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 This macro is run once at program startup and when the optimization
837 options are changed via @code{#pragma GCC optimize} or by using the
838 @code{optimize} attribute.
840 @strong{Do not examine @code{write_symbols} in
841 this macro!} The debugging options are not supposed to alter the
845 @deftypefn {Target Hook} bool TARGET_HELP (void)
846 This hook is called in response to the user invoking
847 @option{--target-help} on the command line. It gives the target a
848 chance to display extra information on the target specific command
849 line options found in its @file{.opt} file.
852 @defmac CAN_DEBUG_WITHOUT_FP
853 Define this macro if debugging can be performed even without a frame
854 pointer. If this macro is defined, GCC will turn on the
855 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
858 @node Per-Function Data
859 @section Defining data structures for per-function information.
860 @cindex per-function data
861 @cindex data structures
863 If the target needs to store information on a per-function basis, GCC
864 provides a macro and a couple of variables to allow this. Note, just
865 using statics to store the information is a bad idea, since GCC supports
866 nested functions, so you can be halfway through encoding one function
867 when another one comes along.
869 GCC defines a data structure called @code{struct function} which
870 contains all of the data specific to an individual function. This
871 structure contains a field called @code{machine} whose type is
872 @code{struct machine_function *}, which can be used by targets to point
873 to their own specific data.
875 If a target needs per-function specific data it should define the type
876 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
877 This macro should be used to initialize the function pointer
878 @code{init_machine_status}. This pointer is explained below.
880 One typical use of per-function, target specific data is to create an
881 RTX to hold the register containing the function's return address. This
882 RTX can then be used to implement the @code{__builtin_return_address}
883 function, for level 0.
885 Note---earlier implementations of GCC used a single data area to hold
886 all of the per-function information. Thus when processing of a nested
887 function began the old per-function data had to be pushed onto a
888 stack, and when the processing was finished, it had to be popped off the
889 stack. GCC used to provide function pointers called
890 @code{save_machine_status} and @code{restore_machine_status} to handle
891 the saving and restoring of the target specific information. Since the
892 single data area approach is no longer used, these pointers are no
895 @defmac INIT_EXPANDERS
896 Macro called to initialize any target specific information. This macro
897 is called once per function, before generation of any RTL has begun.
898 The intention of this macro is to allow the initialization of the
899 function pointer @code{init_machine_status}.
902 @deftypevar {void (*)(struct function *)} init_machine_status
903 If this function pointer is non-@code{NULL} it will be called once per
904 function, before function compilation starts, in order to allow the
905 target to perform any target specific initialization of the
906 @code{struct function} structure. It is intended that this would be
907 used to initialize the @code{machine} of that structure.
909 @code{struct machine_function} structures are expected to be freed by GC@.
910 Generally, any memory that they reference must be allocated by using
911 @code{ggc_alloc}, including the structure itself.
915 @section Storage Layout
916 @cindex storage layout
918 Note that the definitions of the macros in this table which are sizes or
919 alignments measured in bits do not need to be constant. They can be C
920 expressions that refer to static variables, such as the @code{target_flags}.
921 @xref{Run-time Target}.
923 @defmac BITS_BIG_ENDIAN
924 Define this macro to have the value 1 if the most significant bit in a
925 byte has the lowest number; otherwise define it to have the value zero.
926 This means that bit-field instructions count from the most significant
927 bit. If the machine has no bit-field instructions, then this must still
928 be defined, but it doesn't matter which value it is defined to. This
929 macro need not be a constant.
931 This macro does not affect the way structure fields are packed into
932 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
935 @defmac BYTES_BIG_ENDIAN
936 Define this macro to have the value 1 if the most significant byte in a
937 word has the lowest number. This macro need not be a constant.
940 @defmac WORDS_BIG_ENDIAN
941 Define this macro to have the value 1 if, in a multiword object, the
942 most significant word has the lowest number. This applies to both
943 memory locations and registers; GCC fundamentally assumes that the
944 order of words in memory is the same as the order in registers. This
945 macro need not be a constant.
948 @defmac LIBGCC2_WORDS_BIG_ENDIAN
949 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
950 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
951 used only when compiling @file{libgcc2.c}. Typically the value will be set
952 based on preprocessor defines.
955 @defmac FLOAT_WORDS_BIG_ENDIAN
956 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
957 @code{TFmode} floating point numbers are stored in memory with the word
958 containing the sign bit at the lowest address; otherwise define it to
959 have the value 0. This macro need not be a constant.
961 You need not define this macro if the ordering is the same as for
965 @defmac BITS_PER_UNIT
966 Define this macro to be the number of bits in an addressable storage
967 unit (byte). If you do not define this macro the default is 8.
970 @defmac BITS_PER_WORD
971 Number of bits in a word. If you do not define this macro, the default
972 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
975 @defmac MAX_BITS_PER_WORD
976 Maximum number of bits in a word. If this is undefined, the default is
977 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
978 largest value that @code{BITS_PER_WORD} can have at run-time.
981 @defmac UNITS_PER_WORD
982 Number of storage units in a word; normally the size of a general-purpose
983 register, a power of two from 1 or 8.
986 @defmac MIN_UNITS_PER_WORD
987 Minimum number of units in a word. If this is undefined, the default is
988 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
989 smallest value that @code{UNITS_PER_WORD} can have at run-time.
992 @defmac UNITS_PER_SIMD_WORD (@var{mode})
993 Number of units in the vectors that the vectorizer can produce for
994 scalar mode @var{mode}. The default is equal to @code{UNITS_PER_WORD},
995 because the vectorizer can do some transformations even in absence of
996 specialized @acronym{SIMD} hardware.
1000 Width of a pointer, in bits. You must specify a value no wider than the
1001 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
1002 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
1003 a value the default is @code{BITS_PER_WORD}.
1006 @defmac POINTERS_EXTEND_UNSIGNED
1007 A C expression that determines how pointers should be extended from
1008 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
1009 greater than zero if pointers should be zero-extended, zero if they
1010 should be sign-extended, and negative if some other sort of conversion
1011 is needed. In the last case, the extension is done by the target's
1012 @code{ptr_extend} instruction.
1014 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
1015 and @code{word_mode} are all the same width.
1018 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
1019 A macro to update @var{m} and @var{unsignedp} when an object whose type
1020 is @var{type} and which has the specified mode and signedness is to be
1021 stored in a register. This macro is only called when @var{type} is a
1024 On most RISC machines, which only have operations that operate on a full
1025 register, define this macro to set @var{m} to @code{word_mode} if
1026 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
1027 cases, only integer modes should be widened because wider-precision
1028 floating-point operations are usually more expensive than their narrower
1031 For most machines, the macro definition does not change @var{unsignedp}.
1032 However, some machines, have instructions that preferentially handle
1033 either signed or unsigned quantities of certain modes. For example, on
1034 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
1035 sign-extend the result to 64 bits. On such machines, set
1036 @var{unsignedp} according to which kind of extension is more efficient.
1038 Do not define this macro if it would never modify @var{m}.
1041 @defmac PROMOTE_FUNCTION_MODE
1042 Like @code{PROMOTE_MODE}, but is applied to outgoing function arguments or
1043 function return values, as specified by @code{TARGET_PROMOTE_FUNCTION_ARGS}
1044 and @code{TARGET_PROMOTE_FUNCTION_RETURN}, respectively.
1046 The default is @code{PROMOTE_MODE}.
1049 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_ARGS (tree @var{fntype})
1050 This target hook should return @code{true} if the promotion described by
1051 @code{PROMOTE_FUNCTION_MODE} should be done for outgoing function
1055 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_RETURN (tree @var{fntype})
1056 This target hook should return @code{true} if the promotion described by
1057 @code{PROMOTE_FUNCTION_MODE} should be done for the return value of
1060 If this target hook returns @code{true}, @code{TARGET_FUNCTION_VALUE}
1061 must perform the same promotions done by @code{PROMOTE_FUNCTION_MODE}.
1064 @defmac PARM_BOUNDARY
1065 Normal alignment required for function parameters on the stack, in
1066 bits. All stack parameters receive at least this much alignment
1067 regardless of data type. On most machines, this is the same as the
1071 @defmac STACK_BOUNDARY
1072 Define this macro to the minimum alignment enforced by hardware for the
1073 stack pointer on this machine. The definition is a C expression for the
1074 desired alignment (measured in bits). This value is used as a default
1075 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1076 this should be the same as @code{PARM_BOUNDARY}.
1079 @defmac PREFERRED_STACK_BOUNDARY
1080 Define this macro if you wish to preserve a certain alignment for the
1081 stack pointer, greater than what the hardware enforces. The definition
1082 is a C expression for the desired alignment (measured in bits). This
1083 macro must evaluate to a value equal to or larger than
1084 @code{STACK_BOUNDARY}.
1087 @defmac INCOMING_STACK_BOUNDARY
1088 Define this macro if the incoming stack boundary may be different
1089 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
1090 to a value equal to or larger than @code{STACK_BOUNDARY}.
1093 @defmac FUNCTION_BOUNDARY
1094 Alignment required for a function entry point, in bits.
1097 @defmac BIGGEST_ALIGNMENT
1098 Biggest alignment that any data type can require on this machine, in
1099 bits. Note that this is not the biggest alignment that is supported,
1100 just the biggest alignment that, when violated, may cause a fault.
1103 @defmac MALLOC_ABI_ALIGNMENT
1104 Alignment, in bits, a C conformant malloc implementation has to
1105 provide. If not defined, the default value is @code{BITS_PER_WORD}.
1108 @defmac MINIMUM_ATOMIC_ALIGNMENT
1109 If defined, the smallest alignment, in bits, that can be given to an
1110 object that can be referenced in one operation, without disturbing any
1111 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1112 on machines that don't have byte or half-word store operations.
1115 @defmac BIGGEST_FIELD_ALIGNMENT
1116 Biggest alignment that any structure or union field can require on this
1117 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1118 structure and union fields only, unless the field alignment has been set
1119 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1122 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1123 An expression for the alignment of a structure field @var{field} if the
1124 alignment computed in the usual way (including applying of
1125 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1126 alignment) is @var{computed}. It overrides alignment only if the
1127 field alignment has not been set by the
1128 @code{__attribute__ ((aligned (@var{n})))} construct.
1131 @defmac MAX_STACK_ALIGNMENT
1132 Biggest stack alignment guaranteed by the backend. Use this macro
1133 to specify the maximum alignment of a variable on stack.
1135 If not defined, the default value is @code{STACK_BOUNDARY}.
1137 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1138 @c But the fix for PR 32893 indicates that we can only guarantee
1139 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1140 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1143 @defmac MAX_OFILE_ALIGNMENT
1144 Biggest alignment supported by the object file format of this machine.
1145 Use this macro to limit the alignment which can be specified using the
1146 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1147 the default value is @code{BIGGEST_ALIGNMENT}.
1149 On systems that use ELF, the default (in @file{config/elfos.h}) is
1150 the largest supported 32-bit ELF section alignment representable on
1151 a 32-bit host e.g. @samp{(((unsigned HOST_WIDEST_INT) 1 << 28) * 8)}.
1152 On 32-bit ELF the largest supported section alignment in bits is
1153 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1156 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1157 If defined, a C expression to compute the alignment for a variable in
1158 the static store. @var{type} is the data type, and @var{basic-align} is
1159 the alignment that the object would ordinarily have. The value of this
1160 macro is used instead of that alignment to align the object.
1162 If this macro is not defined, then @var{basic-align} is used.
1165 One use of this macro is to increase alignment of medium-size data to
1166 make it all fit in fewer cache lines. Another is to cause character
1167 arrays to be word-aligned so that @code{strcpy} calls that copy
1168 constants to character arrays can be done inline.
1171 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1172 If defined, a C expression to compute the alignment given to a constant
1173 that is being placed in memory. @var{constant} is the constant and
1174 @var{basic-align} is the alignment that the object would ordinarily
1175 have. The value of this macro is used instead of that alignment to
1178 If this macro is not defined, then @var{basic-align} is used.
1180 The typical use of this macro is to increase alignment for string
1181 constants to be word aligned so that @code{strcpy} calls that copy
1182 constants can be done inline.
1185 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1186 If defined, a C expression to compute the alignment for a variable in
1187 the local store. @var{type} is the data type, and @var{basic-align} is
1188 the alignment that the object would ordinarily have. The value of this
1189 macro is used instead of that alignment to align the object.
1191 If this macro is not defined, then @var{basic-align} is used.
1193 One use of this macro is to increase alignment of medium-size data to
1194 make it all fit in fewer cache lines.
1197 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1198 If defined, a C expression to compute the alignment for stack slot.
1199 @var{type} is the data type, @var{mode} is the widest mode available,
1200 and @var{basic-align} is the alignment that the slot would ordinarily
1201 have. The value of this macro is used instead of that alignment to
1204 If this macro is not defined, then @var{basic-align} is used when
1205 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1208 This macro is to set alignment of stack slot to the maximum alignment
1209 of all possible modes which the slot may have.
1212 @defmac EMPTY_FIELD_BOUNDARY
1213 Alignment in bits to be given to a structure bit-field that follows an
1214 empty field such as @code{int : 0;}.
1216 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1219 @defmac STRUCTURE_SIZE_BOUNDARY
1220 Number of bits which any structure or union's size must be a multiple of.
1221 Each structure or union's size is rounded up to a multiple of this.
1223 If you do not define this macro, the default is the same as
1224 @code{BITS_PER_UNIT}.
1227 @defmac STRICT_ALIGNMENT
1228 Define this macro to be the value 1 if instructions will fail to work
1229 if given data not on the nominal alignment. If instructions will merely
1230 go slower in that case, define this macro as 0.
1233 @defmac PCC_BITFIELD_TYPE_MATTERS
1234 Define this if you wish to imitate the way many other C compilers handle
1235 alignment of bit-fields and the structures that contain them.
1237 The behavior is that the type written for a named bit-field (@code{int},
1238 @code{short}, or other integer type) imposes an alignment for the entire
1239 structure, as if the structure really did contain an ordinary field of
1240 that type. In addition, the bit-field is placed within the structure so
1241 that it would fit within such a field, not crossing a boundary for it.
1243 Thus, on most machines, a named bit-field whose type is written as
1244 @code{int} would not cross a four-byte boundary, and would force
1245 four-byte alignment for the whole structure. (The alignment used may
1246 not be four bytes; it is controlled by the other alignment parameters.)
1248 An unnamed bit-field will not affect the alignment of the containing
1251 If the macro is defined, its definition should be a C expression;
1252 a nonzero value for the expression enables this behavior.
1254 Note that if this macro is not defined, or its value is zero, some
1255 bit-fields may cross more than one alignment boundary. The compiler can
1256 support such references if there are @samp{insv}, @samp{extv}, and
1257 @samp{extzv} insns that can directly reference memory.
1259 The other known way of making bit-fields work is to define
1260 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1261 Then every structure can be accessed with fullwords.
1263 Unless the machine has bit-field instructions or you define
1264 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1265 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1267 If your aim is to make GCC use the same conventions for laying out
1268 bit-fields as are used by another compiler, here is how to investigate
1269 what the other compiler does. Compile and run this program:
1288 printf ("Size of foo1 is %d\n",
1289 sizeof (struct foo1));
1290 printf ("Size of foo2 is %d\n",
1291 sizeof (struct foo2));
1296 If this prints 2 and 5, then the compiler's behavior is what you would
1297 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1300 @defmac BITFIELD_NBYTES_LIMITED
1301 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1302 to aligning a bit-field within the structure.
1305 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELDS (void)
1306 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1307 whether unnamed bitfields affect the alignment of the containing
1308 structure. The hook should return true if the structure should inherit
1309 the alignment requirements of an unnamed bitfield's type.
1312 @deftypefn {Target Hook} bool TARGET_NARROW_VOLATILE_BITFIELDS (void)
1313 This target hook should return @code{true} if accesses to volatile bitfields
1314 should use the narrowest mode possible. It should return @code{false} if
1315 these accesses should use the bitfield container type.
1317 The default is @code{!TARGET_STRICT_ALIGN}.
1320 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1321 Return 1 if a structure or array containing @var{field} should be accessed using
1324 If @var{field} is the only field in the structure, @var{mode} is its
1325 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1326 case where structures of one field would require the structure's mode to
1327 retain the field's mode.
1329 Normally, this is not needed.
1332 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1333 Define this macro as an expression for the alignment of a type (given
1334 by @var{type} as a tree node) if the alignment computed in the usual
1335 way is @var{computed} and the alignment explicitly specified was
1338 The default is to use @var{specified} if it is larger; otherwise, use
1339 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1342 @defmac MAX_FIXED_MODE_SIZE
1343 An integer expression for the size in bits of the largest integer
1344 machine mode that should actually be used. All integer machine modes of
1345 this size or smaller can be used for structures and unions with the
1346 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1347 (DImode)} is assumed.
1350 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1351 If defined, an expression of type @code{enum machine_mode} that
1352 specifies the mode of the save area operand of a
1353 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1354 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1355 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1356 having its mode specified.
1358 You need not define this macro if it always returns @code{Pmode}. You
1359 would most commonly define this macro if the
1360 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1364 @defmac STACK_SIZE_MODE
1365 If defined, an expression of type @code{enum machine_mode} that
1366 specifies the mode of the size increment operand of an
1367 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1369 You need not define this macro if it always returns @code{word_mode}.
1370 You would most commonly define this macro if the @code{allocate_stack}
1371 pattern needs to support both a 32- and a 64-bit mode.
1374 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_CMP_RETURN_MODE ()
1375 This target hook should return the mode to be used for the return value
1376 of compare instructions expanded to libgcc calls. If not defined
1377 @code{word_mode} is returned which is the right choice for a majority of
1381 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_SHIFT_COUNT_MODE ()
1382 This target hook should return the mode to be used for the shift count operand
1383 of shift instructions expanded to libgcc calls. If not defined
1384 @code{word_mode} is returned which is the right choice for a majority of
1388 @defmac ROUND_TOWARDS_ZERO
1389 If defined, this macro should be true if the prevailing rounding
1390 mode is towards zero.
1392 Defining this macro only affects the way @file{libgcc.a} emulates
1393 floating-point arithmetic.
1395 Not defining this macro is equivalent to returning zero.
1398 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1399 This macro should return true if floats with @var{size}
1400 bits do not have a NaN or infinity representation, but use the largest
1401 exponent for normal numbers instead.
1403 Defining this macro only affects the way @file{libgcc.a} emulates
1404 floating-point arithmetic.
1406 The default definition of this macro returns false for all sizes.
1409 @deftypefn {Target Hook} bool TARGET_VECTOR_OPAQUE_P (tree @var{type})
1410 This target hook should return @code{true} a vector is opaque. That
1411 is, if no cast is needed when copying a vector value of type
1412 @var{type} into another vector lvalue of the same size. Vector opaque
1413 types cannot be initialized. The default is that there are no such
1417 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1418 This target hook returns @code{true} if bit-fields in the given
1419 @var{record_type} are to be laid out following the rules of Microsoft
1420 Visual C/C++, namely: (i) a bit-field won't share the same storage
1421 unit with the previous bit-field if their underlying types have
1422 different sizes, and the bit-field will be aligned to the highest
1423 alignment of the underlying types of itself and of the previous
1424 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1425 the whole enclosing structure, even if it is unnamed; except that
1426 (iii) a zero-sized bit-field will be disregarded unless it follows
1427 another bit-field of nonzero size. If this hook returns @code{true},
1428 other macros that control bit-field layout are ignored.
1430 When a bit-field is inserted into a packed record, the whole size
1431 of the underlying type is used by one or more same-size adjacent
1432 bit-fields (that is, if its long:3, 32 bits is used in the record,
1433 and any additional adjacent long bit-fields are packed into the same
1434 chunk of 32 bits. However, if the size changes, a new field of that
1435 size is allocated). In an unpacked record, this is the same as using
1436 alignment, but not equivalent when packing.
1438 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1439 the latter will take precedence. If @samp{__attribute__((packed))} is
1440 used on a single field when MS bit-fields are in use, it will take
1441 precedence for that field, but the alignment of the rest of the structure
1442 may affect its placement.
1445 @deftypefn {Target Hook} {bool} TARGET_DECIMAL_FLOAT_SUPPORTED_P (void)
1446 Returns true if the target supports decimal floating point.
1449 @deftypefn {Target Hook} {bool} TARGET_FIXED_POINT_SUPPORTED_P (void)
1450 Returns true if the target supports fixed-point arithmetic.
1453 @deftypefn {Target Hook} void TARGET_EXPAND_TO_RTL_HOOK (void)
1454 This hook is called just before expansion into rtl, allowing the target
1455 to perform additional initializations or analysis before the expansion.
1456 For example, the rs6000 port uses it to allocate a scratch stack slot
1457 for use in copying SDmode values between memory and floating point
1458 registers whenever the function being expanded has any SDmode
1462 @deftypefn {Target Hook} void TARGET_INSTANTIATE_DECLS (void)
1463 This hook allows the backend to perform additional instantiations on rtl
1464 that are not actually in any insns yet, but will be later.
1467 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_TYPE (tree @var{type})
1468 If your target defines any fundamental types, or any types your target
1469 uses should be mangled differently from the default, define this hook
1470 to return the appropriate encoding for these types as part of a C++
1471 mangled name. The @var{type} argument is the tree structure representing
1472 the type to be mangled. The hook may be applied to trees which are
1473 not target-specific fundamental types; it should return @code{NULL}
1474 for all such types, as well as arguments it does not recognize. If the
1475 return value is not @code{NULL}, it must point to a statically-allocated
1478 Target-specific fundamental types might be new fundamental types or
1479 qualified versions of ordinary fundamental types. Encode new
1480 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1481 is the name used for the type in source code, and @var{n} is the
1482 length of @var{name} in decimal. Encode qualified versions of
1483 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1484 @var{name} is the name used for the type qualifier in source code,
1485 @var{n} is the length of @var{name} as above, and @var{code} is the
1486 code used to represent the unqualified version of this type. (See
1487 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1488 codes.) In both cases the spaces are for clarity; do not include any
1489 spaces in your string.
1491 This hook is applied to types prior to typedef resolution. If the mangled
1492 name for a particular type depends only on that type's main variant, you
1493 can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT}
1496 The default version of this hook always returns @code{NULL}, which is
1497 appropriate for a target that does not define any new fundamental
1502 @section Layout of Source Language Data Types
1504 These macros define the sizes and other characteristics of the standard
1505 basic data types used in programs being compiled. Unlike the macros in
1506 the previous section, these apply to specific features of C and related
1507 languages, rather than to fundamental aspects of storage layout.
1509 @defmac INT_TYPE_SIZE
1510 A C expression for the size in bits of the type @code{int} on the
1511 target machine. If you don't define this, the default is one word.
1514 @defmac SHORT_TYPE_SIZE
1515 A C expression for the size in bits of the type @code{short} on the
1516 target machine. If you don't define this, the default is half a word.
1517 (If this would be less than one storage unit, it is rounded up to one
1521 @defmac LONG_TYPE_SIZE
1522 A C expression for the size in bits of the type @code{long} on the
1523 target machine. If you don't define this, the default is one word.
1526 @defmac ADA_LONG_TYPE_SIZE
1527 On some machines, the size used for the Ada equivalent of the type
1528 @code{long} by a native Ada compiler differs from that used by C@. In
1529 that situation, define this macro to be a C expression to be used for
1530 the size of that type. If you don't define this, the default is the
1531 value of @code{LONG_TYPE_SIZE}.
1534 @defmac LONG_LONG_TYPE_SIZE
1535 A C expression for the size in bits of the type @code{long long} on the
1536 target machine. If you don't define this, the default is two
1537 words. If you want to support GNU Ada on your machine, the value of this
1538 macro must be at least 64.
1541 @defmac CHAR_TYPE_SIZE
1542 A C expression for the size in bits of the type @code{char} on the
1543 target machine. If you don't define this, the default is
1544 @code{BITS_PER_UNIT}.
1547 @defmac BOOL_TYPE_SIZE
1548 A C expression for the size in bits of the C++ type @code{bool} and
1549 C99 type @code{_Bool} on the target machine. If you don't define
1550 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1553 @defmac FLOAT_TYPE_SIZE
1554 A C expression for the size in bits of the type @code{float} on the
1555 target machine. If you don't define this, the default is one word.
1558 @defmac DOUBLE_TYPE_SIZE
1559 A C expression for the size in bits of the type @code{double} on the
1560 target machine. If you don't define this, the default is two
1564 @defmac LONG_DOUBLE_TYPE_SIZE
1565 A C expression for the size in bits of the type @code{long double} on
1566 the target machine. If you don't define this, the default is two
1570 @defmac SHORT_FRACT_TYPE_SIZE
1571 A C expression for the size in bits of the type @code{short _Fract} on
1572 the target machine. If you don't define this, the default is
1573 @code{BITS_PER_UNIT}.
1576 @defmac FRACT_TYPE_SIZE
1577 A C expression for the size in bits of the type @code{_Fract} on
1578 the target machine. If you don't define this, the default is
1579 @code{BITS_PER_UNIT * 2}.
1582 @defmac LONG_FRACT_TYPE_SIZE
1583 A C expression for the size in bits of the type @code{long _Fract} on
1584 the target machine. If you don't define this, the default is
1585 @code{BITS_PER_UNIT * 4}.
1588 @defmac LONG_LONG_FRACT_TYPE_SIZE
1589 A C expression for the size in bits of the type @code{long long _Fract} on
1590 the target machine. If you don't define this, the default is
1591 @code{BITS_PER_UNIT * 8}.
1594 @defmac SHORT_ACCUM_TYPE_SIZE
1595 A C expression for the size in bits of the type @code{short _Accum} on
1596 the target machine. If you don't define this, the default is
1597 @code{BITS_PER_UNIT * 2}.
1600 @defmac ACCUM_TYPE_SIZE
1601 A C expression for the size in bits of the type @code{_Accum} on
1602 the target machine. If you don't define this, the default is
1603 @code{BITS_PER_UNIT * 4}.
1606 @defmac LONG_ACCUM_TYPE_SIZE
1607 A C expression for the size in bits of the type @code{long _Accum} on
1608 the target machine. If you don't define this, the default is
1609 @code{BITS_PER_UNIT * 8}.
1612 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1613 A C expression for the size in bits of the type @code{long long _Accum} on
1614 the target machine. If you don't define this, the default is
1615 @code{BITS_PER_UNIT * 16}.
1618 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1619 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1620 if you want routines in @file{libgcc2.a} for a size other than
1621 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1622 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1625 @defmac LIBGCC2_HAS_DF_MODE
1626 Define this macro if neither @code{LIBGCC2_DOUBLE_TYPE_SIZE} nor
1627 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1628 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1629 anyway. If you don't define this and either @code{LIBGCC2_DOUBLE_TYPE_SIZE}
1630 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1634 @defmac LIBGCC2_HAS_XF_MODE
1635 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1636 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1637 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1638 is 80 then the default is 1, otherwise it is 0.
1641 @defmac LIBGCC2_HAS_TF_MODE
1642 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1643 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1644 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1645 is 128 then the default is 1, otherwise it is 0.
1652 Define these macros to be the size in bits of the mantissa of
1653 @code{SFmode}, @code{DFmode}, @code{XFmode} and @code{TFmode} values,
1654 if the defaults in @file{libgcc2.h} are inappropriate. By default,
1655 @code{FLT_MANT_DIG} is used for @code{SF_SIZE}, @code{LDBL_MANT_DIG}
1656 for @code{XF_SIZE} and @code{TF_SIZE}, and @code{DBL_MANT_DIG} or
1657 @code{LDBL_MANT_DIG} for @code{DF_SIZE} according to whether
1658 @code{LIBGCC2_DOUBLE_TYPE_SIZE} or
1659 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64.
1662 @defmac TARGET_FLT_EVAL_METHOD
1663 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1664 assuming, if applicable, that the floating-point control word is in its
1665 default state. If you do not define this macro the value of
1666 @code{FLT_EVAL_METHOD} will be zero.
1669 @defmac WIDEST_HARDWARE_FP_SIZE
1670 A C expression for the size in bits of the widest floating-point format
1671 supported by the hardware. If you define this macro, you must specify a
1672 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1673 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1677 @defmac DEFAULT_SIGNED_CHAR
1678 An expression whose value is 1 or 0, according to whether the type
1679 @code{char} should be signed or unsigned by default. The user can
1680 always override this default with the options @option{-fsigned-char}
1681 and @option{-funsigned-char}.
1684 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1685 This target hook should return true if the compiler should give an
1686 @code{enum} type only as many bytes as it takes to represent the range
1687 of possible values of that type. It should return false if all
1688 @code{enum} types should be allocated like @code{int}.
1690 The default is to return false.
1694 A C expression for a string describing the name of the data type to use
1695 for size values. The typedef name @code{size_t} is defined using the
1696 contents of the string.
1698 The string can contain more than one keyword. If so, separate them with
1699 spaces, and write first any length keyword, then @code{unsigned} if
1700 appropriate, and finally @code{int}. The string must exactly match one
1701 of the data type names defined in the function
1702 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1703 omit @code{int} or change the order---that would cause the compiler to
1706 If you don't define this macro, the default is @code{"long unsigned
1710 @defmac PTRDIFF_TYPE
1711 A C expression for a string describing the name of the data type to use
1712 for the result of subtracting two pointers. The typedef name
1713 @code{ptrdiff_t} is defined using the contents of the string. See
1714 @code{SIZE_TYPE} above for more information.
1716 If you don't define this macro, the default is @code{"long int"}.
1720 A C expression for a string describing the name of the data type to use
1721 for wide characters. The typedef name @code{wchar_t} is defined using
1722 the contents of the string. See @code{SIZE_TYPE} above for more
1725 If you don't define this macro, the default is @code{"int"}.
1728 @defmac WCHAR_TYPE_SIZE
1729 A C expression for the size in bits of the data type for wide
1730 characters. This is used in @code{cpp}, which cannot make use of
1735 A C expression for a string describing the name of the data type to
1736 use for wide characters passed to @code{printf} and returned from
1737 @code{getwc}. The typedef name @code{wint_t} is defined using the
1738 contents of the string. See @code{SIZE_TYPE} above for more
1741 If you don't define this macro, the default is @code{"unsigned int"}.
1745 A C expression for a string describing the name of the data type that
1746 can represent any value of any standard or extended signed integer type.
1747 The typedef name @code{intmax_t} is defined using the contents of the
1748 string. See @code{SIZE_TYPE} above for more information.
1750 If you don't define this macro, the default is the first of
1751 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1752 much precision as @code{long long int}.
1755 @defmac UINTMAX_TYPE
1756 A C expression for a string describing the name of the data type that
1757 can represent any value of any standard or extended unsigned integer
1758 type. The typedef name @code{uintmax_t} is defined using the contents
1759 of the string. See @code{SIZE_TYPE} above for more information.
1761 If you don't define this macro, the default is the first of
1762 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1763 unsigned int"} that has as much precision as @code{long long unsigned
1767 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1768 The C++ compiler represents a pointer-to-member-function with a struct
1775 ptrdiff_t vtable_index;
1782 The C++ compiler must use one bit to indicate whether the function that
1783 will be called through a pointer-to-member-function is virtual.
1784 Normally, we assume that the low-order bit of a function pointer must
1785 always be zero. Then, by ensuring that the vtable_index is odd, we can
1786 distinguish which variant of the union is in use. But, on some
1787 platforms function pointers can be odd, and so this doesn't work. In
1788 that case, we use the low-order bit of the @code{delta} field, and shift
1789 the remainder of the @code{delta} field to the left.
1791 GCC will automatically make the right selection about where to store
1792 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1793 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1794 set such that functions always start at even addresses, but the lowest
1795 bit of pointers to functions indicate whether the function at that
1796 address is in ARM or Thumb mode. If this is the case of your
1797 architecture, you should define this macro to
1798 @code{ptrmemfunc_vbit_in_delta}.
1800 In general, you should not have to define this macro. On architectures
1801 in which function addresses are always even, according to
1802 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1803 @code{ptrmemfunc_vbit_in_pfn}.
1806 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1807 Normally, the C++ compiler uses function pointers in vtables. This
1808 macro allows the target to change to use ``function descriptors''
1809 instead. Function descriptors are found on targets for whom a
1810 function pointer is actually a small data structure. Normally the
1811 data structure consists of the actual code address plus a data
1812 pointer to which the function's data is relative.
1814 If vtables are used, the value of this macro should be the number
1815 of words that the function descriptor occupies.
1818 @defmac TARGET_VTABLE_ENTRY_ALIGN
1819 By default, the vtable entries are void pointers, the so the alignment
1820 is the same as pointer alignment. The value of this macro specifies
1821 the alignment of the vtable entry in bits. It should be defined only
1822 when special alignment is necessary. */
1825 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1826 There are a few non-descriptor entries in the vtable at offsets below
1827 zero. If these entries must be padded (say, to preserve the alignment
1828 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1829 of words in each data entry.
1833 @section Register Usage
1834 @cindex register usage
1836 This section explains how to describe what registers the target machine
1837 has, and how (in general) they can be used.
1839 The description of which registers a specific instruction can use is
1840 done with register classes; see @ref{Register Classes}. For information
1841 on using registers to access a stack frame, see @ref{Frame Registers}.
1842 For passing values in registers, see @ref{Register Arguments}.
1843 For returning values in registers, see @ref{Scalar Return}.
1846 * Register Basics:: Number and kinds of registers.
1847 * Allocation Order:: Order in which registers are allocated.
1848 * Values in Registers:: What kinds of values each reg can hold.
1849 * Leaf Functions:: Renumbering registers for leaf functions.
1850 * Stack Registers:: Handling a register stack such as 80387.
1853 @node Register Basics
1854 @subsection Basic Characteristics of Registers
1856 @c prevent bad page break with this line
1857 Registers have various characteristics.
1859 @defmac FIRST_PSEUDO_REGISTER
1860 Number of hardware registers known to the compiler. They receive
1861 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1862 pseudo register's number really is assigned the number
1863 @code{FIRST_PSEUDO_REGISTER}.
1866 @defmac FIXED_REGISTERS
1867 @cindex fixed register
1868 An initializer that says which registers are used for fixed purposes
1869 all throughout the compiled code and are therefore not available for
1870 general allocation. These would include the stack pointer, the frame
1871 pointer (except on machines where that can be used as a general
1872 register when no frame pointer is needed), the program counter on
1873 machines where that is considered one of the addressable registers,
1874 and any other numbered register with a standard use.
1876 This information is expressed as a sequence of numbers, separated by
1877 commas and surrounded by braces. The @var{n}th number is 1 if
1878 register @var{n} is fixed, 0 otherwise.
1880 The table initialized from this macro, and the table initialized by
1881 the following one, may be overridden at run time either automatically,
1882 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1883 the user with the command options @option{-ffixed-@var{reg}},
1884 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1887 @defmac CALL_USED_REGISTERS
1888 @cindex call-used register
1889 @cindex call-clobbered register
1890 @cindex call-saved register
1891 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1892 clobbered (in general) by function calls as well as for fixed
1893 registers. This macro therefore identifies the registers that are not
1894 available for general allocation of values that must live across
1897 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1898 automatically saves it on function entry and restores it on function
1899 exit, if the register is used within the function.
