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_1
416 Define this macro if the driver program should find the library
417 @file{libgcc.a}. If you do not define this macro, the driver program will pass
418 the argument @option{-lgcc} to tell the linker to do the search.
421 @defmac LINK_GCC_C_SEQUENCE_SPEC
422 The sequence in which libgcc and libc are specified to the linker.
423 By default this is @code{%G %L %G}.
426 @defmac LINK_COMMAND_SPEC
427 A C string constant giving the complete command line need to execute the
428 linker. When you do this, you will need to update your port each time a
429 change is made to the link command line within @file{gcc.c}. Therefore,
430 define this macro only if you need to completely redefine the command
431 line for invoking the linker and there is no other way to accomplish
432 the effect you need. Overriding this macro may be avoidable by overriding
433 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
436 @defmac LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
437 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
438 directories from linking commands. Do not give it a nonzero value if
439 removing duplicate search directories changes the linker's semantics.
442 @defmac MULTILIB_DEFAULTS
443 Define this macro as a C expression for the initializer of an array of
444 string to tell the driver program which options are defaults for this
445 target and thus do not need to be handled specially when using
446 @code{MULTILIB_OPTIONS}.
448 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
449 the target makefile fragment or if none of the options listed in
450 @code{MULTILIB_OPTIONS} are set by default.
451 @xref{Target Fragment}.
454 @defmac RELATIVE_PREFIX_NOT_LINKDIR
455 Define this macro to tell @command{gcc} that it should only translate
456 a @option{-B} prefix into a @option{-L} linker option if the prefix
457 indicates an absolute file name.
460 @defmac MD_EXEC_PREFIX
461 If defined, this macro is an additional prefix to try after
462 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
463 when the @option{-b} option is used, or the compiler is built as a cross
464 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
465 to the list of directories used to find the assembler in @file{configure.in}.
468 @defmac STANDARD_STARTFILE_PREFIX
469 Define this macro as a C string constant if you wish to override the
470 standard choice of @code{libdir} as the default prefix to
471 try when searching for startup files such as @file{crt0.o}.
472 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
473 is built as a cross compiler.
476 @defmac STANDARD_STARTFILE_PREFIX_1
477 Define this macro as a C string constant if you wish to override the
478 standard choice of @code{/lib} as a prefix to try after the default prefix
479 when searching for startup files such as @file{crt0.o}.
480 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
481 is built as a cross compiler.
484 @defmac STANDARD_STARTFILE_PREFIX_2
485 Define this macro as a C string constant if you wish to override the
486 standard choice of @code{/lib} as yet another prefix to try after the
487 default prefix when searching for startup files such as @file{crt0.o}.
488 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
489 is built as a cross compiler.
492 @defmac MD_STARTFILE_PREFIX
493 If defined, this macro supplies an additional prefix to try after the
494 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
495 @option{-b} option is used, or when the compiler is built as a cross
499 @defmac MD_STARTFILE_PREFIX_1
500 If defined, this macro supplies yet another prefix to try after the
501 standard prefixes. It is not searched when the @option{-b} option is
502 used, or when the compiler is built as a cross compiler.
505 @defmac INIT_ENVIRONMENT
506 Define this macro as a C string constant if you wish to set environment
507 variables for programs called by the driver, such as the assembler and
508 loader. The driver passes the value of this macro to @code{putenv} to
509 initialize the necessary environment variables.
512 @defmac LOCAL_INCLUDE_DIR
513 Define this macro as a C string constant if you wish to override the
514 standard choice of @file{/usr/local/include} as the default prefix to
515 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
516 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
518 Cross compilers do not search either @file{/usr/local/include} or its
522 @defmac MODIFY_TARGET_NAME
523 Define this macro if you wish to define command-line switches that
524 modify the default target name.
526 For each switch, you can include a string to be appended to the first
527 part of the configuration name or a string to be deleted from the
528 configuration name, if present. The definition should be an initializer
529 for an array of structures. Each array element should have three
530 elements: the switch name (a string constant, including the initial
531 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
532 indicate whether the string should be inserted or deleted, and the string
533 to be inserted or deleted (a string constant).
535 For example, on a machine where @samp{64} at the end of the
536 configuration name denotes a 64-bit target and you want the @option{-32}
537 and @option{-64} switches to select between 32- and 64-bit targets, you would
541 #define MODIFY_TARGET_NAME \
542 @{ @{ "-32", DELETE, "64"@}, \
543 @{"-64", ADD, "64"@}@}
547 @defmac SYSTEM_INCLUDE_DIR
548 Define this macro as a C string constant if you wish to specify a
549 system-specific directory to search for header files before the standard
550 directory. @code{SYSTEM_INCLUDE_DIR} comes before
551 @code{STANDARD_INCLUDE_DIR} in the search order.
553 Cross compilers do not use this macro and do not search the directory
557 @defmac STANDARD_INCLUDE_DIR
558 Define this macro as a C string constant if you wish to override the
559 standard choice of @file{/usr/include} as the default prefix to
560 try when searching for header files.
562 Cross compilers ignore this macro and do not search either
563 @file{/usr/include} or its replacement.
566 @defmac STANDARD_INCLUDE_COMPONENT
567 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
568 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
569 If you do not define this macro, no component is used.
572 @defmac INCLUDE_DEFAULTS
573 Define this macro if you wish to override the entire default search path
574 for include files. For a native compiler, the default search path
575 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
576 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
577 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
578 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
579 and specify private search areas for GCC@. The directory
580 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
582 The definition should be an initializer for an array of structures.
583 Each array element should have four elements: the directory name (a
584 string constant), the component name (also a string constant), a flag
585 for C++-only directories,
586 and a flag showing that the includes in the directory don't need to be
587 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
588 the array with a null element.
590 The component name denotes what GNU package the include file is part of,
591 if any, in all uppercase letters. For example, it might be @samp{GCC}
592 or @samp{BINUTILS}. If the package is part of a vendor-supplied
593 operating system, code the component name as @samp{0}.
595 For example, here is the definition used for VAX/VMS:
598 #define INCLUDE_DEFAULTS \
600 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
601 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
602 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
609 Here is the order of prefixes tried for exec files:
613 Any prefixes specified by the user with @option{-B}.
616 The environment variable @code{GCC_EXEC_PREFIX}, if any.
619 The directories specified by the environment variable @code{COMPILER_PATH}.
622 The macro @code{STANDARD_EXEC_PREFIX}.
625 @file{/usr/lib/gcc/}.
628 The macro @code{MD_EXEC_PREFIX}, if any.
631 Here is the order of prefixes tried for startfiles:
635 Any prefixes specified by the user with @option{-B}.
638 The environment variable @code{GCC_EXEC_PREFIX}, if any.
641 The directories specified by the environment variable @code{LIBRARY_PATH}
642 (or port-specific name; native only, cross compilers do not use this).
645 The macro @code{STANDARD_EXEC_PREFIX}.
648 @file{/usr/lib/gcc/}.
651 The macro @code{MD_EXEC_PREFIX}, if any.
654 The macro @code{MD_STARTFILE_PREFIX}, if any.
657 The macro @code{STANDARD_STARTFILE_PREFIX}.
666 @node Run-time Target
667 @section Run-time Target Specification
668 @cindex run-time target specification
669 @cindex predefined macros
670 @cindex target specifications
672 @c prevent bad page break with this line
673 Here are run-time target specifications.
675 @defmac TARGET_CPU_CPP_BUILTINS ()
676 This function-like macro expands to a block of code that defines
677 built-in preprocessor macros and assertions for the target cpu, using
678 the functions @code{builtin_define}, @code{builtin_define_std} and
679 @code{builtin_assert}. When the front end
680 calls this macro it provides a trailing semicolon, and since it has
681 finished command line option processing your code can use those
684 @code{builtin_assert} takes a string in the form you pass to the
685 command-line option @option{-A}, such as @code{cpu=mips}, and creates
686 the assertion. @code{builtin_define} takes a string in the form
687 accepted by option @option{-D} and unconditionally defines the macro.
689 @code{builtin_define_std} takes a string representing the name of an
690 object-like macro. If it doesn't lie in the user's namespace,
691 @code{builtin_define_std} defines it unconditionally. Otherwise, it
692 defines a version with two leading underscores, and another version
693 with two leading and trailing underscores, and defines the original
694 only if an ISO standard was not requested on the command line. For
695 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
696 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
697 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
698 defines only @code{_ABI64}.
700 You can also test for the C dialect being compiled. The variable
701 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
702 or @code{clk_objective_c}. Note that if we are preprocessing
703 assembler, this variable will be @code{clk_c} but the function-like
704 macro @code{preprocessing_asm_p()} will return true, so you might want
705 to check for that first. If you need to check for strict ANSI, the
706 variable @code{flag_iso} can be used. The function-like macro
707 @code{preprocessing_trad_p()} can be used to check for traditional
711 @defmac TARGET_OS_CPP_BUILTINS ()
712 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
713 and is used for the target operating system instead.
716 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
717 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
718 and is used for the target object format. @file{elfos.h} uses this
719 macro to define @code{__ELF__}, so you probably do not need to define
723 @deftypevar {extern int} target_flags
724 This variable is declared in @file{options.h}, which is included before
725 any target-specific headers.
728 @deftypevar {Target Hook} int TARGET_DEFAULT_TARGET_FLAGS
729 This variable specifies the initial value of @code{target_flags}.
730 Its default setting is 0.
733 @cindex optional hardware or system features
734 @cindex features, optional, in system conventions
736 @deftypefn {Target Hook} bool TARGET_HANDLE_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
737 This hook is called whenever the user specifies one of the
738 target-specific options described by the @file{.opt} definition files
739 (@pxref{Options}). It has the opportunity to do some option-specific
740 processing and should return true if the option is valid. The default
741 definition does nothing but return true.
743 @var{code} specifies the @code{OPT_@var{name}} enumeration value
744 associated with the selected option; @var{name} is just a rendering of
745 the option name in which non-alphanumeric characters are replaced by
746 underscores. @var{arg} specifies the string argument and is null if
747 no argument was given. If the option is flagged as a @code{UInteger}
748 (@pxref{Option properties}), @var{value} is the numeric value of the
749 argument. Otherwise @var{value} is 1 if the positive form of the
750 option was used and 0 if the ``no-'' form was.
753 @defmac TARGET_VERSION
754 This macro is a C statement to print on @code{stderr} a string
755 describing the particular machine description choice. Every machine
756 description should define @code{TARGET_VERSION}. For example:
760 #define TARGET_VERSION \
761 fprintf (stderr, " (68k, Motorola syntax)");
763 #define TARGET_VERSION \
764 fprintf (stderr, " (68k, MIT syntax)");
769 @defmac OVERRIDE_OPTIONS
770 Sometimes certain combinations of command options do not make sense on
771 a particular target machine. You can define a macro
772 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
773 defined, is executed once just after all the command options have been
776 Don't use this macro to turn on various extra optimizations for
777 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
780 @defmac OPTIMIZATION_OPTIONS (@var{level}, @var{size})
781 Some machines may desire to change what optimizations are performed for
782 various optimization levels. This macro, if defined, is executed once
783 just after the optimization level is determined and before the remainder
784 of the command options have been parsed. Values set in this macro are
785 used as the default values for the other command line options.
787 @var{level} is the optimization level specified; 2 if @option{-O2} is
788 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
790 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
792 You should not use this macro to change options that are not
793 machine-specific. These should uniformly selected by the same
794 optimization level on all supported machines. Use this macro to enable
795 machine-specific optimizations.
797 @strong{Do not examine @code{write_symbols} in
798 this macro!} The debugging options are not supposed to alter the
802 @defmac CAN_DEBUG_WITHOUT_FP
803 Define this macro if debugging can be performed even without a frame
804 pointer. If this macro is defined, GCC will turn on the
805 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
808 @node Per-Function Data
809 @section Defining data structures for per-function information.
810 @cindex per-function data
811 @cindex data structures
813 If the target needs to store information on a per-function basis, GCC
814 provides a macro and a couple of variables to allow this. Note, just
815 using statics to store the information is a bad idea, since GCC supports
816 nested functions, so you can be halfway through encoding one function
817 when another one comes along.
819 GCC defines a data structure called @code{struct function} which
820 contains all of the data specific to an individual function. This
821 structure contains a field called @code{machine} whose type is
822 @code{struct machine_function *}, which can be used by targets to point
823 to their own specific data.
825 If a target needs per-function specific data it should define the type
826 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
827 This macro should be used to initialize the function pointer
828 @code{init_machine_status}. This pointer is explained below.
830 One typical use of per-function, target specific data is to create an
831 RTX to hold the register containing the function's return address. This
832 RTX can then be used to implement the @code{__builtin_return_address}
833 function, for level 0.
835 Note---earlier implementations of GCC used a single data area to hold
836 all of the per-function information. Thus when processing of a nested
837 function began the old per-function data had to be pushed onto a
838 stack, and when the processing was finished, it had to be popped off the
839 stack. GCC used to provide function pointers called
840 @code{save_machine_status} and @code{restore_machine_status} to handle
841 the saving and restoring of the target specific information. Since the
842 single data area approach is no longer used, these pointers are no
845 @defmac INIT_EXPANDERS
846 Macro called to initialize any target specific information. This macro
847 is called once per function, before generation of any RTL has begun.
848 The intention of this macro is to allow the initialization of the
849 function pointer @code{init_machine_status}.
852 @deftypevar {void (*)(struct function *)} init_machine_status
853 If this function pointer is non-@code{NULL} it will be called once per
854 function, before function compilation starts, in order to allow the
855 target to perform any target specific initialization of the
856 @code{struct function} structure. It is intended that this would be
857 used to initialize the @code{machine} of that structure.
859 @code{struct machine_function} structures are expected to be freed by GC@.
860 Generally, any memory that they reference must be allocated by using
861 @code{ggc_alloc}, including the structure itself.
865 @section Storage Layout
866 @cindex storage layout
868 Note that the definitions of the macros in this table which are sizes or
869 alignments measured in bits do not need to be constant. They can be C
870 expressions that refer to static variables, such as the @code{target_flags}.
871 @xref{Run-time Target}.
873 @defmac BITS_BIG_ENDIAN
874 Define this macro to have the value 1 if the most significant bit in a
875 byte has the lowest number; otherwise define it to have the value zero.
876 This means that bit-field instructions count from the most significant
877 bit. If the machine has no bit-field instructions, then this must still
878 be defined, but it doesn't matter which value it is defined to. This
879 macro need not be a constant.
881 This macro does not affect the way structure fields are packed into
882 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
885 @defmac BYTES_BIG_ENDIAN
886 Define this macro to have the value 1 if the most significant byte in a
887 word has the lowest number. This macro need not be a constant.
890 @defmac WORDS_BIG_ENDIAN
891 Define this macro to have the value 1 if, in a multiword object, the
892 most significant word has the lowest number. This applies to both
893 memory locations and registers; GCC fundamentally assumes that the
894 order of words in memory is the same as the order in registers. This
895 macro need not be a constant.
898 @defmac LIBGCC2_WORDS_BIG_ENDIAN
899 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
900 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
901 used only when compiling @file{libgcc2.c}. Typically the value will be set
902 based on preprocessor defines.
905 @defmac FLOAT_WORDS_BIG_ENDIAN
906 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
907 @code{TFmode} floating point numbers are stored in memory with the word
908 containing the sign bit at the lowest address; otherwise define it to
909 have the value 0. This macro need not be a constant.
911 You need not define this macro if the ordering is the same as for
915 @defmac BITS_PER_UNIT
916 Define this macro to be the number of bits in an addressable storage
917 unit (byte). If you do not define this macro the default is 8.
920 @defmac BITS_PER_WORD
921 Number of bits in a word. If you do not define this macro, the default
922 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
925 @defmac MAX_BITS_PER_WORD
926 Maximum number of bits in a word. If this is undefined, the default is
927 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
928 largest value that @code{BITS_PER_WORD} can have at run-time.
931 @defmac UNITS_PER_WORD
932 Number of storage units in a word; normally the size of a general-purpose
933 register, a power of two from 1 or 8.
936 @defmac MIN_UNITS_PER_WORD
937 Minimum number of units in a word. If this is undefined, the default is
938 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
939 smallest value that @code{UNITS_PER_WORD} can have at run-time.
942 @defmac UNITS_PER_SIMD_WORD
943 Number of units in the vectors that the vectorizer can produce.
944 The default is equal to @code{UNITS_PER_WORD}, because the vectorizer
945 can do some transformations even in absence of specialized @acronym{SIMD}
950 Width of a pointer, in bits. You must specify a value no wider than the
951 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
952 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
953 a value the default is @code{BITS_PER_WORD}.
956 @defmac POINTERS_EXTEND_UNSIGNED
957 A C expression whose value is greater than zero if pointers that need to be
958 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
959 be zero-extended and zero if they are to be sign-extended. If the value
960 is less then zero then there must be an "ptr_extend" instruction that
961 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
963 You need not define this macro if the @code{POINTER_SIZE} is equal
964 to the width of @code{Pmode}.
967 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
968 A macro to update @var{m} and @var{unsignedp} when an object whose type
969 is @var{type} and which has the specified mode and signedness is to be
970 stored in a register. This macro is only called when @var{type} is a
973 On most RISC machines, which only have operations that operate on a full
974 register, define this macro to set @var{m} to @code{word_mode} if
975 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
976 cases, only integer modes should be widened because wider-precision
977 floating-point operations are usually more expensive than their narrower
980 For most machines, the macro definition does not change @var{unsignedp}.
981 However, some machines, have instructions that preferentially handle
982 either signed or unsigned quantities of certain modes. For example, on
983 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
984 sign-extend the result to 64 bits. On such machines, set
985 @var{unsignedp} according to which kind of extension is more efficient.
987 Do not define this macro if it would never modify @var{m}.
990 @defmac PROMOTE_FUNCTION_MODE
991 Like @code{PROMOTE_MODE}, but is applied to outgoing function arguments or
992 function return values, as specified by @code{TARGET_PROMOTE_FUNCTION_ARGS}
993 and @code{TARGET_PROMOTE_FUNCTION_RETURN}, respectively.
995 The default is @code{PROMOTE_MODE}.
998 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_ARGS (tree @var{fntype})
999 This target hook should return @code{true} if the promotion described by
1000 @code{PROMOTE_FUNCTION_MODE} should be done for outgoing function
1004 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_RETURN (tree @var{fntype})
1005 This target hook should return @code{true} if the promotion described by
1006 @code{PROMOTE_FUNCTION_MODE} should be done for the return value of
1009 If this target hook returns @code{true}, @code{FUNCTION_VALUE} must
1010 perform the same promotions done by @code{PROMOTE_FUNCTION_MODE}.
1013 @defmac PARM_BOUNDARY
1014 Normal alignment required for function parameters on the stack, in
1015 bits. All stack parameters receive at least this much alignment
1016 regardless of data type. On most machines, this is the same as the
1020 @defmac STACK_BOUNDARY
1021 Define this macro to the minimum alignment enforced by hardware for the
1022 stack pointer on this machine. The definition is a C expression for the
1023 desired alignment (measured in bits). This value is used as a default
1024 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1025 this should be the same as @code{PARM_BOUNDARY}.
1028 @defmac PREFERRED_STACK_BOUNDARY
1029 Define this macro if you wish to preserve a certain alignment for the
1030 stack pointer, greater than what the hardware enforces. The definition
1031 is a C expression for the desired alignment (measured in bits). This
1032 macro must evaluate to a value equal to or larger than
1033 @code{STACK_BOUNDARY}.
1036 @defmac FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1037 A C expression that evaluates true if @code{PREFERRED_STACK_BOUNDARY} is
1038 not guaranteed by the runtime and we should emit code to align the stack
1039 at the beginning of @code{main}.
1041 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
1042 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
1043 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
1044 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
1045 be momentarily unaligned while pushing arguments.
1048 @defmac FUNCTION_BOUNDARY
1049 Alignment required for a function entry point, in bits.
1052 @defmac BIGGEST_ALIGNMENT
1053 Biggest alignment that any data type can require on this machine, in bits.
1056 @defmac MINIMUM_ATOMIC_ALIGNMENT
1057 If defined, the smallest alignment, in bits, that can be given to an
1058 object that can be referenced in one operation, without disturbing any
1059 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1060 on machines that don't have byte or half-word store operations.
1063 @defmac BIGGEST_FIELD_ALIGNMENT
1064 Biggest alignment that any structure or union field can require on this
1065 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1066 structure and union fields only, unless the field alignment has been set
1067 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1070 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1071 An expression for the alignment of a structure field @var{field} if the
1072 alignment computed in the usual way (including applying of
1073 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1074 alignment) is @var{computed}. It overrides alignment only if the
1075 field alignment has not been set by the
1076 @code{__attribute__ ((aligned (@var{n})))} construct.
1079 @defmac MAX_OFILE_ALIGNMENT
1080 Biggest alignment supported by the object file format of this machine.
1081 Use this macro to limit the alignment which can be specified using the
1082 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1083 the default value is @code{BIGGEST_ALIGNMENT}.
1086 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1087 If defined, a C expression to compute the alignment for a variable in
1088 the static store. @var{type} is the data type, and @var{basic-align} is
1089 the alignment that the object would ordinarily have. The value of this
1090 macro is used instead of that alignment to align the object.
1092 If this macro is not defined, then @var{basic-align} is used.
1095 One use of this macro is to increase alignment of medium-size data to
1096 make it all fit in fewer cache lines. Another is to cause character
1097 arrays to be word-aligned so that @code{strcpy} calls that copy
1098 constants to character arrays can be done inline.
1101 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1102 If defined, a C expression to compute the alignment given to a constant
1103 that is being placed in memory. @var{constant} is the constant and
1104 @var{basic-align} is the alignment that the object would ordinarily
1105 have. The value of this macro is used instead of that alignment to
1108 If this macro is not defined, then @var{basic-align} is used.
1110 The typical use of this macro is to increase alignment for string
1111 constants to be word aligned so that @code{strcpy} calls that copy
1112 constants can be done inline.
1115 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1116 If defined, a C expression to compute the alignment for a variable in
1117 the local store. @var{type} is the data type, and @var{basic-align} is
1118 the alignment that the object would ordinarily have. The value of this
1119 macro is used instead of that alignment to align the object.
1121 If this macro is not defined, then @var{basic-align} is used.
1123 One use of this macro is to increase alignment of medium-size data to
1124 make it all fit in fewer cache lines.
1127 @defmac EMPTY_FIELD_BOUNDARY
1128 Alignment in bits to be given to a structure bit-field that follows an
1129 empty field such as @code{int : 0;}.
1131 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1134 @defmac STRUCTURE_SIZE_BOUNDARY
1135 Number of bits which any structure or union's size must be a multiple of.
1136 Each structure or union's size is rounded up to a multiple of this.
1138 If you do not define this macro, the default is the same as
1139 @code{BITS_PER_UNIT}.
1142 @defmac STRICT_ALIGNMENT
1143 Define this macro to be the value 1 if instructions will fail to work
1144 if given data not on the nominal alignment. If instructions will merely
1145 go slower in that case, define this macro as 0.
1148 @defmac PCC_BITFIELD_TYPE_MATTERS
1149 Define this if you wish to imitate the way many other C compilers handle
1150 alignment of bit-fields and the structures that contain them.
1152 The behavior is that the type written for a named bit-field (@code{int},
1153 @code{short}, or other integer type) imposes an alignment for the entire
1154 structure, as if the structure really did contain an ordinary field of
1155 that type. In addition, the bit-field is placed within the structure so
1156 that it would fit within such a field, not crossing a boundary for it.
1158 Thus, on most machines, a named bit-field whose type is written as
1159 @code{int} would not cross a four-byte boundary, and would force
1160 four-byte alignment for the whole structure. (The alignment used may
1161 not be four bytes; it is controlled by the other alignment parameters.)
1163 An unnamed bit-field will not affect the alignment of the containing
1166 If the macro is defined, its definition should be a C expression;
1167 a nonzero value for the expression enables this behavior.
1169 Note that if this macro is not defined, or its value is zero, some
1170 bit-fields may cross more than one alignment boundary. The compiler can
1171 support such references if there are @samp{insv}, @samp{extv}, and
1172 @samp{extzv} insns that can directly reference memory.
1174 The other known way of making bit-fields work is to define
1175 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1176 Then every structure can be accessed with fullwords.
1178 Unless the machine has bit-field instructions or you define
1179 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1180 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1182 If your aim is to make GCC use the same conventions for laying out
1183 bit-fields as are used by another compiler, here is how to investigate
1184 what the other compiler does. Compile and run this program:
1203 printf ("Size of foo1 is %d\n",
1204 sizeof (struct foo1));
1205 printf ("Size of foo2 is %d\n",
1206 sizeof (struct foo2));
1211 If this prints 2 and 5, then the compiler's behavior is what you would
1212 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1215 @defmac BITFIELD_NBYTES_LIMITED
1216 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1217 to aligning a bit-field within the structure.
1220 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELDS (void)
1221 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1222 whether unnamed bitfields affect the alignment of the containing
1223 structure. The hook should return true if the structure should inherit
1224 the alignment requirements of an unnamed bitfield's type.
1227 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1228 Return 1 if a structure or array containing @var{field} should be accessed using
1231 If @var{field} is the only field in the structure, @var{mode} is its
1232 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1233 case where structures of one field would require the structure's mode to
1234 retain the field's mode.
1236 Normally, this is not needed. See the file @file{c4x.h} for an example
1237 of how to use this macro to prevent a structure having a floating point
1238 field from being accessed in an integer mode.
1241 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1242 Define this macro as an expression for the alignment of a type (given
1243 by @var{type} as a tree node) if the alignment computed in the usual
1244 way is @var{computed} and the alignment explicitly specified was
1247 The default is to use @var{specified} if it is larger; otherwise, use
1248 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1251 @defmac MAX_FIXED_MODE_SIZE
1252 An integer expression for the size in bits of the largest integer
1253 machine mode that should actually be used. All integer machine modes of
1254 this size or smaller can be used for structures and unions with the
1255 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1256 (DImode)} is assumed.
1259 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1260 If defined, an expression of type @code{enum machine_mode} that
1261 specifies the mode of the save area operand of a
1262 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1263 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1264 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1265 having its mode specified.
1267 You need not define this macro if it always returns @code{Pmode}. You
1268 would most commonly define this macro if the
1269 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1273 @defmac STACK_SIZE_MODE
1274 If defined, an expression of type @code{enum machine_mode} that
1275 specifies the mode of the size increment operand of an
1276 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1278 You need not define this macro if it always returns @code{word_mode}.
1279 You would most commonly define this macro if the @code{allocate_stack}
1280 pattern needs to support both a 32- and a 64-bit mode.
1283 @defmac TARGET_FLOAT_FORMAT
1284 A code distinguishing the floating point format of the target machine.
1285 There are four defined values:
1288 @item IEEE_FLOAT_FORMAT
1289 This code indicates IEEE floating point. It is the default; there is no
1290 need to define @code{TARGET_FLOAT_FORMAT} when the format is IEEE@.
1292 @item VAX_FLOAT_FORMAT
1293 This code indicates the ``F float'' (for @code{float}) and ``D float''
1294 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
1296 @item IBM_FLOAT_FORMAT
1297 This code indicates the format used on the IBM System/370.
1299 @item C4X_FLOAT_FORMAT
1300 This code indicates the format used on the TMS320C3x/C4x.
1303 If your target uses a floating point format other than these, you must
1304 define a new @var{name}_FLOAT_FORMAT code for it, and add support for
1305 it to @file{real.c}.
1307 The ordering of the component words of floating point values stored in
1308 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1311 @defmac MODE_HAS_NANS (@var{mode})
1312 When defined, this macro should be true if @var{mode} has a NaN
1313 representation. The compiler assumes that NaNs are not equal to
1314 anything (including themselves) and that addition, subtraction,
1315 multiplication and division all return NaNs when one operand is
1318 By default, this macro is true if @var{mode} is a floating-point
1319 mode and the target floating-point format is IEEE@.
1322 @defmac MODE_HAS_INFINITIES (@var{mode})
1323 This macro should be true if @var{mode} can represent infinity. At
1324 present, the compiler uses this macro to decide whether @samp{x - x}
1325 is always defined. By default, the macro is true when @var{mode}
1326 is a floating-point mode and the target format is IEEE@.
1329 @defmac MODE_HAS_SIGNED_ZEROS (@var{mode})
1330 True if @var{mode} distinguishes between positive and negative zero.
1331 The rules are expected to follow the IEEE standard:
1335 @samp{x + x} has the same sign as @samp{x}.
1338 If the sum of two values with opposite sign is zero, the result is
1339 positive for all rounding modes expect towards @minus{}infinity, for
1340 which it is negative.
1343 The sign of a product or quotient is negative when exactly one
1344 of the operands is negative.
1347 The default definition is true if @var{mode} is a floating-point
1348 mode and the target format is IEEE@.
1351 @defmac MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1352 If defined, this macro should be true for @var{mode} if it has at
1353 least one rounding mode in which @samp{x} and @samp{-x} can be
1354 rounded to numbers of different magnitude. Two such modes are
1355 towards @minus{}infinity and towards +infinity.
1357 The default definition of this macro is true if @var{mode} is
1358 a floating-point mode and the target format is IEEE@.
1361 @defmac ROUND_TOWARDS_ZERO
1362 If defined, this macro should be true if the prevailing rounding
1363 mode is towards zero. A true value has the following effects:
1367 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1370 @file{libgcc.a}'s floating-point emulator will round towards zero
1371 rather than towards nearest.
1374 The compiler's floating-point emulator will round towards zero after
1375 doing arithmetic, and when converting from the internal float format to
1379 The macro does not affect the parsing of string literals. When the
1380 primary rounding mode is towards zero, library functions like
1381 @code{strtod} might still round towards nearest, and the compiler's
1382 parser should behave like the target's @code{strtod} where possible.
1384 Not defining this macro is equivalent to returning zero.
1387 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1388 This macro should return true if floats with @var{size}
1389 bits do not have a NaN or infinity representation, but use the largest
1390 exponent for normal numbers instead.
1392 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1393 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1394 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1395 floating-point arithmetic.
1397 The default definition of this macro returns false for all sizes.
1400 @deftypefn {Target Hook} bool TARGET_VECTOR_OPAQUE_P (tree @var{type})
1401 This target hook should return @code{true} a vector is opaque. That
1402 is, if no cast is needed when copying a vector value of type
1403 @var{type} into another vector lvalue of the same size. Vector opaque
1404 types cannot be initialized. The default is that there are no such
1408 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1409 This target hook returns @code{true} if bit-fields in the given
1410 @var{record_type} are to be laid out following the rules of Microsoft
1411 Visual C/C++, namely: (i) a bit-field won't share the same storage
1412 unit with the previous bit-field if their underlying types have
1413 different sizes, and the bit-field will be aligned to the highest
1414 alignment of the underlying types of itself and of the previous
1415 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1416 the whole enclosing structure, even if it is unnamed; except that
1417 (iii) a zero-sized bit-field will be disregarded unless it follows
1418 another bit-field of nonzero size. If this hook returns @code{true},
1419 other macros that control bit-field layout are ignored.
1421 When a bit-field is inserted into a packed record, the whole size
1422 of the underlying type is used by one or more same-size adjacent
1423 bit-fields (that is, if its long:3, 32 bits is used in the record,
1424 and any additional adjacent long bit-fields are packed into the same
1425 chunk of 32 bits. However, if the size changes, a new field of that
1426 size is allocated). In an unpacked record, this is the same as using
1427 alignment, but not equivalent when packing.
1429 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1430 the latter will take precedence. If @samp{__attribute__((packed))} is
1431 used on a single field when MS bit-fields are in use, it will take
1432 precedence for that field, but the alignment of the rest of the structure
1433 may affect its placement.
1436 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_FUNDAMENTAL_TYPE (tree @var{type})
1437 If your target defines any fundamental types, define this hook to
1438 return the appropriate encoding for these types as part of a C++
1439 mangled name. The @var{type} argument is the tree structure
1440 representing the type to be mangled. The hook may be applied to trees
1441 which are not target-specific fundamental types; it should return
1442 @code{NULL} for all such types, as well as arguments it does not
1443 recognize. If the return value is not @code{NULL}, it must point to
1444 a statically-allocated string constant.
1446 Target-specific fundamental types might be new fundamental types or
1447 qualified versions of ordinary fundamental types. Encode new
1448 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1449 is the name used for the type in source code, and @var{n} is the
1450 length of @var{name} in decimal. Encode qualified versions of
1451 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1452 @var{name} is the name used for the type qualifier in source code,
1453 @var{n} is the length of @var{name} as above, and @var{code} is the
1454 code used to represent the unqualified version of this type. (See
1455 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1456 codes.) In both cases the spaces are for clarity; do not include any
1457 spaces in your string.
1459 The default version of this hook always returns @code{NULL}, which is
1460 appropriate for a target that does not define any new fundamental
1465 @section Layout of Source Language Data Types
1467 These macros define the sizes and other characteristics of the standard
1468 basic data types used in programs being compiled. Unlike the macros in
1469 the previous section, these apply to specific features of C and related
1470 languages, rather than to fundamental aspects of storage layout.
1472 @defmac INT_TYPE_SIZE
1473 A C expression for the size in bits of the type @code{int} on the
1474 target machine. If you don't define this, the default is one word.
1477 @defmac SHORT_TYPE_SIZE
1478 A C expression for the size in bits of the type @code{short} on the
1479 target machine. If you don't define this, the default is half a word.
1480 (If this would be less than one storage unit, it is rounded up to one
1484 @defmac LONG_TYPE_SIZE
1485 A C expression for the size in bits of the type @code{long} on the
1486 target machine. If you don't define this, the default is one word.
1489 @defmac ADA_LONG_TYPE_SIZE
1490 On some machines, the size used for the Ada equivalent of the type
1491 @code{long} by a native Ada compiler differs from that used by C@. In
1492 that situation, define this macro to be a C expression to be used for
1493 the size of that type. If you don't define this, the default is the
1494 value of @code{LONG_TYPE_SIZE}.
1497 @defmac LONG_LONG_TYPE_SIZE
1498 A C expression for the size in bits of the type @code{long long} on the
1499 target machine. If you don't define this, the default is two
1500 words. If you want to support GNU Ada on your machine, the value of this
1501 macro must be at least 64.
1504 @defmac CHAR_TYPE_SIZE
1505 A C expression for the size in bits of the type @code{char} on the
1506 target machine. If you don't define this, the default is
1507 @code{BITS_PER_UNIT}.
1510 @defmac BOOL_TYPE_SIZE
1511 A C expression for the size in bits of the C++ type @code{bool} and
1512 C99 type @code{_Bool} on the target machine. If you don't define
1513 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1516 @defmac FLOAT_TYPE_SIZE
1517 A C expression for the size in bits of the type @code{float} on the
1518 target machine. If you don't define this, the default is one word.
1521 @defmac DOUBLE_TYPE_SIZE
1522 A C expression for the size in bits of the type @code{double} on the
1523 target machine. If you don't define this, the default is two
1527 @defmac LONG_DOUBLE_TYPE_SIZE
1528 A C expression for the size in bits of the type @code{long double} on
1529 the target machine. If you don't define this, the default is two
1533 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1534 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1535 if you want routines in @file{libgcc2.a} for a size other than
1536 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1537 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1540 @defmac LIBGCC2_HAS_DF_MODE
1541 Define this macro if neither @code{LIBGCC2_DOUBLE_TYPE_SIZE} nor
1542 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1543 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1544 anyway. If you don't define this and either @code{LIBGCC2_DOUBLE_TYPE_SIZE}
1545 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1549 @defmac LIBGCC2_HAS_XF_MODE
1550 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1551 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1552 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1553 is 80 then the default is 1, otherwise it is 0.
1556 @defmac LIBGCC2_HAS_TF_MODE
1557 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1558 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1559 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1560 is 128 then the default is 1, otherwise it is 0.
1563 @defmac TARGET_FLT_EVAL_METHOD
1564 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1565 assuming, if applicable, that the floating-point control word is in its
1566 default state. If you do not define this macro the value of
1567 @code{FLT_EVAL_METHOD} will be zero.
1570 @defmac WIDEST_HARDWARE_FP_SIZE
1571 A C expression for the size in bits of the widest floating-point format
1572 supported by the hardware. If you define this macro, you must specify a
1573 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1574 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1578 @defmac DEFAULT_SIGNED_CHAR
1579 An expression whose value is 1 or 0, according to whether the type
1580 @code{char} should be signed or unsigned by default. The user can
1581 always override this default with the options @option{-fsigned-char}
1582 and @option{-funsigned-char}.
1585 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1586 This target hook should return true if the compiler should give an
1587 @code{enum} type only as many bytes as it takes to represent the range
1588 of possible values of that type. It should return false if all
1589 @code{enum} types should be allocated like @code{int}.
1591 The default is to return false.
1595 A C expression for a string describing the name of the data type to use
1596 for size values. The typedef name @code{size_t} is defined using the
1597 contents of the string.
1599 The string can contain more than one keyword. If so, separate them with
1600 spaces, and write first any length keyword, then @code{unsigned} if
1601 appropriate, and finally @code{int}. The string must exactly match one
1602 of the data type names defined in the function
1603 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1604 omit @code{int} or change the order---that would cause the compiler to
1607 If you don't define this macro, the default is @code{"long unsigned
1611 @defmac PTRDIFF_TYPE
1612 A C expression for a string describing the name of the data type to use
1613 for the result of subtracting two pointers. The typedef name
1614 @code{ptrdiff_t} is defined using the contents of the string. See
1615 @code{SIZE_TYPE} above for more information.
