value of
9057 @code{POINTER_SIZE}. If it is not equal, you must define the macro
9058 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
9062 @defmac FUNCTION_MODE
9063 An alias for the machine mode used for memory references to functions
9064 being called, in @code{call} RTL expressions. On most machines this
9065 should be @code{QImode}.
9068 @defmac STDC_0_IN_SYSTEM_HEADERS
9069 In normal operation, the preprocessor expands @code{__STDC__} to the
9070 constant 1, to signify that GCC conforms to ISO Standard C@. On some
9071 hosts, like Solaris, the system compiler uses a different convention,
9072 where @code{__STDC__} is normally 0, but is 1 if the user specifies
9073 strict conformance to the C Standard.
9075 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
9076 convention when processing system header files, but when processing user
9077 files @code{__STDC__} will always expand to 1.
9080 @defmac NO_IMPLICIT_EXTERN_C
9081 Define this macro if the system header files support C++ as well as C@.
9082 This macro inhibits the usual method of using system header files in
9083 C++, which is to pretend that the file's contents are enclosed in
9084 @samp{extern "C" @{@dots{}@}}.
9089 @defmac REGISTER_TARGET_PRAGMAS ()
9090 Define this macro if you want to implement any target-specific pragmas.
9091 If defined, it is a C expression which makes a series of calls to
9092 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
9093 for each pragma. The macro may also do any
9094 setup required for the pragmas.
9096 The primary reason to define this macro is to provide compatibility with
9097 other compilers for the same target. In general, we discourage
9098 definition of target-specific pragmas for GCC@.
9100 If the pragma can be implemented by attributes then you should consider
9101 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
9103 Preprocessor macros that appear on pragma lines are not expanded. All
9104 @samp{#pragma} directives that do not match any registered pragma are
9105 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
9108 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9109 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9111 Each call to @code{c_register_pragma} or
9112 @code{c_register_pragma_with_expansion} establishes one pragma. The
9113 @var{callback} routine will be called when the preprocessor encounters a
9117 #pragma [@var{space}] @var{name} @dots{}
9120 @var{space} is the case-sensitive namespace of the pragma, or
9121 @code{NULL} to put the pragma in the global namespace. The callback
9122 routine receives @var{pfile} as its first argument, which can be passed
9123 on to cpplib's functions if necessary. You can lex tokens after the
9124 @var{name} by calling @code{c_lex}. Tokens that are not read by the
9125 callback will be silently ignored. The end of the line is indicated by
9126 a token of type @code{CPP_EOF}. Macro expansion occurs on the
9127 arguments of pragmas registered with
9128 @code{c_register_pragma_with_expansion} but not on the arguments of
9129 pragmas registered with @code{c_register_pragma}.
9131 For an example use of this routine, see @file{c4x.h} and the callback
9132 routines defined in @file{c4x-c.c}.
9134 Note that the use of @code{c_lex} is specific to the C and C++
9135 compilers. It will not work in the Java or Fortran compilers, or any
9136 other language compilers for that matter. Thus if @code{c_lex} is going
9137 to be called from target-specific code, it must only be done so when
9138 building the C and C++ compilers. This can be done by defining the
9139 variables @code{c_target_objs} and @code{cxx_target_objs} in the
9140 target entry in the @file{config.gcc} file. These variables should name
9141 the target-specific, language-specific object file which contains the
9142 code that uses @code{c_lex}. Note it will also be necessary to add a
9143 rule to the makefile fragment pointed to by @code{tmake_file} that shows
9144 how to build this object file.
9149 @defmac HANDLE_SYSV_PRAGMA
9150 Define this macro (to a value of 1) if you want the System V style
9151 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
9152 [=<value>]} to be supported by gcc.
9154 The pack pragma specifies the maximum alignment (in bytes) of fields
9155 within a structure, in much the same way as the @samp{__aligned__} and
9156 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
9157 the behavior to the default.
9159 A subtlety for Microsoft Visual C/C++ style bit-field packing
9160 (e.g.@: -mms-bitfields) for targets that support it:
9161 When a bit-field is inserted into a packed record, the whole size
9162 of the underlying type is used by one or more same-size adjacent
9163 bit-fields (that is, if its long:3, 32 bits is used in the record,
9164 and any additional adjacent long bit-fields are packed into the same
9165 chunk of 32 bits. However, if the size changes, a new field of that
9168 If both MS bit-fields and @samp{__attribute__((packed))} are used,
9169 the latter will take precedence. If @samp{__attribute__((packed))} is
9170 used on a single field when MS bit-fields are in use, it will take
9171 precedence for that field, but the alignment of the rest of the structure
9172 may affect its placement.