1902 @defmac CALL_REALLY_USED_REGISTERS
1903 @cindex call-used register
1904 @cindex call-clobbered register
1905 @cindex call-saved register
1906 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1907 that the entire set of @code{FIXED_REGISTERS} be included.
1908 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1909 This macro is optional. If not specified, it defaults to the value
1910 of @code{CALL_USED_REGISTERS}.
1913 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1914 @cindex call-used register
1915 @cindex call-clobbered register
1916 @cindex call-saved register
1917 A C expression that is nonzero if it is not permissible to store a
1918 value of mode @var{mode} in hard register number @var{regno} across a
1919 call without some part of it being clobbered. For most machines this
1920 macro need not be defined. It is only required for machines that do not
1921 preserve the entire contents of a register across a call.
1925 @findex call_used_regs
1928 @findex reg_class_contents
1929 @defmac CONDITIONAL_REGISTER_USAGE
1930 Zero or more C statements that may conditionally modify five variables
1931 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1932 @code{reg_names}, and @code{reg_class_contents}, to take into account
1933 any dependence of these register sets on target flags. The first three
1934 of these are of type @code{char []} (interpreted as Boolean vectors).
1935 @code{global_regs} is a @code{const char *[]}, and
1936 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1937 called, @code{fixed_regs}, @code{call_used_regs},
1938 @code{reg_class_contents}, and @code{reg_names} have been initialized
1939 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1940 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1941 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1942 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1943 command options have been applied.
1945 You need not define this macro if it has no work to do.
1947 @cindex disabling certain registers
1948 @cindex controlling register usage
1949 If the usage of an entire class of registers depends on the target
1950 flags, you may indicate this to GCC by using this macro to modify
1951 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1952 registers in the classes which should not be used by GCC@. Also define
1953 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1954 to return @code{NO_REGS} if it
1955 is called with a letter for a class that shouldn't be used.
1957 (However, if this class is not included in @code{GENERAL_REGS} and all
1958 of the insn patterns whose constraints permit this class are
1959 controlled by target switches, then GCC will automatically avoid using
1960 these registers when the target switches are opposed to them.)
1963 @defmac INCOMING_REGNO (@var{out})
1964 Define this macro if the target machine has register windows. This C
1965 expression returns the register number as seen by the called function
1966 corresponding to the register number @var{out} as seen by the calling
1967 function. Return @var{out} if register number @var{out} is not an
1971 @defmac OUTGOING_REGNO (@var{in})
1972 Define this macro if the target machine has register windows. This C
1973 expression returns the register number as seen by the calling function
1974 corresponding to the register number @var{in} as seen by the called
1975 function. Return @var{in} if register number @var{in} is not an inbound
1979 @defmac LOCAL_REGNO (@var{regno})
1980 Define this macro if the target machine has register windows. This C
1981 expression returns true if the register is call-saved but is in the
1982 register window. Unlike most call-saved registers, such registers
1983 need not be explicitly restored on function exit or during non-local
1988 If the program counter has a register number, define this as that
1989 register number. Otherwise, do not define it.
1992 @node Allocation Order
1993 @subsection Order of Allocation of Registers
1994 @cindex order of register allocation
1995 @cindex register allocation order
1997 @c prevent bad page break with this line
1998 Registers are allocated in order.
2000 @defmac REG_ALLOC_ORDER
2001 If defined, an initializer for a vector of integers, containing the
2002 numbers of hard registers in the order in which GCC should prefer
2003 to use them (from most preferred to least).
2005 If this macro is not defined, registers are used lowest numbered first
2006 (all else being equal).
2008 One use of this macro is on machines where the highest numbered
2009 registers must always be saved and the save-multiple-registers
2010 instruction supports only sequences of consecutive registers. On such
2011 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
2012 the highest numbered allocable register first.
2015 @defmac ORDER_REGS_FOR_LOCAL_ALLOC
2016 A C statement (sans semicolon) to choose the order in which to allocate
2017 hard registers for pseudo-registers local to a basic block.
2019 Store the desired register order in the array @code{reg_alloc_order}.
2020 Element 0 should be the register to allocate first; element 1, the next
2021 register; and so on.
2023 The macro body should not assume anything about the contents of
2024 @code{reg_alloc_order} before execution of the macro.
2026 On most machines, it is not necessary to define this macro.
2029 @node Values in Registers
2030 @subsection How Values Fit in Registers
2032 This section discusses the macros that describe which kinds of values
2033 (specifically, which machine modes) each register can hold, and how many
2034 consecutive registers are needed for a given mode.
2036 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2037 A C expression for the number of consecutive hard registers, starting
2038 at register number @var{regno}, required to hold a value of mode
2039 @var{mode}. This macro must never return zero, even if a register
2040 cannot hold the requested mode - indicate that with HARD_REGNO_MODE_OK
2041 and/or CANNOT_CHANGE_MODE_CLASS instead.
2043 On a machine where all registers are exactly one word, a suitable
2044 definition of this macro is
2047 #define HARD_REGNO_NREGS(REGNO, MODE) \
2048 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2053 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2054 A C expression that is nonzero if a value of mode @var{mode}, stored
2055 in memory, ends with padding that causes it to take up more space than
2056 in registers starting at register number @var{regno} (as determined by
2057 multiplying GCC's notion of the size of the register when containing
2058 this mode by the number of registers returned by
2059 @code{HARD_REGNO_NREGS}). By default this is zero.
2061 For example, if a floating-point value is stored in three 32-bit
2062 registers but takes up 128 bits in memory, then this would be
2065 This macros only needs to be defined if there are cases where
2066 @code{subreg_get_info}
2067 would otherwise wrongly determine that a @code{subreg} can be
2068 represented by an offset to the register number, when in fact such a
2069 @code{subreg} would contain some of the padding not stored in
2070 registers and so not be representable.
2073 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2074 For values of @var{regno} and @var{mode} for which
2075 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2076 returning the greater number of registers required to hold the value
2077 including any padding. In the example above, the value would be four.
2080 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2081 Define this macro if the natural size of registers that hold values
2082 of mode @var{mode} is not the word size. It is a C expression that
2083 should give the natural size in bytes for the specified mode. It is
2084 used by the register allocator to try to optimize its results. This
2085 happens for example on SPARC 64-bit where the natural size of
2086 floating-point registers is still 32-bit.
2089 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2090 A C expression that is nonzero if it is permissible to store a value
2091 of mode @var{mode} in hard register number @var{regno} (or in several
2092 registers starting with that one). For a machine where all registers
2093 are equivalent, a suitable definition is
2096 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2099 You need not include code to check for the numbers of fixed registers,
2100 because the allocation mechanism considers them to be always occupied.
2102 @cindex register pairs
2103 On some machines, double-precision values must be kept in even/odd
2104 register pairs. You can implement that by defining this macro to reject
2105 odd register numbers for such modes.
2107 The minimum requirement for a mode to be OK in a register is that the
2108 @samp{mov@var{mode}} instruction pattern support moves between the
2109 register and other hard register in the same class and that moving a
2110 value into the register and back out not alter it.
2112 Since the same instruction used to move @code{word_mode} will work for
2113 all narrower integer modes, it is not necessary on any machine for
2114 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2115 you define patterns @samp{movhi}, etc., to take advantage of this. This
2116 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2117 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2120 Many machines have special registers for floating point arithmetic.
2121 Often people assume that floating point machine modes are allowed only
2122 in floating point registers. This is not true. Any registers that
2123 can hold integers can safely @emph{hold} a floating point machine
2124 mode, whether or not floating arithmetic can be done on it in those
2125 registers. Integer move instructions can be used to move the values.
2127 On some machines, though, the converse is true: fixed-point machine
2128 modes may not go in floating registers. This is true if the floating
2129 registers normalize any value stored in them, because storing a
2130 non-floating value there would garble it. In this case,
2131 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2132 floating registers. But if the floating registers do not automatically
2133 normalize, if you can store any bit pattern in one and retrieve it
2134 unchanged without a trap, then any machine mode may go in a floating
2135 register, so you can define this macro to say so.
2137 The primary significance of special floating registers is rather that
2138 they are the registers acceptable in floating point arithmetic
2139 instructions. However, this is of no concern to
2140 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2141 constraints for those instructions.
2143 On some machines, the floating registers are especially slow to access,
2144 so that it is better to store a value in a stack frame than in such a
2145 register if floating point arithmetic is not being done. As long as the
2146 floating registers are not in class @code{GENERAL_REGS}, they will not
2147 be used unless some pattern's constraint asks for one.
2150 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2151 A C expression that is nonzero if it is OK to rename a hard register
2152 @var{from} to another hard register @var{to}.
2154 One common use of this macro is to prevent renaming of a register to
2155 another register that is not saved by a prologue in an interrupt
2158 The default is always nonzero.
2161 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2162 A C expression that is nonzero if a value of mode
2163 @var{mode1} is accessible in mode @var{mode2} without copying.
2165 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2166 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2167 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2168 should be nonzero. If they differ for any @var{r}, you should define
2169 this macro to return zero unless some other mechanism ensures the
2170 accessibility of the value in a narrower mode.
2172 You should define this macro to return nonzero in as many cases as
2173 possible since doing so will allow GCC to perform better register
2177 @deftypefn {Target Hook} bool TARGET_HARD_REGNO_SCRATCH_OK (unsigned int @var{regno})
2178 This target hook should return @code{true} if it is OK to use a hard register
2179 @var{regno} as scratch reg in peephole2.
2181 One common use of this macro is to prevent using of a register that
2182 is not saved by a prologue in an interrupt handler.
2184 The default version of this hook always returns @code{true}.
2187 @defmac AVOID_CCMODE_COPIES
2188 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2189 registers. You should only define this macro if support for copying to/from
2190 @code{CCmode} is incomplete.
2193 @node Leaf Functions
2194 @subsection Handling Leaf Functions
2196 @cindex leaf functions
2197 @cindex functions, leaf
2198 On some machines, a leaf function (i.e., one which makes no calls) can run
2199 more efficiently if it does not make its own register window. Often this
2200 means it is required to receive its arguments in the registers where they
2201 are passed by the caller, instead of the registers where they would
2204 The special treatment for leaf functions generally applies only when
2205 other conditions are met; for example, often they may use only those
2206 registers for its own variables and temporaries. We use the term ``leaf
2207 function'' to mean a function that is suitable for this special
2208 handling, so that functions with no calls are not necessarily ``leaf
2211 GCC assigns register numbers before it knows whether the function is
2212 suitable for leaf function treatment. So it needs to renumber the
2213 registers in order to output a leaf function. The following macros
2216 @defmac LEAF_REGISTERS
2217 Name of a char vector, indexed by hard register number, which
2218 contains 1 for a register that is allowable in a candidate for leaf
2221 If leaf function treatment involves renumbering the registers, then the
2222 registers marked here should be the ones before renumbering---those that
2223 GCC would ordinarily allocate. The registers which will actually be
2224 used in the assembler code, after renumbering, should not be marked with 1
2227 Define this macro only if the target machine offers a way to optimize
2228 the treatment of leaf functions.
2231 @defmac LEAF_REG_REMAP (@var{regno})
2232 A C expression whose value is the register number to which @var{regno}
2233 should be renumbered, when a function is treated as a leaf function.
2235 If @var{regno} is a register number which should not appear in a leaf
2236 function before renumbering, then the expression should yield @minus{}1, which
2237 will cause the compiler to abort.
2239 Define this macro only if the target machine offers a way to optimize the
2240 treatment of leaf functions, and registers need to be renumbered to do
2244 @findex current_function_is_leaf
2245 @findex current_function_uses_only_leaf_regs
2246 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2247 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2248 specially. They can test the C variable @code{current_function_is_leaf}
2249 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2250 set prior to local register allocation and is valid for the remaining
2251 compiler passes. They can also test the C variable
2252 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2253 functions which only use leaf registers.
2254 @code{current_function_uses_only_leaf_regs} is valid after all passes
2255 that modify the instructions have been run and is only useful if
2256 @code{LEAF_REGISTERS} is defined.
2257 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2258 @c of the next paragraph?! --mew 2feb93
2260 @node Stack Registers
2261 @subsection Registers That Form a Stack
2263 There are special features to handle computers where some of the
2264 ``registers'' form a stack. Stack registers are normally written by
2265 pushing onto the stack, and are numbered relative to the top of the
2268 Currently, GCC can only handle one group of stack-like registers, and
2269 they must be consecutively numbered. Furthermore, the existing
2270 support for stack-like registers is specific to the 80387 floating
2271 point coprocessor. If you have a new architecture that uses
2272 stack-like registers, you will need to do substantial work on
2273 @file{reg-stack.c} and write your machine description to cooperate
2274 with it, as well as defining these macros.
2277 Define this if the machine has any stack-like registers.
2280 @defmac FIRST_STACK_REG
2281 The number of the first stack-like register. This one is the top
2285 @defmac LAST_STACK_REG
2286 The number of the last stack-like register. This one is the bottom of
2290 @node Register Classes
2291 @section Register Classes
2292 @cindex register class definitions
2293 @cindex class definitions, register
2295 On many machines, the numbered registers are not all equivalent.
2296 For example, certain registers may not be allowed for indexed addressing;
2297 certain registers may not be allowed in some instructions. These machine
2298 restrictions are described to the compiler using @dfn{register classes}.
2300 You define a number of register classes, giving each one a name and saying
2301 which of the registers belong to it. Then you can specify register classes
2302 that are allowed as operands to particular instruction patterns.
2306 In general, each register will belong to several classes. In fact, one
2307 class must be named @code{ALL_REGS} and contain all the registers. Another
2308 class must be named @code{NO_REGS} and contain no registers. Often the
2309 union of two classes will be another class; however, this is not required.
2311 @findex GENERAL_REGS
2312 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2313 terribly special about the name, but the operand constraint letters
2314 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2315 the same as @code{ALL_REGS}, just define it as a macro which expands
2318 Order the classes so that if class @var{x} is contained in class @var{y}
2319 then @var{x} has a lower class number than @var{y}.
2321 The way classes other than @code{GENERAL_REGS} are specified in operand
2322 constraints is through machine-dependent operand constraint letters.
2323 You can define such letters to correspond to various classes, then use
2324 them in operand constraints.
2326 You should define a class for the union of two classes whenever some
2327 instruction allows both classes. For example, if an instruction allows
2328 either a floating point (coprocessor) register or a general register for a
2329 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2330 which includes both of them. Otherwise you will get suboptimal code.
2332 You must also specify certain redundant information about the register
2333 classes: for each class, which classes contain it and which ones are
2334 contained in it; for each pair of classes, the largest class contained
2337 When a value occupying several consecutive registers is expected in a
2338 certain class, all the registers used must belong to that class.
2339 Therefore, register classes cannot be used to enforce a requirement for
2340 a register pair to start with an even-numbered register. The way to
2341 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2343 Register classes used for input-operands of bitwise-and or shift
2344 instructions have a special requirement: each such class must have, for
2345 each fixed-point machine mode, a subclass whose registers can transfer that
2346 mode to or from memory. For example, on some machines, the operations for
2347 single-byte values (@code{QImode}) are limited to certain registers. When
2348 this is so, each register class that is used in a bitwise-and or shift
2349 instruction must have a subclass consisting of registers from which
2350 single-byte values can be loaded or stored. This is so that
2351 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2353 @deftp {Data type} {enum reg_class}
2354 An enumerated type that must be defined with all the register class names
2355 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2356 must be the last register class, followed by one more enumerated value,
2357 @code{LIM_REG_CLASSES}, which is not a register class but rather
2358 tells how many classes there are.
2360 Each register class has a number, which is the value of casting
2361 the class name to type @code{int}. The number serves as an index
2362 in many of the tables described below.
2365 @defmac N_REG_CLASSES
2366 The number of distinct register classes, defined as follows:
2369 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2373 @defmac REG_CLASS_NAMES
2374 An initializer containing the names of the register classes as C string
2375 constants. These names are used in writing some of the debugging dumps.
2378 @defmac REG_CLASS_CONTENTS
2379 An initializer containing the contents of the register classes, as integers
2380 which are bit masks. The @var{n}th integer specifies the contents of class
2381 @var{n}. The way the integer @var{mask} is interpreted is that
2382 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2384 When the machine has more than 32 registers, an integer does not suffice.
2385 Then the integers are replaced by sub-initializers, braced groupings containing
2386 several integers. Each sub-initializer must be suitable as an initializer
2387 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2388 In this situation, the first integer in each sub-initializer corresponds to
2389 registers 0 through 31, the second integer to registers 32 through 63, and
2393 @defmac REGNO_REG_CLASS (@var{regno})
2394 A C expression whose value is a register class containing hard register
2395 @var{regno}. In general there is more than one such class; choose a class
2396 which is @dfn{minimal}, meaning that no smaller class also contains the
2400 @defmac BASE_REG_CLASS
2401 A macro whose definition is the name of the class to which a valid
2402 base register must belong. A base register is one used in an address
2403 which is the register value plus a displacement.
2406 @defmac MODE_BASE_REG_CLASS (@var{mode})
2407 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2408 the selection of a base register in a mode dependent manner. If
2409 @var{mode} is VOIDmode then it should return the same value as
2410 @code{BASE_REG_CLASS}.
2413 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2414 A C expression whose value is the register class to which a valid
2415 base register must belong in order to be used in a base plus index
2416 register address. You should define this macro if base plus index
2417 addresses have different requirements than other base register uses.
2420 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{outer_code}, @var{index_code})
2421 A C expression whose value is the register class to which a valid
2422 base register must belong. @var{outer_code} and @var{index_code} define the
2423 context in which the base register occurs. @var{outer_code} is the code of
2424 the immediately enclosing expression (@code{MEM} for the top level of an
2425 address, @code{ADDRESS} for something that occurs in an
2426 @code{address_operand}). @var{index_code} is the code of the corresponding
2427 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2430 @defmac INDEX_REG_CLASS
2431 A macro whose definition is the name of the class to which a valid
2432 index register must belong. An index register is one used in an
2433 address where its value is either multiplied by a scale factor or
2434 added to another register (as well as added to a displacement).
2437 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2438 A C expression which is nonzero if register number @var{num} is
2439 suitable for use as a base register in operand addresses. It may be
2440 either a suitable hard register or a pseudo register that has been
2441 allocated such a hard register.
2444 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2445 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2446 that expression may examine the mode of the memory reference in
2447 @var{mode}. You should define this macro if the mode of the memory
2448 reference affects whether a register may be used as a base register. If
2449 you define this macro, the compiler will use it instead of
2450 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2451 addresses that appear outside a @code{MEM}, i.e., as an
2452 @code{address_operand}.
2456 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2457 A C expression which is nonzero if register number @var{num} is suitable for
2458 use as a base register in base plus index operand addresses, accessing
2459 memory in mode @var{mode}. It may be either a suitable hard register or a
2460 pseudo register that has been allocated such a hard register. You should
2461 define this macro if base plus index addresses have different requirements
2462 than other base register uses.
2464 Use of this macro is deprecated; please use the more general
2465 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2468 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{outer_code}, @var{index_code})
2469 A C expression that is just like @code{REGNO_MODE_OK_FOR_BASE_P}, except
2470 that that expression may examine the context in which the register
2471 appears in the memory reference. @var{outer_code} is the code of the
2472 immediately enclosing expression (@code{MEM} if at the top level of the
2473 address, @code{ADDRESS} for something that occurs in an
2474 @code{address_operand}). @var{index_code} is the code of the
2475 corresponding index expression if @var{outer_code} is @code{PLUS};
2476 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2477 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2480 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2481 A C expression which is nonzero if register number @var{num} is
2482 suitable for use as an index register in operand addresses. It may be
2483 either a suitable hard register or a pseudo register that has been
2484 allocated such a hard register.
2486 The difference between an index register and a base register is that
2487 the index register may be scaled. If an address involves the sum of
2488 two registers, neither one of them scaled, then either one may be
2489 labeled the ``base'' and the other the ``index''; but whichever
2490 labeling is used must fit the machine's constraints of which registers
2491 may serve in each capacity. The compiler will try both labelings,
2492 looking for one that is valid, and will reload one or both registers
2493 only if neither labeling works.
2496 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2497 A C expression that places additional restrictions on the register class
2498 to use when it is necessary to copy value @var{x} into a register in class
2499 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2500 another, smaller class. On many machines, the following definition is
2504 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2507 Sometimes returning a more restrictive class makes better code. For
2508 example, on the 68000, when @var{x} is an integer constant that is in range
2509 for a @samp{moveq} instruction, the value of this macro is always
2510 @code{DATA_REGS} as long as @var{class} includes the data registers.
2511 Requiring a data register guarantees that a @samp{moveq} will be used.
2513 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2514 @var{class} is if @var{x} is a legitimate constant which cannot be
2515 loaded into some register class. By returning @code{NO_REGS} you can
2516 force @var{x} into a memory location. For example, rs6000 can load
2517 immediate values into general-purpose registers, but does not have an
2518 instruction for loading an immediate value into a floating-point
2519 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2520 @var{x} is a floating-point constant. If the constant can't be loaded
2521 into any kind of register, code generation will be better if
2522 @code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2523 of using @code{PREFERRED_RELOAD_CLASS}.
2525 If an insn has pseudos in it after register allocation, reload will go
2526 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2527 to find the best one. Returning @code{NO_REGS}, in this case, makes
2528 reload add a @code{!} in front of the constraint: the x86 back-end uses
2529 this feature to discourage usage of 387 registers when math is done in
2530 the SSE registers (and vice versa).
2533 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2534 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2535 input reloads. If you don't define this macro, the default is to use
2536 @var{class}, unchanged.
2538 You can also use @code{PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2539 reload from using some alternatives, like @code{PREFERRED_RELOAD_CLASS}.
2542 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2543 A C expression that places additional restrictions on the register class
2544 to use when it is necessary to be able to hold a value of mode
2545 @var{mode} in a reload register for which class @var{class} would
2548 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2549 there are certain modes that simply can't go in certain reload classes.
2551 The value is a register class; perhaps @var{class}, or perhaps another,
2554 Don't define this macro unless the target machine has limitations which
2555 require the macro to do something nontrivial.
2558 @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})
2559 Many machines have some registers that cannot be copied directly to or
2560 from memory or even from other types of registers. An example is the
2561 @samp{MQ} register, which on most machines, can only be copied to or
2562 from general registers, but not memory. Below, we shall be using the
2563 term 'intermediate register' when a move operation cannot be performed
2564 directly, but has to be done by copying the source into the intermediate
2565 register first, and then copying the intermediate register to the
2566 destination. An intermediate register always has the same mode as
2567 source and destination. Since it holds the actual value being copied,
2568 reload might apply optimizations to re-use an intermediate register
2569 and eliding the copy from the source when it can determine that the
2570 intermediate register still holds the required value.
2572 Another kind of secondary reload is required on some machines which
2573 allow copying all registers to and from memory, but require a scratch
2574 register for stores to some memory locations (e.g., those with symbolic
2575 address on the RT, and those with certain symbolic address on the SPARC
2576 when compiling PIC)@. Scratch registers need not have the same mode
2577 as the value being copied, and usually hold a different value that
2578 that being copied. Special patterns in the md file are needed to
2579 describe how the copy is performed with the help of the scratch register;
2580 these patterns also describe the number, register class(es) and mode(s)
2581 of the scratch register(s).
2583 In some cases, both an intermediate and a scratch register are required.
2585 For input reloads, this target hook is called with nonzero @var{in_p},
2586 and @var{x} is an rtx that needs to be copied to a register of class
2587 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2588 hook is called with zero @var{in_p}, and a register of class @var{reload_class}
2589 needs to be copied to rtx @var{x} in @var{reload_mode}.
2591 If copying a register of @var{reload_class} from/to @var{x} requires
2592 an intermediate register, the hook @code{secondary_reload} should
2593 return the register class required for this intermediate register.
2594 If no intermediate register is required, it should return NO_REGS.
2595 If more than one intermediate register is required, describe the one
2596 that is closest in the copy chain to the reload register.
2598 If scratch registers are needed, you also have to describe how to
2599 perform the copy from/to the reload register to/from this
2600 closest intermediate register. Or if no intermediate register is
2601 required, but still a scratch register is needed, describe the
2602 copy from/to the reload register to/from the reload operand @var{x}.
2604 You do this by setting @code{sri->icode} to the instruction code of a pattern
2605 in the md file which performs the move. Operands 0 and 1 are the output
2606 and input of this copy, respectively. Operands from operand 2 onward are
2607 for scratch operands. These scratch operands must have a mode, and a
2608 single-register-class
2609 @c [later: or memory]
2612 When an intermediate register is used, the @code{secondary_reload}
2613 hook will be called again to determine how to copy the intermediate
2614 register to/from the reload operand @var{x}, so your hook must also
2615 have code to handle the register class of the intermediate operand.
2617 @c [For later: maybe we'll allow multi-alternative reload patterns -
2618 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2619 @c and match the constraints of input and output to determine the required
2620 @c alternative. A restriction would be that constraints used to match
2621 @c against reloads registers would have to be written as register class
2622 @c constraints, or we need a new target macro / hook that tells us if an
2623 @c arbitrary constraint can match an unknown register of a given class.
2624 @c Such a macro / hook would also be useful in other places.]
2627 @var{x} might be a pseudo-register or a @code{subreg} of a
2628 pseudo-register, which could either be in a hard register or in memory.
2629 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2630 in memory and the hard register number if it is in a register.
2632 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2633 currently not supported. For the time being, you will have to continue
2634 to use @code{SECONDARY_MEMORY_NEEDED} for that purpose.
2636 @code{copy_cost} also uses this target hook to find out how values are
2637 copied. If you want it to include some extra cost for the need to allocate
2638 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2639 Or if two dependent moves are supposed to have a lower cost than the sum
2640 of the individual moves due to expected fortuitous scheduling and/or special
2641 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2644 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2645 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2646 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2647 These macros are obsolete, new ports should use the target hook
2648 @code{TARGET_SECONDARY_RELOAD} instead.
2650 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2651 target hook. Older ports still define these macros to indicate to the
2652 reload phase that it may
2653 need to allocate at least one register for a reload in addition to the
2654 register to contain the data. Specifically, if copying @var{x} to a
2655 register @var{class} in @var{mode} requires an intermediate register,
2656 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2657 largest register class all of whose registers can be used as
2658 intermediate registers or scratch registers.
2660 If copying a register @var{class} in @var{mode} to @var{x} requires an
2661 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2662 was supposed to be defined be defined to return the largest register
2663 class required. If the
2664 requirements for input and output reloads were the same, the macro
2665 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2668 The values returned by these macros are often @code{GENERAL_REGS}.
2669 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2670 can be directly copied to or from a register of @var{class} in
2671 @var{mode} without requiring a scratch register. Do not define this
2672 macro if it would always return @code{NO_REGS}.
2674 If a scratch register is required (either with or without an
2675 intermediate register), you were supposed to define patterns for
2676 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2677 (@pxref{Standard Names}. These patterns, which were normally
2678 implemented with a @code{define_expand}, should be similar to the
2679 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2682 These patterns need constraints for the reload register and scratch
2684 contain a single register class. If the original reload register (whose
2685 class is @var{class}) can meet the constraint given in the pattern, the
2686 value returned by these macros is used for the class of the scratch
2687 register. Otherwise, two additional reload registers are required.
2688 Their classes are obtained from the constraints in the insn pattern.
2690 @var{x} might be a pseudo-register or a @code{subreg} of a
2691 pseudo-register, which could either be in a hard register or in memory.
2692 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2693 in memory and the hard register number if it is in a register.
2695 These macros should not be used in the case where a particular class of
2696 registers can only be copied to memory and not to another class of
2697 registers. In that case, secondary reload registers are not needed and
2698 would not be helpful. Instead, a stack location must be used to perform
2699 the copy and the @code{mov@var{m}} pattern should use memory as an
2700 intermediate storage. This case often occurs between floating-point and
2704 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2705 Certain machines have the property that some registers cannot be copied
2706 to some other registers without using memory. Define this macro on
2707 those machines to be a C expression that is nonzero if objects of mode
2708 @var{m} in registers of @var{class1} can only be copied to registers of
2709 class @var{class2} by storing a register of @var{class1} into memory
2710 and loading that memory location into a register of @var{class2}.
2712 Do not define this macro if its value would always be zero.
2715 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2716 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2717 allocates a stack slot for a memory location needed for register copies.
2718 If this macro is defined, the compiler instead uses the memory location
2719 defined by this macro.
2721 Do not define this macro if you do not define
2722 @code{SECONDARY_MEMORY_NEEDED}.
2725 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2726 When the compiler needs a secondary memory location to copy between two
2727 registers of mode @var{mode}, it normally allocates sufficient memory to
2728 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2729 load operations in a mode that many bits wide and whose class is the
2730 same as that of @var{mode}.
2732 This is right thing to do on most machines because it ensures that all
2733 bits of the register are copied and prevents accesses to the registers
2734 in a narrower mode, which some machines prohibit for floating-point
2737 However, this default behavior is not correct on some machines, such as
2738 the DEC Alpha, that store short integers in floating-point registers
2739 differently than in integer registers. On those machines, the default
2740 widening will not work correctly and you must define this macro to
2741 suppress that widening in some cases. See the file @file{alpha.h} for
2744 Do not define this macro if you do not define
2745 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2746 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2749 @defmac SMALL_REGISTER_CLASSES
2750 On some machines, it is risky to let hard registers live across arbitrary
2751 insns. Typically, these machines have instructions that require values
2752 to be in specific registers (like an accumulator), and reload will fail
2753 if the required hard register is used for another purpose across such an
2756 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2757 value on these machines. When this macro has a nonzero value, the
2758 compiler will try to minimize the lifetime of hard registers.
2760 It is always safe to define this macro with a nonzero value, but if you
2761 unnecessarily define it, you will reduce the amount of optimizations
2762 that can be performed in some cases. If you do not define this macro
2763 with a nonzero value when it is required, the compiler will run out of
2764 spill registers and print a fatal error message. For most machines, you
2765 should not define this macro at all.
2768 @defmac CLASS_LIKELY_SPILLED_P (@var{class})
2769 A C expression whose value is nonzero if pseudos that have been assigned
2770 to registers of class @var{class} would likely be spilled because
2771 registers of @var{class} are needed for spill registers.
2773 The default value of this macro returns 1 if @var{class} has exactly one
2774 register and zero otherwise. On most machines, this default should be
2775 used. Only define this macro to some other expression if pseudos
2776 allocated by @file{local-alloc.c} end up in memory because their hard
2777 registers were needed for spill registers. If this macro returns nonzero
2778 for those classes, those pseudos will only be allocated by
2779 @file{global.c}, which knows how to reallocate the pseudo to another
2780 register. If there would not be another register available for
2781 reallocation, you should not change the definition of this macro since
2782 the only effect of such a definition would be to slow down register
2786 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2787 A C expression for the maximum number of consecutive registers
2788 of class @var{class} needed to hold a value of mode @var{mode}.
2790 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2791 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2792 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2793 @var{mode})} for all @var{regno} values in the class @var{class}.
2795 This macro helps control the handling of multiple-word values
2799 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2800 If defined, a C expression that returns nonzero for a @var{class} for which
2801 a change from mode @var{from} to mode @var{to} is invalid.
2803 For the example, loading 32-bit integer or floating-point objects into
2804 floating-point registers on the Alpha extends them to 64 bits.
2805 Therefore loading a 64-bit object and then storing it as a 32-bit object
2806 does not store the low-order 32 bits, as would be the case for a normal
2807 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2811 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2812 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2813 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2817 @node Old Constraints
2818 @section Obsolete Macros for Defining Constraints
2819 @cindex defining constraints, obsolete method
2820 @cindex constraints, defining, obsolete method
2822 Machine-specific constraints can be defined with these macros instead
2823 of the machine description constructs described in @ref{Define
2824 Constraints}. This mechanism is obsolete. New ports should not use
2825 it; old ports should convert to the new mechanism.
2827 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2828 For the constraint at the start of @var{str}, which starts with the letter
2829 @var{c}, return the length. This allows you to have register class /
2830 constant / extra constraints that are longer than a single letter;
2831 you don't need to define this macro if you can do with single-letter
2832 constraints only. The definition of this macro should use
2833 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2834 to handle specially.
2835 There are some sanity checks in genoutput.c that check the constraint lengths
2836 for the md file, so you can also use this macro to help you while you are
2837 transitioning from a byzantine single-letter-constraint scheme: when you
2838 return a negative length for a constraint you want to re-use, genoutput
2839 will complain about every instance where it is used in the md file.
2842 @defmac REG_CLASS_FROM_LETTER (@var{char})
2843 A C expression which defines the machine-dependent operand constraint
2844 letters for register classes. If @var{char} is such a letter, the
2845 value should be the register class corresponding to it. Otherwise,
2846 the value should be @code{NO_REGS}. The register letter @samp{r},
2847 corresponding to class @code{GENERAL_REGS}, will not be passed
2848 to this macro; you do not need to handle it.
2851 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2852 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2853 passed in @var{str}, so that you can use suffixes to distinguish between
2857 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2858 A C expression that defines the machine-dependent operand constraint
2859 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2860 particular ranges of integer values. If @var{c} is one of those
2861 letters, the expression should check that @var{value}, an integer, is in
2862 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2863 not one of those letters, the value should be 0 regardless of
2867 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2868 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2869 string passed in @var{str}, so that you can use suffixes to distinguish
2870 between different variants.
2873 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2874 A C expression that defines the machine-dependent operand constraint
2875 letters that specify particular ranges of @code{const_double} values
2876 (@samp{G} or @samp{H}).
2878 If @var{c} is one of those letters, the expression should check that
2879 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2880 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2881 letters, the value should be 0 regardless of @var{value}.
2883 @code{const_double} is used for all floating-point constants and for
2884 @code{DImode} fixed-point constants. A given letter can accept either
2885 or both kinds of values. It can use @code{GET_MODE} to distinguish
2886 between these kinds.
2889 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2890 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2891 string passed in @var{str}, so that you can use suffixes to distinguish
2892 between different variants.
2895 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2896 A C expression that defines the optional machine-dependent constraint
2897 letters that can be used to segregate specific types of operands, usually
2898 memory references, for the target machine. Any letter that is not
2899 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2900 @code{REG_CLASS_FROM_CONSTRAINT}
2901 may be used. Normally this macro will not be defined.
2903 If it is required for a particular target machine, it should return 1
2904 if @var{value} corresponds to the operand type represented by the
2905 constraint letter @var{c}. If @var{c} is not defined as an extra
2906 constraint, the value returned should be 0 regardless of @var{value}.
2908 For example, on the ROMP, load instructions cannot have their output
2909 in r0 if the memory reference contains a symbolic address. Constraint
2910 letter @samp{Q} is defined as representing a memory address that does
2911 @emph{not} contain a symbolic address. An alternative is specified with
2912 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2913 alternative specifies @samp{m} on the input and a register class that
2914 does not include r0 on the output.
2917 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2918 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2919 in @var{str}, so that you can use suffixes to distinguish between different
2923 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2924 A C expression that defines the optional machine-dependent constraint
2925 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2926 be treated like memory constraints by the reload pass.