1617 If you don't define this macro, the default is @code{"long int"}.
1621 A C expression for a string describing the name of the data type to use
1622 for wide characters. The typedef name @code{wchar_t} is defined using
1623 the contents of the string. See @code{SIZE_TYPE} above for more
1626 If you don't define this macro, the default is @code{"int"}.
1629 @defmac WCHAR_TYPE_SIZE
1630 A C expression for the size in bits of the data type for wide
1631 characters. This is used in @code{cpp}, which cannot make use of
1636 A C expression for a string describing the name of the data type to
1637 use for wide characters passed to @code{printf} and returned from
1638 @code{getwc}. The typedef name @code{wint_t} is defined using the
1639 contents of the string. See @code{SIZE_TYPE} above for more
1642 If you don't define this macro, the default is @code{"unsigned int"}.
1646 A C expression for a string describing the name of the data type that
1647 can represent any value of any standard or extended signed integer type.
1648 The typedef name @code{intmax_t} is defined using the contents of the
1649 string. See @code{SIZE_TYPE} above for more information.
1651 If you don't define this macro, the default is the first of
1652 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1653 much precision as @code{long long int}.
1656 @defmac UINTMAX_TYPE
1657 A C expression for a string describing the name of the data type that
1658 can represent any value of any standard or extended unsigned integer
1659 type. The typedef name @code{uintmax_t} is defined using the contents
1660 of the string. See @code{SIZE_TYPE} above for more information.
1662 If you don't define this macro, the default is the first of
1663 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1664 unsigned int"} that has as much precision as @code{long long unsigned
1668 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1669 The C++ compiler represents a pointer-to-member-function with a struct
1676 ptrdiff_t vtable_index;
1683 The C++ compiler must use one bit to indicate whether the function that
1684 will be called through a pointer-to-member-function is virtual.
1685 Normally, we assume that the low-order bit of a function pointer must
1686 always be zero. Then, by ensuring that the vtable_index is odd, we can
1687 distinguish which variant of the union is in use. But, on some
1688 platforms function pointers can be odd, and so this doesn't work. In
1689 that case, we use the low-order bit of the @code{delta} field, and shift
1690 the remainder of the @code{delta} field to the left.
1692 GCC will automatically make the right selection about where to store
1693 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1694 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1695 set such that functions always start at even addresses, but the lowest
1696 bit of pointers to functions indicate whether the function at that
1697 address is in ARM or Thumb mode. If this is the case of your
1698 architecture, you should define this macro to
1699 @code{ptrmemfunc_vbit_in_delta}.
1701 In general, you should not have to define this macro. On architectures
1702 in which function addresses are always even, according to
1703 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1704 @code{ptrmemfunc_vbit_in_pfn}.
1707 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1708 Normally, the C++ compiler uses function pointers in vtables. This
1709 macro allows the target to change to use ``function descriptors''
1710 instead. Function descriptors are found on targets for whom a
1711 function pointer is actually a small data structure. Normally the
1712 data structure consists of the actual code address plus a data
1713 pointer to which the function's data is relative.
1715 If vtables are used, the value of this macro should be the number
1716 of words that the function descriptor occupies.
1719 @defmac TARGET_VTABLE_ENTRY_ALIGN
1720 By default, the vtable entries are void pointers, the so the alignment
1721 is the same as pointer alignment. The value of this macro specifies
1722 the alignment of the vtable entry in bits. It should be defined only
1723 when special alignment is necessary. */
1726 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1727 There are a few non-descriptor entries in the vtable at offsets below
1728 zero. If these entries must be padded (say, to preserve the alignment
1729 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1730 of words in each data entry.
1734 @section Register Usage
1735 @cindex register usage
1737 This section explains how to describe what registers the target machine
1738 has, and how (in general) they can be used.
1740 The description of which registers a specific instruction can use is
1741 done with register classes; see @ref{Register Classes}. For information
1742 on using registers to access a stack frame, see @ref{Frame Registers}.
1743 For passing values in registers, see @ref{Register Arguments}.
1744 For returning values in registers, see @ref{Scalar Return}.
1747 * Register Basics:: Number and kinds of registers.
1748 * Allocation Order:: Order in which registers are allocated.
1749 * Values in Registers:: What kinds of values each reg can hold.
1750 * Leaf Functions:: Renumbering registers for leaf functions.
1751 * Stack Registers:: Handling a register stack such as 80387.
1754 @node Register Basics
1755 @subsection Basic Characteristics of Registers
1757 @c prevent bad page break with this line
1758 Registers have various characteristics.
1760 @defmac FIRST_PSEUDO_REGISTER
1761 Number of hardware registers known to the compiler. They receive
1762 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1763 pseudo register's number really is assigned the number
1764 @code{FIRST_PSEUDO_REGISTER}.
1767 @defmac FIXED_REGISTERS
1768 @cindex fixed register
1769 An initializer that says which registers are used for fixed purposes
1770 all throughout the compiled code and are therefore not available for
1771 general allocation. These would include the stack pointer, the frame
1772 pointer (except on machines where that can be used as a general
1773 register when no frame pointer is needed), the program counter on
1774 machines where that is considered one of the addressable registers,
1775 and any other numbered register with a standard use.
1777 This information is expressed as a sequence of numbers, separated by
1778 commas and surrounded by braces. The @var{n}th number is 1 if
1779 register @var{n} is fixed, 0 otherwise.
1781 The table initialized from this macro, and the table initialized by
1782 the following one, may be overridden at run time either automatically,
1783 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1784 the user with the command options @option{-ffixed-@var{reg}},
1785 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1788 @defmac CALL_USED_REGISTERS
1789 @cindex call-used register
1790 @cindex call-clobbered register
1791 @cindex call-saved register
1792 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1793 clobbered (in general) by function calls as well as for fixed
1794 registers. This macro therefore identifies the registers that are not
1795 available for general allocation of values that must live across
1798 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1799 automatically saves it on function entry and restores it on function
1800 exit, if the register is used within the function.
1803 @defmac CALL_REALLY_USED_REGISTERS
1804 @cindex call-used register
1805 @cindex call-clobbered register
1806 @cindex call-saved register
1807 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1808 that the entire set of @code{FIXED_REGISTERS} be included.
1809 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1810 This macro is optional. If not specified, it defaults to the value
1811 of @code{CALL_USED_REGISTERS}.
1814 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1815 @cindex call-used register
1816 @cindex call-clobbered register
1817 @cindex call-saved register
1818 A C expression that is nonzero if it is not permissible to store a
1819 value of mode @var{mode} in hard register number @var{regno} across a
1820 call without some part of it being clobbered. For most machines this
1821 macro need not be defined. It is only required for machines that do not
1822 preserve the entire contents of a register across a call.
1826 @findex call_used_regs
1829 @findex reg_class_contents
1830 @defmac CONDITIONAL_REGISTER_USAGE
1831 Zero or more C statements that may conditionally modify five variables
1832 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1833 @code{reg_names}, and @code{reg_class_contents}, to take into account
1834 any dependence of these register sets on target flags. The first three
1835 of these are of type @code{char []} (interpreted as Boolean vectors).
1836 @code{global_regs} is a @code{const char *[]}, and
1837 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1838 called, @code{fixed_regs}, @code{call_used_regs},
1839 @code{reg_class_contents}, and @code{reg_names} have been initialized
1840 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1841 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1842 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1843 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1844 command options have been applied.
1846 You need not define this macro if it has no work to do.
1848 @cindex disabling certain registers
1849 @cindex controlling register usage
1850 If the usage of an entire class of registers depends on the target
1851 flags, you may indicate this to GCC by using this macro to modify
1852 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1853 registers in the classes which should not be used by GCC@. Also define
1854 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1855 to return @code{NO_REGS} if it
1856 is called with a letter for a class that shouldn't be used.
1858 (However, if this class is not included in @code{GENERAL_REGS} and all
1859 of the insn patterns whose constraints permit this class are
1860 controlled by target switches, then GCC will automatically avoid using
1861 these registers when the target switches are opposed to them.)
1864 @defmac INCOMING_REGNO (@var{out})
1865 Define this macro if the target machine has register windows. This C
1866 expression returns the register number as seen by the called function
1867 corresponding to the register number @var{out} as seen by the calling
1868 function. Return @var{out} if register number @var{out} is not an
1872 @defmac OUTGOING_REGNO (@var{in})
1873 Define this macro if the target machine has register windows. This C
1874 expression returns the register number as seen by the calling function
1875 corresponding to the register number @var{in} as seen by the called
1876 function. Return @var{in} if register number @var{in} is not an inbound
1880 @defmac LOCAL_REGNO (@var{regno})
1881 Define this macro if the target machine has register windows. This C
1882 expression returns true if the register is call-saved but is in the
1883 register window. Unlike most call-saved registers, such registers
1884 need not be explicitly restored on function exit or during non-local
1889 If the program counter has a register number, define this as that
1890 register number. Otherwise, do not define it.
1893 @node Allocation Order
1894 @subsection Order of Allocation of Registers
1895 @cindex order of register allocation
1896 @cindex register allocation order
1898 @c prevent bad page break with this line
1899 Registers are allocated in order.
1901 @defmac REG_ALLOC_ORDER
1902 If defined, an initializer for a vector of integers, containing the
1903 numbers of hard registers in the order in which GCC should prefer
1904 to use them (from most preferred to least).
1906 If this macro is not defined, registers are used lowest numbered first
1907 (all else being equal).
1909 One use of this macro is on machines where the highest numbered
1910 registers must always be saved and the save-multiple-registers
1911 instruction supports only sequences of consecutive registers. On such
1912 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1913 the highest numbered allocable register first.
1916 @defmac ORDER_REGS_FOR_LOCAL_ALLOC
1917 A C statement (sans semicolon) to choose the order in which to allocate
1918 hard registers for pseudo-registers local to a basic block.
1920 Store the desired register order in the array @code{reg_alloc_order}.
1921 Element 0 should be the register to allocate first; element 1, the next
1922 register; and so on.
1924 The macro body should not assume anything about the contents of
1925 @code{reg_alloc_order} before execution of the macro.
1927 On most machines, it is not necessary to define this macro.
1930 @node Values in Registers
1931 @subsection How Values Fit in Registers
1933 This section discusses the macros that describe which kinds of values
1934 (specifically, which machine modes) each register can hold, and how many
1935 consecutive registers are needed for a given mode.
1937 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
1938 A C expression for the number of consecutive hard registers, starting
1939 at register number @var{regno}, required to hold a value of mode
1942 On a machine where all registers are exactly one word, a suitable
1943 definition of this macro is
1946 #define HARD_REGNO_NREGS(REGNO, MODE) \
1947 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
1952 @defmac REGMODE_NATURAL_SIZE (@var{mode})
1953 Define this macro if the natural size of registers that hold values
1954 of mode @var{mode} is not the word size. It is a C expression that
1955 should give the natural size in bytes for the specified mode. It is
1956 used by the register allocator to try to optimize its results. This
1957 happens for example on SPARC 64-bit where the natural size of
1958 floating-point registers is still 32-bit.
1961 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
1962 A C expression that is nonzero if it is permissible to store a value
1963 of mode @var{mode} in hard register number @var{regno} (or in several
1964 registers starting with that one). For a machine where all registers
1965 are equivalent, a suitable definition is
1968 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
1971 You need not include code to check for the numbers of fixed registers,
1972 because the allocation mechanism considers them to be always occupied.
1974 @cindex register pairs
1975 On some machines, double-precision values must be kept in even/odd
1976 register pairs. You can implement that by defining this macro to reject
1977 odd register numbers for such modes.
1979 The minimum requirement for a mode to be OK in a register is that the
1980 @samp{mov@var{mode}} instruction pattern support moves between the
1981 register and other hard register in the same class and that moving a
1982 value into the register and back out not alter it.
1984 Since the same instruction used to move @code{word_mode} will work for
1985 all narrower integer modes, it is not necessary on any machine for
1986 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
1987 you define patterns @samp{movhi}, etc., to take advantage of this. This
1988 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
1989 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
1992 Many machines have special registers for floating point arithmetic.
1993 Often people assume that floating point machine modes are allowed only
1994 in floating point registers. This is not true. Any registers that
1995 can hold integers can safely @emph{hold} a floating point machine
1996 mode, whether or not floating arithmetic can be done on it in those
1997 registers. Integer move instructions can be used to move the values.
1999 On some machines, though, the converse is true: fixed-point machine
2000 modes may not go in floating registers. This is true if the floating
2001 registers normalize any value stored in them, because storing a
2002 non-floating value there would garble it. In this case,
2003 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2004 floating registers. But if the floating registers do not automatically
2005 normalize, if you can store any bit pattern in one and retrieve it
2006 unchanged without a trap, then any machine mode may go in a floating
2007 register, so you can define this macro to say so.
2009 The primary significance of special floating registers is rather that
2010 they are the registers acceptable in floating point arithmetic
2011 instructions. However, this is of no concern to
2012 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2013 constraints for those instructions.
2015 On some machines, the floating registers are especially slow to access,
2016 so that it is better to store a value in a stack frame than in such a
2017 register if floating point arithmetic is not being done. As long as the
2018 floating registers are not in class @code{GENERAL_REGS}, they will not
2019 be used unless some pattern's constraint asks for one.
2022 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2023 A C expression that is nonzero if it is OK to rename a hard register
2024 @var{from} to another hard register @var{to}.
2026 One common use of this macro is to prevent renaming of a register to
2027 another register that is not saved by a prologue in an interrupt
2030 The default is always nonzero.
2033 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2034 A C expression that is nonzero if a value of mode
2035 @var{mode1} is accessible in mode @var{mode2} without copying.
2037 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2038 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2039 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2040 should be nonzero. If they differ for any @var{r}, you should define
2041 this macro to return zero unless some other mechanism ensures the
2042 accessibility of the value in a narrower mode.
2044 You should define this macro to return nonzero in as many cases as
2045 possible since doing so will allow GCC to perform better register
2049 @defmac AVOID_CCMODE_COPIES
2050 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2051 registers. You should only define this macro if support for copying to/from
2052 @code{CCmode} is incomplete.
2055 @node Leaf Functions
2056 @subsection Handling Leaf Functions
2058 @cindex leaf functions
2059 @cindex functions, leaf
2060 On some machines, a leaf function (i.e., one which makes no calls) can run
2061 more efficiently if it does not make its own register window. Often this
2062 means it is required to receive its arguments in the registers where they
2063 are passed by the caller, instead of the registers where they would
2066 The special treatment for leaf functions generally applies only when
2067 other conditions are met; for example, often they may use only those
2068 registers for its own variables and temporaries. We use the term ``leaf
2069 function'' to mean a function that is suitable for this special
2070 handling, so that functions with no calls are not necessarily ``leaf
2073 GCC assigns register numbers before it knows whether the function is
2074 suitable for leaf function treatment. So it needs to renumber the
2075 registers in order to output a leaf function. The following macros
2078 @defmac LEAF_REGISTERS
2079 Name of a char vector, indexed by hard register number, which
2080 contains 1 for a register that is allowable in a candidate for leaf
2083 If leaf function treatment involves renumbering the registers, then the
2084 registers marked here should be the ones before renumbering---those that
2085 GCC would ordinarily allocate. The registers which will actually be
2086 used in the assembler code, after renumbering, should not be marked with 1
2089 Define this macro only if the target machine offers a way to optimize
2090 the treatment of leaf functions.
2093 @defmac LEAF_REG_REMAP (@var{regno})
2094 A C expression whose value is the register number to which @var{regno}
2095 should be renumbered, when a function is treated as a leaf function.
2097 If @var{regno} is a register number which should not appear in a leaf
2098 function before renumbering, then the expression should yield @minus{}1, which
2099 will cause the compiler to abort.
2101 Define this macro only if the target machine offers a way to optimize the
2102 treatment of leaf functions, and registers need to be renumbered to do
2106 @findex current_function_is_leaf
2107 @findex current_function_uses_only_leaf_regs
2108 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2109 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2110 specially. They can test the C variable @code{current_function_is_leaf}
2111 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2112 set prior to local register allocation and is valid for the remaining
2113 compiler passes. They can also test the C variable
2114 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2115 functions which only use leaf registers.
2116 @code{current_function_uses_only_leaf_regs} is valid after all passes
2117 that modify the instructions have been run and is only useful if
2118 @code{LEAF_REGISTERS} is defined.
2119 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2120 @c of the next paragraph?! --mew 2feb93
2122 @node Stack Registers
2123 @subsection Registers That Form a Stack
2125 There are special features to handle computers where some of the
2126 ``registers'' form a stack. Stack registers are normally written by
2127 pushing onto the stack, and are numbered relative to the top of the
2130 Currently, GCC can only handle one group of stack-like registers, and
2131 they must be consecutively numbered. Furthermore, the existing
2132 support for stack-like registers is specific to the 80387 floating
2133 point coprocessor. If you have a new architecture that uses
2134 stack-like registers, you will need to do substantial work on
2135 @file{reg-stack.c} and write your machine description to cooperate
2136 with it, as well as defining these macros.
2139 Define this if the machine has any stack-like registers.
2142 @defmac FIRST_STACK_REG
2143 The number of the first stack-like register. This one is the top
2147 @defmac LAST_STACK_REG
2148 The number of the last stack-like register. This one is the bottom of
2152 @node Register Classes
2153 @section Register Classes
2154 @cindex register class definitions
2155 @cindex class definitions, register
2157 On many machines, the numbered registers are not all equivalent.
2158 For example, certain registers may not be allowed for indexed addressing;
2159 certain registers may not be allowed in some instructions. These machine
2160 restrictions are described to the compiler using @dfn{register classes}.
2162 You define a number of register classes, giving each one a name and saying
2163 which of the registers belong to it. Then you can specify register classes
2164 that are allowed as operands to particular instruction patterns.
2168 In general, each register will belong to several classes. In fact, one
2169 class must be named @code{ALL_REGS} and contain all the registers. Another
2170 class must be named @code{NO_REGS} and contain no registers. Often the
2171 union of two classes will be another class; however, this is not required.
2173 @findex GENERAL_REGS
2174 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2175 terribly special about the name, but the operand constraint letters
2176 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2177 the same as @code{ALL_REGS}, just define it as a macro which expands
2180 Order the classes so that if class @var{x} is contained in class @var{y}
2181 then @var{x} has a lower class number than @var{y}.
2183 The way classes other than @code{GENERAL_REGS} are specified in operand
2184 constraints is through machine-dependent operand constraint letters.
2185 You can define such letters to correspond to various classes, then use
2186 them in operand constraints.
2188 You should define a class for the union of two classes whenever some
2189 instruction allows both classes. For example, if an instruction allows
2190 either a floating point (coprocessor) register or a general register for a
2191 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2192 which includes both of them. Otherwise you will get suboptimal code.
2194 You must also specify certain redundant information about the register
2195 classes: for each class, which classes contain it and which ones are
2196 contained in it; for each pair of classes, the largest class contained
2199 When a value occupying several consecutive registers is expected in a
2200 certain class, all the registers used must belong to that class.
2201 Therefore, register classes cannot be used to enforce a requirement for
2202 a register pair to start with an even-numbered register. The way to
2203 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2205 Register classes used for input-operands of bitwise-and or shift
2206 instructions have a special requirement: each such class must have, for
2207 each fixed-point machine mode, a subclass whose registers can transfer that
2208 mode to or from memory. For example, on some machines, the operations for
2209 single-byte values (@code{QImode}) are limited to certain registers. When
2210 this is so, each register class that is used in a bitwise-and or shift
2211 instruction must have a subclass consisting of registers from which
2212 single-byte values can be loaded or stored. This is so that
2213 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2215 @deftp {Data type} {enum reg_class}
2216 An enumerated type that must be defined with all the register class names
2217 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2218 must be the last register class, followed by one more enumerated value,
2219 @code{LIM_REG_CLASSES}, which is not a register class but rather
2220 tells how many classes there are.
2222 Each register class has a number, which is the value of casting
2223 the class name to type @code{int}. The number serves as an index
2224 in many of the tables described below.
2227 @defmac N_REG_CLASSES
2228 The number of distinct register classes, defined as follows:
2231 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2235 @defmac REG_CLASS_NAMES
2236 An initializer containing the names of the register classes as C string
2237 constants. These names are used in writing some of the debugging dumps.
2240 @defmac REG_CLASS_CONTENTS
2241 An initializer containing the contents of the register classes, as integers
2242 which are bit masks. The @var{n}th integer specifies the contents of class
2243 @var{n}. The way the integer @var{mask} is interpreted is that
2244 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2246 When the machine has more than 32 registers, an integer does not suffice.
2247 Then the integers are replaced by sub-initializers, braced groupings containing
2248 several integers. Each sub-initializer must be suitable as an initializer
2249 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2250 In this situation, the first integer in each sub-initializer corresponds to
2251 registers 0 through 31, the second integer to registers 32 through 63, and
2255 @defmac REGNO_REG_CLASS (@var{regno})
2256 A C expression whose value is a register class containing hard register
2257 @var{regno}. In general there is more than one such class; choose a class
2258 which is @dfn{minimal}, meaning that no smaller class also contains the
2262 @defmac BASE_REG_CLASS
2263 A macro whose definition is the name of the class to which a valid
2264 base register must belong. A base register is one used in an address
2265 which is the register value plus a displacement.
2268 @defmac MODE_BASE_REG_CLASS (@var{mode})
2269 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2270 the selection of a base register in a mode dependent manner. If
2271 @var{mode} is VOIDmode then it should return the same value as
2272 @code{BASE_REG_CLASS}.
2275 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2276 A C expression whose value is the register class to which a valid
2277 base register must belong in order to be used in a base plus index
2278 register address. You should define this macro if base plus index
2279 addresses have different requirements than other base register uses.
2282 @defmac INDEX_REG_CLASS
2283 A macro whose definition is the name of the class to which a valid
2284 index register must belong. An index register is one used in an
2285 address where its value is either multiplied by a scale factor or
2286 added to another register (as well as added to a displacement).
2289 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2290 For the constraint at the start of @var{str}, which starts with the letter
2291 @var{c}, return the length. This allows you to have register class /
2292 constant / extra constraints that are longer than a single letter;
2293 you don't need to define this macro if you can do with single-letter
2294 constraints only. The definition of this macro should use
2295 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2296 to handle specially.
2297 There are some sanity checks in genoutput.c that check the constraint lengths
2298 for the md file, so you can also use this macro to help you while you are
2299 transitioning from a byzantine single-letter-constraint scheme: when you
2300 return a negative length for a constraint you want to re-use, genoutput
2301 will complain about every instance where it is used in the md file.
2304 @defmac REG_CLASS_FROM_LETTER (@var{char})
2305 A C expression which defines the machine-dependent operand constraint
2306 letters for register classes. If @var{char} is such a letter, the
2307 value should be the register class corresponding to it. Otherwise,
2308 the value should be @code{NO_REGS}. The register letter @samp{r},
2309 corresponding to class @code{GENERAL_REGS}, will not be passed
2310 to this macro; you do not need to handle it.
2313 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2314 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2315 passed in @var{str}, so that you can use suffixes to distinguish between
2319 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2320 A C expression which is nonzero if register number @var{num} is
2321 suitable for use as a base register in operand addresses. It may be
2322 either a suitable hard register or a pseudo register that has been
2323 allocated such a hard register.
2326 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2327 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2328 that expression may examine the mode of the memory reference in
2329 @var{mode}. You should define this macro if the mode of the memory
2330 reference affects whether a register may be used as a base register. If
2331 you define this macro, the compiler will use it instead of
2332 @code{REGNO_OK_FOR_BASE_P}.
2335 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2336 A C expression which is nonzero if register number @var{num} is suitable for
2337 use as a base register in base plus index operand addresses, accessing
2338 memory in mode @var{mode}. It may be either a suitable hard register or a
2339 pseudo register that has been allocated such a hard register. You should
2340 define this macro if base plus index addresses have different requirements
2341 than other base register uses.
2344 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2345 A C expression which is nonzero if register number @var{num} is
2346 suitable for use as an index register in operand addresses. It may be
2347 either a suitable hard register or a pseudo register that has been
2348 allocated such a hard register.
2350 The difference between an index register and a base register is that
2351 the index register may be scaled. If an address involves the sum of
2352 two registers, neither one of them scaled, then either one may be
2353 labeled the ``base'' and the other the ``index''; but whichever
2354 labeling is used must fit the machine's constraints of which registers
2355 may serve in each capacity. The compiler will try both labelings,
2356 looking for one that is valid, and will reload one or both registers
2357 only if neither labeling works.
2360 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2361 A C expression that places additional restrictions on the register class
2362 to use when it is necessary to copy value @var{x} into a register in class
2363 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2364 another, smaller class. On many machines, the following definition is
2368 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2371 Sometimes returning a more restrictive class makes better code. For
2372 example, on the 68000, when @var{x} is an integer constant that is in range
2373 for a @samp{moveq} instruction, the value of this macro is always
2374 @code{DATA_REGS} as long as @var{class} includes the data registers.
2375 Requiring a data register guarantees that a @samp{moveq} will be used.
2377 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2378 @var{class} is if @var{x} is a legitimate constant which cannot be
2379 loaded into some register class. By returning @code{NO_REGS} you can
2380 force @var{x} into a memory location. For example, rs6000 can load
2381 immediate values into general-purpose registers, but does not have an
2382 instruction for loading an immediate value into a floating-point
2383 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2384 @var{x} is a floating-point constant. If the constant can't be loaded
2385 into any kind of register, code generation will be better if
2386 @code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2387 of using @code{PREFERRED_RELOAD_CLASS}.
2390 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2391 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2392 input reloads. If you don't define this macro, the default is to use
2393 @var{class}, unchanged.
2396 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2397 A C expression that places additional restrictions on the register class
2398 to use when it is necessary to be able to hold a value of mode
2399 @var{mode} in a reload register for which class @var{class} would
2402 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2403 there are certain modes that simply can't go in certain reload classes.
2405 The value is a register class; perhaps @var{class}, or perhaps another,
2408 Don't define this macro unless the target machine has limitations which
2409 require the macro to do something nontrivial.
2412 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2413 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2414 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2415 Many machines have some registers that cannot be copied directly to or
2416 from memory or even from other types of registers. An example is the
2417 @samp{MQ} register, which on most machines, can only be copied to or
2418 from general registers, but not memory. Some machines allow copying all
2419 registers to and from memory, but require a scratch register for stores
2420 to some memory locations (e.g., those with symbolic address on the RT,
2421 and those with certain symbolic address on the SPARC when compiling
2422 PIC)@. In some cases, both an intermediate and a scratch register are
2425 You should define these macros to indicate to the reload phase that it may
2426 need to allocate at least one register for a reload in addition to the
2427 register to contain the data. Specifically, if copying @var{x} to a
2428 register @var{class} in @var{mode} requires an intermediate register,
2429 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2430 largest register class all of whose registers can be used as
2431 intermediate registers or scratch registers.
2433 If copying a register @var{class} in @var{mode} to @var{x} requires an
2434 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2435 should be defined to return the largest register class required. If the
2436 requirements for input and output reloads are the same, the macro
2437 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2440 The values returned by these macros are often @code{GENERAL_REGS}.
2441 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2442 can be directly copied to or from a register of @var{class} in
2443 @var{mode} without requiring a scratch register. Do not define this
2444 macro if it would always return @code{NO_REGS}.
2446 If a scratch register is required (either with or without an
2447 intermediate register), you should define patterns for
2448 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2449 (@pxref{Standard Names}. These patterns, which will normally be
2450 implemented with a @code{define_expand}, should be similar to the
2451 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2454 Define constraints for the reload register and scratch register that
2455 contain a single register class. If the original reload register (whose
2456 class is @var{class}) can meet the constraint given in the pattern, the
2457 value returned by these macros is used for the class of the scratch
2458 register. Otherwise, two additional reload registers are required.
2459 Their classes are obtained from the constraints in the insn pattern.
2461 @var{x} might be a pseudo-register or a @code{subreg} of a
2462 pseudo-register, which could either be in a hard register or in memory.
2463 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2464 in memory and the hard register number if it is in a register.
2466 These macros should not be used in the case where a particular class of
2467 registers can only be copied to memory and not to another class of
2468 registers. In that case, secondary reload registers are not needed and
2469 would not be helpful. Instead, a stack location must be used to perform
2470 the copy and the @code{mov@var{m}} pattern should use memory as an
2471 intermediate storage. This case often occurs between floating-point and
2475 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2476 Certain machines have the property that some registers cannot be copied
2477 to some other registers without using memory. Define this macro on
2478 those machines to be a C expression that is nonzero if objects of mode
2479 @var{m} in registers of @var{class1} can only be copied to registers of
2480 class @var{class2} by storing a register of @var{class1} into memory
2481 and loading that memory location into a register of @var{class2}.
2483 Do not define this macro if its value would always be zero.
2486 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2487 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2488 allocates a stack slot for a memory location needed for register copies.
2489 If this macro is defined, the compiler instead uses the memory location
2490 defined by this macro.
2492 Do not define this macro if you do not define
2493 @code{SECONDARY_MEMORY_NEEDED}.
2496 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2497 When the compiler needs a secondary memory location to copy between two
2498 registers of mode @var{mode}, it normally allocates sufficient memory to
2499 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2500 load operations in a mode that many bits wide and whose class is the
2501 same as that of @var{mode}.
2503 This is right thing to do on most machines because it ensures that all
2504 bits of the register are copied and prevents accesses to the registers
2505 in a narrower mode, which some machines prohibit for floating-point
2508 However, this default behavior is not correct on some machines, such as
2509 the DEC Alpha, that store short integers in floating-point registers
2510 differently than in integer registers. On those machines, the default
2511 widening will not work correctly and you must define this macro to
2512 suppress that widening in some cases. See the file @file{alpha.h} for
2515 Do not define this macro if you do not define
2516 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2517 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2520 @defmac SMALL_REGISTER_CLASSES
2521 On some machines, it is risky to let hard registers live across arbitrary
2522 insns. Typically, these machines have instructions that require values
2523 to be in specific registers (like an accumulator), and reload will fail
2524 if the required hard register is used for another purpose across such an
2527 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2528 value on these machines. When this macro has a nonzero value, the
2529 compiler will try to minimize the lifetime of hard registers.
2531 It is always safe to define this macro with a nonzero value, but if you
2532 unnecessarily define it, you will reduce the amount of optimizations
2533 that can be performed in some cases. If you do not define this macro
2534 with a nonzero value when it is required, the compiler will run out of
2535 spill registers and print a fatal error message. For most machines, you
2536 should not define this macro at all.
2539 @defmac CLASS_LIKELY_SPILLED_P (@var{class})
2540 A C expression whose value is nonzero if pseudos that have been assigned
2541 to registers of class @var{class} would likely be spilled because
2542 registers of @var{class} are needed for spill registers.
2544 The default value of this macro returns 1 if @var{class} has exactly one
2545 register and zero otherwise. On most machines, this default should be
2546 used. Only define this macro to some other expression if pseudos
2547 allocated by @file{local-alloc.c} end up in memory because their hard
2548 registers were needed for spill registers. If this macro returns nonzero
2549 for those classes, those pseudos will only be allocated by
2550 @file{global.c}, which knows how to reallocate the pseudo to another
2551 register. If there would not be another register available for
2552 reallocation, you should not change the definition of this macro since
2553 the only effect of such a definition would be to slow down register
2557 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2558 A C expression for the maximum number of consecutive registers
2559 of class @var{class} needed to hold a value of mode @var{mode}.
2561 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2562 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2563 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2564 @var{mode})} for all @var{regno} values in the class @var{class}.
2566 This macro helps control the handling of multiple-word values
2570 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2571 If defined, a C expression that returns nonzero for a @var{class} for which
2572 a change from mode @var{from} to mode @var{to} is invalid.
2574 For the example, loading 32-bit integer or floating-point objects into
2575 floating-point registers on the Alpha extends them to 64 bits.
2576 Therefore loading a 64-bit object and then storing it as a 32-bit object
2577 does not store the low-order 32 bits, as would be the case for a normal
2578 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2582 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2583 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2584 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2588 Three other special macros describe which operands fit which constraint
2591 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2592 A C expression that defines the machine-dependent operand constraint
2593 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2594 particular ranges of integer values. If @var{c} is one of those
2595 letters, the expression should check that @var{value}, an integer, is in
2596 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2597 not one of those letters, the value should be 0 regardless of
2601 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2602 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2603 string passed in @var{str}, so that you can use suffixes to distinguish
2604 between different variants.
2607 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2608 A C expression that defines the machine-dependent operand constraint
2609 letters that specify particular ranges of @code{const_double} values
2610 (@samp{G} or @samp{H}).
2612 If @var{c} is one of those letters, the expression should check that
2613 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2614 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2615 letters, the value should be 0 regardless of @var{value}.
2617 @code{const_double} is used for all floating-point constants and for
2618 @code{DImode} fixed-point constants. A given letter can accept either
2619 or both kinds of values. It can use @code{GET_MODE} to distinguish
2620 between these kinds.
2623 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2624 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2625 string passed in @var{str}, so that you can use suffixes to distinguish
2626 between different variants.
2629 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2630 A C expression that defines the optional machine-dependent constraint
2631 letters that can be used to segregate specific types of operands, usually
2632 memory references, for the target machine. Any letter that is not
2633 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2634 @code{REG_CLASS_FROM_CONSTRAINT}
2635 may be used. Normally this macro will not be defined.
2637 If it is required for a particular target machine, it should return 1
2638 if @var{value} corresponds to the operand type represented by the
2639 constraint letter @var{c}. If @var{c} is not defined as an extra
2640 constraint, the value returned should be 0 regardless of @var{value}.
2642 For example, on the ROMP, load instructions cannot have their output
2643 in r0 if the memory reference contains a symbolic address. Constraint
2644 letter @samp{Q} is defined as representing a memory address that does
2645 @emph{not} contain a symbolic address. An alternative is specified with
2646 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2647 alternative specifies @samp{m} on the input and a register class that
2648 does not include r0 on the output.
2651 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2652 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2653 in @var{str}, so that you can use suffixes to distinguish between different
2657 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2658 A C expression that defines the optional machine-dependent constraint
2659 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2660 be treated like memory constraints by the reload pass.
2662 It should return 1 if the operand type represented by the constraint
2663 at the start of @var{str}, the first letter of which is the letter @var{c},
2664 comprises a subset of all memory references including
2665 all those whose address is simply a base register. This allows the reload
2666 pass to reload an operand, if it does not directly correspond to the operand
2667 type of @var{c}, by copying its address into a base register.
2669 For example, on the S/390, some instructions do not accept arbitrary
2670 memory references, but only those that do not make use of an index
2671 register. The constraint letter @samp{Q} is defined via
2672 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2673 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2674 a @samp{Q} constraint can handle any memory operand, because the
2675 reload pass knows it can be reloaded by copying the memory address
2676 into a base register if required. This is analogous to the way
2677 a @samp{o} constraint can handle any memory operand.
2680 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2681 A C expression that defines the optional machine-dependent constraint
2682 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2683 @code{EXTRA_CONSTRAINT_STR}, that should
2684 be treated like address constraints by the reload pass.
2686 It should return 1 if the operand type represented by the constraint
2687 at the start of @var{str}, which starts with the letter @var{c}, comprises
2688 a subset of all memory addresses including
2689 all those that consist of just a base register. This allows the reload
2690 pass to reload an operand, if it does not directly correspond to the operand
2691 type of @var{str}, by copying it into a base register.
2693 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2694 be used with the @code{address_operand} predicate. It is treated
2695 analogously to the @samp{p} constraint.
2698 @node Stack and Calling
2699 @section Stack Layout and Calling Conventions
2700 @cindex calling conventions
2702 @c prevent bad page break with this line
2703 This describes the stack layout and calling conventions.
2707 * Exception Handling::
2712 * Register Arguments::
2714 * Aggregate Return::
2722 @subsection Basic Stack Layout
2723 @cindex stack frame layout
2724 @cindex frame layout
2726 @c prevent bad page break with this line
2727 Here is the basic stack layout.
2729 @defmac STACK_GROWS_DOWNWARD
2730 Define this macro if pushing a word onto the stack moves the stack
2731 pointer to a smaller address.
2733 When we say, ``define this macro if @dots{}'', it means that the
2734 compiler checks this macro only with @code{#ifdef} so the precise
2735 definition used does not matter.
2738 @defmac STACK_PUSH_CODE
2739 This macro defines the operation used when something is pushed
2740 on the stack. In RTL, a push operation will be
2741 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2743 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2744 and @code{POST_INC}. Which of these is correct depends on
2745 the stack direction and on whether the stack pointer points
2746 to the last item on the stack or whether it points to the
2747 space for the next item on the stack.
2749 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2750 defined, which is almost always right, and @code{PRE_INC} otherwise,
2751 which is often wrong.
2754 @defmac FRAME_GROWS_DOWNWARD
2755 Define this macro to non-zero value if the addresses of local variable slots
2756 are at negative offsets from the frame pointer.
2759 @defmac ARGS_GROW_DOWNWARD
2760 Define this macro if successive arguments to a function occupy decreasing
2761 addresses on the stack.
2764 @defmac STARTING_FRAME_OFFSET
2765 Offset from the frame pointer to the first local variable slot to be allocated.
2767 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2768 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2769 Otherwise, it is found by adding the length of the first slot to the
2770 value @code{STARTING_FRAME_OFFSET}.