9174 The weak pragma only works if @code{SUPPORTS_WEAK} and
9175 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
9176 of specifically named weak labels, optionally with a value.
9181 @defmac HANDLE_PRAGMA_PACK_PUSH_POP
9182 Define this macro (to a value of 1) if you want to support the Win32
9183 style pragmas @samp{#pragma pack(push[,@var{n}])} and @samp{#pragma
9184 pack(pop)}. The @samp{pack(push,[@var{n}])} pragma specifies the maximum
9185 alignment (in bytes) of fields within a structure, in much the same way as
9186 the @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
9187 pack value of zero resets the behavior to the default. Successive
9188 invocations of this pragma cause the previous values to be stacked, so
9189 that invocations of @samp{#pragma pack(pop)} will return to the previous
9193 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
9194 Define this macro, as well as
9195 @code{HANDLE_SYSV_PRAGMA}, if macros should be expanded in the
9196 arguments of @samp{#pragma pack}.
9199 @defmac TARGET_DEFAULT_PACK_STRUCT
9200 If your target requires a structure packing default other than 0 (meaning
9201 the machine default), define this macro the the necessary value (in bytes).
9202 This must be a value that would also valid to be used with
9203 @samp{#pragma pack()} (that is, a small power of two).
9206 @defmac DOLLARS_IN_IDENTIFIERS
9207 Define this macro to control use of the character @samp{$} in
9208 identifier names for the C family of languages. 0 means @samp{$} is
9209 not allowed by default; 1 means it is allowed. 1 is the default;
9210 there is no need to define this macro in that case.
9213 @defmac NO_DOLLAR_IN_LABEL
9214 Define this macro if the assembler does not accept the character
9215 @samp{$} in label names. By default constructors and destructors in
9216 G++ have @samp{$} in the identifiers. If this macro is defined,
9217 @samp{.} is used instead.
9220 @defmac NO_DOT_IN_LABEL
9221 Define this macro if the assembler does not accept the character
9222 @samp{.} in label names. By default constructors and destructors in G++
9223 have names that use @samp{.}. If this macro is defined, these names
9224 are rewritten to avoid @samp{.}.
9227 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
9228 Define this macro as a C expression that is nonzero if it is safe for the
9229 delay slot scheduler to place instructions in the delay slot of @var{insn},
9230 even if they appear to use a resource set or clobbered in @var{insn}.
9231 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
9232 every @code{call_insn} has this behavior. On machines where some @code{insn}
9233 or @code{jump_insn} is really a function call and hence has this behavior,
9234 you should define this macro.
9236 You need not define this macro if it would always return zero.
9239 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
9240 Define this macro as a C expression that is nonzero if it is safe for the
9241 delay slot scheduler to place instructions in the delay slot of @var{insn},
9242 even if they appear to set or clobber a resource referenced in @var{insn}.
9243 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
9244 some @code{insn} or @code{jump_insn} is really a function call and its operands
9245 are registers whose use is actually in the subroutine it calls, you should
9246 define this macro. Doing so allows the delay slot scheduler to move
9247 instructions which copy arguments into the argument registers into the delay
9250 You need not define this macro if it would always return zero.
9253 @defmac MULTIPLE_SYMBOL_SPACES
9254 Define this macro as a C expression that is nonzero if, in some cases,
9255 global symbols from one translation unit may not be bound to undefined
9256 symbols in another translation unit without user intervention. For
9257 instance, under Microsoft Windows symbols must be explicitly imported
9258 from shared libraries (DLLs).
9260 You need not define this macro if it would always evaluate to zero.
9263 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{clobbers})
9264 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
9265 any hard regs the port wishes to automatically clobber for all asms.
9266 It should return the result of the last @code{tree_cons} used to add a
9270 @defmac MATH_LIBRARY
9271 Define this macro as a C string constant for the linker argument to link
9272 in the system math library, or @samp{""} if the target does not have a
9273 separate math library.
9275 You need only define this macro if the default of @samp{"-lm"} is wrong.
9278 @defmac LIBRARY_PATH_ENV
9279 Define this macro as a C string constant for the environment variable that
9280 specifies where the linker should look for libraries.
9282 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
9286 @defmac TARGET_HAS_F_SETLKW
9287 Define this macro if the target supports file locking with fcntl / F_SETLKW@.
9288 Note that this functionality is part of POSIX@.
9289 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
9290 to use file locking when exiting a program, which avoids race conditions
9291 if the program has forked.
9294 @defmac MAX_CONDITIONAL_EXECUTE
9296 A C expression for the maximum number of instructions to execute via
9297 conditional execution instructions instead of a branch. A value of
9298 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
9299 1 if it does use cc0.