2928 It should return 1 if the operand type represented by the constraint
2929 at the start of @var{str}, the first letter of which is the letter @var{c},
2930 comprises a subset of all memory references including
2931 all those whose address is simply a base register. This allows the reload
2932 pass to reload an operand, if it does not directly correspond to the operand
2933 type of @var{c}, by copying its address into a base register.
2935 For example, on the S/390, some instructions do not accept arbitrary
2936 memory references, but only those that do not make use of an index
2937 register. The constraint letter @samp{Q} is defined via
2938 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2939 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2940 a @samp{Q} constraint can handle any memory operand, because the
2941 reload pass knows it can be reloaded by copying the memory address
2942 into a base register if required. This is analogous to the way
2943 a @samp{o} constraint can handle any memory operand.
2946 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2947 A C expression that defines the optional machine-dependent constraint
2948 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2949 @code{EXTRA_CONSTRAINT_STR}, that should
2950 be treated like address constraints by the reload pass.
2952 It should return 1 if the operand type represented by the constraint
2953 at the start of @var{str}, which starts with the letter @var{c}, comprises
2954 a subset of all memory addresses including
2955 all those that consist of just a base register. This allows the reload
2956 pass to reload an operand, if it does not directly correspond to the operand
2957 type of @var{str}, by copying it into a base register.
2959 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2960 be used with the @code{address_operand} predicate. It is treated
2961 analogously to the @samp{p} constraint.
2964 @node Stack and Calling
2965 @section Stack Layout and Calling Conventions
2966 @cindex calling conventions
2968 @c prevent bad page break with this line
2969 This describes the stack layout and calling conventions.
2973 * Exception Handling::
2978 * Register Arguments::
2980 * Aggregate Return::
2985 * Stack Smashing Protection::
2989 @subsection Basic Stack Layout
2990 @cindex stack frame layout
2991 @cindex frame layout
2993 @c prevent bad page break with this line
2994 Here is the basic stack layout.
2996 @defmac STACK_GROWS_DOWNWARD
2997 Define this macro if pushing a word onto the stack moves the stack
2998 pointer to a smaller address.
3000 When we say, ``define this macro if @dots{}'', it means that the
3001 compiler checks this macro only with @code{#ifdef} so the precise
3002 definition used does not matter.
3005 @defmac STACK_PUSH_CODE
3006 This macro defines the operation used when something is pushed
3007 on the stack. In RTL, a push operation will be
3008 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
3010 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
3011 and @code{POST_INC}. Which of these is correct depends on
3012 the stack direction and on whether the stack pointer points
3013 to the last item on the stack or whether it points to the
3014 space for the next item on the stack.
3016 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
3017 defined, which is almost always right, and @code{PRE_INC} otherwise,
3018 which is often wrong.
3021 @defmac FRAME_GROWS_DOWNWARD
3022 Define this macro to nonzero value if the addresses of local variable slots
3023 are at negative offsets from the frame pointer.
3026 @defmac ARGS_GROW_DOWNWARD
3027 Define this macro if successive arguments to a function occupy decreasing
3028 addresses on the stack.
3031 @defmac STARTING_FRAME_OFFSET
3032 Offset from the frame pointer to the first local variable slot to be allocated.
3034 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
3035 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
3036 Otherwise, it is found by adding the length of the first slot to the
3037 value @code{STARTING_FRAME_OFFSET}.
3038 @c i'm not sure if the above is still correct.. had to change it to get
3039 @c rid of an overfull. --mew 2feb93
3042 @defmac STACK_ALIGNMENT_NEEDED
3043 Define to zero to disable final alignment of the stack during reload.
3044 The nonzero default for this macro is suitable for most ports.
3046 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
3047 is a register save block following the local block that doesn't require
3048 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
3049 stack alignment and do it in the backend.
3052 @defmac STACK_POINTER_OFFSET
3053 Offset from the stack pointer register to the first location at which
3054 outgoing arguments are placed. If not specified, the default value of
3055 zero is used. This is the proper value for most machines.
3057 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3058 the first location at which outgoing arguments are placed.
3061 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3062 Offset from the argument pointer register to the first argument's
3063 address. On some machines it may depend on the data type of the
3066 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3067 the first argument's address.
3070 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3071 Offset from the stack pointer register to an item dynamically allocated
3072 on the stack, e.g., by @code{alloca}.
3074 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3075 length of the outgoing arguments. The default is correct for most
3076 machines. See @file{function.c} for details.
3079 @defmac INITIAL_FRAME_ADDRESS_RTX
3080 A C expression whose value is RTL representing the address of the initial
3081 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3082 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3083 default value will be used. Define this macro in order to make frame pointer
3084 elimination work in the presence of @code{__builtin_frame_address (count)} and
3085 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3088 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3089 A C expression whose value is RTL representing the address in a stack
3090 frame where the pointer to the caller's frame is stored. Assume that
3091 @var{frameaddr} is an RTL expression for the address of the stack frame
3094 If you don't define this macro, the default is to return the value
3095 of @var{frameaddr}---that is, the stack frame address is also the
3096 address of the stack word that points to the previous frame.
3099 @defmac SETUP_FRAME_ADDRESSES
3100 If defined, a C expression that produces the machine-specific code to
3101 setup the stack so that arbitrary frames can be accessed. For example,
3102 on the SPARC, we must flush all of the register windows to the stack
3103 before we can access arbitrary stack frames. You will seldom need to
3107 @deftypefn {Target Hook} bool TARGET_BUILTIN_SETJMP_FRAME_VALUE ()
3108 This target hook should return an rtx that is used to store
3109 the address of the current frame into the built in @code{setjmp} buffer.
3110 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3111 machines. One reason you may need to define this target hook is if
3112 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3115 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3116 A C expression whose value is RTL representing the value of the frame
3117 address for the current frame. @var{frameaddr} is the frame pointer
3118 of the current frame. This is used for __builtin_frame_address.
3119 You need only define this macro if the frame address is not the same
3120 as the frame pointer. Most machines do not need to define it.
3123 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3124 A C expression whose value is RTL representing the value of the return
3125 address for the frame @var{count} steps up from the current frame, after
3126 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3127 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3128 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
3130 The value of the expression must always be the correct address when
3131 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
3132 determine the return address of other frames.
3135 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3136 Define this if the return address of a particular stack frame is accessed
3137 from the frame pointer of the previous stack frame.
3140 @defmac INCOMING_RETURN_ADDR_RTX
3141 A C expression whose value is RTL representing the location of the
3142 incoming return address at the beginning of any function, before the
3143 prologue. This RTL is either a @code{REG}, indicating that the return
3144 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3147 You only need to define this macro if you want to support call frame
3148 debugging information like that provided by DWARF 2.
3150 If this RTL is a @code{REG}, you should also define
3151 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3154 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3155 A C expression whose value is an integer giving a DWARF 2 column
3156 number that may be used as an alternative return column. The column
3157 must not correspond to any gcc hard register (that is, it must not
3158 be in the range of @code{DWARF_FRAME_REGNUM}).
3160 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3161 general register, but an alternative column needs to be used for signal
3162 frames. Some targets have also used different frame return columns
3166 @defmac DWARF_ZERO_REG
3167 A C expression whose value is an integer giving a DWARF 2 register
3168 number that is considered to always have the value zero. This should
3169 only be defined if the target has an architected zero register, and
3170 someone decided it was a good idea to use that register number to
3171 terminate the stack backtrace. New ports should avoid this.
3174 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3175 This target hook allows the backend to emit frame-related insns that
3176 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3177 info engine will invoke it on insns of the form
3179 (set (reg) (unspec [@dots{}] UNSPEC_INDEX))
3183 (set (reg) (unspec_volatile [@dots{}] UNSPECV_INDEX)).
3185 to let the backend emit the call frame instructions. @var{label} is
3186 the CFI label attached to the insn, @var{pattern} is the pattern of
3187 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3190 @defmac INCOMING_FRAME_SP_OFFSET
3191 A C expression whose value is an integer giving the offset, in bytes,
3192 from the value of the stack pointer register to the top of the stack
3193 frame at the beginning of any function, before the prologue. The top of
3194 the frame is defined to be the value of the stack pointer in the
3195 previous frame, just before the call instruction.
3197 You only need to define this macro if you want to support call frame
3198 debugging information like that provided by DWARF 2.
3201 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3202 A C expression whose value is an integer giving the offset, in bytes,
3203 from the argument pointer to the canonical frame address (cfa). The
3204 final value should coincide with that calculated by
3205 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3206 during virtual register instantiation.
3208 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
3209 which is correct for most machines; in general, the arguments are found
3210 immediately before the stack frame. Note that this is not the case on
3211 some targets that save registers into the caller's frame, such as SPARC
3212 and rs6000, and so such targets need to define this macro.
3214 You only need to define this macro if the default is incorrect, and you
3215 want to support call frame debugging information like that provided by
3219 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3220 If defined, a C expression whose value is an integer giving the offset
3221 in bytes from the frame pointer to the canonical frame address (cfa).
3222 The final value should coincide with that calculated by
3223 @code{INCOMING_FRAME_SP_OFFSET}.
3225 Normally the CFA is calculated as an offset from the argument pointer,
3226 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3227 variable due to the ABI, this may not be possible. If this macro is
3228 defined, it implies that the virtual register instantiation should be
3229 based on the frame pointer instead of the argument pointer. Only one
3230 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3234 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3235 If defined, a C expression whose value is an integer giving the offset
3236 in bytes from the canonical frame address (cfa) to the frame base used
3237 in DWARF 2 debug information. The default is zero. A different value
3238 may reduce the size of debug information on some ports.
3241 @node Exception Handling
3242 @subsection Exception Handling Support
3243 @cindex exception handling
3245 @defmac EH_RETURN_DATA_REGNO (@var{N})
3246 A C expression whose value is the @var{N}th register number used for
3247 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3248 @var{N} registers are usable.
3250 The exception handling library routines communicate with the exception
3251 handlers via a set of agreed upon registers. Ideally these registers
3252 should be call-clobbered; it is possible to use call-saved registers,
3253 but may negatively impact code size. The target must support at least
3254 2 data registers, but should define 4 if there are enough free registers.
3256 You must define this macro if you want to support call frame exception
3257 handling like that provided by DWARF 2.
3260 @defmac EH_RETURN_STACKADJ_RTX
3261 A C expression whose value is RTL representing a location in which
3262 to store a stack adjustment to be applied before function return.
3263 This is used to unwind the stack to an exception handler's call frame.
3264 It will be assigned zero on code paths that return normally.
3266 Typically this is a call-clobbered hard register that is otherwise
3267 untouched by the epilogue, but could also be a stack slot.
3269 Do not define this macro if the stack pointer is saved and restored
3270 by the regular prolog and epilog code in the call frame itself; in
3271 this case, the exception handling library routines will update the
3272 stack location to be restored in place. Otherwise, you must define
3273 this macro if you want to support call frame exception handling like
3274 that provided by DWARF 2.
3277 @defmac EH_RETURN_HANDLER_RTX
3278 A C expression whose value is RTL representing a location in which
3279 to store the address of an exception handler to which we should
3280 return. It will not be assigned on code paths that return normally.
3282 Typically this is the location in the call frame at which the normal
3283 return address is stored. For targets that return by popping an
3284 address off the stack, this might be a memory address just below
3285 the @emph{target} call frame rather than inside the current call
3286 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3287 been assigned, so it may be used to calculate the location of the
3290 Some targets have more complex requirements than storing to an
3291 address calculable during initial code generation. In that case
3292 the @code{eh_return} instruction pattern should be used instead.
3294 If you want to support call frame exception handling, you must
3295 define either this macro or the @code{eh_return} instruction pattern.
3298 @defmac RETURN_ADDR_OFFSET
3299 If defined, an integer-valued C expression for which rtl will be generated
3300 to add it to the exception handler address before it is searched in the
3301 exception handling tables, and to subtract it again from the address before
3302 using it to return to the exception handler.
3305 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3306 This macro chooses the encoding of pointers embedded in the exception
3307 handling sections. If at all possible, this should be defined such
3308 that the exception handling section will not require dynamic relocations,
3309 and so may be read-only.
3311 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3312 @var{global} is true if the symbol may be affected by dynamic relocations.
3313 The macro should return a combination of the @code{DW_EH_PE_*} defines
3314 as found in @file{dwarf2.h}.
3316 If this macro is not defined, pointers will not be encoded but
3317 represented directly.
3320 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3321 This macro allows the target to emit whatever special magic is required
3322 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3323 Generic code takes care of pc-relative and indirect encodings; this must
3324 be defined if the target uses text-relative or data-relative encodings.
3326 This is a C statement that branches to @var{done} if the format was
3327 handled. @var{encoding} is the format chosen, @var{size} is the number
3328 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3332 @defmac MD_UNWIND_SUPPORT
3333 A string specifying a file to be #include'd in unwind-dw2.c. The file
3334 so included typically defines @code{MD_FALLBACK_FRAME_STATE_FOR}.
3337 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3338 This macro allows the target to add CPU and operating system specific
3339 code to the call-frame unwinder for use when there is no unwind data
3340 available. The most common reason to implement this macro is to unwind
3341 through signal frames.
3343 This macro is called from @code{uw_frame_state_for} in
3344 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
3345 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3346 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3347 for the address of the code being executed and @code{context->cfa} for
3348 the stack pointer value. If the frame can be decoded, the register
3349 save addresses should be updated in @var{fs} and the macro should
3350 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
3351 the macro should evaluate to @code{_URC_END_OF_STACK}.
3353 For proper signal handling in Java this macro is accompanied by
3354 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3357 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3358 This macro allows the target to add operating system specific code to the
3359 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3360 usually used for signal or interrupt frames.
3362 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3363 @var{context} is an @code{_Unwind_Context};
3364 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3365 for the abi and context in the @code{.unwabi} directive. If the
3366 @code{.unwabi} directive can be handled, the register save addresses should
3367 be updated in @var{fs}.
3370 @defmac TARGET_USES_WEAK_UNWIND_INFO
3371 A C expression that evaluates to true if the target requires unwind
3372 info to be given comdat linkage. Define it to be @code{1} if comdat
3373 linkage is necessary. The default is @code{0}.
3376 @node Stack Checking
3377 @subsection Specifying How Stack Checking is Done
3379 GCC will check that stack references are within the boundaries of the
3380 stack, if the option @option{-fstack-check} is specified, in one of
3385 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3386 will assume that you have arranged for full stack checking to be done
3387 at appropriate places in the configuration files. GCC will not do
3388 other special processing.
3391 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
3392 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
3393 that you have arranged for static stack checking (checking of the
3394 static stack frame of functions) to be done at appropriate places
3395 in the configuration files. GCC will only emit code to do dynamic
3396 stack checking (checking on dynamic stack allocations) using the third
3400 If neither of the above are true, GCC will generate code to periodically
3401 ``probe'' the stack pointer using the values of the macros defined below.
3404 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
3405 GCC will change its allocation strategy for large objects if the option
3406 @option{-fstack-check} is specified: they will always be allocated
3407 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
3409 @defmac STACK_CHECK_BUILTIN
3410 A nonzero value if stack checking is done by the configuration files in a
3411 machine-dependent manner. You should define this macro if stack checking
3412 is require by the ABI of your machine or if you would like to do stack
3413 checking in some more efficient way than the generic approach. The default
3414 value of this macro is zero.
3417 @defmac STACK_CHECK_STATIC_BUILTIN
3418 A nonzero value if static stack checking is done by the configuration files
3419 in a machine-dependent manner. You should define this macro if you would
3420 like to do static stack checking in some more efficient way than the generic
3421 approach. The default value of this macro is zero.
3424 @defmac STACK_CHECK_PROBE_INTERVAL
3425 An integer representing the interval at which GCC must generate stack
3426 probe instructions. You will normally define this macro to be no larger
3427 than the size of the ``guard pages'' at the end of a stack area. The
3428 default value of 4096 is suitable for most systems.
3431 @defmac STACK_CHECK_PROBE_LOAD
3432 An integer which is nonzero if GCC should perform the stack probe
3433 as a load instruction and zero if GCC should use a store instruction.
3434 The default is zero, which is the most efficient choice on most systems.
3437 @defmac STACK_CHECK_PROTECT
3438 The number of bytes of stack needed to recover from a stack overflow,
3439 for languages where such a recovery is supported. The default value of
3440 75 words should be adequate for most machines.
3443 The following macros are relevant only if neither STACK_CHECK_BUILTIN
3444 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
3445 in the opposite case.
3447 @defmac STACK_CHECK_MAX_FRAME_SIZE
3448 The maximum size of a stack frame, in bytes. GCC will generate probe
3449 instructions in non-leaf functions to ensure at least this many bytes of
3450 stack are available. If a stack frame is larger than this size, stack
3451 checking will not be reliable and GCC will issue a warning. The
3452 default is chosen so that GCC only generates one instruction on most
3453 systems. You should normally not change the default value of this macro.
3456 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3457 GCC uses this value to generate the above warning message. It
3458 represents the amount of fixed frame used by a function, not including
3459 space for any callee-saved registers, temporaries and user variables.
3460 You need only specify an upper bound for this amount and will normally
3461 use the default of four words.
3464 @defmac STACK_CHECK_MAX_VAR_SIZE
3465 The maximum size, in bytes, of an object that GCC will place in the
3466 fixed area of the stack frame when the user specifies
3467 @option{-fstack-check}.
3468 GCC computed the default from the values of the above macros and you will
3469 normally not need to override that default.
3473 @node Frame Registers
3474 @subsection Registers That Address the Stack Frame
3476 @c prevent bad page break with this line
3477 This discusses registers that address the stack frame.
3479 @defmac STACK_POINTER_REGNUM
3480 The register number of the stack pointer register, which must also be a
3481 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3482 the hardware determines which register this is.
3485 @defmac FRAME_POINTER_REGNUM
3486 The register number of the frame pointer register, which is used to
3487 access automatic variables in the stack frame. On some machines, the
3488 hardware determines which register this is. On other machines, you can
3489 choose any register you wish for this purpose.
3492 @defmac HARD_FRAME_POINTER_REGNUM
3493 On some machines the offset between the frame pointer and starting
3494 offset of the automatic variables is not known until after register
3495 allocation has been done (for example, because the saved registers are
3496 between these two locations). On those machines, define
3497 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3498 be used internally until the offset is known, and define
3499 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3500 used for the frame pointer.
3502 You should define this macro only in the very rare circumstances when it
3503 is not possible to calculate the offset between the frame pointer and
3504 the automatic variables until after register allocation has been
3505 completed. When this macro is defined, you must also indicate in your
3506 definition of @code{ELIMINABLE_REGS} how to eliminate
3507 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3508 or @code{STACK_POINTER_REGNUM}.
3510 Do not define this macro if it would be the same as
3511 @code{FRAME_POINTER_REGNUM}.
3514 @defmac ARG_POINTER_REGNUM
3515 The register number of the arg pointer register, which is used to access
3516 the function's argument list. On some machines, this is the same as the
3517 frame pointer register. On some machines, the hardware determines which
3518 register this is. On other machines, you can choose any register you
3519 wish for this purpose. If this is not the same register as the frame
3520 pointer register, then you must mark it as a fixed register according to
3521 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3522 (@pxref{Elimination}).
3525 @defmac RETURN_ADDRESS_POINTER_REGNUM
3526 The register number of the return address pointer register, which is used to
3527 access the current function's return address from the stack. On some
3528 machines, the return address is not at a fixed offset from the frame
3529 pointer or stack pointer or argument pointer. This register can be defined
3530 to point to the return address on the stack, and then be converted by
3531 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3533 Do not define this macro unless there is no other way to get the return
3534 address from the stack.
3537 @defmac STATIC_CHAIN_REGNUM
3538 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3539 Register numbers used for passing a function's static chain pointer. If
3540 register windows are used, the register number as seen by the called
3541 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3542 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3543 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3546 The static chain register need not be a fixed register.
3548 If the static chain is passed in memory, these macros should not be
3549 defined; instead, the next two macros should be defined.
3552 @defmac STATIC_CHAIN
3553 @defmacx STATIC_CHAIN_INCOMING
3554 If the static chain is passed in memory, these macros provide rtx giving
3555 @code{mem} expressions that denote where they are stored.
3556 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3557 as seen by the calling and called functions, respectively. Often the former
3558 will be at an offset from the stack pointer and the latter at an offset from
3561 @findex stack_pointer_rtx
3562 @findex frame_pointer_rtx
3563 @findex arg_pointer_rtx
3564 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3565 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3566 macros and should be used to refer to those items.
3568 If the static chain is passed in a register, the two previous macros should
3572 @defmac DWARF_FRAME_REGISTERS
3573 This macro specifies the maximum number of hard registers that can be
3574 saved in a call frame. This is used to size data structures used in
3575 DWARF2 exception handling.
3577 Prior to GCC 3.0, this macro was needed in order to establish a stable
3578 exception handling ABI in the face of adding new hard registers for ISA
3579 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3580 in the number of hard registers. Nevertheless, this macro can still be
3581 used to reduce the runtime memory requirements of the exception handling
3582 routines, which can be substantial if the ISA contains a lot of
3583 registers that are not call-saved.
3585 If this macro is not defined, it defaults to
3586 @code{FIRST_PSEUDO_REGISTER}.
3589 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3591 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3592 for backward compatibility in pre GCC 3.0 compiled code.
3594 If this macro is not defined, it defaults to
3595 @code{DWARF_FRAME_REGISTERS}.
3598 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3600 Define this macro if the target's representation for dwarf registers
3601 is different than the internal representation for unwind column.
3602 Given a dwarf register, this macro should return the internal unwind
3603 column number to use instead.
3605 See the PowerPC's SPE target for an example.
3608 @defmac DWARF_FRAME_REGNUM (@var{regno})
3610 Define this macro if the target's representation for dwarf registers
3611 used in .eh_frame or .debug_frame is different from that used in other
3612 debug info sections. Given a GCC hard register number, this macro
3613 should return the .eh_frame register number. The default is
3614 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3618 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3620 Define this macro to map register numbers held in the call frame info
3621 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3622 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3623 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3624 return @code{@var{regno}}.
3629 @subsection Eliminating Frame Pointer and Arg Pointer
3631 @c prevent bad page break with this line
3632 This is about eliminating the frame pointer and arg pointer.
3634 @defmac FRAME_POINTER_REQUIRED
3635 A C expression which is nonzero if a function must have and use a frame
3636 pointer. This expression is evaluated in the reload pass. If its value is
3637 nonzero the function will have a frame pointer.
3639 The expression can in principle examine the current function and decide
3640 according to the facts, but on most machines the constant 0 or the
3641 constant 1 suffices. Use 0 when the machine allows code to be generated
3642 with no frame pointer, and doing so saves some time or space. Use 1
3643 when there is no possible advantage to avoiding a frame pointer.
3645 In certain cases, the compiler does not know how to produce valid code
3646 without a frame pointer. The compiler recognizes those cases and
3647 automatically gives the function a frame pointer regardless of what
3648 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3651 In a function that does not require a frame pointer, the frame pointer
3652 register can be allocated for ordinary usage, unless you mark it as a
3653 fixed register. See @code{FIXED_REGISTERS} for more information.
3656 @findex get_frame_size
3657 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3658 A C statement to store in the variable @var{depth-var} the difference
3659 between the frame pointer and the stack pointer values immediately after
3660 the function prologue. The value would be computed from information
3661 such as the result of @code{get_frame_size ()} and the tables of
3662 registers @code{regs_ever_live} and @code{call_used_regs}.
3664 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3665 need not be defined. Otherwise, it must be defined even if
3666 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3667 case, you may set @var{depth-var} to anything.
3670 @defmac ELIMINABLE_REGS
3671 If defined, this macro specifies a table of register pairs used to
3672 eliminate unneeded registers that point into the stack frame. If it is not
3673 defined, the only elimination attempted by the compiler is to replace
3674 references to the frame pointer with references to the stack pointer.
3676 The definition of this macro is a list of structure initializations, each
3677 of which specifies an original and replacement register.
3679 On some machines, the position of the argument pointer is not known until
3680 the compilation is completed. In such a case, a separate hard register
3681 must be used for the argument pointer. This register can be eliminated by
3682 replacing it with either the frame pointer or the argument pointer,
3683 depending on whether or not the frame pointer has been eliminated.
3685 In this case, you might specify:
3687 #define ELIMINABLE_REGS \
3688 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3689 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3690 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3693 Note that the elimination of the argument pointer with the stack pointer is
3694 specified first since that is the preferred elimination.
3697 @defmac CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3698 A C expression that returns nonzero if the compiler is allowed to try
3699 to replace register number @var{from-reg} with register number
3700 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3701 is defined, and will usually be the constant 1, since most of the cases
3702 preventing register elimination are things that the compiler already
3706 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3707 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3708 specifies the initial difference between the specified pair of
3709 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3713 @node Stack Arguments
3714 @subsection Passing Function Arguments on the Stack
3715 @cindex arguments on stack
3716 @cindex stack arguments
3718 The macros in this section control how arguments are passed
3719 on the stack. See the following section for other macros that
3720 control passing certain arguments in registers.
3722 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (tree @var{fntype})
3723 This target hook returns @code{true} if an argument declared in a
3724 prototype as an integral type smaller than @code{int} should actually be
3725 passed as an @code{int}. In addition to avoiding errors in certain
3726 cases of mismatch, it also makes for better code on certain machines.
3727 The default is to not promote prototypes.
3731 A C expression. If nonzero, push insns will be used to pass
3733 If the target machine does not have a push instruction, set it to zero.
3734 That directs GCC to use an alternate strategy: to
3735 allocate the entire argument block and then store the arguments into
3736 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3739 @defmac PUSH_ARGS_REVERSED
3740 A C expression. If nonzero, function arguments will be evaluated from
3741 last to first, rather than from first to last. If this macro is not
3742 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3743 and args grow in opposite directions, and 0 otherwise.
3746 @defmac PUSH_ROUNDING (@var{npushed})
3747 A C expression that is the number of bytes actually pushed onto the
3748 stack when an instruction attempts to push @var{npushed} bytes.
3750 On some machines, the definition
3753 #define PUSH_ROUNDING(BYTES) (BYTES)
3757 will suffice. But on other machines, instructions that appear
3758 to push one byte actually push two bytes in an attempt to maintain
3759 alignment. Then the definition should be
3762 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3766 @findex current_function_outgoing_args_size
3767 @defmac ACCUMULATE_OUTGOING_ARGS
3768 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3769 will be computed and placed into the variable
3770 @code{current_function_outgoing_args_size}. No space will be pushed
3771 onto the stack for each call; instead, the function prologue should
3772 increase the stack frame size by this amount.
3774 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3778 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3779 Define this macro if functions should assume that stack space has been
3780 allocated for arguments even when their values are passed in
3783 The value of this macro is the size, in bytes, of the area reserved for
3784 arguments passed in registers for the function represented by @var{fndecl},
3785 which can be zero if GCC is calling a library function.
3787 This space can be allocated by the caller, or be a part of the
3788 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3791 @c above is overfull. not sure what to do. --mew 5feb93 did
3792 @c something, not sure if it looks good. --mew 10feb93
3794 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3795 Define this to a nonzero value if it is the responsibility of the
3796 caller to allocate the area reserved for arguments passed in registers
3797 when calling a function of @var{fntype}. @var{fntype} may be NULL
3798 if the function called is a library function.
3800 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3801 whether the space for these arguments counts in the value of
3802 @code{current_function_outgoing_args_size}.
3805 @defmac STACK_PARMS_IN_REG_PARM_AREA
3806 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3807 stack parameters don't skip the area specified by it.
3808 @c i changed this, makes more sens and it should have taken care of the
3809 @c overfull.. not as specific, tho. --mew 5feb93
3811 Normally, when a parameter is not passed in registers, it is placed on the
3812 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3813 suppresses this behavior and causes the parameter to be passed on the
3814 stack in its natural location.
3817 @defmac RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3818 A C expression that should indicate the number of bytes of its own
3819 arguments that a function pops on returning, or 0 if the
3820 function pops no arguments and the caller must therefore pop them all
3821 after the function returns.
3823 @var{fundecl} is a C variable whose value is a tree node that describes
3824 the function in question. Normally it is a node of type
3825 @code{FUNCTION_DECL} that describes the declaration of the function.
3826 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3828 @var{funtype} is a C variable whose value is a tree node that
3829 describes the function in question. Normally it is a node of type
3830 @code{FUNCTION_TYPE} that describes the data type of the function.
3831 From this it is possible to obtain the data types of the value and
3832 arguments (if known).
3834 When a call to a library function is being considered, @var{fundecl}
3835 will contain an identifier node for the library function. Thus, if
3836 you need to distinguish among various library functions, you can do so
3837 by their names. Note that ``library function'' in this context means
3838 a function used to perform arithmetic, whose name is known specially
3839 in the compiler and was not mentioned in the C code being compiled.
3841 @var{stack-size} is the number of bytes of arguments passed on the
3842 stack. If a variable number of bytes is passed, it is zero, and
3843 argument popping will always be the responsibility of the calling function.
3845 On the VAX, all functions always pop their arguments, so the definition
3846 of this macro is @var{stack-size}. On the 68000, using the standard
3847 calling convention, no functions pop their arguments, so the value of
3848 the macro is always 0 in this case. But an alternative calling
3849 convention is available in which functions that take a fixed number of
3850 arguments pop them but other functions (such as @code{printf}) pop
3851 nothing (the caller pops all). When this convention is in use,
3852 @var{funtype} is examined to determine whether a function takes a fixed
3853 number of arguments.
3856 @defmac CALL_POPS_ARGS (@var{cum})
3857 A C expression that should indicate the number of bytes a call sequence
3858 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3859 when compiling a function call.
3861 @var{cum} is the variable in which all arguments to the called function
3862 have been accumulated.
3864 On certain architectures, such as the SH5, a call trampoline is used
3865 that pops certain registers off the stack, depending on the arguments
3866 that have been passed to the function. Since this is a property of the
3867 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3871 @node Register Arguments
3872 @subsection Passing Arguments in Registers
3873 @cindex arguments in registers
3874 @cindex registers arguments
3876 This section describes the macros which let you control how various
3877 types of arguments are passed in registers or how they are arranged in
3880 @defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3881 A C expression that controls whether a function argument is passed
3882 in a register, and which register.
3884 The arguments are @var{cum}, which summarizes all the previous
3885 arguments; @var{mode}, the machine mode of the argument; @var{type},
3886 the data type of the argument as a tree node or 0 if that is not known
3887 (which happens for C support library functions); and @var{named},
3888 which is 1 for an ordinary argument and 0 for nameless arguments that
3889 correspond to @samp{@dots{}} in the called function's prototype.
3890 @var{type} can be an incomplete type if a syntax error has previously
3893 The value of the expression is usually either a @code{reg} RTX for the
3894 hard register in which to pass the argument, or zero to pass the
3895 argument on the stack.
3897 For machines like the VAX and 68000, where normally all arguments are
3898 pushed, zero suffices as a definition.
3900 The value of the expression can also be a @code{parallel} RTX@. This is
3901 used when an argument is passed in multiple locations. The mode of the
3902 @code{parallel} should be the mode of the entire argument. The
3903 @code{parallel} holds any number of @code{expr_list} pairs; each one
3904 describes where part of the argument is passed. In each
3905 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3906 register in which to pass this part of the argument, and the mode of the
3907 register RTX indicates how large this part of the argument is. The
3908 second operand of the @code{expr_list} is a @code{const_int} which gives
3909 the offset in bytes into the entire argument of where this part starts.
3910 As a special exception the first @code{expr_list} in the @code{parallel}
3911 RTX may have a first operand of zero. This indicates that the entire
3912 argument is also stored on the stack.
3914 The last time this macro is called, it is called with @code{MODE ==
3915 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3916 pattern as operands 2 and 3 respectively.
3918 @cindex @file{stdarg.h} and register arguments
3919 The usual way to make the ISO library @file{stdarg.h} work on a machine
3920 where some arguments are usually passed in registers, is to cause
3921 nameless arguments to be passed on the stack instead. This is done
3922 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3924 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3925 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3926 You may use the hook @code{targetm.calls.must_pass_in_stack}
3927 in the definition of this macro to determine if this argument is of a
3928 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3929 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3930 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3931 defined, the argument will be computed in the stack and then loaded into
3935 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, tree @var{type})
3936 This target hook should return @code{true} if we should not pass @var{type}
3937 solely in registers. The file @file{expr.h} defines a
3938 definition that is usually appropriate, refer to @file{expr.h} for additional
3942 @defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3943 Define this macro if the target machine has ``register windows'', so
3944 that the register in which a function sees an arguments is not
3945 necessarily the same as the one in which the caller passed the
3948 For such machines, @code{FUNCTION_ARG} computes the register in which
3949 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3950 be defined in a similar fashion to tell the function being called
3951 where the arguments will arrive.
3953 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3954 serves both purposes.
3957 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3958 This target hook returns the number of bytes at the beginning of an
3959 argument that must be put in registers. The value must be zero for
3960 arguments that are passed entirely in registers or that are entirely
3961 pushed on the stack.
3963 On some machines, certain arguments must be passed partially in
3964 registers and partially in memory. On these machines, typically the
3965 first few words of arguments are passed in registers, and the rest
3966 on the stack. If a multi-word argument (a @code{double} or a
3967 structure) crosses that boundary, its first few words must be passed
3968 in registers and the rest must be pushed. This macro tells the
3969 compiler when this occurs, and how many bytes should go in registers.
3971 @code{FUNCTION_ARG} for these arguments should return the first
3972 register to be used by the caller for this argument; likewise
3973 @code{FUNCTION_INCOMING_ARG}, for the called function.
3976 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3977 This target hook should return @code{true} if an argument at the
3978 position indicated by @var{cum} should be passed by reference. This
3979 predicate is queried after target independent reasons for being
3980 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
3982 If the hook returns true, a copy of that argument is made in memory and a
3983 pointer to the argument is passed instead of the argument itself.
3984 The pointer is passed in whatever way is appropriate for passing a pointer
3988 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3989 The function argument described by the parameters to this hook is
3990 known to be passed by reference. The hook should return true if the
3991 function argument should be copied by the callee instead of copied
3994 For any argument for which the hook returns true, if it can be
3995 determined that the argument is not modified, then a copy need
3998 The default version of this hook always returns false.
4001 @defmac CUMULATIVE_ARGS
4002 A C type for declaring a variable that is used as the first argument of
4003 @code{FUNCTION_ARG} and other related values. For some target machines,
4004 the type @code{int} suffices and can hold the number of bytes of
4007 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
4008 arguments that have been passed on the stack. The compiler has other
4009 variables to keep track of that. For target machines on which all
4010 arguments are passed on the stack, there is no need to store anything in
4011 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
4012 should not be empty, so use @code{int}.
4015 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
4016 If defined, this macro is called before generating any code for a
4017 function, but after the @var{cfun} descriptor for the function has been
4018 created. The back end may use this macro to update @var{cfun} to
4019 reflect an ABI other than that which would normally be used by default.
4020 If the compiler is generating code for a compiler-generated function,
4021 @var{fndecl} may be @code{NULL}.