2771 @c i'm not sure if the above is still correct.. had to change it to get
2772 @c rid of an overfull. --mew 2feb93
2775 @defmac STACK_ALIGNMENT_NEEDED
2776 Define to zero to disable final alignment of the stack during reload.
2777 The nonzero default for this macro is suitable for most ports.
2779 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
2780 is a register save block following the local block that doesn't require
2781 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2782 stack alignment and do it in the backend.
2785 @defmac STACK_POINTER_OFFSET
2786 Offset from the stack pointer register to the first location at which
2787 outgoing arguments are placed. If not specified, the default value of
2788 zero is used. This is the proper value for most machines.
2790 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2791 the first location at which outgoing arguments are placed.
2794 @defmac FIRST_PARM_OFFSET (@var{fundecl})
2795 Offset from the argument pointer register to the first argument's
2796 address. On some machines it may depend on the data type of the
2799 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2800 the first argument's address.
2803 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
2804 Offset from the stack pointer register to an item dynamically allocated
2805 on the stack, e.g., by @code{alloca}.
2807 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2808 length of the outgoing arguments. The default is correct for most
2809 machines. See @file{function.c} for details.
2812 @defmac INITIAL_FRAME_ADDRESS_RTX
2813 A C expression whose value is RTL representing the address of the initial
2814 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
2815 @code{DYNAMIC_CHAIN_ADDRESS}.
2816 If you don't define this macro, the default is to return
2817 @code{hard_frame_pointer_rtx}.
2818 This default is usually correct unless @code{-fomit-frame-pointer} is in
2820 Define this macro in order to make @code{__builtin_frame_address (0)} and
2821 @code{__builtin_return_address (0)} work even in absence of a hard frame pointer.
2824 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2825 A C expression whose value is RTL representing the address in a stack
2826 frame where the pointer to the caller's frame is stored. Assume that
2827 @var{frameaddr} is an RTL expression for the address of the stack frame
2830 If you don't define this macro, the default is to return the value
2831 of @var{frameaddr}---that is, the stack frame address is also the
2832 address of the stack word that points to the previous frame.
2835 @defmac SETUP_FRAME_ADDRESSES
2836 If defined, a C expression that produces the machine-specific code to
2837 setup the stack so that arbitrary frames can be accessed. For example,
2838 on the SPARC, we must flush all of the register windows to the stack
2839 before we can access arbitrary stack frames. You will seldom need to
2843 @deftypefn {Target Hook} bool TARGET_BUILTIN_SETJMP_FRAME_VALUE ()
2844 This target hook should return an rtx that is used to store
2845 the address of the current frame into the built in @code{setjmp} buffer.
2846 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2847 machines. One reason you may need to define this target hook is if
2848 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2851 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2852 A C expression whose value is RTL representing the value of the return
2853 address for the frame @var{count} steps up from the current frame, after
2854 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2855 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2856 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2858 The value of the expression must always be the correct address when
2859 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2860 determine the return address of other frames.
2863 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
2864 Define this if the return address of a particular stack frame is accessed
2865 from the frame pointer of the previous stack frame.
2868 @defmac INCOMING_RETURN_ADDR_RTX
2869 A C expression whose value is RTL representing the location of the
2870 incoming return address at the beginning of any function, before the
2871 prologue. This RTL is either a @code{REG}, indicating that the return
2872 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2875 You only need to define this macro if you want to support call frame
2876 debugging information like that provided by DWARF 2.
2878 If this RTL is a @code{REG}, you should also define
2879 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2882 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
2883 A C expression whose value is an integer giving a DWARF 2 column
2884 number that may be used as an alternate return column. This should
2885 be defined only if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
2886 general register, but an alternate column needs to be used for
2890 @defmac DWARF_ZERO_REG
2891 A C expression whose value is an integer giving a DWARF 2 register
2892 number that is considered to always have the value zero. This should
2893 only be defined if the target has an architected zero register, and
2894 someone decided it was a good idea to use that register number to
2895 terminate the stack backtrace. New ports should avoid this.
2898 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
2899 This target hook allows the backend to emit frame-related insns that
2900 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
2901 info engine will invoke it on insns of the form
2903 (set (reg) (unspec [...] UNSPEC_INDEX))
2907 (set (reg) (unspec_volatile [...] UNSPECV_INDEX)).
2909 to let the backend emit the call frame instructions. @var{label} is
2910 the CFI label attached to the insn, @var{pattern} is the pattern of
2911 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
2914 @defmac INCOMING_FRAME_SP_OFFSET
2915 A C expression whose value is an integer giving the offset, in bytes,
2916 from the value of the stack pointer register to the top of the stack
2917 frame at the beginning of any function, before the prologue. The top of
2918 the frame is defined to be the value of the stack pointer in the
2919 previous frame, just before the call instruction.
2921 You only need to define this macro if you want to support call frame
2922 debugging information like that provided by DWARF 2.
2925 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
2926 A C expression whose value is an integer giving the offset, in bytes,
2927 from the argument pointer to the canonical frame address (cfa). The
2928 final value should coincide with that calculated by
2929 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2930 during virtual register instantiation.
2932 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
2933 which is correct for most machines; in general, the arguments are found
2934 immediately before the stack frame. Note that this is not the case on
2935 some targets that save registers into the caller's frame, such as SPARC
2936 and rs6000, and so such targets need to define this macro.
2938 You only need to define this macro if the default is incorrect, and you
2939 want to support call frame debugging information like that provided by
2943 @node Exception Handling
2944 @subsection Exception Handling Support
2945 @cindex exception handling
2947 @defmac EH_RETURN_DATA_REGNO (@var{N})
2948 A C expression whose value is the @var{N}th register number used for
2949 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2950 @var{N} registers are usable.
2952 The exception handling library routines communicate with the exception
2953 handlers via a set of agreed upon registers. Ideally these registers
2954 should be call-clobbered; it is possible to use call-saved registers,
2955 but may negatively impact code size. The target must support at least
2956 2 data registers, but should define 4 if there are enough free registers.
2958 You must define this macro if you want to support call frame exception
2959 handling like that provided by DWARF 2.
2962 @defmac EH_RETURN_STACKADJ_RTX
2963 A C expression whose value is RTL representing a location in which
2964 to store a stack adjustment to be applied before function return.
2965 This is used to unwind the stack to an exception handler's call frame.
2966 It will be assigned zero on code paths that return normally.
2968 Typically this is a call-clobbered hard register that is otherwise
2969 untouched by the epilogue, but could also be a stack slot.
2971 Do not define this macro if the stack pointer is saved and restored
2972 by the regular prolog and epilog code in the call frame itself; in
2973 this case, the exception handling library routines will update the
2974 stack location to be restored in place. Otherwise, you must define
2975 this macro if you want to support call frame exception handling like
2976 that provided by DWARF 2.
2979 @defmac EH_RETURN_HANDLER_RTX
2980 A C expression whose value is RTL representing a location in which
2981 to store the address of an exception handler to which we should
2982 return. It will not be assigned on code paths that return normally.
2984 Typically this is the location in the call frame at which the normal
2985 return address is stored. For targets that return by popping an
2986 address off the stack, this might be a memory address just below
2987 the @emph{target} call frame rather than inside the current call
2988 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
2989 been assigned, so it may be used to calculate the location of the
2992 Some targets have more complex requirements than storing to an
2993 address calculable during initial code generation. In that case
2994 the @code{eh_return} instruction pattern should be used instead.
2996 If you want to support call frame exception handling, you must
2997 define either this macro or the @code{eh_return} instruction pattern.
3000 @defmac RETURN_ADDR_OFFSET
3001 If defined, an integer-valued C expression for which rtl will be generated
3002 to add it to the exception handler address before it is searched in the
3003 exception handling tables, and to subtract it again from the address before
3004 using it to return to the exception handler.
3007 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3008 This macro chooses the encoding of pointers embedded in the exception
3009 handling sections. If at all possible, this should be defined such
3010 that the exception handling section will not require dynamic relocations,
3011 and so may be read-only.
3013 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3014 @var{global} is true if the symbol may be affected by dynamic relocations.
3015 The macro should return a combination of the @code{DW_EH_PE_*} defines
3016 as found in @file{dwarf2.h}.
3018 If this macro is not defined, pointers will not be encoded but
3019 represented directly.
3022 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3023 This macro allows the target to emit whatever special magic is required
3024 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3025 Generic code takes care of pc-relative and indirect encodings; this must
3026 be defined if the target uses text-relative or data-relative encodings.
3028 This is a C statement that branches to @var{done} if the format was
3029 handled. @var{encoding} is the format chosen, @var{size} is the number
3030 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3034 @defmac MD_UNWIND_SUPPORT
3035 A string specifying a file to be #include'd in unwind-dw2.c. The file
3036 so included typically defines @code{MD_FALLBACK_FRAME_STATE_FOR}.
3039 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3040 This macro allows the target to add cpu and operating system specific
3041 code to the call-frame unwinder for use when there is no unwind data
3042 available. The most common reason to implement this macro is to unwind
3043 through signal frames.
3045 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
3046 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3047 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3048 for the address of the code being executed and @code{context->cfa} for
3049 the stack pointer value. If the frame can be decoded, the register save
3050 addresses should be updated in @var{fs} and the macro should evaluate to
3051 @code{_URC_NO_REASON}. If the frame cannot be decoded, the macro should
3052 evaluate to @code{_URC_END_OF_STACK}.
3054 For proper signal handling in Java this macro is accompanied by
3055 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3058 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3059 This macro allows the target to add operating system specific code to the
3060 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3061 usually used for signal or interrupt frames.
3063 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3064 @var{context} is an @code{_Unwind_Context};
3065 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3066 for the abi and context in the @code{.unwabi} directive. If the
3067 @code{.unwabi} directive can be handled, the register save addresses should
3068 be updated in @var{fs}.
3071 @defmac TARGET_USES_WEAK_UNWIND_INFO
3072 A C expression that evaluates to true if the target requires unwind
3073 info to be given comdat linkage. Define it to be @code{1} if comdat
3074 linkage is necessary. The default is @code{0}.
3077 @node Stack Checking
3078 @subsection Specifying How Stack Checking is Done
3080 GCC will check that stack references are within the boundaries of
3081 the stack, if the @option{-fstack-check} is specified, in one of three ways:
3085 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3086 will assume that you have arranged for stack checking to be done at
3087 appropriate places in the configuration files, e.g., in
3088 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
3092 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
3093 called @code{check_stack} in your @file{md} file, GCC will call that
3094 pattern with one argument which is the address to compare the stack
3095 value against. You must arrange for this pattern to report an error if
3096 the stack pointer is out of range.
3099 If neither of the above are true, GCC will generate code to periodically
3100 ``probe'' the stack pointer using the values of the macros defined below.
3103 Normally, you will use the default values of these macros, so GCC
3104 will use the third approach.
3106 @defmac STACK_CHECK_BUILTIN
3107 A nonzero value if stack checking is done by the configuration files in a
3108 machine-dependent manner. You should define this macro if stack checking
3109 is require by the ABI of your machine or if you would like to have to stack
3110 checking in some more efficient way than GCC's portable approach.
3111 The default value of this macro is zero.
3114 @defmac STACK_CHECK_PROBE_INTERVAL
3115 An integer representing the interval at which GCC must generate stack
3116 probe instructions. You will normally define this macro to be no larger
3117 than the size of the ``guard pages'' at the end of a stack area. The
3118 default value of 4096 is suitable for most systems.
3121 @defmac STACK_CHECK_PROBE_LOAD
3122 A integer which is nonzero if GCC should perform the stack probe
3123 as a load instruction and zero if GCC should use a store instruction.
3124 The default is zero, which is the most efficient choice on most systems.
3127 @defmac STACK_CHECK_PROTECT
3128 The number of bytes of stack needed to recover from a stack overflow,
3129 for languages where such a recovery is supported. The default value of
3130 75 words should be adequate for most machines.
3133 @defmac STACK_CHECK_MAX_FRAME_SIZE
3134 The maximum size of a stack frame, in bytes. GCC will generate probe
3135 instructions in non-leaf functions to ensure at least this many bytes of
3136 stack are available. If a stack frame is larger than this size, stack
3137 checking will not be reliable and GCC will issue a warning. The
3138 default is chosen so that GCC only generates one instruction on most
3139 systems. You should normally not change the default value of this macro.
3142 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3143 GCC uses this value to generate the above warning message. It
3144 represents the amount of fixed frame used by a function, not including
3145 space for any callee-saved registers, temporaries and user variables.
3146 You need only specify an upper bound for this amount and will normally
3147 use the default of four words.
3150 @defmac STACK_CHECK_MAX_VAR_SIZE
3151 The maximum size, in bytes, of an object that GCC will place in the
3152 fixed area of the stack frame when the user specifies
3153 @option{-fstack-check}.
3154 GCC computed the default from the values of the above macros and you will
3155 normally not need to override that default.
3159 @node Frame Registers
3160 @subsection Registers That Address the Stack Frame
3162 @c prevent bad page break with this line
3163 This discusses registers that address the stack frame.
3165 @defmac STACK_POINTER_REGNUM
3166 The register number of the stack pointer register, which must also be a
3167 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3168 the hardware determines which register this is.
3171 @defmac FRAME_POINTER_REGNUM
3172 The register number of the frame pointer register, which is used to
3173 access automatic variables in the stack frame. On some machines, the
3174 hardware determines which register this is. On other machines, you can
3175 choose any register you wish for this purpose.
3178 @defmac HARD_FRAME_POINTER_REGNUM
3179 On some machines the offset between the frame pointer and starting
3180 offset of the automatic variables is not known until after register
3181 allocation has been done (for example, because the saved registers are
3182 between these two locations). On those machines, define
3183 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3184 be used internally until the offset is known, and define
3185 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3186 used for the frame pointer.
3188 You should define this macro only in the very rare circumstances when it
3189 is not possible to calculate the offset between the frame pointer and
3190 the automatic variables until after register allocation has been
3191 completed. When this macro is defined, you must also indicate in your
3192 definition of @code{ELIMINABLE_REGS} how to eliminate
3193 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3194 or @code{STACK_POINTER_REGNUM}.
3196 Do not define this macro if it would be the same as
3197 @code{FRAME_POINTER_REGNUM}.
3200 @defmac ARG_POINTER_REGNUM
3201 The register number of the arg pointer register, which is used to access
3202 the function's argument list. On some machines, this is the same as the
3203 frame pointer register. On some machines, the hardware determines which
3204 register this is. On other machines, you can choose any register you
3205 wish for this purpose. If this is not the same register as the frame
3206 pointer register, then you must mark it as a fixed register according to
3207 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3208 (@pxref{Elimination}).
3211 @defmac RETURN_ADDRESS_POINTER_REGNUM
3212 The register number of the return address pointer register, which is used to
3213 access the current function's return address from the stack. On some
3214 machines, the return address is not at a fixed offset from the frame
3215 pointer or stack pointer or argument pointer. This register can be defined
3216 to point to the return address on the stack, and then be converted by
3217 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3219 Do not define this macro unless there is no other way to get the return
3220 address from the stack.
3223 @defmac STATIC_CHAIN_REGNUM
3224 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3225 Register numbers used for passing a function's static chain pointer. If
3226 register windows are used, the register number as seen by the called
3227 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3228 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3229 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3232 The static chain register need not be a fixed register.
3234 If the static chain is passed in memory, these macros should not be
3235 defined; instead, the next two macros should be defined.
3238 @defmac STATIC_CHAIN
3239 @defmacx STATIC_CHAIN_INCOMING
3240 If the static chain is passed in memory, these macros provide rtx giving
3241 @code{mem} expressions that denote where they are stored.
3242 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3243 as seen by the calling and called functions, respectively. Often the former
3244 will be at an offset from the stack pointer and the latter at an offset from
3247 @findex stack_pointer_rtx
3248 @findex frame_pointer_rtx
3249 @findex arg_pointer_rtx
3250 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3251 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3252 macros and should be used to refer to those items.
3254 If the static chain is passed in a register, the two previous macros should
3258 @defmac DWARF_FRAME_REGISTERS
3259 This macro specifies the maximum number of hard registers that can be
3260 saved in a call frame. This is used to size data structures used in
3261 DWARF2 exception handling.
3263 Prior to GCC 3.0, this macro was needed in order to establish a stable
3264 exception handling ABI in the face of adding new hard registers for ISA
3265 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3266 in the number of hard registers. Nevertheless, this macro can still be
3267 used to reduce the runtime memory requirements of the exception handling
3268 routines, which can be substantial if the ISA contains a lot of
3269 registers that are not call-saved.
3271 If this macro is not defined, it defaults to
3272 @code{FIRST_PSEUDO_REGISTER}.
3275 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3277 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3278 for backward compatibility in pre GCC 3.0 compiled code.
3280 If this macro is not defined, it defaults to
3281 @code{DWARF_FRAME_REGISTERS}.
3284 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3286 Define this macro if the target's representation for dwarf registers
3287 is different than the internal representation for unwind column.
3288 Given a dwarf register, this macro should return the internal unwind
3289 column number to use instead.
3291 See the PowerPC's SPE target for an example.
3294 @defmac DWARF_FRAME_REGNUM (@var{regno})
3296 Define this macro if the target's representation for dwarf registers
3297 used in .eh_frame or .debug_frame is different from that used in other
3298 debug info sections. Given a GCC hard register number, this macro
3299 should return the .eh_frame register number. The default is
3300 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3304 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3306 Define this macro to map register numbers held in the call frame info
3307 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3308 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3309 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3310 return @code{@var{regno}}.
3315 @subsection Eliminating Frame Pointer and Arg Pointer
3317 @c prevent bad page break with this line
3318 This is about eliminating the frame pointer and arg pointer.
3320 @defmac FRAME_POINTER_REQUIRED
3321 A C expression which is nonzero if a function must have and use a frame
3322 pointer. This expression is evaluated in the reload pass. If its value is
3323 nonzero the function will have a frame pointer.
3325 The expression can in principle examine the current function and decide
3326 according to the facts, but on most machines the constant 0 or the
3327 constant 1 suffices. Use 0 when the machine allows code to be generated
3328 with no frame pointer, and doing so saves some time or space. Use 1
3329 when there is no possible advantage to avoiding a frame pointer.
3331 In certain cases, the compiler does not know how to produce valid code
3332 without a frame pointer. The compiler recognizes those cases and
3333 automatically gives the function a frame pointer regardless of what
3334 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3337 In a function that does not require a frame pointer, the frame pointer
3338 register can be allocated for ordinary usage, unless you mark it as a
3339 fixed register. See @code{FIXED_REGISTERS} for more information.
3342 @findex get_frame_size
3343 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3344 A C statement to store in the variable @var{depth-var} the difference
3345 between the frame pointer and the stack pointer values immediately after
3346 the function prologue. The value would be computed from information
3347 such as the result of @code{get_frame_size ()} and the tables of
3348 registers @code{regs_ever_live} and @code{call_used_regs}.
3350 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3351 need not be defined. Otherwise, it must be defined even if
3352 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3353 case, you may set @var{depth-var} to anything.
3356 @defmac ELIMINABLE_REGS
3357 If defined, this macro specifies a table of register pairs used to
3358 eliminate unneeded registers that point into the stack frame. If it is not
3359 defined, the only elimination attempted by the compiler is to replace
3360 references to the frame pointer with references to the stack pointer.
3362 The definition of this macro is a list of structure initializations, each
3363 of which specifies an original and replacement register.
3365 On some machines, the position of the argument pointer is not known until
3366 the compilation is completed. In such a case, a separate hard register
3367 must be used for the argument pointer. This register can be eliminated by
3368 replacing it with either the frame pointer or the argument pointer,
3369 depending on whether or not the frame pointer has been eliminated.
3371 In this case, you might specify:
3373 #define ELIMINABLE_REGS \
3374 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3375 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3376 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3379 Note that the elimination of the argument pointer with the stack pointer is
3380 specified first since that is the preferred elimination.
3383 @defmac CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3384 A C expression that returns nonzero if the compiler is allowed to try
3385 to replace register number @var{from-reg} with register number
3386 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3387 is defined, and will usually be the constant 1, since most of the cases
3388 preventing register elimination are things that the compiler already
3392 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3393 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3394 specifies the initial difference between the specified pair of
3395 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3399 @node Stack Arguments
3400 @subsection Passing Function Arguments on the Stack
3401 @cindex arguments on stack
3402 @cindex stack arguments
3404 The macros in this section control how arguments are passed
3405 on the stack. See the following section for other macros that
3406 control passing certain arguments in registers.
3408 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (tree @var{fntype})
3409 This target hook returns @code{true} if an argument declared in a
3410 prototype as an integral type smaller than @code{int} should actually be
3411 passed as an @code{int}. In addition to avoiding errors in certain
3412 cases of mismatch, it also makes for better code on certain machines.
3413 The default is to not promote prototypes.
3417 A C expression. If nonzero, push insns will be used to pass
3419 If the target machine does not have a push instruction, set it to zero.
3420 That directs GCC to use an alternate strategy: to
3421 allocate the entire argument block and then store the arguments into
3422 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3425 @defmac PUSH_ARGS_REVERSED
3426 A C expression. If nonzero, function arguments will be evaluated from
3427 last to first, rather than from first to last. If this macro is not
3428 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3429 and args grow in opposite directions, and 0 otherwise.
3432 @defmac PUSH_ROUNDING (@var{npushed})
3433 A C expression that is the number of bytes actually pushed onto the
3434 stack when an instruction attempts to push @var{npushed} bytes.
3436 On some machines, the definition
3439 #define PUSH_ROUNDING(BYTES) (BYTES)
3443 will suffice. But on other machines, instructions that appear
3444 to push one byte actually push two bytes in an attempt to maintain
3445 alignment. Then the definition should be
3448 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3452 @findex current_function_outgoing_args_size
3453 @defmac ACCUMULATE_OUTGOING_ARGS
3454 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3455 will be computed and placed into the variable
3456 @code{current_function_outgoing_args_size}. No space will be pushed
3457 onto the stack for each call; instead, the function prologue should
3458 increase the stack frame size by this amount.
3460 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3464 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3465 Define this macro if functions should assume that stack space has been
3466 allocated for arguments even when their values are passed in
3469 The value of this macro is the size, in bytes, of the area reserved for
3470 arguments passed in registers for the function represented by @var{fndecl},
3471 which can be zero if GCC is calling a library function.
3473 This space can be allocated by the caller, or be a part of the
3474 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3477 @c above is overfull. not sure what to do. --mew 5feb93 did
3478 @c something, not sure if it looks good. --mew 10feb93
3480 @defmac OUTGOING_REG_PARM_STACK_SPACE
3481 Define this if it is the responsibility of the caller to allocate the area
3482 reserved for arguments passed in registers.
3484 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3485 whether the space for these arguments counts in the value of
3486 @code{current_function_outgoing_args_size}.
3489 @defmac STACK_PARMS_IN_REG_PARM_AREA
3490 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3491 stack parameters don't skip the area specified by it.
3492 @c i changed this, makes more sens and it should have taken care of the
3493 @c overfull.. not as specific, tho. --mew 5feb93
3495 Normally, when a parameter is not passed in registers, it is placed on the
3496 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3497 suppresses this behavior and causes the parameter to be passed on the
3498 stack in its natural location.
3501 @defmac RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3502 A C expression that should indicate the number of bytes of its own
3503 arguments that a function pops on returning, or 0 if the
3504 function pops no arguments and the caller must therefore pop them all
3505 after the function returns.
3507 @var{fundecl} is a C variable whose value is a tree node that describes
3508 the function in question. Normally it is a node of type
3509 @code{FUNCTION_DECL} that describes the declaration of the function.
3510 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3512 @var{funtype} is a C variable whose value is a tree node that
3513 describes the function in question. Normally it is a node of type
3514 @code{FUNCTION_TYPE} that describes the data type of the function.
3515 From this it is possible to obtain the data types of the value and
3516 arguments (if known).
3518 When a call to a library function is being considered, @var{fundecl}
3519 will contain an identifier node for the library function. Thus, if
3520 you need to distinguish among various library functions, you can do so
3521 by their names. Note that ``library function'' in this context means
3522 a function used to perform arithmetic, whose name is known specially
3523 in the compiler and was not mentioned in the C code being compiled.
3525 @var{stack-size} is the number of bytes of arguments passed on the
3526 stack. If a variable number of bytes is passed, it is zero, and
3527 argument popping will always be the responsibility of the calling function.
3529 On the VAX, all functions always pop their arguments, so the definition
3530 of this macro is @var{stack-size}. On the 68000, using the standard
3531 calling convention, no functions pop their arguments, so the value of
3532 the macro is always 0 in this case. But an alternative calling
3533 convention is available in which functions that take a fixed number of
3534 arguments pop them but other functions (such as @code{printf}) pop
3535 nothing (the caller pops all). When this convention is in use,
3536 @var{funtype} is examined to determine whether a function takes a fixed
3537 number of arguments.
3540 @defmac CALL_POPS_ARGS (@var{cum})
3541 A C expression that should indicate the number of bytes a call sequence
3542 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3543 when compiling a function call.
3545 @var{cum} is the variable in which all arguments to the called function
3546 have been accumulated.
3548 On certain architectures, such as the SH5, a call trampoline is used
3549 that pops certain registers off the stack, depending on the arguments
3550 that have been passed to the function. Since this is a property of the
3551 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3555 @node Register Arguments
3556 @subsection Passing Arguments in Registers
3557 @cindex arguments in registers
3558 @cindex registers arguments
3560 This section describes the macros which let you control how various
3561 types of arguments are passed in registers or how they are arranged in
3564 @defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3565 A C expression that controls whether a function argument is passed
3566 in a register, and which register.
3568 The arguments are @var{cum}, which summarizes all the previous
3569 arguments; @var{mode}, the machine mode of the argument; @var{type},
3570 the data type of the argument as a tree node or 0 if that is not known
3571 (which happens for C support library functions); and @var{named},
3572 which is 1 for an ordinary argument and 0 for nameless arguments that
3573 correspond to @samp{@dots{}} in the called function's prototype.
3574 @var{type} can be an incomplete type if a syntax error has previously
3577 The value of the expression is usually either a @code{reg} RTX for the
3578 hard register in which to pass the argument, or zero to pass the
3579 argument on the stack.
3581 For machines like the VAX and 68000, where normally all arguments are
3582 pushed, zero suffices as a definition.
3584 The value of the expression can also be a @code{parallel} RTX@. This is
3585 used when an argument is passed in multiple locations. The mode of the
3586 @code{parallel} should be the mode of the entire argument. The
3587 @code{parallel} holds any number of @code{expr_list} pairs; each one
3588 describes where part of the argument is passed. In each
3589 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3590 register in which to pass this part of the argument, and the mode of the
3591 register RTX indicates how large this part of the argument is. The
3592 second operand of the @code{expr_list} is a @code{const_int} which gives
3593 the offset in bytes into the entire argument of where this part starts.
3594 As a special exception the first @code{expr_list} in the @code{parallel}
3595 RTX may have a first operand of zero. This indicates that the entire
3596 argument is also stored on the stack.
3598 The last time this macro is called, it is called with @code{MODE ==
3599 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3600 pattern as operands 2 and 3 respectively.
3602 @cindex @file{stdarg.h} and register arguments
3603 The usual way to make the ISO library @file{stdarg.h} work on a machine
3604 where some arguments are usually passed in registers, is to cause
3605 nameless arguments to be passed on the stack instead. This is done
3606 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3608 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3609 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3610 You may use the hook @code{targetm.calls.must_pass_in_stack}
3611 in the definition of this macro to determine if this argument is of a
3612 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3613 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3614 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3615 defined, the argument will be computed in the stack and then loaded into
3619 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, tree @var{type})
3620 This target hook should return @code{true} if we should not pass @var{type}
3621 solely in registers. The file @file{expr.h} defines a
3622 definition that is usually appropriate, refer to @file{expr.h} for additional
3626 @defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3627 Define this macro if the target machine has ``register windows'', so
3628 that the register in which a function sees an arguments is not
3629 necessarily the same as the one in which the caller passed the
3632 For such machines, @code{FUNCTION_ARG} computes the register in which
3633 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3634 be defined in a similar fashion to tell the function being called
3635 where the arguments will arrive.
3637 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3638 serves both purposes.
3641 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3642 This target hook returns the number of bytes at the beginning of an
3643 argument that must be put in registers. The value must be zero for
3644 arguments that are passed entirely in registers or that are entirely
3645 pushed on the stack.
3647 On some machines, certain arguments must be passed partially in
3648 registers and partially in memory. On these machines, typically the
3649 first few words of arguments are passed in registers, and the rest
3650 on the stack. If a multi-word argument (a @code{double} or a
3651 structure) crosses that boundary, its first few words must be passed
3652 in registers and the rest must be pushed. This macro tells the
3653 compiler when this occurs, and how many bytes should go in registers.
3655 @code{FUNCTION_ARG} for these arguments should return the first
3656 register to be used by the caller for this argument; likewise
3657 @code{FUNCTION_INCOMING_ARG}, for the called function.
3660 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3661 This target hook should return @code{true} if an argument at the
3662 position indicated by @var{cum} should be passed by reference. This
3663 predicate is queried after target independent reasons for being
3664 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
3666 If the hook returns true, a copy of that argument is made in memory and a
3667 pointer to the argument is passed instead of the argument itself.
3668 The pointer is passed in whatever way is appropriate for passing a pointer
3672 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3673 The function argument described by the parameters to this hook is
3674 known to be passed by reference. The hook should return true if the
3675 function argument should be copied by the callee instead of copied
3678 For any argument for which the hook returns true, if it can be
3679 determined that the argument is not modified, then a copy need
3682 The default version of this hook always returns false.
3685 @defmac CUMULATIVE_ARGS
3686 A C type for declaring a variable that is used as the first argument of
3687 @code{FUNCTION_ARG} and other related values. For some target machines,
3688 the type @code{int} suffices and can hold the number of bytes of
3691 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3692 arguments that have been passed on the stack. The compiler has other
3693 variables to keep track of that. For target machines on which all
3694 arguments are passed on the stack, there is no need to store anything in
3695 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3696 should not be empty, so use @code{int}.
3699 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
3700 A C statement (sans semicolon) for initializing the variable
3701 @var{cum} for the state at the beginning of the argument list. The
3702 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
3703 is the tree node for the data type of the function which will receive
3704 the args, or 0 if the args are to a compiler support library function.
3705 For direct calls that are not libcalls, @var{fndecl} contain the
3706 declaration node of the function. @var{fndecl} is also set when
3707 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3708 being compiled. @var{n_named_args} is set to the number of named
3709 arguments, including a structure return address if it is passed as a
3710 parameter, when making a call. When processing incoming arguments,
3711 @var{n_named_args} is set to @minus{}1.
3713 When processing a call to a compiler support library function,
3714 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3715 contains the name of the function, as a string. @var{libname} is 0 when
3716 an ordinary C function call is being processed. Thus, each time this
3717 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3718 never both of them at once.
3721 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3722 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3723 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3724 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3725 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3726 0)} is used instead.
3729 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3730 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3731 finding the arguments for the function being compiled. If this macro is
3732 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3734 The value passed for @var{libname} is always 0, since library routines
3735 with special calling conventions are never compiled with GCC@. The
3736 argument @var{libname} exists for symmetry with
3737 @code{INIT_CUMULATIVE_ARGS}.
3738 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3739 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3742 @defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3743 A C statement (sans semicolon) to update the summarizer variable
3744 @var{cum} to advance past an argument in the argument list. The
3745 values @var{mode}, @var{type} and @var{named} describe that argument.
3746 Once this is done, the variable @var{cum} is suitable for analyzing
3747 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
3749 This macro need not do anything if the argument in question was passed
3750 on the stack. The compiler knows how to track the amount of stack space
3751 used for arguments without any special help.
3754 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3755 If defined, a C expression which determines whether, and in which direction,
3756 to pad out an argument with extra space. The value should be of type
3757 @code{enum direction}: either @code{upward} to pad above the argument,
3758 @code{downward} to pad below, or @code{none} to inhibit padding.
3760 The @emph{amount} of padding is always just enough to reach the next
3761 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3764 This macro has a default definition which is right for most systems.
3765 For little-endian machines, the default is to pad upward. For
3766 big-endian machines, the default is to pad downward for an argument of
3767 constant size shorter than an @code{int}, and upward otherwise.
3770 @defmac PAD_VARARGS_DOWN
3771 If defined, a C expression which determines whether the default
3772 implementation of va_arg will attempt to pad down before reading the
3773 next argument, if that argument is smaller than its aligned space as
3774 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3775 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3778 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
3779 Specify padding for the last element of a block move between registers and
3780 memory. @var{first} is nonzero if this is the only element. Defining this
3781 macro allows better control of register function parameters on big-endian
3782 machines, without using @code{PARALLEL} rtl. In particular,
3783 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
3784 registers, as there is no longer a "wrong" part of a register; For example,
3785 a three byte aggregate may be passed in the high part of a register if so
3789 @defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3790 If defined, a C expression that gives the alignment boundary, in bits,
3791 of an argument with the specified mode and type. If it is not defined,
3792 @code{PARM_BOUNDARY} is used for all arguments.
3795 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
3796 A C expression that is nonzero if @var{regno} is the number of a hard
3797 register in which function arguments are sometimes passed. This does
3798 @emph{not} include implicit arguments such as the static chain and
3799 the structure-value address. On many machines, no registers can be
3800 used for this purpose since all function arguments are pushed on the
3804 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (tree @var{type})
3805 This hook should return true if parameter of type @var{type} are passed
3806 as two scalar parameters. By default, GCC will attempt to pack complex
3807 arguments into the target's word size. Some ABIs require complex arguments
3808 to be split and treated as their individual components. For example, on
3809 AIX64, complex floats should be passed in a pair of floating point
3810 registers, even though a complex float would fit in one 64-bit floating
3813 The default value of this hook is @code{NULL}, which is treated as always
3817 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
3818 This hook returns a type node for @code{va_list} for the target.
3819 The default version of the hook returns @code{void*}.
3822 @deftypefn {Target Hook} tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree @var{valist}, tree @var{type}, tree *@var{pre_p}, tree *@var{post_p})
3823 This hook performs target-specific gimplification of
3824 @code{VA_ARG_EXPR}. The first two parameters correspond to the
3825 arguments to @code{va_arg}; the latter two are as in
3826 @code{gimplify.c:gimplify_expr}.
3829 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (enum machine_mode @var{mode})
3830 Define this to return nonzero if the port can handle pointers
3831 with machine mode @var{mode}. The default version of this
3832 hook returns true for both @code{ptr_mode} and @code{Pmode}.
3835 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
3836 Define this to return nonzero if the port is prepared to handle
3837 insns involving scalar mode @var{mode}. For a scalar mode to be
3838 considered supported, all the basic arithmetic and comparisons
3841 The default version of this hook returns true for any mode
3842 required to handle the basic C types (as defined by the port).
3843 Included here are the double-word arithmetic supported by the
3844 code in @file{optabs.c}.
3847 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
3848 Define this to return nonzero if the port is prepared to handle
3849 insns involving vector mode @var{mode}. At the very least, it
3850 must have move patterns for this mode.
3854 @subsection How Scalar Function Values Are Returned
3855 @cindex return values in registers
3856 @cindex values, returned by functions
3857 @cindex scalars, returned as values
3859 This section discusses the macros that control returning scalars as
3860 values---values that can fit in registers.
3862 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
3863 A C expression to create an RTX representing the place where a
3864 function returns a value of data type @var{valtype}. @var{valtype} is
3865 a tree node representing a data type. Write @code{TYPE_MODE
3866 (@var{valtype})} to get the machine mode used to represent that type.
3867 On many machines, only the mode is relevant. (Actually, on most
3868 machines, scalar values are returned in the same place regardless of
3871 The value of the expression is usually a @code{reg} RTX for the hard
3872 register where the return value is stored. The value can also be a
3873 @code{parallel} RTX, if the return value is in multiple places. See
3874 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3876 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply the same
3877 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3880 If the precise function being called is known, @var{func} is a tree
3881 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3882 pointer. This makes it possible to use a different value-returning
3883 convention for specific functions when all their calls are
3886 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3887 types, because these are returned in another way. See
3888 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
3891 @defmac FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3892 Define this macro if the target machine has ``register windows''
3893 so that the register in which a function returns its value is not
3894 the same as the one in which the caller sees the value.
3896 For such machines, @code{FUNCTION_VALUE} computes the register in which
3897 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3898 defined in a similar fashion to tell the function where to put the
3901 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3902 @code{FUNCTION_VALUE} serves both purposes.
3904 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3905 aggregate data types, because these are returned in another way. See
3906 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
3909 @defmac LIBCALL_VALUE (@var{mode})
3910 A C expression to create an RTX representing the place where a library
3911 function returns a value of mode @var{mode}. If the precise function
3912 being called is known, @var{func} is a tree node
3913 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3914 pointer. This makes it possible to use a different value-returning
3915 convention for specific functions when all their calls are
3918 Note that ``library function'' in this context means a compiler
3919 support routine, used to perform arithmetic, whose name is known
3920 specially by the compiler and was not mentioned in the C code being
3923 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3924 data types, because none of the library functions returns such types.
3927 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
3928 A C expression that is nonzero if @var{regno} is the number of a hard
3929 register in which the values of called function may come back.
3931 A register whose use for returning values is limited to serving as the
3932 second of a pair (for a value of type @code{double}, say) need not be
3933 recognized by this macro. So for most machines, this definition
3937 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3940 If the machine has register windows, so that the caller and the called
3941 function use different registers for the return value, this macro
3942 should recognize only the caller's register numbers.
