1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,
2 @c 2002, 2003, 2004, 2005, 2006, 2007, 2008 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
7 @chapter Target Description Macros and Functions
8 @cindex machine description macros
9 @cindex target description macros
10 @cindex macros, target description
11 @cindex @file{tm.h} macros
13 In addition to the file @file{@var{machine}.md}, a machine description
14 includes a C header file conventionally given the name
15 @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}.
16 The header file defines numerous macros that convey the information
17 about the target machine that does not fit into the scheme of the
18 @file{.md} file. The file @file{tm.h} should be a link to
19 @file{@var{machine}.h}. The header file @file{config.h} includes
20 @file{tm.h} and most compiler source files include @file{config.h}. The
21 source file defines a variable @code{targetm}, which is a structure
22 containing pointers to functions and data relating to the target
23 machine. @file{@var{machine}.c} should also contain their definitions,
24 if they are not defined elsewhere in GCC, and other functions called
25 through the macros defined in the @file{.h} file.
28 * Target Structure:: The @code{targetm} variable.
29 * Driver:: Controlling how the driver runs the compilation passes.
30 * Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}.
31 * Per-Function Data:: Defining data structures for per-function information.
32 * Storage Layout:: Defining sizes and alignments of data.
33 * Type Layout:: Defining sizes and properties of basic user data types.
34 * Registers:: Naming and describing the hardware registers.
35 * Register Classes:: Defining the classes of hardware registers.
36 * Old Constraints:: The old way to define machine-specific constraints.
37 * Stack and Calling:: Defining which way the stack grows and by how much.
38 * Varargs:: Defining the varargs macros.
39 * Trampolines:: Code set up at run time to enter a nested function.
40 * Library Calls:: Controlling how library routines are implicitly called.
41 * Addressing Modes:: Defining addressing modes valid for memory operands.
42 * Anchored Addresses:: Defining how @option{-fsection-anchors} should work.
43 * Condition Code:: Defining how insns update the condition code.
44 * Costs:: Defining relative costs of different operations.
45 * Scheduling:: Adjusting the behavior of the instruction scheduler.
46 * Sections:: Dividing storage into text, data, and other sections.
47 * PIC:: Macros for position independent code.
48 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
49 * Debugging Info:: Defining the format of debugging output.
50 * Floating Point:: Handling floating point for cross-compilers.
51 * Mode Switching:: Insertion of mode-switching instructions.
52 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
53 * Emulated TLS:: Emulated TLS support.
54 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
55 * PCH Target:: Validity checking for precompiled headers.
56 * C++ ABI:: Controlling C++ ABI changes.
57 * Misc:: Everything else.
60 @node Target Structure
61 @section The Global @code{targetm} Variable
63 @cindex target functions
65 @deftypevar {struct gcc_target} targetm
66 The target @file{.c} file must define the global @code{targetm} variable
67 which contains pointers to functions and data relating to the target
68 machine. The variable is declared in @file{target.h};
69 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
70 used to initialize the variable, and macros for the default initializers
71 for elements of the structure. The @file{.c} file should override those
72 macros for which the default definition is inappropriate. For example:
75 #include "target-def.h"
77 /* @r{Initialize the GCC target structure.} */
79 #undef TARGET_COMP_TYPE_ATTRIBUTES
80 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
82 struct gcc_target targetm = TARGET_INITIALIZER;
86 Where a macro should be defined in the @file{.c} file in this manner to
87 form part of the @code{targetm} structure, it is documented below as a
88 ``Target Hook'' with a prototype. Many macros will change in future
89 from being defined in the @file{.h} file to being part of the
90 @code{targetm} structure.
93 @section Controlling the Compilation Driver, @file{gcc}
95 @cindex controlling the compilation driver
97 @c prevent bad page break with this line
98 You can control the compilation driver.
100 @defmac SWITCH_TAKES_ARG (@var{char})
101 A C expression which determines whether the option @option{-@var{char}}
102 takes arguments. The value should be the number of arguments that
103 option takes--zero, for many options.
105 By default, this macro is defined as
106 @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options
107 properly. You need not define @code{SWITCH_TAKES_ARG} unless you
108 wish to add additional options which take arguments. Any redefinition
109 should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for
113 @defmac WORD_SWITCH_TAKES_ARG (@var{name})
114 A C expression which determines whether the option @option{-@var{name}}
115 takes arguments. The value should be the number of arguments that
116 option takes--zero, for many options. This macro rather than
117 @code{SWITCH_TAKES_ARG} is used for multi-character option names.
119 By default, this macro is defined as
120 @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
121 properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
122 wish to add additional options which take arguments. Any redefinition
123 should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
127 @defmac SWITCH_CURTAILS_COMPILATION (@var{char})
128 A C expression which determines whether the option @option{-@var{char}}
129 stops compilation before the generation of an executable. The value is
130 boolean, nonzero if the option does stop an executable from being
131 generated, zero otherwise.
133 By default, this macro is defined as
134 @code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard
135 options properly. You need not define
136 @code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional
137 options which affect the generation of an executable. Any redefinition
138 should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check
139 for additional options.
142 @defmac SWITCHES_NEED_SPACES
143 A string-valued C expression which enumerates the options for which
144 the linker needs a space between the option and its argument.
146 If this macro is not defined, the default value is @code{""}.
149 @defmac TARGET_OPTION_TRANSLATE_TABLE
150 If defined, a list of pairs of strings, the first of which is a
151 potential command line target to the @file{gcc} driver program, and the
152 second of which is a space-separated (tabs and other whitespace are not
153 supported) list of options with which to replace the first option. The
154 target defining this list is responsible for assuring that the results
155 are valid. Replacement options may not be the @code{--opt} style, they
156 must be the @code{-opt} style. It is the intention of this macro to
157 provide a mechanism for substitution that affects the multilibs chosen,
158 such as one option that enables many options, some of which select
159 multilibs. Example nonsensical definition, where @option{-malt-abi},
160 @option{-EB}, and @option{-mspoo} cause different multilibs to be chosen:
163 #define TARGET_OPTION_TRANSLATE_TABLE \
164 @{ "-fast", "-march=fast-foo -malt-abi -I/usr/fast-foo" @}, \
165 @{ "-compat", "-EB -malign=4 -mspoo" @}
169 @defmac DRIVER_SELF_SPECS
170 A list of specs for the driver itself. It should be a suitable
171 initializer for an array of strings, with no surrounding braces.
173 The driver applies these specs to its own command line between loading
174 default @file{specs} files (but not command-line specified ones) and
175 choosing the multilib directory or running any subcommands. It
176 applies them in the order given, so each spec can depend on the
177 options added by earlier ones. It is also possible to remove options
178 using @samp{%<@var{option}} in the usual way.
180 This macro can be useful when a port has several interdependent target
181 options. It provides a way of standardizing the command line so
182 that the other specs are easier to write.
184 Do not define this macro if it does not need to do anything.
187 @defmac OPTION_DEFAULT_SPECS
188 A list of specs used to support configure-time default options (i.e.@:
189 @option{--with} options) in the driver. It should be a suitable initializer
190 for an array of structures, each containing two strings, without the
191 outermost pair of surrounding braces.
193 The first item in the pair is the name of the default. This must match
194 the code in @file{config.gcc} for the target. The second item is a spec
195 to apply if a default with this name was specified. The string
196 @samp{%(VALUE)} in the spec will be replaced by the value of the default
197 everywhere it occurs.
199 The driver will apply these specs to its own command line between loading
200 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
201 the same mechanism as @code{DRIVER_SELF_SPECS}.
203 Do not define this macro if it does not need to do anything.
207 A C string constant that tells the GCC driver program options to
208 pass to CPP@. It can also specify how to translate options you
209 give to GCC into options for GCC to pass to the CPP@.
211 Do not define this macro if it does not need to do anything.
214 @defmac CPLUSPLUS_CPP_SPEC
215 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
216 than C@. If you do not define this macro, then the value of
217 @code{CPP_SPEC} (if any) will be used instead.
221 A C string constant that tells the GCC driver program options to
222 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
224 It can also specify how to translate options you give to GCC into options
225 for GCC to pass to front ends.
227 Do not define this macro if it does not need to do anything.
231 A C string constant that tells the GCC driver program options to
232 pass to @code{cc1plus}. It can also specify how to translate options you
233 give to GCC into options for GCC to pass to the @code{cc1plus}.
235 Do not define this macro if it does not need to do anything.
236 Note that everything defined in CC1_SPEC is already passed to
237 @code{cc1plus} so there is no need to duplicate the contents of
238 CC1_SPEC in CC1PLUS_SPEC@.
242 A C string constant that tells the GCC driver program options to
243 pass to the assembler. It can also specify how to translate options
244 you give to GCC into options for GCC to pass to the assembler.
245 See the file @file{sun3.h} for an example of this.
247 Do not define this macro if it does not need to do anything.
250 @defmac ASM_FINAL_SPEC
251 A C string constant that tells the GCC driver program how to
252 run any programs which cleanup after the normal assembler.
253 Normally, this is not needed. See the file @file{mips.h} for
256 Do not define this macro if it does not need to do anything.
259 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
260 Define this macro, with no value, if the driver should give the assembler
261 an argument consisting of a single dash, @option{-}, to instruct it to
262 read from its standard input (which will be a pipe connected to the
263 output of the compiler proper). This argument is given after any
264 @option{-o} option specifying the name of the output file.
266 If you do not define this macro, the assembler is assumed to read its
267 standard input if given no non-option arguments. If your assembler
268 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
269 see @file{mips.h} for instance.
273 A C string constant that tells the GCC driver program options to
274 pass to the linker. It can also specify how to translate options you
275 give to GCC into options for GCC to pass to the linker.
277 Do not define this macro if it does not need to do anything.
281 Another C string constant used much like @code{LINK_SPEC}. The difference
282 between the two is that @code{LIB_SPEC} is used at the end of the
283 command given to the linker.
285 If this macro is not defined, a default is provided that
286 loads the standard C library from the usual place. See @file{gcc.c}.
290 Another C string constant that tells the GCC driver program
291 how and when to place a reference to @file{libgcc.a} into the
292 linker command line. This constant is placed both before and after
293 the value of @code{LIB_SPEC}.
295 If this macro is not defined, the GCC driver provides a default that
296 passes the string @option{-lgcc} to the linker.
299 @defmac REAL_LIBGCC_SPEC
300 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
301 @code{LIBGCC_SPEC} is not directly used by the driver program but is
302 instead modified to refer to different versions of @file{libgcc.a}
303 depending on the values of the command line flags @option{-static},
304 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
305 targets where these modifications are inappropriate, define
306 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
307 driver how to place a reference to @file{libgcc} on the link command
308 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
311 @defmac USE_LD_AS_NEEDED
312 A macro that controls the modifications to @code{LIBGCC_SPEC}
313 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
314 generated that uses --as-needed and the shared libgcc in place of the
315 static exception handler library, when linking without any of
316 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
320 If defined, this C string constant is added to @code{LINK_SPEC}.
321 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
322 the modifications to @code{LIBGCC_SPEC} mentioned in
323 @code{REAL_LIBGCC_SPEC}.
326 @defmac STARTFILE_SPEC
327 Another C string constant used much like @code{LINK_SPEC}. The
328 difference between the two is that @code{STARTFILE_SPEC} is used at
329 the very beginning of the command given to the linker.
331 If this macro is not defined, a default is provided that loads the
332 standard C startup file from the usual place. See @file{gcc.c}.
336 Another C string constant used much like @code{LINK_SPEC}. The
337 difference between the two is that @code{ENDFILE_SPEC} is used at
338 the very end of the command given to the linker.
340 Do not define this macro if it does not need to do anything.
343 @defmac THREAD_MODEL_SPEC
344 GCC @code{-v} will print the thread model GCC was configured to use.
345 However, this doesn't work on platforms that are multilibbed on thread
346 models, such as AIX 4.3. On such platforms, define
347 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
348 blanks that names one of the recognized thread models. @code{%*}, the
349 default value of this macro, will expand to the value of
350 @code{thread_file} set in @file{config.gcc}.
353 @defmac SYSROOT_SUFFIX_SPEC
354 Define this macro to add a suffix to the target sysroot when GCC is
355 configured with a sysroot. This will cause GCC to search for usr/lib,
356 et al, within sysroot+suffix.
359 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
360 Define this macro to add a headers_suffix to the target sysroot when
361 GCC is configured with a sysroot. This will cause GCC to pass the
362 updated sysroot+headers_suffix to CPP, causing it to search for
363 usr/include, et al, within sysroot+headers_suffix.
367 Define this macro to provide additional specifications to put in the
368 @file{specs} file that can be used in various specifications like
371 The definition should be an initializer for an array of structures,
372 containing a string constant, that defines the specification name, and a
373 string constant that provides the specification.
375 Do not define this macro if it does not need to do anything.
377 @code{EXTRA_SPECS} is useful when an architecture contains several
378 related targets, which have various @code{@dots{}_SPECS} which are similar
379 to each other, and the maintainer would like one central place to keep
382 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
383 define either @code{_CALL_SYSV} when the System V calling sequence is
384 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
387 The @file{config/rs6000/rs6000.h} target file defines:
390 #define EXTRA_SPECS \
391 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
393 #define CPP_SYS_DEFAULT ""
396 The @file{config/rs6000/sysv.h} target file defines:
400 "%@{posix: -D_POSIX_SOURCE @} \
401 %@{mcall-sysv: -D_CALL_SYSV @} \
402 %@{!mcall-sysv: %(cpp_sysv_default) @} \
403 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
405 #undef CPP_SYSV_DEFAULT
406 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
409 while the @file{config/rs6000/eabiaix.h} target file defines
410 @code{CPP_SYSV_DEFAULT} as:
413 #undef CPP_SYSV_DEFAULT
414 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
418 @defmac LINK_LIBGCC_SPECIAL_1
419 Define this macro if the driver program should find the library
420 @file{libgcc.a}. If you do not define this macro, the driver program will pass
421 the argument @option{-lgcc} to tell the linker to do the search.
424 @defmac LINK_GCC_C_SEQUENCE_SPEC
425 The sequence in which libgcc and libc are specified to the linker.
426 By default this is @code{%G %L %G}.
429 @defmac LINK_COMMAND_SPEC
430 A C string constant giving the complete command line need to execute the
431 linker. When you do this, you will need to update your port each time a
432 change is made to the link command line within @file{gcc.c}. Therefore,
433 define this macro only if you need to completely redefine the command
434 line for invoking the linker and there is no other way to accomplish
435 the effect you need. Overriding this macro may be avoidable by overriding
436 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
439 @defmac LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
440 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
441 directories from linking commands. Do not give it a nonzero value if
442 removing duplicate search directories changes the linker's semantics.
445 @defmac MULTILIB_DEFAULTS
446 Define this macro as a C expression for the initializer of an array of
447 string to tell the driver program which options are defaults for this
448 target and thus do not need to be handled specially when using
449 @code{MULTILIB_OPTIONS}.
451 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
452 the target makefile fragment or if none of the options listed in
453 @code{MULTILIB_OPTIONS} are set by default.
454 @xref{Target Fragment}.
457 @defmac RELATIVE_PREFIX_NOT_LINKDIR
458 Define this macro to tell @command{gcc} that it should only translate
459 a @option{-B} prefix into a @option{-L} linker option if the prefix
460 indicates an absolute file name.
463 @defmac MD_EXEC_PREFIX
464 If defined, this macro is an additional prefix to try after
465 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
466 when the @option{-b} option is used, or the compiler is built as a cross
467 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
468 to the list of directories used to find the assembler in @file{configure.in}.
471 @defmac STANDARD_STARTFILE_PREFIX
472 Define this macro as a C string constant if you wish to override the
473 standard choice of @code{libdir} as the default prefix to
474 try when searching for startup files such as @file{crt0.o}.
475 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
476 is built as a cross compiler.
479 @defmac STANDARD_STARTFILE_PREFIX_1
480 Define this macro as a C string constant if you wish to override the
481 standard choice of @code{/lib} as a prefix to try after the default prefix
482 when searching for startup files such as @file{crt0.o}.
483 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
484 is built as a cross compiler.
487 @defmac STANDARD_STARTFILE_PREFIX_2
488 Define this macro as a C string constant if you wish to override the
489 standard choice of @code{/lib} as yet another prefix to try after the
490 default prefix when searching for startup files such as @file{crt0.o}.
491 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
492 is built as a cross compiler.
495 @defmac MD_STARTFILE_PREFIX
496 If defined, this macro supplies an additional prefix to try after the
497 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
498 @option{-b} option is used, or when the compiler is built as a cross
502 @defmac MD_STARTFILE_PREFIX_1
503 If defined, this macro supplies yet another prefix to try after the
504 standard prefixes. It is not searched when the @option{-b} option is
505 used, or when the compiler is built as a cross compiler.
508 @defmac INIT_ENVIRONMENT
509 Define this macro as a C string constant if you wish to set environment
510 variables for programs called by the driver, such as the assembler and
511 loader. The driver passes the value of this macro to @code{putenv} to
512 initialize the necessary environment variables.
515 @defmac LOCAL_INCLUDE_DIR
516 Define this macro as a C string constant if you wish to override the
517 standard choice of @file{/usr/local/include} as the default prefix to
518 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
519 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
521 Cross compilers do not search either @file{/usr/local/include} or its
525 @defmac MODIFY_TARGET_NAME
526 Define this macro if you wish to define command-line switches that
527 modify the default target name.
529 For each switch, you can include a string to be appended to the first
530 part of the configuration name or a string to be deleted from the
531 configuration name, if present. The definition should be an initializer
532 for an array of structures. Each array element should have three
533 elements: the switch name (a string constant, including the initial
534 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
535 indicate whether the string should be inserted or deleted, and the string
536 to be inserted or deleted (a string constant).
538 For example, on a machine where @samp{64} at the end of the
539 configuration name denotes a 64-bit target and you want the @option{-32}
540 and @option{-64} switches to select between 32- and 64-bit targets, you would
544 #define MODIFY_TARGET_NAME \
545 @{ @{ "-32", DELETE, "64"@}, \
546 @{"-64", ADD, "64"@}@}
550 @defmac SYSTEM_INCLUDE_DIR
551 Define this macro as a C string constant if you wish to specify a
552 system-specific directory to search for header files before the standard
553 directory. @code{SYSTEM_INCLUDE_DIR} comes before
554 @code{STANDARD_INCLUDE_DIR} in the search order.
556 Cross compilers do not use this macro and do not search the directory
560 @defmac STANDARD_INCLUDE_DIR
561 Define this macro as a C string constant if you wish to override the
562 standard choice of @file{/usr/include} as the default prefix to
563 try when searching for header files.
565 Cross compilers ignore this macro and do not search either
566 @file{/usr/include} or its replacement.
569 @defmac STANDARD_INCLUDE_COMPONENT
570 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
571 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
572 If you do not define this macro, no component is used.
575 @defmac INCLUDE_DEFAULTS
576 Define this macro if you wish to override the entire default search path
577 for include files. For a native compiler, the default search path
578 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
579 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
580 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
581 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
582 and specify private search areas for GCC@. The directory
583 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
585 The definition should be an initializer for an array of structures.
586 Each array element should have four elements: the directory name (a
587 string constant), the component name (also a string constant), a flag
588 for C++-only directories,
589 and a flag showing that the includes in the directory don't need to be
590 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
591 the array with a null element.
593 The component name denotes what GNU package the include file is part of,
594 if any, in all uppercase letters. For example, it might be @samp{GCC}
595 or @samp{BINUTILS}. If the package is part of a vendor-supplied
596 operating system, code the component name as @samp{0}.
598 For example, here is the definition used for VAX/VMS:
601 #define INCLUDE_DEFAULTS \
603 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
604 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
605 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
612 Here is the order of prefixes tried for exec files:
616 Any prefixes specified by the user with @option{-B}.
619 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
620 is not set and the compiler has not been installed in the configure-time
621 @var{prefix}, the location in which the compiler has actually been installed.
624 The directories specified by the environment variable @code{COMPILER_PATH}.
627 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
628 in the configured-time @var{prefix}.
631 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
634 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
637 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
641 Here is the order of prefixes tried for startfiles:
645 Any prefixes specified by the user with @option{-B}.
648 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
649 value based on the installed toolchain location.
652 The directories specified by the environment variable @code{LIBRARY_PATH}
653 (or port-specific name; native only, cross compilers do not use this).
656 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
657 in the configured @var{prefix} or this is a native compiler.
660 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
663 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
667 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
668 native compiler, or we have a target system root.
671 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
672 native compiler, or we have a target system root.
675 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
676 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
677 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
680 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
681 compiler, or we have a target system root. The default for this macro is
685 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
686 compiler, or we have a target system root. The default for this macro is
690 @node Run-time Target
691 @section Run-time Target Specification
692 @cindex run-time target specification
693 @cindex predefined macros
694 @cindex target specifications
696 @c prevent bad page break with this line
697 Here are run-time target specifications.
699 @defmac TARGET_CPU_CPP_BUILTINS ()
700 This function-like macro expands to a block of code that defines
701 built-in preprocessor macros and assertions for the target CPU, using
702 the functions @code{builtin_define}, @code{builtin_define_std} and
703 @code{builtin_assert}. When the front end
704 calls this macro it provides a trailing semicolon, and since it has
705 finished command line option processing your code can use those
708 @code{builtin_assert} takes a string in the form you pass to the
709 command-line option @option{-A}, such as @code{cpu=mips}, and creates
710 the assertion. @code{builtin_define} takes a string in the form
711 accepted by option @option{-D} and unconditionally defines the macro.
713 @code{builtin_define_std} takes a string representing the name of an
714 object-like macro. If it doesn't lie in the user's namespace,
715 @code{builtin_define_std} defines it unconditionally. Otherwise, it
716 defines a version with two leading underscores, and another version
717 with two leading and trailing underscores, and defines the original
718 only if an ISO standard was not requested on the command line. For
719 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
720 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
721 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
722 defines only @code{_ABI64}.
724 You can also test for the C dialect being compiled. The variable
725 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
726 or @code{clk_objective_c}. Note that if we are preprocessing
727 assembler, this variable will be @code{clk_c} but the function-like
728 macro @code{preprocessing_asm_p()} will return true, so you might want
729 to check for that first. If you need to check for strict ANSI, the
730 variable @code{flag_iso} can be used. The function-like macro
731 @code{preprocessing_trad_p()} can be used to check for traditional
735 @defmac TARGET_OS_CPP_BUILTINS ()
736 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
737 and is used for the target operating system instead.
740 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
741 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
742 and is used for the target object format. @file{elfos.h} uses this
743 macro to define @code{__ELF__}, so you probably do not need to define
747 @deftypevar {extern int} target_flags
748 This variable is declared in @file{options.h}, which is included before
749 any target-specific headers.
752 @deftypevar {Target Hook} int TARGET_DEFAULT_TARGET_FLAGS
753 This variable specifies the initial value of @code{target_flags}.
754 Its default setting is 0.
757 @cindex optional hardware or system features
758 @cindex features, optional, in system conventions
760 @deftypefn {Target Hook} bool TARGET_HANDLE_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
761 This hook is called whenever the user specifies one of the
762 target-specific options described by the @file{.opt} definition files
763 (@pxref{Options}). It has the opportunity to do some option-specific
764 processing and should return true if the option is valid. The default
765 definition does nothing but return true.
767 @var{code} specifies the @code{OPT_@var{name}} enumeration value
768 associated with the selected option; @var{name} is just a rendering of
769 the option name in which non-alphanumeric characters are replaced by
770 underscores. @var{arg} specifies the string argument and is null if
771 no argument was given. If the option is flagged as a @code{UInteger}
772 (@pxref{Option properties}), @var{value} is the numeric value of the
773 argument. Otherwise @var{value} is 1 if the positive form of the
774 option was used and 0 if the ``no-'' form was.
777 @deftypefn {Target Hook} bool TARGET_HANDLE_C_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
778 This target hook is called whenever the user specifies one of the
779 target-specific C language family options described by the @file{.opt}
780 definition files(@pxref{Options}). It has the opportunity to do some
781 option-specific processing and should return true if the option is
782 valid. The default definition does nothing but return false.
784 In general, you should use @code{TARGET_HANDLE_OPTION} to handle
785 options. However, if processing an option requires routines that are
786 only available in the C (and related language) front ends, then you
787 should use @code{TARGET_HANDLE_C_OPTION} instead.
790 @defmac TARGET_VERSION
791 This macro is a C statement to print on @code{stderr} a string
792 describing the particular machine description choice. Every machine
793 description should define @code{TARGET_VERSION}. For example:
797 #define TARGET_VERSION \
798 fprintf (stderr, " (68k, Motorola syntax)");
800 #define TARGET_VERSION \
801 fprintf (stderr, " (68k, MIT syntax)");
806 @defmac OVERRIDE_OPTIONS
807 Sometimes certain combinations of command options do not make sense on
808 a particular target machine. You can define a macro
809 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
810 defined, is executed once just after all the command options have been
813 Don't use this macro to turn on various extra optimizations for
814 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
817 @defmac C_COMMON_OVERRIDE_OPTIONS
818 This is similar to @code{OVERRIDE_OPTIONS} but is only used in the C
819 language frontends (C, Objective-C, C++, Objective-C++) and so can be
820 used to alter option flag variables which only exist in those
824 @defmac OPTIMIZATION_OPTIONS (@var{level}, @var{size})
825 Some machines may desire to change what optimizations are performed for
826 various optimization levels. This macro, if defined, is executed once
827 just after the optimization level is determined and before the remainder
828 of the command options have been parsed. Values set in this macro are
829 used as the default values for the other command line options.
831 @var{level} is the optimization level specified; 2 if @option{-O2} is
832 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
834 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
836 You should not use this macro to change options that are not
837 machine-specific. These should uniformly selected by the same
838 optimization level on all supported machines. Use this macro to enable
839 machine-specific optimizations.
841 @strong{Do not examine @code{write_symbols} in
842 this macro!} The debugging options are not supposed to alter the
846 @deftypefn {Target Hook} bool TARGET_HELP (void)
847 This hook is called in response to the user invoking
848 @option{--target-help} on the command line. It gives the target a
849 chance to display extra information on the target specific command
850 line options found in its @file{.opt} file.
853 @defmac CAN_DEBUG_WITHOUT_FP
854 Define this macro if debugging can be performed even without a frame
855 pointer. If this macro is defined, GCC will turn on the
856 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
859 @node Per-Function Data
860 @section Defining data structures for per-function information.
861 @cindex per-function data
862 @cindex data structures
864 If the target needs to store information on a per-function basis, GCC
865 provides a macro and a couple of variables to allow this. Note, just
866 using statics to store the information is a bad idea, since GCC supports
867 nested functions, so you can be halfway through encoding one function
868 when another one comes along.
870 GCC defines a data structure called @code{struct function} which
871 contains all of the data specific to an individual function. This
872 structure contains a field called @code{machine} whose type is
873 @code{struct machine_function *}, which can be used by targets to point
874 to their own specific data.
876 If a target needs per-function specific data it should define the type
877 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
878 This macro should be used to initialize the function pointer
879 @code{init_machine_status}. This pointer is explained below.
881 One typical use of per-function, target specific data is to create an
882 RTX to hold the register containing the function's return address. This
883 RTX can then be used to implement the @code{__builtin_return_address}
884 function, for level 0.
886 Note---earlier implementations of GCC used a single data area to hold
887 all of the per-function information. Thus when processing of a nested
888 function began the old per-function data had to be pushed onto a
889 stack, and when the processing was finished, it had to be popped off the
890 stack. GCC used to provide function pointers called
891 @code{save_machine_status} and @code{restore_machine_status} to handle
892 the saving and restoring of the target specific information. Since the
893 single data area approach is no longer used, these pointers are no
896 @defmac INIT_EXPANDERS
897 Macro called to initialize any target specific information. This macro
898 is called once per function, before generation of any RTL has begun.
899 The intention of this macro is to allow the initialization of the
900 function pointer @code{init_machine_status}.
903 @deftypevar {void (*)(struct function *)} init_machine_status
904 If this function pointer is non-@code{NULL} it will be called once per
905 function, before function compilation starts, in order to allow the
906 target to perform any target specific initialization of the
907 @code{struct function} structure. It is intended that this would be
908 used to initialize the @code{machine} of that structure.
910 @code{struct machine_function} structures are expected to be freed by GC@.
911 Generally, any memory that they reference must be allocated by using
912 @code{ggc_alloc}, including the structure itself.
916 @section Storage Layout
917 @cindex storage layout
919 Note that the definitions of the macros in this table which are sizes or
920 alignments measured in bits do not need to be constant. They can be C
921 expressions that refer to static variables, such as the @code{target_flags}.
922 @xref{Run-time Target}.
924 @defmac BITS_BIG_ENDIAN
925 Define this macro to have the value 1 if the most significant bit in a
926 byte has the lowest number; otherwise define it to have the value zero.
927 This means that bit-field instructions count from the most significant
928 bit. If the machine has no bit-field instructions, then this must still
929 be defined, but it doesn't matter which value it is defined to. This
930 macro need not be a constant.
932 This macro does not affect the way structure fields are packed into
933 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
936 @defmac BYTES_BIG_ENDIAN
937 Define this macro to have the value 1 if the most significant byte in a
938 word has the lowest number. This macro need not be a constant.
941 @defmac WORDS_BIG_ENDIAN
942 Define this macro to have the value 1 if, in a multiword object, the
943 most significant word has the lowest number. This applies to both
944 memory locations and registers; GCC fundamentally assumes that the
945 order of words in memory is the same as the order in registers. This
946 macro need not be a constant.
949 @defmac LIBGCC2_WORDS_BIG_ENDIAN
950 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
951 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
952 used only when compiling @file{libgcc2.c}. Typically the value will be set
953 based on preprocessor defines.
956 @defmac FLOAT_WORDS_BIG_ENDIAN
957 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
958 @code{TFmode} floating point numbers are stored in memory with the word
959 containing the sign bit at the lowest address; otherwise define it to
960 have the value 0. This macro need not be a constant.
962 You need not define this macro if the ordering is the same as for
966 @defmac BITS_PER_UNIT
967 Define this macro to be the number of bits in an addressable storage
968 unit (byte). If you do not define this macro the default is 8.
971 @defmac BITS_PER_WORD
972 Number of bits in a word. If you do not define this macro, the default
973 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
976 @defmac MAX_BITS_PER_WORD
977 Maximum number of bits in a word. If this is undefined, the default is
978 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
979 largest value that @code{BITS_PER_WORD} can have at run-time.
982 @defmac UNITS_PER_WORD
983 Number of storage units in a word; normally the size of a general-purpose
984 register, a power of two from 1 or 8.
987 @defmac MIN_UNITS_PER_WORD
988 Minimum number of units in a word. If this is undefined, the default is
989 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
990 smallest value that @code{UNITS_PER_WORD} can have at run-time.
993 @defmac UNITS_PER_SIMD_WORD (@var{mode})
994 Number of units in the vectors that the vectorizer can produce for
995 scalar mode @var{mode}. The default is equal to @code{UNITS_PER_WORD},
996 because the vectorizer can do some transformations even in absence of
997 specialized @acronym{SIMD} hardware.
1000 @defmac POINTER_SIZE
1001 Width of a pointer, in bits. You must specify a value no wider than the
1002 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
1003 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
1004 a value the default is @code{BITS_PER_WORD}.
1007 @defmac POINTERS_EXTEND_UNSIGNED
1008 A C expression that determines how pointers should be extended from
1009 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
1010 greater than zero if pointers should be zero-extended, zero if they
1011 should be sign-extended, and negative if some other sort of conversion
1012 is needed. In the last case, the extension is done by the target's
1013 @code{ptr_extend} instruction.
1015 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
1016 and @code{word_mode} are all the same width.
1019 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
1020 A macro to update @var{m} and @var{unsignedp} when an object whose type
1021 is @var{type} and which has the specified mode and signedness is to be
1022 stored in a register. This macro is only called when @var{type} is a
1025 On most RISC machines, which only have operations that operate on a full
1026 register, define this macro to set @var{m} to @code{word_mode} if
1027 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
1028 cases, only integer modes should be widened because wider-precision
1029 floating-point operations are usually more expensive than their narrower
1032 For most machines, the macro definition does not change @var{unsignedp}.
1033 However, some machines, have instructions that preferentially handle
1034 either signed or unsigned quantities of certain modes. For example, on
1035 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
1036 sign-extend the result to 64 bits. On such machines, set
1037 @var{unsignedp} according to which kind of extension is more efficient.
1039 Do not define this macro if it would never modify @var{m}.
1042 @defmac PROMOTE_FUNCTION_MODE
1043 Like @code{PROMOTE_MODE}, but is applied to outgoing function arguments or
1044 function return values, as specified by @code{TARGET_PROMOTE_FUNCTION_ARGS}
1045 and @code{TARGET_PROMOTE_FUNCTION_RETURN}, respectively.
1047 The default is @code{PROMOTE_MODE}.
1050 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_ARGS (tree @var{fntype})
1051 This target hook should return @code{true} if the promotion described by
1052 @code{PROMOTE_FUNCTION_MODE} should be done for outgoing function
1056 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_RETURN (tree @var{fntype})
1057 This target hook should return @code{true} if the promotion described by
1058 @code{PROMOTE_FUNCTION_MODE} should be done for the return value of
1061 If this target hook returns @code{true}, @code{TARGET_FUNCTION_VALUE}
1062 must perform the same promotions done by @code{PROMOTE_FUNCTION_MODE}.
1065 @defmac PARM_BOUNDARY
1066 Normal alignment required for function parameters on the stack, in
1067 bits. All stack parameters receive at least this much alignment
1068 regardless of data type. On most machines, this is the same as the
1072 @defmac STACK_BOUNDARY
1073 Define this macro to the minimum alignment enforced by hardware for the
1074 stack pointer on this machine. The definition is a C expression for the
1075 desired alignment (measured in bits). This value is used as a default
1076 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1077 this should be the same as @code{PARM_BOUNDARY}.
1080 @defmac PREFERRED_STACK_BOUNDARY
1081 Define this macro if you wish to preserve a certain alignment for the
1082 stack pointer, greater than what the hardware enforces. The definition
1083 is a C expression for the desired alignment (measured in bits). This
1084 macro must evaluate to a value equal to or larger than
1085 @code{STACK_BOUNDARY}.
1088 @defmac FUNCTION_BOUNDARY
1089 Alignment required for a function entry point, in bits.
1092 @defmac BIGGEST_ALIGNMENT
1093 Biggest alignment that any data type can require on this machine, in
1094 bits. Note that this is not the biggest alignment that is supported,
1095 just the biggest alignment that, when violated, may cause a fault.
1098 @defmac MINIMUM_ATOMIC_ALIGNMENT
1099 If defined, the smallest alignment, in bits, that can be given to an
1100 object that can be referenced in one operation, without disturbing any
1101 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1102 on machines that don't have byte or half-word store operations.
1105 @defmac BIGGEST_FIELD_ALIGNMENT
1106 Biggest alignment that any structure or union field can require on this
1107 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1108 structure and union fields only, unless the field alignment has been set
1109 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1112 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1113 An expression for the alignment of a structure field @var{field} if the
1114 alignment computed in the usual way (including applying of
1115 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1116 alignment) is @var{computed}. It overrides alignment only if the
1117 field alignment has not been set by the
1118 @code{__attribute__ ((aligned (@var{n})))} construct.
1121 @defmac MAX_OFILE_ALIGNMENT
1122 Biggest alignment supported by the object file format of this machine.
1123 Use this macro to limit the alignment which can be specified using the
1124 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1125 the default value is @code{BIGGEST_ALIGNMENT}.
1127 On systems that use ELF, the default (in @file{config/elfos.h}) is
1128 the largest supported 32-bit ELF section alignment representable on
1129 a 32-bit host e.g. @samp{(((unsigned HOST_WIDEST_INT) 1 << 28) * 8)}.
1130 On 32-bit ELF the largest supported section alignment in bits is
1131 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1134 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1135 If defined, a C expression to compute the alignment for a variable in
1136 the static store. @var{type} is the data type, and @var{basic-align} is
1137 the alignment that the object would ordinarily have. The value of this
1138 macro is used instead of that alignment to align the object.
1140 If this macro is not defined, then @var{basic-align} is used.
1143 One use of this macro is to increase alignment of medium-size data to
1144 make it all fit in fewer cache lines. Another is to cause character
1145 arrays to be word-aligned so that @code{strcpy} calls that copy
1146 constants to character arrays can be done inline.
1149 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1150 If defined, a C expression to compute the alignment given to a constant
1151 that is being placed in memory. @var{constant} is the constant and
1152 @var{basic-align} is the alignment that the object would ordinarily
1153 have. The value of this macro is used instead of that alignment to
1156 If this macro is not defined, then @var{basic-align} is used.
1158 The typical use of this macro is to increase alignment for string
1159 constants to be word aligned so that @code{strcpy} calls that copy
1160 constants can be done inline.
1163 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1164 If defined, a C expression to compute the alignment for a variable in
1165 the local store. @var{type} is the data type, and @var{basic-align} is
1166 the alignment that the object would ordinarily have. The value of this
1167 macro is used instead of that alignment to align the object.
1169 If this macro is not defined, then @var{basic-align} is used.
1171 One use of this macro is to increase alignment of medium-size data to
1172 make it all fit in fewer cache lines.
1175 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1176 If defined, a C expression to compute the alignment for stack slot.
1177 @var{type} is the data type, @var{mode} is the widest mode available,
1178 and @var{basic-align} is the alignment that the slot would ordinarily
1179 have. The value of this macro is used instead of that alignment to
1182 If this macro is not defined, then @var{basic-align} is used when
1183 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1186 This macro is to set alignment of stack slot to the maximum alignment
1187 of all possible modes which the slot may have.
1190 @defmac EMPTY_FIELD_BOUNDARY
1191 Alignment in bits to be given to a structure bit-field that follows an
1192 empty field such as @code{int : 0;}.
1194 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1197 @defmac STRUCTURE_SIZE_BOUNDARY
1198 Number of bits which any structure or union's size must be a multiple of.
1199 Each structure or union's size is rounded up to a multiple of this.
1201 If you do not define this macro, the default is the same as
1202 @code{BITS_PER_UNIT}.
1205 @defmac STRICT_ALIGNMENT
1206 Define this macro to be the value 1 if instructions will fail to work
1207 if given data not on the nominal alignment. If instructions will merely
1208 go slower in that case, define this macro as 0.
1211 @defmac PCC_BITFIELD_TYPE_MATTERS
1212 Define this if you wish to imitate the way many other C compilers handle
1213 alignment of bit-fields and the structures that contain them.
1215 The behavior is that the type written for a named bit-field (@code{int},
1216 @code{short}, or other integer type) imposes an alignment for the entire
1217 structure, as if the structure really did contain an ordinary field of
1218 that type. In addition, the bit-field is placed within the structure so
1219 that it would fit within such a field, not crossing a boundary for it.
1221 Thus, on most machines, a named bit-field whose type is written as
1222 @code{int} would not cross a four-byte boundary, and would force
1223 four-byte alignment for the whole structure. (The alignment used may
1224 not be four bytes; it is controlled by the other alignment parameters.)
1226 An unnamed bit-field will not affect the alignment of the containing
1229 If the macro is defined, its definition should be a C expression;
1230 a nonzero value for the expression enables this behavior.
1232 Note that if this macro is not defined, or its value is zero, some
1233 bit-fields may cross more than one alignment boundary. The compiler can
1234 support such references if there are @samp{insv}, @samp{extv}, and
1235 @samp{extzv} insns that can directly reference memory.
1237 The other known way of making bit-fields work is to define
1238 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1239 Then every structure can be accessed with fullwords.
1241 Unless the machine has bit-field instructions or you define
1242 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1243 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1245 If your aim is to make GCC use the same conventions for laying out
1246 bit-fields as are used by another compiler, here is how to investigate
1247 what the other compiler does. Compile and run this program:
1266 printf ("Size of foo1 is %d\n",
1267 sizeof (struct foo1));
1268 printf ("Size of foo2 is %d\n",
1269 sizeof (struct foo2));
1274 If this prints 2 and 5, then the compiler's behavior is what you would
1275 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1278 @defmac BITFIELD_NBYTES_LIMITED
1279 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1280 to aligning a bit-field within the structure.
1283 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELDS (void)
1284 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1285 whether unnamed bitfields affect the alignment of the containing
1286 structure. The hook should return true if the structure should inherit
1287 the alignment requirements of an unnamed bitfield's type.
1290 @deftypefn {Target Hook} bool TARGET_NARROW_VOLATILE_BITFIELDS (void)
1291 This target hook should return @code{true} if accesses to volatile bitfields
1292 should use the narrowest mode possible. It should return @code{false} if
1293 these accesses should use the bitfield container type.
1295 The default is @code{!TARGET_STRICT_ALIGN}.
1298 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1299 Return 1 if a structure or array containing @var{field} should be accessed using
1302 If @var{field} is the only field in the structure, @var{mode} is its
1303 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1304 case where structures of one field would require the structure's mode to
1305 retain the field's mode.
1307 Normally, this is not needed.
1310 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1311 Define this macro as an expression for the alignment of a type (given
1312 by @var{type} as a tree node) if the alignment computed in the usual
1313 way is @var{computed} and the alignment explicitly specified was
1316 The default is to use @var{specified} if it is larger; otherwise, use
1317 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1320 @defmac MAX_FIXED_MODE_SIZE
1321 An integer expression for the size in bits of the largest integer
1322 machine mode that should actually be used. All integer machine modes of
1323 this size or smaller can be used for structures and unions with the
1324 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1325 (DImode)} is assumed.
1328 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1329 If defined, an expression of type @code{enum machine_mode} that
1330 specifies the mode of the save area operand of a
1331 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1332 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1333 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1334 having its mode specified.
1336 You need not define this macro if it always returns @code{Pmode}. You
1337 would most commonly define this macro if the
1338 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1342 @defmac STACK_SIZE_MODE
1343 If defined, an expression of type @code{enum machine_mode} that
1344 specifies the mode of the size increment operand of an
1345 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1347 You need not define this macro if it always returns @code{word_mode}.
1348 You would most commonly define this macro if the @code{allocate_stack}
1349 pattern needs to support both a 32- and a 64-bit mode.
1352 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_CMP_RETURN_MODE ()
1353 This target hook should return the mode to be used for the return value
1354 of compare instructions expanded to libgcc calls. If not defined
1355 @code{word_mode} is returned which is the right choice for a majority of
1359 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_SHIFT_COUNT_MODE ()
1360 This target hook should return the mode to be used for the shift count operand
1361 of shift instructions expanded to libgcc calls. If not defined
1362 @code{word_mode} is returned which is the right choice for a majority of
1366 @defmac TARGET_FLOAT_FORMAT
1367 A code distinguishing the floating point format of the target machine.
1368 There are two defined values:
1371 @item IEEE_FLOAT_FORMAT
1372 This code indicates IEEE floating point. It is the default; there is no
1373 need to define @code{TARGET_FLOAT_FORMAT} when the format is IEEE@.
1375 @item VAX_FLOAT_FORMAT
1376 This code indicates the ``F float'' (for @code{float}) and ``D float''
1377 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
1380 If your target uses a floating point format other than these, you must
1381 define a new @var{name}_FLOAT_FORMAT code for it, and add support for
1382 it to @file{real.c}.