4024 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
4025 A C statement (sans semicolon) for initializing the variable
4026 @var{cum} for the state at the beginning of the argument list. The
4027 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
4028 is the tree node for the data type of the function which will receive
4029 the args, or 0 if the args are to a compiler support library function.
4030 For direct calls that are not libcalls, @var{fndecl} contain the
4031 declaration node of the function. @var{fndecl} is also set when
4032 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
4033 being compiled. @var{n_named_args} is set to the number of named
4034 arguments, including a structure return address if it is passed as a
4035 parameter, when making a call. When processing incoming arguments,
4036 @var{n_named_args} is set to @minus{}1.
4038 When processing a call to a compiler support library function,
4039 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
4040 contains the name of the function, as a string. @var{libname} is 0 when
4041 an ordinary C function call is being processed. Thus, each time this
4042 macro is called, either @var{libname} or @var{fntype} is nonzero, but
4043 never both of them at once.
4046 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
4047 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
4048 it gets a @code{MODE} argument instead of @var{fntype}, that would be
4049 @code{NULL}. @var{indirect} would always be zero, too. If this macro
4050 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
4051 0)} is used instead.
4054 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
4055 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
4056 finding the arguments for the function being compiled. If this macro is
4057 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
4059 The value passed for @var{libname} is always 0, since library routines
4060 with special calling conventions are never compiled with GCC@. The
4061 argument @var{libname} exists for symmetry with
4062 @code{INIT_CUMULATIVE_ARGS}.
4063 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
4064 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
4067 @defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
4068 A C statement (sans semicolon) to update the summarizer variable
4069 @var{cum} to advance past an argument in the argument list. The
4070 values @var{mode}, @var{type} and @var{named} describe that argument.
4071 Once this is done, the variable @var{cum} is suitable for analyzing
4072 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
4074 This macro need not do anything if the argument in question was passed
4075 on the stack. The compiler knows how to track the amount of stack space
4076 used for arguments without any special help.
4079 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
4080 If defined, a C expression which determines whether, and in which direction,
4081 to pad out an argument with extra space. The value should be of type
4082 @code{enum direction}: either @code{upward} to pad above the argument,
4083 @code{downward} to pad below, or @code{none} to inhibit padding.
4085 The @emph{amount} of padding is always just enough to reach the next
4086 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
4089 This macro has a default definition which is right for most systems.
4090 For little-endian machines, the default is to pad upward. For
4091 big-endian machines, the default is to pad downward for an argument of
4092 constant size shorter than an @code{int}, and upward otherwise.
4095 @defmac PAD_VARARGS_DOWN
4096 If defined, a C expression which determines whether the default
4097 implementation of va_arg will attempt to pad down before reading the
4098 next argument, if that argument is smaller than its aligned space as
4099 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4100 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4103 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4104 Specify padding for the last element of a block move between registers and
4105 memory. @var{first} is nonzero if this is the only element. Defining this
4106 macro allows better control of register function parameters on big-endian
4107 machines, without using @code{PARALLEL} rtl. In particular,
4108 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4109 registers, as there is no longer a "wrong" part of a register; For example,
4110 a three byte aggregate may be passed in the high part of a register if so
4114 @defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
4115 If defined, a C expression that gives the alignment boundary, in bits,
4116 of an argument with the specified mode and type. If it is not defined,
4117 @code{PARM_BOUNDARY} is used for all arguments.
4120 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4121 A C expression that is nonzero if @var{regno} is the number of a hard
4122 register in which function arguments are sometimes passed. This does
4123 @emph{not} include implicit arguments such as the static chain and
4124 the structure-value address. On many machines, no registers can be
4125 used for this purpose since all function arguments are pushed on the
4129 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (tree @var{type})
4130 This hook should return true if parameter of type @var{type} are passed
4131 as two scalar parameters. By default, GCC will attempt to pack complex
4132 arguments into the target's word size. Some ABIs require complex arguments
4133 to be split and treated as their individual components. For example, on
4134 AIX64, complex floats should be passed in a pair of floating point
4135 registers, even though a complex float would fit in one 64-bit floating
4138 The default value of this hook is @code{NULL}, which is treated as always
4142 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
4143 This hook returns a type node for @code{va_list} for the target.
4144 The default version of the hook returns @code{void*}.
4147 @deftypefn {Target Hook} tree TARGET_FN_ABI_VA_LIST (tree @var{fndecl})
4148 This hook returns the va_list type of the calling convention specified by
4150 The default version of this hook returns @code{va_list_type_node}.
4153 @deftypefn {Target Hook} tree TARGET_CANONICAL_VA_LIST_TYPE (tree @var{type})
4154 This hook returns the va_list type of the calling convention specified by the
4155 type of @var{type}. If @var{type} is not a valid va_list type, it returns
4159 @deftypefn {Target Hook} tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree @var{valist}, tree @var{type}, tree *@var{pre_p}, tree *@var{post_p})
4160 This hook performs target-specific gimplification of
4161 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4162 arguments to @code{va_arg}; the latter two are as in
4163 @code{gimplify.c:gimplify_expr}.
4166 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (enum machine_mode @var{mode})
4167 Define this to return nonzero if the port can handle pointers
4168 with machine mode @var{mode}. The default version of this
4169 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4172 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4173 Define this to return nonzero if the port is prepared to handle
4174 insns involving scalar mode @var{mode}. For a scalar mode to be
4175 considered supported, all the basic arithmetic and comparisons
4178 The default version of this hook returns true for any mode
4179 required to handle the basic C types (as defined by the port).
4180 Included here are the double-word arithmetic supported by the
4181 code in @file{optabs.c}.
4184 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4185 Define this to return nonzero if the port is prepared to handle
4186 insns involving vector mode @var{mode}. At the very least, it
4187 must have move patterns for this mode.
4191 @subsection How Scalar Function Values Are Returned
4192 @cindex return values in registers
4193 @cindex values, returned by functions
4194 @cindex scalars, returned as values
4196 This section discusses the macros that control returning scalars as
4197 values---values that can fit in registers.
4199 @deftypefn {Target Hook} rtx TARGET_FUNCTION_VALUE (tree @var{ret_type}, tree @var{fn_decl_or_type}, bool @var{outgoing})
4201 Define this to return an RTX representing the place where a function
4202 returns or receives a value of data type @var{ret_type}, a tree node
4203 node representing a data type. @var{fn_decl_or_type} is a tree node
4204 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4205 function being called. If @var{outgoing} is false, the hook should
4206 compute the register in which the caller will see the return value.
4207 Otherwise, the hook should return an RTX representing the place where
4208 a function returns a value.
4210 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4211 (Actually, on most machines, scalar values are returned in the same
4212 place regardless of mode.) The value of the expression is usually a
4213 @code{reg} RTX for the hard register where the return value is stored.
4214 The value can also be a @code{parallel} RTX, if the return value is in
4215 multiple places. See @code{FUNCTION_ARG} for an explanation of the
4216 @code{parallel} form. Note that the callee will populate every
4217 location specified in the @code{parallel}, but if the first element of
4218 the @code{parallel} contains the whole return value, callers will use
4219 that element as the canonical location and ignore the others. The m68k
4220 port uses this type of @code{parallel} to return pointers in both
4221 @samp{%a0} (the canonical location) and @samp{%d0}.
4223 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4224 the same promotion rules specified in @code{PROMOTE_MODE} if
4225 @var{valtype} is a scalar type.
4227 If the precise function being called is known, @var{func} is a tree
4228 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4229 pointer. This makes it possible to use a different value-returning
4230 convention for specific functions when all their calls are
4233 Some target machines have ``register windows'' so that the register in
4234 which a function returns its value is not the same as the one in which
4235 the caller sees the value. For such machines, you should return
4236 different RTX depending on @var{outgoing}.
4238 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4239 aggregate data types, because these are returned in another way. See
4240 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4243 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4244 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4245 a new target instead.
4248 @defmac FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
4249 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4250 a new target instead.
4253 @defmac LIBCALL_VALUE (@var{mode})
4254 A C expression to create an RTX representing the place where a library
4255 function returns a value of mode @var{mode}. If the precise function
4256 being called is known, @var{func} is a tree node
4257 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4258 pointer. This makes it possible to use a different value-returning
4259 convention for specific functions when all their calls are
4262 Note that ``library function'' in this context means a compiler
4263 support routine, used to perform arithmetic, whose name is known
4264 specially by the compiler and was not mentioned in the C code being
4267 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
4268 data types, because none of the library functions returns such types.
4271 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4272 A C expression that is nonzero if @var{regno} is the number of a hard
4273 register in which the values of called function may come back.
4275 A register whose use for returning values is limited to serving as the
4276 second of a pair (for a value of type @code{double}, say) need not be
4277 recognized by this macro. So for most machines, this definition
4281 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4284 If the machine has register windows, so that the caller and the called
4285 function use different registers for the return value, this macro
4286 should recognize only the caller's register numbers.
4289 @defmac TARGET_ENUM_VA_LIST (@var{idx}, @var{pname}, @var{ptype})
4290 This target macro is used in function @code{c_common_nodes_and_builtins}
4291 to iterate through the target specific builtin types for va_list. The
4292 variable @var{idx} is used as iterator. @var{pname} has to be a pointer
4293 to a @code{const char *} and @var{ptype} a pointer to a @code{tree} typed
4295 The arguments @var{pname} and @var{ptype} are used to store the result of
4296 this macro and are set to the name of the va_list builtin type and its
4298 If the return value of this macro is zero, then there is no more element.
4299 Otherwise the @var{IDX} should be increased for the next call of this
4300 macro to iterate through all types.
4303 @defmac APPLY_RESULT_SIZE
4304 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4305 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4306 saving and restoring an arbitrary return value.
4309 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (tree @var{type})
4310 This hook should return true if values of type @var{type} are returned
4311 at the most significant end of a register (in other words, if they are
4312 padded at the least significant end). You can assume that @var{type}
4313 is returned in a register; the caller is required to check this.
4315 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4316 be able to hold the complete return value. For example, if a 1-, 2-
4317 or 3-byte structure is returned at the most significant end of a
4318 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4322 @node Aggregate Return
4323 @subsection How Large Values Are Returned
4324 @cindex aggregates as return values
4325 @cindex large return values
4326 @cindex returning aggregate values
4327 @cindex structure value address
4329 When a function value's mode is @code{BLKmode} (and in some other
4330 cases), the value is not returned according to
4331 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4332 caller passes the address of a block of memory in which the value
4333 should be stored. This address is called the @dfn{structure value
4336 This section describes how to control returning structure values in
4339 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (tree @var{type}, tree @var{fntype})
4340 This target hook should return a nonzero value to say to return the
4341 function value in memory, just as large structures are always returned.
4342 Here @var{type} will be the data type of the value, and @var{fntype}
4343 will be the type of the function doing the returning, or @code{NULL} for
4346 Note that values of mode @code{BLKmode} must be explicitly handled
4347 by this function. Also, the option @option{-fpcc-struct-return}
4348 takes effect regardless of this macro. On most systems, it is
4349 possible to leave the hook undefined; this causes a default
4350 definition to be used, whose value is the constant 1 for @code{BLKmode}
4351 values, and 0 otherwise.
4353 Do not use this hook to indicate that structures and unions should always
4354 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4358 @defmac DEFAULT_PCC_STRUCT_RETURN
4359 Define this macro to be 1 if all structure and union return values must be
4360 in memory. Since this results in slower code, this should be defined
4361 only if needed for compatibility with other compilers or with an ABI@.
4362 If you define this macro to be 0, then the conventions used for structure
4363 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4366 If not defined, this defaults to the value 1.
4369 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4370 This target hook should return the location of the structure value
4371 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4372 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4373 be @code{NULL}, for libcalls. You do not need to define this target
4374 hook if the address is always passed as an ``invisible'' first
4377 On some architectures the place where the structure value address
4378 is found by the called function is not the same place that the
4379 caller put it. This can be due to register windows, or it could
4380 be because the function prologue moves it to a different place.
4381 @var{incoming} is @code{1} or @code{2} when the location is needed in
4382 the context of the called function, and @code{0} in the context of
4385 If @var{incoming} is nonzero and the address is to be found on the
4386 stack, return a @code{mem} which refers to the frame pointer. If
4387 @var{incoming} is @code{2}, the result is being used to fetch the
4388 structure value address at the beginning of a function. If you need
4389 to emit adjusting code, you should do it at this point.
4392 @defmac PCC_STATIC_STRUCT_RETURN
4393 Define this macro if the usual system convention on the target machine
4394 for returning structures and unions is for the called function to return
4395 the address of a static variable containing the value.
4397 Do not define this if the usual system convention is for the caller to
4398 pass an address to the subroutine.
4400 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4401 nothing when you use @option{-freg-struct-return} mode.
4405 @subsection Caller-Saves Register Allocation
4407 If you enable it, GCC can save registers around function calls. This
4408 makes it possible to use call-clobbered registers to hold variables that
4409 must live across calls.
4411 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4412 A C expression to determine whether it is worthwhile to consider placing
4413 a pseudo-register in a call-clobbered hard register and saving and
4414 restoring it around each function call. The expression should be 1 when
4415 this is worth doing, and 0 otherwise.
4417 If you don't define this macro, a default is used which is good on most
4418 machines: @code{4 * @var{calls} < @var{refs}}.
4421 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4422 A C expression specifying which mode is required for saving @var{nregs}
4423 of a pseudo-register in call-clobbered hard register @var{regno}. If
4424 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4425 returned. For most machines this macro need not be defined since GCC
4426 will select the smallest suitable mode.
4429 @node Function Entry
4430 @subsection Function Entry and Exit
4431 @cindex function entry and exit
4435 This section describes the macros that output function entry
4436 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4438 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4439 If defined, a function that outputs the assembler code for entry to a
4440 function. The prologue is responsible for setting up the stack frame,
4441 initializing the frame pointer register, saving registers that must be
4442 saved, and allocating @var{size} additional bytes of storage for the
4443 local variables. @var{size} is an integer. @var{file} is a stdio
4444 stream to which the assembler code should be output.
4446 The label for the beginning of the function need not be output by this
4447 macro. That has already been done when the macro is run.
4449 @findex regs_ever_live
4450 To determine which registers to save, the macro can refer to the array
4451 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4452 @var{r} is used anywhere within the function. This implies the function
4453 prologue should save register @var{r}, provided it is not one of the
4454 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4455 @code{regs_ever_live}.)
4457 On machines that have ``register windows'', the function entry code does
4458 not save on the stack the registers that are in the windows, even if
4459 they are supposed to be preserved by function calls; instead it takes
4460 appropriate steps to ``push'' the register stack, if any non-call-used
4461 registers are used in the function.
4463 @findex frame_pointer_needed
4464 On machines where functions may or may not have frame-pointers, the
4465 function entry code must vary accordingly; it must set up the frame
4466 pointer if one is wanted, and not otherwise. To determine whether a
4467 frame pointer is in wanted, the macro can refer to the variable
4468 @code{frame_pointer_needed}. The variable's value will be 1 at run
4469 time in a function that needs a frame pointer. @xref{Elimination}.
4471 The function entry code is responsible for allocating any stack space
4472 required for the function. This stack space consists of the regions
4473 listed below. In most cases, these regions are allocated in the
4474 order listed, with the last listed region closest to the top of the
4475 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4476 the highest address if it is not defined). You can use a different order
4477 for a machine if doing so is more convenient or required for
4478 compatibility reasons. Except in cases where required by standard
4479 or by a debugger, there is no reason why the stack layout used by GCC
4480 need agree with that used by other compilers for a machine.
4483 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4484 If defined, a function that outputs assembler code at the end of a
4485 prologue. This should be used when the function prologue is being
4486 emitted as RTL, and you have some extra assembler that needs to be
4487 emitted. @xref{prologue instruction pattern}.
4490 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4491 If defined, a function that outputs assembler code at the start of an
4492 epilogue. This should be used when the function epilogue is being
4493 emitted as RTL, and you have some extra assembler that needs to be
4494 emitted. @xref{epilogue instruction pattern}.
4497 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4498 If defined, a function that outputs the assembler code for exit from a
4499 function. The epilogue is responsible for restoring the saved
4500 registers and stack pointer to their values when the function was
4501 called, and returning control to the caller. This macro takes the
4502 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4503 registers to restore are determined from @code{regs_ever_live} and
4504 @code{CALL_USED_REGISTERS} in the same way.
4506 On some machines, there is a single instruction that does all the work
4507 of returning from the function. On these machines, give that
4508 instruction the name @samp{return} and do not define the macro
4509 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4511 Do not define a pattern named @samp{return} if you want the
4512 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4513 switches to control whether return instructions or epilogues are used,
4514 define a @samp{return} pattern with a validity condition that tests the
4515 target switches appropriately. If the @samp{return} pattern's validity
4516 condition is false, epilogues will be used.
4518 On machines where functions may or may not have frame-pointers, the
4519 function exit code must vary accordingly. Sometimes the code for these
4520 two cases is completely different. To determine whether a frame pointer
4521 is wanted, the macro can refer to the variable
4522 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4523 a function that needs a frame pointer.
4525 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4526 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4527 The C variable @code{current_function_is_leaf} is nonzero for such a
4528 function. @xref{Leaf Functions}.
4530 On some machines, some functions pop their arguments on exit while
4531 others leave that for the caller to do. For example, the 68020 when
4532 given @option{-mrtd} pops arguments in functions that take a fixed
4533 number of arguments.
4535 @findex current_function_pops_args
4536 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4537 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4538 needs to know what was decided. The variable that is called
4539 @code{current_function_pops_args} is the number of bytes of its
4540 arguments that a function should pop. @xref{Scalar Return}.
4541 @c what is the "its arguments" in the above sentence referring to, pray
4542 @c tell? --mew 5feb93
4547 @findex current_function_pretend_args_size
4548 A region of @code{current_function_pretend_args_size} bytes of
4549 uninitialized space just underneath the first argument arriving on the
4550 stack. (This may not be at the very start of the allocated stack region
4551 if the calling sequence has pushed anything else since pushing the stack
4552 arguments. But usually, on such machines, nothing else has been pushed
4553 yet, because the function prologue itself does all the pushing.) This
4554 region is used on machines where an argument may be passed partly in
4555 registers and partly in memory, and, in some cases to support the
4556 features in @code{<stdarg.h>}.
4559 An area of memory used to save certain registers used by the function.
4560 The size of this area, which may also include space for such things as
4561 the return address and pointers to previous stack frames, is
4562 machine-specific and usually depends on which registers have been used
4563 in the function. Machines with register windows often do not require
4567 A region of at least @var{size} bytes, possibly rounded up to an allocation
4568 boundary, to contain the local variables of the function. On some machines,
4569 this region and the save area may occur in the opposite order, with the
4570 save area closer to the top of the stack.
4573 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4574 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4575 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4576 argument lists of the function. @xref{Stack Arguments}.
4579 @defmac EXIT_IGNORE_STACK
4580 Define this macro as a C expression that is nonzero if the return
4581 instruction or the function epilogue ignores the value of the stack
4582 pointer; in other words, if it is safe to delete an instruction to
4583 adjust the stack pointer before a return from the function. The
4586 Note that this macro's value is relevant only for functions for which
4587 frame pointers are maintained. It is never safe to delete a final
4588 stack adjustment in a function that has no frame pointer, and the
4589 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4592 @defmac EPILOGUE_USES (@var{regno})
4593 Define this macro as a C expression that is nonzero for registers that are
4594 used by the epilogue or the @samp{return} pattern. The stack and frame
4595 pointer registers are already assumed to be used as needed.
4598 @defmac EH_USES (@var{regno})
4599 Define this macro as a C expression that is nonzero for registers that are
4600 used by the exception handling mechanism, and so should be considered live
4601 on entry to an exception edge.
4604 @defmac DELAY_SLOTS_FOR_EPILOGUE
4605 Define this macro if the function epilogue contains delay slots to which
4606 instructions from the rest of the function can be ``moved''. The
4607 definition should be a C expression whose value is an integer
4608 representing the number of delay slots there.
4611 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4612 A C expression that returns 1 if @var{insn} can be placed in delay
4613 slot number @var{n} of the epilogue.
4615 The argument @var{n} is an integer which identifies the delay slot now
4616 being considered (since different slots may have different rules of
4617 eligibility). It is never negative and is always less than the number
4618 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4619 If you reject a particular insn for a given delay slot, in principle, it
4620 may be reconsidered for a subsequent delay slot. Also, other insns may
4621 (at least in principle) be considered for the so far unfilled delay
4624 @findex current_function_epilogue_delay_list
4625 @findex final_scan_insn
4626 The insns accepted to fill the epilogue delay slots are put in an RTL
4627 list made with @code{insn_list} objects, stored in the variable
4628 @code{current_function_epilogue_delay_list}. The insn for the first
4629 delay slot comes first in the list. Your definition of the macro
4630 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4631 outputting the insns in this list, usually by calling
4632 @code{final_scan_insn}.
4634 You need not define this macro if you did not define
4635 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4638 @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})
4639 A function that outputs the assembler code for a thunk
4640 function, used to implement C++ virtual function calls with multiple
4641 inheritance. The thunk acts as a wrapper around a virtual function,
4642 adjusting the implicit object parameter before handing control off to
4645 First, emit code to add the integer @var{delta} to the location that
4646 contains the incoming first argument. Assume that this argument
4647 contains a pointer, and is the one used to pass the @code{this} pointer
4648 in C++. This is the incoming argument @emph{before} the function prologue,
4649 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4650 all other incoming arguments.
4652 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4653 made after adding @code{delta}. In particular, if @var{p} is the
4654 adjusted pointer, the following adjustment should be made:
4657 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4660 After the additions, emit code to jump to @var{function}, which is a
4661 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4662 not touch the return address. Hence returning from @var{FUNCTION} will
4663 return to whoever called the current @samp{thunk}.
4665 The effect must be as if @var{function} had been called directly with
4666 the adjusted first argument. This macro is responsible for emitting all
4667 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4668 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4670 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4671 have already been extracted from it.) It might possibly be useful on
4672 some targets, but probably not.
4674 If you do not define this macro, the target-independent code in the C++
4675 front end will generate a less efficient heavyweight thunk that calls
4676 @var{function} instead of jumping to it. The generic approach does
4677 not support varargs.
4680 @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})
4681 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4682 to output the assembler code for the thunk function specified by the
4683 arguments it is passed, and false otherwise. In the latter case, the
4684 generic approach will be used by the C++ front end, with the limitations
4689 @subsection Generating Code for Profiling
4690 @cindex profiling, code generation
4692 These macros will help you generate code for profiling.
4694 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4695 A C statement or compound statement to output to @var{file} some
4696 assembler code to call the profiling subroutine @code{mcount}.
4699 The details of how @code{mcount} expects to be called are determined by
4700 your operating system environment, not by GCC@. To figure them out,
4701 compile a small program for profiling using the system's installed C
4702 compiler and look at the assembler code that results.
4704 Older implementations of @code{mcount} expect the address of a counter
4705 variable to be loaded into some register. The name of this variable is
4706 @samp{LP} followed by the number @var{labelno}, so you would generate
4707 the name using @samp{LP%d} in a @code{fprintf}.
4710 @defmac PROFILE_HOOK
4711 A C statement or compound statement to output to @var{file} some assembly
4712 code to call the profiling subroutine @code{mcount} even the target does
4713 not support profiling.
4716 @defmac NO_PROFILE_COUNTERS
4717 Define this macro to be an expression with a nonzero value if the
4718 @code{mcount} subroutine on your system does not need a counter variable
4719 allocated for each function. This is true for almost all modern
4720 implementations. If you define this macro, you must not use the
4721 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4724 @defmac PROFILE_BEFORE_PROLOGUE
4725 Define this macro if the code for function profiling should come before
4726 the function prologue. Normally, the profiling code comes after.
4730 @subsection Permitting tail calls
4733 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4734 True if it is ok to do sibling call optimization for the specified
4735 call expression @var{exp}. @var{decl} will be the called function,
4736 or @code{NULL} if this is an indirect call.
4738 It is not uncommon for limitations of calling conventions to prevent
4739 tail calls to functions outside the current unit of translation, or
4740 during PIC compilation. The hook is used to enforce these restrictions,
4741 as the @code{sibcall} md pattern can not fail, or fall over to a
4742 ``normal'' call. The criteria for successful sibling call optimization
4743 may vary greatly between different architectures.
4746 @deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap *@var{regs})
4747 Add any hard registers to @var{regs} that are live on entry to the
4748 function. This hook only needs to be defined to provide registers that
4749 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4750 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4751 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4752 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4755 @node Stack Smashing Protection
4756 @subsection Stack smashing protection
4757 @cindex stack smashing protection
4759 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void)
4760 This hook returns a @code{DECL} node for the external variable to use
4761 for the stack protection guard. This variable is initialized by the
4762 runtime to some random value and is used to initialize the guard value
4763 that is placed at the top of the local stack frame. The type of this
4764 variable must be @code{ptr_type_node}.
4766 The default version of this hook creates a variable called
4767 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
4770 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void)
4771 This hook returns a tree expression that alerts the runtime that the
4772 stack protect guard variable has been modified. This expression should
4773 involve a call to a @code{noreturn} function.
4775 The default version of this hook invokes a function called
4776 @samp{__stack_chk_fail}, taking no arguments. This function is
4777 normally defined in @file{libgcc2.c}.
4781 @section Implementing the Varargs Macros
4782 @cindex varargs implementation
4784 GCC comes with an implementation of @code{<varargs.h>} and
4785 @code{<stdarg.h>} that work without change on machines that pass arguments
4786 on the stack. Other machines require their own implementations of
4787 varargs, and the two machine independent header files must have
4788 conditionals to include it.
4790 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4791 the calling convention for @code{va_start}. The traditional
4792 implementation takes just one argument, which is the variable in which
4793 to store the argument pointer. The ISO implementation of
4794 @code{va_start} takes an additional second argument. The user is
4795 supposed to write the last named argument of the function here.
4797 However, @code{va_start} should not use this argument. The way to find
4798 the end of the named arguments is with the built-in functions described
4801 @defmac __builtin_saveregs ()
4802 Use this built-in function to save the argument registers in memory so
4803 that the varargs mechanism can access them. Both ISO and traditional
4804 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4805 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4807 On some machines, @code{__builtin_saveregs} is open-coded under the
4808 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4809 other machines, it calls a routine written in assembler language,
4810 found in @file{libgcc2.c}.
4812 Code generated for the call to @code{__builtin_saveregs} appears at the
4813 beginning of the function, as opposed to where the call to
4814 @code{__builtin_saveregs} is written, regardless of what the code is.
4815 This is because the registers must be saved before the function starts
4816 to use them for its own purposes.
4817 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4821 @defmac __builtin_args_info (@var{category})
4822 Use this built-in function to find the first anonymous arguments in
4825 In general, a machine may have several categories of registers used for
4826 arguments, each for a particular category of data types. (For example,
4827 on some machines, floating-point registers are used for floating-point
4828 arguments while other arguments are passed in the general registers.)
4829 To make non-varargs functions use the proper calling convention, you
4830 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4831 registers in each category have been used so far
4833 @code{__builtin_args_info} accesses the same data structure of type
4834 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4835 with it, with @var{category} specifying which word to access. Thus, the
4836 value indicates the first unused register in a given category.
4838 Normally, you would use @code{__builtin_args_info} in the implementation
4839 of @code{va_start}, accessing each category just once and storing the
4840 value in the @code{va_list} object. This is because @code{va_list} will
4841 have to update the values, and there is no way to alter the
4842 values accessed by @code{__builtin_args_info}.
4845 @defmac __builtin_next_arg (@var{lastarg})
4846 This is the equivalent of @code{__builtin_args_info}, for stack
4847 arguments. It returns the address of the first anonymous stack
4848 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4849 returns the address of the location above the first anonymous stack
4850 argument. Use it in @code{va_start} to initialize the pointer for
4851 fetching arguments from the stack. Also use it in @code{va_start} to
4852 verify that the second parameter @var{lastarg} is the last named argument
4853 of the current function.
4856 @defmac __builtin_classify_type (@var{object})
4857 Since each machine has its own conventions for which data types are
4858 passed in which kind of register, your implementation of @code{va_arg}
4859 has to embody these conventions. The easiest way to categorize the
4860 specified data type is to use @code{__builtin_classify_type} together
4861 with @code{sizeof} and @code{__alignof__}.
4863 @code{__builtin_classify_type} ignores the value of @var{object},
4864 considering only its data type. It returns an integer describing what
4865 kind of type that is---integer, floating, pointer, structure, and so on.
4867 The file @file{typeclass.h} defines an enumeration that you can use to
4868 interpret the values of @code{__builtin_classify_type}.
4871 These machine description macros help implement varargs:
4873 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
4874 If defined, this hook produces the machine-specific code for a call to
4875 @code{__builtin_saveregs}. This code will be moved to the very
4876 beginning of the function, before any parameter access are made. The
4877 return value of this function should be an RTX that contains the value
4878 to use as the return of @code{__builtin_saveregs}.
4881 @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})
4882 This target hook offers an alternative to using
4883 @code{__builtin_saveregs} and defining the hook
4884 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
4885 register arguments into the stack so that all the arguments appear to
4886 have been passed consecutively on the stack. Once this is done, you can
4887 use the standard implementation of varargs that works for machines that
4888 pass all their arguments on the stack.
4890 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
4891 structure, containing the values that are obtained after processing the
4892 named arguments. The arguments @var{mode} and @var{type} describe the
4893 last named argument---its machine mode and its data type as a tree node.
4895 The target hook should do two things: first, push onto the stack all the
4896 argument registers @emph{not} used for the named arguments, and second,
4897 store the size of the data thus pushed into the @code{int}-valued
4898 variable pointed to by @var{pretend_args_size}. The value that you
4899 store here will serve as additional offset for setting up the stack
4902 Because you must generate code to push the anonymous arguments at
4903 compile time without knowing their data types,
4904 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
4905 have just a single category of argument register and use it uniformly
4908 If the argument @var{second_time} is nonzero, it means that the
4909 arguments of the function are being analyzed for the second time. This
4910 happens for an inline function, which is not actually compiled until the
4911 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
4912 not generate any instructions in this case.
4915 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca})
4916 Define this hook to return @code{true} if the location where a function
4917 argument is passed depends on whether or not it is a named argument.
4919 This hook controls how the @var{named} argument to @code{FUNCTION_ARG}
4920 is set for varargs and stdarg functions. If this hook returns
4921 @code{true}, the @var{named} argument is always true for named
4922 arguments, and false for unnamed arguments. If it returns @code{false},
4923 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
4924 then all arguments are treated as named. Otherwise, all named arguments
4925 except the last are treated as named.
4927 You need not define this hook if it always returns zero.
4930 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED
4931 If you need to conditionally change ABIs so that one works with
4932 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
4933 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
4934 defined, then define this hook to return @code{true} if
4935 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
4936 Otherwise, you should not define this hook.
4940 @section Trampolines for Nested Functions
4941 @cindex trampolines for nested functions
4942 @cindex nested functions, trampolines for
4944 A @dfn{trampoline} is a small piece of code that is created at run time
4945 when the address of a nested function is taken. It normally resides on
4946 the stack, in the stack frame of the containing function. These macros
4947 tell GCC how to generate code to allocate and initialize a
4950 The instructions in the trampoline must do two things: load a constant
4951 address into the static chain register, and jump to the real address of
4952 the nested function. On CISC machines such as the m68k, this requires
4953 two instructions, a move immediate and a jump. Then the two addresses
4954 exist in the trampoline as word-long immediate operands. On RISC
4955 machines, it is often necessary to load each address into a register in
4956 two parts. Then pieces of each address form separate immediate
4959 The code generated to initialize the trampoline must store the variable
4960 parts---the static chain value and the function address---into the
4961 immediate operands of the instructions. On a CISC machine, this is
4962 simply a matter of copying each address to a memory reference at the
4963 proper offset from the start of the trampoline. On a RISC machine, it
4964 may be necessary to take out pieces of the address and store them
4967 @defmac TRAMPOLINE_TEMPLATE (@var{file})
4968 A C statement to output, on the stream @var{file}, assembler code for a
4969 block of data that contains the constant parts of a trampoline. This
4970 code should not include a label---the label is taken care of
4973 If you do not define this macro, it means no template is needed
4974 for the target. Do not define this macro on systems where the block move
4975 code to copy the trampoline into place would be larger than the code
4976 to generate it on the spot.
4979 @defmac TRAMPOLINE_SECTION
4980 Return the section into which the trampoline template is to be placed
4981 (@pxref{Sections}). The default value is @code{readonly_data_section}.
4984 @defmac TRAMPOLINE_SIZE
4985 A C expression for the size in bytes of the trampoline, as an integer.
4988 @defmac TRAMPOLINE_ALIGNMENT
4989 Alignment required for trampolines, in bits.
4991 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4992 is used for aligning trampolines.
4995 @defmac INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4996 A C statement to initialize the variable parts of a trampoline.
4997 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4998 an RTX for the address of the nested function; @var{static_chain} is an
4999 RTX for the static chain value that should be passed to the function
5003 @defmac TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
5004 A C statement that should perform any machine-specific adjustment in
5005 the address of the trampoline. Its argument contains the address that
5006 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
5007 used for a function call should be different from the address in which
5008 the template was stored, the different address should be assigned to
5009 @var{addr}. If this macro is not defined, @var{addr} will be used for
5012 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
5013 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
5014 If this macro is not defined, by default the trampoline is allocated as
5015 a stack slot. This default is right for most machines. The exceptions
5016 are machines where it is impossible to execute instructions in the stack
5017 area. On such machines, you may have to implement a separate stack,
5018 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
5019 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
5021 @var{fp} points to a data structure, a @code{struct function}, which
5022 describes the compilation status of the immediate containing function of
5023 the function which the trampoline is for. The stack slot for the
5024 trampoline is in the stack frame of this containing function. Other
5025 allocation strategies probably must do something analogous with this
5029 Implementing trampolines is difficult on many machines because they have
5030 separate instruction and data caches. Writing into a stack location
5031 fails to clear the memory in the instruction cache, so when the program
5032 jumps to that location, it executes the old contents.
5034 Here are two possible solutions. One is to clear the relevant parts of
5035 the instruction cache whenever a trampoline is set up. The other is to
5036 make all trampolines identical, by having them jump to a standard
5037 subroutine. The former technique makes trampoline execution faster; the
5038 latter makes initialization faster.
5040 To clear the instruction cache when a trampoline is initialized, define
5041 the following macro.
5043 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
5044 If defined, expands to a C expression clearing the @emph{instruction
5045 cache} in the specified interval. The definition of this macro would
5046 typically be a series of @code{asm} statements. Both @var{beg} and
5047 @var{end} are both pointer expressions.
5050 The operating system may also require the stack to be made executable
5051 before calling the trampoline. To implement this requirement, define
5052 the following macro.
5054 @defmac ENABLE_EXECUTE_STACK
5055 Define this macro if certain operations must be performed before executing
5056 code located on the stack. The macro should expand to a series of C
5057 file-scope constructs (e.g.@: functions) and provide a unique entry point
5058 named @code{__enable_execute_stack}. The target is responsible for
5059 emitting calls to the entry point in the code, for example from the
5060 @code{INITIALIZE_TRAMPOLINE} macro.