3945 @defmac APPLY_RESULT_SIZE
3946 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3947 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3948 saving and restoring an arbitrary return value.
3951 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (tree @var{type})
3952 This hook should return true if values of type @var{type} are returned
3953 at the most significant end of a register (in other words, if they are
3954 padded at the least significant end). You can assume that @var{type}
3955 is returned in a register; the caller is required to check this.
3957 Note that the register provided by @code{FUNCTION_VALUE} must be able
3958 to hold the complete return value. For example, if a 1-, 2- or 3-byte
3959 structure is returned at the most significant end of a 4-byte register,
3960 @code{FUNCTION_VALUE} should provide an @code{SImode} rtx.
3963 @node Aggregate Return
3964 @subsection How Large Values Are Returned
3965 @cindex aggregates as return values
3966 @cindex large return values
3967 @cindex returning aggregate values
3968 @cindex structure value address
3970 When a function value's mode is @code{BLKmode} (and in some other
3971 cases), the value is not returned according to @code{FUNCTION_VALUE}
3972 (@pxref{Scalar Return}). Instead, the caller passes the address of a
3973 block of memory in which the value should be stored. This address
3974 is called the @dfn{structure value address}.
3976 This section describes how to control returning structure values in
3979 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (tree @var{type}, tree @var{fntype})
3980 This target hook should return a nonzero value to say to return the
3981 function value in memory, just as large structures are always returned.
3982 Here @var{type} will be the data type of the value, and @var{fntype}
3983 will be the type of the function doing the returning, or @code{NULL} for
3986 Note that values of mode @code{BLKmode} must be explicitly handled
3987 by this function. Also, the option @option{-fpcc-struct-return}
3988 takes effect regardless of this macro. On most systems, it is
3989 possible to leave the hook undefined; this causes a default
3990 definition to be used, whose value is the constant 1 for @code{BLKmode}
3991 values, and 0 otherwise.
3993 Do not use this hook to indicate that structures and unions should always
3994 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
3998 @defmac DEFAULT_PCC_STRUCT_RETURN
3999 Define this macro to be 1 if all structure and union return values must be
4000 in memory. Since this results in slower code, this should be defined
4001 only if needed for compatibility with other compilers or with an ABI@.
4002 If you define this macro to be 0, then the conventions used for structure
4003 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4006 If not defined, this defaults to the value 1.
4009 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4010 This target hook should return the location of the structure value
4011 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4012 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4013 be @code{NULL}, for libcalls. You do not need to define this target
4014 hook if the address is always passed as an ``invisible'' first
4017 On some architectures the place where the structure value address
4018 is found by the called function is not the same place that the
4019 caller put it. This can be due to register windows, or it could
4020 be because the function prologue moves it to a different place.
4021 @var{incoming} is @code{true} when the location is needed in
4022 the context of the called function, and @code{false} in the context of
4025 If @var{incoming} is @code{true} and the address is to be found on the
4026 stack, return a @code{mem} which refers to the frame pointer.
4029 @defmac PCC_STATIC_STRUCT_RETURN
4030 Define this macro if the usual system convention on the target machine
4031 for returning structures and unions is for the called function to return
4032 the address of a static variable containing the value.
4034 Do not define this if the usual system convention is for the caller to
4035 pass an address to the subroutine.
4037 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4038 nothing when you use @option{-freg-struct-return} mode.
4042 @subsection Caller-Saves Register Allocation
4044 If you enable it, GCC can save registers around function calls. This
4045 makes it possible to use call-clobbered registers to hold variables that
4046 must live across calls.
4048 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4049 A C expression to determine whether it is worthwhile to consider placing
4050 a pseudo-register in a call-clobbered hard register and saving and
4051 restoring it around each function call. The expression should be 1 when
4052 this is worth doing, and 0 otherwise.
4054 If you don't define this macro, a default is used which is good on most
4055 machines: @code{4 * @var{calls} < @var{refs}}.
4058 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4059 A C expression specifying which mode is required for saving @var{nregs}
4060 of a pseudo-register in call-clobbered hard register @var{regno}. If
4061 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4062 returned. For most machines this macro need not be defined since GCC
4063 will select the smallest suitable mode.
4066 @node Function Entry
4067 @subsection Function Entry and Exit
4068 @cindex function entry and exit
4072 This section describes the macros that output function entry
4073 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4075 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4076 If defined, a function that outputs the assembler code for entry to a
4077 function. The prologue is responsible for setting up the stack frame,
4078 initializing the frame pointer register, saving registers that must be
4079 saved, and allocating @var{size} additional bytes of storage for the
4080 local variables. @var{size} is an integer. @var{file} is a stdio
4081 stream to which the assembler code should be output.
4083 The label for the beginning of the function need not be output by this
4084 macro. That has already been done when the macro is run.
4086 @findex regs_ever_live
4087 To determine which registers to save, the macro can refer to the array
4088 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4089 @var{r} is used anywhere within the function. This implies the function
4090 prologue should save register @var{r}, provided it is not one of the
4091 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4092 @code{regs_ever_live}.)
4094 On machines that have ``register windows'', the function entry code does
4095 not save on the stack the registers that are in the windows, even if
4096 they are supposed to be preserved by function calls; instead it takes
4097 appropriate steps to ``push'' the register stack, if any non-call-used
4098 registers are used in the function.
4100 @findex frame_pointer_needed
4101 On machines where functions may or may not have frame-pointers, the
4102 function entry code must vary accordingly; it must set up the frame
4103 pointer if one is wanted, and not otherwise. To determine whether a
4104 frame pointer is in wanted, the macro can refer to the variable
4105 @code{frame_pointer_needed}. The variable's value will be 1 at run
4106 time in a function that needs a frame pointer. @xref{Elimination}.
4108 The function entry code is responsible for allocating any stack space
4109 required for the function. This stack space consists of the regions
4110 listed below. In most cases, these regions are allocated in the
4111 order listed, with the last listed region closest to the top of the
4112 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4113 the highest address if it is not defined). You can use a different order
4114 for a machine if doing so is more convenient or required for
4115 compatibility reasons. Except in cases where required by standard
4116 or by a debugger, there is no reason why the stack layout used by GCC
4117 need agree with that used by other compilers for a machine.
4120 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4121 If defined, a function that outputs assembler code at the end of a
4122 prologue. This should be used when the function prologue is being
4123 emitted as RTL, and you have some extra assembler that needs to be
4124 emitted. @xref{prologue instruction pattern}.
4127 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4128 If defined, a function that outputs assembler code at the start of an
4129 epilogue. This should be used when the function epilogue is being
4130 emitted as RTL, and you have some extra assembler that needs to be
4131 emitted. @xref{epilogue instruction pattern}.
4134 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4135 If defined, a function that outputs the assembler code for exit from a
4136 function. The epilogue is responsible for restoring the saved
4137 registers and stack pointer to their values when the function was
4138 called, and returning control to the caller. This macro takes the
4139 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4140 registers to restore are determined from @code{regs_ever_live} and
4141 @code{CALL_USED_REGISTERS} in the same way.
4143 On some machines, there is a single instruction that does all the work
4144 of returning from the function. On these machines, give that
4145 instruction the name @samp{return} and do not define the macro
4146 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4148 Do not define a pattern named @samp{return} if you want the
4149 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4150 switches to control whether return instructions or epilogues are used,
4151 define a @samp{return} pattern with a validity condition that tests the
4152 target switches appropriately. If the @samp{return} pattern's validity
4153 condition is false, epilogues will be used.
4155 On machines where functions may or may not have frame-pointers, the
4156 function exit code must vary accordingly. Sometimes the code for these
4157 two cases is completely different. To determine whether a frame pointer
4158 is wanted, the macro can refer to the variable
4159 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4160 a function that needs a frame pointer.
4162 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4163 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4164 The C variable @code{current_function_is_leaf} is nonzero for such a
4165 function. @xref{Leaf Functions}.
4167 On some machines, some functions pop their arguments on exit while
4168 others leave that for the caller to do. For example, the 68020 when
4169 given @option{-mrtd} pops arguments in functions that take a fixed
4170 number of arguments.
4172 @findex current_function_pops_args
4173 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4174 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4175 needs to know what was decided. The variable that is called
4176 @code{current_function_pops_args} is the number of bytes of its
4177 arguments that a function should pop. @xref{Scalar Return}.
4178 @c what is the "its arguments" in the above sentence referring to, pray
4179 @c tell? --mew 5feb93
4184 @findex current_function_pretend_args_size
4185 A region of @code{current_function_pretend_args_size} bytes of
4186 uninitialized space just underneath the first argument arriving on the
4187 stack. (This may not be at the very start of the allocated stack region
4188 if the calling sequence has pushed anything else since pushing the stack
4189 arguments. But usually, on such machines, nothing else has been pushed
4190 yet, because the function prologue itself does all the pushing.) This
4191 region is used on machines where an argument may be passed partly in
4192 registers and partly in memory, and, in some cases to support the
4193 features in @code{<stdarg.h>}.
4196 An area of memory used to save certain registers used by the function.
4197 The size of this area, which may also include space for such things as
4198 the return address and pointers to previous stack frames, is
4199 machine-specific and usually depends on which registers have been used
4200 in the function. Machines with register windows often do not require
4204 A region of at least @var{size} bytes, possibly rounded up to an allocation
4205 boundary, to contain the local variables of the function. On some machines,
4206 this region and the save area may occur in the opposite order, with the
4207 save area closer to the top of the stack.
4210 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4211 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4212 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4213 argument lists of the function. @xref{Stack Arguments}.
4216 @defmac EXIT_IGNORE_STACK
4217 Define this macro as a C expression that is nonzero if the return
4218 instruction or the function epilogue ignores the value of the stack
4219 pointer; in other words, if it is safe to delete an instruction to
4220 adjust the stack pointer before a return from the function. The
4223 Note that this macro's value is relevant only for functions for which
4224 frame pointers are maintained. It is never safe to delete a final
4225 stack adjustment in a function that has no frame pointer, and the
4226 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4229 @defmac EPILOGUE_USES (@var{regno})
4230 Define this macro as a C expression that is nonzero for registers that are
4231 used by the epilogue or the @samp{return} pattern. The stack and frame
4232 pointer registers are already be assumed to be used as needed.
4235 @defmac EH_USES (@var{regno})
4236 Define this macro as a C expression that is nonzero for registers that are
4237 used by the exception handling mechanism, and so should be considered live
4238 on entry to an exception edge.
4241 @defmac DELAY_SLOTS_FOR_EPILOGUE
4242 Define this macro if the function epilogue contains delay slots to which
4243 instructions from the rest of the function can be ``moved''. The
4244 definition should be a C expression whose value is an integer
4245 representing the number of delay slots there.
4248 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4249 A C expression that returns 1 if @var{insn} can be placed in delay
4250 slot number @var{n} of the epilogue.
4252 The argument @var{n} is an integer which identifies the delay slot now
4253 being considered (since different slots may have different rules of
4254 eligibility). It is never negative and is always less than the number
4255 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4256 If you reject a particular insn for a given delay slot, in principle, it
4257 may be reconsidered for a subsequent delay slot. Also, other insns may
4258 (at least in principle) be considered for the so far unfilled delay
4261 @findex current_function_epilogue_delay_list
4262 @findex final_scan_insn
4263 The insns accepted to fill the epilogue delay slots are put in an RTL
4264 list made with @code{insn_list} objects, stored in the variable
4265 @code{current_function_epilogue_delay_list}. The insn for the first
4266 delay slot comes first in the list. Your definition of the macro
4267 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4268 outputting the insns in this list, usually by calling
4269 @code{final_scan_insn}.
4271 You need not define this macro if you did not define
4272 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4275 @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})
4276 A function that outputs the assembler code for a thunk
4277 function, used to implement C++ virtual function calls with multiple
4278 inheritance. The thunk acts as a wrapper around a virtual function,
4279 adjusting the implicit object parameter before handing control off to
4282 First, emit code to add the integer @var{delta} to the location that
4283 contains the incoming first argument. Assume that this argument
4284 contains a pointer, and is the one used to pass the @code{this} pointer
4285 in C++. This is the incoming argument @emph{before} the function prologue,
4286 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4287 all other incoming arguments.
4289 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4290 made after adding @code{delta}. In particular, if @var{p} is the
4291 adjusted pointer, the following adjustment should be made:
4294 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4297 After the additions, emit code to jump to @var{function}, which is a
4298 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4299 not touch the return address. Hence returning from @var{FUNCTION} will
4300 return to whoever called the current @samp{thunk}.
4302 The effect must be as if @var{function} had been called directly with
4303 the adjusted first argument. This macro is responsible for emitting all
4304 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4305 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4307 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4308 have already been extracted from it.) It might possibly be useful on
4309 some targets, but probably not.
4311 If you do not define this macro, the target-independent code in the C++
4312 front end will generate a less efficient heavyweight thunk that calls
4313 @var{function} instead of jumping to it. The generic approach does
4314 not support varargs.
4317 @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})
4318 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4319 to output the assembler code for the thunk function specified by the
4320 arguments it is passed, and false otherwise. In the latter case, the
4321 generic approach will be used by the C++ front end, with the limitations
4326 @subsection Generating Code for Profiling
4327 @cindex profiling, code generation
4329 These macros will help you generate code for profiling.
4331 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4332 A C statement or compound statement to output to @var{file} some
4333 assembler code to call the profiling subroutine @code{mcount}.
4336 The details of how @code{mcount} expects to be called are determined by
4337 your operating system environment, not by GCC@. To figure them out,
4338 compile a small program for profiling using the system's installed C
4339 compiler and look at the assembler code that results.
4341 Older implementations of @code{mcount} expect the address of a counter
4342 variable to be loaded into some register. The name of this variable is
4343 @samp{LP} followed by the number @var{labelno}, so you would generate
4344 the name using @samp{LP%d} in a @code{fprintf}.
4347 @defmac PROFILE_HOOK
4348 A C statement or compound statement to output to @var{file} some assembly
4349 code to call the profiling subroutine @code{mcount} even the target does
4350 not support profiling.
4353 @defmac NO_PROFILE_COUNTERS
4354 Define this macro if the @code{mcount} subroutine on your system does
4355 not need a counter variable allocated for each function. This is true
4356 for almost all modern implementations. If you define this macro, you
4357 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
4360 @defmac PROFILE_BEFORE_PROLOGUE
4361 Define this macro if the code for function profiling should come before
4362 the function prologue. Normally, the profiling code comes after.
4366 @subsection Permitting tail calls
4369 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4370 True if it is ok to do sibling call optimization for the specified
4371 call expression @var{exp}. @var{decl} will be the called function,
4372 or @code{NULL} if this is an indirect call.
4374 It is not uncommon for limitations of calling conventions to prevent
4375 tail calls to functions outside the current unit of translation, or
4376 during PIC compilation. The hook is used to enforce these restrictions,
4377 as the @code{sibcall} md pattern can not fail, or fall over to a
4378 ``normal'' call. The criteria for successful sibling call optimization
4379 may vary greatly between different architectures.
4383 @section Implementing the Varargs Macros
4384 @cindex varargs implementation
4386 GCC comes with an implementation of @code{<varargs.h>} and
4387 @code{<stdarg.h>} that work without change on machines that pass arguments
4388 on the stack. Other machines require their own implementations of
4389 varargs, and the two machine independent header files must have
4390 conditionals to include it.
4392 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4393 the calling convention for @code{va_start}. The traditional
4394 implementation takes just one argument, which is the variable in which
4395 to store the argument pointer. The ISO implementation of
4396 @code{va_start} takes an additional second argument. The user is
4397 supposed to write the last named argument of the function here.
4399 However, @code{va_start} should not use this argument. The way to find
4400 the end of the named arguments is with the built-in functions described
4403 @defmac __builtin_saveregs ()
4404 Use this built-in function to save the argument registers in memory so
4405 that the varargs mechanism can access them. Both ISO and traditional
4406 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4407 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4409 On some machines, @code{__builtin_saveregs} is open-coded under the
4410 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4411 other machines, it calls a routine written in assembler language,
4412 found in @file{libgcc2.c}.
4414 Code generated for the call to @code{__builtin_saveregs} appears at the
4415 beginning of the function, as opposed to where the call to
4416 @code{__builtin_saveregs} is written, regardless of what the code is.
4417 This is because the registers must be saved before the function starts
4418 to use them for its own purposes.
4419 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4423 @defmac __builtin_args_info (@var{category})
4424 Use this built-in function to find the first anonymous arguments in
4427 In general, a machine may have several categories of registers used for
4428 arguments, each for a particular category of data types. (For example,
4429 on some machines, floating-point registers are used for floating-point
4430 arguments while other arguments are passed in the general registers.)
4431 To make non-varargs functions use the proper calling convention, you
4432 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4433 registers in each category have been used so far
4435 @code{__builtin_args_info} accesses the same data structure of type
4436 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4437 with it, with @var{category} specifying which word to access. Thus, the
4438 value indicates the first unused register in a given category.
4440 Normally, you would use @code{__builtin_args_info} in the implementation
4441 of @code{va_start}, accessing each category just once and storing the
4442 value in the @code{va_list} object. This is because @code{va_list} will
4443 have to update the values, and there is no way to alter the
4444 values accessed by @code{__builtin_args_info}.
4447 @defmac __builtin_next_arg (@var{lastarg})
4448 This is the equivalent of @code{__builtin_args_info}, for stack
4449 arguments. It returns the address of the first anonymous stack
4450 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4451 returns the address of the location above the first anonymous stack
4452 argument. Use it in @code{va_start} to initialize the pointer for
4453 fetching arguments from the stack. Also use it in @code{va_start} to
4454 verify that the second parameter @var{lastarg} is the last named argument
4455 of the current function.
4458 @defmac __builtin_classify_type (@var{object})
4459 Since each machine has its own conventions for which data types are
4460 passed in which kind of register, your implementation of @code{va_arg}
4461 has to embody these conventions. The easiest way to categorize the
4462 specified data type is to use @code{__builtin_classify_type} together
4463 with @code{sizeof} and @code{__alignof__}.
4465 @code{__builtin_classify_type} ignores the value of @var{object},
4466 considering only its data type. It returns an integer describing what
4467 kind of type that is---integer, floating, pointer, structure, and so on.
4469 The file @file{typeclass.h} defines an enumeration that you can use to
4470 interpret the values of @code{__builtin_classify_type}.
4473 These machine description macros help implement varargs:
4475 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
4476 If defined, this hook produces the machine-specific code for a call to
4477 @code{__builtin_saveregs}. This code will be moved to the very
4478 beginning of the function, before any parameter access are made. The
4479 return value of this function should be an RTX that contains the value
4480 to use as the return of @code{__builtin_saveregs}.
4483 @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})
4484 This target hook offers an alternative to using
4485 @code{__builtin_saveregs} and defining the hook
4486 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
4487 register arguments into the stack so that all the arguments appear to
4488 have been passed consecutively on the stack. Once this is done, you can
4489 use the standard implementation of varargs that works for machines that
4490 pass all their arguments on the stack.
4492 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
4493 structure, containing the values that are obtained after processing the
4494 named arguments. The arguments @var{mode} and @var{type} describe the
4495 last named argument---its machine mode and its data type as a tree node.
4497 The target hook should do two things: first, push onto the stack all the
4498 argument registers @emph{not} used for the named arguments, and second,
4499 store the size of the data thus pushed into the @code{int}-valued
4500 variable pointed to by @var{pretend_args_size}. The value that you
4501 store here will serve as additional offset for setting up the stack
4504 Because you must generate code to push the anonymous arguments at
4505 compile time without knowing their data types,
4506 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
4507 have just a single category of argument register and use it uniformly
4510 If the argument @var{second_time} is nonzero, it means that the
4511 arguments of the function are being analyzed for the second time. This
4512 happens for an inline function, which is not actually compiled until the
4513 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
4514 not generate any instructions in this case.
4517 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca})
4518 Define this hook to return @code{true} if the location where a function
4519 argument is passed depends on whether or not it is a named argument.
4521 This hook controls how the @var{named} argument to @code{FUNCTION_ARG}
4522 is set for varargs and stdarg functions. If this hook returns
4523 @code{true}, the @var{named} argument is always true for named
4524 arguments, and false for unnamed arguments. If it returns @code{false},
4525 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
4526 then all arguments are treated as named. Otherwise, all named arguments
4527 except the last are treated as named.
4529 You need not define this hook if it always returns zero.
4532 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED
4533 If you need to conditionally change ABIs so that one works with
4534 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
4535 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
4536 defined, then define this hook to return @code{true} if
4537 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
4538 Otherwise, you should not define this hook.
4542 @section Trampolines for Nested Functions
4543 @cindex trampolines for nested functions
4544 @cindex nested functions, trampolines for
4546 A @dfn{trampoline} is a small piece of code that is created at run time
4547 when the address of a nested function is taken. It normally resides on
4548 the stack, in the stack frame of the containing function. These macros
4549 tell GCC how to generate code to allocate and initialize a
4552 The instructions in the trampoline must do two things: load a constant
4553 address into the static chain register, and jump to the real address of
4554 the nested function. On CISC machines such as the m68k, this requires
4555 two instructions, a move immediate and a jump. Then the two addresses
4556 exist in the trampoline as word-long immediate operands. On RISC
4557 machines, it is often necessary to load each address into a register in
4558 two parts. Then pieces of each address form separate immediate
4561 The code generated to initialize the trampoline must store the variable
4562 parts---the static chain value and the function address---into the
4563 immediate operands of the instructions. On a CISC machine, this is
4564 simply a matter of copying each address to a memory reference at the
4565 proper offset from the start of the trampoline. On a RISC machine, it
4566 may be necessary to take out pieces of the address and store them
4569 @defmac TRAMPOLINE_TEMPLATE (@var{file})
4570 A C statement to output, on the stream @var{file}, assembler code for a
4571 block of data that contains the constant parts of a trampoline. This
4572 code should not include a label---the label is taken care of
4575 If you do not define this macro, it means no template is needed
4576 for the target. Do not define this macro on systems where the block move
4577 code to copy the trampoline into place would be larger than the code
4578 to generate it on the spot.
4581 @defmac TRAMPOLINE_SECTION
4582 The name of a subroutine to switch to the section in which the
4583 trampoline template is to be placed (@pxref{Sections}). The default is
4584 a value of @samp{readonly_data_section}, which places the trampoline in
4585 the section containing read-only data.
4588 @defmac TRAMPOLINE_SIZE
4589 A C expression for the size in bytes of the trampoline, as an integer.
4592 @defmac TRAMPOLINE_ALIGNMENT
4593 Alignment required for trampolines, in bits.
4595 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4596 is used for aligning trampolines.
4599 @defmac INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4600 A C statement to initialize the variable parts of a trampoline.
4601 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4602 an RTX for the address of the nested function; @var{static_chain} is an
4603 RTX for the static chain value that should be passed to the function
4607 @defmac TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4608 A C statement that should perform any machine-specific adjustment in
4609 the address of the trampoline. Its argument contains the address that
4610 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4611 used for a function call should be different from the address in which
4612 the template was stored, the different address should be assigned to
4613 @var{addr}. If this macro is not defined, @var{addr} will be used for
4616 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4617 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4618 If this macro is not defined, by default the trampoline is allocated as
4619 a stack slot. This default is right for most machines. The exceptions
4620 are machines where it is impossible to execute instructions in the stack
4621 area. On such machines, you may have to implement a separate stack,
4622 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4623 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4625 @var{fp} points to a data structure, a @code{struct function}, which
4626 describes the compilation status of the immediate containing function of
4627 the function which the trampoline is for. The stack slot for the
4628 trampoline is in the stack frame of this containing function. Other
4629 allocation strategies probably must do something analogous with this
4633 Implementing trampolines is difficult on many machines because they have
4634 separate instruction and data caches. Writing into a stack location
4635 fails to clear the memory in the instruction cache, so when the program
4636 jumps to that location, it executes the old contents.
4638 Here are two possible solutions. One is to clear the relevant parts of
4639 the instruction cache whenever a trampoline is set up. The other is to
4640 make all trampolines identical, by having them jump to a standard
4641 subroutine. The former technique makes trampoline execution faster; the
4642 latter makes initialization faster.
4644 To clear the instruction cache when a trampoline is initialized, define
4645 the following macro.
4647 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
4648 If defined, expands to a C expression clearing the @emph{instruction
4649 cache} in the specified interval. The definition of this macro would
4650 typically be a series of @code{asm} statements. Both @var{beg} and
4651 @var{end} are both pointer expressions.
4654 The operating system may also require the stack to be made executable
4655 before calling the trampoline. To implement this requirement, define
4656 the following macro.
4658 @defmac ENABLE_EXECUTE_STACK
4659 Define this macro if certain operations must be performed before executing
4660 code located on the stack. The macro should expand to a series of C
4661 file-scope constructs (e.g.@: functions) and provide a unique entry point
4662 named @code{__enable_execute_stack}. The target is responsible for
4663 emitting calls to the entry point in the code, for example from the
4664 @code{INITIALIZE_TRAMPOLINE} macro.
4667 To use a standard subroutine, define the following macro. In addition,
4668 you must make sure that the instructions in a trampoline fill an entire
4669 cache line with identical instructions, or else ensure that the
4670 beginning of the trampoline code is always aligned at the same point in
4671 its cache line. Look in @file{m68k.h} as a guide.
4673 @defmac TRANSFER_FROM_TRAMPOLINE
4674 Define this macro if trampolines need a special subroutine to do their
4675 work. The macro should expand to a series of @code{asm} statements
4676 which will be compiled with GCC@. They go in a library function named
4677 @code{__transfer_from_trampoline}.
4679 If you need to avoid executing the ordinary prologue code of a compiled
4680 C function when you jump to the subroutine, you can do so by placing a
4681 special label of your own in the assembler code. Use one @code{asm}
4682 statement to generate an assembler label, and another to make the label
4683 global. Then trampolines can use that label to jump directly to your
4684 special assembler code.
4688 @section Implicit Calls to Library Routines
4689 @cindex library subroutine names
4690 @cindex @file{libgcc.a}
4692 @c prevent bad page break with this line
4693 Here is an explanation of implicit calls to library routines.
4695 @defmac DECLARE_LIBRARY_RENAMES
4696 This macro, if defined, should expand to a piece of C code that will get
4697 expanded when compiling functions for libgcc.a. It can be used to
4698 provide alternate names for GCC's internal library functions if there
4699 are ABI-mandated names that the compiler should provide.
4702 @findex init_one_libfunc
4703 @findex set_optab_libfunc
4704 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
4705 This hook should declare additional library routines or rename
4706 existing ones, using the functions @code{set_optab_libfunc} and
4707 @code{init_one_libfunc} defined in @file{optabs.c}.
4708 @code{init_optabs} calls this macro after initializing all the normal
4711 The default is to do nothing. Most ports don't need to define this hook.
4714 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4715 This macro should return @code{true} if the library routine that
4716 implements the floating point comparison operator @var{comparison} in
4717 mode @var{mode} will return a boolean, and @var{false} if it will
4720 GCC's own floating point libraries return tristates from the
4721 comparison operators, so the default returns false always. Most ports
4722 don't need to define this macro.
4725 @defmac TARGET_LIB_INT_CMP_BIASED
4726 This macro should evaluate to @code{true} if the integer comparison
4727 functions (like @code{__cmpdi2}) return 0 to indicate that the first
4728 operand is smaller than the second, 1 to indicate that they are equal,
4729 and 2 to indicate that the first operand is greater than the second.
4730 If this macro evaluates to @code{false} the comparison functions return
4731 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
4732 in @file{libgcc.a}, you do not need to define this macro.
4735 @cindex US Software GOFAST, floating point emulation library
4736 @cindex floating point emulation library, US Software GOFAST
4737 @cindex GOFAST, floating point emulation library
4738 @findex gofast_maybe_init_libfuncs
4739 @defmac US_SOFTWARE_GOFAST
4740 Define this macro if your system C library uses the US Software GOFAST
4741 library to provide floating point emulation.
4743 In addition to defining this macro, your architecture must set
4744 @code{TARGET_INIT_LIBFUNCS} to @code{gofast_maybe_init_libfuncs}, or
4745 else call that function from its version of that hook. It is defined
4746 in @file{config/gofast.h}, which must be included by your
4747 architecture's @file{@var{cpu}.c} file. See @file{sparc/sparc.c} for
4750 If this macro is defined, the
4751 @code{TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL} target hook must return
4752 false for @code{SFmode} and @code{DFmode} comparisons.
4755 @cindex @code{EDOM}, implicit usage
4758 The value of @code{EDOM} on the target machine, as a C integer constant
4759 expression. If you don't define this macro, GCC does not attempt to
4760 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4761 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4764 If you do not define @code{TARGET_EDOM}, then compiled code reports
4765 domain errors by calling the library function and letting it report the
4766 error. If mathematical functions on your system use @code{matherr} when
4767 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4768 that @code{matherr} is used normally.
4771 @cindex @code{errno}, implicit usage
4772 @defmac GEN_ERRNO_RTX
4773 Define this macro as a C expression to create an rtl expression that
4774 refers to the global ``variable'' @code{errno}. (On certain systems,
4775 @code{errno} may not actually be a variable.) If you don't define this
4776 macro, a reasonable default is used.
4779 @cindex C99 math functions, implicit usage
4780 @defmac TARGET_C99_FUNCTIONS
4781 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
4782 @code{sinf} and similarly for other functions defined by C99 standard. The
4783 default is nonzero that should be proper value for most modern systems, however
4784 number of existing systems lacks support for these functions in the runtime so
4785 they needs this macro to be redefined to 0.
4788 @defmac NEXT_OBJC_RUNTIME
4789 Define this macro to generate code for Objective-C message sending using
4790 the calling convention of the NeXT system. This calling convention
4791 involves passing the object, the selector and the method arguments all
4792 at once to the method-lookup library function.
4794 The default calling convention passes just the object and the selector
4795 to the lookup function, which returns a pointer to the method.
4798 @node Addressing Modes
4799 @section Addressing Modes
4800 @cindex addressing modes
4802 @c prevent bad page break with this line
4803 This is about addressing modes.
4805 @defmac HAVE_PRE_INCREMENT
4806 @defmacx HAVE_PRE_DECREMENT
4807 @defmacx HAVE_POST_INCREMENT
4808 @defmacx HAVE_POST_DECREMENT
4809 A C expression that is nonzero if the machine supports pre-increment,
4810 pre-decrement, post-increment, or post-decrement addressing respectively.
4813 @defmac HAVE_PRE_MODIFY_DISP
4814 @defmacx HAVE_POST_MODIFY_DISP
4815 A C expression that is nonzero if the machine supports pre- or
4816 post-address side-effect generation involving constants other than
4817 the size of the memory operand.
4820 @defmac HAVE_PRE_MODIFY_REG
4821 @defmacx HAVE_POST_MODIFY_REG
4822 A C expression that is nonzero if the machine supports pre- or
4823 post-address side-effect generation involving a register displacement.
4826 @defmac CONSTANT_ADDRESS_P (@var{x})
4827 A C expression that is 1 if the RTX @var{x} is a constant which
4828 is a valid address. On most machines, this can be defined as
4829 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4830 in which constant addresses are supported.
4833 @defmac CONSTANT_P (@var{x})
4834 @code{CONSTANT_P}, which is defined by target-independent code,
4835 accepts integer-values expressions whose values are not explicitly
4836 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
4837 expressions and @code{const} arithmetic expressions, in addition to
4838 @code{const_int} and @code{const_double} expressions.
4841 @defmac MAX_REGS_PER_ADDRESS
4842 A number, the maximum number of registers that can appear in a valid
4843 memory address. Note that it is up to you to specify a value equal to
4844 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4848 @defmac GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4849 A C compound statement with a conditional @code{goto @var{label};}
4850 executed if @var{x} (an RTX) is a legitimate memory address on the
4851 target machine for a memory operand of mode @var{mode}.
4853 It usually pays to define several simpler macros to serve as
4854 subroutines for this one. Otherwise it may be too complicated to
4857 This macro must exist in two variants: a strict variant and a
4858 non-strict one. The strict variant is used in the reload pass. It
4859 must be defined so that any pseudo-register that has not been
4860 allocated a hard register is considered a memory reference. In
4861 contexts where some kind of register is required, a pseudo-register
4862 with no hard register must be rejected.
4864 The non-strict variant is used in other passes. It must be defined to
4865 accept all pseudo-registers in every context where some kind of
4866 register is required.
4868 @findex REG_OK_STRICT
4869 Compiler source files that want to use the strict variant of this
4870 macro define the macro @code{REG_OK_STRICT}. You should use an
4871 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4872 in that case and the non-strict variant otherwise.
4874 Subroutines to check for acceptable registers for various purposes (one
4875 for base registers, one for index registers, and so on) are typically
4876 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4877 Then only these subroutine macros need have two variants; the higher
4878 levels of macros may be the same whether strict or not.
4880 Normally, constant addresses which are the sum of a @code{symbol_ref}
4881 and an integer are stored inside a @code{const} RTX to mark them as
4882 constant. Therefore, there is no need to recognize such sums
4883 specifically as legitimate addresses. Normally you would simply
4884 recognize any @code{const} as legitimate.
4886 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4887 sums that are not marked with @code{const}. It assumes that a naked
4888 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4889 naked constant sums as illegitimate addresses, so that none of them will
4890 be given to @code{PRINT_OPERAND_ADDRESS}.
4892 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
4893 On some machines, whether a symbolic address is legitimate depends on
4894 the section that the address refers to. On these machines, define the
4895 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
4896 into the @code{symbol_ref}, and then check for it here. When you see a
4897 @code{const}, you will have to look inside it to find the
4898 @code{symbol_ref} in order to determine the section. @xref{Assembler
4902 @defmac REG_OK_FOR_BASE_P (@var{x})
4903 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4904 RTX) is valid for use as a base register. For hard registers, it
4905 should always accept those which the hardware permits and reject the
4906 others. Whether the macro accepts or rejects pseudo registers must be
4907 controlled by @code{REG_OK_STRICT} as described above. This usually
4908 requires two variant definitions, of which @code{REG_OK_STRICT}
4909 controls the one actually used.
4912 @defmac REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4913 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4914 that expression may examine the mode of the memory reference in
4915 @var{mode}. You should define this macro if the mode of the memory
4916 reference affects whether a register may be used as a base register. If
4917 you define this macro, the compiler will use it instead of
4918 @code{REG_OK_FOR_BASE_P}.
4921 @defmac REG_MODE_OK_FOR_REG_BASE_P (@var{x}, @var{mode})
4922 A C expression which is nonzero if @var{x} (assumed to be a @code{reg} RTX)
4923 is suitable for use as a base register in base plus index operand addresses,
4924 accessing memory in mode @var{mode}. It may be either a suitable hard
4925 register or a pseudo register that has been allocated such a hard register.
4926 You should define this macro if base plus index addresses have different
4927 requirements than other base register uses.
4930 @defmac REG_OK_FOR_INDEX_P (@var{x})
4931 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4932 RTX) is valid for use as an index register.
4934 The difference between an index register and a base register is that
4935 the index register may be scaled. If an address involves the sum of
4936 two registers, neither one of them scaled, then either one may be
4937 labeled the ``base'' and the other the ``index''; but whichever
4938 labeling is used must fit the machine's constraints of which registers
4939 may serve in each capacity. The compiler will try both labelings,
4940 looking for one that is valid, and will reload one or both registers
4941 only if neither labeling works.
4944 @defmac FIND_BASE_TERM (@var{x})
4945 A C expression to determine the base term of address @var{x}.
4946 This macro is used in only one place: `find_base_term' in alias.c.
4948 It is always safe for this macro to not be defined. It exists so
4949 that alias analysis can understand machine-dependent addresses.
4951 The typical use of this macro is to handle addresses containing
4952 a label_ref or symbol_ref within an UNSPEC@.
4955 @defmac LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4956 A C compound statement that attempts to replace @var{x} with a valid
4957 memory address for an operand of mode @var{mode}. @var{win} will be a
4958 C statement label elsewhere in the code; the macro definition may use
4961 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4965 to avoid further processing if the address has become legitimate.
4967 @findex break_out_memory_refs
4968 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4969 and @var{oldx} will be the operand that was given to that function to produce
4972 The code generated by this macro should not alter the substructure of
4973 @var{x}. If it transforms @var{x} into a more legitimate form, it
4974 should assign @var{x} (which will always be a C variable) a new value.
4976 It is not necessary for this macro to come up with a legitimate
4977 address. The compiler has standard ways of doing so in all cases. In
4978 fact, it is safe to omit this macro. But often a
4979 machine-dependent strategy can generate better code.
4982 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4983 A C compound statement that attempts to replace @var{x}, which is an address
4984 that needs reloading, with a valid memory address for an operand of mode
4985 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4986 It is not necessary to define this macro, but it might be useful for
4987 performance reasons.
4989 For example, on the i386, it is sometimes possible to use a single
4990 reload register instead of two by reloading a sum of two pseudo
4991 registers into a register. On the other hand, for number of RISC
4992 processors offsets are limited so that often an intermediate address
4993 needs to be generated in order to address a stack slot. By defining
4994 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
4995 generated for adjacent some stack slots can be made identical, and thus
4998 @emph{Note}: This macro should be used with caution. It is necessary
4999 to know something of how reload works in order to effectively use this,
5000 and it is quite easy to produce macros that build in too much knowledge
5001 of reload internals.