1384 The ordering of the component words of floating point values stored in
1385 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1388 @defmac MODE_HAS_NANS (@var{mode})
1389 When defined, this macro should be true if @var{mode} has a NaN
1390 representation. The compiler assumes that NaNs are not equal to
1391 anything (including themselves) and that addition, subtraction,
1392 multiplication and division all return NaNs when one operand is
1395 By default, this macro is true if @var{mode} is a floating-point
1396 mode and the target floating-point format is IEEE@.
1399 @defmac MODE_HAS_INFINITIES (@var{mode})
1400 This macro should be true if @var{mode} can represent infinity. At
1401 present, the compiler uses this macro to decide whether @samp{x - x}
1402 is always defined. By default, the macro is true when @var{mode}
1403 is a floating-point mode and the target format is IEEE@.
1406 @defmac MODE_HAS_SIGNED_ZEROS (@var{mode})
1407 True if @var{mode} distinguishes between positive and negative zero.
1408 The rules are expected to follow the IEEE standard:
1412 @samp{x + x} has the same sign as @samp{x}.
1415 If the sum of two values with opposite sign is zero, the result is
1416 positive for all rounding modes expect towards @minus{}infinity, for
1417 which it is negative.
1420 The sign of a product or quotient is negative when exactly one
1421 of the operands is negative.
1424 The default definition is true if @var{mode} is a floating-point
1425 mode and the target format is IEEE@.
1428 @defmac MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1429 If defined, this macro should be true for @var{mode} if it has at
1430 least one rounding mode in which @samp{x} and @samp{-x} can be
1431 rounded to numbers of different magnitude. Two such modes are
1432 towards @minus{}infinity and towards +infinity.
1434 The default definition of this macro is true if @var{mode} is
1435 a floating-point mode and the target format is IEEE@.
1438 @defmac ROUND_TOWARDS_ZERO
1439 If defined, this macro should be true if the prevailing rounding
1440 mode is towards zero. A true value has the following effects:
1444 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1447 @file{libgcc.a}'s floating-point emulator will round towards zero
1448 rather than towards nearest.
1451 The compiler's floating-point emulator will round towards zero after
1452 doing arithmetic, and when converting from the internal float format to
1456 The macro does not affect the parsing of string literals. When the
1457 primary rounding mode is towards zero, library functions like
1458 @code{strtod} might still round towards nearest, and the compiler's
1459 parser should behave like the target's @code{strtod} where possible.
1461 Not defining this macro is equivalent to returning zero.
1464 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1465 This macro should return true if floats with @var{size}
1466 bits do not have a NaN or infinity representation, but use the largest
1467 exponent for normal numbers instead.
1469 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1470 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1471 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1472 floating-point arithmetic.
1474 The default definition of this macro returns false for all sizes.
1477 @deftypefn {Target Hook} bool TARGET_VECTOR_OPAQUE_P (tree @var{type})
1478 This target hook should return @code{true} a vector is opaque. That
1479 is, if no cast is needed when copying a vector value of type
1480 @var{type} into another vector lvalue of the same size. Vector opaque
1481 types cannot be initialized. The default is that there are no such
1485 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1486 This target hook returns @code{true} if bit-fields in the given
1487 @var{record_type} are to be laid out following the rules of Microsoft
1488 Visual C/C++, namely: (i) a bit-field won't share the same storage
1489 unit with the previous bit-field if their underlying types have
1490 different sizes, and the bit-field will be aligned to the highest
1491 alignment of the underlying types of itself and of the previous
1492 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1493 the whole enclosing structure, even if it is unnamed; except that
1494 (iii) a zero-sized bit-field will be disregarded unless it follows
1495 another bit-field of nonzero size. If this hook returns @code{true},
1496 other macros that control bit-field layout are ignored.
1498 When a bit-field is inserted into a packed record, the whole size
1499 of the underlying type is used by one or more same-size adjacent
1500 bit-fields (that is, if its long:3, 32 bits is used in the record,
1501 and any additional adjacent long bit-fields are packed into the same
1502 chunk of 32 bits. However, if the size changes, a new field of that
1503 size is allocated). In an unpacked record, this is the same as using
1504 alignment, but not equivalent when packing.
1506 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1507 the latter will take precedence. If @samp{__attribute__((packed))} is
1508 used on a single field when MS bit-fields are in use, it will take
1509 precedence for that field, but the alignment of the rest of the structure
1510 may affect its placement.
1513 @deftypefn {Target Hook} {bool} TARGET_DECIMAL_FLOAT_SUPPORTED_P (void)
1514 Returns true if the target supports decimal floating point.
1517 @deftypefn {Target Hook} {bool} TARGET_FIXED_POINT_SUPPORTED_P (void)
1518 Returns true if the target supports fixed-point arithmetic.
1521 @deftypefn {Target Hook} void TARGET_EXPAND_TO_RTL_HOOK (void)
1522 This hook is called just before expansion into rtl, allowing the target
1523 to perform additional initializations or analysis before the expansion.
1524 For example, the rs6000 port uses it to allocate a scratch stack slot
1525 for use in copying SDmode values between memory and floating point
1526 registers whenever the function being expanded has any SDmode
1530 @deftypefn {Target Hook} void TARGET_INSTANTIATE_DECLS (void)
1531 This hook allows the backend to perform additional instantiations on rtl
1532 that are not actually in any insns yet, but will be later.
1535 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_TYPE (tree @var{type})
1536 If your target defines any fundamental types, or any types your target
1537 uses should be mangled differently from the default, define this hook
1538 to return the appropriate encoding for these types as part of a C++
1539 mangled name. The @var{type} argument is the tree structure representing
1540 the type to be mangled. The hook may be applied to trees which are
1541 not target-specific fundamental types; it should return @code{NULL}
1542 for all such types, as well as arguments it does not recognize. If the
1543 return value is not @code{NULL}, it must point to a statically-allocated
1546 Target-specific fundamental types might be new fundamental types or
1547 qualified versions of ordinary fundamental types. Encode new
1548 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1549 is the name used for the type in source code, and @var{n} is the
1550 length of @var{name} in decimal. Encode qualified versions of
1551 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1552 @var{name} is the name used for the type qualifier in source code,
1553 @var{n} is the length of @var{name} as above, and @var{code} is the
1554 code used to represent the unqualified version of this type. (See
1555 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1556 codes.) In both cases the spaces are for clarity; do not include any
1557 spaces in your string.
1559 This hook is applied to types prior to typedef resolution. If the mangled
1560 name for a particular type depends only on that type's main variant, you
1561 can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT}
1564 The default version of this hook always returns @code{NULL}, which is
1565 appropriate for a target that does not define any new fundamental
1570 @section Layout of Source Language Data Types
1572 These macros define the sizes and other characteristics of the standard
1573 basic data types used in programs being compiled. Unlike the macros in
1574 the previous section, these apply to specific features of C and related
1575 languages, rather than to fundamental aspects of storage layout.
1577 @defmac INT_TYPE_SIZE
1578 A C expression for the size in bits of the type @code{int} on the
1579 target machine. If you don't define this, the default is one word.
1582 @defmac SHORT_TYPE_SIZE
1583 A C expression for the size in bits of the type @code{short} on the
1584 target machine. If you don't define this, the default is half a word.
1585 (If this would be less than one storage unit, it is rounded up to one
1589 @defmac LONG_TYPE_SIZE
1590 A C expression for the size in bits of the type @code{long} on the
1591 target machine. If you don't define this, the default is one word.
1594 @defmac ADA_LONG_TYPE_SIZE
1595 On some machines, the size used for the Ada equivalent of the type
1596 @code{long} by a native Ada compiler differs from that used by C@. In
1597 that situation, define this macro to be a C expression to be used for
1598 the size of that type. If you don't define this, the default is the
1599 value of @code{LONG_TYPE_SIZE}.
1602 @defmac LONG_LONG_TYPE_SIZE
1603 A C expression for the size in bits of the type @code{long long} on the
1604 target machine. If you don't define this, the default is two
1605 words. If you want to support GNU Ada on your machine, the value of this
1606 macro must be at least 64.
1609 @defmac CHAR_TYPE_SIZE
1610 A C expression for the size in bits of the type @code{char} on the
1611 target machine. If you don't define this, the default is
1612 @code{BITS_PER_UNIT}.
1615 @defmac BOOL_TYPE_SIZE
1616 A C expression for the size in bits of the C++ type @code{bool} and
1617 C99 type @code{_Bool} on the target machine. If you don't define
1618 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1621 @defmac FLOAT_TYPE_SIZE
1622 A C expression for the size in bits of the type @code{float} on the
1623 target machine. If you don't define this, the default is one word.
1626 @defmac DOUBLE_TYPE_SIZE
1627 A C expression for the size in bits of the type @code{double} on the
1628 target machine. If you don't define this, the default is two
1632 @defmac LONG_DOUBLE_TYPE_SIZE
1633 A C expression for the size in bits of the type @code{long double} on
1634 the target machine. If you don't define this, the default is two
1638 @defmac SHORT_FRACT_TYPE_SIZE
1639 A C expression for the size in bits of the type @code{short _Fract} on
1640 the target machine. If you don't define this, the default is
1641 @code{BITS_PER_UNIT}.
1644 @defmac FRACT_TYPE_SIZE
1645 A C expression for the size in bits of the type @code{_Fract} on
1646 the target machine. If you don't define this, the default is
1647 @code{BITS_PER_UNIT * 2}.
1650 @defmac LONG_FRACT_TYPE_SIZE
1651 A C expression for the size in bits of the type @code{long _Fract} on
1652 the target machine. If you don't define this, the default is
1653 @code{BITS_PER_UNIT * 4}.
1656 @defmac LONG_LONG_FRACT_TYPE_SIZE
1657 A C expression for the size in bits of the type @code{long long _Fract} on
1658 the target machine. If you don't define this, the default is
1659 @code{BITS_PER_UNIT * 8}.
1662 @defmac SHORT_ACCUM_TYPE_SIZE
1663 A C expression for the size in bits of the type @code{short _Accum} on
1664 the target machine. If you don't define this, the default is
1665 @code{BITS_PER_UNIT * 2}.
1668 @defmac ACCUM_TYPE_SIZE
1669 A C expression for the size in bits of the type @code{_Accum} on
1670 the target machine. If you don't define this, the default is
1671 @code{BITS_PER_UNIT * 4}.
1674 @defmac LONG_ACCUM_TYPE_SIZE
1675 A C expression for the size in bits of the type @code{long _Accum} on
1676 the target machine. If you don't define this, the default is
1677 @code{BITS_PER_UNIT * 8}.
1680 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1681 A C expression for the size in bits of the type @code{long long _Accum} on
1682 the target machine. If you don't define this, the default is
1683 @code{BITS_PER_UNIT * 16}.
1686 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1687 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1688 if you want routines in @file{libgcc2.a} for a size other than
1689 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1690 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1693 @defmac LIBGCC2_HAS_DF_MODE
1694 Define this macro if neither @code{LIBGCC2_DOUBLE_TYPE_SIZE} nor
1695 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1696 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1697 anyway. If you don't define this and either @code{LIBGCC2_DOUBLE_TYPE_SIZE}
1698 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1702 @defmac LIBGCC2_HAS_XF_MODE
1703 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1704 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1705 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1706 is 80 then the default is 1, otherwise it is 0.
1709 @defmac LIBGCC2_HAS_TF_MODE
1710 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1711 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1712 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1713 is 128 then the default is 1, otherwise it is 0.
1720 Define these macros to be the size in bits of the mantissa of
1721 @code{SFmode}, @code{DFmode}, @code{XFmode} and @code{TFmode} values,
1722 if the defaults in @file{libgcc2.h} are inappropriate. By default,
1723 @code{FLT_MANT_DIG} is used for @code{SF_SIZE}, @code{LDBL_MANT_DIG}
1724 for @code{XF_SIZE} and @code{TF_SIZE}, and @code{DBL_MANT_DIG} or
1725 @code{LDBL_MANT_DIG} for @code{DF_SIZE} according to whether
1726 @code{LIBGCC2_DOUBLE_TYPE_SIZE} or
1727 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64.
1730 @defmac TARGET_FLT_EVAL_METHOD
1731 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1732 assuming, if applicable, that the floating-point control word is in its
1733 default state. If you do not define this macro the value of
1734 @code{FLT_EVAL_METHOD} will be zero.
1737 @defmac WIDEST_HARDWARE_FP_SIZE
1738 A C expression for the size in bits of the widest floating-point format
1739 supported by the hardware. If you define this macro, you must specify a
1740 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1741 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1745 @defmac DEFAULT_SIGNED_CHAR
1746 An expression whose value is 1 or 0, according to whether the type
1747 @code{char} should be signed or unsigned by default. The user can
1748 always override this default with the options @option{-fsigned-char}
1749 and @option{-funsigned-char}.
1752 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1753 This target hook should return true if the compiler should give an
1754 @code{enum} type only as many bytes as it takes to represent the range
1755 of possible values of that type. It should return false if all
1756 @code{enum} types should be allocated like @code{int}.
1758 The default is to return false.
1762 A C expression for a string describing the name of the data type to use
1763 for size values. The typedef name @code{size_t} is defined using the
1764 contents of the string.
1766 The string can contain more than one keyword. If so, separate them with
1767 spaces, and write first any length keyword, then @code{unsigned} if
1768 appropriate, and finally @code{int}. The string must exactly match one
1769 of the data type names defined in the function
1770 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1771 omit @code{int} or change the order---that would cause the compiler to
1774 If you don't define this macro, the default is @code{"long unsigned
1778 @defmac PTRDIFF_TYPE
1779 A C expression for a string describing the name of the data type to use
1780 for the result of subtracting two pointers. The typedef name
1781 @code{ptrdiff_t} is defined using the contents of the string. See
1782 @code{SIZE_TYPE} above for more information.
1784 If you don't define this macro, the default is @code{"long int"}.
1788 A C expression for a string describing the name of the data type to use
1789 for wide characters. The typedef name @code{wchar_t} is defined using
1790 the contents of the string. See @code{SIZE_TYPE} above for more
1793 If you don't define this macro, the default is @code{"int"}.
1796 @defmac WCHAR_TYPE_SIZE
1797 A C expression for the size in bits of the data type for wide
1798 characters. This is used in @code{cpp}, which cannot make use of
1803 A C expression for a string describing the name of the data type to
1804 use for wide characters passed to @code{printf} and returned from
1805 @code{getwc}. The typedef name @code{wint_t} is defined using the
1806 contents of the string. See @code{SIZE_TYPE} above for more
1809 If you don't define this macro, the default is @code{"unsigned int"}.
1813 A C expression for a string describing the name of the data type that
1814 can represent any value of any standard or extended signed integer type.
1815 The typedef name @code{intmax_t} is defined using the contents of the
1816 string. See @code{SIZE_TYPE} above for more information.
1818 If you don't define this macro, the default is the first of
1819 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1820 much precision as @code{long long int}.
1823 @defmac UINTMAX_TYPE
1824 A C expression for a string describing the name of the data type that
1825 can represent any value of any standard or extended unsigned integer
1826 type. The typedef name @code{uintmax_t} is defined using the contents
1827 of the string. See @code{SIZE_TYPE} above for more information.
1829 If you don't define this macro, the default is the first of
1830 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1831 unsigned int"} that has as much precision as @code{long long unsigned
1835 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1836 The C++ compiler represents a pointer-to-member-function with a struct
1843 ptrdiff_t vtable_index;
1850 The C++ compiler must use one bit to indicate whether the function that
1851 will be called through a pointer-to-member-function is virtual.
1852 Normally, we assume that the low-order bit of a function pointer must
1853 always be zero. Then, by ensuring that the vtable_index is odd, we can
1854 distinguish which variant of the union is in use. But, on some
1855 platforms function pointers can be odd, and so this doesn't work. In
1856 that case, we use the low-order bit of the @code{delta} field, and shift
1857 the remainder of the @code{delta} field to the left.
1859 GCC will automatically make the right selection about where to store
1860 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1861 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1862 set such that functions always start at even addresses, but the lowest
1863 bit of pointers to functions indicate whether the function at that
1864 address is in ARM or Thumb mode. If this is the case of your
1865 architecture, you should define this macro to
1866 @code{ptrmemfunc_vbit_in_delta}.
1868 In general, you should not have to define this macro. On architectures
1869 in which function addresses are always even, according to
1870 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1871 @code{ptrmemfunc_vbit_in_pfn}.
1874 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1875 Normally, the C++ compiler uses function pointers in vtables. This
1876 macro allows the target to change to use ``function descriptors''
1877 instead. Function descriptors are found on targets for whom a
1878 function pointer is actually a small data structure. Normally the
1879 data structure consists of the actual code address plus a data
1880 pointer to which the function's data is relative.
1882 If vtables are used, the value of this macro should be the number
1883 of words that the function descriptor occupies.
1886 @defmac TARGET_VTABLE_ENTRY_ALIGN
1887 By default, the vtable entries are void pointers, the so the alignment
1888 is the same as pointer alignment. The value of this macro specifies
1889 the alignment of the vtable entry in bits. It should be defined only
1890 when special alignment is necessary. */
1893 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1894 There are a few non-descriptor entries in the vtable at offsets below
1895 zero. If these entries must be padded (say, to preserve the alignment
1896 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1897 of words in each data entry.
1901 @section Register Usage
1902 @cindex register usage
1904 This section explains how to describe what registers the target machine
1905 has, and how (in general) they can be used.
1907 The description of which registers a specific instruction can use is
1908 done with register classes; see @ref{Register Classes}. For information
1909 on using registers to access a stack frame, see @ref{Frame Registers}.
1910 For passing values in registers, see @ref{Register Arguments}.
1911 For returning values in registers, see @ref{Scalar Return}.
1914 * Register Basics:: Number and kinds of registers.
1915 * Allocation Order:: Order in which registers are allocated.
1916 * Values in Registers:: What kinds of values each reg can hold.
1917 * Leaf Functions:: Renumbering registers for leaf functions.
1918 * Stack Registers:: Handling a register stack such as 80387.
1921 @node Register Basics
1922 @subsection Basic Characteristics of Registers
1924 @c prevent bad page break with this line
1925 Registers have various characteristics.
1927 @defmac FIRST_PSEUDO_REGISTER
1928 Number of hardware registers known to the compiler. They receive
1929 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1930 pseudo register's number really is assigned the number
1931 @code{FIRST_PSEUDO_REGISTER}.
1934 @defmac FIXED_REGISTERS
1935 @cindex fixed register
1936 An initializer that says which registers are used for fixed purposes
1937 all throughout the compiled code and are therefore not available for
1938 general allocation. These would include the stack pointer, the frame
1939 pointer (except on machines where that can be used as a general
1940 register when no frame pointer is needed), the program counter on
1941 machines where that is considered one of the addressable registers,
1942 and any other numbered register with a standard use.
1944 This information is expressed as a sequence of numbers, separated by
1945 commas and surrounded by braces. The @var{n}th number is 1 if
1946 register @var{n} is fixed, 0 otherwise.
1948 The table initialized from this macro, and the table initialized by
1949 the following one, may be overridden at run time either automatically,
1950 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1951 the user with the command options @option{-ffixed-@var{reg}},
1952 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1955 @defmac CALL_USED_REGISTERS
1956 @cindex call-used register
1957 @cindex call-clobbered register
1958 @cindex call-saved register
1959 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1960 clobbered (in general) by function calls as well as for fixed
1961 registers. This macro therefore identifies the registers that are not
1962 available for general allocation of values that must live across
1965 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1966 automatically saves it on function entry and restores it on function
1967 exit, if the register is used within the function.
1970 @defmac CALL_REALLY_USED_REGISTERS
1971 @cindex call-used register
1972 @cindex call-clobbered register
1973 @cindex call-saved register
1974 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1975 that the entire set of @code{FIXED_REGISTERS} be included.
1976 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1977 This macro is optional. If not specified, it defaults to the value
1978 of @code{CALL_USED_REGISTERS}.
1981 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1982 @cindex call-used register
1983 @cindex call-clobbered register
1984 @cindex call-saved register
1985 A C expression that is nonzero if it is not permissible to store a
1986 value of mode @var{mode} in hard register number @var{regno} across a
1987 call without some part of it being clobbered. For most machines this
1988 macro need not be defined. It is only required for machines that do not
1989 preserve the entire contents of a register across a call.
1993 @findex call_used_regs
1996 @findex reg_class_contents
1997 @defmac CONDITIONAL_REGISTER_USAGE
1998 Zero or more C statements that may conditionally modify five variables
1999 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
2000 @code{reg_names}, and @code{reg_class_contents}, to take into account
2001 any dependence of these register sets on target flags. The first three
2002 of these are of type @code{char []} (interpreted as Boolean vectors).
2003 @code{global_regs} is a @code{const char *[]}, and
2004 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
2005 called, @code{fixed_regs}, @code{call_used_regs},
2006 @code{reg_class_contents}, and @code{reg_names} have been initialized
2007 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
2008 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
2009 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
2010 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
2011 command options have been applied.
2013 You need not define this macro if it has no work to do.
2015 @cindex disabling certain registers
2016 @cindex controlling register usage
2017 If the usage of an entire class of registers depends on the target
2018 flags, you may indicate this to GCC by using this macro to modify
2019 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
2020 registers in the classes which should not be used by GCC@. Also define
2021 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
2022 to return @code{NO_REGS} if it
2023 is called with a letter for a class that shouldn't be used.
2025 (However, if this class is not included in @code{GENERAL_REGS} and all
2026 of the insn patterns whose constraints permit this class are
2027 controlled by target switches, then GCC will automatically avoid using
2028 these registers when the target switches are opposed to them.)
2031 @defmac INCOMING_REGNO (@var{out})
2032 Define this macro if the target machine has register windows. This C
2033 expression returns the register number as seen by the called function
2034 corresponding to the register number @var{out} as seen by the calling
2035 function. Return @var{out} if register number @var{out} is not an
2039 @defmac OUTGOING_REGNO (@var{in})
2040 Define this macro if the target machine has register windows. This C
2041 expression returns the register number as seen by the calling function
2042 corresponding to the register number @var{in} as seen by the called
2043 function. Return @var{in} if register number @var{in} is not an inbound
2047 @defmac LOCAL_REGNO (@var{regno})
2048 Define this macro if the target machine has register windows. This C
2049 expression returns true if the register is call-saved but is in the
2050 register window. Unlike most call-saved registers, such registers
2051 need not be explicitly restored on function exit or during non-local
2056 If the program counter has a register number, define this as that
2057 register number. Otherwise, do not define it.
2060 @node Allocation Order
2061 @subsection Order of Allocation of Registers
2062 @cindex order of register allocation
2063 @cindex register allocation order
2065 @c prevent bad page break with this line
2066 Registers are allocated in order.
2068 @defmac REG_ALLOC_ORDER
2069 If defined, an initializer for a vector of integers, containing the
2070 numbers of hard registers in the order in which GCC should prefer
2071 to use them (from most preferred to least).
2073 If this macro is not defined, registers are used lowest numbered first
2074 (all else being equal).
2076 One use of this macro is on machines where the highest numbered
2077 registers must always be saved and the save-multiple-registers
2078 instruction supports only sequences of consecutive registers. On such
2079 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
2080 the highest numbered allocable register first.
2083 @defmac ORDER_REGS_FOR_LOCAL_ALLOC
2084 A C statement (sans semicolon) to choose the order in which to allocate
2085 hard registers for pseudo-registers local to a basic block.
2087 Store the desired register order in the array @code{reg_alloc_order}.
2088 Element 0 should be the register to allocate first; element 1, the next
2089 register; and so on.
2091 The macro body should not assume anything about the contents of
2092 @code{reg_alloc_order} before execution of the macro.
2094 On most machines, it is not necessary to define this macro.
2097 @node Values in Registers
2098 @subsection How Values Fit in Registers
2100 This section discusses the macros that describe which kinds of values
2101 (specifically, which machine modes) each register can hold, and how many
2102 consecutive registers are needed for a given mode.
2104 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2105 A C expression for the number of consecutive hard registers, starting
2106 at register number @var{regno}, required to hold a value of mode
2107 @var{mode}. This macro must never return zero, even if a register
2108 cannot hold the requested mode - indicate that with HARD_REGNO_MODE_OK
2109 and/or CANNOT_CHANGE_MODE_CLASS instead.
2111 On a machine where all registers are exactly one word, a suitable
2112 definition of this macro is
2115 #define HARD_REGNO_NREGS(REGNO, MODE) \
2116 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2121 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2122 A C expression that is nonzero if a value of mode @var{mode}, stored
2123 in memory, ends with padding that causes it to take up more space than
2124 in registers starting at register number @var{regno} (as determined by
2125 multiplying GCC's notion of the size of the register when containing
2126 this mode by the number of registers returned by
2127 @code{HARD_REGNO_NREGS}). By default this is zero.
2129 For example, if a floating-point value is stored in three 32-bit
2130 registers but takes up 128 bits in memory, then this would be
2133 This macros only needs to be defined if there are cases where
2134 @code{subreg_get_info}
2135 would otherwise wrongly determine that a @code{subreg} can be
2136 represented by an offset to the register number, when in fact such a
2137 @code{subreg} would contain some of the padding not stored in
2138 registers and so not be representable.
2141 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2142 For values of @var{regno} and @var{mode} for which
2143 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2144 returning the greater number of registers required to hold the value
2145 including any padding. In the example above, the value would be four.
2148 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2149 Define this macro if the natural size of registers that hold values
2150 of mode @var{mode} is not the word size. It is a C expression that
2151 should give the natural size in bytes for the specified mode. It is
2152 used by the register allocator to try to optimize its results. This
2153 happens for example on SPARC 64-bit where the natural size of
2154 floating-point registers is still 32-bit.
2157 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2158 A C expression that is nonzero if it is permissible to store a value
2159 of mode @var{mode} in hard register number @var{regno} (or in several
2160 registers starting with that one). For a machine where all registers
2161 are equivalent, a suitable definition is
2164 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2167 You need not include code to check for the numbers of fixed registers,
2168 because the allocation mechanism considers them to be always occupied.
2170 @cindex register pairs
2171 On some machines, double-precision values must be kept in even/odd
2172 register pairs. You can implement that by defining this macro to reject
2173 odd register numbers for such modes.
2175 The minimum requirement for a mode to be OK in a register is that the
2176 @samp{mov@var{mode}} instruction pattern support moves between the
2177 register and other hard register in the same class and that moving a
2178 value into the register and back out not alter it.
2180 Since the same instruction used to move @code{word_mode} will work for
2181 all narrower integer modes, it is not necessary on any machine for
2182 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2183 you define patterns @samp{movhi}, etc., to take advantage of this. This
2184 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2185 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2188 Many machines have special registers for floating point arithmetic.
2189 Often people assume that floating point machine modes are allowed only
2190 in floating point registers. This is not true. Any registers that
2191 can hold integers can safely @emph{hold} a floating point machine
2192 mode, whether or not floating arithmetic can be done on it in those
2193 registers. Integer move instructions can be used to move the values.
2195 On some machines, though, the converse is true: fixed-point machine
2196 modes may not go in floating registers. This is true if the floating
2197 registers normalize any value stored in them, because storing a
2198 non-floating value there would garble it. In this case,
2199 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2200 floating registers. But if the floating registers do not automatically
2201 normalize, if you can store any bit pattern in one and retrieve it
2202 unchanged without a trap, then any machine mode may go in a floating
2203 register, so you can define this macro to say so.
2205 The primary significance of special floating registers is rather that
2206 they are the registers acceptable in floating point arithmetic
2207 instructions. However, this is of no concern to
2208 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2209 constraints for those instructions.
2211 On some machines, the floating registers are especially slow to access,
2212 so that it is better to store a value in a stack frame than in such a
2213 register if floating point arithmetic is not being done. As long as the
2214 floating registers are not in class @code{GENERAL_REGS}, they will not
2215 be used unless some pattern's constraint asks for one.
2218 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2219 A C expression that is nonzero if it is OK to rename a hard register
2220 @var{from} to another hard register @var{to}.
2222 One common use of this macro is to prevent renaming of a register to
2223 another register that is not saved by a prologue in an interrupt
2226 The default is always nonzero.
2229 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2230 A C expression that is nonzero if a value of mode
2231 @var{mode1} is accessible in mode @var{mode2} without copying.
2233 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2234 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2235 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2236 should be nonzero. If they differ for any @var{r}, you should define
2237 this macro to return zero unless some other mechanism ensures the
2238 accessibility of the value in a narrower mode.
2240 You should define this macro to return nonzero in as many cases as
2241 possible since doing so will allow GCC to perform better register
2245 @defmac AVOID_CCMODE_COPIES
2246 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2247 registers. You should only define this macro if support for copying to/from
2248 @code{CCmode} is incomplete.
2251 @node Leaf Functions
2252 @subsection Handling Leaf Functions
2254 @cindex leaf functions
2255 @cindex functions, leaf
2256 On some machines, a leaf function (i.e., one which makes no calls) can run
2257 more efficiently if it does not make its own register window. Often this
2258 means it is required to receive its arguments in the registers where they
2259 are passed by the caller, instead of the registers where they would
2262 The special treatment for leaf functions generally applies only when
2263 other conditions are met; for example, often they may use only those
2264 registers for its own variables and temporaries. We use the term ``leaf
2265 function'' to mean a function that is suitable for this special
2266 handling, so that functions with no calls are not necessarily ``leaf
2269 GCC assigns register numbers before it knows whether the function is
2270 suitable for leaf function treatment. So it needs to renumber the
2271 registers in order to output a leaf function. The following macros
2274 @defmac LEAF_REGISTERS
2275 Name of a char vector, indexed by hard register number, which
2276 contains 1 for a register that is allowable in a candidate for leaf
2279 If leaf function treatment involves renumbering the registers, then the
2280 registers marked here should be the ones before renumbering---those that
2281 GCC would ordinarily allocate. The registers which will actually be
2282 used in the assembler code, after renumbering, should not be marked with 1
2285 Define this macro only if the target machine offers a way to optimize
2286 the treatment of leaf functions.
2289 @defmac LEAF_REG_REMAP (@var{regno})
2290 A C expression whose value is the register number to which @var{regno}
2291 should be renumbered, when a function is treated as a leaf function.
2293 If @var{regno} is a register number which should not appear in a leaf
2294 function before renumbering, then the expression should yield @minus{}1, which
2295 will cause the compiler to abort.
2297 Define this macro only if the target machine offers a way to optimize the
2298 treatment of leaf functions, and registers need to be renumbered to do
2302 @findex current_function_is_leaf
2303 @findex current_function_uses_only_leaf_regs
2304 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2305 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2306 specially. They can test the C variable @code{current_function_is_leaf}
2307 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2308 set prior to local register allocation and is valid for the remaining
2309 compiler passes. They can also test the C variable
2310 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2311 functions which only use leaf registers.
2312 @code{current_function_uses_only_leaf_regs} is valid after all passes
2313 that modify the instructions have been run and is only useful if
2314 @code{LEAF_REGISTERS} is defined.
2315 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2316 @c of the next paragraph?! --mew 2feb93
2318 @node Stack Registers
2319 @subsection Registers That Form a Stack
2321 There are special features to handle computers where some of the
2322 ``registers'' form a stack. Stack registers are normally written by
2323 pushing onto the stack, and are numbered relative to the top of the
2326 Currently, GCC can only handle one group of stack-like registers, and
2327 they must be consecutively numbered. Furthermore, the existing
2328 support for stack-like registers is specific to the 80387 floating
2329 point coprocessor. If you have a new architecture that uses
2330 stack-like registers, you will need to do substantial work on
2331 @file{reg-stack.c} and write your machine description to cooperate
2332 with it, as well as defining these macros.
2335 Define this if the machine has any stack-like registers.
2338 @defmac FIRST_STACK_REG
2339 The number of the first stack-like register. This one is the top
2343 @defmac LAST_STACK_REG
2344 The number of the last stack-like register. This one is the bottom of
2348 @node Register Classes
2349 @section Register Classes
2350 @cindex register class definitions
2351 @cindex class definitions, register
2353 On many machines, the numbered registers are not all equivalent.
2354 For example, certain registers may not be allowed for indexed addressing;
2355 certain registers may not be allowed in some instructions. These machine
2356 restrictions are described to the compiler using @dfn{register classes}.
2358 You define a number of register classes, giving each one a name and saying
2359 which of the registers belong to it. Then you can specify register classes
2360 that are allowed as operands to particular instruction patterns.
2364 In general, each register will belong to several classes. In fact, one
2365 class must be named @code{ALL_REGS} and contain all the registers. Another
2366 class must be named @code{NO_REGS} and contain no registers. Often the
2367 union of two classes will be another class; however, this is not required.
2369 @findex GENERAL_REGS
2370 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2371 terribly special about the name, but the operand constraint letters
2372 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2373 the same as @code{ALL_REGS}, just define it as a macro which expands
2376 Order the classes so that if class @var{x} is contained in class @var{y}
2377 then @var{x} has a lower class number than @var{y}.
2379 The way classes other than @code{GENERAL_REGS} are specified in operand
2380 constraints is through machine-dependent operand constraint letters.
2381 You can define such letters to correspond to various classes, then use
2382 them in operand constraints.
2384 You should define a class for the union of two classes whenever some
2385 instruction allows both classes. For example, if an instruction allows
2386 either a floating point (coprocessor) register or a general register for a
2387 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2388 which includes both of them. Otherwise you will get suboptimal code.
2390 You must also specify certain redundant information about the register
2391 classes: for each class, which classes contain it and which ones are
2392 contained in it; for each pair of classes, the largest class contained
2395 When a value occupying several consecutive registers is expected in a
2396 certain class, all the registers used must belong to that class.
2397 Therefore, register classes cannot be used to enforce a requirement for
2398 a register pair to start with an even-numbered register. The way to
2399 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2401 Register classes used for input-operands of bitwise-and or shift
2402 instructions have a special requirement: each such class must have, for
2403 each fixed-point machine mode, a subclass whose registers can transfer that
2404 mode to or from memory. For example, on some machines, the operations for
2405 single-byte values (@code{QImode}) are limited to certain registers. When
2406 this is so, each register class that is used in a bitwise-and or shift
2407 instruction must have a subclass consisting of registers from which
2408 single-byte values can be loaded or stored. This is so that
2409 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2411 @deftp {Data type} {enum reg_class}
2412 An enumerated type that must be defined with all the register class names
2413 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2414 must be the last register class, followed by one more enumerated value,
2415 @code{LIM_REG_CLASSES}, which is not a register class but rather
2416 tells how many classes there are.
2418 Each register class has a number, which is the value of casting
2419 the class name to type @code{int}. The number serves as an index
2420 in many of the tables described below.
2423 @defmac N_REG_CLASSES
2424 The number of distinct register classes, defined as follows:
2427 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2431 @defmac REG_CLASS_NAMES
2432 An initializer containing the names of the register classes as C string
2433 constants. These names are used in writing some of the debugging dumps.
2436 @defmac REG_CLASS_CONTENTS
2437 An initializer containing the contents of the register classes, as integers
2438 which are bit masks. The @var{n}th integer specifies the contents of class
2439 @var{n}. The way the integer @var{mask} is interpreted is that
2440 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2442 When the machine has more than 32 registers, an integer does not suffice.
2443 Then the integers are replaced by sub-initializers, braced groupings containing
2444 several integers. Each sub-initializer must be suitable as an initializer
2445 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2446 In this situation, the first integer in each sub-initializer corresponds to
2447 registers 0 through 31, the second integer to registers 32 through 63, and
2451 @defmac REGNO_REG_CLASS (@var{regno})
2452 A C expression whose value is a register class containing hard register
2453 @var{regno}. In general there is more than one such class; choose a class
2454 which is @dfn{minimal}, meaning that no smaller class also contains the
2458 @defmac BASE_REG_CLASS
2459 A macro whose definition is the name of the class to which a valid
2460 base register must belong. A base register is one used in an address
2461 which is the register value plus a displacement.
2464 @defmac MODE_BASE_REG_CLASS (@var{mode})
2465 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2466 the selection of a base register in a mode dependent manner. If
2467 @var{mode} is VOIDmode then it should return the same value as
2468 @code{BASE_REG_CLASS}.
2471 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2472 A C expression whose value is the register class to which a valid
2473 base register must belong in order to be used in a base plus index
2474 register address. You should define this macro if base plus index
2475 addresses have different requirements than other base register uses.
2478 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{outer_code}, @var{index_code})
2479 A C expression whose value is the register class to which a valid
2480 base register must belong. @var{outer_code} and @var{index_code} define the
2481 context in which the base register occurs. @var{outer_code} is the code of
2482 the immediately enclosing expression (@code{MEM} for the top level of an
2483 address, @code{ADDRESS} for something that occurs in an
2484 @code{address_operand}). @var{index_code} is the code of the corresponding
2485 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2488 @defmac INDEX_REG_CLASS
2489 A macro whose definition is the name of the class to which a valid
2490 index register must belong. An index register is one used in an
2491 address where its value is either multiplied by a scale factor or
2492 added to another register (as well as added to a displacement).
2495 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2496 A C expression which is nonzero if register number @var{num} is
2497 suitable for use as a base register in operand addresses. It may be
2498 either a suitable hard register or a pseudo register that has been
2499 allocated such a hard register.
2502 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2503 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2504 that expression may examine the mode of the memory reference in
2505 @var{mode}. You should define this macro if the mode of the memory
2506 reference affects whether a register may be used as a base register. If
2507 you define this macro, the compiler will use it instead of
2508 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2509 addresses that appear outside a @code{MEM}, i.e., as an
2510 @code{address_operand}.
2514 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2515 A C expression which is nonzero if register number @var{num} is suitable for
2516 use as a base register in base plus index operand addresses, accessing
2517 memory in mode @var{mode}. It may be either a suitable hard register or a
2518 pseudo register that has been allocated such a hard register. You should
2519 define this macro if base plus index addresses have different requirements
2520 than other base register uses.
2522 Use of this macro is deprecated; please use the more general
2523 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2526 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{outer_code}, @var{index_code})
2527 A C expression that is just like @code{REGNO_MODE_OK_FOR_BASE_P}, except
2528 that that expression may examine the context in which the register
2529 appears in the memory reference. @var{outer_code} is the code of the
2530 immediately enclosing expression (@code{MEM} if at the top level of the
2531 address, @code{ADDRESS} for something that occurs in an
2532 @code{address_operand}). @var{index_code} is the code of the
2533 corresponding index expression if @var{outer_code} is @code{PLUS};
2534 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2535 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2538 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2539 A C expression which is nonzero if register number @var{num} is
2540 suitable for use as an index register in operand addresses. It may be
2541 either a suitable hard register or a pseudo register that has been
2542 allocated such a hard register.
2544 The difference between an index register and a base register is that
2545 the index register may be scaled. If an address involves the sum of
2546 two registers, neither one of them scaled, then either one may be
2547 labeled the ``base'' and the other the ``index''; but whichever
2548 labeling is used must fit the machine's constraints of which registers
2549 may serve in each capacity. The compiler will try both labelings,
2550 looking for one that is valid, and will reload one or both registers
2551 only if neither labeling works.
2554 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2555 A C expression that places additional restrictions on the register class
2556 to use when it is necessary to copy value @var{x} into a register in class
2557 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2558 another, smaller class. On many machines, the following definition is
2562 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2565 Sometimes returning a more restrictive class makes better code. For
2566 example, on the 68000, when @var{x} is an integer constant that is in range
2567 for a @samp{moveq} instruction, the value of this macro is always
2568 @code{DATA_REGS} as long as @var{class} includes the data registers.
2569 Requiring a data register guarantees that a @samp{moveq} will be used.
2571 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2572 @var{class} is if @var{x} is a legitimate constant which cannot be
2573 loaded into some register class. By returning @code{NO_REGS} you can
2574 force @var{x} into a memory location. For example, rs6000 can load
2575 immediate values into general-purpose registers, but does not have an
2576 instruction for loading an immediate value into a floating-point
2577 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2578 @var{x} is a floating-point constant. If the constant can't be loaded
2579 into any kind of register, code generation will be better if
2580 @code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2581 of using @code{PREFERRED_RELOAD_CLASS}.
2583 If an insn has pseudos in it after register allocation, reload will go
2584 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2585 to find the best one. Returning @code{NO_REGS}, in this case, makes
2586 reload add a @code{!} in front of the constraint: the x86 back-end uses
2587 this feature to discourage usage of 387 registers when math is done in
2588 the SSE registers (and vice versa).
2591 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2592 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2593 input reloads. If you don't define this macro, the default is to use
2594 @var{class}, unchanged.
2596 You can also use @code{PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2597 reload from using some alternatives, like @code{PREFERRED_RELOAD_CLASS}.
2600 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2601 A C expression that places additional restrictions on the register class
2602 to use when it is necessary to be able to hold a value of mode
2603 @var{mode} in a reload register for which class @var{class} would
2606 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2607 there are certain modes that simply can't go in certain reload classes.
2609 The value is a register class; perhaps @var{class}, or perhaps another,
2612 Don't define this macro unless the target machine has limitations which
2613 require the macro to do something nontrivial.
2616 @deftypefn {Target Hook} {enum reg_class} TARGET_SECONDARY_RELOAD (bool @var{in_p}, rtx @var{x}, enum reg_class @var{reload_class}, enum machine_mode @var{reload_mode}, secondary_reload_info *@var{sri})
2617 Many machines have some registers that cannot be copied directly to or
2618 from memory or even from other types of registers. An example is the
2619 @samp{MQ} register, which on most machines, can only be copied to or
2620 from general registers, but not memory. Below, we shall be using the
2621 term 'intermediate register' when a move operation cannot be performed
2622 directly, but has to be done by copying the source into the intermediate
2623 register first, and then copying the intermediate register to the
2624 destination. An intermediate register always has the same mode as
2625 source and destination. Since it holds the actual value being copied,
2626 reload might apply optimizations to re-use an intermediate register
2627 and eliding the copy from the source when it can determine that the
2628 intermediate register still holds the required value.
2630 Another kind of secondary reload is required on some machines which
2631 allow copying all registers to and from memory, but require a scratch
2632 register for stores to some memory locations (e.g., those with symbolic
2633 address on the RT, and those with certain symbolic address on the SPARC
2634 when compiling PIC)@. Scratch registers need not have the same mode
2635 as the value being copied, and usually hold a different value that
2636 that being copied. Special patterns in the md file are needed to
2637 describe how the copy is performed with the help of the scratch register;
2638 these patterns also describe the number, register class(es) and mode(s)
2639 of the scratch register(s).
2641 In some cases, both an intermediate and a scratch register are required.
2643 For input reloads, this target hook is called with nonzero @var{in_p},
2644 and @var{x} is an rtx that needs to be copied to a register of class
2645 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2646 hook is called with zero @var{in_p}, and a register of class @var{reload_class}
2647 needs to be copied to rtx @var{x} in @var{reload_mode}.
2649 If copying a register of @var{reload_class} from/to @var{x} requires
2650 an intermediate register, the hook @code{secondary_reload} should
2651 return the register class required for this intermediate register.
2652 If no intermediate register is required, it should return NO_REGS.
2653 If more than one intermediate register is required, describe the one
2654 that is closest in the copy chain to the reload register.
2656 If scratch registers are needed, you also have to describe how to
2657 perform the copy from/to the reload register to/from this
2658 closest intermediate register. Or if no intermediate register is
2659 required, but still a scratch register is needed, describe the
2660 copy from/to the reload register to/from the reload operand @var{x}.
2662 You do this by setting @code{sri->icode} to the instruction code of a pattern
2663 in the md file which performs the move. Operands 0 and 1 are the output
2664 and input of this copy, respectively. Operands from operand 2 onward are
2665 for scratch operands. These scratch operands must have a mode, and a
2666 single-register-class
2667 @c [later: or memory]
2670 When an intermediate register is used, the @code{secondary_reload}
2671 hook will be called again to determine how to copy the intermediate
2672 register to/from the reload operand @var{x}, so your hook must also
2673 have code to handle the register class of the intermediate operand.