5063 To use a standard subroutine, define the following macro. In addition,
5064 you must make sure that the instructions in a trampoline fill an entire
5065 cache line with identical instructions, or else ensure that the
5066 beginning of the trampoline code is always aligned at the same point in
5067 its cache line. Look in @file{m68k.h} as a guide.
5069 @defmac TRANSFER_FROM_TRAMPOLINE
5070 Define this macro if trampolines need a special subroutine to do their
5071 work. The macro should expand to a series of @code{asm} statements
5072 which will be compiled with GCC@. They go in a library function named
5073 @code{__transfer_from_trampoline}.
5075 If you need to avoid executing the ordinary prologue code of a compiled
5076 C function when you jump to the subroutine, you can do so by placing a
5077 special label of your own in the assembler code. Use one @code{asm}
5078 statement to generate an assembler label, and another to make the label
5079 global. Then trampolines can use that label to jump directly to your
5080 special assembler code.
5084 @section Implicit Calls to Library Routines
5085 @cindex library subroutine names
5086 @cindex @file{libgcc.a}
5088 @c prevent bad page break with this line
5089 Here is an explanation of implicit calls to library routines.
5091 @defmac DECLARE_LIBRARY_RENAMES
5092 This macro, if defined, should expand to a piece of C code that will get
5093 expanded when compiling functions for libgcc.a. It can be used to
5094 provide alternate names for GCC's internal library functions if there
5095 are ABI-mandated names that the compiler should provide.
5098 @findex init_one_libfunc
5099 @findex set_optab_libfunc
5100 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
5101 This hook should declare additional library routines or rename
5102 existing ones, using the functions @code{set_optab_libfunc} and
5103 @code{init_one_libfunc} defined in @file{optabs.c}.
5104 @code{init_optabs} calls this macro after initializing all the normal
5107 The default is to do nothing. Most ports don't need to define this hook.
5110 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
5111 This macro should return @code{true} if the library routine that
5112 implements the floating point comparison operator @var{comparison} in
5113 mode @var{mode} will return a boolean, and @var{false} if it will
5116 GCC's own floating point libraries return tristates from the
5117 comparison operators, so the default returns false always. Most ports
5118 don't need to define this macro.
5121 @defmac TARGET_LIB_INT_CMP_BIASED
5122 This macro should evaluate to @code{true} if the integer comparison
5123 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5124 operand is smaller than the second, 1 to indicate that they are equal,
5125 and 2 to indicate that the first operand is greater than the second.
5126 If this macro evaluates to @code{false} the comparison functions return
5127 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5128 in @file{libgcc.a}, you do not need to define this macro.
5131 @cindex US Software GOFAST, floating point emulation library
5132 @cindex floating point emulation library, US Software GOFAST
5133 @cindex GOFAST, floating point emulation library
5134 @findex gofast_maybe_init_libfuncs
5135 @defmac US_SOFTWARE_GOFAST
5136 Define this macro if your system C library uses the US Software GOFAST
5137 library to provide floating point emulation.
5139 In addition to defining this macro, your architecture must set
5140 @code{TARGET_INIT_LIBFUNCS} to @code{gofast_maybe_init_libfuncs}, or
5141 else call that function from its version of that hook. It is defined
5142 in @file{config/gofast.h}, which must be included by your
5143 architecture's @file{@var{cpu}.c} file. See @file{sparc/sparc.c} for
5146 If this macro is defined, the
5147 @code{TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL} target hook must return
5148 false for @code{SFmode} and @code{DFmode} comparisons.
5151 @cindex @code{EDOM}, implicit usage
5154 The value of @code{EDOM} on the target machine, as a C integer constant
5155 expression. If you don't define this macro, GCC does not attempt to
5156 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5157 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5160 If you do not define @code{TARGET_EDOM}, then compiled code reports
5161 domain errors by calling the library function and letting it report the
5162 error. If mathematical functions on your system use @code{matherr} when
5163 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5164 that @code{matherr} is used normally.
5167 @cindex @code{errno}, implicit usage
5168 @defmac GEN_ERRNO_RTX
5169 Define this macro as a C expression to create an rtl expression that
5170 refers to the global ``variable'' @code{errno}. (On certain systems,
5171 @code{errno} may not actually be a variable.) If you don't define this
5172 macro, a reasonable default is used.
5175 @cindex C99 math functions, implicit usage
5176 @defmac TARGET_C99_FUNCTIONS
5177 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
5178 @code{sinf} and similarly for other functions defined by C99 standard. The
5179 default is nonzero that should be proper value for most modern systems, however
5180 number of existing systems lacks support for these functions in the runtime so
5181 they needs this macro to be redefined to 0.
5184 @cindex sincos math function, implicit usage
5185 @defmac TARGET_HAS_SINCOS
5186 When this macro is nonzero, GCC will implicitly optimize calls to @code{sin}
5187 and @code{cos} with the same argument to a call to @code{sincos}. The
5188 default is zero. The target has to provide the following functions:
5190 void sincos(double x, double *sin, double *cos);
5191 void sincosf(float x, float *sin, float *cos);
5192 void sincosl(long double x, long double *sin, long double *cos);
5196 @defmac NEXT_OBJC_RUNTIME
5197 Define this macro to generate code for Objective-C message sending using
5198 the calling convention of the NeXT system. This calling convention
5199 involves passing the object, the selector and the method arguments all
5200 at once to the method-lookup library function.
5202 The default calling convention passes just the object and the selector
5203 to the lookup function, which returns a pointer to the method.
5206 @node Addressing Modes
5207 @section Addressing Modes
5208 @cindex addressing modes
5210 @c prevent bad page break with this line
5211 This is about addressing modes.
5213 @defmac HAVE_PRE_INCREMENT
5214 @defmacx HAVE_PRE_DECREMENT
5215 @defmacx HAVE_POST_INCREMENT
5216 @defmacx HAVE_POST_DECREMENT
5217 A C expression that is nonzero if the machine supports pre-increment,
5218 pre-decrement, post-increment, or post-decrement addressing respectively.
5221 @defmac HAVE_PRE_MODIFY_DISP
5222 @defmacx HAVE_POST_MODIFY_DISP
5223 A C expression that is nonzero if the machine supports pre- or
5224 post-address side-effect generation involving constants other than
5225 the size of the memory operand.
5228 @defmac HAVE_PRE_MODIFY_REG
5229 @defmacx HAVE_POST_MODIFY_REG
5230 A C expression that is nonzero if the machine supports pre- or
5231 post-address side-effect generation involving a register displacement.
5234 @defmac CONSTANT_ADDRESS_P (@var{x})
5235 A C expression that is 1 if the RTX @var{x} is a constant which
5236 is a valid address. On most machines, this can be defined as
5237 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
5238 in which constant addresses are supported.
5241 @defmac CONSTANT_P (@var{x})
5242 @code{CONSTANT_P}, which is defined by target-independent code,
5243 accepts integer-values expressions whose values are not explicitly
5244 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5245 expressions and @code{const} arithmetic expressions, in addition to
5246 @code{const_int} and @code{const_double} expressions.
5249 @defmac MAX_REGS_PER_ADDRESS
5250 A number, the maximum number of registers that can appear in a valid
5251 memory address. Note that it is up to you to specify a value equal to
5252 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
5256 @defmac GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5257 A C compound statement with a conditional @code{goto @var{label};}
5258 executed if @var{x} (an RTX) is a legitimate memory address on the
5259 target machine for a memory operand of mode @var{mode}.
5261 It usually pays to define several simpler macros to serve as
5262 subroutines for this one. Otherwise it may be too complicated to
5265 This macro must exist in two variants: a strict variant and a
5266 non-strict one. The strict variant is used in the reload pass. It
5267 must be defined so that any pseudo-register that has not been
5268 allocated a hard register is considered a memory reference. In
5269 contexts where some kind of register is required, a pseudo-register
5270 with no hard register must be rejected.
5272 The non-strict variant is used in other passes. It must be defined to
5273 accept all pseudo-registers in every context where some kind of
5274 register is required.
5276 @findex REG_OK_STRICT
5277 Compiler source files that want to use the strict variant of this
5278 macro define the macro @code{REG_OK_STRICT}. You should use an
5279 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
5280 in that case and the non-strict variant otherwise.
5282 Subroutines to check for acceptable registers for various purposes (one
5283 for base registers, one for index registers, and so on) are typically
5284 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
5285 Then only these subroutine macros need have two variants; the higher
5286 levels of macros may be the same whether strict or not.
5288 Normally, constant addresses which are the sum of a @code{symbol_ref}
5289 and an integer are stored inside a @code{const} RTX to mark them as
5290 constant. Therefore, there is no need to recognize such sums
5291 specifically as legitimate addresses. Normally you would simply
5292 recognize any @code{const} as legitimate.
5294 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5295 sums that are not marked with @code{const}. It assumes that a naked
5296 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5297 naked constant sums as illegitimate addresses, so that none of them will
5298 be given to @code{PRINT_OPERAND_ADDRESS}.
5300 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5301 On some machines, whether a symbolic address is legitimate depends on
5302 the section that the address refers to. On these machines, define the
5303 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5304 into the @code{symbol_ref}, and then check for it here. When you see a
5305 @code{const}, you will have to look inside it to find the
5306 @code{symbol_ref} in order to determine the section. @xref{Assembler
5310 @defmac TARGET_MEM_CONSTRAINT
5311 A single character to be used instead of the default @code{'m'}
5312 character for general memory addresses. This defines the constraint
5313 letter which matches the memory addresses accepted by
5314 @code{GO_IF_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
5315 support new address formats in your back end without changing the
5316 semantics of the @code{'m'} constraint. This is necessary in order to
5317 preserve functionality of inline assembly constructs using the
5318 @code{'m'} constraint.
5321 @defmac FIND_BASE_TERM (@var{x})
5322 A C expression to determine the base term of address @var{x}.
5323 This macro is used in only one place: `find_base_term' in alias.c.
5325 It is always safe for this macro to not be defined. It exists so
5326 that alias analysis can understand machine-dependent addresses.
5328 The typical use of this macro is to handle addresses containing
5329 a label_ref or symbol_ref within an UNSPEC@.
5332 @defmac LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
5333 A C compound statement that attempts to replace @var{x} with a valid
5334 memory address for an operand of mode @var{mode}. @var{win} will be a
5335 C statement label elsewhere in the code; the macro definition may use
5338 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
5342 to avoid further processing if the address has become legitimate.
5344 @findex break_out_memory_refs
5345 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5346 and @var{oldx} will be the operand that was given to that function to produce
5349 The code generated by this macro should not alter the substructure of
5350 @var{x}. If it transforms @var{x} into a more legitimate form, it
5351 should assign @var{x} (which will always be a C variable) a new value.
5353 It is not necessary for this macro to come up with a legitimate
5354 address. The compiler has standard ways of doing so in all cases. In
5355 fact, it is safe to omit this macro. But often a
5356 machine-dependent strategy can generate better code.
5359 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5360 A C compound statement that attempts to replace @var{x}, which is an address
5361 that needs reloading, with a valid memory address for an operand of mode
5362 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5363 It is not necessary to define this macro, but it might be useful for
5364 performance reasons.
5366 For example, on the i386, it is sometimes possible to use a single
5367 reload register instead of two by reloading a sum of two pseudo
5368 registers into a register. On the other hand, for number of RISC
5369 processors offsets are limited so that often an intermediate address
5370 needs to be generated in order to address a stack slot. By defining
5371 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5372 generated for adjacent some stack slots can be made identical, and thus
5375 @emph{Note}: This macro should be used with caution. It is necessary
5376 to know something of how reload works in order to effectively use this,
5377 and it is quite easy to produce macros that build in too much knowledge
5378 of reload internals.
5380 @emph{Note}: This macro must be able to reload an address created by a
5381 previous invocation of this macro. If it fails to handle such addresses
5382 then the compiler may generate incorrect code or abort.
5385 The macro definition should use @code{push_reload} to indicate parts that
5386 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5387 suitable to be passed unaltered to @code{push_reload}.
5389 The code generated by this macro must not alter the substructure of
5390 @var{x}. If it transforms @var{x} into a more legitimate form, it
5391 should assign @var{x} (which will always be a C variable) a new value.
5392 This also applies to parts that you change indirectly by calling
5395 @findex strict_memory_address_p
5396 The macro definition may use @code{strict_memory_address_p} to test if
5397 the address has become legitimate.
5400 If you want to change only a part of @var{x}, one standard way of doing
5401 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5402 single level of rtl. Thus, if the part to be changed is not at the
5403 top level, you'll need to replace first the top level.
5404 It is not necessary for this macro to come up with a legitimate
5405 address; but often a machine-dependent strategy can generate better code.
5408 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5409 A C statement or compound statement with a conditional @code{goto
5410 @var{label};} executed if memory address @var{x} (an RTX) can have
5411 different meanings depending on the machine mode of the memory
5412 reference it is used for or if the address is valid for some modes
5415 Autoincrement and autodecrement addresses typically have mode-dependent
5416 effects because the amount of the increment or decrement is the size
5417 of the operand being addressed. Some machines have other mode-dependent
5418 addresses. Many RISC machines have no mode-dependent addresses.
5420 You may assume that @var{addr} is a valid address for the machine.
5423 @defmac LEGITIMATE_CONSTANT_P (@var{x})
5424 A C expression that is nonzero if @var{x} is a legitimate constant for
5425 an immediate operand on the target machine. You can assume that
5426 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5427 @samp{1} is a suitable definition for this macro on machines where
5428 anything @code{CONSTANT_P} is valid.
5431 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5432 This hook is used to undo the possibly obfuscating effects of the
5433 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5434 macros. Some backend implementations of these macros wrap symbol
5435 references inside an @code{UNSPEC} rtx to represent PIC or similar
5436 addressing modes. This target hook allows GCC's optimizers to understand
5437 the semantics of these opaque @code{UNSPEC}s by converting them back
5438 into their original form.
5441 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (rtx @var{x})
5442 This hook should return true if @var{x} is of a form that cannot (or
5443 should not) be spilled to the constant pool. The default version of
5444 this hook returns false.
5446 The primary reason to define this hook is to prevent reload from
5447 deciding that a non-legitimate constant would be better reloaded
5448 from the constant pool instead of spilling and reloading a register
5449 holding the constant. This restriction is often true of addresses
5450 of TLS symbols for various targets.
5453 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (enum machine_mode @var{mode}, rtx @var{x})
5454 This hook should return true if pool entries for constant @var{x} can
5455 be placed in an @code{object_block} structure. @var{mode} is the mode
5458 The default version returns false for all constants.
5461 @deftypefn {Target Hook} tree TARGET_BUILTIN_RECIPROCAL (enum tree_code @var{fn}, bool @var{tm_fn}, bool @var{sqrt})
5462 This hook should return the DECL of a function that implements reciprocal of
5463 the builtin function with builtin function code @var{fn}, or
5464 @code{NULL_TREE} if such a function is not available. @var{tm_fn} is true
5465 when @var{fn} is a code of a machine-dependent builtin function. When
5466 @var{sqrt} is true, additional optimizations that apply only to the reciprocal
5467 of a square root function are performed, and only reciprocals of @code{sqrt}
5471 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5472 This hook should return the DECL of a function @var{f} that given an
5473 address @var{addr} as an argument returns a mask @var{m} that can be
5474 used to extract from two vectors the relevant data that resides in
5475 @var{addr} in case @var{addr} is not properly aligned.
5477 The autovectorizer, when vectorizing a load operation from an address
5478 @var{addr} that may be unaligned, will generate two vector loads from
5479 the two aligned addresses around @var{addr}. It then generates a
5480 @code{REALIGN_LOAD} operation to extract the relevant data from the
5481 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5482 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5483 the third argument, @var{OFF}, defines how the data will be extracted
5484 from these two vectors: if @var{OFF} is 0, then the returned vector is
5485 @var{v2}; otherwise, the returned vector is composed from the last
5486 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5487 @var{OFF} elements of @var{v2}.
5489 If this hook is defined, the autovectorizer will generate a call
5490 to @var{f} (using the DECL tree that this hook returns) and will
5491 use the return value of @var{f} as the argument @var{OFF} to
5492 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5493 should comply with the semantics expected by @code{REALIGN_LOAD}
5495 If this hook is not defined, then @var{addr} will be used as
5496 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5497 log2(@var{VS})-1 bits of @var{addr} will be considered.
5500 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN (tree @var{x})
5501 This hook should return the DECL of a function @var{f} that implements
5502 widening multiplication of the even elements of two input vectors of type @var{x}.
5504 If this hook is defined, the autovectorizer will use it along with the
5505 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD} target hook when vectorizing
5506 widening multiplication in cases that the order of the results does not have to be
5507 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5508 @code{widen_mult_hi/lo} idioms will be used.
5511 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD (tree @var{x})
5512 This hook should return the DECL of a function @var{f} that implements
5513 widening multiplication of the odd elements of two input vectors of type @var{x}.
5515 If this hook is defined, the autovectorizer will use it along with the
5516 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN} target hook when vectorizing
5517 widening multiplication in cases that the order of the results does not have to be
5518 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5519 @code{widen_mult_hi/lo} idioms will be used.
5522 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_CONVERSION (enum tree_code @var{code}, tree @var{type})
5523 This hook should return the DECL of a function that implements conversion of the
5524 input vector of type @var{type}.
5525 If @var{type} is an integral type, the result of the conversion is a vector of
5526 floating-point type of the same size.
5527 If @var{type} is a floating-point type, the result of the conversion is a vector
5528 of integral type of the same size.
5529 @var{code} specifies how the conversion is to be applied
5530 (truncation, rounding, etc.).
5532 If this hook is defined, the autovectorizer will use the
5533 @code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing
5534 conversion. Otherwise, it will return @code{NULL_TREE}.
5537 @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})
5538 This hook should return the decl of a function that implements the vectorized
5539 variant of the builtin function with builtin function code @var{code} or
5540 @code{NULL_TREE} if such a function is not available. The return type of
5541 the vectorized function shall be of vector type @var{vec_type_out} and the
5542 argument types should be @var{vec_type_in}.
5545 @node Anchored Addresses
5546 @section Anchored Addresses
5547 @cindex anchored addresses
5548 @cindex @option{-fsection-anchors}
5550 GCC usually addresses every static object as a separate entity.
5551 For example, if we have:
5555 int foo (void) @{ return a + b + c; @}
5558 the code for @code{foo} will usually calculate three separate symbolic
5559 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
5560 it would be better to calculate just one symbolic address and access
5561 the three variables relative to it. The equivalent pseudocode would
5567 register int *xr = &x;
5568 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5572 (which isn't valid C). We refer to shared addresses like @code{x} as
5573 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
5575 The hooks below describe the target properties that GCC needs to know
5576 in order to make effective use of section anchors. It won't use
5577 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
5578 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
5580 @deftypevar {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET
5581 The minimum offset that should be applied to a section anchor.
5582 On most targets, it should be the smallest offset that can be
5583 applied to a base register while still giving a legitimate address
5584 for every mode. The default value is 0.
5587 @deftypevar {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET
5588 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
5589 offset that should be applied to section anchors. The default
5593 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x})
5594 Write the assembly code to define section anchor @var{x}, which is a
5595 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
5596 The hook is called with the assembly output position set to the beginning
5597 of @code{SYMBOL_REF_BLOCK (@var{x})}.
5599 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
5600 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
5601 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
5602 is @code{NULL}, which disables the use of section anchors altogether.
5605 @deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (rtx @var{x})
5606 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
5607 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
5608 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
5610 The default version is correct for most targets, but you might need to
5611 intercept this hook to handle things like target-specific attributes
5612 or target-specific sections.
5615 @node Condition Code
5616 @section Condition Code Status
5617 @cindex condition code status
5619 @c prevent bad page break with this line
5620 This describes the condition code status.
5623 The file @file{conditions.h} defines a variable @code{cc_status} to
5624 describe how the condition code was computed (in case the interpretation of
5625 the condition code depends on the instruction that it was set by). This
5626 variable contains the RTL expressions on which the condition code is
5627 currently based, and several standard flags.
5629 Sometimes additional machine-specific flags must be defined in the machine
5630 description header file. It can also add additional machine-specific
5631 information by defining @code{CC_STATUS_MDEP}.
5633 @defmac CC_STATUS_MDEP
5634 C code for a data type which is used for declaring the @code{mdep}
5635 component of @code{cc_status}. It defaults to @code{int}.
5637 This macro is not used on machines that do not use @code{cc0}.
5640 @defmac CC_STATUS_MDEP_INIT
5641 A C expression to initialize the @code{mdep} field to ``empty''.
5642 The default definition does nothing, since most machines don't use
5643 the field anyway. If you want to use the field, you should probably
5644 define this macro to initialize it.
5646 This macro is not used on machines that do not use @code{cc0}.
5649 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5650 A C compound statement to set the components of @code{cc_status}
5651 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5652 this macro's responsibility to recognize insns that set the condition
5653 code as a byproduct of other activity as well as those that explicitly
5656 This macro is not used on machines that do not use @code{cc0}.
5658 If there are insns that do not set the condition code but do alter
5659 other machine registers, this macro must check to see whether they
5660 invalidate the expressions that the condition code is recorded as
5661 reflecting. For example, on the 68000, insns that store in address
5662 registers do not set the condition code, which means that usually
5663 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5664 insns. But suppose that the previous insn set the condition code
5665 based on location @samp{a4@@(102)} and the current insn stores a new
5666 value in @samp{a4}. Although the condition code is not changed by
5667 this, it will no longer be true that it reflects the contents of
5668 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5669 @code{cc_status} in this case to say that nothing is known about the
5670 condition code value.
5672 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5673 with the results of peephole optimization: insns whose patterns are
5674 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5675 constants which are just the operands. The RTL structure of these
5676 insns is not sufficient to indicate what the insns actually do. What
5677 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5678 @code{CC_STATUS_INIT}.
5680 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5681 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5682 @samp{cc}. This avoids having detailed information about patterns in
5683 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5686 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5687 Returns a mode from class @code{MODE_CC} to be used when comparison
5688 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5689 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5690 @pxref{Jump Patterns} for a description of the reason for this
5694 #define SELECT_CC_MODE(OP,X,Y) \
5695 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5696 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5697 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5698 || GET_CODE (X) == NEG) \
5699 ? CC_NOOVmode : CCmode))
5702 You should define this macro if and only if you define extra CC modes
5703 in @file{@var{machine}-modes.def}.
5706 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5707 On some machines not all possible comparisons are defined, but you can
5708 convert an invalid comparison into a valid one. For example, the Alpha
5709 does not have a @code{GT} comparison, but you can use an @code{LT}
5710 comparison instead and swap the order of the operands.
5712 On such machines, define this macro to be a C statement to do any
5713 required conversions. @var{code} is the initial comparison code
5714 and @var{op0} and @var{op1} are the left and right operands of the
5715 comparison, respectively. You should modify @var{code}, @var{op0}, and
5716 @var{op1} as required.
5718 GCC will not assume that the comparison resulting from this macro is
5719 valid but will see if the resulting insn matches a pattern in the
5722 You need not define this macro if it would never change the comparison
5726 @defmac REVERSIBLE_CC_MODE (@var{mode})
5727 A C expression whose value is one if it is always safe to reverse a
5728 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5729 can ever return @var{mode} for a floating-point inequality comparison,
5730 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5732 You need not define this macro if it would always returns zero or if the
5733 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5734 For example, here is the definition used on the SPARC, where floating-point
5735 inequality comparisons are always given @code{CCFPEmode}:
5738 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5742 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
5743 A C expression whose value is reversed condition code of the @var{code} for
5744 comparison done in CC_MODE @var{mode}. The macro is used only in case
5745 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5746 machine has some non-standard way how to reverse certain conditionals. For
5747 instance in case all floating point conditions are non-trapping, compiler may
5748 freely convert unordered compares to ordered one. Then definition may look
5752 #define REVERSE_CONDITION(CODE, MODE) \
5753 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5754 : reverse_condition_maybe_unordered (CODE))
5758 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
5759 A C expression that returns true if the conditional execution predicate
5760 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
5761 versa. Define this to return 0 if the target has conditional execution
5762 predicates that cannot be reversed safely. There is no need to validate
5763 that the arguments of op1 and op2 are the same, this is done separately.
5764 If no expansion is specified, this macro is defined as follows:
5767 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5768 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
5772 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *, unsigned int *)
5773 On targets which do not use @code{(cc0)}, and which use a hard
5774 register rather than a pseudo-register to hold condition codes, the
5775 regular CSE passes are often not able to identify cases in which the
5776 hard register is set to a common value. Use this hook to enable a
5777 small pass which optimizes such cases. This hook should return true
5778 to enable this pass, and it should set the integers to which its
5779 arguments point to the hard register numbers used for condition codes.
5780 When there is only one such register, as is true on most systems, the
5781 integer pointed to by the second argument should be set to
5782 @code{INVALID_REGNUM}.
5784 The default version of this hook returns false.
5787 @deftypefn {Target Hook} enum machine_mode TARGET_CC_MODES_COMPATIBLE (enum machine_mode, enum machine_mode)
5788 On targets which use multiple condition code modes in class
5789 @code{MODE_CC}, it is sometimes the case that a comparison can be
5790 validly done in more than one mode. On such a system, define this
5791 target hook to take two mode arguments and to return a mode in which
5792 both comparisons may be validly done. If there is no such mode,
5793 return @code{VOIDmode}.
5795 The default version of this hook checks whether the modes are the
5796 same. If they are, it returns that mode. If they are different, it
5797 returns @code{VOIDmode}.
5801 @section Describing Relative Costs of Operations
5802 @cindex costs of instructions
5803 @cindex relative costs
5804 @cindex speed of instructions
5806 These macros let you describe the relative speed of various operations
5807 on the target machine.
5809 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5810 A C expression for the cost of moving data of mode @var{mode} from a
5811 register in class @var{from} to one in class @var{to}. The classes are
5812 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5813 value of 2 is the default; other values are interpreted relative to
5816 It is not required that the cost always equal 2 when @var{from} is the
5817 same as @var{to}; on some machines it is expensive to move between
5818 registers if they are not general registers.
5820 If reload sees an insn consisting of a single @code{set} between two
5821 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5822 classes returns a value of 2, reload does not check to ensure that the
5823 constraints of the insn are met. Setting a cost of other than 2 will
5824 allow reload to verify that the constraints are met. You should do this
5825 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5828 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5829 A C expression for the cost of moving data of mode @var{mode} between a
5830 register of class @var{class} and memory; @var{in} is zero if the value
5831 is to be written to memory, nonzero if it is to be read in. This cost
5832 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5833 registers and memory is more expensive than between two registers, you
5834 should define this macro to express the relative cost.
5836 If you do not define this macro, GCC uses a default cost of 4 plus
5837 the cost of copying via a secondary reload register, if one is
5838 needed. If your machine requires a secondary reload register to copy
5839 between memory and a register of @var{class} but the reload mechanism is
5840 more complex than copying via an intermediate, define this macro to
5841 reflect the actual cost of the move.
5843 GCC defines the function @code{memory_move_secondary_cost} if
5844 secondary reloads are needed. It computes the costs due to copying via
5845 a secondary register. If your machine copies from memory using a
5846 secondary register in the conventional way but the default base value of
5847 4 is not correct for your machine, define this macro to add some other
5848 value to the result of that function. The arguments to that function
5849 are the same as to this macro.
5853 A C expression for the cost of a branch instruction. A value of 1 is
5854 the default; other values are interpreted relative to that.
5857 Here are additional macros which do not specify precise relative costs,
5858 but only that certain actions are more expensive than GCC would
5861 @defmac SLOW_BYTE_ACCESS
5862 Define this macro as a C expression which is nonzero if accessing less
5863 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5864 faster than accessing a word of memory, i.e., if such access
5865 require more than one instruction or if there is no difference in cost
5866 between byte and (aligned) word loads.
5868 When this macro is not defined, the compiler will access a field by
5869 finding the smallest containing object; when it is defined, a fullword
5870 load will be used if alignment permits. Unless bytes accesses are
5871 faster than word accesses, using word accesses is preferable since it
5872 may eliminate subsequent memory access if subsequent accesses occur to
5873 other fields in the same word of the structure, but to different bytes.
5876 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5877 Define this macro to be the value 1 if memory accesses described by the
5878 @var{mode} and @var{alignment} parameters have a cost many times greater
5879 than aligned accesses, for example if they are emulated in a trap
5882 When this macro is nonzero, the compiler will act as if
5883 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5884 moves. This can cause significantly more instructions to be produced.
5885 Therefore, do not set this macro nonzero if unaligned accesses only add a
5886 cycle or two to the time for a memory access.
5888 If the value of this macro is always zero, it need not be defined. If
5889 this macro is defined, it should produce a nonzero value when
5890 @code{STRICT_ALIGNMENT} is nonzero.
5894 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5895 which a sequence of insns should be generated instead of a
5896 string move insn or a library call. Increasing the value will always
5897 make code faster, but eventually incurs high cost in increased code size.
5899 Note that on machines where the corresponding move insn is a
5900 @code{define_expand} that emits a sequence of insns, this macro counts
5901 the number of such sequences.
5903 If you don't define this, a reasonable default is used.
5906 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5907 A C expression used to determine whether @code{move_by_pieces} will be used to
5908 copy a chunk of memory, or whether some other block move mechanism
5909 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5910 than @code{MOVE_RATIO}.
5913 @defmac MOVE_MAX_PIECES
5914 A C expression used by @code{move_by_pieces} to determine the largest unit
5915 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5919 The threshold of number of scalar move insns, @emph{below} which a sequence
5920 of insns should be generated to clear memory instead of a string clear insn
5921 or a library call. Increasing the value will always make code faster, but
5922 eventually incurs high cost in increased code size.
5924 If you don't define this, a reasonable default is used.
5927 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5928 A C expression used to determine whether @code{clear_by_pieces} will be used
5929 to clear a chunk of memory, or whether some other block clear mechanism
5930 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5931 than @code{CLEAR_RATIO}.
5935 The threshold of number of scalar move insns, @emph{below} which a sequence
5936 of insns should be generated to set memory to a constant value, instead of
5937 a block set insn or a library call.
5938 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, it defaults to the value of @code{MOVE_RATIO}.
5944 @defmac SET_BY_PIECES_P (@var{size}, @var{alignment})
5945 A C expression used to determine whether @code{store_by_pieces} will be
5946 used to set a chunk of memory to a constant value, or whether some
5947 other mechanism will be used. Used by @code{__builtin_memset} when
5948 storing values other than constant zero.
5949 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5950 than @code{SET_RATIO}.
5953 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
5954 A C expression used to determine whether @code{store_by_pieces} will be
5955 used to set a chunk of memory to a constant string value, or whether some
5956 other mechanism will be used. Used by @code{__builtin_strcpy} when
5957 called with a constant source string.
5958 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5959 than @code{MOVE_RATIO}.
5962 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
5963 A C expression used to determine whether a load postincrement is a good
5964 thing to use for a given mode. Defaults to the value of
5965 @code{HAVE_POST_INCREMENT}.
5968 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
5969 A C expression used to determine whether a load postdecrement is a good
5970 thing to use for a given mode. Defaults to the value of
5971 @code{HAVE_POST_DECREMENT}.
5974 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
5975 A C expression used to determine whether a load preincrement is a good
5976 thing to use for a given mode. Defaults to the value of
5977 @code{HAVE_PRE_INCREMENT}.
5980 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
5981 A C expression used to determine whether a load predecrement is a good
5982 thing to use for a given mode. Defaults to the value of
5983 @code{HAVE_PRE_DECREMENT}.
5986 @defmac USE_STORE_POST_INCREMENT (@var{mode})
5987 A C expression used to determine whether a store postincrement is a good
5988 thing to use for a given mode. Defaults to the value of
5989 @code{HAVE_POST_INCREMENT}.
5992 @defmac USE_STORE_POST_DECREMENT (@var{mode})
5993 A C expression used to determine whether a store postdecrement is a good
5994 thing to use for a given mode. Defaults to the value of
5995 @code{HAVE_POST_DECREMENT}.
5998 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
5999 This macro is used to determine whether a store preincrement is a good
6000 thing to use for a given mode. Defaults to the value of
6001 @code{HAVE_PRE_INCREMENT}.
6004 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
6005 This macro is used to determine whether a store predecrement is a good
6006 thing to use for a given mode. Defaults to the value of
6007 @code{HAVE_PRE_DECREMENT}.
6010 @defmac NO_FUNCTION_CSE
6011 Define this macro if it is as good or better to call a constant
6012 function address than to call an address kept in a register.
6015 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
6016 Define this macro if a non-short-circuit operation produced by
6017 @samp{fold_range_test ()} is optimal. This macro defaults to true if
6018 @code{BRANCH_COST} is greater than or equal to the value 2.
6021 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total})
6022 This target hook describes the relative costs of RTL expressions.
6024 The cost may depend on the precise form of the expression, which is
6025 available for examination in @var{x}, and the rtx code of the expression
6026 in which it is contained, found in @var{outer_code}. @var{code} is the
6027 expression code---redundant, since it can be obtained with
6028 @code{GET_CODE (@var{x})}.
6030 In implementing this hook, you can use the construct
6031 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6034 On entry to the hook, @code{*@var{total}} contains a default estimate
6035 for the cost of the expression. The hook should modify this value as
6036 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
6037 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
6038 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
6040 When optimizing for code size, i.e.@: when @code{optimize_size} is
6041 nonzero, this target hook should be used to estimate the relative
6042 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6044 The hook returns true when all subexpressions of @var{x} have been
6045 processed, and false when @code{rtx_cost} should recurse.
6048 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address})
6049 This hook computes the cost of an addressing mode that contains
6050 @var{address}. If not defined, the cost is computed from
6051 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
6053 For most CISC machines, the default cost is a good approximation of the
6054 true cost of the addressing mode. However, on RISC machines, all
6055 instructions normally have the same length and execution time. Hence
6056 all addresses will have equal costs.
6058 In cases where more than one form of an address is known, the form with
6059 the lowest cost will be used. If multiple forms have the same, lowest,
6060 cost, the one that is the most complex will be used.
6062 For example, suppose an address that is equal to the sum of a register
6063 and a constant is used twice in the same basic block. When this macro
6064 is not defined, the address will be computed in a register and memory
6065 references will be indirect through that register. On machines where
6066 the cost of the addressing mode containing the sum is no higher than
6067 that of a simple indirect reference, this will produce an additional
6068 instruction and possibly require an additional register. Proper
6069 specification of this macro eliminates this overhead for such machines.
6071 This hook is never called with an invalid address.
6073 On machines where an address involving more than one register is as
6074 cheap as an address computation involving only one register, defining
6075 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
6076 be live over a region of code where only one would have been if
6077 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
6078 should be considered in the definition of this macro. Equivalent costs
6079 should probably only be given to addresses with different numbers of
6080 registers on machines with lots of registers.
6084 @section Adjusting the Instruction Scheduler
6086 The instruction scheduler may need a fair amount of machine-specific
6087 adjustment in order to produce good code. GCC provides several target
6088 hooks for this purpose. It is usually enough to define just a few of
6089 them: try the first ones in this list first.