5003 @emph{Note}: This macro must be able to reload an address created by a
5004 previous invocation of this macro. If it fails to handle such addresses
5005 then the compiler may generate incorrect code or abort.
5008 The macro definition should use @code{push_reload} to indicate parts that
5009 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5010 suitable to be passed unaltered to @code{push_reload}.
5012 The code generated by this macro must not alter the substructure of
5013 @var{x}. If it transforms @var{x} into a more legitimate form, it
5014 should assign @var{x} (which will always be a C variable) a new value.
5015 This also applies to parts that you change indirectly by calling
5018 @findex strict_memory_address_p
5019 The macro definition may use @code{strict_memory_address_p} to test if
5020 the address has become legitimate.
5023 If you want to change only a part of @var{x}, one standard way of doing
5024 this is to use @code{copy_rtx}. Note, however, that is unshares only a
5025 single level of rtl. Thus, if the part to be changed is not at the
5026 top level, you'll need to replace first the top level.
5027 It is not necessary for this macro to come up with a legitimate
5028 address; but often a machine-dependent strategy can generate better code.
5031 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5032 A C statement or compound statement with a conditional @code{goto
5033 @var{label};} executed if memory address @var{x} (an RTX) can have
5034 different meanings depending on the machine mode of the memory
5035 reference it is used for or if the address is valid for some modes
5038 Autoincrement and autodecrement addresses typically have mode-dependent
5039 effects because the amount of the increment or decrement is the size
5040 of the operand being addressed. Some machines have other mode-dependent
5041 addresses. Many RISC machines have no mode-dependent addresses.
5043 You may assume that @var{addr} is a valid address for the machine.
5046 @defmac LEGITIMATE_CONSTANT_P (@var{x})
5047 A C expression that is nonzero if @var{x} is a legitimate constant for
5048 an immediate operand on the target machine. You can assume that
5049 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5050 @samp{1} is a suitable definition for this macro on machines where
5051 anything @code{CONSTANT_P} is valid.
5054 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5055 This hook is used to undo the possibly obfuscating effects of the
5056 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5057 macros. Some backend implementations of these macros wrap symbol
5058 references inside an @code{UNSPEC} rtx to represent PIC or similar
5059 addressing modes. This target hook allows GCC's optimizers to understand
5060 the semantics of these opaque @code{UNSPEC}s by converting them back
5061 into their original form.
5064 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (rtx @var{x})
5065 This hook should return true if @var{x} is of a form that cannot (or
5066 should not) be spilled to the constant pool. The default version of
5067 this hook returns false.
5069 The primary reason to define this hook is to prevent reload from
5070 deciding that a non-legitimate constant would be better reloaded
5071 from the constant pool instead of spilling and reloading a register
5072 holding the constant. This restriction is often true of addresses
5073 of TLS symbols for various targets.
5076 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5077 This hook should return the DECL of a function @var{f} that given an
5078 address @var{addr} as an argument returns a mask @var{m} that can be
5079 used to extract from two vectors the relevant data that resides in
5080 @var{addr} in case @var{addr} is not properly aligned.
5082 The autovectrizer, when vectorizing a load operation from an address
5083 @var{addr} that may be unaligned, will generate two vector loads from
5084 the two aligned addresses around @var{addr}. It then generates a
5085 @code{REALIGN_LOAD} operation to extract the relevant data from the
5086 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5087 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5088 the third argument, @var{OFF}, defines how the data will be extracted
5089 from these two vectors: if @var{OFF} is 0, then the returned vector is
5090 @var{v2}; otherwise, the returned vector is composed from the last
5091 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5092 @var{OFF} elements of @var{v2}.
5094 If this hook is defined, the autovectorizer will generate a call
5095 to @var{f} (using the DECL tree that this hook returns) and will
5096 use the return value of @var{f} as the argument @var{OFF} to
5097 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5098 should comply with the semantics expected by @code{REALIGN_LOAD}
5100 If this hook is not defined, then @var{addr} will be used as
5101 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5102 log2(@var{VS})-1 bits of @var{addr} will be considered.
5105 @node Condition Code
5106 @section Condition Code Status
5107 @cindex condition code status
5109 @c prevent bad page break with this line
5110 This describes the condition code status.
5113 The file @file{conditions.h} defines a variable @code{cc_status} to
5114 describe how the condition code was computed (in case the interpretation of
5115 the condition code depends on the instruction that it was set by). This
5116 variable contains the RTL expressions on which the condition code is
5117 currently based, and several standard flags.
5119 Sometimes additional machine-specific flags must be defined in the machine
5120 description header file. It can also add additional machine-specific
5121 information by defining @code{CC_STATUS_MDEP}.
5123 @defmac CC_STATUS_MDEP
5124 C code for a data type which is used for declaring the @code{mdep}
5125 component of @code{cc_status}. It defaults to @code{int}.
5127 This macro is not used on machines that do not use @code{cc0}.
5130 @defmac CC_STATUS_MDEP_INIT
5131 A C expression to initialize the @code{mdep} field to ``empty''.
5132 The default definition does nothing, since most machines don't use
5133 the field anyway. If you want to use the field, you should probably
5134 define this macro to initialize it.
5136 This macro is not used on machines that do not use @code{cc0}.
5139 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5140 A C compound statement to set the components of @code{cc_status}
5141 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5142 this macro's responsibility to recognize insns that set the condition
5143 code as a byproduct of other activity as well as those that explicitly
5146 This macro is not used on machines that do not use @code{cc0}.
5148 If there are insns that do not set the condition code but do alter
5149 other machine registers, this macro must check to see whether they
5150 invalidate the expressions that the condition code is recorded as
5151 reflecting. For example, on the 68000, insns that store in address
5152 registers do not set the condition code, which means that usually
5153 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5154 insns. But suppose that the previous insn set the condition code
5155 based on location @samp{a4@@(102)} and the current insn stores a new
5156 value in @samp{a4}. Although the condition code is not changed by
5157 this, it will no longer be true that it reflects the contents of
5158 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5159 @code{cc_status} in this case to say that nothing is known about the
5160 condition code value.
5162 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5163 with the results of peephole optimization: insns whose patterns are
5164 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5165 constants which are just the operands. The RTL structure of these
5166 insns is not sufficient to indicate what the insns actually do. What
5167 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5168 @code{CC_STATUS_INIT}.
5170 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5171 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5172 @samp{cc}. This avoids having detailed information about patterns in
5173 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5176 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5177 Returns a mode from class @code{MODE_CC} to be used when comparison
5178 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5179 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5180 @pxref{Jump Patterns} for a description of the reason for this
5184 #define SELECT_CC_MODE(OP,X,Y) \
5185 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5186 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5187 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5188 || GET_CODE (X) == NEG) \
5189 ? CC_NOOVmode : CCmode))
5192 You should define this macro if and only if you define extra CC modes
5193 in @file{@var{machine}-modes.def}.
5196 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5197 On some machines not all possible comparisons are defined, but you can
5198 convert an invalid comparison into a valid one. For example, the Alpha
5199 does not have a @code{GT} comparison, but you can use an @code{LT}
5200 comparison instead and swap the order of the operands.
5202 On such machines, define this macro to be a C statement to do any
5203 required conversions. @var{code} is the initial comparison code
5204 and @var{op0} and @var{op1} are the left and right operands of the
5205 comparison, respectively. You should modify @var{code}, @var{op0}, and
5206 @var{op1} as required.
5208 GCC will not assume that the comparison resulting from this macro is
5209 valid but will see if the resulting insn matches a pattern in the
5212 You need not define this macro if it would never change the comparison
5216 @defmac REVERSIBLE_CC_MODE (@var{mode})
5217 A C expression whose value is one if it is always safe to reverse a
5218 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5219 can ever return @var{mode} for a floating-point inequality comparison,
5220 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5222 You need not define this macro if it would always returns zero or if the
5223 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5224 For example, here is the definition used on the SPARC, where floating-point
5225 inequality comparisons are always given @code{CCFPEmode}:
5228 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5232 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
5233 A C expression whose value is reversed condition code of the @var{code} for
5234 comparison done in CC_MODE @var{mode}. The macro is used only in case
5235 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5236 machine has some non-standard way how to reverse certain conditionals. For
5237 instance in case all floating point conditions are non-trapping, compiler may
5238 freely convert unordered compares to ordered one. Then definition may look
5242 #define REVERSE_CONDITION(CODE, MODE) \
5243 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5244 : reverse_condition_maybe_unordered (CODE))
5248 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
5249 A C expression that returns true if the conditional execution predicate
5250 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
5251 versa. Define this to return 0 if the target has conditional execution
5252 predicates that cannot be reversed safely. There is no need to validate
5253 that the arguments of op1 and op2 are the same, this is done separately.
5254 If no expansion is specified, this macro is defined as follows:
5257 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5258 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
5262 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *, unsigned int *)
5263 On targets which do not use @code{(cc0)}, and which use a hard
5264 register rather than a pseudo-register to hold condition codes, the
5265 regular CSE passes are often not able to identify cases in which the
5266 hard register is set to a common value. Use this hook to enable a
5267 small pass which optimizes such cases. This hook should return true
5268 to enable this pass, and it should set the integers to which its
5269 arguments point to the hard register numbers used for condition codes.
5270 When there is only one such register, as is true on most systems, the
5271 integer pointed to by the second argument should be set to
5272 @code{INVALID_REGNUM}.
5274 The default version of this hook returns false.
5277 @deftypefn {Target Hook} enum machine_mode TARGET_CC_MODES_COMPATIBLE (enum machine_mode, enum machine_mode)
5278 On targets which use multiple condition code modes in class
5279 @code{MODE_CC}, it is sometimes the case that a comparison can be
5280 validly done in more than one mode. On such a system, define this
5281 target hook to take two mode arguments and to return a mode in which
5282 both comparisons may be validly done. If there is no such mode,
5283 return @code{VOIDmode}.
5285 The default version of this hook checks whether the modes are the
5286 same. If they are, it returns that mode. If they are different, it
5287 returns @code{VOIDmode}.
5291 @section Describing Relative Costs of Operations
5292 @cindex costs of instructions
5293 @cindex relative costs
5294 @cindex speed of instructions
5296 These macros let you describe the relative speed of various operations
5297 on the target machine.
5299 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5300 A C expression for the cost of moving data of mode @var{mode} from a
5301 register in class @var{from} to one in class @var{to}. The classes are
5302 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5303 value of 2 is the default; other values are interpreted relative to
5306 It is not required that the cost always equal 2 when @var{from} is the
5307 same as @var{to}; on some machines it is expensive to move between
5308 registers if they are not general registers.
5310 If reload sees an insn consisting of a single @code{set} between two
5311 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5312 classes returns a value of 2, reload does not check to ensure that the
5313 constraints of the insn are met. Setting a cost of other than 2 will
5314 allow reload to verify that the constraints are met. You should do this
5315 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5318 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5319 A C expression for the cost of moving data of mode @var{mode} between a
5320 register of class @var{class} and memory; @var{in} is zero if the value
5321 is to be written to memory, nonzero if it is to be read in. This cost
5322 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5323 registers and memory is more expensive than between two registers, you
5324 should define this macro to express the relative cost.
5326 If you do not define this macro, GCC uses a default cost of 4 plus
5327 the cost of copying via a secondary reload register, if one is
5328 needed. If your machine requires a secondary reload register to copy
5329 between memory and a register of @var{class} but the reload mechanism is
5330 more complex than copying via an intermediate, define this macro to
5331 reflect the actual cost of the move.
5333 GCC defines the function @code{memory_move_secondary_cost} if
5334 secondary reloads are needed. It computes the costs due to copying via
5335 a secondary register. If your machine copies from memory using a
5336 secondary register in the conventional way but the default base value of
5337 4 is not correct for your machine, define this macro to add some other
5338 value to the result of that function. The arguments to that function
5339 are the same as to this macro.
5343 A C expression for the cost of a branch instruction. A value of 1 is
5344 the default; other values are interpreted relative to that.
5347 Here are additional macros which do not specify precise relative costs,
5348 but only that certain actions are more expensive than GCC would
5351 @defmac SLOW_BYTE_ACCESS
5352 Define this macro as a C expression which is nonzero if accessing less
5353 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5354 faster than accessing a word of memory, i.e., if such access
5355 require more than one instruction or if there is no difference in cost
5356 between byte and (aligned) word loads.
5358 When this macro is not defined, the compiler will access a field by
5359 finding the smallest containing object; when it is defined, a fullword
5360 load will be used if alignment permits. Unless bytes accesses are
5361 faster than word accesses, using word accesses is preferable since it
5362 may eliminate subsequent memory access if subsequent accesses occur to
5363 other fields in the same word of the structure, but to different bytes.
5366 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5367 Define this macro to be the value 1 if memory accesses described by the
5368 @var{mode} and @var{alignment} parameters have a cost many times greater
5369 than aligned accesses, for example if they are emulated in a trap
5372 When this macro is nonzero, the compiler will act as if
5373 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5374 moves. This can cause significantly more instructions to be produced.
5375 Therefore, do not set this macro nonzero if unaligned accesses only add a
5376 cycle or two to the time for a memory access.
5378 If the value of this macro is always zero, it need not be defined. If
5379 this macro is defined, it should produce a nonzero value when
5380 @code{STRICT_ALIGNMENT} is nonzero.
5384 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5385 which a sequence of insns should be generated instead of a
5386 string move insn or a library call. Increasing the value will always
5387 make code faster, but eventually incurs high cost in increased code size.
5389 Note that on machines where the corresponding move insn is a
5390 @code{define_expand} that emits a sequence of insns, this macro counts
5391 the number of such sequences.
5393 If you don't define this, a reasonable default is used.
5396 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5397 A C expression used to determine whether @code{move_by_pieces} will be used to
5398 copy a chunk of memory, or whether some other block move mechanism
5399 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5400 than @code{MOVE_RATIO}.
5403 @defmac MOVE_MAX_PIECES
5404 A C expression used by @code{move_by_pieces} to determine the largest unit
5405 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5409 The threshold of number of scalar move insns, @emph{below} which a sequence
5410 of insns should be generated to clear memory instead of a string clear insn
5411 or a library call. Increasing the value will always make code faster, but
5412 eventually incurs high cost in increased code size.
5414 If you don't define this, a reasonable default is used.
5417 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5418 A C expression used to determine whether @code{clear_by_pieces} will be used
5419 to clear a chunk of memory, or whether some other block clear mechanism
5420 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5421 than @code{CLEAR_RATIO}.
5424 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
5425 A C expression used to determine whether @code{store_by_pieces} will be
5426 used to set a chunk of memory to a constant value, or whether some other
5427 mechanism will be used. Used by @code{__builtin_memset} when storing
5428 values other than constant zero and by @code{__builtin_strcpy} when
5429 when called with a constant source string.
5430 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5431 than @code{MOVE_RATIO}.
5434 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
5435 A C expression used to determine whether a load postincrement is a good
5436 thing to use for a given mode. Defaults to the value of
5437 @code{HAVE_POST_INCREMENT}.
5440 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
5441 A C expression used to determine whether a load postdecrement is a good
5442 thing to use for a given mode. Defaults to the value of
5443 @code{HAVE_POST_DECREMENT}.
5446 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
5447 A C expression used to determine whether a load preincrement is a good
5448 thing to use for a given mode. Defaults to the value of
5449 @code{HAVE_PRE_INCREMENT}.
5452 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
5453 A C expression used to determine whether a load predecrement is a good
5454 thing to use for a given mode. Defaults to the value of
5455 @code{HAVE_PRE_DECREMENT}.
5458 @defmac USE_STORE_POST_INCREMENT (@var{mode})
5459 A C expression used to determine whether a store postincrement is a good
5460 thing to use for a given mode. Defaults to the value of
5461 @code{HAVE_POST_INCREMENT}.
5464 @defmac USE_STORE_POST_DECREMENT (@var{mode})
5465 A C expression used to determine whether a store postdecrement is a good
5466 thing to use for a given mode. Defaults to the value of
5467 @code{HAVE_POST_DECREMENT}.
5470 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
5471 This macro is used to determine whether a store preincrement is a good
5472 thing to use for a given mode. Defaults to the value of
5473 @code{HAVE_PRE_INCREMENT}.
5476 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
5477 This macro is used to determine whether a store predecrement is a good
5478 thing to use for a given mode. Defaults to the value of
5479 @code{HAVE_PRE_DECREMENT}.
5482 @defmac NO_FUNCTION_CSE
5483 Define this macro if it is as good or better to call a constant
5484 function address than to call an address kept in a register.
5487 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
5488 Define this macro if a non-short-circuit operation produced by
5489 @samp{fold_range_test ()} is optimal. This macro defaults to true if
5490 @code{BRANCH_COST} is greater than or equal to the value 2.
5493 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total})
5494 This target hook describes the relative costs of RTL expressions.
5496 The cost may depend on the precise form of the expression, which is
5497 available for examination in @var{x}, and the rtx code of the expression
5498 in which it is contained, found in @var{outer_code}. @var{code} is the
5499 expression code---redundant, since it can be obtained with
5500 @code{GET_CODE (@var{x})}.
5502 In implementing this hook, you can use the construct
5503 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
5506 On entry to the hook, @code{*@var{total}} contains a default estimate
5507 for the cost of the expression. The hook should modify this value as
5508 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
5509 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
5510 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
5512 When optimizing for code size, i.e.@: when @code{optimize_size} is
5513 nonzero, this target hook should be used to estimate the relative
5514 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
5516 The hook returns true when all subexpressions of @var{x} have been
5517 processed, and false when @code{rtx_cost} should recurse.
5520 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address})
5521 This hook computes the cost of an addressing mode that contains
5522 @var{address}. If not defined, the cost is computed from
5523 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
5525 For most CISC machines, the default cost is a good approximation of the
5526 true cost of the addressing mode. However, on RISC machines, all
5527 instructions normally have the same length and execution time. Hence
5528 all addresses will have equal costs.
5530 In cases where more than one form of an address is known, the form with
5531 the lowest cost will be used. If multiple forms have the same, lowest,
5532 cost, the one that is the most complex will be used.
5534 For example, suppose an address that is equal to the sum of a register
5535 and a constant is used twice in the same basic block. When this macro
5536 is not defined, the address will be computed in a register and memory
5537 references will be indirect through that register. On machines where
5538 the cost of the addressing mode containing the sum is no higher than
5539 that of a simple indirect reference, this will produce an additional
5540 instruction and possibly require an additional register. Proper
5541 specification of this macro eliminates this overhead for such machines.
5543 This hook is never called with an invalid address.
5545 On machines where an address involving more than one register is as
5546 cheap as an address computation involving only one register, defining
5547 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
5548 be live over a region of code where only one would have been if
5549 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
5550 should be considered in the definition of this macro. Equivalent costs
5551 should probably only be given to addresses with different numbers of
5552 registers on machines with lots of registers.
5556 @section Adjusting the Instruction Scheduler
5558 The instruction scheduler may need a fair amount of machine-specific
5559 adjustment in order to produce good code. GCC provides several target
5560 hooks for this purpose. It is usually enough to define just a few of
5561 them: try the first ones in this list first.
5563 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
5564 This hook returns the maximum number of instructions that can ever
5565 issue at the same time on the target machine. The default is one.
5566 Although the insn scheduler can define itself the possibility of issue
5567 an insn on the same cycle, the value can serve as an additional
5568 constraint to issue insns on the same simulated processor cycle (see
5569 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
5570 This value must be constant over the entire compilation. If you need
5571 it to vary depending on what the instructions are, you must use
5572 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
5574 You could define this hook to return the value of the macro
5575 @code{MAX_DFA_ISSUE_RATE}.
5578 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
5579 This hook is executed by the scheduler after it has scheduled an insn
5580 from the ready list. It should return the number of insns which can
5581 still be issued in the current cycle. The default is
5582 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
5583 @code{USE}, which normally are not counted against the issue rate.
5584 You should define this hook if some insns take more machine resources
5585 than others, so that fewer insns can follow them in the same cycle.
5586 @var{file} is either a null pointer, or a stdio stream to write any
5587 debug output to. @var{verbose} is the verbose level provided by
5588 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
5592 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
5593 This function corrects the value of @var{cost} based on the
5594 relationship between @var{insn} and @var{dep_insn} through the
5595 dependence @var{link}. It should return the new value. The default
5596 is to make no adjustment to @var{cost}. This can be used for example
5597 to specify to the scheduler using the traditional pipeline description
5598 that an output- or anti-dependence does not incur the same cost as a
5599 data-dependence. If the scheduler using the automaton based pipeline
5600 description, the cost of anti-dependence is zero and the cost of
5601 output-dependence is maximum of one and the difference of latency
5602 times of the first and the second insns. If these values are not
5603 acceptable, you could use the hook to modify them too. See also
5604 @pxref{Processor pipeline description}.
5607 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
5608 This hook adjusts the integer scheduling priority @var{priority} of
5609 @var{insn}. It should return the new priority. Reduce the priority to
5610 execute @var{insn} earlier, increase the priority to execute @var{insn}
5611 later. Do not define this hook if you do not need to adjust the
5612 scheduling priorities of insns.
5615 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
5616 This hook is executed by the scheduler after it has scheduled the ready
5617 list, to allow the machine description to reorder it (for example to
5618 combine two small instructions together on @samp{VLIW} machines).
5619 @var{file} is either a null pointer, or a stdio stream to write any
5620 debug output to. @var{verbose} is the verbose level provided by
5621 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
5622 list of instructions that are ready to be scheduled. @var{n_readyp} is
5623 a pointer to the number of elements in the ready list. The scheduler
5624 reads the ready list in reverse order, starting with
5625 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
5626 is the timer tick of the scheduler. You may modify the ready list and
5627 the number of ready insns. The return value is the number of insns that
5628 can issue this cycle; normally this is just @code{issue_rate}. See also
5629 @samp{TARGET_SCHED_REORDER2}.
5632 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
5633 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
5634 function is called whenever the scheduler starts a new cycle. This one
5635 is called once per iteration over a cycle, immediately after
5636 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
5637 return the number of insns to be scheduled in the same cycle. Defining
5638 this hook can be useful if there are frequent situations where
5639 scheduling one insn causes other insns to become ready in the same
5640 cycle. These other insns can then be taken into account properly.
5643 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
5644 This hook is called after evaluation forward dependencies of insns in
5645 chain given by two parameter values (@var{head} and @var{tail}
5646 correspondingly) but before insns scheduling of the insn chain. For
5647 example, it can be used for better insn classification if it requires
5648 analysis of dependencies. This hook can use backward and forward
5649 dependencies of the insn scheduler because they are already
5653 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
5654 This hook is executed by the scheduler at the beginning of each block of
5655 instructions that are to be scheduled. @var{file} is either a null
5656 pointer, or a stdio stream to write any debug output to. @var{verbose}
5657 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5658 @var{max_ready} is the maximum number of insns in the current scheduling
5659 region that can be live at the same time. This can be used to allocate
5660 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
5663 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
5664 This hook is executed by the scheduler at the end of each block of
5665 instructions that are to be scheduled. It can be used to perform
5666 cleanup of any actions done by the other scheduling hooks. @var{file}
5667 is either a null pointer, or a stdio stream to write any debug output
5668 to. @var{verbose} is the verbose level provided by
5669 @option{-fsched-verbose-@var{n}}.
5672 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
5673 This hook is executed by the scheduler after function level initializations.
5674 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
5675 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5676 @var{old_max_uid} is the maximum insn uid when scheduling begins.
5679 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
5680 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
5681 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
5682 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5685 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
5686 The hook returns an RTL insn. The automaton state used in the
5687 pipeline hazard recognizer is changed as if the insn were scheduled
5688 when the new simulated processor cycle starts. Usage of the hook may
5689 simplify the automaton pipeline description for some @acronym{VLIW}
5690 processors. If the hook is defined, it is used only for the automaton
5691 based pipeline description. The default is not to change the state
5692 when the new simulated processor cycle starts.
5695 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
5696 The hook can be used to initialize data used by the previous hook.
5699 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
5700 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
5701 to changed the state as if the insn were scheduled when the new
5702 simulated processor cycle finishes.
5705 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
5706 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
5707 used to initialize data used by the previous hook.
5710 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
5711 This hook controls better choosing an insn from the ready insn queue
5712 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
5713 chooses the first insn from the queue. If the hook returns a positive
5714 value, an additional scheduler code tries all permutations of
5715 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
5716 subsequent ready insns to choose an insn whose issue will result in
5717 maximal number of issued insns on the same cycle. For the
5718 @acronym{VLIW} processor, the code could actually solve the problem of
5719 packing simple insns into the @acronym{VLIW} insn. Of course, if the
5720 rules of @acronym{VLIW} packing are described in the automaton.
5722 This code also could be used for superscalar @acronym{RISC}
5723 processors. Let us consider a superscalar @acronym{RISC} processor
5724 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
5725 @var{B}, some insns can be executed only in pipelines @var{B} or
5726 @var{C}, and one insn can be executed in pipeline @var{B}. The
5727 processor may issue the 1st insn into @var{A} and the 2nd one into
5728 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
5729 until the next cycle. If the scheduler issues the 3rd insn the first,
5730 the processor could issue all 3 insns per cycle.
5732 Actually this code demonstrates advantages of the automaton based
5733 pipeline hazard recognizer. We try quickly and easy many insn
5734 schedules to choose the best one.
5736 The default is no multipass scheduling.
5739 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
5741 This hook controls what insns from the ready insn queue will be
5742 considered for the multipass insn scheduling. If the hook returns
5743 zero for insn passed as the parameter, the insn will be not chosen to
5746 The default is that any ready insns can be chosen to be issued.
5749 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
5751 This hook is called by the insn scheduler before issuing insn passed
5752 as the third parameter on given cycle. If the hook returns nonzero,
5753 the insn is not issued on given processors cycle. Instead of that,
5754 the processor cycle is advanced. If the value passed through the last
5755 parameter is zero, the insn ready queue is not sorted on the new cycle
5756 start as usually. The first parameter passes file for debugging
5757 output. The second one passes the scheduler verbose level of the
5758 debugging output. The forth and the fifth parameter values are
5759 correspondingly processor cycle on which the previous insn has been
5760 issued and the current processor cycle.
5763 @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})
5764 This hook is used to define which dependences are considered costly by
5765 the target, so costly that it is not advisable to schedule the insns that
5766 are involved in the dependence too close to one another. The parameters
5767 to this hook are as follows: The second parameter @var{insn2} is dependent
5768 upon the first parameter @var{insn1}. The dependence between @var{insn1}
5769 and @var{insn2} is represented by the third parameter @var{dep_link}. The
5770 fourth parameter @var{cost} is the cost of the dependence, and the fifth
5771 parameter @var{distance} is the distance in cycles between the two insns.
5772 The hook returns @code{true} if considering the distance between the two
5773 insns the dependence between them is considered costly by the target,
5774 and @code{false} otherwise.
5776 Defining this hook can be useful in multiple-issue out-of-order machines,
5777 where (a) it's practically hopeless to predict the actual data/resource
5778 delays, however: (b) there's a better chance to predict the actual grouping
5779 that will be formed, and (c) correctly emulating the grouping can be very
5780 important. In such targets one may want to allow issuing dependent insns
5781 closer to one another---i.e., closer than the dependence distance; however,
5782 not in cases of "costly dependences", which this hooks allows to define.
5785 Macros in the following table are generated by the program
5786 @file{genattr} and can be useful for writing the hooks.
5788 @defmac MAX_DFA_ISSUE_RATE
5789 The macro definition is generated in the automaton based pipeline
5790 description interface. Its value is calculated from the automaton
5791 based pipeline description and is equal to maximal number of all insns
5792 described in constructions @samp{define_insn_reservation} which can be
5793 issued on the same processor cycle.
5797 @section Dividing the Output into Sections (Texts, Data, @dots{})
5798 @c the above section title is WAY too long. maybe cut the part between
5799 @c the (...)? --mew 10feb93
5801 An object file is divided into sections containing different types of
5802 data. In the most common case, there are three sections: the @dfn{text
5803 section}, which holds instructions and read-only data; the @dfn{data
5804 section}, which holds initialized writable data; and the @dfn{bss
5805 section}, which holds uninitialized data. Some systems have other kinds
5808 The compiler must tell the assembler when to switch sections. These
5809 macros control what commands to output to tell the assembler this. You
5810 can also define additional sections.
5812 @defmac TEXT_SECTION_ASM_OP
5813 A C expression whose value is a string, including spacing, containing the
5814 assembler operation that should precede instructions and read-only data.
5815 Normally @code{"\t.text"} is right.
5818 @defmac HOT_TEXT_SECTION_NAME
5819 If defined, a C string constant for the name of the section containing most
5820 frequently executed functions of the program. If not defined, GCC will provide
5821 a default definition if the target supports named sections.
5824 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5825 If defined, a C string constant for the name of the section containing unlikely
5826 executed functions in the program.
5829 @defmac DATA_SECTION_ASM_OP
5830 A C expression whose value is a string, including spacing, containing the
5831 assembler operation to identify the following data as writable initialized
5832 data. Normally @code{"\t.data"} is right.
5835 @defmac READONLY_DATA_SECTION_ASM_OP
5836 A C expression whose value is a string, including spacing, containing the
5837 assembler operation to identify the following data as read-only initialized
5841 @defmac READONLY_DATA_SECTION
5842 A macro naming a function to call to switch to the proper section for
5843 read-only data. The default is to use @code{READONLY_DATA_SECTION_ASM_OP}
5844 if defined, else fall back to @code{text_section}.
5846 The most common definition will be @code{data_section}, if the target
5847 does not have a special read-only data section, and does not put data
5848 in the text section.
5851 @defmac BSS_SECTION_ASM_OP
5852 If defined, a C expression whose value is a string, including spacing,
5853 containing the assembler operation to identify the following data as
5854 uninitialized global data. If not defined, and neither
5855 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5856 uninitialized global data will be output in the data section if
5857 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5861 @defmac INIT_SECTION_ASM_OP
5862 If defined, a C expression whose value is a string, including spacing,
5863 containing the assembler operation to identify the following data as
5864 initialization code. If not defined, GCC will assume such a section does
5868 @defmac FINI_SECTION_ASM_OP
5869 If defined, a C expression whose value is a string, including spacing,
5870 containing the assembler operation to identify the following data as
5871 finalization code. If not defined, GCC will assume such a section does
5875 @defmac INIT_ARRAY_SECTION_ASM_OP
5876 If defined, a C expression whose value is a string, including spacing,
5877 containing the assembler operation to identify the following data as
5878 part of the @code{.init_array} (or equivalent) section. If not
5879 defined, GCC will assume such a section does not exist. Do not define
5880 both this macro and @code{INIT_SECTION_ASM_OP}.
5883 @defmac FINI_ARRAY_SECTION_ASM_OP
5884 If defined, a C expression whose value is a string, including spacing,
5885 containing the assembler operation to identify the following data as
5886 part of the @code{.fini_array} (or equivalent) section. If not
5887 defined, GCC will assume such a section does not exist. Do not define
5888 both this macro and @code{FINI_SECTION_ASM_OP}.
5891 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
5892 If defined, an ASM statement that switches to a different section
5893 via @var{section_op}, calls @var{function}, and switches back to
5894 the text section. This is used in @file{crtstuff.c} if
5895 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
5896 to initialization and finalization functions from the init and fini
5897 sections. By default, this macro uses a simple function call. Some
5898 ports need hand-crafted assembly code to avoid dependencies on
5899 registers initialized in the function prologue or to ensure that
5900 constant pools don't end up too far way in the text section.
5903 @defmac FORCE_CODE_SECTION_ALIGN
5904 If defined, an ASM statement that aligns a code section to some
5905 arbitrary boundary. This is used to force all fragments of the
5906 @code{.init} and @code{.fini} sections to have to same alignment
5907 and thus prevent the linker from having to add any padding.
5912 @defmac EXTRA_SECTIONS
5913 A list of names for sections other than the standard two, which are
5914 @code{in_text} and @code{in_data}. You need not define this macro
5915 on a system with no other sections (that GCC needs to use).
5918 @findex text_section
5919 @findex data_section
5920 @defmac EXTRA_SECTION_FUNCTIONS
5921 One or more functions to be defined in @file{varasm.c}. These
5922 functions should do jobs analogous to those of @code{text_section} and
5923 @code{data_section}, for your additional sections. Do not define this
5924 macro if you do not define @code{EXTRA_SECTIONS}.
5927 @defmac JUMP_TABLES_IN_TEXT_SECTION
5928 Define this macro to be an expression with a nonzero value if jump
5929 tables (for @code{tablejump} insns) should be output in the text
5930 section, along with the assembler instructions. Otherwise, the
5931 readonly data section is used.
5933 This macro is irrelevant if there is no separate readonly data section.
5936 @deftypefn {Target Hook} void TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
5937 Switches to the appropriate section for output of @var{exp}. You can
5938 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
5939 some sort. @var{reloc} indicates whether the initial value of @var{exp}
5940 requires link-time relocations. Bit 0 is set when variable contains
5941 local relocations only, while bit 1 is set for global relocations.
5942 Select the section by calling @code{data_section} or one of the
5943 alternatives for other sections. @var{align} is the constant alignment
5946 The default version of this function takes care of putting read-only
5947 variables in @code{readonly_data_section}.
5949 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
5952 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
5953 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
5954 for @code{FUNCTION_DECL}s as well as for variables and constants.
5956 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
5957 function has been determined to be likely to be called, and nonzero if
5958 it is unlikely to be called.
5961 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
5962 Build up a unique section name, expressed as a @code{STRING_CST} node,
5963 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
5964 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
5965 the initial value of @var{exp} requires link-time relocations.
5967 The default version of this function appends the symbol name to the
5968 ELF section name that would normally be used for the symbol. For
5969 example, the function @code{foo} would be placed in @code{.text.foo}.
5970 Whatever the actual target object format, this is often good enough.
5973 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
5974 Switches to a readonly data section associated with
5975 @samp{DECL_SECTION_NAME (@var{decl})}.
5976 The default version of this function switches to @code{.gnu.linkonce.r.name}
5977 section if function's section is @code{.gnu.linkonce.t.name}, to
5978 @code{.rodata.name} if function is in @code{.text.name} section
5979 and otherwise switches to the normal readonly data section.
5982 @deftypefn {Target Hook} void TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
5983 Switches to the appropriate section for output of constant pool entry
5984 @var{x} in @var{mode}. You can assume that @var{x} is some kind of
5985 constant in RTL@. The argument @var{mode} is redundant except in the
5986 case of a @code{const_int} rtx. Select the section by calling
5987 @code{readonly_data_section} or one of the alternatives for other
5988 sections. @var{align} is the constant alignment in bits.
5990 The default version of this function takes care of putting symbolic
5991 constants in @code{flag_pic} mode in @code{data_section} and everything
5992 else in @code{readonly_data_section}.
5995 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
5996 Define this hook if references to a symbol or a constant must be
5997 treated differently depending on something about the variable or
5998 function named by the symbol (such as what section it is in).
6000 The hook is executed immediately after rtl has been created for
6001 @var{decl}, which may be a variable or function declaration or
6002 an entry in the constant pool. In either case, @var{rtl} is the
6003 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
6004 in this hook; that field may not have been initialized yet.
6006 In the case of a constant, it is safe to assume that the rtl is
6007 a @code{mem} whose address is a @code{symbol_ref}. Most decls
6008 will also have this form, but that is not guaranteed. Global
6009 register variables, for instance, will have a @code{reg} for their
6010 rtl. (Normally the right thing to do with such unusual rtl is
6013 The @var{new_decl_p} argument will be true if this is the first time
6014 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
6015 be false for subsequent invocations, which will happen for duplicate
6016 declarations. Whether or not anything must be done for the duplicate
6017 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
6018 @var{new_decl_p} is always true when the hook is called for a constant.
6020 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
6021 The usual thing for this hook to do is to record flags in the
6022 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
6023 Historically, the name string was modified if it was necessary to
6024 encode more than one bit of information, but this practice is now
6025 discouraged; use @code{SYMBOL_REF_FLAGS}.
6027 The default definition of this hook, @code{default_encode_section_info}
6028 in @file{varasm.c}, sets a number of commonly-useful bits in
6029 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
6030 before overriding it.
6033 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
6034 Decode @var{name} and return the real name part, sans
6035 the characters that @code{TARGET_ENCODE_SECTION_INFO}
6039 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
6040 Returns true if @var{exp} should be placed into a ``small data'' section.
6041 The default version of this hook always returns false.
6044 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
6045 Contains the value true if the target places read-only
6046 ``small data'' into a separate section. The default value is false.
6049 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
6050 Returns true if @var{exp} names an object for which name resolution
6051 rules must resolve to the current ``module'' (dynamic shared library
6052 or executable image).
6054 The default version of this hook implements the name resolution rules
6055 for ELF, which has a looser model of global name binding than other
6056 currently supported object file formats.
6059 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
6060 Contains the value true if the target supports thread-local storage.
6061 The default value is false.
6066 @section Position Independent Code
6067 @cindex position independent code
6070 This section describes macros that help implement generation of position
6071 independent code. Simply defining these macros is not enough to
6072 generate valid PIC; you must also add support to the macros
6073 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
6074 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
6075 @samp{movsi} to do something appropriate when the source operand
6076 contains a symbolic address. You may also need to alter the handling of
6077 switch statements so that they use relative addresses.