2675 @c [For later: maybe we'll allow multi-alternative reload patterns -
2676 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2677 @c and match the constraints of input and output to determine the required
2678 @c alternative. A restriction would be that constraints used to match
2679 @c against reloads registers would have to be written as register class
2680 @c constraints, or we need a new target macro / hook that tells us if an
2681 @c arbitrary constraint can match an unknown register of a given class.
2682 @c Such a macro / hook would also be useful in other places.]
2685 @var{x} might be a pseudo-register or a @code{subreg} of a
2686 pseudo-register, which could either be in a hard register or in memory.
2687 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2688 in memory and the hard register number if it is in a register.
2690 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2691 currently not supported. For the time being, you will have to continue
2692 to use @code{SECONDARY_MEMORY_NEEDED} for that purpose.
2694 @code{copy_cost} also uses this target hook to find out how values are
2695 copied. If you want it to include some extra cost for the need to allocate
2696 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2697 Or if two dependent moves are supposed to have a lower cost than the sum
2698 of the individual moves due to expected fortuitous scheduling and/or special
2699 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2702 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2703 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2704 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2705 These macros are obsolete, new ports should use the target hook
2706 @code{TARGET_SECONDARY_RELOAD} instead.
2708 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2709 target hook. Older ports still define these macros to indicate to the
2710 reload phase that it may
2711 need to allocate at least one register for a reload in addition to the
2712 register to contain the data. Specifically, if copying @var{x} to a
2713 register @var{class} in @var{mode} requires an intermediate register,
2714 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2715 largest register class all of whose registers can be used as
2716 intermediate registers or scratch registers.
2718 If copying a register @var{class} in @var{mode} to @var{x} requires an
2719 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2720 was supposed to be defined be defined to return the largest register
2721 class required. If the
2722 requirements for input and output reloads were the same, the macro
2723 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2726 The values returned by these macros are often @code{GENERAL_REGS}.
2727 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2728 can be directly copied to or from a register of @var{class} in
2729 @var{mode} without requiring a scratch register. Do not define this
2730 macro if it would always return @code{NO_REGS}.
2732 If a scratch register is required (either with or without an
2733 intermediate register), you were supposed to define patterns for
2734 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2735 (@pxref{Standard Names}. These patterns, which were normally
2736 implemented with a @code{define_expand}, should be similar to the
2737 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2740 These patterns need constraints for the reload register and scratch
2742 contain a single register class. If the original reload register (whose
2743 class is @var{class}) can meet the constraint given in the pattern, the
2744 value returned by these macros is used for the class of the scratch
2745 register. Otherwise, two additional reload registers are required.
2746 Their classes are obtained from the constraints in the insn pattern.
2748 @var{x} might be a pseudo-register or a @code{subreg} of a
2749 pseudo-register, which could either be in a hard register or in memory.
2750 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2751 in memory and the hard register number if it is in a register.
2753 These macros should not be used in the case where a particular class of
2754 registers can only be copied to memory and not to another class of
2755 registers. In that case, secondary reload registers are not needed and
2756 would not be helpful. Instead, a stack location must be used to perform
2757 the copy and the @code{mov@var{m}} pattern should use memory as an
2758 intermediate storage. This case often occurs between floating-point and
2762 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2763 Certain machines have the property that some registers cannot be copied
2764 to some other registers without using memory. Define this macro on
2765 those machines to be a C expression that is nonzero if objects of mode
2766 @var{m} in registers of @var{class1} can only be copied to registers of
2767 class @var{class2} by storing a register of @var{class1} into memory
2768 and loading that memory location into a register of @var{class2}.
2770 Do not define this macro if its value would always be zero.
2773 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2774 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2775 allocates a stack slot for a memory location needed for register copies.
2776 If this macro is defined, the compiler instead uses the memory location
2777 defined by this macro.
2779 Do not define this macro if you do not define
2780 @code{SECONDARY_MEMORY_NEEDED}.
2783 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2784 When the compiler needs a secondary memory location to copy between two
2785 registers of mode @var{mode}, it normally allocates sufficient memory to
2786 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2787 load operations in a mode that many bits wide and whose class is the
2788 same as that of @var{mode}.
2790 This is right thing to do on most machines because it ensures that all
2791 bits of the register are copied and prevents accesses to the registers
2792 in a narrower mode, which some machines prohibit for floating-point
2795 However, this default behavior is not correct on some machines, such as
2796 the DEC Alpha, that store short integers in floating-point registers
2797 differently than in integer registers. On those machines, the default
2798 widening will not work correctly and you must define this macro to
2799 suppress that widening in some cases. See the file @file{alpha.h} for
2802 Do not define this macro if you do not define
2803 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2804 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2807 @defmac SMALL_REGISTER_CLASSES
2808 On some machines, it is risky to let hard registers live across arbitrary
2809 insns. Typically, these machines have instructions that require values
2810 to be in specific registers (like an accumulator), and reload will fail
2811 if the required hard register is used for another purpose across such an
2814 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2815 value on these machines. When this macro has a nonzero value, the
2816 compiler will try to minimize the lifetime of hard registers.
2818 It is always safe to define this macro with a nonzero value, but if you
2819 unnecessarily define it, you will reduce the amount of optimizations
2820 that can be performed in some cases. If you do not define this macro
2821 with a nonzero value when it is required, the compiler will run out of
2822 spill registers and print a fatal error message. For most machines, you
2823 should not define this macro at all.
2826 @defmac CLASS_LIKELY_SPILLED_P (@var{class})
2827 A C expression whose value is nonzero if pseudos that have been assigned
2828 to registers of class @var{class} would likely be spilled because
2829 registers of @var{class} are needed for spill registers.
2831 The default value of this macro returns 1 if @var{class} has exactly one
2832 register and zero otherwise. On most machines, this default should be
2833 used. Only define this macro to some other expression if pseudos
2834 allocated by @file{local-alloc.c} end up in memory because their hard
2835 registers were needed for spill registers. If this macro returns nonzero
2836 for those classes, those pseudos will only be allocated by
2837 @file{global.c}, which knows how to reallocate the pseudo to another
2838 register. If there would not be another register available for
2839 reallocation, you should not change the definition of this macro since
2840 the only effect of such a definition would be to slow down register
2844 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2845 A C expression for the maximum number of consecutive registers
2846 of class @var{class} needed to hold a value of mode @var{mode}.
2848 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2849 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2850 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2851 @var{mode})} for all @var{regno} values in the class @var{class}.
2853 This macro helps control the handling of multiple-word values
2857 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2858 If defined, a C expression that returns nonzero for a @var{class} for which
2859 a change from mode @var{from} to mode @var{to} is invalid.
2861 For the example, loading 32-bit integer or floating-point objects into
2862 floating-point registers on the Alpha extends them to 64 bits.
2863 Therefore loading a 64-bit object and then storing it as a 32-bit object
2864 does not store the low-order 32 bits, as would be the case for a normal
2865 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2869 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2870 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2871 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2875 @node Old Constraints
2876 @section Obsolete Macros for Defining Constraints
2877 @cindex defining constraints, obsolete method
2878 @cindex constraints, defining, obsolete method
2880 Machine-specific constraints can be defined with these macros instead
2881 of the machine description constructs described in @ref{Define
2882 Constraints}. This mechanism is obsolete. New ports should not use
2883 it; old ports should convert to the new mechanism.
2885 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2886 For the constraint at the start of @var{str}, which starts with the letter
2887 @var{c}, return the length. This allows you to have register class /
2888 constant / extra constraints that are longer than a single letter;
2889 you don't need to define this macro if you can do with single-letter
2890 constraints only. The definition of this macro should use
2891 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2892 to handle specially.
2893 There are some sanity checks in genoutput.c that check the constraint lengths
2894 for the md file, so you can also use this macro to help you while you are
2895 transitioning from a byzantine single-letter-constraint scheme: when you
2896 return a negative length for a constraint you want to re-use, genoutput
2897 will complain about every instance where it is used in the md file.
2900 @defmac REG_CLASS_FROM_LETTER (@var{char})
2901 A C expression which defines the machine-dependent operand constraint
2902 letters for register classes. If @var{char} is such a letter, the
2903 value should be the register class corresponding to it. Otherwise,
2904 the value should be @code{NO_REGS}. The register letter @samp{r},
2905 corresponding to class @code{GENERAL_REGS}, will not be passed
2906 to this macro; you do not need to handle it.
2909 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2910 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2911 passed in @var{str}, so that you can use suffixes to distinguish between
2915 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2916 A C expression that defines the machine-dependent operand constraint
2917 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2918 particular ranges of integer values. If @var{c} is one of those
2919 letters, the expression should check that @var{value}, an integer, is in
2920 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2921 not one of those letters, the value should be 0 regardless of
2925 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2926 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2927 string passed in @var{str}, so that you can use suffixes to distinguish
2928 between different variants.
2931 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2932 A C expression that defines the machine-dependent operand constraint
2933 letters that specify particular ranges of @code{const_double} values
2934 (@samp{G} or @samp{H}).
2936 If @var{c} is one of those letters, the expression should check that
2937 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2938 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2939 letters, the value should be 0 regardless of @var{value}.
2941 @code{const_double} is used for all floating-point constants and for
2942 @code{DImode} fixed-point constants. A given letter can accept either
2943 or both kinds of values. It can use @code{GET_MODE} to distinguish
2944 between these kinds.
2947 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2948 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2949 string passed in @var{str}, so that you can use suffixes to distinguish
2950 between different variants.
2953 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2954 A C expression that defines the optional machine-dependent constraint
2955 letters that can be used to segregate specific types of operands, usually
2956 memory references, for the target machine. Any letter that is not
2957 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2958 @code{REG_CLASS_FROM_CONSTRAINT}
2959 may be used. Normally this macro will not be defined.
2961 If it is required for a particular target machine, it should return 1
2962 if @var{value} corresponds to the operand type represented by the
2963 constraint letter @var{c}. If @var{c} is not defined as an extra
2964 constraint, the value returned should be 0 regardless of @var{value}.
2966 For example, on the ROMP, load instructions cannot have their output
2967 in r0 if the memory reference contains a symbolic address. Constraint
2968 letter @samp{Q} is defined as representing a memory address that does
2969 @emph{not} contain a symbolic address. An alternative is specified with
2970 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2971 alternative specifies @samp{m} on the input and a register class that
2972 does not include r0 on the output.
2975 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2976 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2977 in @var{str}, so that you can use suffixes to distinguish between different
2981 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2982 A C expression that defines the optional machine-dependent constraint
2983 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2984 be treated like memory constraints by the reload pass.
2986 It should return 1 if the operand type represented by the constraint
2987 at the start of @var{str}, the first letter of which is the letter @var{c},
2988 comprises a subset of all memory references including
2989 all those whose address is simply a base register. This allows the reload
2990 pass to reload an operand, if it does not directly correspond to the operand
2991 type of @var{c}, by copying its address into a base register.
2993 For example, on the S/390, some instructions do not accept arbitrary
2994 memory references, but only those that do not make use of an index
2995 register. The constraint letter @samp{Q} is defined via
2996 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2997 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2998 a @samp{Q} constraint can handle any memory operand, because the
2999 reload pass knows it can be reloaded by copying the memory address
3000 into a base register if required. This is analogous to the way
3001 a @samp{o} constraint can handle any memory operand.
3004 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
3005 A C expression that defines the optional machine-dependent constraint
3006 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
3007 @code{EXTRA_CONSTRAINT_STR}, that should
3008 be treated like address constraints by the reload pass.
3010 It should return 1 if the operand type represented by the constraint
3011 at the start of @var{str}, which starts with the letter @var{c}, comprises
3012 a subset of all memory addresses including
3013 all those that consist of just a base register. This allows the reload
3014 pass to reload an operand, if it does not directly correspond to the operand
3015 type of @var{str}, by copying it into a base register.
3017 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
3018 be used with the @code{address_operand} predicate. It is treated
3019 analogously to the @samp{p} constraint.
3022 @node Stack and Calling
3023 @section Stack Layout and Calling Conventions
3024 @cindex calling conventions
3026 @c prevent bad page break with this line
3027 This describes the stack layout and calling conventions.
3031 * Exception Handling::
3036 * Register Arguments::
3038 * Aggregate Return::
3043 * Stack Smashing Protection::
3047 @subsection Basic Stack Layout
3048 @cindex stack frame layout
3049 @cindex frame layout
3051 @c prevent bad page break with this line
3052 Here is the basic stack layout.
3054 @defmac STACK_GROWS_DOWNWARD
3055 Define this macro if pushing a word onto the stack moves the stack
3056 pointer to a smaller address.
3058 When we say, ``define this macro if @dots{}'', it means that the
3059 compiler checks this macro only with @code{#ifdef} so the precise
3060 definition used does not matter.
3063 @defmac STACK_PUSH_CODE
3064 This macro defines the operation used when something is pushed
3065 on the stack. In RTL, a push operation will be
3066 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
3068 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
3069 and @code{POST_INC}. Which of these is correct depends on
3070 the stack direction and on whether the stack pointer points
3071 to the last item on the stack or whether it points to the
3072 space for the next item on the stack.
3074 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
3075 defined, which is almost always right, and @code{PRE_INC} otherwise,
3076 which is often wrong.
3079 @defmac FRAME_GROWS_DOWNWARD
3080 Define this macro to nonzero value if the addresses of local variable slots
3081 are at negative offsets from the frame pointer.
3084 @defmac ARGS_GROW_DOWNWARD
3085 Define this macro if successive arguments to a function occupy decreasing
3086 addresses on the stack.
3089 @defmac STARTING_FRAME_OFFSET
3090 Offset from the frame pointer to the first local variable slot to be allocated.
3092 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
3093 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
3094 Otherwise, it is found by adding the length of the first slot to the
3095 value @code{STARTING_FRAME_OFFSET}.
3096 @c i'm not sure if the above is still correct.. had to change it to get
3097 @c rid of an overfull. --mew 2feb93
3100 @defmac STACK_ALIGNMENT_NEEDED
3101 Define to zero to disable final alignment of the stack during reload.
3102 The nonzero default for this macro is suitable for most ports.
3104 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
3105 is a register save block following the local block that doesn't require
3106 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
3107 stack alignment and do it in the backend.
3110 @defmac STACK_POINTER_OFFSET
3111 Offset from the stack pointer register to the first location at which
3112 outgoing arguments are placed. If not specified, the default value of
3113 zero is used. This is the proper value for most machines.
3115 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3116 the first location at which outgoing arguments are placed.
3119 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3120 Offset from the argument pointer register to the first argument's
3121 address. On some machines it may depend on the data type of the
3124 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3125 the first argument's address.
3128 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3129 Offset from the stack pointer register to an item dynamically allocated
3130 on the stack, e.g., by @code{alloca}.
3132 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3133 length of the outgoing arguments. The default is correct for most
3134 machines. See @file{function.c} for details.
3137 @defmac INITIAL_FRAME_ADDRESS_RTX
3138 A C expression whose value is RTL representing the address of the initial
3139 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3140 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3141 default value will be used. Define this macro in order to make frame pointer
3142 elimination work in the presence of @code{__builtin_frame_address (count)} and
3143 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3146 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3147 A C expression whose value is RTL representing the address in a stack
3148 frame where the pointer to the caller's frame is stored. Assume that
3149 @var{frameaddr} is an RTL expression for the address of the stack frame
3152 If you don't define this macro, the default is to return the value
3153 of @var{frameaddr}---that is, the stack frame address is also the
3154 address of the stack word that points to the previous frame.
3157 @defmac SETUP_FRAME_ADDRESSES
3158 If defined, a C expression that produces the machine-specific code to
3159 setup the stack so that arbitrary frames can be accessed. For example,
3160 on the SPARC, we must flush all of the register windows to the stack
3161 before we can access arbitrary stack frames. You will seldom need to
3165 @deftypefn {Target Hook} bool TARGET_BUILTIN_SETJMP_FRAME_VALUE ()
3166 This target hook should return an rtx that is used to store
3167 the address of the current frame into the built in @code{setjmp} buffer.
3168 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3169 machines. One reason you may need to define this target hook is if
3170 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3173 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3174 A C expression whose value is RTL representing the value of the frame
3175 address for the current frame. @var{frameaddr} is the frame pointer
3176 of the current frame. This is used for __builtin_frame_address.
3177 You need only define this macro if the frame address is not the same
3178 as the frame pointer. Most machines do not need to define it.
3181 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3182 A C expression whose value is RTL representing the value of the return
3183 address for the frame @var{count} steps up from the current frame, after
3184 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3185 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3186 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
3188 The value of the expression must always be the correct address when
3189 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
3190 determine the return address of other frames.
3193 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3194 Define this if the return address of a particular stack frame is accessed
3195 from the frame pointer of the previous stack frame.
3198 @defmac INCOMING_RETURN_ADDR_RTX
3199 A C expression whose value is RTL representing the location of the
3200 incoming return address at the beginning of any function, before the
3201 prologue. This RTL is either a @code{REG}, indicating that the return
3202 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3205 You only need to define this macro if you want to support call frame
3206 debugging information like that provided by DWARF 2.
3208 If this RTL is a @code{REG}, you should also define
3209 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3212 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3213 A C expression whose value is an integer giving a DWARF 2 column
3214 number that may be used as an alternative return column. The column
3215 must not correspond to any gcc hard register (that is, it must not
3216 be in the range of @code{DWARF_FRAME_REGNUM}).
3218 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3219 general register, but an alternative column needs to be used for signal
3220 frames. Some targets have also used different frame return columns
3224 @defmac DWARF_ZERO_REG
3225 A C expression whose value is an integer giving a DWARF 2 register
3226 number that is considered to always have the value zero. This should
3227 only be defined if the target has an architected zero register, and
3228 someone decided it was a good idea to use that register number to
3229 terminate the stack backtrace. New ports should avoid this.
3232 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3233 This target hook allows the backend to emit frame-related insns that
3234 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3235 info engine will invoke it on insns of the form
3237 (set (reg) (unspec [@dots{}] UNSPEC_INDEX))
3241 (set (reg) (unspec_volatile [@dots{}] UNSPECV_INDEX)).
3243 to let the backend emit the call frame instructions. @var{label} is
3244 the CFI label attached to the insn, @var{pattern} is the pattern of
3245 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3248 @defmac INCOMING_FRAME_SP_OFFSET
3249 A C expression whose value is an integer giving the offset, in bytes,
3250 from the value of the stack pointer register to the top of the stack
3251 frame at the beginning of any function, before the prologue. The top of
3252 the frame is defined to be the value of the stack pointer in the
3253 previous frame, just before the call instruction.
3255 You only need to define this macro if you want to support call frame
3256 debugging information like that provided by DWARF 2.
3259 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3260 A C expression whose value is an integer giving the offset, in bytes,
3261 from the argument pointer to the canonical frame address (cfa). The
3262 final value should coincide with that calculated by
3263 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3264 during virtual register instantiation.
3266 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
3267 which is correct for most machines; in general, the arguments are found
3268 immediately before the stack frame. Note that this is not the case on
3269 some targets that save registers into the caller's frame, such as SPARC
3270 and rs6000, and so such targets need to define this macro.
3272 You only need to define this macro if the default is incorrect, and you
3273 want to support call frame debugging information like that provided by
3277 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3278 If defined, a C expression whose value is an integer giving the offset
3279 in bytes from the frame pointer to the canonical frame address (cfa).
3280 The final value should coincide with that calculated by
3281 @code{INCOMING_FRAME_SP_OFFSET}.
3283 Normally the CFA is calculated as an offset from the argument pointer,
3284 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3285 variable due to the ABI, this may not be possible. If this macro is
3286 defined, it implies that the virtual register instantiation should be
3287 based on the frame pointer instead of the argument pointer. Only one
3288 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3292 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3293 If defined, a C expression whose value is an integer giving the offset
3294 in bytes from the canonical frame address (cfa) to the frame base used
3295 in DWARF 2 debug information. The default is zero. A different value
3296 may reduce the size of debug information on some ports.
3299 @node Exception Handling
3300 @subsection Exception Handling Support
3301 @cindex exception handling
3303 @defmac EH_RETURN_DATA_REGNO (@var{N})
3304 A C expression whose value is the @var{N}th register number used for
3305 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3306 @var{N} registers are usable.
3308 The exception handling library routines communicate with the exception
3309 handlers via a set of agreed upon registers. Ideally these registers
3310 should be call-clobbered; it is possible to use call-saved registers,
3311 but may negatively impact code size. The target must support at least
3312 2 data registers, but should define 4 if there are enough free registers.
3314 You must define this macro if you want to support call frame exception
3315 handling like that provided by DWARF 2.
3318 @defmac EH_RETURN_STACKADJ_RTX
3319 A C expression whose value is RTL representing a location in which
3320 to store a stack adjustment to be applied before function return.
3321 This is used to unwind the stack to an exception handler's call frame.
3322 It will be assigned zero on code paths that return normally.
3324 Typically this is a call-clobbered hard register that is otherwise
3325 untouched by the epilogue, but could also be a stack slot.
3327 Do not define this macro if the stack pointer is saved and restored
3328 by the regular prolog and epilog code in the call frame itself; in
3329 this case, the exception handling library routines will update the
3330 stack location to be restored in place. Otherwise, you must define
3331 this macro if you want to support call frame exception handling like
3332 that provided by DWARF 2.
3335 @defmac EH_RETURN_HANDLER_RTX
3336 A C expression whose value is RTL representing a location in which
3337 to store the address of an exception handler to which we should
3338 return. It will not be assigned on code paths that return normally.
3340 Typically this is the location in the call frame at which the normal
3341 return address is stored. For targets that return by popping an
3342 address off the stack, this might be a memory address just below
3343 the @emph{target} call frame rather than inside the current call
3344 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3345 been assigned, so it may be used to calculate the location of the
3348 Some targets have more complex requirements than storing to an
3349 address calculable during initial code generation. In that case
3350 the @code{eh_return} instruction pattern should be used instead.
3352 If you want to support call frame exception handling, you must
3353 define either this macro or the @code{eh_return} instruction pattern.
3356 @defmac RETURN_ADDR_OFFSET
3357 If defined, an integer-valued C expression for which rtl will be generated
3358 to add it to the exception handler address before it is searched in the
3359 exception handling tables, and to subtract it again from the address before
3360 using it to return to the exception handler.
3363 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3364 This macro chooses the encoding of pointers embedded in the exception
3365 handling sections. If at all possible, this should be defined such
3366 that the exception handling section will not require dynamic relocations,
3367 and so may be read-only.
3369 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3370 @var{global} is true if the symbol may be affected by dynamic relocations.
3371 The macro should return a combination of the @code{DW_EH_PE_*} defines
3372 as found in @file{dwarf2.h}.
3374 If this macro is not defined, pointers will not be encoded but
3375 represented directly.
3378 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3379 This macro allows the target to emit whatever special magic is required
3380 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3381 Generic code takes care of pc-relative and indirect encodings; this must
3382 be defined if the target uses text-relative or data-relative encodings.
3384 This is a C statement that branches to @var{done} if the format was
3385 handled. @var{encoding} is the format chosen, @var{size} is the number
3386 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3390 @defmac MD_UNWIND_SUPPORT
3391 A string specifying a file to be #include'd in unwind-dw2.c. The file
3392 so included typically defines @code{MD_FALLBACK_FRAME_STATE_FOR}.
3395 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3396 This macro allows the target to add CPU and operating system specific
3397 code to the call-frame unwinder for use when there is no unwind data
3398 available. The most common reason to implement this macro is to unwind
3399 through signal frames.
3401 This macro is called from @code{uw_frame_state_for} in
3402 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
3403 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3404 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3405 for the address of the code being executed and @code{context->cfa} for
3406 the stack pointer value. If the frame can be decoded, the register
3407 save addresses should be updated in @var{fs} and the macro should
3408 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
3409 the macro should evaluate to @code{_URC_END_OF_STACK}.
3411 For proper signal handling in Java this macro is accompanied by
3412 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3415 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3416 This macro allows the target to add operating system specific code to the
3417 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3418 usually used for signal or interrupt frames.
3420 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3421 @var{context} is an @code{_Unwind_Context};
3422 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3423 for the abi and context in the @code{.unwabi} directive. If the
3424 @code{.unwabi} directive can be handled, the register save addresses should
3425 be updated in @var{fs}.
3428 @defmac TARGET_USES_WEAK_UNWIND_INFO
3429 A C expression that evaluates to true if the target requires unwind
3430 info to be given comdat linkage. Define it to be @code{1} if comdat
3431 linkage is necessary. The default is @code{0}.
3434 @node Stack Checking
3435 @subsection Specifying How Stack Checking is Done
3437 GCC will check that stack references are within the boundaries of
3438 the stack, if the @option{-fstack-check} is specified, in one of three ways:
3442 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3443 will assume that you have arranged for stack checking to be done at
3444 appropriate places in the configuration files, e.g., in
3445 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
3449 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
3450 called @code{check_stack} in your @file{md} file, GCC will call that
3451 pattern with one argument which is the address to compare the stack
3452 value against. You must arrange for this pattern to report an error if
3453 the stack pointer is out of range.
3456 If neither of the above are true, GCC will generate code to periodically
3457 ``probe'' the stack pointer using the values of the macros defined below.
3460 Normally, you will use the default values of these macros, so GCC
3461 will use the third approach.
3463 @defmac STACK_CHECK_BUILTIN
3464 A nonzero value if stack checking is done by the configuration files in a
3465 machine-dependent manner. You should define this macro if stack checking
3466 is require by the ABI of your machine or if you would like to have to stack
3467 checking in some more efficient way than GCC's portable approach.
3468 The default value of this macro is zero.
3471 @defmac STACK_CHECK_PROBE_INTERVAL
3472 An integer representing the interval at which GCC must generate stack
3473 probe instructions. You will normally define this macro to be no larger
3474 than the size of the ``guard pages'' at the end of a stack area. The
3475 default value of 4096 is suitable for most systems.
3478 @defmac STACK_CHECK_PROBE_LOAD
3479 A integer which is nonzero if GCC should perform the stack probe
3480 as a load instruction and zero if GCC should use a store instruction.
3481 The default is zero, which is the most efficient choice on most systems.
3484 @defmac STACK_CHECK_PROTECT
3485 The number of bytes of stack needed to recover from a stack overflow,
3486 for languages where such a recovery is supported. The default value of
3487 75 words should be adequate for most machines.
3490 @defmac STACK_CHECK_MAX_FRAME_SIZE
3491 The maximum size of a stack frame, in bytes. GCC will generate probe
3492 instructions in non-leaf functions to ensure at least this many bytes of
3493 stack are available. If a stack frame is larger than this size, stack
3494 checking will not be reliable and GCC will issue a warning. The
3495 default is chosen so that GCC only generates one instruction on most
3496 systems. You should normally not change the default value of this macro.
3499 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3500 GCC uses this value to generate the above warning message. It
3501 represents the amount of fixed frame used by a function, not including
3502 space for any callee-saved registers, temporaries and user variables.
3503 You need only specify an upper bound for this amount and will normally
3504 use the default of four words.
3507 @defmac STACK_CHECK_MAX_VAR_SIZE
3508 The maximum size, in bytes, of an object that GCC will place in the
3509 fixed area of the stack frame when the user specifies
3510 @option{-fstack-check}.
3511 GCC computed the default from the values of the above macros and you will
3512 normally not need to override that default.
3516 @node Frame Registers
3517 @subsection Registers That Address the Stack Frame
3519 @c prevent bad page break with this line
3520 This discusses registers that address the stack frame.
3522 @defmac STACK_POINTER_REGNUM
3523 The register number of the stack pointer register, which must also be a
3524 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3525 the hardware determines which register this is.
3528 @defmac FRAME_POINTER_REGNUM
3529 The register number of the frame pointer register, which is used to
3530 access automatic variables in the stack frame. On some machines, the
3531 hardware determines which register this is. On other machines, you can
3532 choose any register you wish for this purpose.
3535 @defmac HARD_FRAME_POINTER_REGNUM
3536 On some machines the offset between the frame pointer and starting
3537 offset of the automatic variables is not known until after register
3538 allocation has been done (for example, because the saved registers are
3539 between these two locations). On those machines, define
3540 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3541 be used internally until the offset is known, and define
3542 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3543 used for the frame pointer.
3545 You should define this macro only in the very rare circumstances when it
3546 is not possible to calculate the offset between the frame pointer and
3547 the automatic variables until after register allocation has been
3548 completed. When this macro is defined, you must also indicate in your
3549 definition of @code{ELIMINABLE_REGS} how to eliminate
3550 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3551 or @code{STACK_POINTER_REGNUM}.
3553 Do not define this macro if it would be the same as
3554 @code{FRAME_POINTER_REGNUM}.
3557 @defmac ARG_POINTER_REGNUM
3558 The register number of the arg pointer register, which is used to access
3559 the function's argument list. On some machines, this is the same as the
3560 frame pointer register. On some machines, the hardware determines which
3561 register this is. On other machines, you can choose any register you
3562 wish for this purpose. If this is not the same register as the frame
3563 pointer register, then you must mark it as a fixed register according to
3564 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3565 (@pxref{Elimination}).
3568 @defmac RETURN_ADDRESS_POINTER_REGNUM
3569 The register number of the return address pointer register, which is used to
3570 access the current function's return address from the stack. On some
3571 machines, the return address is not at a fixed offset from the frame
3572 pointer or stack pointer or argument pointer. This register can be defined
3573 to point to the return address on the stack, and then be converted by
3574 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3576 Do not define this macro unless there is no other way to get the return
3577 address from the stack.
3580 @defmac STATIC_CHAIN_REGNUM
3581 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3582 Register numbers used for passing a function's static chain pointer. If
3583 register windows are used, the register number as seen by the called
3584 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3585 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3586 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3589 The static chain register need not be a fixed register.
3591 If the static chain is passed in memory, these macros should not be
3592 defined; instead, the next two macros should be defined.
3595 @defmac STATIC_CHAIN
3596 @defmacx STATIC_CHAIN_INCOMING
3597 If the static chain is passed in memory, these macros provide rtx giving
3598 @code{mem} expressions that denote where they are stored.
3599 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3600 as seen by the calling and called functions, respectively. Often the former
3601 will be at an offset from the stack pointer and the latter at an offset from
3604 @findex stack_pointer_rtx
3605 @findex frame_pointer_rtx
3606 @findex arg_pointer_rtx
3607 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3608 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3609 macros and should be used to refer to those items.
3611 If the static chain is passed in a register, the two previous macros should
3615 @defmac DWARF_FRAME_REGISTERS
3616 This macro specifies the maximum number of hard registers that can be
3617 saved in a call frame. This is used to size data structures used in
3618 DWARF2 exception handling.
3620 Prior to GCC 3.0, this macro was needed in order to establish a stable
3621 exception handling ABI in the face of adding new hard registers for ISA
3622 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3623 in the number of hard registers. Nevertheless, this macro can still be
3624 used to reduce the runtime memory requirements of the exception handling
3625 routines, which can be substantial if the ISA contains a lot of
3626 registers that are not call-saved.
3628 If this macro is not defined, it defaults to
3629 @code{FIRST_PSEUDO_REGISTER}.
3632 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3634 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3635 for backward compatibility in pre GCC 3.0 compiled code.
3637 If this macro is not defined, it defaults to
3638 @code{DWARF_FRAME_REGISTERS}.
3641 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3643 Define this macro if the target's representation for dwarf registers
3644 is different than the internal representation for unwind column.
3645 Given a dwarf register, this macro should return the internal unwind
3646 column number to use instead.
3648 See the PowerPC's SPE target for an example.
3651 @defmac DWARF_FRAME_REGNUM (@var{regno})
3653 Define this macro if the target's representation for dwarf registers
3654 used in .eh_frame or .debug_frame is different from that used in other
3655 debug info sections. Given a GCC hard register number, this macro
3656 should return the .eh_frame register number. The default is
3657 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3661 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3663 Define this macro to map register numbers held in the call frame info
3664 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3665 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3666 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3667 return @code{@var{regno}}.
3672 @subsection Eliminating Frame Pointer and Arg Pointer
3674 @c prevent bad page break with this line
3675 This is about eliminating the frame pointer and arg pointer.
3677 @defmac FRAME_POINTER_REQUIRED
3678 A C expression which is nonzero if a function must have and use a frame
3679 pointer. This expression is evaluated in the reload pass. If its value is
3680 nonzero the function will have a frame pointer.
3682 The expression can in principle examine the current function and decide
3683 according to the facts, but on most machines the constant 0 or the
3684 constant 1 suffices. Use 0 when the machine allows code to be generated
3685 with no frame pointer, and doing so saves some time or space. Use 1
3686 when there is no possible advantage to avoiding a frame pointer.
3688 In certain cases, the compiler does not know how to produce valid code
3689 without a frame pointer. The compiler recognizes those cases and
3690 automatically gives the function a frame pointer regardless of what
3691 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3694 In a function that does not require a frame pointer, the frame pointer
3695 register can be allocated for ordinary usage, unless you mark it as a
3696 fixed register. See @code{FIXED_REGISTERS} for more information.
3699 @findex get_frame_size
3700 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3701 A C statement to store in the variable @var{depth-var} the difference
3702 between the frame pointer and the stack pointer values immediately after
3703 the function prologue. The value would be computed from information
3704 such as the result of @code{get_frame_size ()} and the tables of
3705 registers @code{regs_ever_live} and @code{call_used_regs}.
3707 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3708 need not be defined. Otherwise, it must be defined even if
3709 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3710 case, you may set @var{depth-var} to anything.
3713 @defmac ELIMINABLE_REGS
3714 If defined, this macro specifies a table of register pairs used to
3715 eliminate unneeded registers that point into the stack frame. If it is not
3716 defined, the only elimination attempted by the compiler is to replace
3717 references to the frame pointer with references to the stack pointer.
3719 The definition of this macro is a list of structure initializations, each
3720 of which specifies an original and replacement register.
3722 On some machines, the position of the argument pointer is not known until
3723 the compilation is completed. In such a case, a separate hard register
3724 must be used for the argument pointer. This register can be eliminated by
3725 replacing it with either the frame pointer or the argument pointer,
3726 depending on whether or not the frame pointer has been eliminated.
3728 In this case, you might specify:
3730 #define ELIMINABLE_REGS \
3731 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3732 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3733 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3736 Note that the elimination of the argument pointer with the stack pointer is
3737 specified first since that is the preferred elimination.
3740 @defmac CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3741 A C expression that returns nonzero if the compiler is allowed to try
3742 to replace register number @var{from-reg} with register number
3743 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3744 is defined, and will usually be the constant 1, since most of the cases
3745 preventing register elimination are things that the compiler already
3749 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3750 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3751 specifies the initial difference between the specified pair of
3752 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3756 @node Stack Arguments
3757 @subsection Passing Function Arguments on the Stack
3758 @cindex arguments on stack
3759 @cindex stack arguments
3761 The macros in this section control how arguments are passed
3762 on the stack. See the following section for other macros that
3763 control passing certain arguments in registers.
3765 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (tree @var{fntype})
3766 This target hook returns @code{true} if an argument declared in a
3767 prototype as an integral type smaller than @code{int} should actually be
3768 passed as an @code{int}. In addition to avoiding errors in certain
3769 cases of mismatch, it also makes for better code on certain machines.
3770 The default is to not promote prototypes.
3774 A C expression. If nonzero, push insns will be used to pass
3776 If the target machine does not have a push instruction, set it to zero.
3777 That directs GCC to use an alternate strategy: to
3778 allocate the entire argument block and then store the arguments into
3779 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3782 @defmac PUSH_ARGS_REVERSED
3783 A C expression. If nonzero, function arguments will be evaluated from
3784 last to first, rather than from first to last. If this macro is not
3785 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3786 and args grow in opposite directions, and 0 otherwise.
3789 @defmac PUSH_ROUNDING (@var{npushed})
3790 A C expression that is the number of bytes actually pushed onto the
3791 stack when an instruction attempts to push @var{npushed} bytes.
3793 On some machines, the definition
3796 #define PUSH_ROUNDING(BYTES) (BYTES)
3800 will suffice. But on other machines, instructions that appear
3801 to push one byte actually push two bytes in an attempt to maintain
3802 alignment. Then the definition should be
3805 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3809 @findex current_function_outgoing_args_size
3810 @defmac ACCUMULATE_OUTGOING_ARGS
3811 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3812 will be computed and placed into the variable
3813 @code{current_function_outgoing_args_size}. No space will be pushed
3814 onto the stack for each call; instead, the function prologue should
3815 increase the stack frame size by this amount.
3817 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3821 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3822 Define this macro if functions should assume that stack space has been
3823 allocated for arguments even when their values are passed in
3826 The value of this macro is the size, in bytes, of the area reserved for
3827 arguments passed in registers for the function represented by @var{fndecl},
3828 which can be zero if GCC is calling a library function.
3830 This space can be allocated by the caller, or be a part of the
3831 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3834 @c above is overfull. not sure what to do. --mew 5feb93 did
3835 @c something, not sure if it looks good. --mew 10feb93
3837 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3838 Define this to a nonzero value if it is the responsibility of the
3839 caller to allocate the area reserved for arguments passed in registers
3840 when calling a function of @var{fntype}. @var{fntype} may be NULL
3841 if the function called is a library function.
3843 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3844 whether the space for these arguments counts in the value of
3845 @code{current_function_outgoing_args_size}.
3848 @defmac STACK_PARMS_IN_REG_PARM_AREA
3849 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3850 stack parameters don't skip the area specified by it.
3851 @c i changed this, makes more sens and it should have taken care of the
3852 @c overfull.. not as specific, tho. --mew 5feb93
3854 Normally, when a parameter is not passed in registers, it is placed on the
3855 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3856 suppresses this behavior and causes the parameter to be passed on the
3857 stack in its natural location.
3860 @defmac RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3861 A C expression that should indicate the number of bytes of its own
3862 arguments that a function pops on returning, or 0 if the
3863 function pops no arguments and the caller must therefore pop them all
3864 after the function returns.
3866 @var{fundecl} is a C variable whose value is a tree node that describes
3867 the function in question. Normally it is a node of type
3868 @code{FUNCTION_DECL} that describes the declaration of the function.
3869 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3871 @var{funtype} is a C variable whose value is a tree node that
3872 describes the function in question. Normally it is a node of type
3873 @code{FUNCTION_TYPE} that describes the data type of the function.
3874 From this it is possible to obtain the data types of the value and
3875 arguments (if known).
3877 When a call to a library function is being considered, @var{fundecl}
3878 will contain an identifier node for the library function. Thus, if
3879 you need to distinguish among various library functions, you can do so
3880 by their names. Note that ``library function'' in this context means
3881 a function used to perform arithmetic, whose name is known specially
3882 in the compiler and was not mentioned in the C code being compiled.
3884 @var{stack-size} is the number of bytes of arguments passed on the
3885 stack. If a variable number of bytes is passed, it is zero, and
3886 argument popping will always be the responsibility of the calling function.
3888 On the VAX, all functions always pop their arguments, so the definition
3889 of this macro is @var{stack-size}. On the 68000, using the standard
3890 calling convention, no functions pop their arguments, so the value of
3891 the macro is always 0 in this case. But an alternative calling
3892 convention is available in which functions that take a fixed number of
3893 arguments pop them but other functions (such as @code{printf}) pop
3894 nothing (the caller pops all). When this convention is in use,
3895 @var{funtype} is examined to determine whether a function takes a fixed
3896 number of arguments.
3899 @defmac CALL_POPS_ARGS (@var{cum})
3900 A C expression that should indicate the number of bytes a call sequence
3901 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3902 when compiling a function call.
3904 @var{cum} is the variable in which all arguments to the called function
3905 have been accumulated.
3907 On certain architectures, such as the SH5, a call trampoline is used
3908 that pops certain registers off the stack, depending on the arguments
3909 that have been passed to the function. Since this is a property of the
3910 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3914 @node Register Arguments
3915 @subsection Passing Arguments in Registers
3916 @cindex arguments in registers
3917 @cindex registers arguments
3919 This section describes the macros which let you control how various
3920 types of arguments are passed in registers or how they are arranged in
3923 @defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3924 A C expression that controls whether a function argument is passed
3925 in a register, and which register.
3927 The arguments are @var{cum}, which summarizes all the previous
3928 arguments; @var{mode}, the machine mode of the argument; @var{type},
3929 the data type of the argument as a tree node or 0 if that is not known
3930 (which happens for C support library functions); and @var{named},
3931 which is 1 for an ordinary argument and 0 for nameless arguments that
3932 correspond to @samp{@dots{}} in the called function's prototype.
3933 @var{type} can be an incomplete type if a syntax error has previously
3936 The value of the expression is usually either a @code{reg} RTX for the
3937 hard register in which to pass the argument, or zero to pass the
3938 argument on the stack.
3940 For machines like the VAX and 68000, where normally all arguments are
3941 pushed, zero suffices as a definition.
3943 The value of the expression can also be a @code{parallel} RTX@. This is
3944 used when an argument is passed in multiple locations. The mode of the
3945 @code{parallel} should be the mode of the entire argument. The
3946 @code{parallel} holds any number of @code{expr_list} pairs; each one
3947 describes where part of the argument is passed. In each
3948 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3949 register in which to pass this part of the argument, and the mode of the
3950 register RTX indicates how large this part of the argument is. The
3951 second operand of the @code{expr_list} is a @code{const_int} which gives
3952 the offset in bytes into the entire argument of where this part starts.
3953 As a special exception the first @code{expr_list} in the @code{parallel}
3954 RTX may have a first operand of zero. This indicates that the entire
3955 argument is also stored on the stack.
3957 The last time this macro is called, it is called with @code{MODE ==
3958 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3959 pattern as operands 2 and 3 respectively.
3961 @cindex @file{stdarg.h} and register arguments
3962 The usual way to make the ISO library @file{stdarg.h} work on a machine
3963 where some arguments are usually passed in registers, is to cause
3964 nameless arguments to be passed on the stack instead. This is done
3965 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3967 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3968 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3969 You may use the hook @code{targetm.calls.must_pass_in_stack}
3970 in the definition of this macro to determine if this argument is of a
3971 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3972 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3973 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3974 defined, the argument will be computed in the stack and then loaded into
3978 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, tree @var{type})
3979 This target hook should return @code{true} if we should not pass @var{type}
3980 solely in registers. The file @file{expr.h} defines a
3981 definition that is usually appropriate, refer to @file{expr.h} for additional
3985 @defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3986 Define this macro if the target machine has ``register windows'', so
3987 that the register in which a function sees an arguments is not
3988 necessarily the same as the one in which the caller passed the
3991 For such machines, @code{FUNCTION_ARG} computes the register in which
3992 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3993 be defined in a similar fashion to tell the function being called
3994 where the arguments will arrive.
3996 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3997 serves both purposes.
4000 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
4001 This target hook returns the number of bytes at the beginning of an
4002 argument that must be put in registers. The value must be zero for
4003 arguments that are passed entirely in registers or that are entirely
4004 pushed on the stack.
4006 On some machines, certain arguments must be passed partially in
4007 registers and partially in memory. On these machines, typically the
4008 first few words of arguments are passed in registers, and the rest
4009 on the stack. If a multi-word argument (a @code{double} or a
4010 structure) crosses that boundary, its first few words must be passed
4011 in registers and the rest must be pushed. This macro tells the
4012 compiler when this occurs, and how many bytes should go in registers.