6091 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
6092 This hook returns the maximum number of instructions that can ever
6093 issue at the same time on the target machine. The default is one.
6094 Although the insn scheduler can define itself the possibility of issue
6095 an insn on the same cycle, the value can serve as an additional
6096 constraint to issue insns on the same simulated processor cycle (see
6097 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
6098 This value must be constant over the entire compilation. If you need
6099 it to vary depending on what the instructions are, you must use
6100 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
6103 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
6104 This hook is executed by the scheduler after it has scheduled an insn
6105 from the ready list. It should return the number of insns which can
6106 still be issued in the current cycle. The default is
6107 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
6108 @code{USE}, which normally are not counted against the issue rate.
6109 You should define this hook if some insns take more machine resources
6110 than others, so that fewer insns can follow them in the same cycle.
6111 @var{file} is either a null pointer, or a stdio stream to write any
6112 debug output to. @var{verbose} is the verbose level provided by
6113 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
6117 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
6118 This function corrects the value of @var{cost} based on the
6119 relationship between @var{insn} and @var{dep_insn} through the
6120 dependence @var{link}. It should return the new value. The default
6121 is to make no adjustment to @var{cost}. This can be used for example
6122 to specify to the scheduler using the traditional pipeline description
6123 that an output- or anti-dependence does not incur the same cost as a
6124 data-dependence. If the scheduler using the automaton based pipeline
6125 description, the cost of anti-dependence is zero and the cost of
6126 output-dependence is maximum of one and the difference of latency
6127 times of the first and the second insns. If these values are not
6128 acceptable, you could use the hook to modify them too. See also
6129 @pxref{Processor pipeline description}.
6132 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
6133 This hook adjusts the integer scheduling priority @var{priority} of
6134 @var{insn}. It should return the new priority. Increase the priority to
6135 execute @var{insn} earlier, reduce the priority to execute @var{insn}
6136 later. Do not define this hook if you do not need to adjust the
6137 scheduling priorities of insns.
6140 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
6141 This hook is executed by the scheduler after it has scheduled the ready
6142 list, to allow the machine description to reorder it (for example to
6143 combine two small instructions together on @samp{VLIW} machines).
6144 @var{file} is either a null pointer, or a stdio stream to write any
6145 debug output to. @var{verbose} is the verbose level provided by
6146 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
6147 list of instructions that are ready to be scheduled. @var{n_readyp} is
6148 a pointer to the number of elements in the ready list. The scheduler
6149 reads the ready list in reverse order, starting with
6150 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
6151 is the timer tick of the scheduler. You may modify the ready list and
6152 the number of ready insns. The return value is the number of insns that
6153 can issue this cycle; normally this is just @code{issue_rate}. See also
6154 @samp{TARGET_SCHED_REORDER2}.
6157 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
6158 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
6159 function is called whenever the scheduler starts a new cycle. This one
6160 is called once per iteration over a cycle, immediately after
6161 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
6162 return the number of insns to be scheduled in the same cycle. Defining
6163 this hook can be useful if there are frequent situations where
6164 scheduling one insn causes other insns to become ready in the same
6165 cycle. These other insns can then be taken into account properly.
6168 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
6169 This hook is called after evaluation forward dependencies of insns in
6170 chain given by two parameter values (@var{head} and @var{tail}
6171 correspondingly) but before insns scheduling of the insn chain. For
6172 example, it can be used for better insn classification if it requires
6173 analysis of dependencies. This hook can use backward and forward
6174 dependencies of the insn scheduler because they are already
6178 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
6179 This hook is executed by the scheduler at the beginning of each block of
6180 instructions that are to be scheduled. @var{file} is either a null
6181 pointer, or a stdio stream to write any debug output to. @var{verbose}
6182 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6183 @var{max_ready} is the maximum number of insns in the current scheduling
6184 region that can be live at the same time. This can be used to allocate
6185 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
6188 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
6189 This hook is executed by the scheduler at the end of each block of
6190 instructions that are to be scheduled. It can be used to perform
6191 cleanup of any actions done by the other scheduling hooks. @var{file}
6192 is either a null pointer, or a stdio stream to write any debug output
6193 to. @var{verbose} is the verbose level provided by
6194 @option{-fsched-verbose-@var{n}}.
6197 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
6198 This hook is executed by the scheduler after function level initializations.
6199 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6200 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6201 @var{old_max_uid} is the maximum insn uid when scheduling begins.
6204 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
6205 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
6206 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6207 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6210 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
6211 The hook returns an RTL insn. The automaton state used in the
6212 pipeline hazard recognizer is changed as if the insn were scheduled
6213 when the new simulated processor cycle starts. Usage of the hook may
6214 simplify the automaton pipeline description for some @acronym{VLIW}
6215 processors. If the hook is defined, it is used only for the automaton
6216 based pipeline description. The default is not to change the state
6217 when the new simulated processor cycle starts.
6220 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
6221 The hook can be used to initialize data used by the previous hook.
6224 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
6225 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6226 to changed the state as if the insn were scheduled when the new
6227 simulated processor cycle finishes.
6230 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
6231 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
6232 used to initialize data used by the previous hook.
6235 @deftypefn {Target Hook} void TARGET_SCHED_DFA_PRE_CYCLE_ADVANCE (void)
6236 The hook to notify target that the current simulated cycle is about to finish.
6237 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6238 to change the state in more complicated situations - e.g., when advancing
6239 state on a single insn is not enough.
6242 @deftypefn {Target Hook} void TARGET_SCHED_DFA_POST_CYCLE_ADVANCE (void)
6243 The hook to notify target that new simulated cycle has just started.
6244 The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used
6245 to change the state in more complicated situations - e.g., when advancing
6246 state on a single insn is not enough.
6249 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
6250 This hook controls better choosing an insn from the ready insn queue
6251 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
6252 chooses the first insn from the queue. If the hook returns a positive
6253 value, an additional scheduler code tries all permutations of
6254 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
6255 subsequent ready insns to choose an insn whose issue will result in
6256 maximal number of issued insns on the same cycle. For the
6257 @acronym{VLIW} processor, the code could actually solve the problem of
6258 packing simple insns into the @acronym{VLIW} insn. Of course, if the
6259 rules of @acronym{VLIW} packing are described in the automaton.
6261 This code also could be used for superscalar @acronym{RISC}
6262 processors. Let us consider a superscalar @acronym{RISC} processor
6263 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
6264 @var{B}, some insns can be executed only in pipelines @var{B} or
6265 @var{C}, and one insn can be executed in pipeline @var{B}. The
6266 processor may issue the 1st insn into @var{A} and the 2nd one into
6267 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
6268 until the next cycle. If the scheduler issues the 3rd insn the first,
6269 the processor could issue all 3 insns per cycle.
6271 Actually this code demonstrates advantages of the automaton based
6272 pipeline hazard recognizer. We try quickly and easy many insn
6273 schedules to choose the best one.
6275 The default is no multipass scheduling.
6278 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
6280 This hook controls what insns from the ready insn queue will be
6281 considered for the multipass insn scheduling. If the hook returns
6282 zero for insn passed as the parameter, the insn will be not chosen to
6285 The default is that any ready insns can be chosen to be issued.
6288 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
6290 This hook is called by the insn scheduler before issuing insn passed
6291 as the third parameter on given cycle. If the hook returns nonzero,
6292 the insn is not issued on given processors cycle. Instead of that,
6293 the processor cycle is advanced. If the value passed through the last
6294 parameter is zero, the insn ready queue is not sorted on the new cycle
6295 start as usually. The first parameter passes file for debugging
6296 output. The second one passes the scheduler verbose level of the
6297 debugging output. The forth and the fifth parameter values are
6298 correspondingly processor cycle on which the previous insn has been
6299 issued and the current processor cycle.
6302 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct dep_def *@var{_dep}, int @var{cost}, int @var{distance})
6303 This hook is used to define which dependences are considered costly by
6304 the target, so costly that it is not advisable to schedule the insns that
6305 are involved in the dependence too close to one another. The parameters
6306 to this hook are as follows: The first parameter @var{_dep} is the dependence
6307 being evaluated. The second parameter @var{cost} is the cost of the
6308 dependence, and the third
6309 parameter @var{distance} is the distance in cycles between the two insns.
6310 The hook returns @code{true} if considering the distance between the two
6311 insns the dependence between them is considered costly by the target,
6312 and @code{false} otherwise.
6314 Defining this hook can be useful in multiple-issue out-of-order machines,
6315 where (a) it's practically hopeless to predict the actual data/resource
6316 delays, however: (b) there's a better chance to predict the actual grouping
6317 that will be formed, and (c) correctly emulating the grouping can be very
6318 important. In such targets one may want to allow issuing dependent insns
6319 closer to one another---i.e., closer than the dependence distance; however,
6320 not in cases of "costly dependences", which this hooks allows to define.
6323 @deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void)
6324 This hook is called by the insn scheduler after emitting a new instruction to
6325 the instruction stream. The hook notifies a target backend to extend its
6326 per instruction data structures.
6329 @deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx @var{insn}, int @var{request}, rtx *@var{new_pat})
6330 This hook is called by the insn scheduler when @var{insn} has only
6331 speculative dependencies and therefore can be scheduled speculatively.
6332 The hook is used to check if the pattern of @var{insn} has a speculative
6333 version and, in case of successful check, to generate that speculative
6334 pattern. The hook should return 1, if the instruction has a speculative form,
6335 or -1, if it doesn't. @var{request} describes the type of requested
6336 speculation. If the return value equals 1 then @var{new_pat} is assigned
6337 the generated speculative pattern.
6340 @deftypefn {Target Hook} int TARGET_SCHED_NEEDS_BLOCK_P (rtx @var{insn})
6341 This hook is called by the insn scheduler during generation of recovery code
6342 for @var{insn}. It should return nonzero, if the corresponding check
6343 instruction should branch to recovery code, or zero otherwise.
6346 @deftypefn {Target Hook} rtx TARGET_SCHED_GEN_CHECK (rtx @var{insn}, rtx @var{label}, int @var{mutate_p})
6347 This hook is called by the insn scheduler to generate a pattern for recovery
6348 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
6349 speculative instruction for which the check should be generated.
6350 @var{label} is either a label of a basic block, where recovery code should
6351 be emitted, or a null pointer, when requested check doesn't branch to
6352 recovery code (a simple check). If @var{mutate_p} is nonzero, then
6353 a pattern for a branchy check corresponding to a simple check denoted by
6354 @var{insn} should be generated. In this case @var{label} can't be null.
6357 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC (rtx @var{insn})
6358 This hook is used as a workaround for
6359 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD} not being
6360 called on the first instruction of the ready list. The hook is used to
6361 discard speculative instruction that stand first in the ready list from
6362 being scheduled on the current cycle. For non-speculative instructions,
6363 the hook should always return nonzero. For example, in the ia64 backend
6364 the hook is used to cancel data speculative insns when the ALAT table
6368 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (unsigned int *@var{flags}, spec_info_t @var{spec_info})
6369 This hook is used by the insn scheduler to find out what features should be
6370 enabled/used. @var{flags} initially may have either the SCHED_RGN or SCHED_EBB
6371 bit set. This denotes the scheduler pass for which the data should be
6372 provided. The target backend should modify @var{flags} by modifying
6373 the bits corresponding to the following features: USE_DEPS_LIST, USE_GLAT,
6374 DETACH_LIFE_INFO, and DO_SPECULATION@. For the DO_SPECULATION feature
6375 an additional structure @var{spec_info} should be filled by the target.
6376 The structure describes speculation types that can be used in the scheduler.
6379 @deftypefn {Target Hook} int TARGET_SCHED_SMS_RES_MII (struct ddg *@var{g})
6380 This hook is called by the swing modulo scheduler to calculate a
6381 resource-based lower bound which is based on the resources available in
6382 the machine and the resources required by each instruction. The target
6383 backend can use @var{g} to calculate such bound. A very simple lower
6384 bound will be used in case this hook is not implemented: the total number
6385 of instructions divided by the issue rate.
6389 @section Dividing the Output into Sections (Texts, Data, @dots{})
6390 @c the above section title is WAY too long. maybe cut the part between
6391 @c the (...)? --mew 10feb93
6393 An object file is divided into sections containing different types of
6394 data. In the most common case, there are three sections: the @dfn{text
6395 section}, which holds instructions and read-only data; the @dfn{data
6396 section}, which holds initialized writable data; and the @dfn{bss
6397 section}, which holds uninitialized data. Some systems have other kinds
6400 @file{varasm.c} provides several well-known sections, such as
6401 @code{text_section}, @code{data_section} and @code{bss_section}.
6402 The normal way of controlling a @code{@var{foo}_section} variable
6403 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
6404 as described below. The macros are only read once, when @file{varasm.c}
6405 initializes itself, so their values must be run-time constants.
6406 They may however depend on command-line flags.
6408 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
6409 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
6410 to be string literals.
6412 Some assemblers require a different string to be written every time a
6413 section is selected. If your assembler falls into this category, you
6414 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
6415 @code{get_unnamed_section} to set up the sections.
6417 You must always create a @code{text_section}, either by defining
6418 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
6419 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
6420 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
6421 create a distinct @code{readonly_data_section}, the default is to
6422 reuse @code{text_section}.
6424 All the other @file{varasm.c} sections are optional, and are null
6425 if the target does not provide them.
6427 @defmac TEXT_SECTION_ASM_OP
6428 A C expression whose value is a string, including spacing, containing the
6429 assembler operation that should precede instructions and read-only data.
6430 Normally @code{"\t.text"} is right.
6433 @defmac HOT_TEXT_SECTION_NAME
6434 If defined, a C string constant for the name of the section containing most
6435 frequently executed functions of the program. If not defined, GCC will provide
6436 a default definition if the target supports named sections.
6439 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
6440 If defined, a C string constant for the name of the section containing unlikely
6441 executed functions in the program.
6444 @defmac DATA_SECTION_ASM_OP
6445 A C expression whose value is a string, including spacing, containing the
6446 assembler operation to identify the following data as writable initialized
6447 data. Normally @code{"\t.data"} is right.
6450 @defmac SDATA_SECTION_ASM_OP
6451 If defined, a C expression whose value is a string, including spacing,
6452 containing the assembler operation to identify the following data as
6453 initialized, writable small data.
6456 @defmac READONLY_DATA_SECTION_ASM_OP
6457 A C expression whose value is a string, including spacing, containing the
6458 assembler operation to identify the following data as read-only initialized
6462 @defmac BSS_SECTION_ASM_OP
6463 If defined, a C expression whose value is a string, including spacing,
6464 containing the assembler operation to identify the following data as
6465 uninitialized global data. If not defined, and neither
6466 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
6467 uninitialized global data will be output in the data section if
6468 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
6472 @defmac SBSS_SECTION_ASM_OP
6473 If defined, a C expression whose value is a string, including spacing,
6474 containing the assembler operation to identify the following data as
6475 uninitialized, writable small data.
6478 @defmac INIT_SECTION_ASM_OP
6479 If defined, a C expression whose value is a string, including spacing,
6480 containing the assembler operation to identify the following data as
6481 initialization code. If not defined, GCC will assume such a section does
6482 not exist. This section has no corresponding @code{init_section}
6483 variable; it is used entirely in runtime code.
6486 @defmac FINI_SECTION_ASM_OP
6487 If defined, a C expression whose value is a string, including spacing,
6488 containing the assembler operation to identify the following data as
6489 finalization code. If not defined, GCC will assume such a section does
6490 not exist. This section has no corresponding @code{fini_section}
6491 variable; it is used entirely in runtime code.
6494 @defmac INIT_ARRAY_SECTION_ASM_OP
6495 If defined, a C expression whose value is a string, including spacing,
6496 containing the assembler operation to identify the following data as
6497 part of the @code{.init_array} (or equivalent) section. If not
6498 defined, GCC will assume such a section does not exist. Do not define
6499 both this macro and @code{INIT_SECTION_ASM_OP}.
6502 @defmac FINI_ARRAY_SECTION_ASM_OP
6503 If defined, a C expression whose value is a string, including spacing,
6504 containing the assembler operation to identify the following data as
6505 part of the @code{.fini_array} (or equivalent) section. If not
6506 defined, GCC will assume such a section does not exist. Do not define
6507 both this macro and @code{FINI_SECTION_ASM_OP}.
6510 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
6511 If defined, an ASM statement that switches to a different section
6512 via @var{section_op}, calls @var{function}, and switches back to
6513 the text section. This is used in @file{crtstuff.c} if
6514 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
6515 to initialization and finalization functions from the init and fini
6516 sections. By default, this macro uses a simple function call. Some
6517 ports need hand-crafted assembly code to avoid dependencies on
6518 registers initialized in the function prologue or to ensure that
6519 constant pools don't end up too far way in the text section.
6522 @defmac TARGET_LIBGCC_SDATA_SECTION
6523 If defined, a string which names the section into which small
6524 variables defined in crtstuff and libgcc should go. This is useful
6525 when the target has options for optimizing access to small data, and
6526 you want the crtstuff and libgcc routines to be conservative in what
6527 they expect of your application yet liberal in what your application
6528 expects. For example, for targets with a @code{.sdata} section (like
6529 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
6530 require small data support from your application, but use this macro
6531 to put small data into @code{.sdata} so that your application can
6532 access these variables whether it uses small data or not.
6535 @defmac FORCE_CODE_SECTION_ALIGN
6536 If defined, an ASM statement that aligns a code section to some
6537 arbitrary boundary. This is used to force all fragments of the
6538 @code{.init} and @code{.fini} sections to have to same alignment
6539 and thus prevent the linker from having to add any padding.
6542 @defmac JUMP_TABLES_IN_TEXT_SECTION
6543 Define this macro to be an expression with a nonzero value if jump
6544 tables (for @code{tablejump} insns) should be output in the text
6545 section, along with the assembler instructions. Otherwise, the
6546 readonly data section is used.
6548 This macro is irrelevant if there is no separate readonly data section.
6551 @deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void)
6552 Define this hook if you need to do something special to set up the
6553 @file{varasm.c} sections, or if your target has some special sections
6554 of its own that you need to create.
6556 GCC calls this hook after processing the command line, but before writing
6557 any assembly code, and before calling any of the section-returning hooks
6561 @deftypefn {Target Hook} TARGET_ASM_RELOC_RW_MASK (void)
6562 Return a mask describing how relocations should be treated when
6563 selecting sections. Bit 1 should be set if global relocations
6564 should be placed in a read-write section; bit 0 should be set if
6565 local relocations should be placed in a read-write section.
6567 The default version of this function returns 3 when @option{-fpic}
6568 is in effect, and 0 otherwise. The hook is typically redefined
6569 when the target cannot support (some kinds of) dynamic relocations
6570 in read-only sections even in executables.
6573 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
6574 Return the section into which @var{exp} should be placed. You can
6575 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
6576 some sort. @var{reloc} indicates whether the initial value of @var{exp}
6577 requires link-time relocations. Bit 0 is set when variable contains
6578 local relocations only, while bit 1 is set for global relocations.
6579 @var{align} is the constant alignment in bits.
6581 The default version of this function takes care of putting read-only
6582 variables in @code{readonly_data_section}.
6584 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
6587 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
6588 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
6589 for @code{FUNCTION_DECL}s as well as for variables and constants.
6591 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
6592 function has been determined to be likely to be called, and nonzero if
6593 it is unlikely to be called.
6596 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
6597 Build up a unique section name, expressed as a @code{STRING_CST} node,
6598 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
6599 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
6600 the initial value of @var{exp} requires link-time relocations.
6602 The default version of this function appends the symbol name to the
6603 ELF section name that would normally be used for the symbol. For
6604 example, the function @code{foo} would be placed in @code{.text.foo}.
6605 Whatever the actual target object format, this is often good enough.
6608 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
6609 Return the readonly data section associated with
6610 @samp{DECL_SECTION_NAME (@var{decl})}.
6611 The default version of this function selects @code{.gnu.linkonce.r.name} if
6612 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
6613 if function is in @code{.text.name}, and the normal readonly-data section
6617 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
6618 Return the section into which a constant @var{x}, of mode @var{mode},
6619 should be placed. You can assume that @var{x} is some kind of
6620 constant in RTL@. The argument @var{mode} is redundant except in the
6621 case of a @code{const_int} rtx. @var{align} is the constant alignment
6624 The default version of this function takes care of putting symbolic
6625 constants in @code{flag_pic} mode in @code{data_section} and everything
6626 else in @code{readonly_data_section}.
6629 @deftypefn {Target Hook} void TARGET_MANGLE_DECL_ASSEMBLER_NAME (tree @var{decl}, tree @var{id})
6630 Define this hook if you need to postprocess the assembler name generated
6631 by target-independent code. The @var{id} provided to this hook will be
6632 the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C,
6633 or the mangled name of the @var{decl} in C++). The return value of the
6634 hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on
6635 your target system. The default implementation of this hook just
6636 returns the @var{id} provided.
6639 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
6640 Define this hook if references to a symbol or a constant must be
6641 treated differently depending on something about the variable or
6642 function named by the symbol (such as what section it is in).
6644 The hook is executed immediately after rtl has been created for
6645 @var{decl}, which may be a variable or function declaration or
6646 an entry in the constant pool. In either case, @var{rtl} is the
6647 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
6648 in this hook; that field may not have been initialized yet.
6650 In the case of a constant, it is safe to assume that the rtl is
6651 a @code{mem} whose address is a @code{symbol_ref}. Most decls
6652 will also have this form, but that is not guaranteed. Global
6653 register variables, for instance, will have a @code{reg} for their
6654 rtl. (Normally the right thing to do with such unusual rtl is
6657 The @var{new_decl_p} argument will be true if this is the first time
6658 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
6659 be false for subsequent invocations, which will happen for duplicate
6660 declarations. Whether or not anything must be done for the duplicate
6661 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
6662 @var{new_decl_p} is always true when the hook is called for a constant.
6664 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
6665 The usual thing for this hook to do is to record flags in the
6666 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
6667 Historically, the name string was modified if it was necessary to
6668 encode more than one bit of information, but this practice is now
6669 discouraged; use @code{SYMBOL_REF_FLAGS}.
6671 The default definition of this hook, @code{default_encode_section_info}
6672 in @file{varasm.c}, sets a number of commonly-useful bits in
6673 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
6674 before overriding it.
6677 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
6678 Decode @var{name} and return the real name part, sans
6679 the characters that @code{TARGET_ENCODE_SECTION_INFO}
6683 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
6684 Returns true if @var{exp} should be placed into a ``small data'' section.
6685 The default version of this hook always returns false.
6688 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
6689 Contains the value true if the target places read-only
6690 ``small data'' into a separate section. The default value is false.
6693 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
6694 Returns true if @var{exp} names an object for which name resolution
6695 rules must resolve to the current ``module'' (dynamic shared library
6696 or executable image).
6698 The default version of this hook implements the name resolution rules
6699 for ELF, which has a looser model of global name binding than other
6700 currently supported object file formats.
6703 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
6704 Contains the value true if the target supports thread-local storage.
6705 The default value is false.
6710 @section Position Independent Code
6711 @cindex position independent code
6714 This section describes macros that help implement generation of position
6715 independent code. Simply defining these macros is not enough to
6716 generate valid PIC; you must also add support to the macros
6717 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
6718 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
6719 @samp{movsi} to do something appropriate when the source operand
6720 contains a symbolic address. You may also need to alter the handling of
6721 switch statements so that they use relative addresses.
6722 @c i rearranged the order of the macros above to try to force one of
6723 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
6725 @defmac PIC_OFFSET_TABLE_REGNUM
6726 The register number of the register used to address a table of static
6727 data addresses in memory. In some cases this register is defined by a
6728 processor's ``application binary interface'' (ABI)@. When this macro
6729 is defined, RTL is generated for this register once, as with the stack
6730 pointer and frame pointer registers. If this macro is not defined, it
6731 is up to the machine-dependent files to allocate such a register (if
6732 necessary). Note that this register must be fixed when in use (e.g.@:
6733 when @code{flag_pic} is true).
6736 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6737 Define this macro if the register defined by
6738 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
6739 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
6742 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
6743 A C expression that is nonzero if @var{x} is a legitimate immediate
6744 operand on the target machine when generating position independent code.
6745 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6746 check this. You can also assume @var{flag_pic} is true, so you need not
6747 check it either. You need not define this macro if all constants
6748 (including @code{SYMBOL_REF}) can be immediate operands when generating
6749 position independent code.
6752 @node Assembler Format
6753 @section Defining the Output Assembler Language
6755 This section describes macros whose principal purpose is to describe how
6756 to write instructions in assembler language---rather than what the
6760 * File Framework:: Structural information for the assembler file.
6761 * Data Output:: Output of constants (numbers, strings, addresses).
6762 * Uninitialized Data:: Output of uninitialized variables.
6763 * Label Output:: Output and generation of labels.
6764 * Initialization:: General principles of initialization
6765 and termination routines.
6766 * Macros for Initialization::
6767 Specific macros that control the handling of
6768 initialization and termination routines.
6769 * Instruction Output:: Output of actual instructions.
6770 * Dispatch Tables:: Output of jump tables.
6771 * Exception Region Output:: Output of exception region code.
6772 * Alignment Output:: Pseudo ops for alignment and skipping data.
6775 @node File Framework
6776 @subsection The Overall Framework of an Assembler File
6777 @cindex assembler format
6778 @cindex output of assembler code
6780 @c prevent bad page break with this line
6781 This describes the overall framework of an assembly file.
6783 @deftypefn {Target Hook} void TARGET_ASM_FILE_START ()
6784 @findex default_file_start
6785 Output to @code{asm_out_file} any text which the assembler expects to
6786 find at the beginning of a file. The default behavior is controlled
6787 by two flags, documented below. Unless your target's assembler is
6788 quite unusual, if you override the default, you should call
6789 @code{default_file_start} at some point in your target hook. This
6790 lets other target files rely on these variables.
6793 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
6794 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
6795 printed as the very first line in the assembly file, unless
6796 @option{-fverbose-asm} is in effect. (If that macro has been defined
6797 to the empty string, this variable has no effect.) With the normal
6798 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
6799 assembler that it need not bother stripping comments or extra
6800 whitespace from its input. This allows it to work a bit faster.
6802 The default is false. You should not set it to true unless you have
6803 verified that your port does not generate any extra whitespace or
6804 comments that will cause GAS to issue errors in NO_APP mode.
6807 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
6808 If this flag is true, @code{output_file_directive} will be called
6809 for the primary source file, immediately after printing
6810 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
6811 this to be done. The default is false.
6814 @deftypefn {Target Hook} void TARGET_ASM_FILE_END ()
6815 Output to @code{asm_out_file} any text which the assembler expects
6816 to find at the end of a file. The default is to output nothing.
6819 @deftypefun void file_end_indicate_exec_stack ()
6820 Some systems use a common convention, the @samp{.note.GNU-stack}
6821 special section, to indicate whether or not an object file relies on
6822 the stack being executable. If your system uses this convention, you
6823 should define @code{TARGET_ASM_FILE_END} to this function. If you
6824 need to do other things in that hook, have your hook function call
6828 @defmac ASM_COMMENT_START
6829 A C string constant describing how to begin a comment in the target
6830 assembler language. The compiler assumes that the comment will end at
6831 the end of the line.
6835 A C string constant for text to be output before each @code{asm}
6836 statement or group of consecutive ones. Normally this is
6837 @code{"#APP"}, which is a comment that has no effect on most
6838 assemblers but tells the GNU assembler that it must check the lines
6839 that follow for all valid assembler constructs.
6843 A C string constant for text to be output after each @code{asm}
6844 statement or group of consecutive ones. Normally this is
6845 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6846 time-saving assumptions that are valid for ordinary compiler output.
6849 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6850 A C statement to output COFF information or DWARF debugging information
6851 which indicates that filename @var{name} is the current source file to
6852 the stdio stream @var{stream}.
6854 This macro need not be defined if the standard form of output
6855 for the file format in use is appropriate.
6858 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6859 A C statement to output the string @var{string} to the stdio stream
6860 @var{stream}. If you do not call the function @code{output_quoted_string}
6861 in your config files, GCC will only call it to output filenames to
6862 the assembler source. So you can use it to canonicalize the format
6863 of the filename using this macro.
6866 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6867 A C statement to output something to the assembler file to handle a
6868 @samp{#ident} directive containing the text @var{string}. If this
6869 macro is not defined, nothing is output for a @samp{#ident} directive.
6872 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6873 Output assembly directives to switch to section @var{name}. The section
6874 should have attributes as specified by @var{flags}, which is a bit mask
6875 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6876 is nonzero, it contains an alignment in bytes to be used for the section,
6877 otherwise some target default should be used. Only targets that must
6878 specify an alignment within the section directive need pay attention to
6879 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6882 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6883 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6886 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
6887 @deftypefn {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
6888 This flag is true if we can create zeroed data by switching to a BSS
6889 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
6890 This is true on most ELF targets.
6893 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6894 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6895 based on a variable or function decl, a section name, and whether or not the
6896 declaration's initializer may contain runtime relocations. @var{decl} may be
6897 null, in which case read-write data should be assumed.
6899 The default version of this function handles choosing code vs data,
6900 read-only vs read-write data, and @code{flag_pic}. You should only
6901 need to override this if your target has special flags that might be
6902 set via @code{__attribute__}.
6905 @deftypefn {Target Hook} {int} TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type @var{type}, const char * @var{text})
6906 Provides the target with the ability to record the gcc command line
6907 switches that have been passed to the compiler, and options that are
6908 enabled. The @var{type} argument specifies what is being recorded.
6909 It can take the following values:
6912 @item SWITCH_TYPE_PASSED
6913 @var{text} is a command line switch that has been set by the user.
6915 @item SWITCH_TYPE_ENABLED
6916 @var{text} is an option which has been enabled. This might be as a
6917 direct result of a command line switch, or because it is enabled by
6918 default or because it has been enabled as a side effect of a different
6919 command line switch. For example, the @option{-O2} switch enables
6920 various different individual optimization passes.
6922 @item SWITCH_TYPE_DESCRIPTIVE
6923 @var{text} is either NULL or some descriptive text which should be
6924 ignored. If @var{text} is NULL then it is being used to warn the
6925 target hook that either recording is starting or ending. The first
6926 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
6927 warning is for start up and the second time the warning is for
6928 wind down. This feature is to allow the target hook to make any
6929 necessary preparations before it starts to record switches and to
6930 perform any necessary tidying up after it has finished recording
6933 @item SWITCH_TYPE_LINE_START
6934 This option can be ignored by this target hook.
6936 @item SWITCH_TYPE_LINE_END
6937 This option can be ignored by this target hook.
6940 The hook's return value must be zero. Other return values may be
6941 supported in the future.
6943 By default this hook is set to NULL, but an example implementation is
6944 provided for ELF based targets. Called @var{elf_record_gcc_switches},
6945 it records the switches as ASCII text inside a new, string mergeable
6946 section in the assembler output file. The name of the new section is
6947 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
6951 @deftypefn {Target Hook} {const char *} TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
6952 This is the name of the section that will be created by the example
6953 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
6959 @subsection Output of Data
6962 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6963 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6964 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6965 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6966 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6967 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6968 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6969 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6970 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6971 These hooks specify assembly directives for creating certain kinds
6972 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6973 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6974 aligned two-byte object, and so on. Any of the hooks may be
6975 @code{NULL}, indicating that no suitable directive is available.
6977 The compiler will print these strings at the start of a new line,
6978 followed immediately by the object's initial value. In most cases,
6979 the string should contain a tab, a pseudo-op, and then another tab.
6982 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6983 The @code{assemble_integer} function uses this hook to output an
6984 integer object. @var{x} is the object's value, @var{size} is its size
6985 in bytes and @var{aligned_p} indicates whether it is aligned. The
6986 function should return @code{true} if it was able to output the
6987 object. If it returns false, @code{assemble_integer} will try to
6988 split the object into smaller parts.
6990 The default implementation of this hook will use the
6991 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6992 when the relevant string is @code{NULL}.
6995 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6996 A C statement to recognize @var{rtx} patterns that
6997 @code{output_addr_const} can't deal with, and output assembly code to
6998 @var{stream} corresponding to the pattern @var{x}. This may be used to
6999 allow machine-dependent @code{UNSPEC}s to appear within constants.
7001 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
7002 @code{goto fail}, so that a standard error message is printed. If it
7003 prints an error message itself, by calling, for example,
7004 @code{output_operand_lossage}, it may just complete normally.
7007 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
7008 A C statement to output to the stdio stream @var{stream} an assembler
7009 instruction to assemble a string constant containing the @var{len}
7010 bytes at @var{ptr}. @var{ptr} will be a C expression of type
7011 @code{char *} and @var{len} a C expression of type @code{int}.
7013 If the assembler has a @code{.ascii} pseudo-op as found in the
7014 Berkeley Unix assembler, do not define the macro
7015 @code{ASM_OUTPUT_ASCII}.
7018 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
7019 A C statement to output word @var{n} of a function descriptor for
7020 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
7021 is defined, and is otherwise unused.
7024 @defmac CONSTANT_POOL_BEFORE_FUNCTION
7025 You may define this macro as a C expression. You should define the
7026 expression to have a nonzero value if GCC should output the constant
7027 pool for a function before the code for the function, or a zero value if
7028 GCC should output the constant pool after the function. If you do
7029 not define this macro, the usual case, GCC will output the constant
7030 pool before the function.
7033 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
7034 A C statement to output assembler commands to define the start of the
7035 constant pool for a function. @var{funname} is a string giving
7036 the name of the function. Should the return type of the function
7037 be required, it can be obtained via @var{fundecl}. @var{size}
7038 is the size, in bytes, of the constant pool that will be written
7039 immediately after this call.
7041 If no constant-pool prefix is required, the usual case, this macro need
7045 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
7046 A C statement (with or without semicolon) to output a constant in the
7047 constant pool, if it needs special treatment. (This macro need not do
7048 anything for RTL expressions that can be output normally.)
7050 The argument @var{file} is the standard I/O stream to output the
7051 assembler code on. @var{x} is the RTL expression for the constant to
7052 output, and @var{mode} is the machine mode (in case @var{x} is a
7053 @samp{const_int}). @var{align} is the required alignment for the value
7054 @var{x}; you should output an assembler directive to force this much
7057 The argument @var{labelno} is a number to use in an internal label for
7058 the address of this pool entry. The definition of this macro is
7059 responsible for outputting the label definition at the proper place.
7060 Here is how to do this:
7063 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
7066 When you output a pool entry specially, you should end with a
7067 @code{goto} to the label @var{jumpto}. This will prevent the same pool
7068 entry from being output a second time in the usual manner.
7070 You need not define this macro if it would do nothing.
7073 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
7074 A C statement to output assembler commands to at the end of the constant
7075 pool for a function. @var{funname} is a string giving the name of the
7076 function. Should the return type of the function be required, you can
7077 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
7078 constant pool that GCC wrote immediately before this call.
7080 If no constant-pool epilogue is required, the usual case, you need not
7084 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
7085 Define this macro as a C expression which is nonzero if @var{C} is
7086 used as a logical line separator by the assembler. @var{STR} points
7087 to the position in the string where @var{C} was found; this can be used if
7088 a line separator uses multiple characters.