9302 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
9303 Used if the target needs to perform machine-dependent modifications on the
9304 conditionals used for turning basic blocks into conditionally executed code.
9305 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
9306 contains information about the currently processed blocks. @var{true_expr}
9307 and @var{false_expr} are the tests that are used for converting the
9308 then-block and the else-block, respectively. Set either @var{true_expr} or
9309 @var{false_expr} to a null pointer if the tests cannot be converted.
9312 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
9313 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
9314 if-statements into conditions combined by @code{and} and @code{or} operations.
9315 @var{bb} contains the basic block that contains the test that is currently
9316 being processed and about to be turned into a condition.
9319 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
9320 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
9321 be converted to conditional execution format. @var{ce_info} points to
9322 a data structure, @code{struct ce_if_block}, which contains information
9323 about the currently processed blocks.
9326 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
9327 A C expression to perform any final machine dependent modifications in
9328 converting code to conditional execution. The involved basic blocks
9329 can be found in the @code{struct ce_if_block} structure that is pointed
9330 to by @var{ce_info}.
9333 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
9334 A C expression to cancel any machine dependent modifications in
9335 converting code to conditional execution. The involved basic blocks
9336 can be found in the @code{struct ce_if_block} structure that is pointed
9337 to by @var{ce_info}.
9340 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
9341 A C expression to initialize any extra fields in a @code{struct ce_if_block}
9342 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
9345 @defmac IFCVT_EXTRA_FIELDS
9346 If defined, it should expand to a set of field declarations that will be
9347 added to the @code{struct ce_if_block} structure. These should be initialized
9348 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
9351 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG ()
9352 If non-null, this hook performs a target-specific pass over the
9353 instruction stream. The compiler will run it at all optimization levels,
9354 just before the point at which it normally does delayed-branch scheduling.
9356 The exact purpose of the hook varies from target to target. Some use
9357 it to do transformations that are necessary for correctness, such as
9358 laying out in-function constant pools or avoiding hardware hazards.
9359 Others use it as an opportunity to do some machine-dependent optimizations.
9361 You need not implement the hook if it has nothing to do. The default
9365 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
9366 Define this hook if you have any machine-specific built-in functions
9367 that need to be defined. It should be a function that performs the
9370 Machine specific built-in functions can be useful to expand special machine
9371 instructions that would otherwise not normally be generated because
9372 they have no equivalent in the source language (for example, SIMD vector
9373 instructions or prefetch instructions).
9375 To create a built-in function, call the function
9376 @code{lang_hooks.builtin_function}
9377 which is defined by the language front end. You can use any type nodes set
9378 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
9379 only language front ends that use those two functions will call
9380 @samp{TARGET_INIT_BUILTINS}.
9383 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
9385 Expand a call to a machine specific built-in function that was set up by
9386 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
9387 function call; the result should go to @var{target} if that is
9388 convenient, and have mode @var{mode} if that is convenient.
9389 @var{subtarget} may be used as the target for computing one of
9390 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
9391 ignored. This function should return the result of the call to the
9395 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{exp}, bool @var{ignore})
9397 Expand a call to a machine specific built-in function that was set up by
9398 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
9399 function call; the result is another tree containing a simplified
9400 expression for the call's result. If @var{ignore} is true the
9401 value will be ignored.
9404 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
9406 Take a branch insn in @var{branch1} and another in @var{branch2}.
9407 Return true if redirecting @var{branch1} to the destination of
9408 @var{branch2} is possible.
9410 On some targets, branches may have a limited range. Optimizing the
9411 filling of delay slots can result in branches being redirected, and this
9412 may in turn cause a branch offset to overflow.
9415 @defmac ALLOCATE_INITIAL_VALUE (@var{hard_reg})
9417 When the initial value of a hard register has been copied in a pseudo
9418 register, it is often not necessary to actually allocate another register
9419 to this pseudo register, because the original hard register or a stack slot
9420 it has been saved into can be used. @code{ALLOCATE_INITIAL_VALUE}, if
9421 defined, is called at the start of register allocation once for each
9422 hard register that had its initial value copied by using
9423 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
9424 Possible values are @code{NULL_RTX}, if you don't want
9425 to do any special allocation, a @code{REG} rtx---that would typically be
9426 the hard register itself, if it is known not to be clobbered---or a
9428 If you are returning a @code{MEM}, this is only a hint for the allocator;
9429 it might decide to use another register anyways.
9430 You may use @code{current_function_leaf_function} in the definition of the
9431 macro, functions that use @code{REG_N_SETS}, to determine if the hard
9432 register in question will not be clobbered.
9435 @defmac TARGET_OBJECT_SUFFIX
9436 Define this macro to be a C string representing the suffix for object
9437 files on your target machine. If you do not define this macro, GCC will
9438 use @samp{.o} as the suffix for object files.