4014 @code{FUNCTION_ARG} for these arguments should return the first
4015 register to be used by the caller for this argument; likewise
4016 @code{FUNCTION_INCOMING_ARG}, for the called function.
4019 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
4020 This target hook should return @code{true} if an argument at the
4021 position indicated by @var{cum} should be passed by reference. This
4022 predicate is queried after target independent reasons for being
4023 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
4025 If the hook returns true, a copy of that argument is made in memory and a
4026 pointer to the argument is passed instead of the argument itself.
4027 The pointer is passed in whatever way is appropriate for passing a pointer
4031 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
4032 The function argument described by the parameters to this hook is
4033 known to be passed by reference. The hook should return true if the
4034 function argument should be copied by the callee instead of copied
4037 For any argument for which the hook returns true, if it can be
4038 determined that the argument is not modified, then a copy need
4041 The default version of this hook always returns false.
4044 @defmac CUMULATIVE_ARGS
4045 A C type for declaring a variable that is used as the first argument of
4046 @code{FUNCTION_ARG} and other related values. For some target machines,
4047 the type @code{int} suffices and can hold the number of bytes of
4050 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
4051 arguments that have been passed on the stack. The compiler has other
4052 variables to keep track of that. For target machines on which all
4053 arguments are passed on the stack, there is no need to store anything in
4054 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
4055 should not be empty, so use @code{int}.
4058 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
4059 If defined, this macro is called before generating any code for a
4060 function, but after the @var{cfun} descriptor for the function has been
4061 created. The back end may use this macro to update @var{cfun} to
4062 reflect an ABI other than that which would normally be used by default.
4063 If the compiler is generating code for a compiler-generated function,
4064 @var{fndecl} may be @code{NULL}.
4067 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
4068 A C statement (sans semicolon) for initializing the variable
4069 @var{cum} for the state at the beginning of the argument list. The
4070 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
4071 is the tree node for the data type of the function which will receive
4072 the args, or 0 if the args are to a compiler support library function.
4073 For direct calls that are not libcalls, @var{fndecl} contain the
4074 declaration node of the function. @var{fndecl} is also set when
4075 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
4076 being compiled. @var{n_named_args} is set to the number of named
4077 arguments, including a structure return address if it is passed as a
4078 parameter, when making a call. When processing incoming arguments,
4079 @var{n_named_args} is set to @minus{}1.
4081 When processing a call to a compiler support library function,
4082 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
4083 contains the name of the function, as a string. @var{libname} is 0 when
4084 an ordinary C function call is being processed. Thus, each time this
4085 macro is called, either @var{libname} or @var{fntype} is nonzero, but
4086 never both of them at once.
4089 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
4090 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
4091 it gets a @code{MODE} argument instead of @var{fntype}, that would be
4092 @code{NULL}. @var{indirect} would always be zero, too. If this macro
4093 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
4094 0)} is used instead.
4097 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
4098 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
4099 finding the arguments for the function being compiled. If this macro is
4100 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
4102 The value passed for @var{libname} is always 0, since library routines
4103 with special calling conventions are never compiled with GCC@. The
4104 argument @var{libname} exists for symmetry with
4105 @code{INIT_CUMULATIVE_ARGS}.
4106 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
4107 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
4110 @defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
4111 A C statement (sans semicolon) to update the summarizer variable
4112 @var{cum} to advance past an argument in the argument list. The
4113 values @var{mode}, @var{type} and @var{named} describe that argument.
4114 Once this is done, the variable @var{cum} is suitable for analyzing
4115 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
4117 This macro need not do anything if the argument in question was passed
4118 on the stack. The compiler knows how to track the amount of stack space
4119 used for arguments without any special help.
4122 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
4123 If defined, a C expression which determines whether, and in which direction,
4124 to pad out an argument with extra space. The value should be of type
4125 @code{enum direction}: either @code{upward} to pad above the argument,
4126 @code{downward} to pad below, or @code{none} to inhibit padding.
4128 The @emph{amount} of padding is always just enough to reach the next
4129 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
4132 This macro has a default definition which is right for most systems.
4133 For little-endian machines, the default is to pad upward. For
4134 big-endian machines, the default is to pad downward for an argument of
4135 constant size shorter than an @code{int}, and upward otherwise.
4138 @defmac PAD_VARARGS_DOWN
4139 If defined, a C expression which determines whether the default
4140 implementation of va_arg will attempt to pad down before reading the
4141 next argument, if that argument is smaller than its aligned space as
4142 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4143 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4146 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4147 Specify padding for the last element of a block move between registers and
4148 memory. @var{first} is nonzero if this is the only element. Defining this
4149 macro allows better control of register function parameters on big-endian
4150 machines, without using @code{PARALLEL} rtl. In particular,
4151 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4152 registers, as there is no longer a "wrong" part of a register; For example,
4153 a three byte aggregate may be passed in the high part of a register if so
4157 @defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
4158 If defined, a C expression that gives the alignment boundary, in bits,
4159 of an argument with the specified mode and type. If it is not defined,
4160 @code{PARM_BOUNDARY} is used for all arguments.
4163 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4164 A C expression that is nonzero if @var{regno} is the number of a hard
4165 register in which function arguments are sometimes passed. This does
4166 @emph{not} include implicit arguments such as the static chain and
4167 the structure-value address. On many machines, no registers can be
4168 used for this purpose since all function arguments are pushed on the
4172 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (tree @var{type})
4173 This hook should return true if parameter of type @var{type} are passed
4174 as two scalar parameters. By default, GCC will attempt to pack complex
4175 arguments into the target's word size. Some ABIs require complex arguments
4176 to be split and treated as their individual components. For example, on
4177 AIX64, complex floats should be passed in a pair of floating point
4178 registers, even though a complex float would fit in one 64-bit floating
4181 The default value of this hook is @code{NULL}, which is treated as always
4185 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
4186 This hook returns a type node for @code{va_list} for the target.
4187 The default version of the hook returns @code{void*}.
4190 @deftypefn {Target Hook} tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree @var{valist}, tree @var{type}, tree *@var{pre_p}, tree *@var{post_p})
4191 This hook performs target-specific gimplification of
4192 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4193 arguments to @code{va_arg}; the latter two are as in
4194 @code{gimplify.c:gimplify_expr}.
4197 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (enum machine_mode @var{mode})
4198 Define this to return nonzero if the port can handle pointers
4199 with machine mode @var{mode}. The default version of this
4200 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4203 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4204 Define this to return nonzero if the port is prepared to handle
4205 insns involving scalar mode @var{mode}. For a scalar mode to be
4206 considered supported, all the basic arithmetic and comparisons
4209 The default version of this hook returns true for any mode
4210 required to handle the basic C types (as defined by the port).
4211 Included here are the double-word arithmetic supported by the
4212 code in @file{optabs.c}.
4215 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4216 Define this to return nonzero if the port is prepared to handle
4217 insns involving vector mode @var{mode}. At the very least, it
4218 must have move patterns for this mode.
4222 @subsection How Scalar Function Values Are Returned
4223 @cindex return values in registers
4224 @cindex values, returned by functions
4225 @cindex scalars, returned as values
4227 This section discusses the macros that control returning scalars as
4228 values---values that can fit in registers.
4230 @deftypefn {Target Hook} rtx TARGET_FUNCTION_VALUE (tree @var{ret_type}, tree @var{fn_decl_or_type}, bool @var{outgoing})
4232 Define this to return an RTX representing the place where a function
4233 returns or receives a value of data type @var{ret_type}, a tree node
4234 node representing a data type. @var{fn_decl_or_type} is a tree node
4235 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4236 function being called. If @var{outgoing} is false, the hook should
4237 compute the register in which the caller will see the return value.
4238 Otherwise, the hook should return an RTX representing the place where
4239 a function returns a value.
4241 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4242 (Actually, on most machines, scalar values are returned in the same
4243 place regardless of mode.) The value of the expression is usually a
4244 @code{reg} RTX for the hard register where the return value is stored.
4245 The value can also be a @code{parallel} RTX, if the return value is in
4246 multiple places. See @code{FUNCTION_ARG} for an explanation of the
4247 @code{parallel} form. Note that the callee will populate every
4248 location specified in the @code{parallel}, but if the first element of
4249 the @code{parallel} contains the whole return value, callers will use
4250 that element as the canonical location and ignore the others. The m68k
4251 port uses this type of @code{parallel} to return pointers in both
4252 @samp{%a0} (the canonical location) and @samp{%d0}.
4254 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4255 the same promotion rules specified in @code{PROMOTE_MODE} if
4256 @var{valtype} is a scalar type.
4258 If the precise function being called is known, @var{func} is a tree
4259 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4260 pointer. This makes it possible to use a different value-returning
4261 convention for specific functions when all their calls are
4264 Some target machines have ``register windows'' so that the register in
4265 which a function returns its value is not the same as the one in which
4266 the caller sees the value. For such machines, you should return
4267 different RTX depending on @var{outgoing}.
4269 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4270 aggregate data types, because these are returned in another way. See
4271 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4274 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4275 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4276 a new target instead.
4279 @defmac FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
4280 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4281 a new target instead.
4284 @defmac LIBCALL_VALUE (@var{mode})
4285 A C expression to create an RTX representing the place where a library
4286 function returns a value of mode @var{mode}. If the precise function
4287 being called is known, @var{func} is a tree node
4288 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4289 pointer. This makes it possible to use a different value-returning
4290 convention for specific functions when all their calls are
4293 Note that ``library function'' in this context means a compiler
4294 support routine, used to perform arithmetic, whose name is known
4295 specially by the compiler and was not mentioned in the C code being
4298 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
4299 data types, because none of the library functions returns such types.
4302 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4303 A C expression that is nonzero if @var{regno} is the number of a hard
4304 register in which the values of called function may come back.
4306 A register whose use for returning values is limited to serving as the
4307 second of a pair (for a value of type @code{double}, say) need not be
4308 recognized by this macro. So for most machines, this definition
4312 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4315 If the machine has register windows, so that the caller and the called
4316 function use different registers for the return value, this macro
4317 should recognize only the caller's register numbers.
4320 @defmac APPLY_RESULT_SIZE
4321 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4322 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4323 saving and restoring an arbitrary return value.
4326 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (tree @var{type})
4327 This hook should return true if values of type @var{type} are returned
4328 at the most significant end of a register (in other words, if they are
4329 padded at the least significant end). You can assume that @var{type}
4330 is returned in a register; the caller is required to check this.
4332 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4333 be able to hold the complete return value. For example, if a 1-, 2-
4334 or 3-byte structure is returned at the most significant end of a
4335 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4339 @node Aggregate Return
4340 @subsection How Large Values Are Returned
4341 @cindex aggregates as return values
4342 @cindex large return values
4343 @cindex returning aggregate values
4344 @cindex structure value address
4346 When a function value's mode is @code{BLKmode} (and in some other
4347 cases), the value is not returned according to
4348 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4349 caller passes the address of a block of memory in which the value
4350 should be stored. This address is called the @dfn{structure value
4353 This section describes how to control returning structure values in
4356 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (tree @var{type}, tree @var{fntype})
4357 This target hook should return a nonzero value to say to return the
4358 function value in memory, just as large structures are always returned.
4359 Here @var{type} will be the data type of the value, and @var{fntype}
4360 will be the type of the function doing the returning, or @code{NULL} for
4363 Note that values of mode @code{BLKmode} must be explicitly handled
4364 by this function. Also, the option @option{-fpcc-struct-return}
4365 takes effect regardless of this macro. On most systems, it is
4366 possible to leave the hook undefined; this causes a default
4367 definition to be used, whose value is the constant 1 for @code{BLKmode}
4368 values, and 0 otherwise.
4370 Do not use this hook to indicate that structures and unions should always
4371 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4375 @defmac DEFAULT_PCC_STRUCT_RETURN
4376 Define this macro to be 1 if all structure and union return values must be
4377 in memory. Since this results in slower code, this should be defined
4378 only if needed for compatibility with other compilers or with an ABI@.
4379 If you define this macro to be 0, then the conventions used for structure
4380 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4383 If not defined, this defaults to the value 1.
4386 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4387 This target hook should return the location of the structure value
4388 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4389 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4390 be @code{NULL}, for libcalls. You do not need to define this target
4391 hook if the address is always passed as an ``invisible'' first
4394 On some architectures the place where the structure value address
4395 is found by the called function is not the same place that the
4396 caller put it. This can be due to register windows, or it could
4397 be because the function prologue moves it to a different place.
4398 @var{incoming} is @code{1} or @code{2} when the location is needed in
4399 the context of the called function, and @code{0} in the context of
4402 If @var{incoming} is nonzero and the address is to be found on the
4403 stack, return a @code{mem} which refers to the frame pointer. If
4404 @var{incoming} is @code{2}, the result is being used to fetch the
4405 structure value address at the beginning of a function. If you need
4406 to emit adjusting code, you should do it at this point.
4409 @defmac PCC_STATIC_STRUCT_RETURN
4410 Define this macro if the usual system convention on the target machine
4411 for returning structures and unions is for the called function to return
4412 the address of a static variable containing the value.
4414 Do not define this if the usual system convention is for the caller to
4415 pass an address to the subroutine.
4417 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4418 nothing when you use @option{-freg-struct-return} mode.
4422 @subsection Caller-Saves Register Allocation
4424 If you enable it, GCC can save registers around function calls. This
4425 makes it possible to use call-clobbered registers to hold variables that
4426 must live across calls.
4428 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4429 A C expression to determine whether it is worthwhile to consider placing
4430 a pseudo-register in a call-clobbered hard register and saving and
4431 restoring it around each function call. The expression should be 1 when
4432 this is worth doing, and 0 otherwise.
4434 If you don't define this macro, a default is used which is good on most
4435 machines: @code{4 * @var{calls} < @var{refs}}.
4438 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4439 A C expression specifying which mode is required for saving @var{nregs}
4440 of a pseudo-register in call-clobbered hard register @var{regno}. If
4441 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4442 returned. For most machines this macro need not be defined since GCC
4443 will select the smallest suitable mode.
4446 @node Function Entry
4447 @subsection Function Entry and Exit
4448 @cindex function entry and exit
4452 This section describes the macros that output function entry
4453 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4455 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4456 If defined, a function that outputs the assembler code for entry to a
4457 function. The prologue is responsible for setting up the stack frame,
4458 initializing the frame pointer register, saving registers that must be
4459 saved, and allocating @var{size} additional bytes of storage for the
4460 local variables. @var{size} is an integer. @var{file} is a stdio
4461 stream to which the assembler code should be output.
4463 The label for the beginning of the function need not be output by this
4464 macro. That has already been done when the macro is run.
4466 @findex regs_ever_live
4467 To determine which registers to save, the macro can refer to the array
4468 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4469 @var{r} is used anywhere within the function. This implies the function
4470 prologue should save register @var{r}, provided it is not one of the
4471 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4472 @code{regs_ever_live}.)
4474 On machines that have ``register windows'', the function entry code does
4475 not save on the stack the registers that are in the windows, even if
4476 they are supposed to be preserved by function calls; instead it takes
4477 appropriate steps to ``push'' the register stack, if any non-call-used
4478 registers are used in the function.
4480 @findex frame_pointer_needed
4481 On machines where functions may or may not have frame-pointers, the
4482 function entry code must vary accordingly; it must set up the frame
4483 pointer if one is wanted, and not otherwise. To determine whether a
4484 frame pointer is in wanted, the macro can refer to the variable
4485 @code{frame_pointer_needed}. The variable's value will be 1 at run
4486 time in a function that needs a frame pointer. @xref{Elimination}.
4488 The function entry code is responsible for allocating any stack space
4489 required for the function. This stack space consists of the regions
4490 listed below. In most cases, these regions are allocated in the
4491 order listed, with the last listed region closest to the top of the
4492 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4493 the highest address if it is not defined). You can use a different order
4494 for a machine if doing so is more convenient or required for
4495 compatibility reasons. Except in cases where required by standard
4496 or by a debugger, there is no reason why the stack layout used by GCC
4497 need agree with that used by other compilers for a machine.
4500 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4501 If defined, a function that outputs assembler code at the end of a
4502 prologue. This should be used when the function prologue is being
4503 emitted as RTL, and you have some extra assembler that needs to be
4504 emitted. @xref{prologue instruction pattern}.
4507 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4508 If defined, a function that outputs assembler code at the start of an
4509 epilogue. This should be used when the function epilogue is being
4510 emitted as RTL, and you have some extra assembler that needs to be
4511 emitted. @xref{epilogue instruction pattern}.
4514 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4515 If defined, a function that outputs the assembler code for exit from a
4516 function. The epilogue is responsible for restoring the saved
4517 registers and stack pointer to their values when the function was
4518 called, and returning control to the caller. This macro takes the
4519 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4520 registers to restore are determined from @code{regs_ever_live} and
4521 @code{CALL_USED_REGISTERS} in the same way.
4523 On some machines, there is a single instruction that does all the work
4524 of returning from the function. On these machines, give that
4525 instruction the name @samp{return} and do not define the macro
4526 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4528 Do not define a pattern named @samp{return} if you want the
4529 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4530 switches to control whether return instructions or epilogues are used,
4531 define a @samp{return} pattern with a validity condition that tests the
4532 target switches appropriately. If the @samp{return} pattern's validity
4533 condition is false, epilogues will be used.
4535 On machines where functions may or may not have frame-pointers, the
4536 function exit code must vary accordingly. Sometimes the code for these
4537 two cases is completely different. To determine whether a frame pointer
4538 is wanted, the macro can refer to the variable
4539 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4540 a function that needs a frame pointer.
4542 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4543 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4544 The C variable @code{current_function_is_leaf} is nonzero for such a
4545 function. @xref{Leaf Functions}.
4547 On some machines, some functions pop their arguments on exit while
4548 others leave that for the caller to do. For example, the 68020 when
4549 given @option{-mrtd} pops arguments in functions that take a fixed
4550 number of arguments.
4552 @findex current_function_pops_args
4553 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4554 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4555 needs to know what was decided. The variable that is called
4556 @code{current_function_pops_args} is the number of bytes of its
4557 arguments that a function should pop. @xref{Scalar Return}.
4558 @c what is the "its arguments" in the above sentence referring to, pray
4559 @c tell? --mew 5feb93
4564 @findex current_function_pretend_args_size
4565 A region of @code{current_function_pretend_args_size} bytes of
4566 uninitialized space just underneath the first argument arriving on the
4567 stack. (This may not be at the very start of the allocated stack region
4568 if the calling sequence has pushed anything else since pushing the stack
4569 arguments. But usually, on such machines, nothing else has been pushed
4570 yet, because the function prologue itself does all the pushing.) This
4571 region is used on machines where an argument may be passed partly in
4572 registers and partly in memory, and, in some cases to support the
4573 features in @code{<stdarg.h>}.
4576 An area of memory used to save certain registers used by the function.
4577 The size of this area, which may also include space for such things as
4578 the return address and pointers to previous stack frames, is
4579 machine-specific and usually depends on which registers have been used
4580 in the function. Machines with register windows often do not require
4584 A region of at least @var{size} bytes, possibly rounded up to an allocation
4585 boundary, to contain the local variables of the function. On some machines,
4586 this region and the save area may occur in the opposite order, with the
4587 save area closer to the top of the stack.
4590 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4591 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4592 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4593 argument lists of the function. @xref{Stack Arguments}.
4596 @defmac EXIT_IGNORE_STACK
4597 Define this macro as a C expression that is nonzero if the return
4598 instruction or the function epilogue ignores the value of the stack
4599 pointer; in other words, if it is safe to delete an instruction to
4600 adjust the stack pointer before a return from the function. The
4603 Note that this macro's value is relevant only for functions for which
4604 frame pointers are maintained. It is never safe to delete a final
4605 stack adjustment in a function that has no frame pointer, and the
4606 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4609 @defmac EPILOGUE_USES (@var{regno})
4610 Define this macro as a C expression that is nonzero for registers that are
4611 used by the epilogue or the @samp{return} pattern. The stack and frame
4612 pointer registers are already assumed to be used as needed.
4615 @defmac EH_USES (@var{regno})
4616 Define this macro as a C expression that is nonzero for registers that are
4617 used by the exception handling mechanism, and so should be considered live
4618 on entry to an exception edge.
4621 @defmac DELAY_SLOTS_FOR_EPILOGUE
4622 Define this macro if the function epilogue contains delay slots to which
4623 instructions from the rest of the function can be ``moved''. The
4624 definition should be a C expression whose value is an integer
4625 representing the number of delay slots there.
4628 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4629 A C expression that returns 1 if @var{insn} can be placed in delay
4630 slot number @var{n} of the epilogue.
4632 The argument @var{n} is an integer which identifies the delay slot now
4633 being considered (since different slots may have different rules of
4634 eligibility). It is never negative and is always less than the number
4635 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4636 If you reject a particular insn for a given delay slot, in principle, it
4637 may be reconsidered for a subsequent delay slot. Also, other insns may
4638 (at least in principle) be considered for the so far unfilled delay
4641 @findex current_function_epilogue_delay_list
4642 @findex final_scan_insn
4643 The insns accepted to fill the epilogue delay slots are put in an RTL
4644 list made with @code{insn_list} objects, stored in the variable
4645 @code{current_function_epilogue_delay_list}. The insn for the first
4646 delay slot comes first in the list. Your definition of the macro
4647 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4648 outputting the insns in this list, usually by calling
4649 @code{final_scan_insn}.
4651 You need not define this macro if you did not define
4652 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4655 @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})
4656 A function that outputs the assembler code for a thunk
4657 function, used to implement C++ virtual function calls with multiple
4658 inheritance. The thunk acts as a wrapper around a virtual function,
4659 adjusting the implicit object parameter before handing control off to
4662 First, emit code to add the integer @var{delta} to the location that
4663 contains the incoming first argument. Assume that this argument
4664 contains a pointer, and is the one used to pass the @code{this} pointer
4665 in C++. This is the incoming argument @emph{before} the function prologue,
4666 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4667 all other incoming arguments.
4669 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4670 made after adding @code{delta}. In particular, if @var{p} is the
4671 adjusted pointer, the following adjustment should be made:
4674 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4677 After the additions, emit code to jump to @var{function}, which is a
4678 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4679 not touch the return address. Hence returning from @var{FUNCTION} will
4680 return to whoever called the current @samp{thunk}.
4682 The effect must be as if @var{function} had been called directly with
4683 the adjusted first argument. This macro is responsible for emitting all
4684 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4685 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4687 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4688 have already been extracted from it.) It might possibly be useful on
4689 some targets, but probably not.
4691 If you do not define this macro, the target-independent code in the C++
4692 front end will generate a less efficient heavyweight thunk that calls
4693 @var{function} instead of jumping to it. The generic approach does
4694 not support varargs.
4697 @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})
4698 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4699 to output the assembler code for the thunk function specified by the
4700 arguments it is passed, and false otherwise. In the latter case, the
4701 generic approach will be used by the C++ front end, with the limitations
4706 @subsection Generating Code for Profiling
4707 @cindex profiling, code generation
4709 These macros will help you generate code for profiling.
4711 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4712 A C statement or compound statement to output to @var{file} some
4713 assembler code to call the profiling subroutine @code{mcount}.
4716 The details of how @code{mcount} expects to be called are determined by
4717 your operating system environment, not by GCC@. To figure them out,
4718 compile a small program for profiling using the system's installed C
4719 compiler and look at the assembler code that results.
4721 Older implementations of @code{mcount} expect the address of a counter
4722 variable to be loaded into some register. The name of this variable is
4723 @samp{LP} followed by the number @var{labelno}, so you would generate
4724 the name using @samp{LP%d} in a @code{fprintf}.
4727 @defmac PROFILE_HOOK
4728 A C statement or compound statement to output to @var{file} some assembly
4729 code to call the profiling subroutine @code{mcount} even the target does
4730 not support profiling.
4733 @defmac NO_PROFILE_COUNTERS
4734 Define this macro to be an expression with a nonzero value if the
4735 @code{mcount} subroutine on your system does not need a counter variable
4736 allocated for each function. This is true for almost all modern
4737 implementations. If you define this macro, you must not use the
4738 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4741 @defmac PROFILE_BEFORE_PROLOGUE
4742 Define this macro if the code for function profiling should come before
4743 the function prologue. Normally, the profiling code comes after.
4747 @subsection Permitting tail calls
4750 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4751 True if it is ok to do sibling call optimization for the specified
4752 call expression @var{exp}. @var{decl} will be the called function,
4753 or @code{NULL} if this is an indirect call.
4755 It is not uncommon for limitations of calling conventions to prevent
4756 tail calls to functions outside the current unit of translation, or
4757 during PIC compilation. The hook is used to enforce these restrictions,
4758 as the @code{sibcall} md pattern can not fail, or fall over to a
4759 ``normal'' call. The criteria for successful sibling call optimization
4760 may vary greatly between different architectures.
4763 @deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap *@var{regs})
4764 Add any hard registers to @var{regs} that are live on entry to the
4765 function. This hook only needs to be defined to provide registers that
4766 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4767 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4768 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4769 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4772 @node Stack Smashing Protection
4773 @subsection Stack smashing protection
4774 @cindex stack smashing protection
4776 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void)
4777 This hook returns a @code{DECL} node for the external variable to use
4778 for the stack protection guard. This variable is initialized by the
4779 runtime to some random value and is used to initialize the guard value
4780 that is placed at the top of the local stack frame. The type of this
4781 variable must be @code{ptr_type_node}.
4783 The default version of this hook creates a variable called
4784 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
4787 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void)
4788 This hook returns a tree expression that alerts the runtime that the
4789 stack protect guard variable has been modified. This expression should
4790 involve a call to a @code{noreturn} function.
4792 The default version of this hook invokes a function called
4793 @samp{__stack_chk_fail}, taking no arguments. This function is
4794 normally defined in @file{libgcc2.c}.
4798 @section Implementing the Varargs Macros
4799 @cindex varargs implementation
4801 GCC comes with an implementation of @code{<varargs.h>} and
4802 @code{<stdarg.h>} that work without change on machines that pass arguments
4803 on the stack. Other machines require their own implementations of
4804 varargs, and the two machine independent header files must have
4805 conditionals to include it.
4807 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4808 the calling convention for @code{va_start}. The traditional
4809 implementation takes just one argument, which is the variable in which
4810 to store the argument pointer. The ISO implementation of
4811 @code{va_start} takes an additional second argument. The user is
4812 supposed to write the last named argument of the function here.
4814 However, @code{va_start} should not use this argument. The way to find
4815 the end of the named arguments is with the built-in functions described
4818 @defmac __builtin_saveregs ()
4819 Use this built-in function to save the argument registers in memory so
4820 that the varargs mechanism can access them. Both ISO and traditional
4821 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4822 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4824 On some machines, @code{__builtin_saveregs} is open-coded under the
4825 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4826 other machines, it calls a routine written in assembler language,
4827 found in @file{libgcc2.c}.
4829 Code generated for the call to @code{__builtin_saveregs} appears at the
4830 beginning of the function, as opposed to where the call to
4831 @code{__builtin_saveregs} is written, regardless of what the code is.
4832 This is because the registers must be saved before the function starts
4833 to use them for its own purposes.
4834 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4838 @defmac __builtin_args_info (@var{category})
4839 Use this built-in function to find the first anonymous arguments in
4842 In general, a machine may have several categories of registers used for
4843 arguments, each for a particular category of data types. (For example,
4844 on some machines, floating-point registers are used for floating-point
4845 arguments while other arguments are passed in the general registers.)
4846 To make non-varargs functions use the proper calling convention, you
4847 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4848 registers in each category have been used so far
4850 @code{__builtin_args_info} accesses the same data structure of type
4851 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4852 with it, with @var{category} specifying which word to access. Thus, the
4853 value indicates the first unused register in a given category.
4855 Normally, you would use @code{__builtin_args_info} in the implementation
4856 of @code{va_start}, accessing each category just once and storing the
4857 value in the @code{va_list} object. This is because @code{va_list} will
4858 have to update the values, and there is no way to alter the
4859 values accessed by @code{__builtin_args_info}.
4862 @defmac __builtin_next_arg (@var{lastarg})
4863 This is the equivalent of @code{__builtin_args_info}, for stack
4864 arguments. It returns the address of the first anonymous stack
4865 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4866 returns the address of the location above the first anonymous stack
4867 argument. Use it in @code{va_start} to initialize the pointer for
4868 fetching arguments from the stack. Also use it in @code{va_start} to
4869 verify that the second parameter @var{lastarg} is the last named argument
4870 of the current function.
4873 @defmac __builtin_classify_type (@var{object})
4874 Since each machine has its own conventions for which data types are
4875 passed in which kind of register, your implementation of @code{va_arg}
4876 has to embody these conventions. The easiest way to categorize the
4877 specified data type is to use @code{__builtin_classify_type} together
4878 with @code{sizeof} and @code{__alignof__}.
4880 @code{__builtin_classify_type} ignores the value of @var{object},
4881 considering only its data type. It returns an integer describing what
4882 kind of type that is---integer, floating, pointer, structure, and so on.
4884 The file @file{typeclass.h} defines an enumeration that you can use to
4885 interpret the values of @code{__builtin_classify_type}.
4888 These machine description macros help implement varargs:
4890 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
4891 If defined, this hook produces the machine-specific code for a call to
4892 @code{__builtin_saveregs}. This code will be moved to the very
4893 beginning of the function, before any parameter access are made. The
4894 return value of this function should be an RTX that contains the value
4895 to use as the return of @code{__builtin_saveregs}.
4898 @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})
4899 This target hook offers an alternative to using
4900 @code{__builtin_saveregs} and defining the hook
4901 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
4902 register arguments into the stack so that all the arguments appear to
4903 have been passed consecutively on the stack. Once this is done, you can
4904 use the standard implementation of varargs that works for machines that
4905 pass all their arguments on the stack.
4907 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
4908 structure, containing the values that are obtained after processing the
4909 named arguments. The arguments @var{mode} and @var{type} describe the
4910 last named argument---its machine mode and its data type as a tree node.
4912 The target hook should do two things: first, push onto the stack all the
4913 argument registers @emph{not} used for the named arguments, and second,
4914 store the size of the data thus pushed into the @code{int}-valued
4915 variable pointed to by @var{pretend_args_size}. The value that you
4916 store here will serve as additional offset for setting up the stack
4919 Because you must generate code to push the anonymous arguments at
4920 compile time without knowing their data types,
4921 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
4922 have just a single category of argument register and use it uniformly
4925 If the argument @var{second_time} is nonzero, it means that the
4926 arguments of the function are being analyzed for the second time. This
4927 happens for an inline function, which is not actually compiled until the
4928 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
4929 not generate any instructions in this case.
4932 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca})
4933 Define this hook to return @code{true} if the location where a function
4934 argument is passed depends on whether or not it is a named argument.
4936 This hook controls how the @var{named} argument to @code{FUNCTION_ARG}
4937 is set for varargs and stdarg functions. If this hook returns
4938 @code{true}, the @var{named} argument is always true for named
4939 arguments, and false for unnamed arguments. If it returns @code{false},
4940 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
4941 then all arguments are treated as named. Otherwise, all named arguments
4942 except the last are treated as named.
4944 You need not define this hook if it always returns zero.
4947 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED
4948 If you need to conditionally change ABIs so that one works with
4949 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
4950 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
4951 defined, then define this hook to return @code{true} if
4952 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
4953 Otherwise, you should not define this hook.
4957 @section Trampolines for Nested Functions
4958 @cindex trampolines for nested functions
4959 @cindex nested functions, trampolines for
4961 A @dfn{trampoline} is a small piece of code that is created at run time
4962 when the address of a nested function is taken. It normally resides on
4963 the stack, in the stack frame of the containing function. These macros
4964 tell GCC how to generate code to allocate and initialize a
4967 The instructions in the trampoline must do two things: load a constant
4968 address into the static chain register, and jump to the real address of
4969 the nested function. On CISC machines such as the m68k, this requires
4970 two instructions, a move immediate and a jump. Then the two addresses
4971 exist in the trampoline as word-long immediate operands. On RISC
4972 machines, it is often necessary to load each address into a register in
4973 two parts. Then pieces of each address form separate immediate
4976 The code generated to initialize the trampoline must store the variable
4977 parts---the static chain value and the function address---into the
4978 immediate operands of the instructions. On a CISC machine, this is
4979 simply a matter of copying each address to a memory reference at the
4980 proper offset from the start of the trampoline. On a RISC machine, it
4981 may be necessary to take out pieces of the address and store them
4984 @defmac TRAMPOLINE_TEMPLATE (@var{file})
4985 A C statement to output, on the stream @var{file}, assembler code for a
4986 block of data that contains the constant parts of a trampoline. This
4987 code should not include a label---the label is taken care of
4990 If you do not define this macro, it means no template is needed
4991 for the target. Do not define this macro on systems where the block move
4992 code to copy the trampoline into place would be larger than the code
4993 to generate it on the spot.
4996 @defmac TRAMPOLINE_SECTION
4997 Return the section into which the trampoline template is to be placed
4998 (@pxref{Sections}). The default value is @code{readonly_data_section}.
5001 @defmac TRAMPOLINE_SIZE
5002 A C expression for the size in bytes of the trampoline, as an integer.
5005 @defmac TRAMPOLINE_ALIGNMENT
5006 Alignment required for trampolines, in bits.
5008 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
5009 is used for aligning trampolines.
5012 @defmac INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
5013 A C statement to initialize the variable parts of a trampoline.
5014 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
5015 an RTX for the address of the nested function; @var{static_chain} is an
5016 RTX for the static chain value that should be passed to the function
5020 @defmac TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
5021 A C statement that should perform any machine-specific adjustment in
5022 the address of the trampoline. Its argument contains the address that
5023 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
5024 used for a function call should be different from the address in which
5025 the template was stored, the different address should be assigned to
5026 @var{addr}. If this macro is not defined, @var{addr} will be used for
5029 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
5030 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
5031 If this macro is not defined, by default the trampoline is allocated as
5032 a stack slot. This default is right for most machines. The exceptions
5033 are machines where it is impossible to execute instructions in the stack
5034 area. On such machines, you may have to implement a separate stack,
5035 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
5036 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
5038 @var{fp} points to a data structure, a @code{struct function}, which
5039 describes the compilation status of the immediate containing function of
5040 the function which the trampoline is for. The stack slot for the
5041 trampoline is in the stack frame of this containing function. Other
5042 allocation strategies probably must do something analogous with this
5046 Implementing trampolines is difficult on many machines because they have
5047 separate instruction and data caches. Writing into a stack location
5048 fails to clear the memory in the instruction cache, so when the program
5049 jumps to that location, it executes the old contents.
5051 Here are two possible solutions. One is to clear the relevant parts of
5052 the instruction cache whenever a trampoline is set up. The other is to
5053 make all trampolines identical, by having them jump to a standard
5054 subroutine. The former technique makes trampoline execution faster; the
5055 latter makes initialization faster.
5057 To clear the instruction cache when a trampoline is initialized, define
5058 the following macro.
5060 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
5061 If defined, expands to a C expression clearing the @emph{instruction
5062 cache} in the specified interval. The definition of this macro would
5063 typically be a series of @code{asm} statements. Both @var{beg} and
5064 @var{end} are both pointer expressions.
5067 The operating system may also require the stack to be made executable
5068 before calling the trampoline. To implement this requirement, define
5069 the following macro.
5071 @defmac ENABLE_EXECUTE_STACK
5072 Define this macro if certain operations must be performed before executing
5073 code located on the stack. The macro should expand to a series of C
5074 file-scope constructs (e.g.@: functions) and provide a unique entry point
5075 named @code{__enable_execute_stack}. The target is responsible for
5076 emitting calls to the entry point in the code, for example from the
5077 @code{INITIALIZE_TRAMPOLINE} macro.
5080 To use a standard subroutine, define the following macro. In addition,
5081 you must make sure that the instructions in a trampoline fill an entire
5082 cache line with identical instructions, or else ensure that the
5083 beginning of the trampoline code is always aligned at the same point in
5084 its cache line. Look in @file{m68k.h} as a guide.
5086 @defmac TRANSFER_FROM_TRAMPOLINE
5087 Define this macro if trampolines need a special subroutine to do their
5088 work. The macro should expand to a series of @code{asm} statements
5089 which will be compiled with GCC@. They go in a library function named
5090 @code{__transfer_from_trampoline}.
5092 If you need to avoid executing the ordinary prologue code of a compiled
5093 C function when you jump to the subroutine, you can do so by placing a
5094 special label of your own in the assembler code. Use one @code{asm}
5095 statement to generate an assembler label, and another to make the label
5096 global. Then trampolines can use that label to jump directly to your
5097 special assembler code.
5101 @section Implicit Calls to Library Routines
5102 @cindex library subroutine names
5103 @cindex @file{libgcc.a}
5105 @c prevent bad page break with this line
5106 Here is an explanation of implicit calls to library routines.
5108 @defmac DECLARE_LIBRARY_RENAMES
5109 This macro, if defined, should expand to a piece of C code that will get
5110 expanded when compiling functions for libgcc.a. It can be used to
5111 provide alternate names for GCC's internal library functions if there
5112 are ABI-mandated names that the compiler should provide.
5115 @findex init_one_libfunc
5116 @findex set_optab_libfunc
5117 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
5118 This hook should declare additional library routines or rename
5119 existing ones, using the functions @code{set_optab_libfunc} and
5120 @code{init_one_libfunc} defined in @file{optabs.c}.
5121 @code{init_optabs} calls this macro after initializing all the normal
5124 The default is to do nothing. Most ports don't need to define this hook.
5127 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
5128 This macro should return @code{true} if the library routine that
5129 implements the floating point comparison operator @var{comparison} in
5130 mode @var{mode} will return a boolean, and @var{false} if it will
5133 GCC's own floating point libraries return tristates from the
5134 comparison operators, so the default returns false always. Most ports
5135 don't need to define this macro.
5138 @defmac TARGET_LIB_INT_CMP_BIASED
5139 This macro should evaluate to @code{true} if the integer comparison
5140 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5141 operand is smaller than the second, 1 to indicate that they are equal,
5142 and 2 to indicate that the first operand is greater than the second.
5143 If this macro evaluates to @code{false} the comparison functions return
5144 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5145 in @file{libgcc.a}, you do not need to define this macro.
5148 @cindex US Software GOFAST, floating point emulation library
5149 @cindex floating point emulation library, US Software GOFAST
5150 @cindex GOFAST, floating point emulation library
5151 @findex gofast_maybe_init_libfuncs
5152 @defmac US_SOFTWARE_GOFAST
5153 Define this macro if your system C library uses the US Software GOFAST
5154 library to provide floating point emulation.
5156 In addition to defining this macro, your architecture must set
5157 @code{TARGET_INIT_LIBFUNCS} to @code{gofast_maybe_init_libfuncs}, or
5158 else call that function from its version of that hook. It is defined
5159 in @file{config/gofast.h}, which must be included by your
5160 architecture's @file{@var{cpu}.c} file. See @file{sparc/sparc.c} for
5163 If this macro is defined, the
5164 @code{TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL} target hook must return
5165 false for @code{SFmode} and @code{DFmode} comparisons.
5168 @cindex @code{EDOM}, implicit usage
5171 The value of @code{EDOM} on the target machine, as a C integer constant
5172 expression. If you don't define this macro, GCC does not attempt to
5173 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5174 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5177 If you do not define @code{TARGET_EDOM}, then compiled code reports
5178 domain errors by calling the library function and letting it report the
5179 error. If mathematical functions on your system use @code{matherr} when
5180 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5181 that @code{matherr} is used normally.
5184 @cindex @code{errno}, implicit usage
5185 @defmac GEN_ERRNO_RTX
5186 Define this macro as a C expression to create an rtl expression that
5187 refers to the global ``variable'' @code{errno}. (On certain systems,
5188 @code{errno} may not actually be a variable.) If you don't define this
5189 macro, a reasonable default is used.
5192 @cindex C99 math functions, implicit usage
5193 @defmac TARGET_C99_FUNCTIONS
5194 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
5195 @code{sinf} and similarly for other functions defined by C99 standard. The
5196 default is nonzero that should be proper value for most modern systems, however
5197 number of existing systems lacks support for these functions in the runtime so
5198 they needs this macro to be redefined to 0.
5201 @cindex sincos math function, implicit usage
5202 @defmac TARGET_HAS_SINCOS
5203 When this macro is nonzero, GCC will implicitly optimize calls to @code{sin}
5204 and @code{cos} with the same argument to a call to @code{sincos}. The
5205 default is zero. The target has to provide the following functions:
5207 void sincos(double x, double *sin, double *cos);
5208 void sincosf(float x, float *sin, float *cos);
5209 void sincosl(long double x, long double *sin, long double *cos);
5213 @defmac NEXT_OBJC_RUNTIME
5214 Define this macro to generate code for Objective-C message sending using
5215 the calling convention of the NeXT system. This calling convention
5216 involves passing the object, the selector and the method arguments all
5217 at once to the method-lookup library function.
5219 The default calling convention passes just the object and the selector
5220 to the lookup function, which returns a pointer to the method.
5223 @node Addressing Modes
5224 @section Addressing Modes
5225 @cindex addressing modes
5227 @c prevent bad page break with this line
5228 This is about addressing modes.
5230 @defmac HAVE_PRE_INCREMENT
5231 @defmacx HAVE_PRE_DECREMENT
5232 @defmacx HAVE_POST_INCREMENT
5233 @defmacx HAVE_POST_DECREMENT
5234 A C expression that is nonzero if the machine supports pre-increment,
5235 pre-decrement, post-increment, or post-decrement addressing respectively.
5238 @defmac HAVE_PRE_MODIFY_DISP
5239 @defmacx HAVE_POST_MODIFY_DISP
5240 A C expression that is nonzero if the machine supports pre- or
5241 post-address side-effect generation involving constants other than
5242 the size of the memory operand.
5245 @defmac HAVE_PRE_MODIFY_REG
5246 @defmacx HAVE_POST_MODIFY_REG
5247 A C expression that is nonzero if the machine supports pre- or
5248 post-address side-effect generation involving a register displacement.
5251 @defmac CONSTANT_ADDRESS_P (@var{x})
5252 A C expression that is 1 if the RTX @var{x} is a constant which
5253 is a valid address. On most machines, this can be defined as
5254 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
5255 in which constant addresses are supported.
5258 @defmac CONSTANT_P (@var{x})
5259 @code{CONSTANT_P}, which is defined by target-independent code,
5260 accepts integer-values expressions whose values are not explicitly
5261 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5262 expressions and @code{const} arithmetic expressions, in addition to
5263 @code{const_int} and @code{const_double} expressions.
5266 @defmac MAX_REGS_PER_ADDRESS
5267 A number, the maximum number of registers that can appear in a valid
5268 memory address. Note that it is up to you to specify a value equal to
5269 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
5273 @defmac GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5274 A C compound statement with a conditional @code{goto @var{label};}
5275 executed if @var{x} (an RTX) is a legitimate memory address on the
5276 target machine for a memory operand of mode @var{mode}.
5278 It usually pays to define several simpler macros to serve as
5279 subroutines for this one. Otherwise it may be too complicated to
5282 This macro must exist in two variants: a strict variant and a
5283 non-strict one. The strict variant is used in the reload pass. It
5284 must be defined so that any pseudo-register that has not been
5285 allocated a hard register is considered a memory reference. In
5286 contexts where some kind of register is required, a pseudo-register
5287 with no hard register must be rejected.
5289 The non-strict variant is used in other passes. It must be defined to
5290 accept all pseudo-registers in every context where some kind of
5291 register is required.