7090 If you do not define this macro, the default is that only
7091 the character @samp{;} is treated as a logical line separator.
7094 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
7095 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
7096 These target hooks are C string constants, describing the syntax in the
7097 assembler for grouping arithmetic expressions. If not overridden, they
7098 default to normal parentheses, which is correct for most assemblers.
7101 These macros are provided by @file{real.h} for writing the definitions
7102 of @code{ASM_OUTPUT_DOUBLE} and the like:
7104 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
7105 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
7106 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
7107 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
7108 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
7109 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
7110 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
7111 target's floating point representation, and store its bit pattern in
7112 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
7113 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
7114 simple @code{long int}. For the others, it should be an array of
7115 @code{long int}. The number of elements in this array is determined
7116 by the size of the desired target floating point data type: 32 bits of
7117 it go in each @code{long int} array element. Each array element holds
7118 32 bits of the result, even if @code{long int} is wider than 32 bits
7119 on the host machine.
7121 The array element values are designed so that you can print them out
7122 using @code{fprintf} in the order they should appear in the target
7126 @node Uninitialized Data
7127 @subsection Output of Uninitialized Variables
7129 Each of the macros in this section is used to do the whole job of
7130 outputting a single uninitialized variable.
7132 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
7133 A C statement (sans semicolon) to output to the stdio stream
7134 @var{stream} the assembler definition of a common-label named
7135 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7136 is the size rounded up to whatever alignment the caller wants.
7138 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7139 output the name itself; before and after that, output the additional
7140 assembler syntax for defining the name, and a newline.
7142 This macro controls how the assembler definitions of uninitialized
7143 common global variables are output.
7146 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
7147 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
7148 separate, explicit argument. If you define this macro, it is used in
7149 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
7150 handling the required alignment of the variable. The alignment is specified
7151 as the number of bits.
7154 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7155 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
7156 variable to be output, if there is one, or @code{NULL_TREE} if there
7157 is no corresponding variable. If you define this macro, GCC will use it
7158 in place of both @code{ASM_OUTPUT_COMMON} and
7159 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
7160 the variable's decl in order to chose what to output.
7163 @defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
7164 A C statement (sans semicolon) to output to the stdio stream
7165 @var{stream} the assembler definition of uninitialized global @var{decl} named
7166 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7167 is the size rounded up to whatever alignment the caller wants.
7169 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
7170 defining this macro. If unable, use the expression
7171 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
7172 before and after that, output the additional assembler syntax for defining
7173 the name, and a newline.
7175 There are two ways of handling global BSS@. One is to define either
7176 this macro or its aligned counterpart, @code{ASM_OUTPUT_ALIGNED_BSS}.
7177 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
7178 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
7179 You do not need to do both.
7181 Some languages do not have @code{common} data, and require a
7182 non-common form of global BSS in order to handle uninitialized globals
7183 efficiently. C++ is one example of this. However, if the target does
7184 not support global BSS, the front end may choose to make globals
7185 common in order to save space in the object file.
7188 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7189 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
7190 separate, explicit argument. If you define this macro, it is used in
7191 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
7192 handling the required alignment of the variable. The alignment is specified
7193 as the number of bits.
7195 Try to use function @code{asm_output_aligned_bss} defined in file
7196 @file{varasm.c} when defining this macro.
7199 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
7200 A C statement (sans semicolon) to output to the stdio stream
7201 @var{stream} the assembler definition of a local-common-label named
7202 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7203 is the size rounded up to whatever alignment the caller wants.
7205 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7206 output the name itself; before and after that, output the additional
7207 assembler syntax for defining the name, and a newline.
7209 This macro controls how the assembler definitions of uninitialized
7210 static variables are output.
7213 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
7214 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
7215 separate, explicit argument. If you define this macro, it is used in
7216 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
7217 handling the required alignment of the variable. The alignment is specified
7218 as the number of bits.
7221 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7222 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
7223 variable to be output, if there is one, or @code{NULL_TREE} if there
7224 is no corresponding variable. If you define this macro, GCC will use it
7225 in place of both @code{ASM_OUTPUT_DECL} and
7226 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
7227 the variable's decl in order to chose what to output.
7231 @subsection Output and Generation of Labels
7233 @c prevent bad page break with this line
7234 This is about outputting labels.
7236 @findex assemble_name
7237 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
7238 A C statement (sans semicolon) to output to the stdio stream
7239 @var{stream} the assembler definition of a label named @var{name}.
7240 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7241 output the name itself; before and after that, output the additional
7242 assembler syntax for defining the name, and a newline. A default
7243 definition of this macro is provided which is correct for most systems.
7246 @findex assemble_name_raw
7247 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
7248 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
7249 to refer to a compiler-generated label. The default definition uses
7250 @code{assemble_name_raw}, which is like @code{assemble_name} except
7251 that it is more efficient.
7255 A C string containing the appropriate assembler directive to specify the
7256 size of a symbol, without any arguments. On systems that use ELF, the
7257 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
7258 systems, the default is not to define this macro.
7260 Define this macro only if it is correct to use the default definitions
7261 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
7262 for your system. If you need your own custom definitions of those
7263 macros, or if you do not need explicit symbol sizes at all, do not
7267 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
7268 A C statement (sans semicolon) to output to the stdio stream
7269 @var{stream} a directive telling the assembler that the size of the
7270 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
7271 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7275 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
7276 A C statement (sans semicolon) to output to the stdio stream
7277 @var{stream} a directive telling the assembler to calculate the size of
7278 the symbol @var{name} by subtracting its address from the current
7281 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7282 provided. The default assumes that the assembler recognizes a special
7283 @samp{.} symbol as referring to the current address, and can calculate
7284 the difference between this and another symbol. If your assembler does
7285 not recognize @samp{.} or cannot do calculations with it, you will need
7286 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
7290 A C string containing the appropriate assembler directive to specify the
7291 type of a symbol, without any arguments. On systems that use ELF, the
7292 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
7293 systems, the default is not to define this macro.
7295 Define this macro only if it is correct to use the default definition of
7296 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7297 custom definition of this macro, or if you do not need explicit symbol
7298 types at all, do not define this macro.
7301 @defmac TYPE_OPERAND_FMT
7302 A C string which specifies (using @code{printf} syntax) the format of
7303 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
7304 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
7305 the default is not to define this macro.
7307 Define this macro only if it is correct to use the default definition of
7308 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7309 custom definition of this macro, or if you do not need explicit symbol
7310 types at all, do not define this macro.
7313 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
7314 A C statement (sans semicolon) to output to the stdio stream
7315 @var{stream} a directive telling the assembler that the type of the
7316 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
7317 that string is always either @samp{"function"} or @samp{"object"}, but
7318 you should not count on this.
7320 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
7321 definition of this macro is provided.
7324 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
7325 A C statement (sans semicolon) to output to the stdio stream
7326 @var{stream} any text necessary for declaring the name @var{name} of a
7327 function which is being defined. This macro is responsible for
7328 outputting the label definition (perhaps using
7329 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
7330 @code{FUNCTION_DECL} tree node representing the function.
7332 If this macro is not defined, then the function name is defined in the
7333 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7335 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7339 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
7340 A C statement (sans semicolon) to output to the stdio stream
7341 @var{stream} any text necessary for declaring the size of a function
7342 which is being defined. The argument @var{name} is the name of the
7343 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
7344 representing the function.
7346 If this macro is not defined, then the function size is not defined.
7348 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
7352 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
7353 A C statement (sans semicolon) to output to the stdio stream
7354 @var{stream} any text necessary for declaring the name @var{name} of an
7355 initialized variable which is being defined. This macro must output the
7356 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
7357 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
7359 If this macro is not defined, then the variable name is defined in the
7360 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7362 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
7363 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
7366 @defmac ASM_DECLARE_CONSTANT_NAME (@var{stream}, @var{name}, @var{exp}, @var{size})
7367 A C statement (sans semicolon) to output to the stdio stream
7368 @var{stream} any text necessary for declaring the name @var{name} of a
7369 constant which is being defined. This macro is responsible for
7370 outputting the label definition (perhaps using
7371 @code{ASM_OUTPUT_LABEL}). The argument @var{exp} is the
7372 value of the constant, and @var{size} is the size of the constant
7373 in bytes. @var{name} will be an internal label.
7375 If this macro is not defined, then the @var{name} is defined in the
7376 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7378 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7382 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
7383 A C statement (sans semicolon) to output to the stdio stream
7384 @var{stream} any text necessary for claiming a register @var{regno}
7385 for a global variable @var{decl} with name @var{name}.
7387 If you don't define this macro, that is equivalent to defining it to do
7391 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
7392 A C statement (sans semicolon) to finish up declaring a variable name
7393 once the compiler has processed its initializer fully and thus has had a
7394 chance to determine the size of an array when controlled by an
7395 initializer. This is used on systems where it's necessary to declare
7396 something about the size of the object.
7398 If you don't define this macro, that is equivalent to defining it to do
7401 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
7402 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
7405 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
7406 This target hook is a function to output to the stdio stream
7407 @var{stream} some commands that will make the label @var{name} global;
7408 that is, available for reference from other files.
7410 The default implementation relies on a proper definition of
7411 @code{GLOBAL_ASM_OP}.
7414 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *@var{stream}, tree @var{decl})
7415 This target hook is a function to output to the stdio stream
7416 @var{stream} some commands that will make the name associated with @var{decl}
7417 global; that is, available for reference from other files.
7419 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
7422 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
7423 A C statement (sans semicolon) to output to the stdio stream
7424 @var{stream} some commands that will make the label @var{name} weak;
7425 that is, available for reference from other files but only used if
7426 no other definition is available. Use the expression
7427 @code{assemble_name (@var{stream}, @var{name})} to output the name
7428 itself; before and after that, output the additional assembler syntax
7429 for making that name weak, and a newline.
7431 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
7432 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
7436 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
7437 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
7438 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
7439 or variable decl. If @var{value} is not @code{NULL}, this C statement
7440 should output to the stdio stream @var{stream} assembler code which
7441 defines (equates) the weak symbol @var{name} to have the value
7442 @var{value}. If @var{value} is @code{NULL}, it should output commands
7443 to make @var{name} weak.
7446 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
7447 Outputs a directive that enables @var{name} to be used to refer to
7448 symbol @var{value} with weak-symbol semantics. @code{decl} is the
7449 declaration of @code{name}.
7452 @defmac SUPPORTS_WEAK
7453 A C expression which evaluates to true if the target supports weak symbols.
7455 If you don't define this macro, @file{defaults.h} provides a default
7456 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
7457 is defined, the default definition is @samp{1}; otherwise, it is
7458 @samp{0}. Define this macro if you want to control weak symbol support
7459 with a compiler flag such as @option{-melf}.
7462 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
7463 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
7464 public symbol such that extra copies in multiple translation units will
7465 be discarded by the linker. Define this macro if your object file
7466 format provides support for this concept, such as the @samp{COMDAT}
7467 section flags in the Microsoft Windows PE/COFF format, and this support
7468 requires changes to @var{decl}, such as putting it in a separate section.
7471 @defmac SUPPORTS_ONE_ONLY
7472 A C expression which evaluates to true if the target supports one-only
7475 If you don't define this macro, @file{varasm.c} provides a default
7476 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
7477 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
7478 you want to control one-only symbol support with a compiler flag, or if
7479 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
7480 be emitted as one-only.
7483 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
7484 This target hook is a function to output to @var{asm_out_file} some
7485 commands that will make the symbol(s) associated with @var{decl} have
7486 hidden, protected or internal visibility as specified by @var{visibility}.
7489 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
7490 A C expression that evaluates to true if the target's linker expects
7491 that weak symbols do not appear in a static archive's table of contents.
7492 The default is @code{0}.
7494 Leaving weak symbols out of an archive's table of contents means that,
7495 if a symbol will only have a definition in one translation unit and
7496 will have undefined references from other translation units, that
7497 symbol should not be weak. Defining this macro to be nonzero will
7498 thus have the effect that certain symbols that would normally be weak
7499 (explicit template instantiations, and vtables for polymorphic classes
7500 with noninline key methods) will instead be nonweak.
7502 The C++ ABI requires this macro to be zero. Define this macro for
7503 targets where full C++ ABI compliance is impossible and where linker
7504 restrictions require weak symbols to be left out of a static archive's
7508 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
7509 A C statement (sans semicolon) to output to the stdio stream
7510 @var{stream} any text necessary for declaring the name of an external
7511 symbol named @var{name} which is referenced in this compilation but
7512 not defined. The value of @var{decl} is the tree node for the
7515 This macro need not be defined if it does not need to output anything.
7516 The GNU assembler and most Unix assemblers don't require anything.
7519 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
7520 This target hook is a function to output to @var{asm_out_file} an assembler
7521 pseudo-op to declare a library function name external. The name of the
7522 library function is given by @var{symref}, which is a @code{symbol_ref}.
7525 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (tree @var{decl})
7526 This target hook is a function to output to @var{asm_out_file} an assembler
7527 directive to annotate used symbol. Darwin target use .no_dead_code_strip
7531 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
7532 A C statement (sans semicolon) to output to the stdio stream
7533 @var{stream} a reference in assembler syntax to a label named
7534 @var{name}. This should add @samp{_} to the front of the name, if that
7535 is customary on your operating system, as it is in most Berkeley Unix
7536 systems. This macro is used in @code{assemble_name}.
7539 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
7540 A C statement (sans semicolon) to output a reference to
7541 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
7542 will be used to output the name of the symbol. This macro may be used
7543 to modify the way a symbol is referenced depending on information
7544 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
7547 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
7548 A C statement (sans semicolon) to output a reference to @var{buf}, the
7549 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
7550 @code{assemble_name} will be used to output the name of the symbol.
7551 This macro is not used by @code{output_asm_label}, or the @code{%l}
7552 specifier that calls it; the intention is that this macro should be set
7553 when it is necessary to output a label differently when its address is
7557 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
7558 A function to output to the stdio stream @var{stream} a label whose
7559 name is made from the string @var{prefix} and the number @var{labelno}.
7561 It is absolutely essential that these labels be distinct from the labels
7562 used for user-level functions and variables. Otherwise, certain programs
7563 will have name conflicts with internal labels.
7565 It is desirable to exclude internal labels from the symbol table of the
7566 object file. Most assemblers have a naming convention for labels that
7567 should be excluded; on many systems, the letter @samp{L} at the
7568 beginning of a label has this effect. You should find out what
7569 convention your system uses, and follow it.
7571 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
7574 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
7575 A C statement to output to the stdio stream @var{stream} a debug info
7576 label whose name is made from the string @var{prefix} and the number
7577 @var{num}. This is useful for VLIW targets, where debug info labels
7578 may need to be treated differently than branch target labels. On some
7579 systems, branch target labels must be at the beginning of instruction
7580 bundles, but debug info labels can occur in the middle of instruction
7583 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
7587 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
7588 A C statement to store into the string @var{string} a label whose name
7589 is made from the string @var{prefix} and the number @var{num}.
7591 This string, when output subsequently by @code{assemble_name}, should
7592 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
7593 with the same @var{prefix} and @var{num}.
7595 If the string begins with @samp{*}, then @code{assemble_name} will
7596 output the rest of the string unchanged. It is often convenient for
7597 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
7598 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
7599 to output the string, and may change it. (Of course,
7600 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
7601 you should know what it does on your machine.)
7604 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
7605 A C expression to assign to @var{outvar} (which is a variable of type
7606 @code{char *}) a newly allocated string made from the string
7607 @var{name} and the number @var{number}, with some suitable punctuation
7608 added. Use @code{alloca} to get space for the string.
7610 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
7611 produce an assembler label for an internal static variable whose name is
7612 @var{name}. Therefore, the string must be such as to result in valid
7613 assembler code. The argument @var{number} is different each time this
7614 macro is executed; it prevents conflicts between similarly-named
7615 internal static variables in different scopes.
7617 Ideally this string should not be a valid C identifier, to prevent any
7618 conflict with the user's own symbols. Most assemblers allow periods
7619 or percent signs in assembler symbols; putting at least one of these
7620 between the name and the number will suffice.
7622 If this macro is not defined, a default definition will be provided
7623 which is correct for most systems.
7626 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
7627 A C statement to output to the stdio stream @var{stream} assembler code
7628 which defines (equates) the symbol @var{name} to have the value @var{value}.
7631 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7632 correct for most systems.
7635 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
7636 A C statement to output to the stdio stream @var{stream} assembler code
7637 which defines (equates) the symbol whose tree node is @var{decl_of_name}
7638 to have the value of the tree node @var{decl_of_value}. This macro will
7639 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
7640 the tree nodes are available.
7643 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7644 correct for most systems.
7647 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
7648 A C statement that evaluates to true if the assembler code which defines
7649 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
7650 of the tree node @var{decl_of_value} should be emitted near the end of the
7651 current compilation unit. The default is to not defer output of defines.
7652 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
7653 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
7656 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
7657 A C statement to output to the stdio stream @var{stream} assembler code
7658 which defines (equates) the weak symbol @var{name} to have the value
7659 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
7660 an undefined weak symbol.
7662 Define this macro if the target only supports weak aliases; define
7663 @code{ASM_OUTPUT_DEF} instead if possible.
7666 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
7667 Define this macro to override the default assembler names used for
7668 Objective-C methods.
7670 The default name is a unique method number followed by the name of the
7671 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
7672 the category is also included in the assembler name (e.g.@:
7675 These names are safe on most systems, but make debugging difficult since
7676 the method's selector is not present in the name. Therefore, particular
7677 systems define other ways of computing names.
7679 @var{buf} is an expression of type @code{char *} which gives you a
7680 buffer in which to store the name; its length is as long as
7681 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
7682 50 characters extra.
7684 The argument @var{is_inst} specifies whether the method is an instance
7685 method or a class method; @var{class_name} is the name of the class;
7686 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
7687 in a category); and @var{sel_name} is the name of the selector.
7689 On systems where the assembler can handle quoted names, you can use this
7690 macro to provide more human-readable names.
7693 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
7694 A C statement (sans semicolon) to output to the stdio stream
7695 @var{stream} commands to declare that the label @var{name} is an
7696 Objective-C class reference. This is only needed for targets whose
7697 linkers have special support for NeXT-style runtimes.
7700 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
7701 A C statement (sans semicolon) to output to the stdio stream
7702 @var{stream} commands to declare that the label @var{name} is an
7703 unresolved Objective-C class reference. This is only needed for targets
7704 whose linkers have special support for NeXT-style runtimes.
7707 @node Initialization
7708 @subsection How Initialization Functions Are Handled
7709 @cindex initialization routines
7710 @cindex termination routines
7711 @cindex constructors, output of
7712 @cindex destructors, output of
7714 The compiled code for certain languages includes @dfn{constructors}
7715 (also called @dfn{initialization routines})---functions to initialize
7716 data in the program when the program is started. These functions need
7717 to be called before the program is ``started''---that is to say, before
7718 @code{main} is called.
7720 Compiling some languages generates @dfn{destructors} (also called
7721 @dfn{termination routines}) that should be called when the program
7724 To make the initialization and termination functions work, the compiler
7725 must output something in the assembler code to cause those functions to
7726 be called at the appropriate time. When you port the compiler to a new
7727 system, you need to specify how to do this.
7729 There are two major ways that GCC currently supports the execution of
7730 initialization and termination functions. Each way has two variants.
7731 Much of the structure is common to all four variations.
7733 @findex __CTOR_LIST__
7734 @findex __DTOR_LIST__
7735 The linker must build two lists of these functions---a list of
7736 initialization functions, called @code{__CTOR_LIST__}, and a list of
7737 termination functions, called @code{__DTOR_LIST__}.
7739 Each list always begins with an ignored function pointer (which may hold
7740 0, @minus{}1, or a count of the function pointers after it, depending on
7741 the environment). This is followed by a series of zero or more function
7742 pointers to constructors (or destructors), followed by a function
7743 pointer containing zero.
7745 Depending on the operating system and its executable file format, either
7746 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
7747 time and exit time. Constructors are called in reverse order of the
7748 list; destructors in forward order.
7750 The best way to handle static constructors works only for object file
7751 formats which provide arbitrarily-named sections. A section is set
7752 aside for a list of constructors, and another for a list of destructors.
7753 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
7754 object file that defines an initialization function also puts a word in
7755 the constructor section to point to that function. The linker
7756 accumulates all these words into one contiguous @samp{.ctors} section.
7757 Termination functions are handled similarly.
7759 This method will be chosen as the default by @file{target-def.h} if
7760 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
7761 support arbitrary sections, but does support special designated
7762 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
7763 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
7765 When arbitrary sections are available, there are two variants, depending
7766 upon how the code in @file{crtstuff.c} is called. On systems that
7767 support a @dfn{.init} section which is executed at program startup,
7768 parts of @file{crtstuff.c} are compiled into that section. The
7769 program is linked by the @command{gcc} driver like this:
7772 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
7775 The prologue of a function (@code{__init}) appears in the @code{.init}
7776 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
7777 for the function @code{__fini} in the @dfn{.fini} section. Normally these
7778 files are provided by the operating system or by the GNU C library, but
7779 are provided by GCC for a few targets.
7781 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
7782 compiled from @file{crtstuff.c}. They contain, among other things, code
7783 fragments within the @code{.init} and @code{.fini} sections that branch
7784 to routines in the @code{.text} section. The linker will pull all parts
7785 of a section together, which results in a complete @code{__init} function
7786 that invokes the routines we need at startup.
7788 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
7791 If no init section is available, when GCC compiles any function called
7792 @code{main} (or more accurately, any function designated as a program
7793 entry point by the language front end calling @code{expand_main_function}),
7794 it inserts a procedure call to @code{__main} as the first executable code
7795 after the function prologue. The @code{__main} function is defined
7796 in @file{libgcc2.c} and runs the global constructors.
7798 In file formats that don't support arbitrary sections, there are again
7799 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
7800 and an `a.out' format must be used. In this case,
7801 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
7802 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
7803 and with the address of the void function containing the initialization
7804 code as its value. The GNU linker recognizes this as a request to add
7805 the value to a @dfn{set}; the values are accumulated, and are eventually
7806 placed in the executable as a vector in the format described above, with
7807 a leading (ignored) count and a trailing zero element.
7808 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
7809 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
7810 the compilation of @code{main} to call @code{__main} as above, starting
7811 the initialization process.
7813 The last variant uses neither arbitrary sections nor the GNU linker.
7814 This is preferable when you want to do dynamic linking and when using
7815 file formats which the GNU linker does not support, such as `ECOFF'@. In
7816 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
7817 termination functions are recognized simply by their names. This requires
7818 an extra program in the linkage step, called @command{collect2}. This program
7819 pretends to be the linker, for use with GCC; it does its job by running
7820 the ordinary linker, but also arranges to include the vectors of
7821 initialization and termination functions. These functions are called
7822 via @code{__main} as described above. In order to use this method,
7823 @code{use_collect2} must be defined in the target in @file{config.gcc}.
7826 The following section describes the specific macros that control and
7827 customize the handling of initialization and termination functions.
7830 @node Macros for Initialization
7831 @subsection Macros Controlling Initialization Routines
7833 Here are the macros that control how the compiler handles initialization
7834 and termination functions:
7836 @defmac INIT_SECTION_ASM_OP
7837 If defined, a C string constant, including spacing, for the assembler
7838 operation to identify the following data as initialization code. If not
7839 defined, GCC will assume such a section does not exist. When you are
7840 using special sections for initialization and termination functions, this
7841 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7842 run the initialization functions.
7845 @defmac HAS_INIT_SECTION
7846 If defined, @code{main} will not call @code{__main} as described above.
7847 This macro should be defined for systems that control start-up code
7848 on a symbol-by-symbol basis, such as OSF/1, and should not
7849 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7852 @defmac LD_INIT_SWITCH
7853 If defined, a C string constant for a switch that tells the linker that
7854 the following symbol is an initialization routine.
7857 @defmac LD_FINI_SWITCH
7858 If defined, a C string constant for a switch that tells the linker that
7859 the following symbol is a finalization routine.
7862 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7863 If defined, a C statement that will write a function that can be
7864 automatically called when a shared library is loaded. The function
7865 should call @var{func}, which takes no arguments. If not defined, and
7866 the object format requires an explicit initialization function, then a
7867 function called @code{_GLOBAL__DI} will be generated.
7869 This function and the following one are used by collect2 when linking a
7870 shared library that needs constructors or destructors, or has DWARF2
7871 exception tables embedded in the code.
7874 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7875 If defined, a C statement that will write a function that can be
7876 automatically called when a shared library is unloaded. The function
7877 should call @var{func}, which takes no arguments. If not defined, and
7878 the object format requires an explicit finalization function, then a
7879 function called @code{_GLOBAL__DD} will be generated.
7882 @defmac INVOKE__main
7883 If defined, @code{main} will call @code{__main} despite the presence of
7884 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7885 where the init section is not actually run automatically, but is still
7886 useful for collecting the lists of constructors and destructors.
7889 @defmac SUPPORTS_INIT_PRIORITY
7890 If nonzero, the C++ @code{init_priority} attribute is supported and the
7891 compiler should emit instructions to control the order of initialization
7892 of objects. If zero, the compiler will issue an error message upon
7893 encountering an @code{init_priority} attribute.
7896 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7897 This value is true if the target supports some ``native'' method of
7898 collecting constructors and destructors to be run at startup and exit.
7899 It is false if we must use @command{collect2}.
7902 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7903 If defined, a function that outputs assembler code to arrange to call
7904 the function referenced by @var{symbol} at initialization time.
7906 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7907 no arguments and with no return value. If the target supports initialization
7908 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7909 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7911 If this macro is not defined by the target, a suitable default will
7912 be chosen if (1) the target supports arbitrary section names, (2) the
7913 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7917 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7918 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7919 functions rather than initialization functions.
7922 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7923 generated for the generated object file will have static linkage.
7925 If your system uses @command{collect2} as the means of processing
7926 constructors, then that program normally uses @command{nm} to scan
7927 an object file for constructor functions to be called.
7929 On certain kinds of systems, you can define this macro to make
7930 @command{collect2} work faster (and, in some cases, make it work at all):
7932 @defmac OBJECT_FORMAT_COFF
7933 Define this macro if the system uses COFF (Common Object File Format)
7934 object files, so that @command{collect2} can assume this format and scan
7935 object files directly for dynamic constructor/destructor functions.
7937 This macro is effective only in a native compiler; @command{collect2} as
7938 part of a cross compiler always uses @command{nm} for the target machine.
7941 @defmac REAL_NM_FILE_NAME
7942 Define this macro as a C string constant containing the file name to use
7943 to execute @command{nm}. The default is to search the path normally for
7946 If your system supports shared libraries and has a program to list the
7947 dynamic dependencies of a given library or executable, you can define
7948 these macros to enable support for running initialization and
7949 termination functions in shared libraries:
7953 Define this macro to a C string constant containing the name of the program
7954 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7957 @defmac PARSE_LDD_OUTPUT (@var{ptr})
7958 Define this macro to be C code that extracts filenames from the output
7959 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7960 of type @code{char *} that points to the beginning of a line of output
7961 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7962 code must advance @var{ptr} to the beginning of the filename on that
7963 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7966 @defmac SHLIB_SUFFIX
7967 Define this macro to a C string constant containing the default shared
7968 library extension of the target (e.g., @samp{".so"}). @command{collect2}
7969 strips version information after this suffix when generating global
7970 constructor and destructor names. This define is only needed on targets
7971 that use @command{collect2} to process constructors and destructors.
7974 @node Instruction Output
7975 @subsection Output of Assembler Instructions
7977 @c prevent bad page break with this line
7978 This describes assembler instruction output.
7980 @defmac REGISTER_NAMES
7981 A C initializer containing the assembler's names for the machine
7982 registers, each one as a C string constant. This is what translates
7983 register numbers in the compiler into assembler language.
7986 @defmac ADDITIONAL_REGISTER_NAMES
7987 If defined, a C initializer for an array of structures containing a name
7988 and a register number. This macro defines additional names for hard
7989 registers, thus allowing the @code{asm} option in declarations to refer
7990 to registers using alternate names.
7993 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7994 Define this macro if you are using an unusual assembler that
7995 requires different names for the machine instructions.
7997 The definition is a C statement or statements which output an
7998 assembler instruction opcode to the stdio stream @var{stream}. The
7999 macro-operand @var{ptr} is a variable of type @code{char *} which
8000 points to the opcode name in its ``internal'' form---the form that is
8001 written in the machine description. The definition should output the
8002 opcode name to @var{stream}, performing any translation you desire, and
8003 increment the variable @var{ptr} to point at the end of the opcode
8004 so that it will not be output twice.
8006 In fact, your macro definition may process less than the entire opcode
8007 name, or more than the opcode name; but if you want to process text
8008 that includes @samp{%}-sequences to substitute operands, you must take
8009 care of the substitution yourself. Just be sure to increment
8010 @var{ptr} over whatever text should not be output normally.
8012 @findex recog_data.operand
8013 If you need to look at the operand values, they can be found as the
8014 elements of @code{recog_data.operand}.
8016 If the macro definition does nothing, the instruction is output
8020 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
8021 If defined, a C statement to be executed just prior to the output of
8022 assembler code for @var{insn}, to modify the extracted operands so
8023 they will be output differently.
8025 Here the argument @var{opvec} is the vector containing the operands
8026 extracted from @var{insn}, and @var{noperands} is the number of
8027 elements of the vector which contain meaningful data for this insn.
8028 The contents of this vector are what will be used to convert the insn
8029 template into assembler code, so you can change the assembler output
8030 by changing the contents of the vector.
8032 This macro is useful when various assembler syntaxes share a single
8033 file of instruction patterns; by defining this macro differently, you
8034 can cause a large class of instructions to be output differently (such
8035 as with rearranged operands). Naturally, variations in assembler
8036 syntax affecting individual insn patterns ought to be handled by
8037 writing conditional output routines in those patterns.
8039 If this macro is not defined, it is equivalent to a null statement.
8042 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
8043 A C compound statement to output to stdio stream @var{stream} the
8044 assembler syntax for an instruction operand @var{x}. @var{x} is an
8047 @var{code} is a value that can be used to specify one of several ways
8048 of printing the operand. It is used when identical operands must be
8049 printed differently depending on the context. @var{code} comes from
8050 the @samp{%} specification that was used to request printing of the
8051 operand. If the specification was just @samp{%@var{digit}} then
8052 @var{code} is 0; if the specification was @samp{%@var{ltr}
8053 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
8056 If @var{x} is a register, this macro should print the register's name.
8057 The names can be found in an array @code{reg_names} whose type is
8058 @code{char *[]}. @code{reg_names} is initialized from
8059 @code{REGISTER_NAMES}.
8061 When the machine description has a specification @samp{%@var{punct}}
8062 (a @samp{%} followed by a punctuation character), this macro is called
8063 with a null pointer for @var{x} and the punctuation character for
8067 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
8068 A C expression which evaluates to true if @var{code} is a valid
8069 punctuation character for use in the @code{PRINT_OPERAND} macro. If
8070 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
8071 punctuation characters (except for the standard one, @samp{%}) are used
8075 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
8076 A C compound statement to output to stdio stream @var{stream} the
8077 assembler syntax for an instruction operand that is a memory reference
8078 whose address is @var{x}. @var{x} is an RTL expression.
8080 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
8081 On some machines, the syntax for a symbolic address depends on the
8082 section that the address refers to. On these machines, define the hook
8083 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
8084 @code{symbol_ref}, and then check for it here. @xref{Assembler
8088 @findex dbr_sequence_length
8089 @defmac DBR_OUTPUT_SEQEND (@var{file})
8090 A C statement, to be executed after all slot-filler instructions have
8091 been output. If necessary, call @code{dbr_sequence_length} to
8092 determine the number of slots filled in a sequence (zero if not
8093 currently outputting a sequence), to decide how many no-ops to output,
8096 Don't define this macro if it has nothing to do, but it is helpful in
8097 reading assembly output if the extent of the delay sequence is made
8098 explicit (e.g.@: with white space).
8101 @findex final_sequence
8102 Note that output routines for instructions with delay slots must be
8103 prepared to deal with not being output as part of a sequence
8104 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
8105 found.) The variable @code{final_sequence} is null when not
8106 processing a sequence, otherwise it contains the @code{sequence} rtx
8110 @defmac REGISTER_PREFIX
8111 @defmacx LOCAL_LABEL_PREFIX
8112 @defmacx USER_LABEL_PREFIX
8113 @defmacx IMMEDIATE_PREFIX
8114 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
8115 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
8116 @file{final.c}). These are useful when a single @file{md} file must
8117 support multiple assembler formats. In that case, the various @file{tm.h}
8118 files can define these macros differently.
8121 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
8122 If defined this macro should expand to a series of @code{case}
8123 statements which will be parsed inside the @code{switch} statement of
8124 the @code{asm_fprintf} function. This allows targets to define extra
8125 printf formats which may useful when generating their assembler
8126 statements. Note that uppercase letters are reserved for future
8127 generic extensions to asm_fprintf, and so are not available to target
8128 specific code. The output file is given by the parameter @var{file}.
8129 The varargs input pointer is @var{argptr} and the rest of the format
8130 string, starting the character after the one that is being switched
8131 upon, is pointed to by @var{format}.
8134 @defmac ASSEMBLER_DIALECT
8135 If your target supports multiple dialects of assembler language (such as
8136 different opcodes), define this macro as a C expression that gives the
8137 numeric index of the assembler language dialect to use, with zero as the
8140 If this macro is defined, you may use constructs of the form
8142 @samp{@{option0|option1|option2@dots{}@}}
8145 in the output templates of patterns (@pxref{Output Template}) or in the
8146 first argument of @code{asm_fprintf}. This construct outputs
8147 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
8148 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
8149 within these strings retain their usual meaning. If there are fewer
8150 alternatives within the braces than the value of
8151 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
8153 If you do not define this macro, the characters @samp{@{}, @samp{|} and
8154 @samp{@}} do not have any special meaning when used in templates or
8155 operands to @code{asm_fprintf}.
8157 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
8158 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
8159 the variations in assembler language syntax with that mechanism. Define
8160 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
8161 if the syntax variant are larger and involve such things as different
8162 opcodes or operand order.
8165 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
8166 A C expression to output to @var{stream} some assembler code
8167 which will push hard register number @var{regno} onto the stack.
8168 The code need not be optimal, since this macro is used only when
8172 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
8173 A C expression to output to @var{stream} some assembler code
8174 which will pop hard register number @var{regno} off of the stack.
8175 The code need not be optimal, since this macro is used only when
8179 @node Dispatch Tables
8180 @subsection Output of Dispatch Tables
8182 @c prevent bad page break with this line
8183 This concerns dispatch tables.
8185 @cindex dispatch table
8186 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
8187 A C statement to output to the stdio stream @var{stream} an assembler
8188 pseudo-instruction to generate a difference between two labels.
8189 @var{value} and @var{rel} are the numbers of two internal labels. The
8190 definitions of these labels are output using
8191 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
8192 way here. For example,
8195 fprintf (@var{stream}, "\t.word L%d-L%d\n",
8196 @var{value}, @var{rel})
8199 You must provide this macro on machines where the addresses in a
8200 dispatch table are relative to the table's own address. If defined, GCC
8201 will also use this macro on all machines when producing PIC@.