6078 @c i rearranged the order of the macros above to try to force one of
6079 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
6081 @defmac PIC_OFFSET_TABLE_REGNUM
6082 The register number of the register used to address a table of static
6083 data addresses in memory. In some cases this register is defined by a
6084 processor's ``application binary interface'' (ABI)@. When this macro
6085 is defined, RTL is generated for this register once, as with the stack
6086 pointer and frame pointer registers. If this macro is not defined, it
6087 is up to the machine-dependent files to allocate such a register (if
6088 necessary). Note that this register must be fixed when in use (e.g.@:
6089 when @code{flag_pic} is true).
6092 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6093 Define this macro if the register defined by
6094 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
6095 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
6098 @defmac FINALIZE_PIC
6099 By generating position-independent code, when two different programs (A
6100 and B) share a common library (libC.a), the text of the library can be
6101 shared whether or not the library is linked at the same address for both
6102 programs. In some of these environments, position-independent code
6103 requires not only the use of different addressing modes, but also
6104 special code to enable the use of these addressing modes.
6106 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
6107 codes once the function is being compiled into assembly code, but not
6108 before. (It is not done before, because in the case of compiling an
6109 inline function, it would lead to multiple PIC prologues being
6110 included in functions which used inline functions and were compiled to
6114 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
6115 A C expression that is nonzero if @var{x} is a legitimate immediate
6116 operand on the target machine when generating position independent code.
6117 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6118 check this. You can also assume @var{flag_pic} is true, so you need not
6119 check it either. You need not define this macro if all constants
6120 (including @code{SYMBOL_REF}) can be immediate operands when generating
6121 position independent code.
6124 @node Assembler Format
6125 @section Defining the Output Assembler Language
6127 This section describes macros whose principal purpose is to describe how
6128 to write instructions in assembler language---rather than what the
6132 * File Framework:: Structural information for the assembler file.
6133 * Data Output:: Output of constants (numbers, strings, addresses).
6134 * Uninitialized Data:: Output of uninitialized variables.
6135 * Label Output:: Output and generation of labels.
6136 * Initialization:: General principles of initialization
6137 and termination routines.
6138 * Macros for Initialization::
6139 Specific macros that control the handling of
6140 initialization and termination routines.
6141 * Instruction Output:: Output of actual instructions.
6142 * Dispatch Tables:: Output of jump tables.
6143 * Exception Region Output:: Output of exception region code.
6144 * Alignment Output:: Pseudo ops for alignment and skipping data.
6147 @node File Framework
6148 @subsection The Overall Framework of an Assembler File
6149 @cindex assembler format
6150 @cindex output of assembler code
6152 @c prevent bad page break with this line
6153 This describes the overall framework of an assembly file.
6155 @deftypefn {Target Hook} void TARGET_ASM_FILE_START ()
6156 @findex default_file_start
6157 Output to @code{asm_out_file} any text which the assembler expects to
6158 find at the beginning of a file. The default behavior is controlled
6159 by two flags, documented below. Unless your target's assembler is
6160 quite unusual, if you override the default, you should call
6161 @code{default_file_start} at some point in your target hook. This
6162 lets other target files rely on these variables.
6165 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
6166 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
6167 printed as the very first line in the assembly file, unless
6168 @option{-fverbose-asm} is in effect. (If that macro has been defined
6169 to the empty string, this variable has no effect.) With the normal
6170 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
6171 assembler that it need not bother stripping comments or extra
6172 whitespace from its input. This allows it to work a bit faster.
6174 The default is false. You should not set it to true unless you have
6175 verified that your port does not generate any extra whitespace or
6176 comments that will cause GAS to issue errors in NO_APP mode.
6179 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
6180 If this flag is true, @code{output_file_directive} will be called
6181 for the primary source file, immediately after printing
6182 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
6183 this to be done. The default is false.
6186 @deftypefn {Target Hook} void TARGET_ASM_FILE_END ()
6187 Output to @code{asm_out_file} any text which the assembler expects
6188 to find at the end of a file. The default is to output nothing.
6191 @deftypefun void file_end_indicate_exec_stack ()
6192 Some systems use a common convention, the @samp{.note.GNU-stack}
6193 special section, to indicate whether or not an object file relies on
6194 the stack being executable. If your system uses this convention, you
6195 should define @code{TARGET_ASM_FILE_END} to this function. If you
6196 need to do other things in that hook, have your hook function call
6200 @defmac ASM_COMMENT_START
6201 A C string constant describing how to begin a comment in the target
6202 assembler language. The compiler assumes that the comment will end at
6203 the end of the line.
6207 A C string constant for text to be output before each @code{asm}
6208 statement or group of consecutive ones. Normally this is
6209 @code{"#APP"}, which is a comment that has no effect on most
6210 assemblers but tells the GNU assembler that it must check the lines
6211 that follow for all valid assembler constructs.
6215 A C string constant for text to be output after each @code{asm}
6216 statement or group of consecutive ones. Normally this is
6217 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6218 time-saving assumptions that are valid for ordinary compiler output.
6221 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6222 A C statement to output COFF information or DWARF debugging information
6223 which indicates that filename @var{name} is the current source file to
6224 the stdio stream @var{stream}.
6226 This macro need not be defined if the standard form of output
6227 for the file format in use is appropriate.
6230 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6231 A C statement to output the string @var{string} to the stdio stream
6232 @var{stream}. If you do not call the function @code{output_quoted_string}
6233 in your config files, GCC will only call it to output filenames to
6234 the assembler source. So you can use it to canonicalize the format
6235 of the filename using this macro.
6238 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6239 A C statement to output something to the assembler file to handle a
6240 @samp{#ident} directive containing the text @var{string}. If this
6241 macro is not defined, nothing is output for a @samp{#ident} directive.
6244 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6245 Output assembly directives to switch to section @var{name}. The section
6246 should have attributes as specified by @var{flags}, which is a bit mask
6247 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6248 is nonzero, it contains an alignment in bytes to be used for the section,
6249 otherwise some target default should be used. Only targets that must
6250 specify an alignment within the section directive need pay attention to
6251 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6254 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6255 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6258 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6259 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6260 based on a variable or function decl, a section name, and whether or not the
6261 declaration's initializer may contain runtime relocations. @var{decl} may be
6262 null, in which case read-write data should be assumed.
6264 The default version if this function handles choosing code vs data,
6265 read-only vs read-write data, and @code{flag_pic}. You should only
6266 need to override this if your target has special flags that might be
6267 set via @code{__attribute__}.
6272 @subsection Output of Data
6275 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6276 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6277 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6278 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6279 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6280 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6281 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6282 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6283 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6284 These hooks specify assembly directives for creating certain kinds
6285 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6286 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6287 aligned two-byte object, and so on. Any of the hooks may be
6288 @code{NULL}, indicating that no suitable directive is available.
6290 The compiler will print these strings at the start of a new line,
6291 followed immediately by the object's initial value. In most cases,
6292 the string should contain a tab, a pseudo-op, and then another tab.
6295 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6296 The @code{assemble_integer} function uses this hook to output an
6297 integer object. @var{x} is the object's value, @var{size} is its size
6298 in bytes and @var{aligned_p} indicates whether it is aligned. The
6299 function should return @code{true} if it was able to output the
6300 object. If it returns false, @code{assemble_integer} will try to
6301 split the object into smaller parts.
6303 The default implementation of this hook will use the
6304 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6305 when the relevant string is @code{NULL}.
6308 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6309 A C statement to recognize @var{rtx} patterns that
6310 @code{output_addr_const} can't deal with, and output assembly code to
6311 @var{stream} corresponding to the pattern @var{x}. This may be used to
6312 allow machine-dependent @code{UNSPEC}s to appear within constants.
6314 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6315 @code{goto fail}, so that a standard error message is printed. If it
6316 prints an error message itself, by calling, for example,
6317 @code{output_operand_lossage}, it may just complete normally.
6320 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6321 A C statement to output to the stdio stream @var{stream} an assembler
6322 instruction to assemble a string constant containing the @var{len}
6323 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6324 @code{char *} and @var{len} a C expression of type @code{int}.
6326 If the assembler has a @code{.ascii} pseudo-op as found in the
6327 Berkeley Unix assembler, do not define the macro
6328 @code{ASM_OUTPUT_ASCII}.
6331 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6332 A C statement to output word @var{n} of a function descriptor for
6333 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6334 is defined, and is otherwise unused.
6337 @defmac CONSTANT_POOL_BEFORE_FUNCTION
6338 You may define this macro as a C expression. You should define the
6339 expression to have a nonzero value if GCC should output the constant
6340 pool for a function before the code for the function, or a zero value if
6341 GCC should output the constant pool after the function. If you do
6342 not define this macro, the usual case, GCC will output the constant
6343 pool before the function.
6346 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
6347 A C statement to output assembler commands to define the start of the
6348 constant pool for a function. @var{funname} is a string giving
6349 the name of the function. Should the return type of the function
6350 be required, it can be obtained via @var{fundecl}. @var{size}
6351 is the size, in bytes, of the constant pool that will be written
6352 immediately after this call.
6354 If no constant-pool prefix is required, the usual case, this macro need
6358 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
6359 A C statement (with or without semicolon) to output a constant in the
6360 constant pool, if it needs special treatment. (This macro need not do
6361 anything for RTL expressions that can be output normally.)
6363 The argument @var{file} is the standard I/O stream to output the
6364 assembler code on. @var{x} is the RTL expression for the constant to
6365 output, and @var{mode} is the machine mode (in case @var{x} is a
6366 @samp{const_int}). @var{align} is the required alignment for the value
6367 @var{x}; you should output an assembler directive to force this much
6370 The argument @var{labelno} is a number to use in an internal label for
6371 the address of this pool entry. The definition of this macro is
6372 responsible for outputting the label definition at the proper place.
6373 Here is how to do this:
6376 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
6379 When you output a pool entry specially, you should end with a
6380 @code{goto} to the label @var{jumpto}. This will prevent the same pool
6381 entry from being output a second time in the usual manner.
6383 You need not define this macro if it would do nothing.
6386 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
6387 A C statement to output assembler commands to at the end of the constant
6388 pool for a function. @var{funname} is a string giving the name of the
6389 function. Should the return type of the function be required, you can
6390 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
6391 constant pool that GCC wrote immediately before this call.
6393 If no constant-pool epilogue is required, the usual case, you need not
6397 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
6398 Define this macro as a C expression which is nonzero if @var{C} is
6399 used as a logical line separator by the assembler.
6401 If you do not define this macro, the default is that only
6402 the character @samp{;} is treated as a logical line separator.
6405 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
6406 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
6407 These target hooks are C string constants, describing the syntax in the
6408 assembler for grouping arithmetic expressions. If not overridden, they
6409 default to normal parentheses, which is correct for most assemblers.
6412 These macros are provided by @file{real.h} for writing the definitions
6413 of @code{ASM_OUTPUT_DOUBLE} and the like:
6415 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
6416 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
6417 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
6418 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
6419 floating point representation, and store its bit pattern in the variable
6420 @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE}, this variable should
6421 be a simple @code{long int}. For the others, it should be an array of
6422 @code{long int}. The number of elements in this array is determined by
6423 the size of the desired target floating point data type: 32 bits of it
6424 go in each @code{long int} array element. Each array element holds 32
6425 bits of the result, even if @code{long int} is wider than 32 bits on the
6428 The array element values are designed so that you can print them out
6429 using @code{fprintf} in the order they should appear in the target
6433 @node Uninitialized Data
6434 @subsection Output of Uninitialized Variables
6436 Each of the macros in this section is used to do the whole job of
6437 outputting a single uninitialized variable.
6439 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6440 A C statement (sans semicolon) to output to the stdio stream
6441 @var{stream} the assembler definition of a common-label named
6442 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6443 is the size rounded up to whatever alignment the caller wants.
6445 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6446 output the name itself; before and after that, output the additional
6447 assembler syntax for defining the name, and a newline.
6449 This macro controls how the assembler definitions of uninitialized
6450 common global variables are output.
6453 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
6454 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
6455 separate, explicit argument. If you define this macro, it is used in
6456 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
6457 handling the required alignment of the variable. The alignment is specified
6458 as the number of bits.
6461 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6462 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
6463 variable to be output, if there is one, or @code{NULL_TREE} if there
6464 is no corresponding variable. If you define this macro, GCC will use it
6465 in place of both @code{ASM_OUTPUT_COMMON} and
6466 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
6467 the variable's decl in order to chose what to output.
6470 @defmac ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6471 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
6472 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
6476 @defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6477 A C statement (sans semicolon) to output to the stdio stream
6478 @var{stream} the assembler definition of uninitialized global @var{decl} named
6479 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6480 is the size rounded up to whatever alignment the caller wants.
6482 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
6483 defining this macro. If unable, use the expression
6484 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
6485 before and after that, output the additional assembler syntax for defining
6486 the name, and a newline.
6488 This macro controls how the assembler definitions of uninitialized global
6489 variables are output. This macro exists to properly support languages like
6490 C++ which do not have @code{common} data. However, this macro currently
6491 is not defined for all targets. If this macro and
6492 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
6493 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
6494 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
6497 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6498 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
6499 separate, explicit argument. If you define this macro, it is used in
6500 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
6501 handling the required alignment of the variable. The alignment is specified
6502 as the number of bits.
6504 Try to use function @code{asm_output_aligned_bss} defined in file
6505 @file{varasm.c} when defining this macro.
6508 @defmac ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6509 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
6510 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
6514 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6515 A C statement (sans semicolon) to output to the stdio stream
6516 @var{stream} the assembler definition of a local-common-label named
6517 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6518 is the size rounded up to whatever alignment the caller wants.
6520 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6521 output the name itself; before and after that, output the additional
6522 assembler syntax for defining the name, and a newline.
6524 This macro controls how the assembler definitions of uninitialized
6525 static variables are output.
6528 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
6529 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
6530 separate, explicit argument. If you define this macro, it is used in
6531 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
6532 handling the required alignment of the variable. The alignment is specified
6533 as the number of bits.
6536 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6537 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
6538 variable to be output, if there is one, or @code{NULL_TREE} if there
6539 is no corresponding variable. If you define this macro, GCC will use it
6540 in place of both @code{ASM_OUTPUT_DECL} and
6541 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
6542 the variable's decl in order to chose what to output.
6545 @defmac ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6546 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
6547 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
6552 @subsection Output and Generation of Labels
6554 @c prevent bad page break with this line
6555 This is about outputting labels.
6557 @findex assemble_name
6558 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
6559 A C statement (sans semicolon) to output to the stdio stream
6560 @var{stream} the assembler definition of a label named @var{name}.
6561 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6562 output the name itself; before and after that, output the additional
6563 assembler syntax for defining the name, and a newline. A default
6564 definition of this macro is provided which is correct for most systems.
6567 @findex assemble_name_raw
6568 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
6569 Identical to @code{ASM_OUTPUT_lABEL}, except that @var{name} is known
6570 to refer to a compiler-generated label. The default definition uses
6571 @code{assemble_name_raw}, which is like @code{assemble_name} except
6572 that it is more efficient.
6576 A C string containing the appropriate assembler directive to specify the
6577 size of a symbol, without any arguments. On systems that use ELF, the
6578 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
6579 systems, the default is not to define this macro.
6581 Define this macro only if it is correct to use the default definitions
6582 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
6583 for your system. If you need your own custom definitions of those
6584 macros, or if you do not need explicit symbol sizes at all, do not
6588 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
6589 A C statement (sans semicolon) to output to the stdio stream
6590 @var{stream} a directive telling the assembler that the size of the
6591 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
6592 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6596 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
6597 A C statement (sans semicolon) to output to the stdio stream
6598 @var{stream} a directive telling the assembler to calculate the size of
6599 the symbol @var{name} by subtracting its address from the current
6602 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6603 provided. The default assumes that the assembler recognizes a special
6604 @samp{.} symbol as referring to the current address, and can calculate
6605 the difference between this and another symbol. If your assembler does
6606 not recognize @samp{.} or cannot do calculations with it, you will need
6607 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
6611 A C string containing the appropriate assembler directive to specify the
6612 type of a symbol, without any arguments. On systems that use ELF, the
6613 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
6614 systems, the default is not to define this macro.
6616 Define this macro only if it is correct to use the default definition of
6617 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6618 custom definition of this macro, or if you do not need explicit symbol
6619 types at all, do not define this macro.
6622 @defmac TYPE_OPERAND_FMT
6623 A C string which specifies (using @code{printf} syntax) the format of
6624 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
6625 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
6626 the default is not to define this macro.
6628 Define this macro only if it is correct to use the default definition of
6629 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6630 custom definition of this macro, or if you do not need explicit symbol
6631 types at all, do not define this macro.
6634 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
6635 A C statement (sans semicolon) to output to the stdio stream
6636 @var{stream} a directive telling the assembler that the type of the
6637 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
6638 that string is always either @samp{"function"} or @samp{"object"}, but
6639 you should not count on this.
6641 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
6642 definition of this macro is provided.
6645 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
6646 A C statement (sans semicolon) to output to the stdio stream
6647 @var{stream} any text necessary for declaring the name @var{name} of a
6648 function which is being defined. This macro is responsible for
6649 outputting the label definition (perhaps using
6650 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
6651 @code{FUNCTION_DECL} tree node representing the function.
6653 If this macro is not defined, then the function name is defined in the
6654 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6656 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6660 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
6661 A C statement (sans semicolon) to output to the stdio stream
6662 @var{stream} any text necessary for declaring the size of a function
6663 which is being defined. The argument @var{name} is the name of the
6664 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
6665 representing the function.
6667 If this macro is not defined, then the function size is not defined.
6669 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
6673 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6674 A C statement (sans semicolon) to output to the stdio stream
6675 @var{stream} any text necessary for declaring the name @var{name} of an
6676 initialized variable which is being defined. This macro must output the
6677 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6678 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6680 If this macro is not defined, then the variable name is defined in the
6681 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6683 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
6684 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
6687 @defmac ASM_DECLARE_CONSTANT_NAME (@var{stream}, @var{name}, @var{exp}, @var{size})
6688 A C statement (sans semicolon) to output to the stdio stream
6689 @var{stream} any text necessary for declaring the name @var{name} of a
6690 constant which is being defined. This macro is responsible for
6691 outputting the label definition (perhaps using
6692 @code{ASM_OUTPUT_LABEL}). The argument @var{exp} is the
6693 value of the constant, and @var{size} is the size of the constant
6694 in bytes. @var{name} will be an internal label.
6696 If this macro is not defined, then the @var{name} is defined in the
6697 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6699 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6703 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6704 A C statement (sans semicolon) to output to the stdio stream
6705 @var{stream} any text necessary for claiming a register @var{regno}
6706 for a global variable @var{decl} with name @var{name}.
6708 If you don't define this macro, that is equivalent to defining it to do
6712 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6713 A C statement (sans semicolon) to finish up declaring a variable name
6714 once the compiler has processed its initializer fully and thus has had a
6715 chance to determine the size of an array when controlled by an
6716 initializer. This is used on systems where it's necessary to declare
6717 something about the size of the object.
6719 If you don't define this macro, that is equivalent to defining it to do
6722 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
6723 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
6726 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
6727 This target hook is a function to output to the stdio stream
6728 @var{stream} some commands that will make the label @var{name} global;
6729 that is, available for reference from other files.
6731 The default implementation relies on a proper definition of
6732 @code{GLOBAL_ASM_OP}.
6735 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
6736 A C statement (sans semicolon) to output to the stdio stream
6737 @var{stream} some commands that will make the label @var{name} weak;
6738 that is, available for reference from other files but only used if
6739 no other definition is available. Use the expression
6740 @code{assemble_name (@var{stream}, @var{name})} to output the name
6741 itself; before and after that, output the additional assembler syntax
6742 for making that name weak, and a newline.
6744 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
6745 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
6749 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
6750 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
6751 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
6752 or variable decl. If @var{value} is not @code{NULL}, this C statement
6753 should output to the stdio stream @var{stream} assembler code which
6754 defines (equates) the weak symbol @var{name} to have the value
6755 @var{value}. If @var{value} is @code{NULL}, it should output commands
6756 to make @var{name} weak.
6759 @defmac SUPPORTS_WEAK
6760 A C expression which evaluates to true if the target supports weak symbols.
6762 If you don't define this macro, @file{defaults.h} provides a default
6763 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
6764 is defined, the default definition is @samp{1}; otherwise, it is
6765 @samp{0}. Define this macro if you want to control weak symbol support
6766 with a compiler flag such as @option{-melf}.
6769 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
6770 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
6771 public symbol such that extra copies in multiple translation units will
6772 be discarded by the linker. Define this macro if your object file
6773 format provides support for this concept, such as the @samp{COMDAT}
6774 section flags in the Microsoft Windows PE/COFF format, and this support
6775 requires changes to @var{decl}, such as putting it in a separate section.
6778 @defmac SUPPORTS_ONE_ONLY
6779 A C expression which evaluates to true if the target supports one-only
6782 If you don't define this macro, @file{varasm.c} provides a default
6783 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6784 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6785 you want to control one-only symbol support with a compiler flag, or if
6786 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6787 be emitted as one-only.
6790 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
6791 This target hook is a function to output to @var{asm_out_file} some
6792 commands that will make the symbol(s) associated with @var{decl} have
6793 hidden, protected or internal visibility as specified by @var{visibility}.
6796 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
6797 A C expression that evaluates to true if the target's linker expects
6798 that weak symbols do not appear in a static archive's table of contents.
6799 The default is @code{0}.
6801 Leaving weak symbols out of an archive's table of contents means that,
6802 if a symbol will only have a definition in one translation unit and
6803 will have undefined references from other translation units, that
6804 symbol should not be weak. Defining this macro to be nonzero will
6805 thus have the effect that certain symbols that would normally be weak
6806 (explicit template instantiations, and vtables for polymorphic classes
6807 with noninline key methods) will instead be nonweak.
6809 The C++ ABI requires this macro to be zero. Define this macro for
6810 targets where full C++ ABI compliance is impossible and where linker
6811 restrictions require weak symbols to be left out of a static archive's
6815 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6816 A C statement (sans semicolon) to output to the stdio stream
6817 @var{stream} any text necessary for declaring the name of an external
6818 symbol named @var{name} which is referenced in this compilation but
6819 not defined. The value of @var{decl} is the tree node for the
6822 This macro need not be defined if it does not need to output anything.
6823 The GNU assembler and most Unix assemblers don't require anything.
6826 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
6827 This target hook is a function to output to @var{asm_out_file} an assembler
6828 pseudo-op to declare a library function name external. The name of the
6829 library function is given by @var{symref}, which is a @code{symbol_ref}.
6832 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (tree @var{decl})
6833 This target hook is a function to output to @var{asm_out_file} an assembler
6834 directive to annotate used symbol. Darwin target use .no_dead_code_strip
6838 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6839 A C statement (sans semicolon) to output to the stdio stream
6840 @var{stream} a reference in assembler syntax to a label named
6841 @var{name}. This should add @samp{_} to the front of the name, if that
6842 is customary on your operating system, as it is in most Berkeley Unix
6843 systems. This macro is used in @code{assemble_name}.
6846 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6847 A C statement (sans semicolon) to output a reference to
6848 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
6849 will be used to output the name of the symbol. This macro may be used
6850 to modify the way a symbol is referenced depending on information
6851 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
6854 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
6855 A C statement (sans semicolon) to output a reference to @var{buf}, the
6856 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
6857 @code{assemble_name} will be used to output the name of the symbol.
6858 This macro is not used by @code{output_asm_label}, or the @code{%l}
6859 specifier that calls it; the intention is that this macro should be set
6860 when it is necessary to output a label differently when its address is
6864 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
6865 A function to output to the stdio stream @var{stream} a label whose
6866 name is made from the string @var{prefix} and the number @var{labelno}.
6868 It is absolutely essential that these labels be distinct from the labels
6869 used for user-level functions and variables. Otherwise, certain programs
6870 will have name conflicts with internal labels.
6872 It is desirable to exclude internal labels from the symbol table of the
6873 object file. Most assemblers have a naming convention for labels that
6874 should be excluded; on many systems, the letter @samp{L} at the
6875 beginning of a label has this effect. You should find out what
6876 convention your system uses, and follow it.
6878 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
6881 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
6882 A C statement to output to the stdio stream @var{stream} a debug info
6883 label whose name is made from the string @var{prefix} and the number
6884 @var{num}. This is useful for VLIW targets, where debug info labels
6885 may need to be treated differently than branch target labels. On some
6886 systems, branch target labels must be at the beginning of instruction
6887 bundles, but debug info labels can occur in the middle of instruction
6890 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
6894 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
6895 A C statement to store into the string @var{string} a label whose name
6896 is made from the string @var{prefix} and the number @var{num}.
6898 This string, when output subsequently by @code{assemble_name}, should
6899 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
6900 with the same @var{prefix} and @var{num}.
6902 If the string begins with @samp{*}, then @code{assemble_name} will
6903 output the rest of the string unchanged. It is often convenient for
6904 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
6905 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
6906 to output the string, and may change it. (Of course,
6907 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
6908 you should know what it does on your machine.)
6911 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
6912 A C expression to assign to @var{outvar} (which is a variable of type
6913 @code{char *}) a newly allocated string made from the string
6914 @var{name} and the number @var{number}, with some suitable punctuation
6915 added. Use @code{alloca} to get space for the string.
6917 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
6918 produce an assembler label for an internal static variable whose name is
6919 @var{name}. Therefore, the string must be such as to result in valid
6920 assembler code. The argument @var{number} is different each time this
6921 macro is executed; it prevents conflicts between similarly-named
6922 internal static variables in different scopes.
6924 Ideally this string should not be a valid C identifier, to prevent any
6925 conflict with the user's own symbols. Most assemblers allow periods
6926 or percent signs in assembler symbols; putting at least one of these
6927 between the name and the number will suffice.
6929 If this macro is not defined, a default definition will be provided
6930 which is correct for most systems.
6933 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
6934 A C statement to output to the stdio stream @var{stream} assembler code
6935 which defines (equates) the symbol @var{name} to have the value @var{value}.
6938 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6939 correct for most systems.
6942 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
6943 A C statement to output to the stdio stream @var{stream} assembler code
6944 which defines (equates) the symbol whose tree node is @var{decl_of_name}
6945 to have the value of the tree node @var{decl_of_value}. This macro will
6946 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
6947 the tree nodes are available.
6950 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6951 correct for most systems.
6954 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
6955 A C statement that evaluates to true if the assembler code which defines
6956 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
6957 of the tree node @var{decl_of_value} should be emitted near the end of the
6958 current compilation unit. The default is to not defer output of defines.
6959 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
6960 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
6963 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
6964 A C statement to output to the stdio stream @var{stream} assembler code
6965 which defines (equates) the weak symbol @var{name} to have the value
6966 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
6967 an undefined weak symbol.
6969 Define this macro if the target only supports weak aliases; define
6970 @code{ASM_OUTPUT_DEF} instead if possible.
6973 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
6974 Define this macro to override the default assembler names used for
6975 Objective-C methods.
6977 The default name is a unique method number followed by the name of the
6978 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
6979 the category is also included in the assembler name (e.g.@:
6982 These names are safe on most systems, but make debugging difficult since
6983 the method's selector is not present in the name. Therefore, particular
6984 systems define other ways of computing names.
6986 @var{buf} is an expression of type @code{char *} which gives you a
6987 buffer in which to store the name; its length is as long as
6988 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
6989 50 characters extra.
6991 The argument @var{is_inst} specifies whether the method is an instance
6992 method or a class method; @var{class_name} is the name of the class;
6993 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
6994 in a category); and @var{sel_name} is the name of the selector.
6996 On systems where the assembler can handle quoted names, you can use this
6997 macro to provide more human-readable names.
7000 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
7001 A C statement (sans semicolon) to output to the stdio stream
7002 @var{stream} commands to declare that the label @var{name} is an
7003 Objective-C class reference. This is only needed for targets whose
7004 linkers have special support for NeXT-style runtimes.
7007 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
7008 A C statement (sans semicolon) to output to the stdio stream
7009 @var{stream} commands to declare that the label @var{name} is an
7010 unresolved Objective-C class reference. This is only needed for targets
7011 whose linkers have special support for NeXT-style runtimes.
7014 @node Initialization
7015 @subsection How Initialization Functions Are Handled
7016 @cindex initialization routines
7017 @cindex termination routines
7018 @cindex constructors, output of
7019 @cindex destructors, output of
7021 The compiled code for certain languages includes @dfn{constructors}
7022 (also called @dfn{initialization routines})---functions to initialize
7023 data in the program when the program is started. These functions need
7024 to be called before the program is ``started''---that is to say, before
7025 @code{main} is called.
7027 Compiling some languages generates @dfn{destructors} (also called
7028 @dfn{termination routines}) that should be called when the program
7031 To make the initialization and termination functions work, the compiler
7032 must output something in the assembler code to cause those functions to
7033 be called at the appropriate time. When you port the compiler to a new
7034 system, you need to specify how to do this.
7036 There are two major ways that GCC currently supports the execution of
7037 initialization and termination functions. Each way has two variants.
7038 Much of the structure is common to all four variations.
7040 @findex __CTOR_LIST__
7041 @findex __DTOR_LIST__
7042 The linker must build two lists of these functions---a list of
7043 initialization functions, called @code{__CTOR_LIST__}, and a list of
7044 termination functions, called @code{__DTOR_LIST__}.
7046 Each list always begins with an ignored function pointer (which may hold
7047 0, @minus{}1, or a count of the function pointers after it, depending on
7048 the environment). This is followed by a series of zero or more function
7049 pointers to constructors (or destructors), followed by a function
7050 pointer containing zero.
7052 Depending on the operating system and its executable file format, either
7053 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
7054 time and exit time. Constructors are called in reverse order of the
7055 list; destructors in forward order.
7057 The best way to handle static constructors works only for object file
7058 formats which provide arbitrarily-named sections. A section is set
7059 aside for a list of constructors, and another for a list of destructors.
7060 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
7061 object file that defines an initialization function also puts a word in
7062 the constructor section to point to that function. The linker
7063 accumulates all these words into one contiguous @samp{.ctors} section.
7064 Termination functions are handled similarly.
7066 This method will be chosen as the default by @file{target-def.h} if
7067 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
7068 support arbitrary sections, but does support special designated
7069 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
7070 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
7072 When arbitrary sections are available, there are two variants, depending
7073 upon how the code in @file{crtstuff.c} is called. On systems that
7074 support a @dfn{.init} section which is executed at program startup,
7075 parts of @file{crtstuff.c} are compiled into that section. The
7076 program is linked by the @command{gcc} driver like this:
7079 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
7082 The prologue of a function (@code{__init}) appears in the @code{.init}
7083 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
7084 for the function @code{__fini} in the @dfn{.fini} section. Normally these
7085 files are provided by the operating system or by the GNU C library, but
7086 are provided by GCC for a few targets.
7088 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
7089 compiled from @file{crtstuff.c}. They contain, among other things, code
7090 fragments within the @code{.init} and @code{.fini} sections that branch
7091 to routines in the @code{.text} section. The linker will pull all parts
7092 of a section together, which results in a complete @code{__init} function
7093 that invokes the routines we need at startup.
7095 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
7098 If no init section is available, when GCC compiles any function called
7099 @code{main} (or more accurately, any function designated as a program
7100 entry point by the language front end calling @code{expand_main_function}),
7101 it inserts a procedure call to @code{__main} as the first executable code
7102 after the function prologue. The @code{__main} function is defined
7103 in @file{libgcc2.c} and runs the global constructors.
7105 In file formats that don't support arbitrary sections, there are again
7106 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
7107 and an `a.out' format must be used. In this case,
7108 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
7109 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
7110 and with the address of the void function containing the initialization
7111 code as its value. The GNU linker recognizes this as a request to add
7112 the value to a @dfn{set}; the values are accumulated, and are eventually
7113 placed in the executable as a vector in the format described above, with
7114 a leading (ignored) count and a trailing zero element.
7115 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
7116 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
7117 the compilation of @code{main} to call @code{__main} as above, starting
7118 the initialization process.
7120 The last variant uses neither arbitrary sections nor the GNU linker.
7121 This is preferable when you want to do dynamic linking and when using
7122 file formats which the GNU linker does not support, such as `ECOFF'@. In
7123 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
7124 termination functions are recognized simply by their names. This requires
7125 an extra program in the linkage step, called @command{collect2}. This program
7126 pretends to be the linker, for use with GCC; it does its job by running
7127 the ordinary linker, but also arranges to include the vectors of
7128 initialization and termination functions. These functions are called
7129 via @code{__main} as described above. In order to use this method,
7130 @code{use_collect2} must be defined in the target in @file{config.gcc}.
7133 The following section describes the specific macros that control and
7134 customize the handling of initialization and termination functions.
7137 @node Macros for Initialization
7138 @subsection Macros Controlling Initialization Routines
7140 Here are the macros that control how the compiler handles initialization
7141 and termination functions:
7143 @defmac INIT_SECTION_ASM_OP
7144 If defined, a C string constant, including spacing, for the assembler
7145 operation to identify the following data as initialization code. If not
7146 defined, GCC will assume such a section does not exist. When you are
7147 using special sections for initialization and termination functions, this
7148 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7149 run the initialization functions.
7152 @defmac HAS_INIT_SECTION
7153 If defined, @code{main} will not call @code{__main} as described above.
7154 This macro should be defined for systems that control start-up code
7155 on a symbol-by-symbol basis, such as OSF/1, and should not
7156 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7159 @defmac LD_INIT_SWITCH
7160 If defined, a C string constant for a switch that tells the linker that
7161 the following symbol is an initialization routine.
7164 @defmac LD_FINI_SWITCH
7165 If defined, a C string constant for a switch that tells the linker that
7166 the following symbol is a finalization routine.
7169 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7170 If defined, a C statement that will write a function that can be
7171 automatically called when a shared library is loaded. The function
7172 should call @var{func}, which takes no arguments. If not defined, and
7173 the object format requires an explicit initialization function, then a
7174 function called @code{_GLOBAL__DI} will be generated.
7176 This function and the following one are used by collect2 when linking a
7177 shared library that needs constructors or destructors, or has DWARF2
7178 exception tables embedded in the code.
7181 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7182 If defined, a C statement that will write a function that can be
7183 automatically called when a shared library is unloaded. The function
7184 should call @var{func}, which takes no arguments. If not defined, and
7185 the object format requires an explicit finalization function, then a
7186 function called @code{_GLOBAL__DD} will be generated.
7189 @defmac INVOKE__main
7190 If defined, @code{main} will call @code{__main} despite the presence of
7191 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7192 where the init section is not actually run automatically, but is still
7193 useful for collecting the lists of constructors and destructors.
7196 @defmac SUPPORTS_INIT_PRIORITY
7197 If nonzero, the C++ @code{init_priority} attribute is supported and the
7198 compiler should emit instructions to control the order of initialization
7199 of objects. If zero, the compiler will issue an error message upon
7200 encountering an @code{init_priority} attribute.
7203 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7204 This value is true if the target supports some ``native'' method of
7205 collecting constructors and destructors to be run at startup and exit.
7206 It is false if we must use @command{collect2}.
7209 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7210 If defined, a function that outputs assembler code to arrange to call
7211 the function referenced by @var{symbol} at initialization time.
7213 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7214 no arguments and with no return value. If the target supports initialization
7215 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7216 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7218 If this macro is not defined by the target, a suitable default will
7219 be chosen if (1) the target supports arbitrary section names, (2) the
7220 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7224 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7225 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7226 functions rather than initialization functions.
7229 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7230 generated for the generated object file will have static linkage.
7232 If your system uses @command{collect2} as the means of processing
7233 constructors, then that program normally uses @command{nm} to scan
7234 an object file for constructor functions to be called.
7236 On certain kinds of systems, you can define this macro to make
7237 @command{collect2} work faster (and, in some cases, make it work at all):
7239 @defmac OBJECT_FORMAT_COFF
7240 Define this macro if the system uses COFF (Common Object File Format)
7241 object files, so that @command{collect2} can assume this format and scan
7242 object files directly for dynamic constructor/destructor functions.
7244 This macro is effective only in a native compiler; @command{collect2} as
7245 part of a cross compiler always uses @command{nm} for the target machine.
7248 @defmac REAL_NM_FILE_NAME
7249 Define this macro as a C string constant containing the file name to use
7250 to execute @command{nm}. The default is to search the path normally for
7253 If your system supports shared libraries and has a program to list the
7254 dynamic dependencies of a given library or executable, you can define
7255 these macros to enable support for running initialization and
7256 termination functions in shared libraries:
7260 Define this macro to a C string constant containing the name of the program
7261 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7264 @defmac PARSE_LDD_OUTPUT (@var{ptr})
7265 Define this macro to be C code that extracts filenames from the output
7266 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7267 of type @code{char *} that points to the beginning of a line of output
7268 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7269 code must advance @var{ptr} to the beginning of the filename on that
7270 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7273 @node Instruction Output
7274 @subsection Output of Assembler Instructions
7276 @c prevent bad page break with this line
7277 This describes assembler instruction output.
7279 @defmac REGISTER_NAMES
7280 A C initializer containing the assembler's names for the machine
7281 registers, each one as a C string constant. This is what translates
7282 register numbers in the compiler into assembler language.
7285 @defmac ADDITIONAL_REGISTER_NAMES
7286 If defined, a C initializer for an array of structures containing a name
7287 and a register number. This macro defines additional names for hard
7288 registers, thus allowing the @code{asm} option in declarations to refer
7289 to registers using alternate names.
7292 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7293 Define this macro if you are using an unusual assembler that
7294 requires different names for the machine instructions.