5293 @findex REG_OK_STRICT
5294 Compiler source files that want to use the strict variant of this
5295 macro define the macro @code{REG_OK_STRICT}. You should use an
5296 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
5297 in that case and the non-strict variant otherwise.
5299 Subroutines to check for acceptable registers for various purposes (one
5300 for base registers, one for index registers, and so on) are typically
5301 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
5302 Then only these subroutine macros need have two variants; the higher
5303 levels of macros may be the same whether strict or not.
5305 Normally, constant addresses which are the sum of a @code{symbol_ref}
5306 and an integer are stored inside a @code{const} RTX to mark them as
5307 constant. Therefore, there is no need to recognize such sums
5308 specifically as legitimate addresses. Normally you would simply
5309 recognize any @code{const} as legitimate.
5311 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5312 sums that are not marked with @code{const}. It assumes that a naked
5313 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5314 naked constant sums as illegitimate addresses, so that none of them will
5315 be given to @code{PRINT_OPERAND_ADDRESS}.
5317 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5318 On some machines, whether a symbolic address is legitimate depends on
5319 the section that the address refers to. On these machines, define the
5320 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5321 into the @code{symbol_ref}, and then check for it here. When you see a
5322 @code{const}, you will have to look inside it to find the
5323 @code{symbol_ref} in order to determine the section. @xref{Assembler
5327 @defmac TARGET_MEM_CONSTRAINT
5328 A single character to be used instead of the default @code{'m'}
5329 character for general memory addresses. This defines the constraint
5330 letter which matches the memory addresses accepted by
5331 @code{GO_IF_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
5332 support new address formats in your back end without changing the
5333 semantics of the @code{'m'} constraint. This is necessary in order to
5334 preserve functionality of inline assembly constructs using the
5335 @code{'m'} constraint.
5338 @defmac FIND_BASE_TERM (@var{x})
5339 A C expression to determine the base term of address @var{x}.
5340 This macro is used in only one place: `find_base_term' in alias.c.
5342 It is always safe for this macro to not be defined. It exists so
5343 that alias analysis can understand machine-dependent addresses.
5345 The typical use of this macro is to handle addresses containing
5346 a label_ref or symbol_ref within an UNSPEC@.
5349 @defmac LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
5350 A C compound statement that attempts to replace @var{x} with a valid
5351 memory address for an operand of mode @var{mode}. @var{win} will be a
5352 C statement label elsewhere in the code; the macro definition may use
5355 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
5359 to avoid further processing if the address has become legitimate.
5361 @findex break_out_memory_refs
5362 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5363 and @var{oldx} will be the operand that was given to that function to produce
5366 The code generated by this macro should not alter the substructure of
5367 @var{x}. If it transforms @var{x} into a more legitimate form, it
5368 should assign @var{x} (which will always be a C variable) a new value.
5370 It is not necessary for this macro to come up with a legitimate
5371 address. The compiler has standard ways of doing so in all cases. In
5372 fact, it is safe to omit this macro. But often a
5373 machine-dependent strategy can generate better code.
5376 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5377 A C compound statement that attempts to replace @var{x}, which is an address
5378 that needs reloading, with a valid memory address for an operand of mode
5379 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5380 It is not necessary to define this macro, but it might be useful for
5381 performance reasons.
5383 For example, on the i386, it is sometimes possible to use a single
5384 reload register instead of two by reloading a sum of two pseudo
5385 registers into a register. On the other hand, for number of RISC
5386 processors offsets are limited so that often an intermediate address
5387 needs to be generated in order to address a stack slot. By defining
5388 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5389 generated for adjacent some stack slots can be made identical, and thus
5392 @emph{Note}: This macro should be used with caution. It is necessary
5393 to know something of how reload works in order to effectively use this,
5394 and it is quite easy to produce macros that build in too much knowledge
5395 of reload internals.
5397 @emph{Note}: This macro must be able to reload an address created by a
5398 previous invocation of this macro. If it fails to handle such addresses
5399 then the compiler may generate incorrect code or abort.
5402 The macro definition should use @code{push_reload} to indicate parts that
5403 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5404 suitable to be passed unaltered to @code{push_reload}.
5406 The code generated by this macro must not alter the substructure of
5407 @var{x}. If it transforms @var{x} into a more legitimate form, it
5408 should assign @var{x} (which will always be a C variable) a new value.
5409 This also applies to parts that you change indirectly by calling
5412 @findex strict_memory_address_p
5413 The macro definition may use @code{strict_memory_address_p} to test if
5414 the address has become legitimate.
5417 If you want to change only a part of @var{x}, one standard way of doing
5418 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5419 single level of rtl. Thus, if the part to be changed is not at the
5420 top level, you'll need to replace first the top level.
5421 It is not necessary for this macro to come up with a legitimate
5422 address; but often a machine-dependent strategy can generate better code.
5425 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5426 A C statement or compound statement with a conditional @code{goto
5427 @var{label};} executed if memory address @var{x} (an RTX) can have
5428 different meanings depending on the machine mode of the memory
5429 reference it is used for or if the address is valid for some modes
5432 Autoincrement and autodecrement addresses typically have mode-dependent
5433 effects because the amount of the increment or decrement is the size
5434 of the operand being addressed. Some machines have other mode-dependent
5435 addresses. Many RISC machines have no mode-dependent addresses.
5437 You may assume that @var{addr} is a valid address for the machine.
5440 @defmac LEGITIMATE_CONSTANT_P (@var{x})
5441 A C expression that is nonzero if @var{x} is a legitimate constant for
5442 an immediate operand on the target machine. You can assume that
5443 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5444 @samp{1} is a suitable definition for this macro on machines where
5445 anything @code{CONSTANT_P} is valid.
5448 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5449 This hook is used to undo the possibly obfuscating effects of the
5450 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5451 macros. Some backend implementations of these macros wrap symbol
5452 references inside an @code{UNSPEC} rtx to represent PIC or similar
5453 addressing modes. This target hook allows GCC's optimizers to understand
5454 the semantics of these opaque @code{UNSPEC}s by converting them back
5455 into their original form.
5458 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (rtx @var{x})
5459 This hook should return true if @var{x} is of a form that cannot (or
5460 should not) be spilled to the constant pool. The default version of
5461 this hook returns false.
5463 The primary reason to define this hook is to prevent reload from
5464 deciding that a non-legitimate constant would be better reloaded
5465 from the constant pool instead of spilling and reloading a register
5466 holding the constant. This restriction is often true of addresses
5467 of TLS symbols for various targets.
5470 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (enum machine_mode @var{mode}, rtx @var{x})
5471 This hook should return true if pool entries for constant @var{x} can
5472 be placed in an @code{object_block} structure. @var{mode} is the mode
5475 The default version returns false for all constants.
5478 @deftypefn {Target Hook} tree TARGET_BUILTIN_RECIPROCAL (enum tree_code @var{fn}, bool @var{tm_fn}, bool @var{sqrt})
5479 This hook should return the DECL of a function that implements reciprocal of
5480 the builtin function with builtin function code @var{fn}, or
5481 @code{NULL_TREE} if such a function is not available. @var{tm_fn} is true
5482 when @var{fn} is a code of a machine-dependent builtin function. When
5483 @var{sqrt} is true, additional optimizations that apply only to the reciprocal
5484 of a square root function are performed, and only reciprocals of @code{sqrt}
5488 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5489 This hook should return the DECL of a function @var{f} that given an
5490 address @var{addr} as an argument returns a mask @var{m} that can be
5491 used to extract from two vectors the relevant data that resides in
5492 @var{addr} in case @var{addr} is not properly aligned.
5494 The autovectorizer, when vectorizing a load operation from an address
5495 @var{addr} that may be unaligned, will generate two vector loads from
5496 the two aligned addresses around @var{addr}. It then generates a
5497 @code{REALIGN_LOAD} operation to extract the relevant data from the
5498 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5499 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5500 the third argument, @var{OFF}, defines how the data will be extracted
5501 from these two vectors: if @var{OFF} is 0, then the returned vector is
5502 @var{v2}; otherwise, the returned vector is composed from the last
5503 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5504 @var{OFF} elements of @var{v2}.
5506 If this hook is defined, the autovectorizer will generate a call
5507 to @var{f} (using the DECL tree that this hook returns) and will
5508 use the return value of @var{f} as the argument @var{OFF} to
5509 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5510 should comply with the semantics expected by @code{REALIGN_LOAD}
5512 If this hook is not defined, then @var{addr} will be used as
5513 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5514 log2(@var{VS})-1 bits of @var{addr} will be considered.
5517 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN (tree @var{x})
5518 This hook should return the DECL of a function @var{f} that implements
5519 widening multiplication of the even elements of two input vectors of type @var{x}.
5521 If this hook is defined, the autovectorizer will use it along with the
5522 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD} target hook when vectorizing
5523 widening multiplication in cases that the order of the results does not have to be
5524 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5525 @code{widen_mult_hi/lo} idioms will be used.
5528 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD (tree @var{x})
5529 This hook should return the DECL of a function @var{f} that implements
5530 widening multiplication of the odd elements of two input vectors of type @var{x}.
5532 If this hook is defined, the autovectorizer will use it along with the
5533 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN} target hook when vectorizing
5534 widening multiplication in cases that the order of the results does not have to be
5535 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5536 @code{widen_mult_hi/lo} idioms will be used.
5539 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_CONVERSION (enum tree_code @var{code}, tree @var{type})
5540 This hook should return the DECL of a function that implements conversion of the
5541 input vector of type @var{type}.
5542 If @var{type} is an integral type, the result of the conversion is a vector of
5543 floating-point type of the same size.
5544 If @var{type} is a floating-point type, the result of the conversion is a vector
5545 of integral type of the same size.
5546 @var{code} specifies how the conversion is to be applied
5547 (truncation, rounding, etc.).
5549 If this hook is defined, the autovectorizer will use the
5550 @code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing
5551 conversion. Otherwise, it will return @code{NULL_TREE}.
5554 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION (enum built_in_function @var{code}, tree @var{vec_type_out}, tree @var{vec_type_in})
5555 This hook should return the decl of a function that implements the vectorized
5556 variant of the builtin function with builtin function code @var{code} or
5557 @code{NULL_TREE} if such a function is not available. The return type of
5558 the vectorized function shall be of vector type @var{vec_type_out} and the
5559 argument types should be @var{vec_type_in}.
5562 @node Anchored Addresses
5563 @section Anchored Addresses
5564 @cindex anchored addresses
5565 @cindex @option{-fsection-anchors}
5567 GCC usually addresses every static object as a separate entity.
5568 For example, if we have:
5572 int foo (void) @{ return a + b + c; @}
5575 the code for @code{foo} will usually calculate three separate symbolic
5576 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
5577 it would be better to calculate just one symbolic address and access
5578 the three variables relative to it. The equivalent pseudocode would
5584 register int *xr = &x;
5585 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5589 (which isn't valid C). We refer to shared addresses like @code{x} as
5590 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
5592 The hooks below describe the target properties that GCC needs to know
5593 in order to make effective use of section anchors. It won't use
5594 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
5595 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
5597 @deftypevar {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET
5598 The minimum offset that should be applied to a section anchor.
5599 On most targets, it should be the smallest offset that can be
5600 applied to a base register while still giving a legitimate address
5601 for every mode. The default value is 0.
5604 @deftypevar {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET
5605 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
5606 offset that should be applied to section anchors. The default
5610 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x})
5611 Write the assembly code to define section anchor @var{x}, which is a
5612 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
5613 The hook is called with the assembly output position set to the beginning
5614 of @code{SYMBOL_REF_BLOCK (@var{x})}.
5616 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
5617 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
5618 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
5619 is @code{NULL}, which disables the use of section anchors altogether.
5622 @deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (rtx @var{x})
5623 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
5624 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
5625 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
5627 The default version is correct for most targets, but you might need to
5628 intercept this hook to handle things like target-specific attributes
5629 or target-specific sections.
5632 @node Condition Code
5633 @section Condition Code Status
5634 @cindex condition code status
5636 @c prevent bad page break with this line
5637 This describes the condition code status.
5640 The file @file{conditions.h} defines a variable @code{cc_status} to
5641 describe how the condition code was computed (in case the interpretation of
5642 the condition code depends on the instruction that it was set by). This
5643 variable contains the RTL expressions on which the condition code is
5644 currently based, and several standard flags.
5646 Sometimes additional machine-specific flags must be defined in the machine
5647 description header file. It can also add additional machine-specific
5648 information by defining @code{CC_STATUS_MDEP}.
5650 @defmac CC_STATUS_MDEP
5651 C code for a data type which is used for declaring the @code{mdep}
5652 component of @code{cc_status}. It defaults to @code{int}.
5654 This macro is not used on machines that do not use @code{cc0}.
5657 @defmac CC_STATUS_MDEP_INIT
5658 A C expression to initialize the @code{mdep} field to ``empty''.
5659 The default definition does nothing, since most machines don't use
5660 the field anyway. If you want to use the field, you should probably
5661 define this macro to initialize it.
5663 This macro is not used on machines that do not use @code{cc0}.
5666 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5667 A C compound statement to set the components of @code{cc_status}
5668 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5669 this macro's responsibility to recognize insns that set the condition
5670 code as a byproduct of other activity as well as those that explicitly
5673 This macro is not used on machines that do not use @code{cc0}.
5675 If there are insns that do not set the condition code but do alter
5676 other machine registers, this macro must check to see whether they
5677 invalidate the expressions that the condition code is recorded as
5678 reflecting. For example, on the 68000, insns that store in address
5679 registers do not set the condition code, which means that usually
5680 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5681 insns. But suppose that the previous insn set the condition code
5682 based on location @samp{a4@@(102)} and the current insn stores a new
5683 value in @samp{a4}. Although the condition code is not changed by
5684 this, it will no longer be true that it reflects the contents of
5685 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5686 @code{cc_status} in this case to say that nothing is known about the
5687 condition code value.
5689 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5690 with the results of peephole optimization: insns whose patterns are
5691 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5692 constants which are just the operands. The RTL structure of these
5693 insns is not sufficient to indicate what the insns actually do. What
5694 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5695 @code{CC_STATUS_INIT}.
5697 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5698 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5699 @samp{cc}. This avoids having detailed information about patterns in
5700 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5703 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5704 Returns a mode from class @code{MODE_CC} to be used when comparison
5705 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5706 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5707 @pxref{Jump Patterns} for a description of the reason for this
5711 #define SELECT_CC_MODE(OP,X,Y) \
5712 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5713 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5714 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5715 || GET_CODE (X) == NEG) \
5716 ? CC_NOOVmode : CCmode))
5719 You should define this macro if and only if you define extra CC modes
5720 in @file{@var{machine}-modes.def}.
5723 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5724 On some machines not all possible comparisons are defined, but you can
5725 convert an invalid comparison into a valid one. For example, the Alpha
5726 does not have a @code{GT} comparison, but you can use an @code{LT}
5727 comparison instead and swap the order of the operands.
5729 On such machines, define this macro to be a C statement to do any
5730 required conversions. @var{code} is the initial comparison code
5731 and @var{op0} and @var{op1} are the left and right operands of the
5732 comparison, respectively. You should modify @var{code}, @var{op0}, and
5733 @var{op1} as required.
5735 GCC will not assume that the comparison resulting from this macro is
5736 valid but will see if the resulting insn matches a pattern in the
5739 You need not define this macro if it would never change the comparison
5743 @defmac REVERSIBLE_CC_MODE (@var{mode})
5744 A C expression whose value is one if it is always safe to reverse a
5745 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5746 can ever return @var{mode} for a floating-point inequality comparison,
5747 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5749 You need not define this macro if it would always returns zero or if the
5750 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5751 For example, here is the definition used on the SPARC, where floating-point
5752 inequality comparisons are always given @code{CCFPEmode}:
5755 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5759 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
5760 A C expression whose value is reversed condition code of the @var{code} for
5761 comparison done in CC_MODE @var{mode}. The macro is used only in case
5762 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5763 machine has some non-standard way how to reverse certain conditionals. For
5764 instance in case all floating point conditions are non-trapping, compiler may
5765 freely convert unordered compares to ordered one. Then definition may look
5769 #define REVERSE_CONDITION(CODE, MODE) \
5770 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5771 : reverse_condition_maybe_unordered (CODE))
5775 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
5776 A C expression that returns true if the conditional execution predicate
5777 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
5778 versa. Define this to return 0 if the target has conditional execution
5779 predicates that cannot be reversed safely. There is no need to validate
5780 that the arguments of op1 and op2 are the same, this is done separately.
5781 If no expansion is specified, this macro is defined as follows:
5784 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5785 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
5789 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *, unsigned int *)
5790 On targets which do not use @code{(cc0)}, and which use a hard
5791 register rather than a pseudo-register to hold condition codes, the
5792 regular CSE passes are often not able to identify cases in which the
5793 hard register is set to a common value. Use this hook to enable a
5794 small pass which optimizes such cases. This hook should return true
5795 to enable this pass, and it should set the integers to which its
5796 arguments point to the hard register numbers used for condition codes.
5797 When there is only one such register, as is true on most systems, the
5798 integer pointed to by the second argument should be set to
5799 @code{INVALID_REGNUM}.
5801 The default version of this hook returns false.
5804 @deftypefn {Target Hook} enum machine_mode TARGET_CC_MODES_COMPATIBLE (enum machine_mode, enum machine_mode)
5805 On targets which use multiple condition code modes in class
5806 @code{MODE_CC}, it is sometimes the case that a comparison can be
5807 validly done in more than one mode. On such a system, define this
5808 target hook to take two mode arguments and to return a mode in which
5809 both comparisons may be validly done. If there is no such mode,
5810 return @code{VOIDmode}.
5812 The default version of this hook checks whether the modes are the
5813 same. If they are, it returns that mode. If they are different, it
5814 returns @code{VOIDmode}.
5818 @section Describing Relative Costs of Operations
5819 @cindex costs of instructions
5820 @cindex relative costs
5821 @cindex speed of instructions
5823 These macros let you describe the relative speed of various operations
5824 on the target machine.
5826 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5827 A C expression for the cost of moving data of mode @var{mode} from a
5828 register in class @var{from} to one in class @var{to}. The classes are
5829 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5830 value of 2 is the default; other values are interpreted relative to
5833 It is not required that the cost always equal 2 when @var{from} is the
5834 same as @var{to}; on some machines it is expensive to move between
5835 registers if they are not general registers.
5837 If reload sees an insn consisting of a single @code{set} between two
5838 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5839 classes returns a value of 2, reload does not check to ensure that the
5840 constraints of the insn are met. Setting a cost of other than 2 will
5841 allow reload to verify that the constraints are met. You should do this
5842 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5845 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5846 A C expression for the cost of moving data of mode @var{mode} between a
5847 register of class @var{class} and memory; @var{in} is zero if the value
5848 is to be written to memory, nonzero if it is to be read in. This cost
5849 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5850 registers and memory is more expensive than between two registers, you
5851 should define this macro to express the relative cost.
5853 If you do not define this macro, GCC uses a default cost of 4 plus
5854 the cost of copying via a secondary reload register, if one is
5855 needed. If your machine requires a secondary reload register to copy
5856 between memory and a register of @var{class} but the reload mechanism is
5857 more complex than copying via an intermediate, define this macro to
5858 reflect the actual cost of the move.
5860 GCC defines the function @code{memory_move_secondary_cost} if
5861 secondary reloads are needed. It computes the costs due to copying via
5862 a secondary register. If your machine copies from memory using a
5863 secondary register in the conventional way but the default base value of
5864 4 is not correct for your machine, define this macro to add some other
5865 value to the result of that function. The arguments to that function
5866 are the same as to this macro.
5870 A C expression for the cost of a branch instruction. A value of 1 is
5871 the default; other values are interpreted relative to that.
5874 Here are additional macros which do not specify precise relative costs,
5875 but only that certain actions are more expensive than GCC would
5878 @defmac SLOW_BYTE_ACCESS
5879 Define this macro as a C expression which is nonzero if accessing less
5880 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5881 faster than accessing a word of memory, i.e., if such access
5882 require more than one instruction or if there is no difference in cost
5883 between byte and (aligned) word loads.
5885 When this macro is not defined, the compiler will access a field by
5886 finding the smallest containing object; when it is defined, a fullword
5887 load will be used if alignment permits. Unless bytes accesses are
5888 faster than word accesses, using word accesses is preferable since it
5889 may eliminate subsequent memory access if subsequent accesses occur to
5890 other fields in the same word of the structure, but to different bytes.
5893 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5894 Define this macro to be the value 1 if memory accesses described by the
5895 @var{mode} and @var{alignment} parameters have a cost many times greater
5896 than aligned accesses, for example if they are emulated in a trap
5899 When this macro is nonzero, the compiler will act as if
5900 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5901 moves. This can cause significantly more instructions to be produced.
5902 Therefore, do not set this macro nonzero if unaligned accesses only add a
5903 cycle or two to the time for a memory access.
5905 If the value of this macro is always zero, it need not be defined. If
5906 this macro is defined, it should produce a nonzero value when
5907 @code{STRICT_ALIGNMENT} is nonzero.
5911 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5912 which a sequence of insns should be generated instead of a
5913 string move insn or a library call. Increasing the value will always
5914 make code faster, but eventually incurs high cost in increased code size.
5916 Note that on machines where the corresponding move insn is a
5917 @code{define_expand} that emits a sequence of insns, this macro counts
5918 the number of such sequences.
5920 If you don't define this, a reasonable default is used.
5923 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5924 A C expression used to determine whether @code{move_by_pieces} will be used to
5925 copy a chunk of memory, or whether some other block move mechanism
5926 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5927 than @code{MOVE_RATIO}.
5930 @defmac MOVE_MAX_PIECES
5931 A C expression used by @code{move_by_pieces} to determine the largest unit
5932 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5936 The threshold of number of scalar move insns, @emph{below} which a sequence
5937 of insns should be generated to clear memory instead of a string clear insn
5938 or a library call. Increasing the value will always make code faster, but
5939 eventually incurs high cost in increased code size.
5941 If you don't define this, a reasonable default is used.
5944 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5945 A C expression used to determine whether @code{clear_by_pieces} will be used
5946 to clear a chunk of memory, or whether some other block clear mechanism
5947 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5948 than @code{CLEAR_RATIO}.
5952 The threshold of number of scalar move insns, @emph{below} which a sequence
5953 of insns should be generated to set memory to a constant value, instead of
5954 a block set insn or a library call.
5955 Increasing the value will always make code faster, but
5956 eventually incurs high cost in increased code size.
5958 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
5961 @defmac SET_BY_PIECES_P (@var{size}, @var{alignment})
5962 A C expression used to determine whether @code{store_by_pieces} will be
5963 used to set a chunk of memory to a constant value, or whether some
5964 other mechanism will be used. Used by @code{__builtin_memset} when
5965 storing values other than constant zero.
5966 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5967 than @code{SET_RATIO}.
5970 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
5971 A C expression used to determine whether @code{store_by_pieces} will be
5972 used to set a chunk of memory to a constant string value, or whether some
5973 other mechanism will be used. Used by @code{__builtin_strcpy} when
5974 called with a constant source string.
5975 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5976 than @code{MOVE_RATIO}.
5979 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
5980 A C expression used to determine whether a load postincrement is a good
5981 thing to use for a given mode. Defaults to the value of
5982 @code{HAVE_POST_INCREMENT}.
5985 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
5986 A C expression used to determine whether a load postdecrement is a good
5987 thing to use for a given mode. Defaults to the value of
5988 @code{HAVE_POST_DECREMENT}.
5991 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
5992 A C expression used to determine whether a load preincrement is a good
5993 thing to use for a given mode. Defaults to the value of
5994 @code{HAVE_PRE_INCREMENT}.
5997 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
5998 A C expression used to determine whether a load predecrement is a good
5999 thing to use for a given mode. Defaults to the value of
6000 @code{HAVE_PRE_DECREMENT}.
6003 @defmac USE_STORE_POST_INCREMENT (@var{mode})
6004 A C expression used to determine whether a store postincrement is a good
6005 thing to use for a given mode. Defaults to the value of
6006 @code{HAVE_POST_INCREMENT}.
6009 @defmac USE_STORE_POST_DECREMENT (@var{mode})
6010 A C expression used to determine whether a store postdecrement is a good
6011 thing to use for a given mode. Defaults to the value of
6012 @code{HAVE_POST_DECREMENT}.
6015 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
6016 This macro is used to determine whether a store preincrement is a good
6017 thing to use for a given mode. Defaults to the value of
6018 @code{HAVE_PRE_INCREMENT}.
6021 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
6022 This macro is used to determine whether a store predecrement is a good
6023 thing to use for a given mode. Defaults to the value of
6024 @code{HAVE_PRE_DECREMENT}.
6027 @defmac NO_FUNCTION_CSE
6028 Define this macro if it is as good or better to call a constant
6029 function address than to call an address kept in a register.
6032 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
6033 Define this macro if a non-short-circuit operation produced by
6034 @samp{fold_range_test ()} is optimal. This macro defaults to true if
6035 @code{BRANCH_COST} is greater than or equal to the value 2.
6038 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total})
6039 This target hook describes the relative costs of RTL expressions.
6041 The cost may depend on the precise form of the expression, which is
6042 available for examination in @var{x}, and the rtx code of the expression
6043 in which it is contained, found in @var{outer_code}. @var{code} is the
6044 expression code---redundant, since it can be obtained with
6045 @code{GET_CODE (@var{x})}.
6047 In implementing this hook, you can use the construct
6048 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6051 On entry to the hook, @code{*@var{total}} contains a default estimate
6052 for the cost of the expression. The hook should modify this value as
6053 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
6054 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
6055 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
6057 When optimizing for code size, i.e.@: when @code{optimize_size} is
6058 nonzero, this target hook should be used to estimate the relative
6059 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6061 The hook returns true when all subexpressions of @var{x} have been
6062 processed, and false when @code{rtx_cost} should recurse.
6065 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address})
6066 This hook computes the cost of an addressing mode that contains
6067 @var{address}. If not defined, the cost is computed from
6068 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
6070 For most CISC machines, the default cost is a good approximation of the
6071 true cost of the addressing mode. However, on RISC machines, all
6072 instructions normally have the same length and execution time. Hence
6073 all addresses will have equal costs.
6075 In cases where more than one form of an address is known, the form with
6076 the lowest cost will be used. If multiple forms have the same, lowest,
6077 cost, the one that is the most complex will be used.
6079 For example, suppose an address that is equal to the sum of a register
6080 and a constant is used twice in the same basic block. When this macro
6081 is not defined, the address will be computed in a register and memory
6082 references will be indirect through that register. On machines where
6083 the cost of the addressing mode containing the sum is no higher than
6084 that of a simple indirect reference, this will produce an additional
6085 instruction and possibly require an additional register. Proper
6086 specification of this macro eliminates this overhead for such machines.
6088 This hook is never called with an invalid address.
6090 On machines where an address involving more than one register is as
6091 cheap as an address computation involving only one register, defining
6092 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
6093 be live over a region of code where only one would have been if
6094 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
6095 should be considered in the definition of this macro. Equivalent costs
6096 should probably only be given to addresses with different numbers of
6097 registers on machines with lots of registers.
6101 @section Adjusting the Instruction Scheduler
6103 The instruction scheduler may need a fair amount of machine-specific
6104 adjustment in order to produce good code. GCC provides several target
6105 hooks for this purpose. It is usually enough to define just a few of
6106 them: try the first ones in this list first.
6108 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
6109 This hook returns the maximum number of instructions that can ever
6110 issue at the same time on the target machine. The default is one.
6111 Although the insn scheduler can define itself the possibility of issue
6112 an insn on the same cycle, the value can serve as an additional
6113 constraint to issue insns on the same simulated processor cycle (see
6114 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
6115 This value must be constant over the entire compilation. If you need
6116 it to vary depending on what the instructions are, you must use
6117 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
6120 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
6121 This hook is executed by the scheduler after it has scheduled an insn
6122 from the ready list. It should return the number of insns which can
6123 still be issued in the current cycle. The default is
6124 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
6125 @code{USE}, which normally are not counted against the issue rate.
6126 You should define this hook if some insns take more machine resources
6127 than others, so that fewer insns can follow them in the same cycle.
6128 @var{file} is either a null pointer, or a stdio stream to write any
6129 debug output to. @var{verbose} is the verbose level provided by
6130 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
6134 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
6135 This function corrects the value of @var{cost} based on the
6136 relationship between @var{insn} and @var{dep_insn} through the
6137 dependence @var{link}. It should return the new value. The default
6138 is to make no adjustment to @var{cost}. This can be used for example
6139 to specify to the scheduler using the traditional pipeline description
6140 that an output- or anti-dependence does not incur the same cost as a
6141 data-dependence. If the scheduler using the automaton based pipeline
6142 description, the cost of anti-dependence is zero and the cost of
6143 output-dependence is maximum of one and the difference of latency
6144 times of the first and the second insns. If these values are not
6145 acceptable, you could use the hook to modify them too. See also
6146 @pxref{Processor pipeline description}.
6149 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
6150 This hook adjusts the integer scheduling priority @var{priority} of
6151 @var{insn}. It should return the new priority. Increase the priority to
6152 execute @var{insn} earlier, reduce the priority to execute @var{insn}
6153 later. Do not define this hook if you do not need to adjust the
6154 scheduling priorities of insns.
6157 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
6158 This hook is executed by the scheduler after it has scheduled the ready
6159 list, to allow the machine description to reorder it (for example to
6160 combine two small instructions together on @samp{VLIW} machines).
6161 @var{file} is either a null pointer, or a stdio stream to write any
6162 debug output to. @var{verbose} is the verbose level provided by
6163 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
6164 list of instructions that are ready to be scheduled. @var{n_readyp} is
6165 a pointer to the number of elements in the ready list. The scheduler
6166 reads the ready list in reverse order, starting with
6167 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
6168 is the timer tick of the scheduler. You may modify the ready list and
6169 the number of ready insns. The return value is the number of insns that
6170 can issue this cycle; normally this is just @code{issue_rate}. See also
6171 @samp{TARGET_SCHED_REORDER2}.
6174 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
6175 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
6176 function is called whenever the scheduler starts a new cycle. This one
6177 is called once per iteration over a cycle, immediately after
6178 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
6179 return the number of insns to be scheduled in the same cycle. Defining
6180 this hook can be useful if there are frequent situations where
6181 scheduling one insn causes other insns to become ready in the same
6182 cycle. These other insns can then be taken into account properly.
6185 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
6186 This hook is called after evaluation forward dependencies of insns in
6187 chain given by two parameter values (@var{head} and @var{tail}
6188 correspondingly) but before insns scheduling of the insn chain. For
6189 example, it can be used for better insn classification if it requires
6190 analysis of dependencies. This hook can use backward and forward
6191 dependencies of the insn scheduler because they are already
6195 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
6196 This hook is executed by the scheduler at the beginning of each block of
6197 instructions that are to be scheduled. @var{file} is either a null
6198 pointer, or a stdio stream to write any debug output to. @var{verbose}
6199 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6200 @var{max_ready} is the maximum number of insns in the current scheduling
6201 region that can be live at the same time. This can be used to allocate
6202 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
6205 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
6206 This hook is executed by the scheduler at the end of each block of
6207 instructions that are to be scheduled. It can be used to perform
6208 cleanup of any actions done by the other scheduling hooks. @var{file}
6209 is either a null pointer, or a stdio stream to write any debug output
6210 to. @var{verbose} is the verbose level provided by
6211 @option{-fsched-verbose-@var{n}}.
6214 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
6215 This hook is executed by the scheduler after function level initializations.
6216 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6217 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6218 @var{old_max_uid} is the maximum insn uid when scheduling begins.
6221 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
6222 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
6223 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6224 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6227 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
6228 The hook returns an RTL insn. The automaton state used in the
6229 pipeline hazard recognizer is changed as if the insn were scheduled
6230 when the new simulated processor cycle starts. Usage of the hook may
6231 simplify the automaton pipeline description for some @acronym{VLIW}
6232 processors. If the hook is defined, it is used only for the automaton
6233 based pipeline description. The default is not to change the state
6234 when the new simulated processor cycle starts.
6237 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
6238 The hook can be used to initialize data used by the previous hook.
6241 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
6242 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6243 to changed the state as if the insn were scheduled when the new
6244 simulated processor cycle finishes.
6247 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
6248 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
6249 used to initialize data used by the previous hook.
6252 @deftypefn {Target Hook} void TARGET_SCHED_DFA_PRE_CYCLE_ADVANCE (void)
6253 The hook to notify target that the current simulated cycle is about to finish.
6254 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6255 to change the state in more complicated situations - e.g., when advancing
6256 state on a single insn is not enough.
6259 @deftypefn {Target Hook} void TARGET_SCHED_DFA_POST_CYCLE_ADVANCE (void)
6260 The hook to notify target that new simulated cycle has just started.
6261 The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used
6262 to change the state in more complicated situations - e.g., when advancing
6263 state on a single insn is not enough.
6266 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
6267 This hook controls better choosing an insn from the ready insn queue
6268 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
6269 chooses the first insn from the queue. If the hook returns a positive
6270 value, an additional scheduler code tries all permutations of
6271 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
6272 subsequent ready insns to choose an insn whose issue will result in
6273 maximal number of issued insns on the same cycle. For the
6274 @acronym{VLIW} processor, the code could actually solve the problem of
6275 packing simple insns into the @acronym{VLIW} insn. Of course, if the
6276 rules of @acronym{VLIW} packing are described in the automaton.
6278 This code also could be used for superscalar @acronym{RISC}
6279 processors. Let us consider a superscalar @acronym{RISC} processor
6280 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
6281 @var{B}, some insns can be executed only in pipelines @var{B} or
6282 @var{C}, and one insn can be executed in pipeline @var{B}. The
6283 processor may issue the 1st insn into @var{A} and the 2nd one into
6284 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
6285 until the next cycle. If the scheduler issues the 3rd insn the first,
6286 the processor could issue all 3 insns per cycle.
6288 Actually this code demonstrates advantages of the automaton based
6289 pipeline hazard recognizer. We try quickly and easy many insn
6290 schedules to choose the best one.
6292 The default is no multipass scheduling.
6295 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
6297 This hook controls what insns from the ready insn queue will be
6298 considered for the multipass insn scheduling. If the hook returns
6299 zero for insn passed as the parameter, the insn will be not chosen to
6302 The default is that any ready insns can be chosen to be issued.
6305 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
6307 This hook is called by the insn scheduler before issuing insn passed
6308 as the third parameter on given cycle. If the hook returns nonzero,
6309 the insn is not issued on given processors cycle. Instead of that,
6310 the processor cycle is advanced. If the value passed through the last
6311 parameter is zero, the insn ready queue is not sorted on the new cycle
6312 start as usually. The first parameter passes file for debugging
6313 output. The second one passes the scheduler verbose level of the
6314 debugging output. The forth and the fifth parameter values are
6315 correspondingly processor cycle on which the previous insn has been
6316 issued and the current processor cycle.
6319 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct dep_def *@var{_dep}, int @var{cost}, int @var{distance})
6320 This hook is used to define which dependences are considered costly by
6321 the target, so costly that it is not advisable to schedule the insns that
6322 are involved in the dependence too close to one another. The parameters
6323 to this hook are as follows: The first parameter @var{_dep} is the dependence
6324 being evaluated. The second parameter @var{cost} is the cost of the
6325 dependence, and the third
6326 parameter @var{distance} is the distance in cycles between the two insns.
6327 The hook returns @code{true} if considering the distance between the two
6328 insns the dependence between them is considered costly by the target,
6329 and @code{false} otherwise.
6331 Defining this hook can be useful in multiple-issue out-of-order machines,
6332 where (a) it's practically hopeless to predict the actual data/resource
6333 delays, however: (b) there's a better chance to predict the actual grouping
6334 that will be formed, and (c) correctly emulating the grouping can be very
6335 important. In such targets one may want to allow issuing dependent insns
6336 closer to one another---i.e., closer than the dependence distance; however,
6337 not in cases of "costly dependences", which this hooks allows to define.
6340 @deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void)
6341 This hook is called by the insn scheduler after emitting a new instruction to
6342 the instruction stream. The hook notifies a target backend to extend its
6343 per instruction data structures.
6346 @deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx @var{insn}, int @var{request}, rtx *@var{new_pat})
6347 This hook is called by the insn scheduler when @var{insn} has only
6348 speculative dependencies and therefore can be scheduled speculatively.
6349 The hook is used to check if the pattern of @var{insn} has a speculative
6350 version and, in case of successful check, to generate that speculative
6351 pattern. The hook should return 1, if the instruction has a speculative form,
6352 or -1, if it doesn't. @var{request} describes the type of requested
6353 speculation. If the return value equals 1 then @var{new_pat} is assigned
6354 the generated speculative pattern.
6357 @deftypefn {Target Hook} int TARGET_SCHED_NEEDS_BLOCK_P (rtx @var{insn})
6358 This hook is called by the insn scheduler during generation of recovery code
6359 for @var{insn}. It should return nonzero, if the corresponding check
6360 instruction should branch to recovery code, or zero otherwise.
6363 @deftypefn {Target Hook} rtx TARGET_SCHED_GEN_CHECK (rtx @var{insn}, rtx @var{label}, int @var{mutate_p})
6364 This hook is called by the insn scheduler to generate a pattern for recovery
6365 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
6366 speculative instruction for which the check should be generated.
6367 @var{label} is either a label of a basic block, where recovery code should
6368 be emitted, or a null pointer, when requested check doesn't branch to
6369 recovery code (a simple check). If @var{mutate_p} is nonzero, then
6370 a pattern for a branchy check corresponding to a simple check denoted by
6371 @var{insn} should be generated. In this case @var{label} can't be null.
6374 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC (rtx @var{insn})
6375 This hook is used as a workaround for
6376 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD} not being
6377 called on the first instruction of the ready list. The hook is used to
6378 discard speculative instruction that stand first in the ready list from
6379 being scheduled on the current cycle. For non-speculative instructions,
6380 the hook should always return nonzero. For example, in the ia64 backend
6381 the hook is used to cancel data speculative insns when the ALAT table
6385 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (unsigned int *@var{flags}, spec_info_t @var{spec_info})
6386 This hook is used by the insn scheduler to find out what features should be
6387 enabled/used. @var{flags} initially may have either the SCHED_RGN or SCHED_EBB
6388 bit set. This denotes the scheduler pass for which the data should be
6389 provided. The target backend should modify @var{flags} by modifying
6390 the bits corresponding to the following features: USE_DEPS_LIST, USE_GLAT,
6391 DETACH_LIFE_INFO, and DO_SPECULATION@. For the DO_SPECULATION feature
6392 an additional structure @var{spec_info} should be filled by the target.
6393 The structure describes speculation types that can be used in the scheduler.
6396 @deftypefn {Target Hook} int TARGET_SCHED_SMS_RES_MII (struct ddg *@var{g})
6397 This hook is called by the swing modulo scheduler to calculate a
6398 resource-based lower bound which is based on the resources available in
6399 the machine and the resources required by each instruction. The target
6400 backend can use @var{g} to calculate such bound. A very simple lower
6401 bound will be used in case this hook is not implemented: the total number
6402 of instructions divided by the issue rate.
6406 @section Dividing the Output into Sections (Texts, Data, @dots{})
6407 @c the above section title is WAY too long. maybe cut the part between
6408 @c the (...)? --mew 10feb93
6410 An object file is divided into sections containing different types of
6411 data. In the most common case, there are three sections: the @dfn{text
6412 section}, which holds instructions and read-only data; the @dfn{data
6413 section}, which holds initialized writable data; and the @dfn{bss
6414 section}, which holds uninitialized data. Some systems have other kinds
6417 @file{varasm.c} provides several well-known sections, such as
6418 @code{text_section}, @code{data_section} and @code{bss_section}.
6419 The normal way of controlling a @code{@var{foo}_section} variable
6420 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
6421 as described below. The macros are only read once, when @file{varasm.c}
6422 initializes itself, so their values must be run-time constants.
6423 They may however depend on command-line flags.
6425 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
6426 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
6427 to be string literals.
6429 Some assemblers require a different string to be written every time a
6430 section is selected. If your assembler falls into this category, you
6431 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
6432 @code{get_unnamed_section} to set up the sections.
6434 You must always create a @code{text_section}, either by defining
6435 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
6436 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
6437 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
6438 create a distinct @code{readonly_data_section}, the default is to
6439 reuse @code{text_section}.
6441 All the other @file{varasm.c} sections are optional, and are null
6442 if the target does not provide them.
6444 @defmac TEXT_SECTION_ASM_OP
6445 A C expression whose value is a string, including spacing, containing the
6446 assembler operation that should precede instructions and read-only data.
6447 Normally @code{"\t.text"} is right.
6450 @defmac HOT_TEXT_SECTION_NAME
6451 If defined, a C string constant for the name of the section containing most
6452 frequently executed functions of the program. If not defined, GCC will provide
6453 a default definition if the target supports named sections.
6456 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
6457 If defined, a C string constant for the name of the section containing unlikely
6458 executed functions in the program.
6461 @defmac DATA_SECTION_ASM_OP
6462 A C expression whose value is a string, including spacing, containing the
6463 assembler operation to identify the following data as writable initialized
6464 data. Normally @code{"\t.data"} is right.
6467 @defmac SDATA_SECTION_ASM_OP
6468 If defined, a C expression whose value is a string, including spacing,
6469 containing the assembler operation to identify the following data as
6470 initialized, writable small data.
6473 @defmac READONLY_DATA_SECTION_ASM_OP
6474 A C expression whose value is a string, including spacing, containing the
6475 assembler operation to identify the following data as read-only initialized
6479 @defmac BSS_SECTION_ASM_OP
6480 If defined, a C expression whose value is a string, including spacing,
6481 containing the assembler operation to identify the following data as
6482 uninitialized global data. If not defined, and neither
6483 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
6484 uninitialized global data will be output in the data section if
6485 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
6489 @defmac SBSS_SECTION_ASM_OP
6490 If defined, a C expression whose value is a string, including spacing,
6491 containing the assembler operation to identify the following data as
6492 uninitialized, writable small data.
6495 @defmac INIT_SECTION_ASM_OP
6496 If defined, a C expression whose value is a string, including spacing,
6497 containing the assembler operation to identify the following data as
6498 initialization code. If not defined, GCC will assume such a section does
6499 not exist. This section has no corresponding @code{init_section}
6500 variable; it is used entirely in runtime code.
6503 @defmac FINI_SECTION_ASM_OP
6504 If defined, a C expression whose value is a string, including spacing,
6505 containing the assembler operation to identify the following data as
6506 finalization code. If not defined, GCC will assume such a section does
6507 not exist. This section has no corresponding @code{fini_section}
6508 variable; it is used entirely in runtime code.
6511 @defmac INIT_ARRAY_SECTION_ASM_OP
6512 If defined, a C expression whose value is a string, including spacing,
6513 containing the assembler operation to identify the following data as
6514 part of the @code{.init_array} (or equivalent) section. If not
6515 defined, GCC will assume such a section does not exist. Do not define
6516 both this macro and @code{INIT_SECTION_ASM_OP}.
6519 @defmac FINI_ARRAY_SECTION_ASM_OP
6520 If defined, a C expression whose value is a string, including spacing,
6521 containing the assembler operation to identify the following data as
6522 part of the @code{.fini_array} (or equivalent) section. If not
6523 defined, GCC will assume such a section does not exist. Do not define
6524 both this macro and @code{FINI_SECTION_ASM_OP}.