9441 @defmac TARGET_EXECUTABLE_SUFFIX
9442 Define this macro to be a C string representing the suffix to be
9443 automatically added to executable files on your target machine. If you
9444 do not define this macro, GCC will use the null string as the suffix for
9448 @defmac COLLECT_EXPORT_LIST
9449 If defined, @code{collect2} will scan the individual object files
9450 specified on its command line and create an export list for the linker.
9451 Define this macro for systems like AIX, where the linker discards
9452 object files that are not referenced from @code{main} and uses export
9456 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
9457 Define this macro to a C expression representing a variant of the
9458 method call @var{mdecl}, if Java Native Interface (JNI) methods
9459 must be invoked differently from other methods on your target.
9460 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
9461 the @code{stdcall} calling convention and this macro is then
9462 defined as this expression:
9465 build_type_attribute_variant (@var{mdecl},
9467 (get_identifier ("stdcall"),
9472 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
9473 This target hook returns @code{true} past the point in which new jump
9474 instructions could be created. On machines that require a register for
9475 every jump such as the SHmedia ISA of SH5, this point would typically be
9476 reload, so this target hook should be defined to a function such as:
9480 cannot_modify_jumps_past_reload_p ()
9482 return (reload_completed || reload_in_progress);
9487 @deftypefn {Target Hook} int TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
9488 This target hook returns a register class for which branch target register
9489 optimizations should be applied. All registers in this class should be
9490 usable interchangeably. After reload, registers in this class will be
9491 re-allocated and loads will be hoisted out of loops and be subjected
9492 to inter-block scheduling.
9495 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
9496 Branch target register optimization will by default exclude callee-saved
9498 that are not already live during the current function; if this target hook
9499 returns true, they will be included. The target code must than make sure
9500 that all target registers in the class returned by
9501 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
9502 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
9503 epilogues have already been generated. Note, even if you only return
9504 true when @var{after_prologue_epilogue_gen} is false, you still are likely
9505 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
9506 to reserve space for caller-saved target registers.
9509 @defmac POWI_MAX_MULTS
9510 If defined, this macro is interpreted as a signed integer C expression
9511 that specifies the maximum number of floating point multiplications
9512 that should be emitted when expanding exponentiation by an integer
9513 constant inline. When this value is defined, exponentiation requiring
9514 more than this number of multiplications is implemented by calling the
9515 system library's @code{pow}, @code{powf} or @code{powl} routines.
9516 The default value places no upper bound on the multiplication count.
9519 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
9520 This target hook should register any extra include files for the
9521 target. The parameter @var{stdinc} indicates if normal include files
9522 are present. The parameter @var{sysroot} is the system root directory.
9523 The parameter @var{iprefix} is the prefix for the gcc directory.
9526 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
9527 This target hook should register any extra include files for the
9528 target before any standard headers. The parameter @var{stdinc}
9529 indicates if normal include files are present. The parameter
9530 @var{sysroot} is the system root directory. The parameter
9531 @var{iprefix} is the prefix for the gcc directory.
9534 @deftypefn Macro void TARGET_OPTF (char *@var{path})
9535 This target hook should register special include paths for the target.
9536 The parameter @var{path} is the include to register. On Darwin
9537 systems, this is used for Framework includes, which have semantics
9538 that are different from @option{-I}.
9541 @deftypefn {Target Hook} bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
9542 This target hook returns @code{true} if it is safe to use a local alias
9543 for a virtual function @var{fndecl} when constructing thunks,
9544 @code{false} otherwise. By default, the hook returns @code{true} for all
9545 functions, if a target supports aliases (i.e.@: defines
9546 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
9549 @defmac TARGET_FORMAT_TYPES
9550 If defined, this macro is the name of a global variable containing
9551 target-specific format checking information for the @option{-Wformat}
9552 option. The default is to have no target-specific format checks.
9555 @defmac TARGET_N_FORMAT_TYPES
9556 If defined, this macro is the number of entries in
9557 @code{TARGET_FORMAT_TYPES}.
9560 @deftypefn {Target Hook} bool TARGET_RELAXED_ORDERING
9561 If set to @code{true}, means that the target's memory model does not
9562 guarantee that loads which do not depend on one another will access
9563 main memory in the order of the instruction stream; if ordering is
9564 important, an explicit memory barrier must be used. This is true of
9565 many recent processors which implement a policy of ``relaxed,''
9566 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
9567 and ia64. The default is @code{false}.
9570 @defmac TARGET_USE_JCR_SECTION
9571 This macro determines whether to use the JCR section to register Java
9572 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
9573 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.