7296 The definition is a C statement or statements which output an
7297 assembler instruction opcode to the stdio stream @var{stream}. The
7298 macro-operand @var{ptr} is a variable of type @code{char *} which
7299 points to the opcode name in its ``internal'' form---the form that is
7300 written in the machine description. The definition should output the
7301 opcode name to @var{stream}, performing any translation you desire, and
7302 increment the variable @var{ptr} to point at the end of the opcode
7303 so that it will not be output twice.
7305 In fact, your macro definition may process less than the entire opcode
7306 name, or more than the opcode name; but if you want to process text
7307 that includes @samp{%}-sequences to substitute operands, you must take
7308 care of the substitution yourself. Just be sure to increment
7309 @var{ptr} over whatever text should not be output normally.
7311 @findex recog_data.operand
7312 If you need to look at the operand values, they can be found as the
7313 elements of @code{recog_data.operand}.
7315 If the macro definition does nothing, the instruction is output
7319 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7320 If defined, a C statement to be executed just prior to the output of
7321 assembler code for @var{insn}, to modify the extracted operands so
7322 they will be output differently.
7324 Here the argument @var{opvec} is the vector containing the operands
7325 extracted from @var{insn}, and @var{noperands} is the number of
7326 elements of the vector which contain meaningful data for this insn.
7327 The contents of this vector are what will be used to convert the insn
7328 template into assembler code, so you can change the assembler output
7329 by changing the contents of the vector.
7331 This macro is useful when various assembler syntaxes share a single
7332 file of instruction patterns; by defining this macro differently, you
7333 can cause a large class of instructions to be output differently (such
7334 as with rearranged operands). Naturally, variations in assembler
7335 syntax affecting individual insn patterns ought to be handled by
7336 writing conditional output routines in those patterns.
7338 If this macro is not defined, it is equivalent to a null statement.
7341 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
7342 A C compound statement to output to stdio stream @var{stream} the
7343 assembler syntax for an instruction operand @var{x}. @var{x} is an
7346 @var{code} is a value that can be used to specify one of several ways
7347 of printing the operand. It is used when identical operands must be
7348 printed differently depending on the context. @var{code} comes from
7349 the @samp{%} specification that was used to request printing of the
7350 operand. If the specification was just @samp{%@var{digit}} then
7351 @var{code} is 0; if the specification was @samp{%@var{ltr}
7352 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
7355 If @var{x} is a register, this macro should print the register's name.
7356 The names can be found in an array @code{reg_names} whose type is
7357 @code{char *[]}. @code{reg_names} is initialized from
7358 @code{REGISTER_NAMES}.
7360 When the machine description has a specification @samp{%@var{punct}}
7361 (a @samp{%} followed by a punctuation character), this macro is called
7362 with a null pointer for @var{x} and the punctuation character for
7366 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
7367 A C expression which evaluates to true if @var{code} is a valid
7368 punctuation character for use in the @code{PRINT_OPERAND} macro. If
7369 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
7370 punctuation characters (except for the standard one, @samp{%}) are used
7374 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
7375 A C compound statement to output to stdio stream @var{stream} the
7376 assembler syntax for an instruction operand that is a memory reference
7377 whose address is @var{x}. @var{x} is an RTL expression.
7379 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
7380 On some machines, the syntax for a symbolic address depends on the
7381 section that the address refers to. On these machines, define the hook
7382 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
7383 @code{symbol_ref}, and then check for it here. @xref{Assembler
7387 @findex dbr_sequence_length
7388 @defmac DBR_OUTPUT_SEQEND (@var{file})
7389 A C statement, to be executed after all slot-filler instructions have
7390 been output. If necessary, call @code{dbr_sequence_length} to
7391 determine the number of slots filled in a sequence (zero if not
7392 currently outputting a sequence), to decide how many no-ops to output,
7395 Don't define this macro if it has nothing to do, but it is helpful in
7396 reading assembly output if the extent of the delay sequence is made
7397 explicit (e.g.@: with white space).
7400 @findex final_sequence
7401 Note that output routines for instructions with delay slots must be
7402 prepared to deal with not being output as part of a sequence
7403 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
7404 found.) The variable @code{final_sequence} is null when not
7405 processing a sequence, otherwise it contains the @code{sequence} rtx
7409 @defmac REGISTER_PREFIX
7410 @defmacx LOCAL_LABEL_PREFIX
7411 @defmacx USER_LABEL_PREFIX
7412 @defmacx IMMEDIATE_PREFIX
7413 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
7414 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
7415 @file{final.c}). These are useful when a single @file{md} file must
7416 support multiple assembler formats. In that case, the various @file{tm.h}
7417 files can define these macros differently.
7420 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
7421 If defined this macro should expand to a series of @code{case}
7422 statements which will be parsed inside the @code{switch} statement of
7423 the @code{asm_fprintf} function. This allows targets to define extra
7424 printf formats which may useful when generating their assembler
7425 statements. Note that uppercase letters are reserved for future
7426 generic extensions to asm_fprintf, and so are not available to target
7427 specific code. The output file is given by the parameter @var{file}.
7428 The varargs input pointer is @var{argptr} and the rest of the format
7429 string, starting the character after the one that is being switched
7430 upon, is pointed to by @var{format}.
7433 @defmac ASSEMBLER_DIALECT
7434 If your target supports multiple dialects of assembler language (such as
7435 different opcodes), define this macro as a C expression that gives the
7436 numeric index of the assembler language dialect to use, with zero as the
7439 If this macro is defined, you may use constructs of the form
7441 @samp{@{option0|option1|option2@dots{}@}}
7444 in the output templates of patterns (@pxref{Output Template}) or in the
7445 first argument of @code{asm_fprintf}. This construct outputs
7446 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
7447 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
7448 within these strings retain their usual meaning. If there are fewer
7449 alternatives within the braces than the value of
7450 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
7452 If you do not define this macro, the characters @samp{@{}, @samp{|} and
7453 @samp{@}} do not have any special meaning when used in templates or
7454 operands to @code{asm_fprintf}.
7456 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
7457 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
7458 the variations in assembler language syntax with that mechanism. Define
7459 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
7460 if the syntax variant are larger and involve such things as different
7461 opcodes or operand order.
7464 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
7465 A C expression to output to @var{stream} some assembler code
7466 which will push hard register number @var{regno} onto the stack.
7467 The code need not be optimal, since this macro is used only when
7471 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
7472 A C expression to output to @var{stream} some assembler code
7473 which will pop hard register number @var{regno} off of the stack.
7474 The code need not be optimal, since this macro is used only when
7478 @node Dispatch Tables
7479 @subsection Output of Dispatch Tables
7481 @c prevent bad page break with this line
7482 This concerns dispatch tables.
7484 @cindex dispatch table
7485 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
7486 A C statement to output to the stdio stream @var{stream} an assembler
7487 pseudo-instruction to generate a difference between two labels.
7488 @var{value} and @var{rel} are the numbers of two internal labels. The
7489 definitions of these labels are output using
7490 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
7491 way here. For example,
7494 fprintf (@var{stream}, "\t.word L%d-L%d\n",
7495 @var{value}, @var{rel})
7498 You must provide this macro on machines where the addresses in a
7499 dispatch table are relative to the table's own address. If defined, GCC
7500 will also use this macro on all machines when producing PIC@.
7501 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
7502 mode and flags can be read.
7505 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
7506 This macro should be provided on machines where the addresses
7507 in a dispatch table are absolute.
7509 The definition should be a C statement to output to the stdio stream
7510 @var{stream} an assembler pseudo-instruction to generate a reference to
7511 a label. @var{value} is the number of an internal label whose
7512 definition is output using @code{(*targetm.asm_out.internal_label)}.
7516 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
7520 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
7521 Define this if the label before a jump-table needs to be output
7522 specially. The first three arguments are the same as for
7523 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
7524 jump-table which follows (a @code{jump_insn} containing an
7525 @code{addr_vec} or @code{addr_diff_vec}).
7527 This feature is used on system V to output a @code{swbeg} statement
7530 If this macro is not defined, these labels are output with
7531 @code{(*targetm.asm_out.internal_label)}.
7534 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
7535 Define this if something special must be output at the end of a
7536 jump-table. The definition should be a C statement to be executed
7537 after the assembler code for the table is written. It should write
7538 the appropriate code to stdio stream @var{stream}. The argument
7539 @var{table} is the jump-table insn, and @var{num} is the label-number
7540 of the preceding label.
7542 If this macro is not defined, nothing special is output at the end of
7546 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (@var{stream}, @var{decl}, @var{for_eh}, @var{empty})
7547 This target hook emits a label at the beginning of each FDE@. It
7548 should be defined on targets where FDEs need special labels, and it
7549 should write the appropriate label, for the FDE associated with the
7550 function declaration @var{decl}, to the stdio stream @var{stream}.
7551 The third argument, @var{for_eh}, is a boolean: true if this is for an
7552 exception table. The fourth argument, @var{empty}, is a boolean:
7553 true if this is a placeholder label for an omitted FDE@.
7555 The default is that FDEs are not given nonlocal labels.
7558 @deftypefn {Taget Hook} void TARGET_UNWIND_EMIT (FILE * @var{stream}, rtx @var{insn})
7559 This target hook emits and assembly directives required to unwind the
7560 given instruction. This is only used when TARGET_UNWIND_INFO is set.
7563 @node Exception Region Output
7564 @subsection Assembler Commands for Exception Regions
7566 @c prevent bad page break with this line
7568 This describes commands marking the start and the end of an exception
7571 @defmac EH_FRAME_SECTION_NAME
7572 If defined, a C string constant for the name of the section containing
7573 exception handling frame unwind information. If not defined, GCC will
7574 provide a default definition if the target supports named sections.
7575 @file{crtstuff.c} uses this macro to switch to the appropriate section.
7577 You should define this symbol if your target supports DWARF 2 frame
7578 unwind information and the default definition does not work.
7581 @defmac EH_FRAME_IN_DATA_SECTION
7582 If defined, DWARF 2 frame unwind information will be placed in the
7583 data section even though the target supports named sections. This
7584 might be necessary, for instance, if the system linker does garbage
7585 collection and sections cannot be marked as not to be collected.
7587 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
7591 @defmac EH_TABLES_CAN_BE_READ_ONLY
7592 Define this macro to 1 if your target is such that no frame unwind
7593 information encoding used with non-PIC code will ever require a
7594 runtime relocation, but the linker may not support merging read-only
7595 and read-write sections into a single read-write section.
7598 @defmac MASK_RETURN_ADDR
7599 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
7600 that it does not contain any extraneous set bits in it.
7603 @defmac DWARF2_UNWIND_INFO
7604 Define this macro to 0 if your target supports DWARF 2 frame unwind
7605 information, but it does not yet work with exception handling.
7606 Otherwise, if your target supports this information (if it defines
7607 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
7608 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
7611 If @code{TARGET_UNWIND_INFO} is defined, the target specific unwinder
7612 will be used in all cases. Defining this macro will enable the generation
7613 of DWARF 2 frame debugging information.
7615 If @code{TARGET_UNWIND_INFO} is not defined, and this macro is defined to 1,
7616 the DWARF 2 unwinder will be the default exception handling mechanism;
7617 otherwise, @code{setjmp}/@code{longjmp} will be used by default.
7620 @defmac TARGET_UNWIND_INFO
7621 Define this macro if your target has ABI specified unwind tables. Usually
7622 these will be output by @code{TARGET_UNWIND_EMIT}.
7625 @defmac MUST_USE_SJLJ_EXCEPTIONS
7626 This macro need only be defined if @code{DWARF2_UNWIND_INFO} is
7627 runtime-variable. In that case, @file{except.h} cannot correctly
7628 determine the corresponding definition of
7629 @code{MUST_USE_SJLJ_EXCEPTIONS}, so the target must provide it directly.
7632 @defmac DWARF_CIE_DATA_ALIGNMENT
7633 This macro need only be defined if the target might save registers in the
7634 function prologue at an offset to the stack pointer that is not aligned to
7635 @code{UNITS_PER_WORD}. The definition should be the negative minimum
7636 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
7637 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
7638 the target supports DWARF 2 frame unwind information.
7641 @deftypefn {Target Hook} void TARGET_ASM_EXCEPTION_SECTION ()
7642 If defined, a function that switches to the section in which the main
7643 exception table is to be placed (@pxref{Sections}). The default is a
7644 function that switches to a section named @code{.gcc_except_table} on
7645 machines that support named sections via
7646 @code{TARGET_ASM_NAMED_SECTION}, otherwise if @option{-fpic} or
7647 @option{-fPIC} is in effect, the @code{data_section}, otherwise the
7648 @code{readonly_data_section}.
7651 @deftypefn {Target Hook} void TARGET_ASM_EH_FRAME_SECTION ()
7652 If defined, a function that switches to the section in which the DWARF 2
7653 frame unwind information to be placed (@pxref{Sections}). The default
7654 is a function that outputs a standard GAS section directive, if
7655 @code{EH_FRAME_SECTION_NAME} is defined, or else a data section
7656 directive followed by a synthetic label.
7659 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
7660 Contains the value true if the target should add a zero word onto the
7661 end of a Dwarf-2 frame info section when used for exception handling.
7662 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
7666 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
7667 Given a register, this hook should return a parallel of registers to
7668 represent where to find the register pieces. Define this hook if the
7669 register and its mode are represented in Dwarf in non-contiguous
7670 locations, or if the register should be represented in more than one
7671 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
7672 If not defined, the default is to return @code{NULL_RTX}.
7675 @node Alignment Output
7676 @subsection Assembler Commands for Alignment
7678 @c prevent bad page break with this line
7679 This describes commands for alignment.
7681 @defmac JUMP_ALIGN (@var{label})
7682 The alignment (log base 2) to put in front of @var{label}, which is
7683 a common destination of jumps and has no fallthru incoming edge.
7685 This macro need not be defined if you don't want any special alignment
7686 to be done at such a time. Most machine descriptions do not currently
7689 Unless it's necessary to inspect the @var{label} parameter, it is better
7690 to set the variable @var{align_jumps} in the target's
7691 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7692 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
7695 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
7696 The alignment (log base 2) to put in front of @var{label}, which follows
7699 This macro need not be defined if you don't want any special alignment
7700 to be done at such a time. Most machine descriptions do not currently
7704 @defmac LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7705 The maximum number of bytes to skip when applying
7706 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
7707 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7710 @defmac LOOP_ALIGN (@var{label})
7711 The alignment (log base 2) to put in front of @var{label}, which follows
7712 a @code{NOTE_INSN_LOOP_BEG} note.
7714 This macro need not be defined if you don't want any special alignment
7715 to be done at such a time. Most machine descriptions do not currently
7718 Unless it's necessary to inspect the @var{label} parameter, it is better
7719 to set the variable @code{align_loops} in the target's
7720 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7721 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
7724 @defmac LOOP_ALIGN_MAX_SKIP
7725 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
7726 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7729 @defmac LABEL_ALIGN (@var{label})
7730 The alignment (log base 2) to put in front of @var{label}.
7731 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
7732 the maximum of the specified values is used.
7734 Unless it's necessary to inspect the @var{label} parameter, it is better
7735 to set the variable @code{align_labels} in the target's
7736 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7737 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
7740 @defmac LABEL_ALIGN_MAX_SKIP
7741 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
7742 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7745 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
7746 A C statement to output to the stdio stream @var{stream} an assembler
7747 instruction to advance the location counter by @var{nbytes} bytes.
7748 Those bytes should be zero when loaded. @var{nbytes} will be a C
7749 expression of type @code{int}.
7752 @defmac ASM_NO_SKIP_IN_TEXT
7753 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
7754 text section because it fails to put zeros in the bytes that are skipped.
7755 This is true on many Unix systems, where the pseudo--op to skip bytes
7756 produces no-op instructions rather than zeros when used in the text
7760 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
7761 A C statement to output to the stdio stream @var{stream} an assembler
7762 command to advance the location counter to a multiple of 2 to the
7763 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
7766 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
7767 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
7768 for padding, if necessary.
7771 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
7772 A C statement to output to the stdio stream @var{stream} an assembler
7773 command to advance the location counter to a multiple of 2 to the
7774 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
7775 satisfy the alignment request. @var{power} and @var{max_skip} will be
7776 a C expression of type @code{int}.
7780 @node Debugging Info
7781 @section Controlling Debugging Information Format
7783 @c prevent bad page break with this line
7784 This describes how to specify debugging information.
7787 * All Debuggers:: Macros that affect all debugging formats uniformly.
7788 * DBX Options:: Macros enabling specific options in DBX format.
7789 * DBX Hooks:: Hook macros for varying DBX format.
7790 * File Names and DBX:: Macros controlling output of file names in DBX format.
7791 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
7792 * VMS Debug:: Macros for VMS debug format.
7796 @subsection Macros Affecting All Debugging Formats
7798 @c prevent bad page break with this line
7799 These macros affect all debugging formats.
7801 @defmac DBX_REGISTER_NUMBER (@var{regno})
7802 A C expression that returns the DBX register number for the compiler
7803 register number @var{regno}. In the default macro provided, the value
7804 of this expression will be @var{regno} itself. But sometimes there are
7805 some registers that the compiler knows about and DBX does not, or vice
7806 versa. In such cases, some register may need to have one number in the
7807 compiler and another for DBX@.
7809 If two registers have consecutive numbers inside GCC, and they can be
7810 used as a pair to hold a multiword value, then they @emph{must} have
7811 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
7812 Otherwise, debuggers will be unable to access such a pair, because they
7813 expect register pairs to be consecutive in their own numbering scheme.
7815 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
7816 does not preserve register pairs, then what you must do instead is
7817 redefine the actual register numbering scheme.
7820 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
7821 A C expression that returns the integer offset value for an automatic
7822 variable having address @var{x} (an RTL expression). The default
7823 computation assumes that @var{x} is based on the frame-pointer and
7824 gives the offset from the frame-pointer. This is required for targets
7825 that produce debugging output for DBX or COFF-style debugging output
7826 for SDB and allow the frame-pointer to be eliminated when the
7827 @option{-g} options is used.
7830 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
7831 A C expression that returns the integer offset value for an argument
7832 having address @var{x} (an RTL expression). The nominal offset is
7836 @defmac PREFERRED_DEBUGGING_TYPE
7837 A C expression that returns the type of debugging output GCC should
7838 produce when the user specifies just @option{-g}. Define
7839 this if you have arranged for GCC to support more than one format of
7840 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7841 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
7842 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
7844 When the user specifies @option{-ggdb}, GCC normally also uses the
7845 value of this macro to select the debugging output format, but with two
7846 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
7847 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7848 defined, GCC uses @code{DBX_DEBUG}.
7850 The value of this macro only affects the default debugging output; the
7851 user can always get a specific type of output by using @option{-gstabs},
7852 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
7856 @subsection Specific Options for DBX Output
7858 @c prevent bad page break with this line
7859 These are specific options for DBX output.
7861 @defmac DBX_DEBUGGING_INFO
7862 Define this macro if GCC should produce debugging output for DBX
7863 in response to the @option{-g} option.
7866 @defmac XCOFF_DEBUGGING_INFO
7867 Define this macro if GCC should produce XCOFF format debugging output
7868 in response to the @option{-g} option. This is a variant of DBX format.
7871 @defmac DEFAULT_GDB_EXTENSIONS
7872 Define this macro to control whether GCC should by default generate
7873 GDB's extended version of DBX debugging information (assuming DBX-format
7874 debugging information is enabled at all). If you don't define the
7875 macro, the default is 1: always generate the extended information
7876 if there is any occasion to.
7879 @defmac DEBUG_SYMS_TEXT
7880 Define this macro if all @code{.stabs} commands should be output while
7881 in the text section.
7884 @defmac ASM_STABS_OP
7885 A C string constant, including spacing, naming the assembler pseudo op to
7886 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
7887 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
7888 applies only to DBX debugging information format.
7891 @defmac ASM_STABD_OP
7892 A C string constant, including spacing, naming the assembler pseudo op to
7893 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
7894 value is the current location. If you don't define this macro,
7895 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
7899 @defmac ASM_STABN_OP
7900 A C string constant, including spacing, naming the assembler pseudo op to
7901 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
7902 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
7903 macro applies only to DBX debugging information format.
7906 @defmac DBX_NO_XREFS
7907 Define this macro if DBX on your system does not support the construct
7908 @samp{xs@var{tagname}}. On some systems, this construct is used to
7909 describe a forward reference to a structure named @var{tagname}.
7910 On other systems, this construct is not supported at all.
7913 @defmac DBX_CONTIN_LENGTH
7914 A symbol name in DBX-format debugging information is normally
7915 continued (split into two separate @code{.stabs} directives) when it
7916 exceeds a certain length (by default, 80 characters). On some
7917 operating systems, DBX requires this splitting; on others, splitting
7918 must not be done. You can inhibit splitting by defining this macro
7919 with the value zero. You can override the default splitting-length by
7920 defining this macro as an expression for the length you desire.
7923 @defmac DBX_CONTIN_CHAR
7924 Normally continuation is indicated by adding a @samp{\} character to
7925 the end of a @code{.stabs} string when a continuation follows. To use
7926 a different character instead, define this macro as a character
7927 constant for the character you want to use. Do not define this macro
7928 if backslash is correct for your system.
7931 @defmac DBX_STATIC_STAB_DATA_SECTION
7932 Define this macro if it is necessary to go to the data section before
7933 outputting the @samp{.stabs} pseudo-op for a non-global static
7937 @defmac DBX_TYPE_DECL_STABS_CODE
7938 The value to use in the ``code'' field of the @code{.stabs} directive
7939 for a typedef. The default is @code{N_LSYM}.
7942 @defmac DBX_STATIC_CONST_VAR_CODE
7943 The value to use in the ``code'' field of the @code{.stabs} directive
7944 for a static variable located in the text section. DBX format does not
7945 provide any ``right'' way to do this. The default is @code{N_FUN}.
7948 @defmac DBX_REGPARM_STABS_CODE
7949 The value to use in the ``code'' field of the @code{.stabs} directive
7950 for a parameter passed in registers. DBX format does not provide any
7951 ``right'' way to do this. The default is @code{N_RSYM}.
7954 @defmac DBX_REGPARM_STABS_LETTER
7955 The letter to use in DBX symbol data to identify a symbol as a parameter
7956 passed in registers. DBX format does not customarily provide any way to
7957 do this. The default is @code{'P'}.
7960 @defmac DBX_FUNCTION_FIRST
7961 Define this macro if the DBX information for a function and its
7962 arguments should precede the assembler code for the function. Normally,
7963 in DBX format, the debugging information entirely follows the assembler
7967 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
7968 Define this macro, with value 1, if the value of a symbol describing
7969 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
7970 relative to the start of the enclosing function. Normally, GCC uses
7971 an absolute address.
7974 @defmac DBX_LINES_FUNCTION_RELATIVE
7975 Define this macro, with value 1, if the value of a symbol indicating
7976 the current line number (@code{N_SLINE}) should be relative to the
7977 start of the enclosing function. Normally, GCC uses an absolute address.
7980 @defmac DBX_USE_BINCL
7981 Define this macro if GCC should generate @code{N_BINCL} and
7982 @code{N_EINCL} stabs for included header files, as on Sun systems. This
7983 macro also directs GCC to output a type number as a pair of a file
7984 number and a type number within the file. Normally, GCC does not
7985 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
7986 number for a type number.
7990 @subsection Open-Ended Hooks for DBX Format
7992 @c prevent bad page break with this line
7993 These are hooks for DBX format.
7995 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
7996 Define this macro to say how to output to @var{stream} the debugging
7997 information for the start of a scope level for variable names. The
7998 argument @var{name} is the name of an assembler symbol (for use with
7999 @code{assemble_name}) whose value is the address where the scope begins.
8002 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
8003 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
8006 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
8007 Define this macro if the target machine requires special handling to
8008 output an @code{N_FUN} entry for the function @var{decl}.
8011 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
8012 A C statement to output DBX debugging information before code for line
8013 number @var{line} of the current source file to the stdio stream
8014 @var{stream}. @var{counter} is the number of time the macro was
8015 invoked, including the current invocation; it is intended to generate
8016 unique labels in the assembly output.
8018 This macro should not be defined if the default output is correct, or
8019 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
8022 @defmac NO_DBX_FUNCTION_END
8023 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
8024 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
8025 On those machines, define this macro to turn this feature off without
8026 disturbing the rest of the gdb extensions.
8029 @defmac NO_DBX_BNSYM_ENSYM
8030 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
8031 extension construct. On those machines, define this macro to turn this
8032 feature off without disturbing the rest of the gdb extensions.
8035 @node File Names and DBX
8036 @subsection File Names in DBX Format
8038 @c prevent bad page break with this line
8039 This describes file names in DBX format.
8041 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
8042 A C statement to output DBX debugging information to the stdio stream
8043 @var{stream}, which indicates that file @var{name} is the main source
8044 file---the file specified as the input file for compilation.
8045 This macro is called only once, at the beginning of compilation.
8047 This macro need not be defined if the standard form of output
8048 for DBX debugging information is appropriate.
8050 It may be necessary to refer to a label equal to the beginning of the
8051 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
8052 to do so. If you do this, you must also set the variable
8053 @var{used_ltext_label_name} to @code{true}.
8056 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
8057 Define this macro, with value 1, if GCC should not emit an indication
8058 of the current directory for compilation and current source language at
8059 the beginning of the file.
8062 @defmac NO_DBX_GCC_MARKER
8063 Define this macro, with value 1, if GCC should not emit an indication
8064 that this object file was compiled by GCC@. The default is to emit
8065 an @code{N_OPT} stab at the beginning of every source file, with
8066 @samp{gcc2_compiled.} for the string and value 0.
8069 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
8070 A C statement to output DBX debugging information at the end of
8071 compilation of the main source file @var{name}. Output should be
8072 written to the stdio stream @var{stream}.
8074 If you don't define this macro, nothing special is output at the end
8075 of compilation, which is correct for most machines.
8078 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
8079 Define this macro @emph{instead of} defining
8080 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
8081 the end of compilation is a @code{N_SO} stab with an empty string,
8082 whose value is the highest absolute text address in the file.
8087 @subsection Macros for SDB and DWARF Output
8089 @c prevent bad page break with this line
8090 Here are macros for SDB and DWARF output.
8092 @defmac SDB_DEBUGGING_INFO
8093 Define this macro if GCC should produce COFF-style debugging output
8094 for SDB in response to the @option{-g} option.
8097 @defmac DWARF2_DEBUGGING_INFO
8098 Define this macro if GCC should produce dwarf version 2 format
8099 debugging output in response to the @option{-g} option.
8101 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (tree @var{function})
8102 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
8103 be emitted for each function. Instead of an integer return the enum
8104 value for the @code{DW_CC_} tag.
8107 To support optional call frame debugging information, you must also
8108 define @code{INCOMING_RETURN_ADDR_RTX} and either set
8109 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
8110 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
8111 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
8114 @defmac DWARF2_FRAME_INFO
8115 Define this macro to a nonzero value if GCC should always output
8116 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
8117 (@pxref{Exception Region Output} is nonzero, GCC will output this
8118 information not matter how you define @code{DWARF2_FRAME_INFO}.
8121 @defmac DWARF2_ASM_LINE_DEBUG_INFO
8122 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8123 line debug info sections. This will result in much more compact line number
8124 tables, and hence is desirable if it works.
8127 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8128 A C statement to issue assembly directives that create a difference
8129 between the two given labels, using an integer of the given size.
8132 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label})
8133 A C statement to issue assembly directives that create a
8134 section-relative reference to the given label, using an integer of the
8138 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8139 A C statement to issue assembly directives that create a self-relative
8140 reference to the given label, using an integer of the given size.
8143 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{FILE}, int @var{size}, rtx @var{x})
8144 If defined, this target hook is a function which outputs a DTP-relative
8145 reference to the given TLS symbol of the specified size.
8148 @defmac PUT_SDB_@dots{}
8149 Define these macros to override the assembler syntax for the special
8150 SDB assembler directives. See @file{sdbout.c} for a list of these
8151 macros and their arguments. If the standard syntax is used, you need
8152 not define them yourself.
8156 Some assemblers do not support a semicolon as a delimiter, even between
8157 SDB assembler directives. In that case, define this macro to be the
8158 delimiter to use (usually @samp{\n}). It is not necessary to define
8159 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8163 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
8164 Define this macro to allow references to unknown structure,
8165 union, or enumeration tags to be emitted. Standard COFF does not
8166 allow handling of unknown references, MIPS ECOFF has support for
8170 @defmac SDB_ALLOW_FORWARD_REFERENCES
8171 Define this macro to allow references to structure, union, or
8172 enumeration tags that have not yet been seen to be handled. Some
8173 assemblers choke if forward tags are used, while some require it.
8176 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
8177 A C statement to output SDB debugging information before code for line
8178 number @var{line} of the current source file to the stdio stream
8179 @var{stream}. The default is to emit an @code{.ln} directive.
8184 @subsection Macros for VMS Debug Format
8186 @c prevent bad page break with this line
8187 Here are macros for VMS debug format.
8189 @defmac VMS_DEBUGGING_INFO
8190 Define this macro if GCC should produce debugging output for VMS
8191 in response to the @option{-g} option. The default behavior for VMS
8192 is to generate minimal debug info for a traceback in the absence of
8193 @option{-g} unless explicitly overridden with @option{-g0}. This
8194 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8195 @code{OVERRIDE_OPTIONS}.
8198 @node Floating Point
8199 @section Cross Compilation and Floating Point
8200 @cindex cross compilation and floating point
8201 @cindex floating point and cross compilation
8203 While all modern machines use twos-complement representation for integers,
8204 there are a variety of representations for floating point numbers. This
8205 means that in a cross-compiler the representation of floating point numbers
8206 in the compiled program may be different from that used in the machine
8207 doing the compilation.
8209 Because different representation systems may offer different amounts of
8210 range and precision, all floating point constants must be represented in
8211 the target machine's format. Therefore, the cross compiler cannot
8212 safely use the host machine's floating point arithmetic; it must emulate
8213 the target's arithmetic. To ensure consistency, GCC always uses
8214 emulation to work with floating point values, even when the host and
8215 target floating point formats are identical.
8217 The following macros are provided by @file{real.h} for the compiler to
8218 use. All parts of the compiler which generate or optimize
8219 floating-point calculations must use these macros. They may evaluate
8220 their operands more than once, so operands must not have side effects.
8222 @defmac REAL_VALUE_TYPE
8223 The C data type to be used to hold a floating point value in the target
8224 machine's format. Typically this is a @code{struct} containing an
8225 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8229 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8230 Compares for equality the two values, @var{x} and @var{y}. If the target
8231 floating point format supports negative zeroes and/or NaNs,
8232 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8233 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8236 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8237 Tests whether @var{x} is less than @var{y}.
8240 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8241 Truncates @var{x} to a signed integer, rounding toward zero.
8244 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8245 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8246 @var{x} is negative, returns zero.
8249 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8250 Converts @var{string} into a floating point number in the target machine's
8251 representation for mode @var{mode}. This routine can handle both
8252 decimal and hexadecimal floating point constants, using the syntax
8253 defined by the C language for both.
8256 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8257 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8260 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8261 Determines whether @var{x} represents infinity (positive or negative).
8264 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8265 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8268 @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})
8269 Calculates an arithmetic operation on the two floating point values
8270 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8273 The operation to be performed is specified by @var{code}. Only the
8274 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8275 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8277 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8278 target's floating point format cannot represent infinity, it will call
8279 @code{abort}. Callers should check for this situation first, using
8280 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8283 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8284 Returns the negative of the floating point value @var{x}.
8287 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8288 Returns the absolute value of @var{x}.
8291 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8292 Truncates the floating point value @var{x} to fit in @var{mode}. The
8293 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8294 appropriate bit pattern to be output asa floating constant whose
8295 precision accords with mode @var{mode}.
8298 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8299 Converts a floating point value @var{x} into a double-precision integer
8300 which is then stored into @var{low} and @var{high}. If the value is not
8301 integral, it is truncated.
8304 @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})
8305 Converts a double-precision integer found in @var{low} and @var{high},
8306 into a floating point value which is then stored into @var{x}. The
8307 value is truncated to fit in mode @var{mode}.
8310 @node Mode Switching
8311 @section Mode Switching Instructions
8312 @cindex mode switching
8313 The following macros control mode switching optimizations:
8315 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
8316 Define this macro if the port needs extra instructions inserted for mode
8317 switching in an optimizing compilation.
8319 For an example, the SH4 can perform both single and double precision
8320 floating point operations, but to perform a single precision operation,
8321 the FPSCR PR bit has to be cleared, while for a double precision
8322 operation, this bit has to be set. Changing the PR bit requires a general
8323 purpose register as a scratch register, hence these FPSCR sets have to
8324 be inserted before reload, i.e.@: you can't put this into instruction emitting
8325 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
8327 You can have multiple entities that are mode-switched, and select at run time
8328 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
8329 return nonzero for any @var{entity} that needs mode-switching.
8330 If you define this macro, you also have to define
8331 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
8332 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
8333 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
8337 @defmac NUM_MODES_FOR_MODE_SWITCHING
8338 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
8339 initializer for an array of integers. Each initializer element
8340 N refers to an entity that needs mode switching, and specifies the number
8341 of different modes that might need to be set for this entity.
8342 The position of the initializer in the initializer---starting counting at
8343 zero---determines the integer that is used to refer to the mode-switched
8345 In macros that take mode arguments / yield a mode result, modes are
8346 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
8347 switch is needed / supplied.
8350 @defmac MODE_NEEDED (@var{entity}, @var{insn})
8351 @var{entity} is an integer specifying a mode-switched entity. If
8352 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
8353 return an integer value not larger than the corresponding element in
8354 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
8355 be switched into prior to the execution of @var{insn}.
8358 @defmac MODE_AFTER (@var{mode}, @var{insn})
8359 If this macro is defined, it is evaluated for every @var{insn} during
8360 mode switching. It determines the mode that an insn results in (if
8361 different from the incoming mode).
8364 @defmac MODE_ENTRY (@var{entity})
8365 If this macro is defined, it is evaluated for every @var{entity} that needs
8366 mode switching. It should evaluate to an integer, which is a mode that
8367 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
8368 is defined then @code{MODE_EXIT} must be defined.
8371 @defmac MODE_EXIT (@var{entity})
8372 If this macro is defined, it is evaluated for every @var{entity} that needs
8373 mode switching. It should evaluate to an integer, which is a mode that
8374 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
8375 is defined then @code{MODE_ENTRY} must be defined.
8378 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
8379 This macro specifies the order in which modes for @var{entity} are processed.
8380 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
8381 lowest. The value of the macro should be an integer designating a mode
8382 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
8383 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
8384 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
8387 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
8388 Generate one or more insns to set @var{entity} to @var{mode}.
8389 @var{hard_reg_live} is the set of hard registers live at the point where
8390 the insn(s) are to be inserted.
8393 @node Target Attributes
8394 @section Defining target-specific uses of @code{__attribute__}
8395 @cindex target attributes
8396 @cindex machine attributes
8397 @cindex attributes, target-specific
8399 Target-specific attributes may be defined for functions, data and types.
8400 These are described using the following target hooks; they also need to
8401 be documented in @file{extend.texi}.
8403 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
8404 If defined, this target hook points to an array of @samp{struct
8405 attribute_spec} (defined in @file{tree.h}) specifying the machine
8406 specific attributes for this target and some of the restrictions on the
8407 entities to which these attributes are applied and the arguments they
8411 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8412 If defined, this target hook is a function which returns zero if the attributes on
8413 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8414 and two if they are nearly compatible (which causes a warning to be
8415 generated). If this is not defined, machine-specific attributes are
8416 supposed always to be compatible.
8419 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
8420 If defined, this target hook is a function which assigns default attributes to
8421 newly defined @var{type}.
8424 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8425 Define this target hook if the merging of type attributes needs special
8426 handling. If defined, the result is a list of the combined
8427 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
8428 that @code{comptypes} has already been called and returned 1. This
8429 function may call @code{merge_attributes} to handle machine-independent
8433 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
8434 Define this target hook if the merging of decl attributes needs special
8435 handling. If defined, the result is a list of the combined
8436 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
8437 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
8438 when this is needed are when one attribute overrides another, or when an
8439 attribute is nullified by a subsequent definition. This function may
8440 call @code{merge_attributes} to handle machine-independent merging.
8442 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
8443 If the only target-specific handling you require is @samp{dllimport}
8444 for Microsoft Windows targets, you should define the macro
8445 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
8446 will then define a function called
8447 @code{merge_dllimport_decl_attributes} which can then be defined as
8448 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
8449 add @code{handle_dll_attribute} in the attribute table for your port
8450 to perform initial processing of the @samp{dllimport} and
8451 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
8452 @file{i386/i386.c}, for example.
8455 @defmac TARGET_DECLSPEC
8456 Define this macro to a nonzero value if you want to treat
8457 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
8458 default, this behavior is enabled only for targets that define
8459 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
8460 of @code{__declspec} is via a built-in macro, but you should not rely
8461 on this implementation detail.
8464 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
8465 Define this target hook if you want to be able to add attributes to a decl
8466 when it is being created. This is normally useful for back ends which
8467 wish to implement a pragma by using the attributes which correspond to
8468 the pragma's effect. The @var{node} argument is the decl which is being
8469 created. The @var{attr_ptr} argument is a pointer to the attribute list
8470 for this decl. The list itself should not be modified, since it may be
8471 shared with other decls, but attributes may be chained on the head of
8472 the list and @code{*@var{attr_ptr}} modified to point to the new
8473 attributes, or a copy of the list may be made if further changes are
8477 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
8479 This target hook returns @code{true} if it is ok to inline @var{fndecl}
8480 into the current function, despite its having target-specific
8481 attributes, @code{false} otherwise. By default, if a function has a
8482 target specific attribute attached to it, it will not be inlined.