6527 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
6528 If defined, an ASM statement that switches to a different section
6529 via @var{section_op}, calls @var{function}, and switches back to
6530 the text section. This is used in @file{crtstuff.c} if
6531 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
6532 to initialization and finalization functions from the init and fini
6533 sections. By default, this macro uses a simple function call. Some
6534 ports need hand-crafted assembly code to avoid dependencies on
6535 registers initialized in the function prologue or to ensure that
6536 constant pools don't end up too far way in the text section.
6539 @defmac TARGET_LIBGCC_SDATA_SECTION
6540 If defined, a string which names the section into which small
6541 variables defined in crtstuff and libgcc should go. This is useful
6542 when the target has options for optimizing access to small data, and
6543 you want the crtstuff and libgcc routines to be conservative in what
6544 they expect of your application yet liberal in what your application
6545 expects. For example, for targets with a @code{.sdata} section (like
6546 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
6547 require small data support from your application, but use this macro
6548 to put small data into @code{.sdata} so that your application can
6549 access these variables whether it uses small data or not.
6552 @defmac FORCE_CODE_SECTION_ALIGN
6553 If defined, an ASM statement that aligns a code section to some
6554 arbitrary boundary. This is used to force all fragments of the
6555 @code{.init} and @code{.fini} sections to have to same alignment
6556 and thus prevent the linker from having to add any padding.
6559 @defmac JUMP_TABLES_IN_TEXT_SECTION
6560 Define this macro to be an expression with a nonzero value if jump
6561 tables (for @code{tablejump} insns) should be output in the text
6562 section, along with the assembler instructions. Otherwise, the
6563 readonly data section is used.
6565 This macro is irrelevant if there is no separate readonly data section.
6568 @deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void)
6569 Define this hook if you need to do something special to set up the
6570 @file{varasm.c} sections, or if your target has some special sections
6571 of its own that you need to create.
6573 GCC calls this hook after processing the command line, but before writing
6574 any assembly code, and before calling any of the section-returning hooks
6578 @deftypefn {Target Hook} TARGET_ASM_RELOC_RW_MASK (void)
6579 Return a mask describing how relocations should be treated when
6580 selecting sections. Bit 1 should be set if global relocations
6581 should be placed in a read-write section; bit 0 should be set if
6582 local relocations should be placed in a read-write section.
6584 The default version of this function returns 3 when @option{-fpic}
6585 is in effect, and 0 otherwise. The hook is typically redefined
6586 when the target cannot support (some kinds of) dynamic relocations
6587 in read-only sections even in executables.
6590 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
6591 Return the section into which @var{exp} should be placed. You can
6592 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
6593 some sort. @var{reloc} indicates whether the initial value of @var{exp}
6594 requires link-time relocations. Bit 0 is set when variable contains
6595 local relocations only, while bit 1 is set for global relocations.
6596 @var{align} is the constant alignment in bits.
6598 The default version of this function takes care of putting read-only
6599 variables in @code{readonly_data_section}.
6601 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
6604 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
6605 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
6606 for @code{FUNCTION_DECL}s as well as for variables and constants.
6608 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
6609 function has been determined to be likely to be called, and nonzero if
6610 it is unlikely to be called.
6613 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
6614 Build up a unique section name, expressed as a @code{STRING_CST} node,
6615 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
6616 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
6617 the initial value of @var{exp} requires link-time relocations.
6619 The default version of this function appends the symbol name to the
6620 ELF section name that would normally be used for the symbol. For
6621 example, the function @code{foo} would be placed in @code{.text.foo}.
6622 Whatever the actual target object format, this is often good enough.
6625 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
6626 Return the readonly data section associated with
6627 @samp{DECL_SECTION_NAME (@var{decl})}.
6628 The default version of this function selects @code{.gnu.linkonce.r.name} if
6629 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
6630 if function is in @code{.text.name}, and the normal readonly-data section
6634 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
6635 Return the section into which a constant @var{x}, of mode @var{mode},
6636 should be placed. You can assume that @var{x} is some kind of
6637 constant in RTL@. The argument @var{mode} is redundant except in the
6638 case of a @code{const_int} rtx. @var{align} is the constant alignment
6641 The default version of this function takes care of putting symbolic
6642 constants in @code{flag_pic} mode in @code{data_section} and everything
6643 else in @code{readonly_data_section}.
6646 @deftypefn {Target Hook} void TARGET_MANGLE_DECL_ASSEMBLER_NAME (tree @var{decl}, tree @var{id})
6647 Define this hook if you need to postprocess the assembler name generated
6648 by target-independent code. The @var{id} provided to this hook will be
6649 the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C,
6650 or the mangled name of the @var{decl} in C++). The return value of the
6651 hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on
6652 your target system. The default implementation of this hook just
6653 returns the @var{id} provided.
6656 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
6657 Define this hook if references to a symbol or a constant must be
6658 treated differently depending on something about the variable or
6659 function named by the symbol (such as what section it is in).
6661 The hook is executed immediately after rtl has been created for
6662 @var{decl}, which may be a variable or function declaration or
6663 an entry in the constant pool. In either case, @var{rtl} is the
6664 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
6665 in this hook; that field may not have been initialized yet.
6667 In the case of a constant, it is safe to assume that the rtl is
6668 a @code{mem} whose address is a @code{symbol_ref}. Most decls
6669 will also have this form, but that is not guaranteed. Global
6670 register variables, for instance, will have a @code{reg} for their
6671 rtl. (Normally the right thing to do with such unusual rtl is
6674 The @var{new_decl_p} argument will be true if this is the first time
6675 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
6676 be false for subsequent invocations, which will happen for duplicate
6677 declarations. Whether or not anything must be done for the duplicate
6678 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
6679 @var{new_decl_p} is always true when the hook is called for a constant.
6681 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
6682 The usual thing for this hook to do is to record flags in the
6683 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
6684 Historically, the name string was modified if it was necessary to
6685 encode more than one bit of information, but this practice is now
6686 discouraged; use @code{SYMBOL_REF_FLAGS}.
6688 The default definition of this hook, @code{default_encode_section_info}
6689 in @file{varasm.c}, sets a number of commonly-useful bits in
6690 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
6691 before overriding it.
6694 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
6695 Decode @var{name} and return the real name part, sans
6696 the characters that @code{TARGET_ENCODE_SECTION_INFO}
6700 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
6701 Returns true if @var{exp} should be placed into a ``small data'' section.
6702 The default version of this hook always returns false.
6705 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
6706 Contains the value true if the target places read-only
6707 ``small data'' into a separate section. The default value is false.
6710 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
6711 Returns true if @var{exp} names an object for which name resolution
6712 rules must resolve to the current ``module'' (dynamic shared library
6713 or executable image).
6715 The default version of this hook implements the name resolution rules
6716 for ELF, which has a looser model of global name binding than other
6717 currently supported object file formats.
6720 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
6721 Contains the value true if the target supports thread-local storage.
6722 The default value is false.
6727 @section Position Independent Code
6728 @cindex position independent code
6731 This section describes macros that help implement generation of position
6732 independent code. Simply defining these macros is not enough to
6733 generate valid PIC; you must also add support to the macros
6734 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
6735 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
6736 @samp{movsi} to do something appropriate when the source operand
6737 contains a symbolic address. You may also need to alter the handling of
6738 switch statements so that they use relative addresses.
6739 @c i rearranged the order of the macros above to try to force one of
6740 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
6742 @defmac PIC_OFFSET_TABLE_REGNUM
6743 The register number of the register used to address a table of static
6744 data addresses in memory. In some cases this register is defined by a
6745 processor's ``application binary interface'' (ABI)@. When this macro
6746 is defined, RTL is generated for this register once, as with the stack
6747 pointer and frame pointer registers. If this macro is not defined, it
6748 is up to the machine-dependent files to allocate such a register (if
6749 necessary). Note that this register must be fixed when in use (e.g.@:
6750 when @code{flag_pic} is true).
6753 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6754 Define this macro if the register defined by
6755 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
6756 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
6759 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
6760 A C expression that is nonzero if @var{x} is a legitimate immediate
6761 operand on the target machine when generating position independent code.
6762 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6763 check this. You can also assume @var{flag_pic} is true, so you need not
6764 check it either. You need not define this macro if all constants
6765 (including @code{SYMBOL_REF}) can be immediate operands when generating
6766 position independent code.
6769 @node Assembler Format
6770 @section Defining the Output Assembler Language
6772 This section describes macros whose principal purpose is to describe how
6773 to write instructions in assembler language---rather than what the
6777 * File Framework:: Structural information for the assembler file.
6778 * Data Output:: Output of constants (numbers, strings, addresses).
6779 * Uninitialized Data:: Output of uninitialized variables.
6780 * Label Output:: Output and generation of labels.
6781 * Initialization:: General principles of initialization
6782 and termination routines.
6783 * Macros for Initialization::
6784 Specific macros that control the handling of
6785 initialization and termination routines.
6786 * Instruction Output:: Output of actual instructions.
6787 * Dispatch Tables:: Output of jump tables.
6788 * Exception Region Output:: Output of exception region code.
6789 * Alignment Output:: Pseudo ops for alignment and skipping data.
6792 @node File Framework
6793 @subsection The Overall Framework of an Assembler File
6794 @cindex assembler format
6795 @cindex output of assembler code
6797 @c prevent bad page break with this line
6798 This describes the overall framework of an assembly file.
6800 @deftypefn {Target Hook} void TARGET_ASM_FILE_START ()
6801 @findex default_file_start
6802 Output to @code{asm_out_file} any text which the assembler expects to
6803 find at the beginning of a file. The default behavior is controlled
6804 by two flags, documented below. Unless your target's assembler is
6805 quite unusual, if you override the default, you should call
6806 @code{default_file_start} at some point in your target hook. This
6807 lets other target files rely on these variables.
6810 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
6811 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
6812 printed as the very first line in the assembly file, unless
6813 @option{-fverbose-asm} is in effect. (If that macro has been defined
6814 to the empty string, this variable has no effect.) With the normal
6815 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
6816 assembler that it need not bother stripping comments or extra
6817 whitespace from its input. This allows it to work a bit faster.
6819 The default is false. You should not set it to true unless you have
6820 verified that your port does not generate any extra whitespace or
6821 comments that will cause GAS to issue errors in NO_APP mode.
6824 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
6825 If this flag is true, @code{output_file_directive} will be called
6826 for the primary source file, immediately after printing
6827 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
6828 this to be done. The default is false.
6831 @deftypefn {Target Hook} void TARGET_ASM_FILE_END ()
6832 Output to @code{asm_out_file} any text which the assembler expects
6833 to find at the end of a file. The default is to output nothing.
6836 @deftypefun void file_end_indicate_exec_stack ()
6837 Some systems use a common convention, the @samp{.note.GNU-stack}
6838 special section, to indicate whether or not an object file relies on
6839 the stack being executable. If your system uses this convention, you
6840 should define @code{TARGET_ASM_FILE_END} to this function. If you
6841 need to do other things in that hook, have your hook function call
6845 @defmac ASM_COMMENT_START
6846 A C string constant describing how to begin a comment in the target
6847 assembler language. The compiler assumes that the comment will end at
6848 the end of the line.
6852 A C string constant for text to be output before each @code{asm}
6853 statement or group of consecutive ones. Normally this is
6854 @code{"#APP"}, which is a comment that has no effect on most
6855 assemblers but tells the GNU assembler that it must check the lines
6856 that follow for all valid assembler constructs.
6860 A C string constant for text to be output after each @code{asm}
6861 statement or group of consecutive ones. Normally this is
6862 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6863 time-saving assumptions that are valid for ordinary compiler output.
6866 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6867 A C statement to output COFF information or DWARF debugging information
6868 which indicates that filename @var{name} is the current source file to
6869 the stdio stream @var{stream}.
6871 This macro need not be defined if the standard form of output
6872 for the file format in use is appropriate.
6875 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6876 A C statement to output the string @var{string} to the stdio stream
6877 @var{stream}. If you do not call the function @code{output_quoted_string}
6878 in your config files, GCC will only call it to output filenames to
6879 the assembler source. So you can use it to canonicalize the format
6880 of the filename using this macro.
6883 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6884 A C statement to output something to the assembler file to handle a
6885 @samp{#ident} directive containing the text @var{string}. If this
6886 macro is not defined, nothing is output for a @samp{#ident} directive.
6889 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6890 Output assembly directives to switch to section @var{name}. The section
6891 should have attributes as specified by @var{flags}, which is a bit mask
6892 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6893 is nonzero, it contains an alignment in bytes to be used for the section,
6894 otherwise some target default should be used. Only targets that must
6895 specify an alignment within the section directive need pay attention to
6896 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6899 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6900 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6903 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
6904 @deftypefn {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
6905 This flag is true if we can create zeroed data by switching to a BSS
6906 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
6907 This is true on most ELF targets.
6910 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6911 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6912 based on a variable or function decl, a section name, and whether or not the
6913 declaration's initializer may contain runtime relocations. @var{decl} may be
6914 null, in which case read-write data should be assumed.
6916 The default version of this function handles choosing code vs data,
6917 read-only vs read-write data, and @code{flag_pic}. You should only
6918 need to override this if your target has special flags that might be
6919 set via @code{__attribute__}.
6922 @deftypefn {Target Hook} {int} TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type @var{type}, const char * @var{text})
6923 Provides the target with the ability to record the gcc command line
6924 switches that have been passed to the compiler, and options that are
6925 enabled. The @var{type} argument specifies what is being recorded.
6926 It can take the following values:
6929 @item SWITCH_TYPE_PASSED
6930 @var{text} is a command line switch that has been set by the user.
6932 @item SWITCH_TYPE_ENABLED
6933 @var{text} is an option which has been enabled. This might be as a
6934 direct result of a command line switch, or because it is enabled by
6935 default or because it has been enabled as a side effect of a different
6936 command line switch. For example, the @option{-O2} switch enables
6937 various different individual optimization passes.
6939 @item SWITCH_TYPE_DESCRIPTIVE
6940 @var{text} is either NULL or some descriptive text which should be
6941 ignored. If @var{text} is NULL then it is being used to warn the
6942 target hook that either recording is starting or ending. The first
6943 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
6944 warning is for start up and the second time the warning is for
6945 wind down. This feature is to allow the target hook to make any
6946 necessary preparations before it starts to record switches and to
6947 perform any necessary tidying up after it has finished recording
6950 @item SWITCH_TYPE_LINE_START
6951 This option can be ignored by this target hook.
6953 @item SWITCH_TYPE_LINE_END
6954 This option can be ignored by this target hook.
6957 The hook's return value must be zero. Other return values may be
6958 supported in the future.
6960 By default this hook is set to NULL, but an example implementation is
6961 provided for ELF based targets. Called @var{elf_record_gcc_switches},
6962 it records the switches as ASCII text inside a new, string mergeable
6963 section in the assembler output file. The name of the new section is
6964 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
6968 @deftypefn {Target Hook} {const char *} TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
6969 This is the name of the section that will be created by the example
6970 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
6976 @subsection Output of Data
6979 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6980 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6981 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6982 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6983 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6984 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6985 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6986 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6987 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6988 These hooks specify assembly directives for creating certain kinds
6989 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6990 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6991 aligned two-byte object, and so on. Any of the hooks may be
6992 @code{NULL}, indicating that no suitable directive is available.
6994 The compiler will print these strings at the start of a new line,
6995 followed immediately by the object's initial value. In most cases,
6996 the string should contain a tab, a pseudo-op, and then another tab.
6999 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
7000 The @code{assemble_integer} function uses this hook to output an
7001 integer object. @var{x} is the object's value, @var{size} is its size
7002 in bytes and @var{aligned_p} indicates whether it is aligned. The
7003 function should return @code{true} if it was able to output the
7004 object. If it returns false, @code{assemble_integer} will try to
7005 split the object into smaller parts.
7007 The default implementation of this hook will use the
7008 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
7009 when the relevant string is @code{NULL}.
7012 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
7013 A C statement to recognize @var{rtx} patterns that
7014 @code{output_addr_const} can't deal with, and output assembly code to
7015 @var{stream} corresponding to the pattern @var{x}. This may be used to
7016 allow machine-dependent @code{UNSPEC}s to appear within constants.
7018 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
7019 @code{goto fail}, so that a standard error message is printed. If it
7020 prints an error message itself, by calling, for example,
7021 @code{output_operand_lossage}, it may just complete normally.
7024 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
7025 A C statement to output to the stdio stream @var{stream} an assembler
7026 instruction to assemble a string constant containing the @var{len}
7027 bytes at @var{ptr}. @var{ptr} will be a C expression of type
7028 @code{char *} and @var{len} a C expression of type @code{int}.
7030 If the assembler has a @code{.ascii} pseudo-op as found in the
7031 Berkeley Unix assembler, do not define the macro
7032 @code{ASM_OUTPUT_ASCII}.
7035 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
7036 A C statement to output word @var{n} of a function descriptor for
7037 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
7038 is defined, and is otherwise unused.
7041 @defmac CONSTANT_POOL_BEFORE_FUNCTION
7042 You may define this macro as a C expression. You should define the
7043 expression to have a nonzero value if GCC should output the constant
7044 pool for a function before the code for the function, or a zero value if
7045 GCC should output the constant pool after the function. If you do
7046 not define this macro, the usual case, GCC will output the constant
7047 pool before the function.
7050 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
7051 A C statement to output assembler commands to define the start of the
7052 constant pool for a function. @var{funname} is a string giving
7053 the name of the function. Should the return type of the function
7054 be required, it can be obtained via @var{fundecl}. @var{size}
7055 is the size, in bytes, of the constant pool that will be written
7056 immediately after this call.
7058 If no constant-pool prefix is required, the usual case, this macro need
7062 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
7063 A C statement (with or without semicolon) to output a constant in the
7064 constant pool, if it needs special treatment. (This macro need not do
7065 anything for RTL expressions that can be output normally.)
7067 The argument @var{file} is the standard I/O stream to output the
7068 assembler code on. @var{x} is the RTL expression for the constant to
7069 output, and @var{mode} is the machine mode (in case @var{x} is a
7070 @samp{const_int}). @var{align} is the required alignment for the value
7071 @var{x}; you should output an assembler directive to force this much
7074 The argument @var{labelno} is a number to use in an internal label for
7075 the address of this pool entry. The definition of this macro is
7076 responsible for outputting the label definition at the proper place.
7077 Here is how to do this:
7080 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
7083 When you output a pool entry specially, you should end with a
7084 @code{goto} to the label @var{jumpto}. This will prevent the same pool
7085 entry from being output a second time in the usual manner.
7087 You need not define this macro if it would do nothing.
7090 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
7091 A C statement to output assembler commands to at the end of the constant
7092 pool for a function. @var{funname} is a string giving the name of the
7093 function. Should the return type of the function be required, you can
7094 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
7095 constant pool that GCC wrote immediately before this call.
7097 If no constant-pool epilogue is required, the usual case, you need not
7101 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
7102 Define this macro as a C expression which is nonzero if @var{C} is
7103 used as a logical line separator by the assembler. @var{STR} points
7104 to the position in the string where @var{C} was found; this can be used if
7105 a line separator uses multiple characters.
7107 If you do not define this macro, the default is that only
7108 the character @samp{;} is treated as a logical line separator.
7111 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
7112 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
7113 These target hooks are C string constants, describing the syntax in the
7114 assembler for grouping arithmetic expressions. If not overridden, they
7115 default to normal parentheses, which is correct for most assemblers.
7118 These macros are provided by @file{real.h} for writing the definitions
7119 of @code{ASM_OUTPUT_DOUBLE} and the like:
7121 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
7122 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
7123 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
7124 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
7125 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
7126 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
7127 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
7128 target's floating point representation, and store its bit pattern in
7129 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
7130 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
7131 simple @code{long int}. For the others, it should be an array of
7132 @code{long int}. The number of elements in this array is determined
7133 by the size of the desired target floating point data type: 32 bits of
7134 it go in each @code{long int} array element. Each array element holds
7135 32 bits of the result, even if @code{long int} is wider than 32 bits
7136 on the host machine.
7138 The array element values are designed so that you can print them out
7139 using @code{fprintf} in the order they should appear in the target
7143 @node Uninitialized Data
7144 @subsection Output of Uninitialized Variables
7146 Each of the macros in this section is used to do the whole job of
7147 outputting a single uninitialized variable.
7149 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
7150 A C statement (sans semicolon) to output to the stdio stream
7151 @var{stream} the assembler definition of a common-label named
7152 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7153 is the size rounded up to whatever alignment the caller wants.
7155 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7156 output the name itself; before and after that, output the additional
7157 assembler syntax for defining the name, and a newline.
7159 This macro controls how the assembler definitions of uninitialized
7160 common global variables are output.
7163 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
7164 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
7165 separate, explicit argument. If you define this macro, it is used in
7166 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
7167 handling the required alignment of the variable. The alignment is specified
7168 as the number of bits.
7171 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7172 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
7173 variable to be output, if there is one, or @code{NULL_TREE} if there
7174 is no corresponding variable. If you define this macro, GCC will use it
7175 in place of both @code{ASM_OUTPUT_COMMON} and
7176 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
7177 the variable's decl in order to chose what to output.
7180 @defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
7181 A C statement (sans semicolon) to output to the stdio stream
7182 @var{stream} the assembler definition of uninitialized global @var{decl} named
7183 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7184 is the size rounded up to whatever alignment the caller wants.
7186 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
7187 defining this macro. If unable, use the expression
7188 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
7189 before and after that, output the additional assembler syntax for defining
7190 the name, and a newline.
7192 There are two ways of handling global BSS@. One is to define either
7193 this macro or its aligned counterpart, @code{ASM_OUTPUT_ALIGNED_BSS}.
7194 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
7195 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
7196 You do not need to do both.
7198 Some languages do not have @code{common} data, and require a
7199 non-common form of global BSS in order to handle uninitialized globals
7200 efficiently. C++ is one example of this. However, if the target does
7201 not support global BSS, the front end may choose to make globals
7202 common in order to save space in the object file.
7205 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7206 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
7207 separate, explicit argument. If you define this macro, it is used in
7208 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
7209 handling the required alignment of the variable. The alignment is specified
7210 as the number of bits.
7212 Try to use function @code{asm_output_aligned_bss} defined in file
7213 @file{varasm.c} when defining this macro.
7216 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
7217 A C statement (sans semicolon) to output to the stdio stream
7218 @var{stream} the assembler definition of a local-common-label named
7219 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7220 is the size rounded up to whatever alignment the caller wants.
7222 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7223 output the name itself; before and after that, output the additional
7224 assembler syntax for defining the name, and a newline.
7226 This macro controls how the assembler definitions of uninitialized
7227 static variables are output.
7230 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
7231 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
7232 separate, explicit argument. If you define this macro, it is used in
7233 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
7234 handling the required alignment of the variable. The alignment is specified
7235 as the number of bits.
7238 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7239 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
7240 variable to be output, if there is one, or @code{NULL_TREE} if there
7241 is no corresponding variable. If you define this macro, GCC will use it
7242 in place of both @code{ASM_OUTPUT_DECL} and
7243 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
7244 the variable's decl in order to chose what to output.
7248 @subsection Output and Generation of Labels
7250 @c prevent bad page break with this line
7251 This is about outputting labels.
7253 @findex assemble_name
7254 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
7255 A C statement (sans semicolon) to output to the stdio stream
7256 @var{stream} the assembler definition of a label named @var{name}.
7257 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7258 output the name itself; before and after that, output the additional
7259 assembler syntax for defining the name, and a newline. A default
7260 definition of this macro is provided which is correct for most systems.
7263 @findex assemble_name_raw
7264 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
7265 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
7266 to refer to a compiler-generated label. The default definition uses
7267 @code{assemble_name_raw}, which is like @code{assemble_name} except
7268 that it is more efficient.
7272 A C string containing the appropriate assembler directive to specify the
7273 size of a symbol, without any arguments. On systems that use ELF, the
7274 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
7275 systems, the default is not to define this macro.
7277 Define this macro only if it is correct to use the default definitions
7278 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
7279 for your system. If you need your own custom definitions of those
7280 macros, or if you do not need explicit symbol sizes at all, do not
7284 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
7285 A C statement (sans semicolon) to output to the stdio stream
7286 @var{stream} a directive telling the assembler that the size of the
7287 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
7288 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7292 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
7293 A C statement (sans semicolon) to output to the stdio stream
7294 @var{stream} a directive telling the assembler to calculate the size of
7295 the symbol @var{name} by subtracting its address from the current
7298 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7299 provided. The default assumes that the assembler recognizes a special
7300 @samp{.} symbol as referring to the current address, and can calculate
7301 the difference between this and another symbol. If your assembler does
7302 not recognize @samp{.} or cannot do calculations with it, you will need
7303 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
7307 A C string containing the appropriate assembler directive to specify the
7308 type of a symbol, without any arguments. On systems that use ELF, the
7309 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
7310 systems, the default is not to define this macro.
7312 Define this macro only if it is correct to use the default definition of
7313 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7314 custom definition of this macro, or if you do not need explicit symbol
7315 types at all, do not define this macro.
7318 @defmac TYPE_OPERAND_FMT
7319 A C string which specifies (using @code{printf} syntax) the format of
7320 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
7321 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
7322 the default is not to define this macro.
7324 Define this macro only if it is correct to use the default definition of
7325 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7326 custom definition of this macro, or if you do not need explicit symbol
7327 types at all, do not define this macro.
7330 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
7331 A C statement (sans semicolon) to output to the stdio stream
7332 @var{stream} a directive telling the assembler that the type of the
7333 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
7334 that string is always either @samp{"function"} or @samp{"object"}, but
7335 you should not count on this.
7337 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
7338 definition of this macro is provided.
7341 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
7342 A C statement (sans semicolon) to output to the stdio stream
7343 @var{stream} any text necessary for declaring the name @var{name} of a
7344 function which is being defined. This macro is responsible for
7345 outputting the label definition (perhaps using
7346 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
7347 @code{FUNCTION_DECL} tree node representing the function.
7349 If this macro is not defined, then the function name is defined in the
7350 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7352 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7356 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
7357 A C statement (sans semicolon) to output to the stdio stream
7358 @var{stream} any text necessary for declaring the size of a function
7359 which is being defined. The argument @var{name} is the name of the
7360 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
7361 representing the function.
7363 If this macro is not defined, then the function size is not defined.
7365 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
7369 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
7370 A C statement (sans semicolon) to output to the stdio stream
7371 @var{stream} any text necessary for declaring the name @var{name} of an
7372 initialized variable which is being defined. This macro must output the
7373 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
7374 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
7376 If this macro is not defined, then the variable name is defined in the
7377 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7379 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
7380 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
7383 @defmac ASM_DECLARE_CONSTANT_NAME (@var{stream}, @var{name}, @var{exp}, @var{size})
7384 A C statement (sans semicolon) to output to the stdio stream
7385 @var{stream} any text necessary for declaring the name @var{name} of a
7386 constant which is being defined. This macro is responsible for
7387 outputting the label definition (perhaps using
7388 @code{ASM_OUTPUT_LABEL}). The argument @var{exp} is the
7389 value of the constant, and @var{size} is the size of the constant
7390 in bytes. @var{name} will be an internal label.
7392 If this macro is not defined, then the @var{name} is defined in the
7393 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7395 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7399 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
7400 A C statement (sans semicolon) to output to the stdio stream
7401 @var{stream} any text necessary for claiming a register @var{regno}
7402 for a global variable @var{decl} with name @var{name}.
7404 If you don't define this macro, that is equivalent to defining it to do
7408 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
7409 A C statement (sans semicolon) to finish up declaring a variable name
7410 once the compiler has processed its initializer fully and thus has had a
7411 chance to determine the size of an array when controlled by an
7412 initializer. This is used on systems where it's necessary to declare
7413 something about the size of the object.
7415 If you don't define this macro, that is equivalent to defining it to do
7418 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
7419 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
7422 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
7423 This target hook is a function to output to the stdio stream
7424 @var{stream} some commands that will make the label @var{name} global;
7425 that is, available for reference from other files.
7427 The default implementation relies on a proper definition of
7428 @code{GLOBAL_ASM_OP}.
7431 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *@var{stream}, tree @var{decl})
7432 This target hook is a function to output to the stdio stream
7433 @var{stream} some commands that will make the name associated with @var{decl}
7434 global; that is, available for reference from other files.
7436 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
7439 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
7440 A C statement (sans semicolon) to output to the stdio stream
7441 @var{stream} some commands that will make the label @var{name} weak;
7442 that is, available for reference from other files but only used if
7443 no other definition is available. Use the expression
7444 @code{assemble_name (@var{stream}, @var{name})} to output the name
7445 itself; before and after that, output the additional assembler syntax
7446 for making that name weak, and a newline.
7448 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
7449 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
7453 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
7454 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
7455 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
7456 or variable decl. If @var{value} is not @code{NULL}, this C statement
7457 should output to the stdio stream @var{stream} assembler code which
7458 defines (equates) the weak symbol @var{name} to have the value
7459 @var{value}. If @var{value} is @code{NULL}, it should output commands
7460 to make @var{name} weak.
7463 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
7464 Outputs a directive that enables @var{name} to be used to refer to
7465 symbol @var{value} with weak-symbol semantics. @code{decl} is the
7466 declaration of @code{name}.
7469 @defmac SUPPORTS_WEAK
7470 A C expression which evaluates to true if the target supports weak symbols.
7472 If you don't define this macro, @file{defaults.h} provides a default
7473 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
7474 is defined, the default definition is @samp{1}; otherwise, it is
7475 @samp{0}. Define this macro if you want to control weak symbol support
7476 with a compiler flag such as @option{-melf}.
7479 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
7480 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
7481 public symbol such that extra copies in multiple translation units will
7482 be discarded by the linker. Define this macro if your object file
7483 format provides support for this concept, such as the @samp{COMDAT}
7484 section flags in the Microsoft Windows PE/COFF format, and this support
7485 requires changes to @var{decl}, such as putting it in a separate section.
7488 @defmac SUPPORTS_ONE_ONLY
7489 A C expression which evaluates to true if the target supports one-only
7492 If you don't define this macro, @file{varasm.c} provides a default
7493 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
7494 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
7495 you want to control one-only symbol support with a compiler flag, or if
7496 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
7497 be emitted as one-only.
7500 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
7501 This target hook is a function to output to @var{asm_out_file} some
7502 commands that will make the symbol(s) associated with @var{decl} have
7503 hidden, protected or internal visibility as specified by @var{visibility}.
7506 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
7507 A C expression that evaluates to true if the target's linker expects
7508 that weak symbols do not appear in a static archive's table of contents.
7509 The default is @code{0}.
7511 Leaving weak symbols out of an archive's table of contents means that,
7512 if a symbol will only have a definition in one translation unit and
7513 will have undefined references from other translation units, that
7514 symbol should not be weak. Defining this macro to be nonzero will
7515 thus have the effect that certain symbols that would normally be weak
7516 (explicit template instantiations, and vtables for polymorphic classes
7517 with noninline key methods) will instead be nonweak.
7519 The C++ ABI requires this macro to be zero. Define this macro for
7520 targets where full C++ ABI compliance is impossible and where linker
7521 restrictions require weak symbols to be left out of a static archive's
7525 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
7526 A C statement (sans semicolon) to output to the stdio stream
7527 @var{stream} any text necessary for declaring the name of an external
7528 symbol named @var{name} which is referenced in this compilation but
7529 not defined. The value of @var{decl} is the tree node for the
7532 This macro need not be defined if it does not need to output anything.
7533 The GNU assembler and most Unix assemblers don't require anything.
7536 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
7537 This target hook is a function to output to @var{asm_out_file} an assembler
7538 pseudo-op to declare a library function name external. The name of the
7539 library function is given by @var{symref}, which is a @code{symbol_ref}.
7542 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (tree @var{decl})
7543 This target hook is a function to output to @var{asm_out_file} an assembler
7544 directive to annotate used symbol. Darwin target use .no_dead_code_strip
7548 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
7549 A C statement (sans semicolon) to output to the stdio stream
7550 @var{stream} a reference in assembler syntax to a label named
7551 @var{name}. This should add @samp{_} to the front of the name, if that
7552 is customary on your operating system, as it is in most Berkeley Unix
7553 systems. This macro is used in @code{assemble_name}.
7556 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
7557 A C statement (sans semicolon) to output a reference to
7558 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
7559 will be used to output the name of the symbol. This macro may be used
7560 to modify the way a symbol is referenced depending on information
7561 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
7564 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
7565 A C statement (sans semicolon) to output a reference to @var{buf}, the
7566 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
7567 @code{assemble_name} will be used to output the name of the symbol.
7568 This macro is not used by @code{output_asm_label}, or the @code{%l}
7569 specifier that calls it; the intention is that this macro should be set
7570 when it is necessary to output a label differently when its address is
7574 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
7575 A function to output to the stdio stream @var{stream} a label whose
7576 name is made from the string @var{prefix} and the number @var{labelno}.
7578 It is absolutely essential that these labels be distinct from the labels
7579 used for user-level functions and variables. Otherwise, certain programs
7580 will have name conflicts with internal labels.
7582 It is desirable to exclude internal labels from the symbol table of the
7583 object file. Most assemblers have a naming convention for labels that
7584 should be excluded; on many systems, the letter @samp{L} at the
7585 beginning of a label has this effect. You should find out what
7586 convention your system uses, and follow it.
7588 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
7591 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
7592 A C statement to output to the stdio stream @var{stream} a debug info
7593 label whose name is made from the string @var{prefix} and the number
7594 @var{num}. This is useful for VLIW targets, where debug info labels
7595 may need to be treated differently than branch target labels. On some
7596 systems, branch target labels must be at the beginning of instruction
7597 bundles, but debug info labels can occur in the middle of instruction
7600 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
7604 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
7605 A C statement to store into the string @var{string} a label whose name
7606 is made from the string @var{prefix} and the number @var{num}.
7608 This string, when output subsequently by @code{assemble_name}, should
7609 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
7610 with the same @var{prefix} and @var{num}.
7612 If the string begins with @samp{*}, then @code{assemble_name} will
7613 output the rest of the string unchanged. It is often convenient for
7614 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
7615 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
7616 to output the string, and may change it. (Of course,
7617 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
7618 you should know what it does on your machine.)
7621 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
7622 A C expression to assign to @var{outvar} (which is a variable of type
7623 @code{char *}) a newly allocated string made from the string
7624 @var{name} and the number @var{number}, with some suitable punctuation
7625 added. Use @code{alloca} to get space for the string.
7627 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
7628 produce an assembler label for an internal static variable whose name is
7629 @var{name}. Therefore, the string must be such as to result in valid
7630 assembler code. The argument @var{number} is different each time this
7631 macro is executed; it prevents conflicts between similarly-named
7632 internal static variables in different scopes.
7634 Ideally this string should not be a valid C identifier, to prevent any
7635 conflict with the user's own symbols. Most assemblers allow periods
7636 or percent signs in assembler symbols; putting at least one of these
7637 between the name and the number will suffice.
7639 If this macro is not defined, a default definition will be provided
7640 which is correct for most systems.
7643 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
7644 A C statement to output to the stdio stream @var{stream} assembler code
7645 which defines (equates) the symbol @var{name} to have the value @var{value}.
7648 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7649 correct for most systems.
7652 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
7653 A C statement to output to the stdio stream @var{stream} assembler code
7654 which defines (equates) the symbol whose tree node is @var{decl_of_name}
7655 to have the value of the tree node @var{decl_of_value}. This macro will
7656 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
7657 the tree nodes are available.
7660 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7661 correct for most systems.
7664 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
7665 A C statement that evaluates to true if the assembler code which defines
7666 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
7667 of the tree node @var{decl_of_value} should be emitted near the end of the
7668 current compilation unit. The default is to not defer output of defines.
7669 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
7670 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
7673 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
7674 A C statement to output to the stdio stream @var{stream} assembler code
7675 which defines (equates) the weak symbol @var{name} to have the value
7676 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
7677 an undefined weak symbol.
7679 Define this macro if the target only supports weak aliases; define
7680 @code{ASM_OUTPUT_DEF} instead if possible.
7683 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
7684 Define this macro to override the default assembler names used for
7685 Objective-C methods.
7687 The default name is a unique method number followed by the name of the
7688 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
7689 the category is also included in the assembler name (e.g.@:
7692 These names are safe on most systems, but make debugging difficult since
7693 the method's selector is not present in the name. Therefore, particular
7694 systems define other ways of computing names.
7696 @var{buf} is an expression of type @code{char *} which gives you a
7697 buffer in which to store the name; its length is as long as
7698 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
7699 50 characters extra.
7701 The argument @var{is_inst} specifies whether the method is an instance
7702 method or a class method; @var{class_name} is the name of the class;
7703 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
7704 in a category); and @var{sel_name} is the name of the selector.
7706 On systems where the assembler can handle quoted names, you can use this
7707 macro to provide more human-readable names.
7710 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
7711 A C statement (sans semicolon) to output to the stdio stream
7712 @var{stream} commands to declare that the label @var{name} is an
7713 Objective-C class reference. This is only needed for targets whose
7714 linkers have special support for NeXT-style runtimes.
7717 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
7718 A C statement (sans semicolon) to output to the stdio stream
7719 @var{stream} commands to declare that the label @var{name} is an
7720 unresolved Objective-C class reference. This is only needed for targets
7721 whose linkers have special support for NeXT-style runtimes.
7724 @node Initialization
7725 @subsection How Initialization Functions Are Handled
7726 @cindex initialization routines
7727 @cindex termination routines
7728 @cindex constructors, output of
7729 @cindex destructors, output of
7731 The compiled code for certain languages includes @dfn{constructors}
7732 (also called @dfn{initialization routines})---functions to initialize
7733 data in the program when the program is started. These functions need
7734 to be called before the program is ``started''---that is to say, before
7735 @code{main} is called.
7737 Compiling some languages generates @dfn{destructors} (also called
7738 @dfn{termination routines}) that should be called when the program
7741 To make the initialization and termination functions work, the compiler
7742 must output something in the assembler code to cause those functions to
7743 be called at the appropriate time. When you port the compiler to a new
7744 system, you need to specify how to do this.
7746 There are two major ways that GCC currently supports the execution of
7747 initialization and termination functions. Each way has two variants.
7748 Much of the structure is common to all four variations.
7750 @findex __CTOR_LIST__
7751 @findex __DTOR_LIST__
7752 The linker must build two lists of these functions---a list of
7753 initialization functions, called @code{__CTOR_LIST__}, and a list of
7754 termination functions, called @code{__DTOR_LIST__}.
7756 Each list always begins with an ignored function pointer (which may hold
7757 0, @minus{}1, or a count of the function pointers after it, depending on
7758 the environment). This is followed by a series of zero or more function
7759 pointers to constructors (or destructors), followed by a function
7760 pointer containing zero.
7762 Depending on the operating system and its executable file format, either
7763 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
7764 time and exit time. Constructors are called in reverse order of the
7765 list; destructors in forward order.
7767 The best way to handle static constructors works only for object file
7768 formats which provide arbitrarily-named sections. A section is set
7769 aside for a list of constructors, and another for a list of destructors.
7770 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
7771 object file that defines an initialization function also puts a word in
7772 the constructor section to point to that function. The linker
7773 accumulates all these words into one contiguous @samp{.ctors} section.
7774 Termination functions are handled similarly.
7776 This method will be chosen as the default by @file{target-def.h} if
7777 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
7778 support arbitrary sections, but does support special designated
7779 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
7780 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
7782 When arbitrary sections are available, there are two variants, depending
7783 upon how the code in @file{crtstuff.c} is called. On systems that
7784 support a @dfn{.init} section which is executed at program startup,
7785 parts of @file{crtstuff.c} are compiled into that section. The
7786 program is linked by the @command{gcc} driver like this:
7789 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
7792 The prologue of a function (@code{__init}) appears in the @code{.init}
7793 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
7794 for the function @code{__fini} in the @dfn{.fini} section. Normally these
7795 files are provided by the operating system or by the GNU C library, but
7796 are provided by GCC for a few targets.
7798 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
7799 compiled from @file{crtstuff.c}. They contain, among other things, code
7800 fragments within the @code{.init} and @code{.fini} sections that branch
7801 to routines in the @code{.text} section. The linker will pull all parts
7802 of a section together, which results in a complete @code{__init} function
7803 that invokes the routines we need at startup.
7805 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
7808 If no init section is available, when GCC compiles any function called
7809 @code{main} (or more accurately, any function designated as a program
7810 entry point by the language front end calling @code{expand_main_function}),
7811 it inserts a procedure call to @code{__main} as the first executable code
7812 after the function prologue. The @code{__main} function is defined
7813 in @file{libgcc2.c} and runs the global constructors.
7815 In file formats that don't support arbitrary sections, there are again
7816 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
7817 and an `a.out' format must be used. In this case,
7818 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
7819 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
7820 and with the address of the void function containing the initialization
7821 code as its value. The GNU linker recognizes this as a request to add
7822 the value to a @dfn{set}; the values are accumulated, and are eventually
7823 placed in the executable as a vector in the format described above, with
7824 a leading (ignored) count and a trailing zero element.
7825 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
7826 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
7827 the compilation of @code{main} to call @code{__main} as above, starting
7828 the initialization process.
7830 The last variant uses neither arbitrary sections nor the GNU linker.
7831 This is preferable when you want to do dynamic linking and when using
7832 file formats which the GNU linker does not support, such as `ECOFF'@. In
7833 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
7834 termination functions are recognized simply by their names. This requires
7835 an extra program in the linkage step, called @command{collect2}. This program
7836 pretends to be the linker, for use with GCC; it does its job by running
7837 the ordinary linker, but also arranges to include the vectors of
7838 initialization and termination functions. These functions are called
7839 via @code{__main} as described above. In order to use this method,
7840 @code{use_collect2} must be defined in the target in @file{config.gcc}.
7843 The following section describes the specific macros that control and
7844 customize the handling of initialization and termination functions.
7847 @node Macros for Initialization
7848 @subsection Macros Controlling Initialization Routines
7850 Here are the macros that control how the compiler handles initialization
7851 and termination functions:
7853 @defmac INIT_SECTION_ASM_OP
7854 If defined, a C string constant, including spacing, for the assembler
7855 operation to identify the following data as initialization code. If not
7856 defined, GCC will assume such a section does not exist. When you are
7857 using special sections for initialization and termination functions, this
7858 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7859 run the initialization functions.
7862 @defmac HAS_INIT_SECTION
7863 If defined, @code{main} will not call @code{__main} as described above.
7864 This macro should be defined for systems that control start-up code
7865 on a symbol-by-symbol basis, such as OSF/1, and should not
7866 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7869 @defmac LD_INIT_SWITCH
7870 If defined, a C string constant for a switch that tells the linker that
7871 the following symbol is an initialization routine.
7874 @defmac LD_FINI_SWITCH
7875 If defined, a C string constant for a switch that tells the linker that
7876 the following symbol is a finalization routine.
7879 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7880 If defined, a C statement that will write a function that can be
7881 automatically called when a shared library is loaded. The function
7882 should call @var{func}, which takes no arguments. If not defined, and
7883 the object format requires an explicit initialization function, then a
7884 function called @code{_GLOBAL__DI} will be generated.
7886 This function and the following one are used by collect2 when linking a
7887 shared library that needs constructors or destructors, or has DWARF2
7888 exception tables embedded in the code.
7891 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7892 If defined, a C statement that will write a function that can be
7893 automatically called when a shared library is unloaded. The function
7894 should call @var{func}, which takes no arguments. If not defined, and
7895 the object format requires an explicit finalization function, then a
7896 function called @code{_GLOBAL__DD} will be generated.
7899 @defmac INVOKE__main
7900 If defined, @code{main} will call @code{__main} despite the presence of
7901 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7902 where the init section is not actually run automatically, but is still
7903 useful for collecting the lists of constructors and destructors.
7906 @defmac SUPPORTS_INIT_PRIORITY
7907 If nonzero, the C++ @code{init_priority} attribute is supported and the
7908 compiler should emit instructions to control the order of initialization
7909 of objects. If zero, the compiler will issue an error message upon
7910 encountering an @code{init_priority} attribute.
7913 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7914 This value is true if the target supports some ``native'' method of
7915 collecting constructors and destructors to be run at startup and exit.
7916 It is false if we must use @command{collect2}.
7919 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7920 If defined, a function that outputs assembler code to arrange to call
7921 the function referenced by @var{symbol} at initialization time.
7923 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7924 no arguments and with no return value. If the target supports initialization
7925 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7926 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7928 If this macro is not defined by the target, a suitable default will
7929 be chosen if (1) the target supports arbitrary section names, (2) the
7930 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7934 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7935 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7936 functions rather than initialization functions.
7939 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7940 generated for the generated object file will have static linkage.
7942 If your system uses @command{collect2} as the means of processing
7943 constructors, then that program normally uses @command{nm} to scan
7944 an object file for constructor functions to be called.
7946 On certain kinds of systems, you can define this macro to make
7947 @command{collect2} work faster (and, in some cases, make it work at all):
7949 @defmac OBJECT_FORMAT_COFF
7950 Define this macro if the system uses COFF (Common Object File Format)
7951 object files, so that @command{collect2} can assume this format and scan
7952 object files directly for dynamic constructor/destructor functions.
7954 This macro is effective only in a native compiler; @command{collect2} as
7955 part of a cross compiler always uses @command{nm} for the target machine.
7958 @defmac REAL_NM_FILE_NAME
7959 Define this macro as a C string constant containing the file name to use
7960 to execute @command{nm}. The default is to search the path normally for
7963 If your system supports shared libraries and has a program to list the
7964 dynamic dependencies of a given library or executable, you can define
7965 these macros to enable support for running initialization and
7966 termination functions in shared libraries:
7970 Define this macro to a C string constant containing the name of the program
7971 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7974 @defmac PARSE_LDD_OUTPUT (@var{ptr})
7975 Define this macro to be C code that extracts filenames from the output
7976 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7977 of type @code{char *} that points to the beginning of a line of output
7978 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7979 code must advance @var{ptr} to the beginning of the filename on that
7980 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7983 @defmac SHLIB_SUFFIX
7984 Define this macro to a C string constant containing the default shared
7985 library extension of the target (e.g., @samp{".so"}). @command{collect2}
7986 strips version information after this suffix when generating global
7987 constructor and destructor names. This define is only needed on targets
7988 that use @command{collect2} to process constructors and destructors.
7991 @node Instruction Output
7992 @subsection Output of Assembler Instructions
7994 @c prevent bad page break with this line
7995 This describes assembler instruction output.
7997 @defmac REGISTER_NAMES
7998 A C initializer containing the assembler's names for the machine
7999 registers, each one as a C string constant. This is what translates
8000 register numbers in the compiler into assembler language.
8003 @defmac ADDITIONAL_REGISTER_NAMES
8004 If defined, a C initializer for an array of structures containing a name
8005 and a register number. This macro defines additional names for hard
8006 registers, thus allowing the @code{asm} option in declarations to refer
8007 to registers using alternate names.
8010 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
8011 Define this macro if you are using an unusual assembler that
8012 requires different names for the machine instructions.
8014 The definition is a C statement or statements which output an
8015 assembler instruction opcode to the stdio stream @var{stream}. The
8016 macro-operand @var{ptr} is a variable of type @code{char *} which
8017 points to the opcode name in its ``internal'' form---the form that is
8018 written in the machine description. The definition should output the
8019 opcode name to @var{stream}, performing any translation you desire, and
8020 increment the variable @var{ptr} to point at the end of the opcode
8021 so that it will not be output twice.
8023 In fact, your macro definition may process less than the entire opcode
8024 name, or more than the opcode name; but if you want to process text
8025 that includes @samp{%}-sequences to substitute operands, you must take
8026 care of the substitution yourself. Just be sure to increment
8027 @var{ptr} over whatever text should not be output normally.
8029 @findex recog_data.operand
8030 If you need to look at the operand values, they can be found as the
8031 elements of @code{recog_data.operand}.
8033 If the macro definition does nothing, the instruction is output
8037 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
8038 If defined, a C statement to be executed just prior to the output of
8039 assembler code for @var{insn}, to modify the extracted operands so
8040 they will be output differently.
8042 Here the argument @var{opvec} is the vector containing the operands
8043 extracted from @var{insn}, and @var{noperands} is the number of
8044 elements of the vector which contain meaningful data for this insn.
8045 The contents of this vector are what will be used to convert the insn
8046 template into assembler code, so you can change the assembler output
8047 by changing the contents of the vector.
8049 This macro is useful when various assembler syntaxes share a single
8050 file of instruction patterns; by defining this macro differently, you
8051 can cause a large class of instructions to be output differently (such
8052 as with rearranged operands). Naturally, variations in assembler
8053 syntax affecting individual insn patterns ought to be handled by
8054 writing conditional output routines in those patterns.
8056 If this macro is not defined, it is equivalent to a null statement.
8059 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
8060 A C compound statement to output to stdio stream @var{stream} the
8061 assembler syntax for an instruction operand @var{x}. @var{x} is an
8064 @var{code} is a value that can be used to specify one of several ways
8065 of printing the operand. It is used when identical operands must be
8066 printed differently depending on the context. @var{code} comes from
8067 the @samp{%} specification that was used to request printing of the
8068 operand. If the specification was just @samp{%@var{digit}} then
8069 @var{code} is 0; if the specification was @samp{%@var{ltr}
8070 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
8073 If @var{x} is a register, this macro should print the register's name.
8074 The names can be found in an array @code{reg_names} whose type is
8075 @code{char *[]}. @code{reg_names} is initialized from
8076 @code{REGISTER_NAMES}.
8078 When the machine description has a specification @samp{%@var{punct}}
8079 (a @samp{%} followed by a punctuation character), this macro is called
8080 with a null pointer for @var{x} and the punctuation character for
8084 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
8085 A C expression which evaluates to true if @var{code} is a valid
8086 punctuation character for use in the @code{PRINT_OPERAND} macro. If
8087 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
8088 punctuation characters (except for the standard one, @samp{%}) are used
8092 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
8093 A C compound statement to output to stdio stream @var{stream} the
8094 assembler syntax for an instruction operand that is a memory reference
8095 whose address is @var{x}. @var{x} is an RTL expression.
8097 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
8098 On some machines, the syntax for a symbolic address depends on the
8099 section that the address refers to. On these machines, define the hook
8100 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
8101 @code{symbol_ref}, and then check for it here. @xref{Assembler
8105 @findex dbr_sequence_length
8106 @defmac DBR_OUTPUT_SEQEND (@var{file})
8107 A C statement, to be executed after all slot-filler instructions have
8108 been output. If necessary, call @code{dbr_sequence_length} to
8109 determine the number of slots filled in a sequence (zero if not
8110 currently outputting a sequence), to decide how many no-ops to output,
8113 Don't define this macro if it has nothing to do, but it is helpful in
8114 reading assembly output if the extent of the delay sequence is made
8115 explicit (e.g.@: with white space).
8118 @findex final_sequence
8119 Note that output routines for instructions with delay slots must be
8120 prepared to deal with not being output as part of a sequence
8121 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
8122 found.) The variable @code{final_sequence} is null when not
8123 processing a sequence, otherwise it contains the @code{sequence} rtx
8127 @defmac REGISTER_PREFIX
8128 @defmacx LOCAL_LABEL_PREFIX
8129 @defmacx USER_LABEL_PREFIX
8130 @defmacx IMMEDIATE_PREFIX
8131 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
8132 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
8133 @file{final.c}). These are useful when a single @file{md} file must
8134 support multiple assembler formats. In that case, the various @file{tm.h}
8135 files can define these macros differently.
8138 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
8139 If defined this macro should expand to a series of @code{case}
8140 statements which will be parsed inside the @code{switch} statement of
8141 the @code{asm_fprintf} function. This allows targets to define extra
8142 printf formats which may useful when generating their assembler
8143 statements. Note that uppercase letters are reserved for future
8144 generic extensions to asm_fprintf, and so are not available to target
8145 specific code. The output file is given by the parameter @var{file}.
8146 The varargs input pointer is @var{argptr} and the rest of the format
8147 string, starting the character after the one that is being switched
8148 upon, is pointed to by @var{format}.
8151 @defmac ASSEMBLER_DIALECT
8152 If your target supports multiple dialects of assembler language (such as
8153 different opcodes), define this macro as a C expression that gives the
8154 numeric index of the assembler language dialect to use, with zero as the
8157 If this macro is defined, you may use constructs of the form
8159 @samp{@{option0|option1|option2@dots{}@}}
8162 in the output templates of patterns (@pxref{Output Template}) or in the
8163 first argument of @code{asm_fprintf}. This construct outputs
8164 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
8165 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
8166 within these strings retain their usual meaning. If there are fewer
8167 alternatives within the braces than the value of
8168 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
8170 If you do not define this macro, the characters @samp{@{}, @samp{|} and
8171 @samp{@}} do not have any special meaning when used in templates or
8172 operands to @code{asm_fprintf}.
8174 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
8175 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
8176 the variations in assembler language syntax with that mechanism. Define
8177 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
8178 if the syntax variant are larger and involve such things as different
8179 opcodes or operand order.
8182 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
8183 A C expression to output to @var{stream} some assembler code
8184 which will push hard register number @var{regno} onto the stack.
8185 The code need not be optimal, since this macro is used only when
8189 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
8190 A C expression to output to @var{stream} some assembler code
8191 which will pop hard register number @var{regno} off of the stack.
8192 The code need not be optimal, since this macro is used only when
8196 @node Dispatch Tables
8197 @subsection Output of Dispatch Tables
8199 @c prevent bad page break with this line
8200 This concerns dispatch tables.
8202 @cindex dispatch table
8203 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
8204 A C statement to output to the stdio stream @var{stream} an assembler
8205 pseudo-instruction to generate a difference between two labels.
8206 @var{value} and @var{rel} are the numbers of two internal labels. The
8207 definitions of these labels are output using
8208 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
8209 way here. For example,
8212 fprintf (@var{stream}, "\t.word L%d-L%d\n",
8213 @var{value}, @var{rel})
8216 You must provide this macro on machines where the addresses in a
8217 dispatch table are relative to the table's own address. If defined, GCC
8218 will also use this macro on all machines when producing PIC@.
8219 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
8220 mode and flags can be read.
8223 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
8224 This macro should be provided on machines where the addresses
8225 in a dispatch table are absolute.
8227 The definition should be a C statement to output to the stdio stream
8228 @var{stream} an assembler pseudo-instruction to generate a reference to
8229 a label. @var{value} is the number of an internal label whose
8230 definition is output using @code{(*targetm.asm_out.internal_label)}.
8234 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
8238 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
8239 Define this if the label before a jump-table needs to be output
8240 specially. The first three arguments are the same as for
8241 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
8242 jump-table which follows (a @code{jump_insn} containing an
8243 @code{addr_vec} or @code{addr_diff_vec}).
8245 This feature is used on system V to output a @code{swbeg} statement
8248 If this macro is not defined, these labels are output with
8249 @code{(*targetm.asm_out.internal_label)}.
8252 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
8253 Define this if something special must be output at the end of a
8254 jump-table. The definition should be a C statement to be executed
8255 after the assembler code for the table is written. It should write
8256 the appropriate code to stdio stream @var{stream}. The argument
8257 @var{table} is the jump-table insn, and @var{num} is the label-number
8258 of the preceding label.
8260 If this macro is not defined, nothing special is output at the end of
8264 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (@var{stream}, @var{decl}, @var{for_eh}, @var{empty})
8265 This target hook emits a label at the beginning of each FDE@. It
8266 should be defined on targets where FDEs need special labels, and it
8267 should write the appropriate label, for the FDE associated with the
8268 function declaration @var{decl}, to the stdio stream @var{stream}.
8269 The third argument, @var{for_eh}, is a boolean: true if this is for an
8270 exception table. The fourth argument, @var{empty}, is a boolean:
8271 true if this is a placeholder label for an omitted FDE@.
8273 The default is that FDEs are not given nonlocal labels.
8276 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (@var{stream})
8277 This target hook emits a label at the beginning of the exception table.
8278 It should be defined on targets where it is desirable for the table
8279 to be broken up according to function.
8281 The default is that no label is emitted.
8284 @deftypefn {Target Hook} void TARGET_UNWIND_EMIT (FILE * @var{stream}, rtx @var{insn})
8285 This target hook emits and assembly directives required to unwind the
8286 given instruction. This is only used when TARGET_UNWIND_INFO is set.
8289 @node Exception Region Output
8290 @subsection Assembler Commands for Exception Regions
8292 @c prevent bad page break with this line
8294 This describes commands marking the start and the end of an exception
8297 @defmac EH_FRAME_SECTION_NAME
8298 If defined, a C string constant for the name of the section containing
8299 exception handling frame unwind information. If not defined, GCC will
8300 provide a default definition if the target supports named sections.
8301 @file{crtstuff.c} uses this macro to switch to the appropriate section.
8303 You should define this symbol if your target supports DWARF 2 frame
8304 unwind information and the default definition does not work.
8307 @defmac EH_FRAME_IN_DATA_SECTION
8308 If defined, DWARF 2 frame unwind information will be placed in the
8309 data section even though the target supports named sections. This
8310 might be necessary, for instance, if the system linker does garbage
8311 collection and sections cannot be marked as not to be collected.
8313 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
8317 @defmac EH_TABLES_CAN_BE_READ_ONLY
8318 Define this macro to 1 if your target is such that no frame unwind
8319 information encoding used with non-PIC code will ever require a
8320 runtime relocation, but the linker may not support merging read-only
8321 and read-write sections into a single read-write section.
8324 @defmac MASK_RETURN_ADDR
8325 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
8326 that it does not contain any extraneous set bits in it.
8329 @defmac DWARF2_UNWIND_INFO
8330 Define this macro to 0 if your target supports DWARF 2 frame unwind
8331 information, but it does not yet work with exception handling.
8332 Otherwise, if your target supports this information (if it defines
8333 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
8334 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of 1.
8336 If @code{TARGET_UNWIND_INFO} is defined, the target specific unwinder
8337 will be used in all cases. Defining this macro will enable the generation
8338 of DWARF 2 frame debugging information.
8340 If @code{TARGET_UNWIND_INFO} is not defined, and this macro is defined to 1,
8341 the DWARF 2 unwinder will be the default exception handling mechanism;
8342 otherwise, the @code{setjmp}/@code{longjmp}-based scheme will be used by
8346 @defmac TARGET_UNWIND_INFO
8347 Define this macro if your target has ABI specified unwind tables. Usually
8348 these will be output by @code{TARGET_UNWIND_EMIT}.
8351 @deftypevar {Target Hook} bool TARGET_UNWIND_TABLES_DEFAULT
8352 This variable should be set to @code{true} if the target ABI requires unwinding
8353 tables even when exceptions are not used.
8356 @defmac MUST_USE_SJLJ_EXCEPTIONS
8357 This macro need only be defined if @code{DWARF2_UNWIND_INFO} is
8358 runtime-variable. In that case, @file{except.h} cannot correctly
8359 determine the corresponding definition of @code{MUST_USE_SJLJ_EXCEPTIONS},
8360 so the target must provide it directly.
8363 @defmac DONT_USE_BUILTIN_SETJMP
8364 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
8365 should use the @code{setjmp}/@code{longjmp} functions from the C library
8366 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
8369 @defmac DWARF_CIE_DATA_ALIGNMENT
8370 This macro need only be defined if the target might save registers in the
8371 function prologue at an offset to the stack pointer that is not aligned to
8372 @code{UNITS_PER_WORD}. The definition should be the negative minimum
8373 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
8374 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
8375 the target supports DWARF 2 frame unwind information.
8378 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
8379 Contains the value true if the target should add a zero word onto the
8380 end of a Dwarf-2 frame info section when used for exception handling.
8381 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
8385 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
8386 Given a register, this hook should return a parallel of registers to
8387 represent where to find the register pieces. Define this hook if the
8388 register and its mode are represented in Dwarf in non-contiguous
8389 locations, or if the register should be represented in more than one
8390 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
8391 If not defined, the default is to return @code{NULL_RTX}.
8394 @deftypefn {Target Hook} void TARGET_INIT_DWARF_REG_SIZES_EXTRA (tree @var{address})
8395 If some registers are represented in Dwarf-2 unwind information in
8396 multiple pieces, define this hook to fill in information about the
8397 sizes of those pieces in the table used by the unwinder at runtime.
8398 It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after
8399 filling in a single size corresponding to each hard register;
8400 @var{address} is the address of the table.
8403 @deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym})
8404 This hook is used to output a reference from a frame unwinding table to
8405 the type_info object identified by @var{sym}. It should return @code{true}
8406 if the reference was output. Returning @code{false} will cause the
8407 reference to be output using the normal Dwarf2 routines.
8410 @deftypefn {Target Hook} bool TARGET_ARM_EABI_UNWINDER
8411 This hook should be set to @code{true} on targets that use an ARM EABI
8412 based unwinding library, and @code{false} on other targets. This effects
8413 the format of unwinding tables, and how the unwinder in entered after
8414 running a cleanup. The default is @code{false}.
8417 @node Alignment Output
8418 @subsection Assembler Commands for Alignment
8420 @c prevent bad page break with this line
8421 This describes commands for alignment.
8423 @defmac JUMP_ALIGN (@var{label})
8424 The alignment (log base 2) to put in front of @var{label}, which is
8425 a common destination of jumps and has no fallthru incoming edge.
8427 This macro need not be defined if you don't want any special alignment
8428 to be done at such a time. Most machine descriptions do not currently
8431 Unless it's necessary to inspect the @var{label} parameter, it is better
8432 to set the variable @var{align_jumps} in the target's
8433 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8434 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
8437 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
8438 The alignment (log base 2) to put in front of @var{label}, which follows
8441 This macro need not be defined if you don't want any special alignment
8442 to be done at such a time. Most machine descriptions do not currently
8446 @defmac LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
8447 The maximum number of bytes to skip when applying
8448 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
8449 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8452 @defmac LOOP_ALIGN (@var{label})
8453 The alignment (log base 2) to put in front of @var{label}, which follows
8454 a @code{NOTE_INSN_LOOP_BEG} note.
8456 This macro need not be defined if you don't want any special alignment
8457 to be done at such a time. Most machine descriptions do not currently
8460 Unless it's necessary to inspect the @var{label} parameter, it is better
8461 to set the variable @code{align_loops} in the target's
8462 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8463 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
8466 @defmac LOOP_ALIGN_MAX_SKIP
8467 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
8468 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8471 @defmac LABEL_ALIGN (@var{label})
8472 The alignment (log base 2) to put in front of @var{label}.
8473 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
8474 the maximum of the specified values is used.
8476 Unless it's necessary to inspect the @var{label} parameter, it is better
8477 to set the variable @code{align_labels} in the target's
8478 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8479 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
8482 @defmac LABEL_ALIGN_MAX_SKIP
8483 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
8484 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8487 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
8488 A C statement to output to the stdio stream @var{stream} an assembler
8489 instruction to advance the location counter by @var{nbytes} bytes.
8490 Those bytes should be zero when loaded. @var{nbytes} will be a C
8491 expression of type @code{unsigned HOST_WIDE_INT}.
8494 @defmac ASM_NO_SKIP_IN_TEXT
8495 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
8496 text section because it fails to put zeros in the bytes that are skipped.
8497 This is true on many Unix systems, where the pseudo--op to skip bytes
8498 produces no-op instructions rather than zeros when used in the text
8502 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
8503 A C statement to output to the stdio stream @var{stream} an assembler
8504 command to advance the location counter to a multiple of 2 to the
8505 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
8508 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
8509 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
8510 for padding, if necessary.
8513 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
8514 A C statement to output to the stdio stream @var{stream} an assembler
8515 command to advance the location counter to a multiple of 2 to the
8516 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
8517 satisfy the alignment request. @var{power} and @var{max_skip} will be
8518 a C expression of type @code{int}.
8522 @node Debugging Info
8523 @section Controlling Debugging Information Format
8525 @c prevent bad page break with this line
8526 This describes how to specify debugging information.
8529 * All Debuggers:: Macros that affect all debugging formats uniformly.
8530 * DBX Options:: Macros enabling specific options in DBX format.
8531 * DBX Hooks:: Hook macros for varying DBX format.
8532 * File Names and DBX:: Macros controlling output of file names in DBX format.
8533 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
8534 * VMS Debug:: Macros for VMS debug format.
8538 @subsection Macros Affecting All Debugging Formats
8540 @c prevent bad page break with this line
8541 These macros affect all debugging formats.
8543 @defmac DBX_REGISTER_NUMBER (@var{regno})
8544 A C expression that returns the DBX register number for the compiler
8545 register number @var{regno}. In the default macro provided, the value
8546 of this expression will be @var{regno} itself. But sometimes there are
8547 some registers that the compiler knows about and DBX does not, or vice
8548 versa. In such cases, some register may need to have one number in the
8549 compiler and another for DBX@.
8551 If two registers have consecutive numbers inside GCC, and they can be
8552 used as a pair to hold a multiword value, then they @emph{must} have
8553 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
8554 Otherwise, debuggers will be unable to access such a pair, because they
8555 expect register pairs to be consecutive in their own numbering scheme.
8557 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
8558 does not preserve register pairs, then what you must do instead is
8559 redefine the actual register numbering scheme.
8562 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
8563 A C expression that returns the integer offset value for an automatic
8564 variable having address @var{x} (an RTL expression). The default
8565 computation assumes that @var{x} is based on the frame-pointer and
8566 gives the offset from the frame-pointer. This is required for targets
8567 that produce debugging output for DBX or COFF-style debugging output
8568 for SDB and allow the frame-pointer to be eliminated when the
8569 @option{-g} options is used.
8572 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
8573 A C expression that returns the integer offset value for an argument
8574 having address @var{x} (an RTL expression). The nominal offset is
8578 @defmac PREFERRED_DEBUGGING_TYPE
8579 A C expression that returns the type of debugging output GCC should
8580 produce when the user specifies just @option{-g}. Define
8581 this if you have arranged for GCC to support more than one format of
8582 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
8583 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
8584 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
8586 When the user specifies @option{-ggdb}, GCC normally also uses the
8587 value of this macro to select the debugging output format, but with two
8588 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
8589 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
8590 defined, GCC uses @code{DBX_DEBUG}.
8592 The value of this macro only affects the default debugging output; the
8593 user can always get a specific type of output by using @option{-gstabs},
8594 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
8598 @subsection Specific Options for DBX Output
8600 @c prevent bad page break with this line
8601 These are specific options for DBX output.
8603 @defmac DBX_DEBUGGING_INFO
8604 Define this macro if GCC should produce debugging output for DBX
8605 in response to the @option{-g} option.
8608 @defmac XCOFF_DEBUGGING_INFO
8609 Define this macro if GCC should produce XCOFF format debugging output
8610 in response to the @option{-g} option. This is a variant of DBX format.
8613 @defmac DEFAULT_GDB_EXTENSIONS
8614 Define this macro to control whether GCC should by default generate
8615 GDB's extended version of DBX debugging information (assuming DBX-format
8616 debugging information is enabled at all). If you don't define the
8617 macro, the default is 1: always generate the extended information
8618 if there is any occasion to.
8621 @defmac DEBUG_SYMS_TEXT
8622 Define this macro if all @code{.stabs} commands should be output while
8623 in the text section.
8626 @defmac ASM_STABS_OP
8627 A C string constant, including spacing, naming the assembler pseudo op to
8628 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
8629 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
8630 applies only to DBX debugging information format.
8633 @defmac ASM_STABD_OP
8634 A C string constant, including spacing, naming the assembler pseudo op to
8635 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
8636 value is the current location. If you don't define this macro,
8637 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
8641 @defmac ASM_STABN_OP
8642 A C string constant, including spacing, naming the assembler pseudo op to
8643 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
8644 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
8645 macro applies only to DBX debugging information format.
8648 @defmac DBX_NO_XREFS
8649 Define this macro if DBX on your system does not support the construct
8650 @samp{xs@var{tagname}}. On some systems, this construct is used to
8651 describe a forward reference to a structure named @var{tagname}.
8652 On other systems, this construct is not supported at all.
8655 @defmac DBX_CONTIN_LENGTH
8656 A symbol name in DBX-format debugging information is normally
8657 continued (split into two separate @code{.stabs} directives) when it
8658 exceeds a certain length (by default, 80 characters). On some
8659 operating systems, DBX requires this splitting; on others, splitting
8660 must not be done. You can inhibit splitting by defining this macro
8661 with the value zero. You can override the default splitting-length by
8662 defining this macro as an expression for the length you desire.
8665 @defmac DBX_CONTIN_CHAR
8666 Normally continuation is indicated by adding a @samp{\} character to
8667 the end of a @code{.stabs} string when a continuation follows. To use
8668 a different character instead, define this macro as a character
8669 constant for the character you want to use. Do not define this macro
8670 if backslash is correct for your system.
8673 @defmac DBX_STATIC_STAB_DATA_SECTION
8674 Define this macro if it is necessary to go to the data section before
8675 outputting the @samp{.stabs} pseudo-op for a non-global static
8679 @defmac DBX_TYPE_DECL_STABS_CODE
8680 The value to use in the ``code'' field of the @code{.stabs} directive
8681 for a typedef. The default is @code{N_LSYM}.
8684 @defmac DBX_STATIC_CONST_VAR_CODE
8685 The value to use in the ``code'' field of the @code{.stabs} directive
8686 for a static variable located in the text section. DBX format does not
8687 provide any ``right'' way to do this. The default is @code{N_FUN}.
8690 @defmac DBX_REGPARM_STABS_CODE
8691 The value to use in the ``code'' field of the @code{.stabs} directive
8692 for a parameter passed in registers. DBX format does not provide any
8693 ``right'' way to do this. The default is @code{N_RSYM}.
8696 @defmac DBX_REGPARM_STABS_LETTER
8697 The letter to use in DBX symbol data to identify a symbol as a parameter
8698 passed in registers. DBX format does not customarily provide any way to
8699 do this. The default is @code{'P'}.
8702 @defmac DBX_FUNCTION_FIRST
8703 Define this macro if the DBX information for a function and its
8704 arguments should precede the assembler code for the function. Normally,
8705 in DBX format, the debugging information entirely follows the assembler
8709 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
8710 Define this macro, with value 1, if the value of a symbol describing
8711 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
8712 relative to the start of the enclosing function. Normally, GCC uses
8713 an absolute address.
8716 @defmac DBX_LINES_FUNCTION_RELATIVE
8717 Define this macro, with value 1, if the value of a symbol indicating
8718 the current line number (@code{N_SLINE}) should be relative to the
8719 start of the enclosing function. Normally, GCC uses an absolute address.
8722 @defmac DBX_USE_BINCL
8723 Define this macro if GCC should generate @code{N_BINCL} and
8724 @code{N_EINCL} stabs for included header files, as on Sun systems. This
8725 macro also directs GCC to output a type number as a pair of a file
8726 number and a type number within the file. Normally, GCC does not
8727 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
8728 number for a type number.
8732 @subsection Open-Ended Hooks for DBX Format
8734 @c prevent bad page break with this line
8735 These are hooks for DBX format.
8737 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
8738 Define this macro to say how to output to @var{stream} the debugging
8739 information for the start of a scope level for variable names. The
8740 argument @var{name} is the name of an assembler symbol (for use with
8741 @code{assemble_name}) whose value is the address where the scope begins.
8744 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
8745 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
8748 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
8749 Define this macro if the target machine requires special handling to
8750 output an @code{N_FUN} entry for the function @var{decl}.
8753 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
8754 A C statement to output DBX debugging information before code for line
8755 number @var{line} of the current source file to the stdio stream
8756 @var{stream}. @var{counter} is the number of time the macro was
8757 invoked, including the current invocation; it is intended to generate
8758 unique labels in the assembly output.
8760 This macro should not be defined if the default output is correct, or
8761 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
8764 @defmac NO_DBX_FUNCTION_END
8765 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
8766 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
8767 On those machines, define this macro to turn this feature off without
8768 disturbing the rest of the gdb extensions.
8771 @defmac NO_DBX_BNSYM_ENSYM
8772 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
8773 extension construct. On those machines, define this macro to turn this
8774 feature off without disturbing the rest of the gdb extensions.
8777 @node File Names and DBX
8778 @subsection File Names in DBX Format
8780 @c prevent bad page break with this line
8781 This describes file names in DBX format.
8783 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
8784 A C statement to output DBX debugging information to the stdio stream
8785 @var{stream}, which indicates that file @var{name} is the main source
8786 file---the file specified as the input file for compilation.
8787 This macro is called only once, at the beginning of compilation.
8789 This macro need not be defined if the standard form of output
8790 for DBX debugging information is appropriate.
8792 It may be necessary to refer to a label equal to the beginning of the
8793 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
8794 to do so. If you do this, you must also set the variable
8795 @var{used_ltext_label_name} to @code{true}.
8798 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
8799 Define this macro, with value 1, if GCC should not emit an indication
8800 of the current directory for compilation and current source language at
8801 the beginning of the file.
8804 @defmac NO_DBX_GCC_MARKER
8805 Define this macro, with value 1, if GCC should not emit an indication
8806 that this object file was compiled by GCC@. The default is to emit
8807 an @code{N_OPT} stab at the beginning of every source file, with
8808 @samp{gcc2_compiled.} for the string and value 0.
8811 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
8812 A C statement to output DBX debugging information at the end of
8813 compilation of the main source file @var{name}. Output should be
8814 written to the stdio stream @var{stream}.
8816 If you don't define this macro, nothing special is output at the end
8817 of compilation, which is correct for most machines.
8820 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
8821 Define this macro @emph{instead of} defining
8822 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
8823 the end of compilation is a @code{N_SO} stab with an empty string,
8824 whose value is the highest absolute text address in the file.
8829 @subsection Macros for SDB and DWARF Output
8831 @c prevent bad page break with this line
8832 Here are macros for SDB and DWARF output.
8834 @defmac SDB_DEBUGGING_INFO
8835 Define this macro if GCC should produce COFF-style debugging output
8836 for SDB in response to the @option{-g} option.
8839 @defmac DWARF2_DEBUGGING_INFO
8840 Define this macro if GCC should produce dwarf version 2 format
8841 debugging output in response to the @option{-g} option.
8843 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (tree @var{function})
8844 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
8845 be emitted for each function. Instead of an integer return the enum
8846 value for the @code{DW_CC_} tag.
8849 To support optional call frame debugging information, you must also
8850 define @code{INCOMING_RETURN_ADDR_RTX} and either set
8851 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
8852 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
8853 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
8856 @defmac DWARF2_FRAME_INFO
8857 Define this macro to a nonzero value if GCC should always output
8858 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
8859 (@pxref{Exception Region Output} is nonzero, GCC will output this
8860 information not matter how you define @code{DWARF2_FRAME_INFO}.
8863 @defmac DWARF2_ASM_LINE_DEBUG_INFO
8864 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8865 line debug info sections. This will result in much more compact line number
8866 tables, and hence is desirable if it works.
8869 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8870 A C statement to issue assembly directives that create a difference
8871 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
8874 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
8875 A C statement to issue assembly directives that create a
8876 section-relative reference to the given @var{label}, using an integer of the
8877 given @var{size}. The label is known to be defined in the given @var{section}.
8880 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8881 A C statement to issue assembly directives that create a self-relative
8882 reference to the given @var{label}, using an integer of the given @var{size}.
8885 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{FILE}, int @var{size}, rtx @var{x})
8886 If defined, this target hook is a function which outputs a DTP-relative
8887 reference to the given TLS symbol of the specified size.
8890 @defmac PUT_SDB_@dots{}
8891 Define these macros to override the assembler syntax for the special
8892 SDB assembler directives. See @file{sdbout.c} for a list of these
8893 macros and their arguments. If the standard syntax is used, you need
8894 not define them yourself.
8898 Some assemblers do not support a semicolon as a delimiter, even between
8899 SDB assembler directives. In that case, define this macro to be the
8900 delimiter to use (usually @samp{\n}). It is not necessary to define
8901 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8905 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
8906 Define this macro to allow references to unknown structure,
8907 union, or enumeration tags to be emitted. Standard COFF does not
8908 allow handling of unknown references, MIPS ECOFF has support for
8912 @defmac SDB_ALLOW_FORWARD_REFERENCES
8913 Define this macro to allow references to structure, union, or
8914 enumeration tags that have not yet been seen to be handled. Some
8915 assemblers choke if forward tags are used, while some require it.
8918 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
8919 A C statement to output SDB debugging information before code for line
8920 number @var{line} of the current source file to the stdio stream
8921 @var{stream}. The default is to emit an @code{.ln} directive.
8926 @subsection Macros for VMS Debug Format
8928 @c prevent bad page break with this line
8929 Here are macros for VMS debug format.
8931 @defmac VMS_DEBUGGING_INFO
8932 Define this macro if GCC should produce debugging output for VMS
8933 in response to the @option{-g} option. The default behavior for VMS
8934 is to generate minimal debug info for a traceback in the absence of
8935 @option{-g} unless explicitly overridden with @option{-g0}. This
8936 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8937 @code{OVERRIDE_OPTIONS}.
8940 @node Floating Point
8941 @section Cross Compilation and Floating Point
8942 @cindex cross compilation and floating point
8943 @cindex floating point and cross compilation
8945 While all modern machines use twos-complement representation for integers,
8946 there are a variety of representations for floating point numbers. This
8947 means that in a cross-compiler the representation of floating point numbers
8948 in the compiled program may be different from that used in the machine
8949 doing the compilation.
8951 Because different representation systems may offer different amounts of
8952 range and precision, all floating point constants must be represented in
8953 the target machine's format. Therefore, the cross compiler cannot
8954 safely use the host machine's floating point arithmetic; it must emulate
8955 the target's arithmetic. To ensure consistency, GCC always uses
8956 emulation to work with floating point values, even when the host and
8957 target floating point formats are identical.
8959 The following macros are provided by @file{real.h} for the compiler to
8960 use. All parts of the compiler which generate or optimize
8961 floating-point calculations must use these macros. They may evaluate
8962 their operands more than once, so operands must not have side effects.
8964 @defmac REAL_VALUE_TYPE
8965 The C data type to be used to hold a floating point value in the target
8966 machine's format. Typically this is a @code{struct} containing an
8967 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8971 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8972 Compares for equality the two values, @var{x} and @var{y}. If the target
8973 floating point format supports negative zeroes and/or NaNs,
8974 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8975 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8978 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8979 Tests whether @var{x} is less than @var{y}.
8982 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8983 Truncates @var{x} to a signed integer, rounding toward zero.
8986 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8987 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8988 @var{x} is negative, returns zero.
8991 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8992 Converts @var{string} into a floating point number in the target machine's
8993 representation for mode @var{mode}. This routine can handle both
8994 decimal and hexadecimal floating point constants, using the syntax
8995 defined by the C language for both.
8998 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8999 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
9002 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
9003 Determines whether @var{x} represents infinity (positive or negative).
9006 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
9007 Determines whether @var{x} represents a ``NaN'' (not-a-number).
9010 @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})
9011 Calculates an arithmetic operation on the two floating point values
9012 @var{x} and @var{y}, storing the result in @var{output} (which must be a
9015 The operation to be performed is specified by @var{code}. Only the
9016 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
9017 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
9019 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
9020 target's floating point format cannot represent infinity, it will call
9021 @code{abort}. Callers should check for this situation first, using
9022 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
9025 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
9026 Returns the negative of the floating point value @var{x}.
9029 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
9030 Returns the absolute value of @var{x}.
9033 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
9034 Truncates the floating point value @var{x} to fit in @var{mode}. The
9035 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
9036 appropriate bit pattern to be output as a floating constant whose
9037 precision accords with mode @var{mode}.
9040 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
9041 Converts a floating point value @var{x} into a double-precision integer
9042 which is then stored into @var{low} and @var{high}. If the value is not
9043 integral, it is truncated.
9046 @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})
9047 Converts a double-precision integer found in @var{low} and @var{high},
9048 into a floating point value which is then stored into @var{x}. The
9049 value is truncated to fit in mode @var{mode}.
9052 @node Mode Switching
9053 @section Mode Switching Instructions
9054 @cindex mode switching
9055 The following macros control mode switching optimizations:
9057 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
9058 Define this macro if the port needs extra instructions inserted for mode
9059 switching in an optimizing compilation.
9061 For an example, the SH4 can perform both single and double precision
9062 floating point operations, but to perform a single precision operation,
9063 the FPSCR PR bit has to be cleared, while for a double precision
9064 operation, this bit has to be set. Changing the PR bit requires a general
9065 purpose register as a scratch register, hence these FPSCR sets have to
9066 be inserted before reload, i.e.@: you can't put this into instruction emitting
9067 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
9069 You can have multiple entities that are mode-switched, and select at run time
9070 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
9071 return nonzero for any @var{entity} that needs mode-switching.
9072 If you define this macro, you also have to define
9073 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
9074 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
9075 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
9079 @defmac NUM_MODES_FOR_MODE_SWITCHING
9080 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
9081 initializer for an array of integers. Each initializer element
9082 N refers to an entity that needs mode switching, and specifies the number
9083 of different modes that might need to be set for this entity.
9084 The position of the initializer in the initializer---starting counting at
9085 zero---determines the integer that is used to refer to the mode-switched
9087 In macros that take mode arguments / yield a mode result, modes are
9088 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
9089 switch is needed / supplied.
9092 @defmac MODE_NEEDED (@var{entity}, @var{insn})
9093 @var{entity} is an integer specifying a mode-switched entity. If
9094 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
9095 return an integer value not larger than the corresponding element in
9096 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
9097 be switched into prior to the execution of @var{insn}.
9100 @defmac MODE_AFTER (@var{mode}, @var{insn})
9101 If this macro is defined, it is evaluated for every @var{insn} during
9102 mode switching. It determines the mode that an insn results in (if
9103 different from the incoming mode).
9106 @defmac MODE_ENTRY (@var{entity})
9107 If this macro is defined, it is evaluated for every @var{entity} that needs
9108 mode switching. It should evaluate to an integer, which is a mode that
9109 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
9110 is defined then @code{MODE_EXIT} must be defined.
9113 @defmac MODE_EXIT (@var{entity})
9114 If this macro is defined, it is evaluated for every @var{entity} that needs
9115 mode switching. It should evaluate to an integer, which is a mode that
9116 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
9117 is defined then @code{MODE_ENTRY} must be defined.
9120 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
9121 This macro specifies the order in which modes for @var{entity} are processed.
9122 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
9123 lowest. The value of the macro should be an integer designating a mode
9124 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
9125 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
9126 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
9129 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
9130 Generate one or more insns to set @var{entity} to @var{mode}.
9131 @var{hard_reg_live} is the set of hard registers live at the point where
9132 the insn(s) are to be inserted.
9135 @node Target Attributes
9136 @section Defining target-specific uses of @code{__attribute__}
9137 @cindex target attributes
9138 @cindex machine attributes
9139 @cindex attributes, target-specific
9141 Target-specific attributes may be defined for functions, data and types.
9142 These are described using the following target hooks; they also need to
9143 be documented in @file{extend.texi}.
9145 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
9146 If defined, this target hook points to an array of @samp{struct
9147 attribute_spec} (defined in @file{tree.h}) specifying the machine
9148 specific attributes for this target and some of the restrictions on the
9149 entities to which these attributes are applied and the arguments they
9153 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
9154 If defined, this target hook is a function which returns zero if the attributes on
9155 @var{type1} and @var{type2} are incompatible, one if they are compatible,
9156 and two if they are nearly compatible (which causes a warning to be
9157 generated). If this is not defined, machine-specific attributes are
9158 supposed always to be compatible.
9161 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
9162 If defined, this target hook is a function which assigns default attributes to
9163 newly defined @var{type}.
9166 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
9167 Define this target hook if the merging of type attributes needs special
9168 handling. If defined, the result is a list of the combined
9169 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
9170 that @code{comptypes} has already been called and returned 1. This
9171 function may call @code{merge_attributes} to handle machine-independent
9175 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
9176 Define this target hook if the merging of decl attributes needs special
9177 handling. If defined, the result is a list of the combined
9178 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
9179 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
9180 when this is needed are when one attribute overrides another, or when an
9181 attribute is nullified by a subsequent definition. This function may
9182 call @code{merge_attributes} to handle machine-independent merging.
9184 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
9185 If the only target-specific handling you require is @samp{dllimport}
9186 for Microsoft Windows targets, you should define the macro
9187 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
9188 will then define a function called
9189 @code{merge_dllimport_decl_attributes} which can then be defined as
9190 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
9191 add @code{handle_dll_attribute} in the attribute table for your port
9192 to perform initial processing of the @samp{dllimport} and
9193 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
9194 @file{i386/i386.c}, for example.
9197 @deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (tree @var{decl})
9198 @var{decl} is a variable or function with @code{__attribute__((dllimport))}
9199 specified. Use this hook if the target needs to add extra validation
9200 checks to @code{handle_dll_attribute}.
9203 @defmac TARGET_DECLSPEC
9204 Define this macro to a nonzero value if you want to treat
9205 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
9206 default, this behavior is enabled only for targets that define
9207 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
9208 of @code{__declspec} is via a built-in macro, but you should not rely
9209 on this implementation detail.
9212 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
9213 Define this target hook if you want to be able to add attributes to a decl
9214 when it is being created. This is normally useful for back ends which
9215 wish to implement a pragma by using the attributes which correspond to
9216 the pragma's effect. The @var{node} argument is the decl which is being
9217 created. The @var{attr_ptr} argument is a pointer to the attribute list
9218 for this decl. The list itself should not be modified, since it may be
9219 shared with other decls, but attributes may be chained on the head of
9220 the list and @code{*@var{attr_ptr}} modified to point to the new
9221 attributes, or a copy of the list may be made if further changes are
9225 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
9227 This target hook returns @code{true} if it is ok to inline @var{fndecl}
9228 into the current function, despite its having target-specific
9229 attributes, @code{false} otherwise. By default, if a function has a
9230 target specific attribute attached to it, it will not be inlined.
9234 @section Emulating TLS
9235 @cindex Emulated TLS
9237 For targets whose psABI does not provide Thread Local Storage via
9238 specific relocations and instruction sequences, an emulation layer is
9239 used. A set of target hooks allows this emulation layer to be
9240 configured for the requirements of a particular target. For instance
9241 the psABI may infact specify TLS support in terms of an emulation
9244 The emulation layer works by creating a control object for every TLS
9245 object. To access the TLS object, a lookup function is provided
9246 which, when given the address of the control object, will return the
9247 address of the current thread's instance of the TLS object.
9249 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_GET_ADDRESS
9250 Contains the name of the helper function that uses a TLS control
9251 object to locate a TLS instance. The default causes libgcc's
9252 emulated TLS helper function to be used.
9255 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_REGISTER_COMMON
9256 Contains the name of the helper function that should be used at
9257 program startup to register TLS objects that are implicitly
9258 initialized to zero. If this is @code{NULL}, all TLS objects will
9259 have explicit initializers. The default causes libgcc's emulated TLS
9260 registration function to be used.