8485 @node MIPS Coprocessors
8486 @section Defining coprocessor specifics for MIPS targets.
8487 @cindex MIPS coprocessor-definition macros
8489 The MIPS specification allows MIPS implementations to have as many as 4
8490 coprocessors, each with as many as 32 private registers. GCC supports
8491 accessing these registers and transferring values between the registers
8492 and memory using asm-ized variables. For example:
8495 register unsigned int cp0count asm ("c0r1");
8501 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
8502 names may be added as described below, or the default names may be
8503 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
8505 Coprocessor registers are assumed to be epilogue-used; sets to them will
8506 be preserved even if it does not appear that the register is used again
8507 later in the function.
8509 Another note: according to the MIPS spec, coprocessor 1 (if present) is
8510 the FPU@. One accesses COP1 registers through standard mips
8511 floating-point support; they are not included in this mechanism.
8513 There is one macro used in defining the MIPS coprocessor interface which
8514 you may want to override in subtargets; it is described below.
8516 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
8517 A comma-separated list (with leading comma) of pairs describing the
8518 alternate names of coprocessor registers. The format of each entry should be
8520 @{ @var{alternatename}, @var{register_number}@}
8526 @section Parameters for Precompiled Header Validity Checking
8527 @cindex parameters, precompiled headers
8529 @deftypefn {Target Hook} void *TARGET_GET_PCH_VALIDITY (size_t *@var{sz})
8530 This hook returns the data needed by @code{TARGET_PCH_VALID_P} and sets
8531 @samp{*@var{sz}} to the size of the data in bytes.
8534 @deftypefn {Target Hook} const char *TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz})
8535 This hook checks whether the options used to create a PCH file are
8536 compatible with the current settings. It returns @code{NULL}
8537 if so and a suitable error message if not. Error messages will
8538 be presented to the user and must be localized using @samp{_(@var{msg})}.
8540 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
8541 when the PCH file was created and @var{sz} is the size of that data in bytes.
8542 It's safe to assume that the data was created by the same version of the
8543 compiler, so no format checking is needed.
8545 The default definition of @code{default_pch_valid_p} should be
8546 suitable for most targets.
8549 @deftypefn {Target Hook} const char *TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags})
8550 If this hook is nonnull, the default implementation of
8551 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
8552 of @code{target_flags}. @var{pch_flags} specifies the value that
8553 @code{target_flags} had when the PCH file was created. The return
8554 value is the same as for @code{TARGET_PCH_VALID_P}.
8558 @section C++ ABI parameters
8559 @cindex parameters, c++ abi
8561 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
8562 Define this hook to override the integer type used for guard variables.
8563 These are used to implement one-time construction of static objects. The
8564 default is long_long_integer_type_node.
8567 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
8568 This hook determines how guard variables are used. It should return
8569 @code{false} (the default) if first byte should be used. A return value of
8570 @code{true} indicates the least significant bit should be used.
8573 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
8574 This hook returns the size of the cookie to use when allocating an array
8575 whose elements have the indicated @var{type}. Assumes that it is already
8576 known that a cookie is needed. The default is
8577 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
8578 IA64/Generic C++ ABI@.
8581 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
8582 This hook should return @code{true} if the element size should be stored in
8583 array cookies. The default is to return @code{false}.
8586 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
8587 If defined by a backend this hook allows the decision made to export
8588 class @var{type} to be overruled. Upon entry @var{import_export}
8589 will contain 1 if the class is going to be exported, @minus{}1 if it is going
8590 to be imported and 0 otherwise. This function should return the
8591 modified value and perform any other actions necessary to support the
8592 backend's targeted operating system.
8595 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
8596 This hook should return @code{true} if constructors and destructors return
8597 the address of the object created/destroyed. The default is to return
8601 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
8602 This hook returns true if the key method for a class (i.e., the method
8603 which, if defined in the current translation unit, causes the virtual
8604 table to be emitted) may be an inline function. Under the standard
8605 Itanium C++ ABI the key method may be an inline function so long as
8606 the function is not declared inline in the class definition. Under
8607 some variants of the ABI, an inline function can never be the key
8608 method. The default is to return @code{true}.
8611 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
8612 @var{decl} is a virtual table, virtual table table, typeinfo object,
8613 or other similar implicit class data object that will be emitted with
8614 external linkage in this translation unit. No ELF visibility has been
8615 explicitly specified. If the target needs to specify a visibility
8616 other than that of the containing class, use this hook to set
8617 @code{DECL_VISIBILITY} and @code{DECL_VISIBILITY_SPECIFIED}.
8620 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
8621 This hook returns true (the default) if virtual tables and other
8622 similar implicit class data objects are always COMDAT if they have
8623 external linkage. If this hook returns false, then class data for
8624 classes whose virtual table will be emitted in only one translation
8625 unit will not be COMDAT.
8628 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
8629 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
8630 should be used to register static destructors when @option{-fuse-cxa-atexit}
8631 is in effect. The default is to return false to use @code{__cxa_atexit}.
8635 @section Miscellaneous Parameters
8636 @cindex parameters, miscellaneous
8638 @c prevent bad page break with this line
8639 Here are several miscellaneous parameters.
8641 @defmac PREDICATE_CODES
8642 Define this if you have defined special-purpose predicates in the file
8643 @file{@var{machine}.c}. This macro is called within an initializer of an
8644 array of structures. The first field in the structure is the name of a
8645 predicate and the second field is an array of rtl codes. For each
8646 predicate, list all rtl codes that can be in expressions matched by the
8647 predicate. The list should have a trailing comma. Here is an example
8648 of two entries in the list for a typical RISC machine:
8651 #define PREDICATE_CODES \
8652 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
8653 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
8656 Defining this macro does not affect the generated code (however,
8657 incorrect definitions that omit an rtl code that may be matched by the
8658 predicate can cause the compiler to malfunction). Instead, it allows
8659 the table built by @file{genrecog} to be more compact and efficient,
8660 thus speeding up the compiler. The most important predicates to include
8661 in the list specified by this macro are those used in the most insn
8664 For each predicate function named in @code{PREDICATE_CODES}, a
8665 declaration will be generated in @file{insn-codes.h}.
8667 Use of this macro is deprecated; use @code{define_predicate} instead.
8668 @xref{Defining Predicates}.
8671 @defmac SPECIAL_MODE_PREDICATES
8672 Define this if you have special predicates that know special things
8673 about modes. Genrecog will warn about certain forms of
8674 @code{match_operand} without a mode; if the operand predicate is
8675 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
8678 Here is an example from the IA-32 port (@code{ext_register_operand}
8679 specially checks for @code{HImode} or @code{SImode} in preparation
8680 for a byte extraction from @code{%ah} etc.).
8683 #define SPECIAL_MODE_PREDICATES \
8684 "ext_register_operand",
8687 Use of this macro is deprecated; use @code{define_special_predicate}
8688 instead. @xref{Defining Predicates}.
8691 @defmac HAS_LONG_COND_BRANCH
8692 Define this boolean macro to indicate whether or not your architecture
8693 has conditional branches that can span all of memory. It is used in
8694 conjunction with an optimization that partitions hot and cold basic
8695 blocks into separate sections of the executable. If this macro is
8696 set to false, gcc will convert any conditional branches that attempt
8697 to cross between sections into unconditional branches or indirect jumps.
8700 @defmac HAS_LONG_UNCOND_BRANCH
8701 Define this boolean macro to indicate whether or not your architecture
8702 has unconditional branches that can span all of memory. It is used in
8703 conjunction with an optimization that partitions hot and cold basic
8704 blocks into separate sections of the executable. If this macro is
8705 set to false, gcc will convert any unconditional branches that attempt
8706 to cross between sections into indirect jumps.
8709 @defmac CASE_VECTOR_MODE
8710 An alias for a machine mode name. This is the machine mode that
8711 elements of a jump-table should have.
8714 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
8715 Optional: return the preferred mode for an @code{addr_diff_vec}
8716 when the minimum and maximum offset are known. If you define this,
8717 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
8718 To make this work, you also have to define @code{INSN_ALIGN} and
8719 make the alignment for @code{addr_diff_vec} explicit.
8720 The @var{body} argument is provided so that the offset_unsigned and scale
8721 flags can be updated.
8724 @defmac CASE_VECTOR_PC_RELATIVE
8725 Define this macro to be a C expression to indicate when jump-tables
8726 should contain relative addresses. You need not define this macro if
8727 jump-tables never contain relative addresses, or jump-tables should
8728 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
8732 @defmac CASE_VALUES_THRESHOLD
8733 Define this to be the smallest number of different values for which it
8734 is best to use a jump-table instead of a tree of conditional branches.
8735 The default is four for machines with a @code{casesi} instruction and
8736 five otherwise. This is best for most machines.
8739 @defmac CASE_USE_BIT_TESTS
8740 Define this macro to be a C expression to indicate whether C switch
8741 statements may be implemented by a sequence of bit tests. This is
8742 advantageous on processors that can efficiently implement left shift
8743 of 1 by the number of bits held in a register, but inappropriate on
8744 targets that would require a loop. By default, this macro returns
8745 @code{true} if the target defines an @code{ashlsi3} pattern, and
8746 @code{false} otherwise.
8749 @defmac WORD_REGISTER_OPERATIONS
8750 Define this macro if operations between registers with integral mode
8751 smaller than a word are always performed on the entire register.
8752 Most RISC machines have this property and most CISC machines do not.
8755 @defmac LOAD_EXTEND_OP (@var{mem_mode})
8756 Define this macro to be a C expression indicating when insns that read
8757 memory in @var{mem_mode}, an integral mode narrower than a word, set the
8758 bits outside of @var{mem_mode} to be either the sign-extension or the
8759 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
8760 of @var{mem_mode} for which the
8761 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
8762 @code{UNKNOWN} for other modes.
8764 This macro is not called with @var{mem_mode} non-integral or with a width
8765 greater than or equal to @code{BITS_PER_WORD}, so you may return any
8766 value in this case. Do not define this macro if it would always return
8767 @code{UNKNOWN}. On machines where this macro is defined, you will normally
8768 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
8770 You may return a non-@code{UNKNOWN} value even if for some hard registers
8771 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
8772 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
8773 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
8774 integral mode larger than this but not larger than @code{word_mode}.
8776 You must return @code{UNKNOWN} if for some hard registers that allow this
8777 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
8778 @code{word_mode}, but that they can change to another integral mode that
8779 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
8782 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
8783 Define this macro if loading short immediate values into registers sign
8787 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
8788 Define this macro if the same instructions that convert a floating
8789 point number to a signed fixed point number also convert validly to an
8794 The maximum number of bytes that a single instruction can move quickly
8795 between memory and registers or between two memory locations.
8798 @defmac MAX_MOVE_MAX
8799 The maximum number of bytes that a single instruction can move quickly
8800 between memory and registers or between two memory locations. If this
8801 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
8802 constant value that is the largest value that @code{MOVE_MAX} can have
8806 @defmac SHIFT_COUNT_TRUNCATED
8807 A C expression that is nonzero if on this machine the number of bits
8808 actually used for the count of a shift operation is equal to the number
8809 of bits needed to represent the size of the object being shifted. When
8810 this macro is nonzero, the compiler will assume that it is safe to omit
8811 a sign-extend, zero-extend, and certain bitwise `and' instructions that
8812 truncates the count of a shift operation. On machines that have
8813 instructions that act on bit-fields at variable positions, which may
8814 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
8815 also enables deletion of truncations of the values that serve as
8816 arguments to bit-field instructions.
8818 If both types of instructions truncate the count (for shifts) and
8819 position (for bit-field operations), or if no variable-position bit-field
8820 instructions exist, you should define this macro.
8822 However, on some machines, such as the 80386 and the 680x0, truncation
8823 only applies to shift operations and not the (real or pretended)
8824 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
8825 such machines. Instead, add patterns to the @file{md} file that include
8826 the implied truncation of the shift instructions.
8828 You need not define this macro if it would always have the value of zero.
8831 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
8832 @deftypefn {Target Hook} int TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode})
8833 This function describes how the standard shift patterns for @var{mode}
8834 deal with shifts by negative amounts or by more than the width of the mode.
8835 @xref{shift patterns}.
8837 On many machines, the shift patterns will apply a mask @var{m} to the
8838 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
8839 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
8840 this is true for mode @var{mode}, the function should return @var{m},
8841 otherwise it should return 0. A return value of 0 indicates that no
8842 particular behavior is guaranteed.
8844 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
8845 @emph{not} apply to general shift rtxes; it applies only to instructions
8846 that are generated by the named shift patterns.
8848 The default implementation of this function returns
8849 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
8850 and 0 otherwise. This definition is always safe, but if
8851 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
8852 nevertheless truncate the shift count, you may get better code
8856 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
8857 A C expression which is nonzero if on this machine it is safe to
8858 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
8859 bits (where @var{outprec} is smaller than @var{inprec}) by merely
8860 operating on it as if it had only @var{outprec} bits.
8862 On many machines, this expression can be 1.
8864 @c rearranged this, removed the phrase "it is reported that". this was
8865 @c to fix an overfull hbox. --mew 10feb93
8866 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
8867 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
8868 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
8869 such cases may improve things.
8872 @defmac STORE_FLAG_VALUE
8873 A C expression describing the value returned by a comparison operator
8874 with an integral mode and stored by a store-flag instruction
8875 (@samp{s@var{cond}}) when the condition is true. This description must
8876 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
8877 comparison operators whose results have a @code{MODE_INT} mode.
8879 A value of 1 or @minus{}1 means that the instruction implementing the
8880 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
8881 and 0 when the comparison is false. Otherwise, the value indicates
8882 which bits of the result are guaranteed to be 1 when the comparison is
8883 true. This value is interpreted in the mode of the comparison
8884 operation, which is given by the mode of the first operand in the
8885 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
8886 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
8889 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
8890 generate code that depends only on the specified bits. It can also
8891 replace comparison operators with equivalent operations if they cause
8892 the required bits to be set, even if the remaining bits are undefined.
8893 For example, on a machine whose comparison operators return an
8894 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
8895 @samp{0x80000000}, saying that just the sign bit is relevant, the
8899 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
8906 (ashift:SI @var{x} (const_int @var{n}))
8910 where @var{n} is the appropriate shift count to move the bit being
8911 tested into the sign bit.
8913 There is no way to describe a machine that always sets the low-order bit
8914 for a true value, but does not guarantee the value of any other bits,
8915 but we do not know of any machine that has such an instruction. If you
8916 are trying to port GCC to such a machine, include an instruction to
8917 perform a logical-and of the result with 1 in the pattern for the
8918 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
8920 Often, a machine will have multiple instructions that obtain a value
8921 from a comparison (or the condition codes). Here are rules to guide the
8922 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
8927 Use the shortest sequence that yields a valid definition for
8928 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
8929 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
8930 comparison operators to do so because there may be opportunities to
8931 combine the normalization with other operations.
8934 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
8935 slightly preferred on machines with expensive jumps and 1 preferred on
8939 As a second choice, choose a value of @samp{0x80000001} if instructions
8940 exist that set both the sign and low-order bits but do not define the
8944 Otherwise, use a value of @samp{0x80000000}.
8947 Many machines can produce both the value chosen for
8948 @code{STORE_FLAG_VALUE} and its negation in the same number of
8949 instructions. On those machines, you should also define a pattern for
8950 those cases, e.g., one matching
8953 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
8956 Some machines can also perform @code{and} or @code{plus} operations on
8957 condition code values with less instructions than the corresponding
8958 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
8959 machines, define the appropriate patterns. Use the names @code{incscc}
8960 and @code{decscc}, respectively, for the patterns which perform
8961 @code{plus} or @code{minus} operations on condition code values. See
8962 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
8963 find such instruction sequences on other machines.
8965 If this macro is not defined, the default value, 1, is used. You need
8966 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
8967 instructions, or if the value generated by these instructions is 1.
8970 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
8971 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
8972 returned when comparison operators with floating-point results are true.
8973 Define this macro on machines that have comparison operations that return
8974 floating-point values. If there are no such operations, do not define
8978 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
8979 A C expression that gives a rtx representing the non-zero true element
8980 for vector comparisons. The returned rtx should be valid for the inner
8981 mode of @var{mode} which is guaranteed to be a vector mode. Define
8982 this macro on machines that have vector comparison operations that
8983 return a vector result. If there are no such operations, do not define
8984 this macro. Typically, this macro is defined as @code{const1_rtx} or
8985 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
8986 the compiler optimizing such vector comparison operations for the
8990 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
8991 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
8992 A C expression that evaluates to true if the architecture defines a value
8993 for @code{clz} or @code{ctz} with a zero operand. If so, @var{value}
8994 should be set to this value. If this macro is not defined, the value of
8995 @code{clz} or @code{ctz} is assumed to be undefined.
8997 This macro must be defined if the target's expansion for @code{ffs}
8998 relies on a particular value to get correct results. Otherwise it
8999 is not necessary, though it may be used to optimize some corner cases.
9001 Note that regardless of this macro the ``definedness'' of @code{clz}
9002 and @code{ctz} at zero do @emph{not} extend to the builtin functions
9003 visible to the user. Thus one may be free to adjust the value at will
9004 to match the target expansion of these operations without fear of
9009 An alias for the machine mode for pointers. On most machines, define
9010 this to be the integer mode corresponding to the width of a hardware
9011 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
9012 On some machines you must define this to be one of the partial integer
9013 modes, such as @code{PSImode}.
9015 The width of @code{Pmode} must be at least as large as the value of
9016 @code{POINTER_SIZE}. If it is not equal, you must define the macro
9017 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
9021 @defmac FUNCTION_MODE
9022 An alias for the machine mode used for memory references to functions
9023 being called, in @code{call} RTL expressions. On most machines this
9024 should be @code{QImode}.
9027 @defmac STDC_0_IN_SYSTEM_HEADERS
9028 In normal operation, the preprocessor expands @code{__STDC__} to the
9029 constant 1, to signify that GCC conforms to ISO Standard C@. On some
9030 hosts, like Solaris, the system compiler uses a different convention,
9031 where @code{__STDC__} is normally 0, but is 1 if the user specifies
9032 strict conformance to the C Standard.
9034 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
9035 convention when processing system header files, but when processing user
9036 files @code{__STDC__} will always expand to 1.
9039 @defmac NO_IMPLICIT_EXTERN_C
9040 Define this macro if the system header files support C++ as well as C@.
9041 This macro inhibits the usual method of using system header files in
9042 C++, which is to pretend that the file's contents are enclosed in
9043 @samp{extern "C" @{@dots{}@}}.
9048 @defmac REGISTER_TARGET_PRAGMAS ()
9049 Define this macro if you want to implement any target-specific pragmas.
9050 If defined, it is a C expression which makes a series of calls to
9051 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
9052 for each pragma. The macro may also do any
9053 setup required for the pragmas.
9055 The primary reason to define this macro is to provide compatibility with
9056 other compilers for the same target. In general, we discourage
9057 definition of target-specific pragmas for GCC@.
9059 If the pragma can be implemented by attributes then you should consider
9060 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
9062 Preprocessor macros that appear on pragma lines are not expanded. All
9063 @samp{#pragma} directives that do not match any registered pragma are
9064 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
9067 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9068 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9070 Each call to @code{c_register_pragma} or
9071 @code{c_register_pragma_with_expansion} establishes one pragma. The
9072 @var{callback} routine will be called when the preprocessor encounters a
9076 #pragma [@var{space}] @var{name} @dots{}
9079 @var{space} is the case-sensitive namespace of the pragma, or
9080 @code{NULL} to put the pragma in the global namespace. The callback
9081 routine receives @var{pfile} as its first argument, which can be passed
9082 on to cpplib's functions if necessary. You can lex tokens after the
9083 @var{name} by calling @code{c_lex}. Tokens that are not read by the
9084 callback will be silently ignored. The end of the line is indicated by
9085 a token of type @code{CPP_EOF}. Macro expansion occurs on the
9086 arguments of pragmas registered with
9087 @code{c_register_pragma_with_expansion} but not on the arguments of
9088 pragmas registered with @code{c_register_pragma}.
9090 For an example use of this routine, see @file{c4x.h} and the callback
9091 routines defined in @file{c4x-c.c}.
9093 Note that the use of @code{c_lex} is specific to the C and C++
9094 compilers. It will not work in the Java or Fortran compilers, or any
9095 other language compilers for that matter. Thus if @code{c_lex} is going
9096 to be called from target-specific code, it must only be done so when
9097 building the C and C++ compilers. This can be done by defining the
9098 variables @code{c_target_objs} and @code{cxx_target_objs} in the
9099 target entry in the @file{config.gcc} file. These variables should name
9100 the target-specific, language-specific object file which contains the
9101 code that uses @code{c_lex}. Note it will also be necessary to add a
9102 rule to the makefile fragment pointed to by @code{tmake_file} that shows
9103 how to build this object file.
9108 @defmac HANDLE_SYSV_PRAGMA
9109 Define this macro (to a value of 1) if you want the System V style
9110 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
9111 [=<value>]} to be supported by gcc.
9113 The pack pragma specifies the maximum alignment (in bytes) of fields
9114 within a structure, in much the same way as the @samp{__aligned__} and
9115 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
9116 the behavior to the default.
9118 A subtlety for Microsoft Visual C/C++ style bit-field packing
9119 (e.g.@: -mms-bitfields) for targets that support it:
9120 When a bit-field is inserted into a packed record, the whole size
9121 of the underlying type is used by one or more same-size adjacent
9122 bit-fields (that is, if its long:3, 32 bits is used in the record,
9123 and any additional adjacent long bit-fields are packed into the same
9124 chunk of 32 bits. However, if the size changes, a new field of that
9127 If both MS bit-fields and @samp{__attribute__((packed))} are used,
9128 the latter will take precedence. If @samp{__attribute__((packed))} is
9129 used on a single field when MS bit-fields are in use, it will take
9130 precedence for that field, but the alignment of the rest of the structure
9131 may affect its placement.
9133 The weak pragma only works if @code{SUPPORTS_WEAK} and
9134 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
9135 of specifically named weak labels, optionally with a value.
9140 @defmac HANDLE_PRAGMA_PACK_PUSH_POP
9141 Define this macro (to a value of 1) if you want to support the Win32
9142 style pragmas @samp{#pragma pack(push[,@var{n}])} and @samp{#pragma
9143 pack(pop)}. The @samp{pack(push,[@var{n}])} pragma specifies the maximum
9144 alignment (in bytes) of fields within a structure, in much the same way as
9145 the @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
9146 pack value of zero resets the behavior to the default. Successive
9147 invocations of this pragma cause the previous values to be stacked, so
9148 that invocations of @samp{#pragma pack(pop)} will return to the previous
9152 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
9153 Define this macro, as well as
9154 @code{HANDLE_SYSV_PRAGMA}, if macros should be expanded in the
9155 arguments of @samp{#pragma pack}.
9158 @defmac TARGET_DEFAULT_PACK_STRUCT
9159 If your target requires a structure packing default other than 0 (meaning
9160 the machine default), define this macro to the necessary value (in bytes).
9161 This must be a value that would also valid to be used with
9162 @samp{#pragma pack()} (that is, a small power of two).
9165 @defmac DOLLARS_IN_IDENTIFIERS
9166 Define this macro to control use of the character @samp{$} in
9167 identifier names for the C family of languages. 0 means @samp{$} is
9168 not allowed by default; 1 means it is allowed. 1 is the default;
9169 there is no need to define this macro in that case.
9172 @defmac NO_DOLLAR_IN_LABEL
9173 Define this macro if the assembler does not accept the character
9174 @samp{$} in label names. By default constructors and destructors in
9175 G++ have @samp{$} in the identifiers. If this macro is defined,
9176 @samp{.} is used instead.
9179 @defmac NO_DOT_IN_LABEL
9180 Define this macro if the assembler does not accept the character
9181 @samp{.} in label names. By default constructors and destructors in G++
9182 have names that use @samp{.}. If this macro is defined, these names
9183 are rewritten to avoid @samp{.}.
9186 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
9187 Define this macro as a C expression that is nonzero if it is safe for the
9188 delay slot scheduler to place instructions in the delay slot of @var{insn},
9189 even if they appear to use a resource set or clobbered in @var{insn}.
9190 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
9191 every @code{call_insn} has this behavior. On machines where some @code{insn}
9192 or @code{jump_insn} is really a function call and hence has this behavior,
9193 you should define this macro.
9195 You need not define this macro if it would always return zero.
9198 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
9199 Define this macro as a C expression that is nonzero if it is safe for the
9200 delay slot scheduler to place instructions in the delay slot of @var{insn},
9201 even if they appear to set or clobber a resource referenced in @var{insn}.
9202 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
9203 some @code{insn} or @code{jump_insn} is really a function call and its operands
9204 are registers whose use is actually in the subroutine it calls, you should
9205 define this macro. Doing so allows the delay slot scheduler to move
9206 instructions which copy arguments into the argument registers into the delay
9209 You need not define this macro if it would always return zero.
9212 @defmac MULTIPLE_SYMBOL_SPACES
9213 Define this macro as a C expression that is nonzero if, in some cases,
9214 global symbols from one translation unit may not be bound to undefined
9215 symbols in another translation unit without user intervention. For
9216 instance, under Microsoft Windows symbols must be explicitly imported
9217 from shared libraries (DLLs).
9219 You need not define this macro if it would always evaluate to zero.
9222 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{outputs}, tree @var{inputs}, tree @var{clobbers})
9223 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
9224 any hard regs the port wishes to automatically clobber for an asm.
9225 It should return the result of the last @code{tree_cons} used to add a
9226 clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the
9227 corresponding parameters to the asm and may be inspected to avoid
9228 clobbering a register that is an input or output of the asm. You can use
9229 @code{decl_overlaps_hard_reg_set_p}, declared in @file{tree.h}, to test
9230 for overlap with regards to asm-declared registers.
9233 @defmac MATH_LIBRARY
9234 Define this macro as a C string constant for the linker argument to link
9235 in the system math library, or @samp{""} if the target does not have a
9236 separate math library.
9238 You need only define this macro if the default of @samp{"-lm"} is wrong.
9241 @defmac LIBRARY_PATH_ENV
9242 Define this macro as a C string constant for the environment variable that
9243 specifies where the linker should look for libraries.
9245 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
9249 @defmac TARGET_HAS_F_SETLKW
9250 Define this macro if the target supports file locking with fcntl / F_SETLKW@.
9251 Note that this functionality is part of POSIX@.
9252 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
9253 to use file locking when exiting a program, which avoids race conditions
9254 if the program has forked.
9257 @defmac MAX_CONDITIONAL_EXECUTE
9259 A C expression for the maximum number of instructions to execute via
9260 conditional execution instructions instead of a branch. A value of
9261 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
9262 1 if it does use cc0.
9265 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
9266 Used if the target needs to perform machine-dependent modifications on the
9267 conditionals used for turning basic blocks into conditionally executed code.
9268 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
9269 contains information about the currently processed blocks. @var{true_expr}
9270 and @var{false_expr} are the tests that are used for converting the
9271 then-block and the else-block, respectively. Set either @var{true_expr} or
9272 @var{false_expr} to a null pointer if the tests cannot be converted.
9275 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
9276 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
9277 if-statements into conditions combined by @code{and} and @code{or} operations.
9278 @var{bb} contains the basic block that contains the test that is currently
9279 being processed and about to be turned into a condition.
9282 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
9283 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
9284 be converted to conditional execution format. @var{ce_info} points to
9285 a data structure, @code{struct ce_if_block}, which contains information
9286 about the currently processed blocks.
9289 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
9290 A C expression to perform any final machine dependent modifications in
9291 converting code to conditional execution. The involved basic blocks
9292 can be found in the @code{struct ce_if_block} structure that is pointed
9293 to by @var{ce_info}.
9296 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
9297 A C expression to cancel any machine dependent modifications in
9298 converting code to conditional execution. The involved basic blocks
9299 can be found in the @code{struct ce_if_block} structure that is pointed
9300 to by @var{ce_info}.
9303 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
9304 A C expression to initialize any extra fields in a @code{struct ce_if_block}
9305 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
9308 @defmac IFCVT_EXTRA_FIELDS
9309 If defined, it should expand to a set of field declarations that will be
9310 added to the @code{struct ce_if_block} structure. These should be initialized
9311 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
9314 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG ()
9315 If non-null, this hook performs a target-specific pass over the
9316 instruction stream. The compiler will run it at all optimization levels,
9317 just before the point at which it normally does delayed-branch scheduling.
9319 The exact purpose of the hook varies from target to target. Some use
9320 it to do transformations that are necessary for correctness, such as
9321 laying out in-function constant pools or avoiding hardware hazards.
9322 Others use it as an opportunity to do some machine-dependent optimizations.
9324 You need not implement the hook if it has nothing to do. The default
9328 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
9329 Define this hook if you have any machine-specific built-in functions
9330 that need to be defined. It should be a function that performs the
9333 Machine specific built-in functions can be useful to expand special machine
9334 instructions that would otherwise not normally be generated because
9335 they have no equivalent in the source language (for example, SIMD vector
9336 instructions or prefetch instructions).
9338 To create a built-in function, call the function
9339 @code{lang_hooks.builtin_function}
9340 which is defined by the language front end. You can use any type nodes set
9341 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
9342 only language front ends that use those two functions will call
9343 @samp{TARGET_INIT_BUILTINS}.
9346 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
9348 Expand a call to a machine specific built-in function that was set up by
9349 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
9350 function call; the result should go to @var{target} if that is
9351 convenient, and have mode @var{mode} if that is convenient.
9352 @var{subtarget} may be used as the target for computing one of
9353 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
9354 ignored. This function should return the result of the call to the
9358 @deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (tree @var{fndecl}, tree @var{arglist})
9360 Select a replacement for a machine specific built-in function that
9361 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
9362 @emph{before} regular type checking, and so allows the target to
9363 implement a crude form of function overloading. @var{fndecl} is the
9364 declaration of the built-in function. @var{arglist} is the list of
9365 arguments passed to the built-in function. The result is a
9366 complete expression that implements the operation, usually
9367 another @code{CALL_EXPR}.
9370 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, tree @var{arglist}, bool @var{ignore})
9372 Fold a call to a machine specific built-in function that was set up by
9373 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
9374 built-in function. @var{arglist} is the list of arguments passed to
9375 the built-in function. The result is another tree containing a
9376 simplified expression for the call's result. If @var{ignore} is true
9377 the value will be ignored.
9380 @deftypefn {Target Hook} const char * TARGET_INVALID_WITHIN_DOLOOP (rtx @var{insn})
9382 Take an instruction in @var{insn} and return NULL if it is valid within a
9383 low-overhead loop, otherwise return a string why doloop could not be applied.
9385 Many targets use special registers for low-overhead looping. For any
9386 instruction that clobbers these this function should return a string indicating
9387 the reason why the doloop could not be applied.
9388 By default, the RTL loop optimizer does not use a present doloop pattern for
9389 loops containing function calls or branch on table instructions.
9392 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
9394 Take a branch insn in @var{branch1} and another in @var{branch2}.
9395 Return true if redirecting @var{branch1} to the destination of
9396 @var{branch2} is possible.
9398 On some targets, branches may have a limited range. Optimizing the
9399 filling of delay slots can result in branches being redirected, and this
9400 may in turn cause a branch offset to overflow.
9403 @defmac ALLOCATE_INITIAL_VALUE (@var{hard_reg})
9405 When the initial value of a hard register has been copied in a pseudo
9406 register, it is often not necessary to actually allocate another register
9407 to this pseudo register, because the original hard register or a stack slot
9408 it has been saved into can be used. @code{ALLOCATE_INITIAL_VALUE}, if
9409 defined, is called at the start of register allocation once for each
9410 hard register that had its initial value copied by using
9411 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
9412 Possible values are @code{NULL_RTX}, if you don't want
9413 to do any special allocation, a @code{REG} rtx---that would typically be
9414 the hard register itself, if it is known not to be clobbered---or a
9416 If you are returning a @code{MEM}, this is only a hint for the allocator;
9417 it might decide to use another register anyways.
9418 You may use @code{current_function_leaf_function} in the definition of the
9419 macro, functions that use @code{REG_N_SETS}, to determine if the hard
9420 register in question will not be clobbered.
9423 @defmac TARGET_OBJECT_SUFFIX
9424 Define this macro to be a C string representing the suffix for object
9425 files on your target machine. If you do not define this macro, GCC will
9426 use @samp{.o} as the suffix for object files.
9429 @defmac TARGET_EXECUTABLE_SUFFIX
9430 Define this macro to be a C string representing the suffix to be
9431 automatically added to executable files on your target machine. If you
9432 do not define this macro, GCC will use the null string as the suffix for
9436 @defmac COLLECT_EXPORT_LIST
9437 If defined, @code{collect2} will scan the individual object files
9438 specified on its command line and create an export list for the linker.
9439 Define this macro for systems like AIX, where the linker discards
9440 object files that are not referenced from @code{main} and uses export
9444 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
9445 Define this macro to a C expression representing a variant of the
9446 method call @var{mdecl}, if Java Native Interface (JNI) methods
9447 must be invoked differently from other methods on your target.
9448 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
9449 the @code{stdcall} calling convention and this macro is then
9450 defined as this expression:
9453 build_type_attribute_variant (@var{mdecl},
9455 (get_identifier ("stdcall"),
9460 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
9461 This target hook returns @code{true} past the point in which new jump
9462 instructions could be created. On machines that require a register for
9463 every jump such as the SHmedia ISA of SH5, this point would typically be
9464 reload, so this target hook should be defined to a function such as:
9468 cannot_modify_jumps_past_reload_p ()
9470 return (reload_completed || reload_in_progress);
9475 @deftypefn {Target Hook} int TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
9476 This target hook returns a register class for which branch target register
9477 optimizations should be applied. All registers in this class should be
9478 usable interchangeably. After reload, registers in this class will be
9479 re-allocated and loads will be hoisted out of loops and be subjected
9480 to inter-block scheduling.
9483 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
9484 Branch target register optimization will by default exclude callee-saved
9486 that are not already live during the current function; if this target hook
9487 returns true, they will be included. The target code must than make sure
9488 that all target registers in the class returned by
9489 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
9490 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
9491 epilogues have already been generated. Note, even if you only return
9492 true when @var{after_prologue_epilogue_gen} is false, you still are likely
9493 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
9494 to reserve space for caller-saved target registers.
9497 @defmac POWI_MAX_MULTS
9498 If defined, this macro is interpreted as a signed integer C expression
9499 that specifies the maximum number of floating point multiplications
9500 that should be emitted when expanding exponentiation by an integer
9501 constant inline. When this value is defined, exponentiation requiring
9502 more than this number of multiplications is implemented by calling the
9503 system library's @code{pow}, @code{powf} or @code{powl} routines.
9504 The default value places no upper bound on the multiplication count.
9507 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
9508 This target hook should register any extra include files for the
9509 target. The parameter @var{stdinc} indicates if normal include files
9510 are present. The parameter @var{sysroot} is the system root directory.
9511 The parameter @var{iprefix} is the prefix for the gcc directory.
9514 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
9515 This target hook should register any extra include files for the
9516 target before any standard headers. The parameter @var{stdinc}
9517 indicates if normal include files are present. The parameter
9518 @var{sysroot} is the system root directory. The parameter
9519 @var{iprefix} is the prefix for the gcc directory.
9522 @deftypefn Macro void TARGET_OPTF (char *@var{path})
9523 This target hook should register special include paths for the target.
9524 The parameter @var{path} is the include to register. On Darwin
9525 systems, this is used for Framework includes, which have semantics
9526 that are different from @option{-I}.
9529 @deftypefn {Target Hook} bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
9530 This target hook returns @code{true} if it is safe to use a local alias
9531 for a virtual function @var{fndecl} when constructing thunks,
9532 @code{false} otherwise. By default, the hook returns @code{true} for all
9533 functions, if a target supports aliases (i.e.@: defines
9534 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
9537 @defmac TARGET_FORMAT_TYPES
9538 If defined, this macro is the name of a global variable containing
9539 target-specific format checking information for the @option{-Wformat}
9540 option. The default is to have no target-specific format checks.
9543 @defmac TARGET_N_FORMAT_TYPES
9544 If defined, this macro is the number of entries in
9545 @code{TARGET_FORMAT_TYPES}.
9548 @deftypefn {Target Hook} bool TARGET_RELAXED_ORDERING
9549 If set to @code{true}, means that the target's memory model does not
9550 guarantee that loads which do not depend on one another will access
9551 main memory in the order of the instruction stream; if ordering is
9552 important, an explicit memory barrier must be used. This is true of
9553 many recent processors which implement a policy of ``relaxed,''
9554 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
9555 and ia64. The default is @code{false}.
9558 @deftypefn {Target Hook} const char *TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (tree @var{typelist}, tree @var{funcdecl}, tree @var{val})
9559 If defined, this macro returns the diagnostic message when it is
9560 illegal to pass argument @var{val} to function @var{funcdecl}
9561 with prototype @var{typelist}.
9564 @defmac TARGET_USE_JCR_SECTION
9565 This macro determines whether to use the JCR section to register Java
9566 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
9567 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.