9263 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_SECTION
9264 Contains the name of the section in which TLS control variables should
9265 be placed. The default of @code{NULL} allows these to be placed in
9269 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_SECTION
9270 Contains the name of the section in which TLS initializers should be
9271 placed. The default of @code{NULL} allows these to be placed in any
9275 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_PREFIX
9276 Contains the prefix to be prepended to TLS control variable names.
9277 The default of @code{NULL} uses a target-specific prefix.
9280 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_PREFIX
9281 Contains the prefix to be prepended to TLS initializer objects. The
9282 default of @code{NULL} uses a target-specific prefix.
9285 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_FIELDS (tree @var{type}, tree *@var{name})
9286 Specifies a function that generates the FIELD_DECLs for a TLS control
9287 object type. @var{type} is the RECORD_TYPE the fields are for and
9288 @var{name} should be filled with the structure tag, if the default of
9289 @code{__emutls_object} is unsuitable. The default creates a type suitable
9290 for libgcc's emulated TLS function.
9293 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_INIT (tree @var{var}, tree @var{decl}, tree @var{tmpl_addr})
9294 Specifies a function that generates the CONSTRUCTOR to initialize a
9295 TLS control object. @var{var} is the TLS control object, @var{decl}
9296 is the TLS object and @var{tmpl_addr} is the address of the
9297 initializer. The default initializes libgcc's emulated TLS control object.
9300 @deftypevr {Target Hook} {bool} TARGET_EMUTLS_VAR_ALIGN_FIXED
9301 Specifies whether the alignment of TLS control variable objects is
9302 fixed and should not be increased as some backends may do to optimize
9303 single objects. The default is false.
9306 @deftypevr {Target Hook} {bool} TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
9307 Specifies whether a DWARF @code{DW_OP_form_tls_address} location descriptor
9308 may be used to describe emulated TLS control objects.
9311 @node MIPS Coprocessors
9312 @section Defining coprocessor specifics for MIPS targets.
9313 @cindex MIPS coprocessor-definition macros
9315 The MIPS specification allows MIPS implementations to have as many as 4
9316 coprocessors, each with as many as 32 private registers. GCC supports
9317 accessing these registers and transferring values between the registers
9318 and memory using asm-ized variables. For example:
9321 register unsigned int cp0count asm ("c0r1");
9327 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
9328 names may be added as described below, or the default names may be
9329 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
9331 Coprocessor registers are assumed to be epilogue-used; sets to them will
9332 be preserved even if it does not appear that the register is used again
9333 later in the function.
9335 Another note: according to the MIPS spec, coprocessor 1 (if present) is
9336 the FPU@. One accesses COP1 registers through standard mips
9337 floating-point support; they are not included in this mechanism.
9339 There is one macro used in defining the MIPS coprocessor interface which
9340 you may want to override in subtargets; it is described below.
9342 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
9343 A comma-separated list (with leading comma) of pairs describing the
9344 alternate names of coprocessor registers. The format of each entry should be
9346 @{ @var{alternatename}, @var{register_number}@}
9352 @section Parameters for Precompiled Header Validity Checking
9353 @cindex parameters, precompiled headers
9355 @deftypefn {Target Hook} void *TARGET_GET_PCH_VALIDITY (size_t *@var{sz})
9356 This hook returns the data needed by @code{TARGET_PCH_VALID_P} and sets
9357 @samp{*@var{sz}} to the size of the data in bytes.
9360 @deftypefn {Target Hook} const char *TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz})
9361 This hook checks whether the options used to create a PCH file are
9362 compatible with the current settings. It returns @code{NULL}
9363 if so and a suitable error message if not. Error messages will
9364 be presented to the user and must be localized using @samp{_(@var{msg})}.
9366 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
9367 when the PCH file was created and @var{sz} is the size of that data in bytes.
9368 It's safe to assume that the data was created by the same version of the
9369 compiler, so no format checking is needed.
9371 The default definition of @code{default_pch_valid_p} should be
9372 suitable for most targets.
9375 @deftypefn {Target Hook} const char *TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags})
9376 If this hook is nonnull, the default implementation of
9377 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
9378 of @code{target_flags}. @var{pch_flags} specifies the value that
9379 @code{target_flags} had when the PCH file was created. The return
9380 value is the same as for @code{TARGET_PCH_VALID_P}.
9384 @section C++ ABI parameters
9385 @cindex parameters, c++ abi
9387 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
9388 Define this hook to override the integer type used for guard variables.
9389 These are used to implement one-time construction of static objects. The
9390 default is long_long_integer_type_node.
9393 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
9394 This hook determines how guard variables are used. It should return
9395 @code{false} (the default) if first byte should be used. A return value of
9396 @code{true} indicates the least significant bit should be used.
9399 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
9400 This hook returns the size of the cookie to use when allocating an array
9401 whose elements have the indicated @var{type}. Assumes that it is already
9402 known that a cookie is needed. The default is
9403 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
9404 IA64/Generic C++ ABI@.
9407 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
9408 This hook should return @code{true} if the element size should be stored in
9409 array cookies. The default is to return @code{false}.
9412 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
9413 If defined by a backend this hook allows the decision made to export
9414 class @var{type} to be overruled. Upon entry @var{import_export}
9415 will contain 1 if the class is going to be exported, @minus{}1 if it is going
9416 to be imported and 0 otherwise. This function should return the
9417 modified value and perform any other actions necessary to support the
9418 backend's targeted operating system.
9421 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
9422 This hook should return @code{true} if constructors and destructors return
9423 the address of the object created/destroyed. The default is to return
9427 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
9428 This hook returns true if the key method for a class (i.e., the method
9429 which, if defined in the current translation unit, causes the virtual
9430 table to be emitted) may be an inline function. Under the standard
9431 Itanium C++ ABI the key method may be an inline function so long as
9432 the function is not declared inline in the class definition. Under
9433 some variants of the ABI, an inline function can never be the key
9434 method. The default is to return @code{true}.
9437 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
9438 @var{decl} is a virtual table, virtual table table, typeinfo object,
9439 or other similar implicit class data object that will be emitted with
9440 external linkage in this translation unit. No ELF visibility has been
9441 explicitly specified. If the target needs to specify a visibility
9442 other than that of the containing class, use this hook to set
9443 @code{DECL_VISIBILITY} and @code{DECL_VISIBILITY_SPECIFIED}.
9446 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
9447 This hook returns true (the default) if virtual tables and other
9448 similar implicit class data objects are always COMDAT if they have
9449 external linkage. If this hook returns false, then class data for
9450 classes whose virtual table will be emitted in only one translation
9451 unit will not be COMDAT.
9454 @deftypefn {Target Hook} bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void)
9455 This hook returns true (the default) if the RTTI information for
9456 the basic types which is defined in the C++ runtime should always
9457 be COMDAT, false if it should not be COMDAT.
9460 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
9461 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
9462 should be used to register static destructors when @option{-fuse-cxa-atexit}
9463 is in effect. The default is to return false to use @code{__cxa_atexit}.
9466 @deftypefn {Target Hook} bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void)
9467 This hook returns true if the target @code{atexit} function can be used
9468 in the same manner as @code{__cxa_atexit} to register C++ static
9469 destructors. This requires that @code{atexit}-registered functions in
9470 shared libraries are run in the correct order when the libraries are
9471 unloaded. The default is to return false.
9474 @deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type})
9475 @var{type} is a C++ class (i.e., RECORD_TYPE or UNION_TYPE) that has just been
9476 defined. Use this hook to make adjustments to the class (eg, tweak
9477 visibility or perform any other required target modifications).
9481 @section Miscellaneous Parameters
9482 @cindex parameters, miscellaneous
9484 @c prevent bad page break with this line
9485 Here are several miscellaneous parameters.
9487 @defmac HAS_LONG_COND_BRANCH
9488 Define this boolean macro to indicate whether or not your architecture
9489 has conditional branches that can span all of memory. It is used in
9490 conjunction with an optimization that partitions hot and cold basic
9491 blocks into separate sections of the executable. If this macro is
9492 set to false, gcc will convert any conditional branches that attempt
9493 to cross between sections into unconditional branches or indirect jumps.
9496 @defmac HAS_LONG_UNCOND_BRANCH
9497 Define this boolean macro to indicate whether or not your architecture
9498 has unconditional branches that can span all of memory. It is used in
9499 conjunction with an optimization that partitions hot and cold basic
9500 blocks into separate sections of the executable. If this macro is
9501 set to false, gcc will convert any unconditional branches that attempt
9502 to cross between sections into indirect jumps.
9505 @defmac CASE_VECTOR_MODE
9506 An alias for a machine mode name. This is the machine mode that
9507 elements of a jump-table should have.
9510 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
9511 Optional: return the preferred mode for an @code{addr_diff_vec}
9512 when the minimum and maximum offset are known. If you define this,
9513 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
9514 To make this work, you also have to define @code{INSN_ALIGN} and
9515 make the alignment for @code{addr_diff_vec} explicit.
9516 The @var{body} argument is provided so that the offset_unsigned and scale
9517 flags can be updated.
9520 @defmac CASE_VECTOR_PC_RELATIVE
9521 Define this macro to be a C expression to indicate when jump-tables
9522 should contain relative addresses. You need not define this macro if
9523 jump-tables never contain relative addresses, or jump-tables should
9524 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
9528 @defmac CASE_VALUES_THRESHOLD
9529 Define this to be the smallest number of different values for which it
9530 is best to use a jump-table instead of a tree of conditional branches.
9531 The default is four for machines with a @code{casesi} instruction and
9532 five otherwise. This is best for most machines.
9535 @defmac CASE_USE_BIT_TESTS
9536 Define this macro to be a C expression to indicate whether C switch
9537 statements may be implemented by a sequence of bit tests. This is
9538 advantageous on processors that can efficiently implement left shift
9539 of 1 by the number of bits held in a register, but inappropriate on
9540 targets that would require a loop. By default, this macro returns
9541 @code{true} if the target defines an @code{ashlsi3} pattern, and
9542 @code{false} otherwise.
9545 @defmac WORD_REGISTER_OPERATIONS
9546 Define this macro if operations between registers with integral mode
9547 smaller than a word are always performed on the entire register.
9548 Most RISC machines have this property and most CISC machines do not.
9551 @defmac LOAD_EXTEND_OP (@var{mem_mode})
9552 Define this macro to be a C expression indicating when insns that read
9553 memory in @var{mem_mode}, an integral mode narrower than a word, set the
9554 bits outside of @var{mem_mode} to be either the sign-extension or the
9555 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
9556 of @var{mem_mode} for which the
9557 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
9558 @code{UNKNOWN} for other modes.
9560 This macro is not called with @var{mem_mode} non-integral or with a width
9561 greater than or equal to @code{BITS_PER_WORD}, so you may return any
9562 value in this case. Do not define this macro if it would always return
9563 @code{UNKNOWN}. On machines where this macro is defined, you will normally
9564 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
9566 You may return a non-@code{UNKNOWN} value even if for some hard registers
9567 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
9568 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
9569 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
9570 integral mode larger than this but not larger than @code{word_mode}.
9572 You must return @code{UNKNOWN} if for some hard registers that allow this
9573 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
9574 @code{word_mode}, but that they can change to another integral mode that
9575 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
9578 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
9579 Define this macro if loading short immediate values into registers sign
9583 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
9584 Define this macro if the same instructions that convert a floating
9585 point number to a signed fixed point number also convert validly to an
9589 @deftypefn {Target Hook} int TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (enum machine_mode @var{mode})
9590 When @option{-ffast-math} is in effect, GCC tries to optimize
9591 divisions by the same divisor, by turning them into multiplications by
9592 the reciprocal. This target hook specifies the minimum number of divisions
9593 that should be there for GCC to perform the optimization for a variable
9594 of mode @var{mode}. The default implementation returns 3 if the machine
9595 has an instruction for the division, and 2 if it does not.
9599 The maximum number of bytes that a single instruction can move quickly
9600 between memory and registers or between two memory locations.
9603 @defmac MAX_MOVE_MAX
9604 The maximum number of bytes that a single instruction can move quickly
9605 between memory and registers or between two memory locations. If this
9606 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
9607 constant value that is the largest value that @code{MOVE_MAX} can have
9611 @defmac SHIFT_COUNT_TRUNCATED
9612 A C expression that is nonzero if on this machine the number of bits
9613 actually used for the count of a shift operation is equal to the number
9614 of bits needed to represent the size of the object being shifted. When
9615 this macro is nonzero, the compiler will assume that it is safe to omit
9616 a sign-extend, zero-extend, and certain bitwise `and' instructions that
9617 truncates the count of a shift operation. On machines that have
9618 instructions that act on bit-fields at variable positions, which may
9619 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
9620 also enables deletion of truncations of the values that serve as
9621 arguments to bit-field instructions.
9623 If both types of instructions truncate the count (for shifts) and
9624 position (for bit-field operations), or if no variable-position bit-field
9625 instructions exist, you should define this macro.
9627 However, on some machines, such as the 80386 and the 680x0, truncation
9628 only applies to shift operations and not the (real or pretended)
9629 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
9630 such machines. Instead, add patterns to the @file{md} file that include
9631 the implied truncation of the shift instructions.
9633 You need not define this macro if it would always have the value of zero.
9636 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
9637 @deftypefn {Target Hook} int TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode})
9638 This function describes how the standard shift patterns for @var{mode}
9639 deal with shifts by negative amounts or by more than the width of the mode.
9640 @xref{shift patterns}.
9642 On many machines, the shift patterns will apply a mask @var{m} to the
9643 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
9644 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
9645 this is true for mode @var{mode}, the function should return @var{m},
9646 otherwise it should return 0. A return value of 0 indicates that no
9647 particular behavior is guaranteed.
9649 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
9650 @emph{not} apply to general shift rtxes; it applies only to instructions
9651 that are generated by the named shift patterns.
9653 The default implementation of this function returns
9654 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
9655 and 0 otherwise. This definition is always safe, but if
9656 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
9657 nevertheless truncate the shift count, you may get better code
9661 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
9662 A C expression which is nonzero if on this machine it is safe to
9663 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
9664 bits (where @var{outprec} is smaller than @var{inprec}) by merely
9665 operating on it as if it had only @var{outprec} bits.
9667 On many machines, this expression can be 1.
9669 @c rearranged this, removed the phrase "it is reported that". this was
9670 @c to fix an overfull hbox. --mew 10feb93
9671 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
9672 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
9673 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
9674 such cases may improve things.
9677 @deftypefn {Target Hook} int TARGET_MODE_REP_EXTENDED (enum machine_mode @var{mode}, enum machine_mode @var{rep_mode})
9678 The representation of an integral mode can be such that the values
9679 are always extended to a wider integral mode. Return
9680 @code{SIGN_EXTEND} if values of @var{mode} are represented in
9681 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
9682 otherwise. (Currently, none of the targets use zero-extended
9683 representation this way so unlike @code{LOAD_EXTEND_OP},
9684 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
9685 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
9686 @var{mode} to @var{mode_rep} so that @var{mode_rep} is not the next
9687 widest integral mode and currently we take advantage of this fact.)
9689 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
9690 value even if the extension is not performed on certain hard registers
9691 as long as for the @code{REGNO_REG_CLASS} of these hard registers
9692 @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero.
9694 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
9695 describe two related properties. If you define
9696 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
9697 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
9700 In order to enforce the representation of @code{mode},
9701 @code{TRULY_NOOP_TRUNCATION} should return false when truncating to
9705 @defmac STORE_FLAG_VALUE
9706 A C expression describing the value returned by a comparison operator
9707 with an integral mode and stored by a store-flag instruction
9708 (@samp{s@var{cond}}) when the condition is true. This description must
9709 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
9710 comparison operators whose results have a @code{MODE_INT} mode.
9712 A value of 1 or @minus{}1 means that the instruction implementing the
9713 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
9714 and 0 when the comparison is false. Otherwise, the value indicates
9715 which bits of the result are guaranteed to be 1 when the comparison is
9716 true. This value is interpreted in the mode of the comparison
9717 operation, which is given by the mode of the first operand in the
9718 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
9719 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
9722 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
9723 generate code that depends only on the specified bits. It can also
9724 replace comparison operators with equivalent operations if they cause
9725 the required bits to be set, even if the remaining bits are undefined.
9726 For example, on a machine whose comparison operators return an
9727 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
9728 @samp{0x80000000}, saying that just the sign bit is relevant, the
9732 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
9739 (ashift:SI @var{x} (const_int @var{n}))
9743 where @var{n} is the appropriate shift count to move the bit being
9744 tested into the sign bit.
9746 There is no way to describe a machine that always sets the low-order bit
9747 for a true value, but does not guarantee the value of any other bits,
9748 but we do not know of any machine that has such an instruction. If you
9749 are trying to port GCC to such a machine, include an instruction to
9750 perform a logical-and of the result with 1 in the pattern for the
9751 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
9753 Often, a machine will have multiple instructions that obtain a value
9754 from a comparison (or the condition codes). Here are rules to guide the
9755 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
9760 Use the shortest sequence that yields a valid definition for
9761 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
9762 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
9763 comparison operators to do so because there may be opportunities to
9764 combine the normalization with other operations.
9767 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
9768 slightly preferred on machines with expensive jumps and 1 preferred on
9772 As a second choice, choose a value of @samp{0x80000001} if instructions
9773 exist that set both the sign and low-order bits but do not define the
9777 Otherwise, use a value of @samp{0x80000000}.
9780 Many machines can produce both the value chosen for
9781 @code{STORE_FLAG_VALUE} and its negation in the same number of
9782 instructions. On those machines, you should also define a pattern for
9783 those cases, e.g., one matching
9786 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
9789 Some machines can also perform @code{and} or @code{plus} operations on
9790 condition code values with less instructions than the corresponding
9791 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
9792 machines, define the appropriate patterns. Use the names @code{incscc}
9793 and @code{decscc}, respectively, for the patterns which perform
9794 @code{plus} or @code{minus} operations on condition code values. See
9795 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
9796 find such instruction sequences on other machines.
9798 If this macro is not defined, the default value, 1, is used. You need
9799 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
9800 instructions, or if the value generated by these instructions is 1.
9803 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
9804 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
9805 returned when comparison operators with floating-point results are true.
9806 Define this macro on machines that have comparison operations that return
9807 floating-point values. If there are no such operations, do not define
9811 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
9812 A C expression that gives a rtx representing the nonzero true element
9813 for vector comparisons. The returned rtx should be valid for the inner
9814 mode of @var{mode} which is guaranteed to be a vector mode. Define
9815 this macro on machines that have vector comparison operations that
9816 return a vector result. If there are no such operations, do not define
9817 this macro. Typically, this macro is defined as @code{const1_rtx} or
9818 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
9819 the compiler optimizing such vector comparison operations for the
9823 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
9824 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
9825 A C expression that indicates whether the architecture defines a value
9826 for @code{clz} or @code{ctz} with a zero operand.
9827 A result of @code{0} indicates the value is undefined.
9828 If the value is defined for only the RTL expression, the macro should
9829 evaluate to @code{1}; if the value applies also to the corresponding optab
9830 entry (which is normally the case if it expands directly into
9831 the corresponding RTL), then the macro should evaluate to @code{2}.
9832 In the cases where the value is defined, @var{value} should be set to
9835 If this macro is not defined, the value of @code{clz} or
9836 @code{ctz} at zero is assumed to be undefined.
9838 This macro must be defined if the target's expansion for @code{ffs}
9839 relies on a particular value to get correct results. Otherwise it
9840 is not necessary, though it may be used to optimize some corner cases, and
9841 to provide a default expansion for the @code{ffs} optab.
9843 Note that regardless of this macro the ``definedness'' of @code{clz}
9844 and @code{ctz} at zero do @emph{not} extend to the builtin functions
9845 visible to the user. Thus one may be free to adjust the value at will
9846 to match the target expansion of these operations without fear of
9851 An alias for the machine mode for pointers. On most machines, define
9852 this to be the integer mode corresponding to the width of a hardware
9853 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
9854 On some machines you must define this to be one of the partial integer
9855 modes, such as @code{PSImode}.
9857 The width of @code{Pmode} must be at least as large as the value of
9858 @code{POINTER_SIZE}. If it is not equal, you must define the macro
9859 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
9863 @defmac FUNCTION_MODE
9864 An alias for the machine mode used for memory references to functions
9865 being called, in @code{call} RTL expressions. On most CISC machines,
9866 where an instruction can begin at any byte address, this should be
9867 @code{QImode}. On most RISC machines, where all instructions have fixed
9868 size and alignment, this should be a mode with the same size and alignment
9869 as the machine instruction words - typically @code{SImode} or @code{HImode}.
9872 @defmac STDC_0_IN_SYSTEM_HEADERS
9873 In normal operation, the preprocessor expands @code{__STDC__} to the
9874 constant 1, to signify that GCC conforms to ISO Standard C@. On some
9875 hosts, like Solaris, the system compiler uses a different convention,
9876 where @code{__STDC__} is normally 0, but is 1 if the user specifies
9877 strict conformance to the C Standard.
9879 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
9880 convention when processing system header files, but when processing user
9881 files @code{__STDC__} will always expand to 1.
9884 @defmac NO_IMPLICIT_EXTERN_C
9885 Define this macro if the system header files support C++ as well as C@.
9886 This macro inhibits the usual method of using system header files in
9887 C++, which is to pretend that the file's contents are enclosed in
9888 @samp{extern "C" @{@dots{}@}}.
9893 @defmac REGISTER_TARGET_PRAGMAS ()
9894 Define this macro if you want to implement any target-specific pragmas.
9895 If defined, it is a C expression which makes a series of calls to
9896 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
9897 for each pragma. The macro may also do any
9898 setup required for the pragmas.
9900 The primary reason to define this macro is to provide compatibility with
9901 other compilers for the same target. In general, we discourage
9902 definition of target-specific pragmas for GCC@.
9904 If the pragma can be implemented by attributes then you should consider
9905 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
9907 Preprocessor macros that appear on pragma lines are not expanded. All
9908 @samp{#pragma} directives that do not match any registered pragma are
9909 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
9912 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9913 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9915 Each call to @code{c_register_pragma} or
9916 @code{c_register_pragma_with_expansion} establishes one pragma. The
9917 @var{callback} routine will be called when the preprocessor encounters a
9921 #pragma [@var{space}] @var{name} @dots{}
9924 @var{space} is the case-sensitive namespace of the pragma, or
9925 @code{NULL} to put the pragma in the global namespace. The callback
9926 routine receives @var{pfile} as its first argument, which can be passed
9927 on to cpplib's functions if necessary. You can lex tokens after the
9928 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
9929 callback will be silently ignored. The end of the line is indicated by
9930 a token of type @code{CPP_EOF}. Macro expansion occurs on the
9931 arguments of pragmas registered with
9932 @code{c_register_pragma_with_expansion} but not on the arguments of
9933 pragmas registered with @code{c_register_pragma}.
9935 Note that the use of @code{pragma_lex} is specific to the C and C++
9936 compilers. It will not work in the Java or Fortran compilers, or any
9937 other language compilers for that matter. Thus if @code{pragma_lex} is going
9938 to be called from target-specific code, it must only be done so when
9939 building the C and C++ compilers. This can be done by defining the
9940 variables @code{c_target_objs} and @code{cxx_target_objs} in the
9941 target entry in the @file{config.gcc} file. These variables should name
9942 the target-specific, language-specific object file which contains the
9943 code that uses @code{pragma_lex}. Note it will also be necessary to add a
9944 rule to the makefile fragment pointed to by @code{tmake_file} that shows
9945 how to build this object file.
9950 @defmac HANDLE_SYSV_PRAGMA
9951 Define this macro (to a value of 1) if you want the System V style
9952 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
9953 [=<value>]} to be supported by gcc.
9955 The pack pragma specifies the maximum alignment (in bytes) of fields
9956 within a structure, in much the same way as the @samp{__aligned__} and
9957 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
9958 the behavior to the default.
9960 A subtlety for Microsoft Visual C/C++ style bit-field packing
9961 (e.g.@: -mms-bitfields) for targets that support it:
9962 When a bit-field is inserted into a packed record, the whole size
9963 of the underlying type is used by one or more same-size adjacent
9964 bit-fields (that is, if its long:3, 32 bits is used in the record,
9965 and any additional adjacent long bit-fields are packed into the same
9966 chunk of 32 bits. However, if the size changes, a new field of that
9969 If both MS bit-fields and @samp{__attribute__((packed))} are used,
9970 the latter will take precedence. If @samp{__attribute__((packed))} is
9971 used on a single field when MS bit-fields are in use, it will take
9972 precedence for that field, but the alignment of the rest of the structure
9973 may affect its placement.
9975 The weak pragma only works if @code{SUPPORTS_WEAK} and
9976 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
9977 of specifically named weak labels, optionally with a value.
9982 @defmac HANDLE_PRAGMA_PACK_PUSH_POP
9983 Define this macro (to a value of 1) if you want to support the Win32
9984 style pragmas @samp{#pragma pack(push[,@var{n}])} and @samp{#pragma
9985 pack(pop)}. The @samp{pack(push,[@var{n}])} pragma specifies the maximum
9986 alignment (in bytes) of fields within a structure, in much the same way as
9987 the @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
9988 pack value of zero resets the behavior to the default. Successive
9989 invocations of this pragma cause the previous values to be stacked, so
9990 that invocations of @samp{#pragma pack(pop)} will return to the previous
9994 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
9995 Define this macro, as well as
9996 @code{HANDLE_SYSV_PRAGMA}, if macros should be expanded in the
9997 arguments of @samp{#pragma pack}.
10000 @defmac TARGET_DEFAULT_PACK_STRUCT
10001 If your target requires a structure packing default other than 0 (meaning
10002 the machine default), define this macro to the necessary value (in bytes).
10003 This must be a value that would also be valid to use with
10004 @samp{#pragma pack()} (that is, a small power of two).
10009 @defmac HANDLE_PRAGMA_PUSH_POP_MACRO
10010 Define this macro if you want to support the Win32 style pragmas
10011 @samp{#pragma push_macro(macro-name-as-string)} and @samp{#pragma
10012 pop_macro(macro-name-as-string)}. The @samp{#pragma push_macro(
10013 macro-name-as-string)} pragma saves the named macro and via
10014 @samp{#pragma pop_macro(macro-name-as-string)} it will return to the
10019 @defmac DOLLARS_IN_IDENTIFIERS
10020 Define this macro to control use of the character @samp{$} in
10021 identifier names for the C family of languages. 0 means @samp{$} is
10022 not allowed by default; 1 means it is allowed. 1 is the default;
10023 there is no need to define this macro in that case.
10026 @defmac NO_DOLLAR_IN_LABEL
10027 Define this macro if the assembler does not accept the character
10028 @samp{$} in label names. By default constructors and destructors in
10029 G++ have @samp{$} in the identifiers. If this macro is defined,
10030 @samp{.} is used instead.
10033 @defmac NO_DOT_IN_LABEL
10034 Define this macro if the assembler does not accept the character
10035 @samp{.} in label names. By default constructors and destructors in G++
10036 have names that use @samp{.}. If this macro is defined, these names
10037 are rewritten to avoid @samp{.}.
10040 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
10041 Define this macro as a C expression that is nonzero if it is safe for the
10042 delay slot scheduler to place instructions in the delay slot of @var{insn},
10043 even if they appear to use a resource set or clobbered in @var{insn}.
10044 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
10045 every @code{call_insn} has this behavior. On machines where some @code{insn}
10046 or @code{jump_insn} is really a function call and hence has this behavior,
10047 you should define this macro.
10049 You need not define this macro if it would always return zero.
10052 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
10053 Define this macro as a C expression that is nonzero if it is safe for the
10054 delay slot scheduler to place instructions in the delay slot of @var{insn},
10055 even if they appear to set or clobber a resource referenced in @var{insn}.
10056 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
10057 some @code{insn} or @code{jump_insn} is really a function call and its operands
10058 are registers whose use is actually in the subroutine it calls, you should
10059 define this macro. Doing so allows the delay slot scheduler to move
10060 instructions which copy arguments into the argument registers into the delay
10061 slot of @var{insn}.
10063 You need not define this macro if it would always return zero.
10066 @defmac MULTIPLE_SYMBOL_SPACES
10067 Define this macro as a C expression that is nonzero if, in some cases,
10068 global symbols from one translation unit may not be bound to undefined
10069 symbols in another translation unit without user intervention. For
10070 instance, under Microsoft Windows symbols must be explicitly imported
10071 from shared libraries (DLLs).
10073 You need not define this macro if it would always evaluate to zero.
10076 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{outputs}, tree @var{inputs}, tree @var{clobbers})
10077 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
10078 any hard regs the port wishes to automatically clobber for an asm.
10079 It should return the result of the last @code{tree_cons} used to add a
10080 clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the
10081 corresponding parameters to the asm and may be inspected to avoid
10082 clobbering a register that is an input or output of the asm. You can use
10083 @code{tree_overlaps_hard_reg_set}, declared in @file{tree.h}, to test
10084 for overlap with regards to asm-declared registers.
10087 @defmac MATH_LIBRARY
10088 Define this macro as a C string constant for the linker argument to link
10089 in the system math library, or @samp{""} if the target does not have a
10090 separate math library.
10092 You need only define this macro if the default of @samp{"-lm"} is wrong.
10095 @defmac LIBRARY_PATH_ENV
10096 Define this macro as a C string constant for the environment variable that
10097 specifies where the linker should look for libraries.
10099 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
10103 @defmac TARGET_POSIX_IO
10104 Define this macro if the target supports the following POSIX@ file
10105 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
10106 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
10107 to use file locking when exiting a program, which avoids race conditions
10108 if the program has forked. It will also create directories at run-time
10109 for cross-profiling.
10112 @defmac MAX_CONDITIONAL_EXECUTE
10114 A C expression for the maximum number of instructions to execute via
10115 conditional execution instructions instead of a branch. A value of
10116 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
10117 1 if it does use cc0.
10120 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
10121 Used if the target needs to perform machine-dependent modifications on the
10122 conditionals used for turning basic blocks into conditionally executed code.
10123 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
10124 contains information about the currently processed blocks. @var{true_expr}
10125 and @var{false_expr} are the tests that are used for converting the
10126 then-block and the else-block, respectively. Set either @var{true_expr} or
10127 @var{false_expr} to a null pointer if the tests cannot be converted.
10130 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
10131 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
10132 if-statements into conditions combined by @code{and} and @code{or} operations.
10133 @var{bb} contains the basic block that contains the test that is currently
10134 being processed and about to be turned into a condition.
10137 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
10138 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
10139 be converted to conditional execution format. @var{ce_info} points to
10140 a data structure, @code{struct ce_if_block}, which contains information
10141 about the currently processed blocks.
10144 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
10145 A C expression to perform any final machine dependent modifications in
10146 converting code to conditional execution. The involved basic blocks
10147 can be found in the @code{struct ce_if_block} structure that is pointed
10148 to by @var{ce_info}.
10151 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
10152 A C expression to cancel any machine dependent modifications in
10153 converting code to conditional execution. The involved basic blocks
10154 can be found in the @code{struct ce_if_block} structure that is pointed
10155 to by @var{ce_info}.
10158 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
10159 A C expression to initialize any extra fields in a @code{struct ce_if_block}
10160 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
10163 @defmac IFCVT_EXTRA_FIELDS
10164 If defined, it should expand to a set of field declarations that will be
10165 added to the @code{struct ce_if_block} structure. These should be initialized
10166 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
10169 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG ()
10170 If non-null, this hook performs a target-specific pass over the
10171 instruction stream. The compiler will run it at all optimization levels,
10172 just before the point at which it normally does delayed-branch scheduling.
10174 The exact purpose of the hook varies from target to target. Some use
10175 it to do transformations that are necessary for correctness, such as
10176 laying out in-function constant pools or avoiding hardware hazards.
10177 Others use it as an opportunity to do some machine-dependent optimizations.
10179 You need not implement the hook if it has nothing to do. The default
10180 definition is null.
10183 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
10184 Define this hook if you have any machine-specific built-in functions
10185 that need to be defined. It should be a function that performs the
10188 Machine specific built-in functions can be useful to expand special machine
10189 instructions that would otherwise not normally be generated because
10190 they have no equivalent in the source language (for example, SIMD vector
10191 instructions or prefetch instructions).
10193 To create a built-in function, call the function
10194 @code{lang_hooks.builtin_function}
10195 which is defined by the language front end. You can use any type nodes set
10196 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
10197 only language front ends that use those two functions will call
10198 @samp{TARGET_INIT_BUILTINS}.
10201 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
10203 Expand a call to a machine specific built-in function that was set up by
10204 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
10205 function call; the result should go to @var{target} if that is
10206 convenient, and have mode @var{mode} if that is convenient.
10207 @var{subtarget} may be used as the target for computing one of
10208 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
10209 ignored. This function should return the result of the call to the
10213 @deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (tree @var{fndecl}, tree @var{arglist})
10215 Select a replacement for a machine specific built-in function that
10216 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
10217 @emph{before} regular type checking, and so allows the target to
10218 implement a crude form of function overloading. @var{fndecl} is the
10219 declaration of the built-in function. @var{arglist} is the list of
10220 arguments passed to the built-in function. The result is a
10221 complete expression that implements the operation, usually
10222 another @code{CALL_EXPR}.
10225 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, tree @var{arglist}, bool @var{ignore})
10227 Fold a call to a machine specific built-in function that was set up by
10228 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
10229 built-in function. @var{arglist} is the list of arguments passed to
10230 the built-in function. The result is another tree containing a
10231 simplified expression for the call's result. If @var{ignore} is true
10232 the value will be ignored.
10235 @deftypefn {Target Hook} const char * TARGET_INVALID_WITHIN_DOLOOP (rtx @var{insn})
10237 Take an instruction in @var{insn} and return NULL if it is valid within a
10238 low-overhead loop, otherwise return a string why doloop could not be applied.
10240 Many targets use special registers for low-overhead looping. For any
10241 instruction that clobbers these this function should return a string indicating
10242 the reason why the doloop could not be applied.
10243 By default, the RTL loop optimizer does not use a present doloop pattern for
10244 loops containing function calls or branch on table instructions.
10247 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
10249 Take a branch insn in @var{branch1} and another in @var{branch2}.
10250 Return true if redirecting @var{branch1} to the destination of
10251 @var{branch2} is possible.
10253 On some targets, branches may have a limited range. Optimizing the
10254 filling of delay slots can result in branches being redirected, and this
10255 may in turn cause a branch offset to overflow.
10258 @deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (rtx @var{x}, @var{outer_code})
10259 This target hook returns @code{true} if @var{x} is considered to be commutative.
10260 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
10261 PLUS to be commutative inside a MEM@. @var{outer_code} is the rtx code
10262 of the enclosing rtl, if known, otherwise it is UNKNOWN.
10265 @deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg})
10267 When the initial value of a hard register has been copied in a pseudo
10268 register, it is often not necessary to actually allocate another register
10269 to this pseudo register, because the original hard register or a stack slot
10270 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
10271 is called at the start of register allocation once for each hard register
10272 that had its initial value copied by using
10273 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
10274 Possible values are @code{NULL_RTX}, if you don't want
10275 to do any special allocation, a @code{REG} rtx---that would typically be
10276 the hard register itself, if it is known not to be clobbered---or a
10278 If you are returning a @code{MEM}, this is only a hint for the allocator;
10279 it might decide to use another register anyways.
10280 You may use @code{current_function_leaf_function} in the hook, functions
10281 that use @code{REG_N_SETS}, to determine if the hard
10282 register in question will not be clobbered.
10283 The default value of this hook is @code{NULL}, which disables any special
10287 @deftypefn {Target Hook} int TARGET_UNSPEC_MAY_TRAP_P (const_rtx @var{x}, unsigned @var{flags})
10288 This target hook returns nonzero if @var{x}, an @code{unspec} or
10289 @code{unspec_volatile} operation, might cause a trap. Targets can use
10290 this hook to enhance precision of analysis for @code{unspec} and
10291 @code{unspec_volatile} operations. You may call @code{may_trap_p_1}
10292 to analyze inner elements of @var{x} in which case @var{flags} should be
10296 @deftypefn {Target Hook} void TARGET_SET_CURRENT_FUNCTION (tree @var{decl})
10297 The compiler invokes this hook whenever it changes its current function
10298 context (@code{cfun}). You can define this function if
10299 the back end needs to perform any initialization or reset actions on a
10300 per-function basis. For example, it may be used to implement function
10301 attributes that affect register usage or code generation patterns.
10302 The argument @var{decl} is the declaration for the new function context,
10303 and may be null to indicate that the compiler has left a function context
10304 and is returning to processing at the top level.
10305 The default hook function does nothing.
10307 GCC sets @code{cfun} to a dummy function context during initialization of
10308 some parts of the back end. The hook function is not invoked in this
10309 situation; you need not worry about the hook being invoked recursively,
10310 or when the back end is in a partially-initialized state.
10313 @defmac TARGET_OBJECT_SUFFIX
10314 Define this macro to be a C string representing the suffix for object
10315 files on your target machine. If you do not define this macro, GCC will
10316 use @samp{.o} as the suffix for object files.
10319 @defmac TARGET_EXECUTABLE_SUFFIX
10320 Define this macro to be a C string representing the suffix to be
10321 automatically added to executable files on your target machine. If you
10322 do not define this macro, GCC will use the null string as the suffix for
10326 @defmac COLLECT_EXPORT_LIST
10327 If defined, @code{collect2} will scan the individual object files
10328 specified on its command line and create an export list for the linker.
10329 Define this macro for systems like AIX, where the linker discards
10330 object files that are not referenced from @code{main} and uses export
10334 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
10335 Define this macro to a C expression representing a variant of the
10336 method call @var{mdecl}, if Java Native Interface (JNI) methods
10337 must be invoked differently from other methods on your target.
10338 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
10339 the @code{stdcall} calling convention and this macro is then
10340 defined as this expression:
10343 build_type_attribute_variant (@var{mdecl},
10345 (get_identifier ("stdcall"),
10350 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
10351 This target hook returns @code{true} past the point in which new jump
10352 instructions could be created. On machines that require a register for
10353 every jump such as the SHmedia ISA of SH5, this point would typically be
10354 reload, so this target hook should be defined to a function such as:
10358 cannot_modify_jumps_past_reload_p ()
10360 return (reload_completed || reload_in_progress);
10365 @deftypefn {Target Hook} int TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
10366 This target hook returns a register class for which branch target register
10367 optimizations should be applied. All registers in this class should be
10368 usable interchangeably. After reload, registers in this class will be
10369 re-allocated and loads will be hoisted out of loops and be subjected
10370 to inter-block scheduling.
10373 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
10374 Branch target register optimization will by default exclude callee-saved
10376 that are not already live during the current function; if this target hook
10377 returns true, they will be included. The target code must than make sure
10378 that all target registers in the class returned by
10379 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
10380 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
10381 epilogues have already been generated. Note, even if you only return
10382 true when @var{after_prologue_epilogue_gen} is false, you still are likely
10383 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
10384 to reserve space for caller-saved target registers.
10387 @defmac POWI_MAX_MULTS
10388 If defined, this macro is interpreted as a signed integer C expression
10389 that specifies the maximum number of floating point multiplications
10390 that should be emitted when expanding exponentiation by an integer
10391 constant inline. When this value is defined, exponentiation requiring
10392 more than this number of multiplications is implemented by calling the
10393 system library's @code{pow}, @code{powf} or @code{powl} routines.
10394 The default value places no upper bound on the multiplication count.
10397 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
10398 This target hook should register any extra include files for the
10399 target. The parameter @var{stdinc} indicates if normal include files
10400 are present. The parameter @var{sysroot} is the system root directory.
10401 The parameter @var{iprefix} is the prefix for the gcc directory.
10404 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
10405 This target hook should register any extra include files for the
10406 target before any standard headers. The parameter @var{stdinc}
10407 indicates if normal include files are present. The parameter
10408 @var{sysroot} is the system root directory. The parameter
10409 @var{iprefix} is the prefix for the gcc directory.
10412 @deftypefn Macro void TARGET_OPTF (char *@var{path})
10413 This target hook should register special include paths for the target.
10414 The parameter @var{path} is the include to register. On Darwin
10415 systems, this is used for Framework includes, which have semantics
10416 that are different from @option{-I}.
10419 @deftypefn {Target Hook} bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
10420 This target hook returns @code{true} if it is safe to use a local alias
10421 for a virtual function @var{fndecl} when constructing thunks,
10422 @code{false} otherwise. By default, the hook returns @code{true} for all
10423 functions, if a target supports aliases (i.e.@: defines
10424 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
10427 @defmac TARGET_FORMAT_TYPES
10428 If defined, this macro is the name of a global variable containing
10429 target-specific format checking information for the @option{-Wformat}
10430 option. The default is to have no target-specific format checks.
10433 @defmac TARGET_N_FORMAT_TYPES
10434 If defined, this macro is the number of entries in
10435 @code{TARGET_FORMAT_TYPES}.
10438 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
10439 If defined, this macro is the name of a global variable containing
10440 target-specific format overrides for the @option{-Wformat} option. The
10441 default is to have no target-specific format overrides. If defined,
10442 @code{TARGET_FORMAT_TYPES} must be defined, too.
10445 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
10446 If defined, this macro specifies the number of entries in
10447 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
10450 @deftypefn {Target Hook} bool TARGET_RELAXED_ORDERING
10451 If set to @code{true}, means that the target's memory model does not
10452 guarantee that loads which do not depend on one another will access
10453 main memory in the order of the instruction stream; if ordering is
10454 important, an explicit memory barrier must be used. This is true of
10455 many recent processors which implement a policy of ``relaxed,''
10456 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
10457 and ia64. The default is @code{false}.
10460 @deftypefn {Target Hook} const char *TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (tree @var{typelist}, tree @var{funcdecl}, tree @var{val})
10461 If defined, this macro returns the diagnostic message when it is
10462 illegal to pass argument @var{val} to function @var{funcdecl}
10463 with prototype @var{typelist}.
10466 @deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (tree @var{fromtype}, tree @var{totype})
10467 If defined, this macro returns the diagnostic message when it is
10468 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
10469 if validity should be determined by the front end.
10472 @deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, tree @var{type})
10473 If defined, this macro returns the diagnostic message when it is
10474 invalid to apply operation @var{op} (where unary plus is denoted by
10475 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
10476 if validity should be determined by the front end.
10479 @deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, tree @var{type1}, tree @var{type2})
10480 If defined, this macro returns the diagnostic message when it is
10481 invalid to apply operation @var{op} to operands of types @var{type1}
10482 and @var{type2}, or @code{NULL} if validity should be determined by
10486 @defmac TARGET_USE_JCR_SECTION
10487 This macro determines whether to use the JCR section to register Java
10488 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
10489 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.
10493 This macro determines the size of the objective C jump buffer for the
10494 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
10497 @defmac LIBGCC2_UNWIND_ATTRIBUTE
10498 Define this macro if any target-specific attributes need to be attached
10499 to the functions in @file{libgcc} that provide low-level support for
10500 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
10501 and the associated definitions of those functions.
10504 @deftypefn {Target Hook} {bool} TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS (void)
10505 When optimization is disabled, this hook indicates whether or not
10506 arguments should be allocated to stack slots. Normally, GCC allocates
10507 stacks slots for arguments when not optimizing in order to make
10508 debugging easier. However, when a function is declared with
10509 @code{__attribute__((naked))}, there is no stack frame, and the compiler
10510 cannot safely move arguments from the registers in which they are passed
10511 to the stack. Therefore, this hook should return true in general, but
10512 false for naked functions. The default implementation always returns true.