8202 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
8203 mode and flags can be read.
8206 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
8207 This macro should be provided on machines where the addresses
8208 in a dispatch table are absolute.
8210 The definition should be a C statement to output to the stdio stream
8211 @var{stream} an assembler pseudo-instruction to generate a reference to
8212 a label. @var{value} is the number of an internal label whose
8213 definition is output using @code{(*targetm.asm_out.internal_label)}.
8217 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
8221 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
8222 Define this if the label before a jump-table needs to be output
8223 specially. The first three arguments are the same as for
8224 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
8225 jump-table which follows (a @code{jump_insn} containing an
8226 @code{addr_vec} or @code{addr_diff_vec}).
8228 This feature is used on system V to output a @code{swbeg} statement
8231 If this macro is not defined, these labels are output with
8232 @code{(*targetm.asm_out.internal_label)}.
8235 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
8236 Define this if something special must be output at the end of a
8237 jump-table. The definition should be a C statement to be executed
8238 after the assembler code for the table is written. It should write
8239 the appropriate code to stdio stream @var{stream}. The argument
8240 @var{table} is the jump-table insn, and @var{num} is the label-number
8241 of the preceding label.
8243 If this macro is not defined, nothing special is output at the end of
8247 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (@var{stream}, @var{decl}, @var{for_eh}, @var{empty})
8248 This target hook emits a label at the beginning of each FDE@. It
8249 should be defined on targets where FDEs need special labels, and it
8250 should write the appropriate label, for the FDE associated with the
8251 function declaration @var{decl}, to the stdio stream @var{stream}.
8252 The third argument, @var{for_eh}, is a boolean: true if this is for an
8253 exception table. The fourth argument, @var{empty}, is a boolean:
8254 true if this is a placeholder label for an omitted FDE@.
8256 The default is that FDEs are not given nonlocal labels.
8259 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (@var{stream})
8260 This target hook emits a label at the beginning of the exception table.
8261 It should be defined on targets where it is desirable for the table
8262 to be broken up according to function.
8264 The default is that no label is emitted.
8267 @deftypefn {Target Hook} void TARGET_UNWIND_EMIT (FILE * @var{stream}, rtx @var{insn})
8268 This target hook emits and assembly directives required to unwind the
8269 given instruction. This is only used when TARGET_UNWIND_INFO is set.
8272 @node Exception Region Output
8273 @subsection Assembler Commands for Exception Regions
8275 @c prevent bad page break with this line
8277 This describes commands marking the start and the end of an exception
8280 @defmac EH_FRAME_SECTION_NAME
8281 If defined, a C string constant for the name of the section containing
8282 exception handling frame unwind information. If not defined, GCC will
8283 provide a default definition if the target supports named sections.
8284 @file{crtstuff.c} uses this macro to switch to the appropriate section.
8286 You should define this symbol if your target supports DWARF 2 frame
8287 unwind information and the default definition does not work.
8290 @defmac EH_FRAME_IN_DATA_SECTION
8291 If defined, DWARF 2 frame unwind information will be placed in the
8292 data section even though the target supports named sections. This
8293 might be necessary, for instance, if the system linker does garbage
8294 collection and sections cannot be marked as not to be collected.
8296 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
8300 @defmac EH_TABLES_CAN_BE_READ_ONLY
8301 Define this macro to 1 if your target is such that no frame unwind
8302 information encoding used with non-PIC code will ever require a
8303 runtime relocation, but the linker may not support merging read-only
8304 and read-write sections into a single read-write section.
8307 @defmac MASK_RETURN_ADDR
8308 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
8309 that it does not contain any extraneous set bits in it.
8312 @defmac DWARF2_UNWIND_INFO
8313 Define this macro to 0 if your target supports DWARF 2 frame unwind
8314 information, but it does not yet work with exception handling.
8315 Otherwise, if your target supports this information (if it defines
8316 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
8317 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of 1.
8319 If @code{TARGET_UNWIND_INFO} is defined, the target specific unwinder
8320 will be used in all cases. Defining this macro will enable the generation
8321 of DWARF 2 frame debugging information.
8323 If @code{TARGET_UNWIND_INFO} is not defined, and this macro is defined to 1,
8324 the DWARF 2 unwinder will be the default exception handling mechanism;
8325 otherwise, the @code{setjmp}/@code{longjmp}-based scheme will be used by
8329 @defmac TARGET_UNWIND_INFO
8330 Define this macro if your target has ABI specified unwind tables. Usually
8331 these will be output by @code{TARGET_UNWIND_EMIT}.
8334 @deftypevar {Target Hook} bool TARGET_UNWIND_TABLES_DEFAULT
8335 This variable should be set to @code{true} if the target ABI requires unwinding
8336 tables even when exceptions are not used.
8339 @defmac MUST_USE_SJLJ_EXCEPTIONS
8340 This macro need only be defined if @code{DWARF2_UNWIND_INFO} is
8341 runtime-variable. In that case, @file{except.h} cannot correctly
8342 determine the corresponding definition of @code{MUST_USE_SJLJ_EXCEPTIONS},
8343 so the target must provide it directly.
8346 @defmac DONT_USE_BUILTIN_SETJMP
8347 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
8348 should use the @code{setjmp}/@code{longjmp} functions from the C library
8349 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
8352 @defmac DWARF_CIE_DATA_ALIGNMENT
8353 This macro need only be defined if the target might save registers in the
8354 function prologue at an offset to the stack pointer that is not aligned to
8355 @code{UNITS_PER_WORD}. The definition should be the negative minimum
8356 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
8357 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
8358 the target supports DWARF 2 frame unwind information.
8361 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
8362 Contains the value true if the target should add a zero word onto the
8363 end of a Dwarf-2 frame info section when used for exception handling.
8364 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
8368 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
8369 Given a register, this hook should return a parallel of registers to
8370 represent where to find the register pieces. Define this hook if the
8371 register and its mode are represented in Dwarf in non-contiguous
8372 locations, or if the register should be represented in more than one
8373 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
8374 If not defined, the default is to return @code{NULL_RTX}.
8377 @deftypefn {Target Hook} void TARGET_INIT_DWARF_REG_SIZES_EXTRA (tree @var{address})
8378 If some registers are represented in Dwarf-2 unwind information in
8379 multiple pieces, define this hook to fill in information about the
8380 sizes of those pieces in the table used by the unwinder at runtime.
8381 It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after
8382 filling in a single size corresponding to each hard register;
8383 @var{address} is the address of the table.
8386 @deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym})
8387 This hook is used to output a reference from a frame unwinding table to
8388 the type_info object identified by @var{sym}. It should return @code{true}
8389 if the reference was output. Returning @code{false} will cause the
8390 reference to be output using the normal Dwarf2 routines.
8393 @deftypefn {Target Hook} bool TARGET_ARM_EABI_UNWINDER
8394 This hook should be set to @code{true} on targets that use an ARM EABI
8395 based unwinding library, and @code{false} on other targets. This effects
8396 the format of unwinding tables, and how the unwinder in entered after
8397 running a cleanup. The default is @code{false}.
8400 @node Alignment Output
8401 @subsection Assembler Commands for Alignment
8403 @c prevent bad page break with this line
8404 This describes commands for alignment.
8406 @defmac JUMP_ALIGN (@var{label})
8407 The alignment (log base 2) to put in front of @var{label}, which is
8408 a common destination of jumps and has no fallthru incoming edge.
8410 This macro need not be defined if you don't want any special alignment
8411 to be done at such a time. Most machine descriptions do not currently
8414 Unless it's necessary to inspect the @var{label} parameter, it is better
8415 to set the variable @var{align_jumps} in the target's
8416 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8417 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
8420 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
8421 The alignment (log base 2) to put in front of @var{label}, which follows
8424 This macro need not be defined if you don't want any special alignment
8425 to be done at such a time. Most machine descriptions do not currently
8429 @defmac LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
8430 The maximum number of bytes to skip when applying
8431 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
8432 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8435 @defmac LOOP_ALIGN (@var{label})
8436 The alignment (log base 2) to put in front of @var{label}, which follows
8437 a @code{NOTE_INSN_LOOP_BEG} note.
8439 This macro need not be defined if you don't want any special alignment
8440 to be done at such a time. Most machine descriptions do not currently
8443 Unless it's necessary to inspect the @var{label} parameter, it is better
8444 to set the variable @code{align_loops} in the target's
8445 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8446 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
8449 @defmac LOOP_ALIGN_MAX_SKIP
8450 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
8451 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8454 @defmac LABEL_ALIGN (@var{label})
8455 The alignment (log base 2) to put in front of @var{label}.
8456 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
8457 the maximum of the specified values is used.
8459 Unless it's necessary to inspect the @var{label} parameter, it is better
8460 to set the variable @code{align_labels} in the target's
8461 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8462 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
8465 @defmac LABEL_ALIGN_MAX_SKIP
8466 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
8467 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8470 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
8471 A C statement to output to the stdio stream @var{stream} an assembler
8472 instruction to advance the location counter by @var{nbytes} bytes.
8473 Those bytes should be zero when loaded. @var{nbytes} will be a C
8474 expression of type @code{unsigned HOST_WIDE_INT}.
8477 @defmac ASM_NO_SKIP_IN_TEXT
8478 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
8479 text section because it fails to put zeros in the bytes that are skipped.
8480 This is true on many Unix systems, where the pseudo--op to skip bytes
8481 produces no-op instructions rather than zeros when used in the text
8485 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
8486 A C statement to output to the stdio stream @var{stream} an assembler
8487 command to advance the location counter to a multiple of 2 to the
8488 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
8491 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
8492 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
8493 for padding, if necessary.
8496 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
8497 A C statement to output to the stdio stream @var{stream} an assembler
8498 command to advance the location counter to a multiple of 2 to the
8499 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
8500 satisfy the alignment request. @var{power} and @var{max_skip} will be
8501 a C expression of type @code{int}.
8505 @node Debugging Info
8506 @section Controlling Debugging Information Format
8508 @c prevent bad page break with this line
8509 This describes how to specify debugging information.
8512 * All Debuggers:: Macros that affect all debugging formats uniformly.
8513 * DBX Options:: Macros enabling specific options in DBX format.
8514 * DBX Hooks:: Hook macros for varying DBX format.
8515 * File Names and DBX:: Macros controlling output of file names in DBX format.
8516 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
8517 * VMS Debug:: Macros for VMS debug format.
8521 @subsection Macros Affecting All Debugging Formats
8523 @c prevent bad page break with this line
8524 These macros affect all debugging formats.
8526 @defmac DBX_REGISTER_NUMBER (@var{regno})
8527 A C expression that returns the DBX register number for the compiler
8528 register number @var{regno}. In the default macro provided, the value
8529 of this expression will be @var{regno} itself. But sometimes there are
8530 some registers that the compiler knows about and DBX does not, or vice
8531 versa. In such cases, some register may need to have one number in the
8532 compiler and another for DBX@.
8534 If two registers have consecutive numbers inside GCC, and they can be
8535 used as a pair to hold a multiword value, then they @emph{must} have
8536 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
8537 Otherwise, debuggers will be unable to access such a pair, because they
8538 expect register pairs to be consecutive in their own numbering scheme.
8540 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
8541 does not preserve register pairs, then what you must do instead is
8542 redefine the actual register numbering scheme.
8545 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
8546 A C expression that returns the integer offset value for an automatic
8547 variable having address @var{x} (an RTL expression). The default
8548 computation assumes that @var{x} is based on the frame-pointer and
8549 gives the offset from the frame-pointer. This is required for targets
8550 that produce debugging output for DBX or COFF-style debugging output
8551 for SDB and allow the frame-pointer to be eliminated when the
8552 @option{-g} options is used.
8555 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
8556 A C expression that returns the integer offset value for an argument
8557 having address @var{x} (an RTL expression). The nominal offset is
8561 @defmac PREFERRED_DEBUGGING_TYPE
8562 A C expression that returns the type of debugging output GCC should
8563 produce when the user specifies just @option{-g}. Define
8564 this if you have arranged for GCC to support more than one format of
8565 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
8566 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
8567 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
8569 When the user specifies @option{-ggdb}, GCC normally also uses the
8570 value of this macro to select the debugging output format, but with two
8571 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
8572 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
8573 defined, GCC uses @code{DBX_DEBUG}.
8575 The value of this macro only affects the default debugging output; the
8576 user can always get a specific type of output by using @option{-gstabs},
8577 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
8581 @subsection Specific Options for DBX Output
8583 @c prevent bad page break with this line
8584 These are specific options for DBX output.
8586 @defmac DBX_DEBUGGING_INFO
8587 Define this macro if GCC should produce debugging output for DBX
8588 in response to the @option{-g} option.
8591 @defmac XCOFF_DEBUGGING_INFO
8592 Define this macro if GCC should produce XCOFF format debugging output
8593 in response to the @option{-g} option. This is a variant of DBX format.
8596 @defmac DEFAULT_GDB_EXTENSIONS
8597 Define this macro to control whether GCC should by default generate
8598 GDB's extended version of DBX debugging information (assuming DBX-format
8599 debugging information is enabled at all). If you don't define the
8600 macro, the default is 1: always generate the extended information
8601 if there is any occasion to.
8604 @defmac DEBUG_SYMS_TEXT
8605 Define this macro if all @code{.stabs} commands should be output while
8606 in the text section.
8609 @defmac ASM_STABS_OP
8610 A C string constant, including spacing, naming the assembler pseudo op to
8611 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
8612 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
8613 applies only to DBX debugging information format.
8616 @defmac ASM_STABD_OP
8617 A C string constant, including spacing, naming the assembler pseudo op to
8618 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
8619 value is the current location. If you don't define this macro,
8620 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
8624 @defmac ASM_STABN_OP
8625 A C string constant, including spacing, naming the assembler pseudo op to
8626 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
8627 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
8628 macro applies only to DBX debugging information format.
8631 @defmac DBX_NO_XREFS
8632 Define this macro if DBX on your system does not support the construct
8633 @samp{xs@var{tagname}}. On some systems, this construct is used to
8634 describe a forward reference to a structure named @var{tagname}.
8635 On other systems, this construct is not supported at all.
8638 @defmac DBX_CONTIN_LENGTH
8639 A symbol name in DBX-format debugging information is normally
8640 continued (split into two separate @code{.stabs} directives) when it
8641 exceeds a certain length (by default, 80 characters). On some
8642 operating systems, DBX requires this splitting; on others, splitting
8643 must not be done. You can inhibit splitting by defining this macro
8644 with the value zero. You can override the default splitting-length by
8645 defining this macro as an expression for the length you desire.
8648 @defmac DBX_CONTIN_CHAR
8649 Normally continuation is indicated by adding a @samp{\} character to
8650 the end of a @code{.stabs} string when a continuation follows. To use
8651 a different character instead, define this macro as a character
8652 constant for the character you want to use. Do not define this macro
8653 if backslash is correct for your system.
8656 @defmac DBX_STATIC_STAB_DATA_SECTION
8657 Define this macro if it is necessary to go to the data section before
8658 outputting the @samp{.stabs} pseudo-op for a non-global static
8662 @defmac DBX_TYPE_DECL_STABS_CODE
8663 The value to use in the ``code'' field of the @code{.stabs} directive
8664 for a typedef. The default is @code{N_LSYM}.
8667 @defmac DBX_STATIC_CONST_VAR_CODE
8668 The value to use in the ``code'' field of the @code{.stabs} directive
8669 for a static variable located in the text section. DBX format does not
8670 provide any ``right'' way to do this. The default is @code{N_FUN}.
8673 @defmac DBX_REGPARM_STABS_CODE
8674 The value to use in the ``code'' field of the @code{.stabs} directive
8675 for a parameter passed in registers. DBX format does not provide any
8676 ``right'' way to do this. The default is @code{N_RSYM}.
8679 @defmac DBX_REGPARM_STABS_LETTER
8680 The letter to use in DBX symbol data to identify a symbol as a parameter
8681 passed in registers. DBX format does not customarily provide any way to
8682 do this. The default is @code{'P'}.
8685 @defmac DBX_FUNCTION_FIRST
8686 Define this macro if the DBX information for a function and its
8687 arguments should precede the assembler code for the function. Normally,
8688 in DBX format, the debugging information entirely follows the assembler
8692 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
8693 Define this macro, with value 1, if the value of a symbol describing
8694 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
8695 relative to the start of the enclosing function. Normally, GCC uses
8696 an absolute address.
8699 @defmac DBX_LINES_FUNCTION_RELATIVE
8700 Define this macro, with value 1, if the value of a symbol indicating
8701 the current line number (@code{N_SLINE}) should be relative to the
8702 start of the enclosing function. Normally, GCC uses an absolute address.
8705 @defmac DBX_USE_BINCL
8706 Define this macro if GCC should generate @code{N_BINCL} and
8707 @code{N_EINCL} stabs for included header files, as on Sun systems. This
8708 macro also directs GCC to output a type number as a pair of a file
8709 number and a type number within the file. Normally, GCC does not
8710 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
8711 number for a type number.
8715 @subsection Open-Ended Hooks for DBX Format
8717 @c prevent bad page break with this line
8718 These are hooks for DBX format.
8720 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
8721 Define this macro to say how to output to @var{stream} the debugging
8722 information for the start of a scope level for variable names. The
8723 argument @var{name} is the name of an assembler symbol (for use with
8724 @code{assemble_name}) whose value is the address where the scope begins.
8727 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
8728 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
8731 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
8732 Define this macro if the target machine requires special handling to
8733 output an @code{N_FUN} entry for the function @var{decl}.
8736 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
8737 A C statement to output DBX debugging information before code for line
8738 number @var{line} of the current source file to the stdio stream
8739 @var{stream}. @var{counter} is the number of time the macro was
8740 invoked, including the current invocation; it is intended to generate
8741 unique labels in the assembly output.
8743 This macro should not be defined if the default output is correct, or
8744 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
8747 @defmac NO_DBX_FUNCTION_END
8748 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
8749 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
8750 On those machines, define this macro to turn this feature off without
8751 disturbing the rest of the gdb extensions.
8754 @defmac NO_DBX_BNSYM_ENSYM
8755 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
8756 extension construct. On those machines, define this macro to turn this
8757 feature off without disturbing the rest of the gdb extensions.
8760 @node File Names and DBX
8761 @subsection File Names in DBX Format
8763 @c prevent bad page break with this line
8764 This describes file names in DBX format.
8766 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
8767 A C statement to output DBX debugging information to the stdio stream
8768 @var{stream}, which indicates that file @var{name} is the main source
8769 file---the file specified as the input file for compilation.
8770 This macro is called only once, at the beginning of compilation.
8772 This macro need not be defined if the standard form of output
8773 for DBX debugging information is appropriate.
8775 It may be necessary to refer to a label equal to the beginning of the
8776 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
8777 to do so. If you do this, you must also set the variable
8778 @var{used_ltext_label_name} to @code{true}.
8781 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
8782 Define this macro, with value 1, if GCC should not emit an indication
8783 of the current directory for compilation and current source language at
8784 the beginning of the file.
8787 @defmac NO_DBX_GCC_MARKER
8788 Define this macro, with value 1, if GCC should not emit an indication
8789 that this object file was compiled by GCC@. The default is to emit
8790 an @code{N_OPT} stab at the beginning of every source file, with
8791 @samp{gcc2_compiled.} for the string and value 0.
8794 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
8795 A C statement to output DBX debugging information at the end of
8796 compilation of the main source file @var{name}. Output should be
8797 written to the stdio stream @var{stream}.
8799 If you don't define this macro, nothing special is output at the end
8800 of compilation, which is correct for most machines.
8803 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
8804 Define this macro @emph{instead of} defining
8805 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
8806 the end of compilation is a @code{N_SO} stab with an empty string,
8807 whose value is the highest absolute text address in the file.
8812 @subsection Macros for SDB and DWARF Output
8814 @c prevent bad page break with this line
8815 Here are macros for SDB and DWARF output.
8817 @defmac SDB_DEBUGGING_INFO
8818 Define this macro if GCC should produce COFF-style debugging output
8819 for SDB in response to the @option{-g} option.
8822 @defmac DWARF2_DEBUGGING_INFO
8823 Define this macro if GCC should produce dwarf version 2 format
8824 debugging output in response to the @option{-g} option.
8826 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (tree @var{function})
8827 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
8828 be emitted for each function. Instead of an integer return the enum
8829 value for the @code{DW_CC_} tag.
8832 To support optional call frame debugging information, you must also
8833 define @code{INCOMING_RETURN_ADDR_RTX} and either set
8834 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
8835 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
8836 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
8839 @defmac DWARF2_FRAME_INFO
8840 Define this macro to a nonzero value if GCC should always output
8841 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
8842 (@pxref{Exception Region Output} is nonzero, GCC will output this
8843 information not matter how you define @code{DWARF2_FRAME_INFO}.
8846 @defmac DWARF2_ASM_LINE_DEBUG_INFO
8847 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8848 line debug info sections. This will result in much more compact line number
8849 tables, and hence is desirable if it works.
8852 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8853 A C statement to issue assembly directives that create a difference
8854 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
8857 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
8858 A C statement to issue assembly directives that create a
8859 section-relative reference to the given @var{label}, using an integer of the
8860 given @var{size}. The label is known to be defined in the given @var{section}.
8863 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8864 A C statement to issue assembly directives that create a self-relative
8865 reference to the given @var{label}, using an integer of the given @var{size}.
8868 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{FILE}, int @var{size}, rtx @var{x})
8869 If defined, this target hook is a function which outputs a DTP-relative
8870 reference to the given TLS symbol of the specified size.
8873 @defmac PUT_SDB_@dots{}
8874 Define these macros to override the assembler syntax for the special
8875 SDB assembler directives. See @file{sdbout.c} for a list of these
8876 macros and their arguments. If the standard syntax is used, you need
8877 not define them yourself.
8881 Some assemblers do not support a semicolon as a delimiter, even between
8882 SDB assembler directives. In that case, define this macro to be the
8883 delimiter to use (usually @samp{\n}). It is not necessary to define
8884 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8888 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
8889 Define this macro to allow references to unknown structure,
8890 union, or enumeration tags to be emitted. Standard COFF does not
8891 allow handling of unknown references, MIPS ECOFF has support for
8895 @defmac SDB_ALLOW_FORWARD_REFERENCES
8896 Define this macro to allow references to structure, union, or
8897 enumeration tags that have not yet been seen to be handled. Some
8898 assemblers choke if forward tags are used, while some require it.
8901 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
8902 A C statement to output SDB debugging information before code for line
8903 number @var{line} of the current source file to the stdio stream
8904 @var{stream}. The default is to emit an @code{.ln} directive.
8909 @subsection Macros for VMS Debug Format
8911 @c prevent bad page break with this line
8912 Here are macros for VMS debug format.
8914 @defmac VMS_DEBUGGING_INFO
8915 Define this macro if GCC should produce debugging output for VMS
8916 in response to the @option{-g} option. The default behavior for VMS
8917 is to generate minimal debug info for a traceback in the absence of
8918 @option{-g} unless explicitly overridden with @option{-g0}. This
8919 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8920 @code{OVERRIDE_OPTIONS}.
8923 @node Floating Point
8924 @section Cross Compilation and Floating Point
8925 @cindex cross compilation and floating point
8926 @cindex floating point and cross compilation
8928 While all modern machines use twos-complement representation for integers,
8929 there are a variety of representations for floating point numbers. This
8930 means that in a cross-compiler the representation of floating point numbers
8931 in the compiled program may be different from that used in the machine
8932 doing the compilation.
8934 Because different representation systems may offer different amounts of
8935 range and precision, all floating point constants must be represented in
8936 the target machine's format. Therefore, the cross compiler cannot
8937 safely use the host machine's floating point arithmetic; it must emulate
8938 the target's arithmetic. To ensure consistency, GCC always uses
8939 emulation to work with floating point values, even when the host and
8940 target floating point formats are identical.
8942 The following macros are provided by @file{real.h} for the compiler to
8943 use. All parts of the compiler which generate or optimize
8944 floating-point calculations must use these macros. They may evaluate
8945 their operands more than once, so operands must not have side effects.
8947 @defmac REAL_VALUE_TYPE
8948 The C data type to be used to hold a floating point value in the target
8949 machine's format. Typically this is a @code{struct} containing an
8950 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8954 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8955 Compares for equality the two values, @var{x} and @var{y}. If the target
8956 floating point format supports negative zeroes and/or NaNs,
8957 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8958 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8961 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8962 Tests whether @var{x} is less than @var{y}.
8965 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8966 Truncates @var{x} to a signed integer, rounding toward zero.
8969 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8970 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8971 @var{x} is negative, returns zero.
8974 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8975 Converts @var{string} into a floating point number in the target machine's
8976 representation for mode @var{mode}. This routine can handle both
8977 decimal and hexadecimal floating point constants, using the syntax
8978 defined by the C language for both.
8981 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8982 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8985 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8986 Determines whether @var{x} represents infinity (positive or negative).
8989 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8990 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8993 @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})
8994 Calculates an arithmetic operation on the two floating point values
8995 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8998 The operation to be performed is specified by @var{code}. Only the
8999 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
9000 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
9002 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
9003 target's floating point format cannot represent infinity, it will call
9004 @code{abort}. Callers should check for this situation first, using
9005 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
9008 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
9009 Returns the negative of the floating point value @var{x}.
9012 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
9013 Returns the absolute value of @var{x}.
9016 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
9017 Truncates the floating point value @var{x} to fit in @var{mode}. The
9018 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
9019 appropriate bit pattern to be output as a floating constant whose
9020 precision accords with mode @var{mode}.
9023 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
9024 Converts a floating point value @var{x} into a double-precision integer
9025 which is then stored into @var{low} and @var{high}. If the value is not
9026 integral, it is truncated.
9029 @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})
9030 Converts a double-precision integer found in @var{low} and @var{high},
9031 into a floating point value which is then stored into @var{x}. The
9032 value is truncated to fit in mode @var{mode}.
9035 @node Mode Switching
9036 @section Mode Switching Instructions
9037 @cindex mode switching
9038 The following macros control mode switching optimizations:
9040 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
9041 Define this macro if the port needs extra instructions inserted for mode
9042 switching in an optimizing compilation.
9044 For an example, the SH4 can perform both single and double precision
9045 floating point operations, but to perform a single precision operation,
9046 the FPSCR PR bit has to be cleared, while for a double precision
9047 operation, this bit has to be set. Changing the PR bit requires a general
9048 purpose register as a scratch register, hence these FPSCR sets have to
9049 be inserted before reload, i.e.@: you can't put this into instruction emitting
9050 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
9052 You can have multiple entities that are mode-switched, and select at run time
9053 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
9054 return nonzero for any @var{entity} that needs mode-switching.
9055 If you define this macro, you also have to define
9056 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
9057 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
9058 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
9062 @defmac NUM_MODES_FOR_MODE_SWITCHING
9063 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
9064 initializer for an array of integers. Each initializer element
9065 N refers to an entity that needs mode switching, and specifies the number
9066 of different modes that might need to be set for this entity.
9067 The position of the initializer in the initializer---starting counting at
9068 zero---determines the integer that is used to refer to the mode-switched
9070 In macros that take mode arguments / yield a mode result, modes are
9071 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
9072 switch is needed / supplied.
9075 @defmac MODE_NEEDED (@var{entity}, @var{insn})
9076 @var{entity} is an integer specifying a mode-switched entity. If
9077 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
9078 return an integer value not larger than the corresponding element in
9079 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
9080 be switched into prior to the execution of @var{insn}.
9083 @defmac MODE_AFTER (@var{mode}, @var{insn})
9084 If this macro is defined, it is evaluated for every @var{insn} during
9085 mode switching. It determines the mode that an insn results in (if
9086 different from the incoming mode).
9089 @defmac MODE_ENTRY (@var{entity})
9090 If this macro is defined, it is evaluated for every @var{entity} that needs
9091 mode switching. It should evaluate to an integer, which is a mode that
9092 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
9093 is defined then @code{MODE_EXIT} must be defined.
9096 @defmac MODE_EXIT (@var{entity})
9097 If this macro is defined, it is evaluated for every @var{entity} that needs
9098 mode switching. It should evaluate to an integer, which is a mode that
9099 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
9100 is defined then @code{MODE_ENTRY} must be defined.
9103 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
9104 This macro specifies the order in which modes for @var{entity} are processed.
9105 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
9106 lowest. The value of the macro should be an integer designating a mode
9107 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
9108 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
9109 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
9112 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
9113 Generate one or more insns to set @var{entity} to @var{mode}.
9114 @var{hard_reg_live} is the set of hard registers live at the point where
9115 the insn(s) are to be inserted.
9118 @node Target Attributes
9119 @section Defining target-specific uses of @code{__attribute__}
9120 @cindex target attributes
9121 @cindex machine attributes
9122 @cindex attributes, target-specific
9124 Target-specific attributes may be defined for functions, data and types.
9125 These are described using the following target hooks; they also need to
9126 be documented in @file{extend.texi}.
9128 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
9129 If defined, this target hook points to an array of @samp{struct
9130 attribute_spec} (defined in @file{tree.h}) specifying the machine
9131 specific attributes for this target and some of the restrictions on the
9132 entities to which these attributes are applied and the arguments they
9136 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
9137 If defined, this target hook is a function which returns zero if the attributes on
9138 @var{type1} and @var{type2} are incompatible, one if they are compatible,
9139 and two if they are nearly compatible (which causes a warning to be
9140 generated). If this is not defined, machine-specific attributes are
9141 supposed always to be compatible.
9144 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
9145 If defined, this target hook is a function which assigns default attributes to
9146 newly defined @var{type}.
9149 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
9150 Define this target hook if the merging of type attributes needs special
9151 handling. If defined, the result is a list of the combined
9152 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
9153 that @code{comptypes} has already been called and returned 1. This
9154 function may call @code{merge_attributes} to handle machine-independent
9158 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
9159 Define this target hook if the merging of decl attributes needs special
9160 handling. If defined, the result is a list of the combined
9161 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
9162 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
9163 when this is needed are when one attribute overrides another, or when an
9164 attribute is nullified by a subsequent definition. This function may
9165 call @code{merge_attributes} to handle machine-independent merging.
9167 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
9168 If the only target-specific handling you require is @samp{dllimport}
9169 for Microsoft Windows targets, you should define the macro
9170 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
9171 will then define a function called
9172 @code{merge_dllimport_decl_attributes} which can then be defined as
9173 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
9174 add @code{handle_dll_attribute} in the attribute table for your port
9175 to perform initial processing of the @samp{dllimport} and
9176 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
9177 @file{i386/i386.c}, for example.
9180 @deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (tree @var{decl})
9181 @var{decl} is a variable or function with @code{__attribute__((dllimport))}
9182 specified. Use this hook if the target needs to add extra validation
9183 checks to @code{handle_dll_attribute}.
9186 @defmac TARGET_DECLSPEC
9187 Define this macro to a nonzero value if you want to treat
9188 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
9189 default, this behavior is enabled only for targets that define
9190 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
9191 of @code{__declspec} is via a built-in macro, but you should not rely
9192 on this implementation detail.
9195 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
9196 Define this target hook if you want to be able to add attributes to a decl
9197 when it is being created. This is normally useful for back ends which
9198 wish to implement a pragma by using the attributes which correspond to
9199 the pragma's effect. The @var{node} argument is the decl which is being
9200 created. The @var{attr_ptr} argument is a pointer to the attribute list
9201 for this decl. The list itself should not be modified, since it may be
9202 shared with other decls, but attributes may be chained on the head of
9203 the list and @code{*@var{attr_ptr}} modified to point to the new
9204 attributes, or a copy of the list may be made if further changes are
9208 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
9210 This target hook returns @code{true} if it is ok to inline @var{fndecl}
9211 into the current function, despite its having target-specific
9212 attributes, @code{false} otherwise. By default, if a function has a
9213 target specific attribute attached to it, it will not be inlined.
9216 @deftypefn {Target Hook} bool TARGET_VALID_OPTION_ATTRIBUTE_P (tree @var{fndecl}, tree @var{name}, tree @var{args}, int @var{flags})
9217 This hook is called to parse the @code{attribute(option("..."))}, and
9218 it allows the function to set different target machine compile time
9219 options for the current function that might be different than the
9220 options specified on the command line. The hook should return
9221 @code{true} if the options are valid.
9223 The hook should set the @var{DECL_FUNCTION_SPECIFIC_TARGET} field in
9224 the function declaration to hold a pointer to a target specific
9225 @var{struct cl_target_option} structure.
9228 @deftypefn {Target Hook} void TARGET_OPTION_SAVE (struct cl_target_option *@var{ptr})
9229 This hook is called to save any additional target specific information
9230 in the @var{struct cl_target_option} structure for function specific
9232 @xref{Option file format}.
9235 @deftypefn {Target Hook} void TARGET_OPTION_RESTORE (struct cl_target_option *@var{ptr})
9236 This hook is called to restore any additional target specific
9237 information in the @var{struct cl_target_option} structure for
9238 function specific options.
9241 @deftypefn {Target Hook} void TARGET_OPTION_PRINT (struct cl_target_option *@var{ptr})
9242 This hook is called to print any additional target specific
9243 information in the @var{struct cl_target_option} structure for
9244 function specific options.
9247 @deftypefn {Target Hook} bool TARGET_OPTION_PRAGMA_PARSE (target @var{args})
9248 This target hook parses the options for @code{#pragma GCC option} to
9249 set the machine specific options for functions that occur later in the
9250 input stream. The options should be the same as handled by the
9251 @code{TARGET_VALID_OPTION_ATTRIBUTE_P} hook.
9254 @deftypefn {Target Hook} bool TARGET_CAN_INLINE_P (tree @var{caller}, tree @var{callee})
9255 This target hook returns @code{false} if the @var{caller} function
9256 cannot inline @var{callee}, based on target specific information. By
9257 default, inlining is not allowed if the callee function has function
9258 specific target options and the caller does not use the same options.
9262 @section Emulating TLS
9263 @cindex Emulated TLS
9265 For targets whose psABI does not provide Thread Local Storage via
9266 specific relocations and instruction sequences, an emulation layer is
9267 used. A set of target hooks allows this emulation layer to be
9268 configured for the requirements of a particular target. For instance
9269 the psABI may infact specify TLS support in terms of an emulation
9272 The emulation layer works by creating a control object for every TLS
9273 object. To access the TLS object, a lookup function is provided
9274 which, when given the address of the control object, will return the
9275 address of the current thread's instance of the TLS object.
9277 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_GET_ADDRESS
9278 Contains the name of the helper function that uses a TLS control
9279 object to locate a TLS instance. The default causes libgcc's
9280 emulated TLS helper function to be used.
9283 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_REGISTER_COMMON
9284 Contains the name of the helper function that should be used at
9285 program startup to register TLS objects that are implicitly
9286 initialized to zero. If this is @code{NULL}, all TLS objects will
9287 have explicit initializers. The default causes libgcc's emulated TLS
9288 registration function to be used.
9291 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_SECTION
9292 Contains the name of the section in which TLS control variables should
9293 be placed. The default of @code{NULL} allows these to be placed in
9297 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_SECTION
9298 Contains the name of the section in which TLS initializers should be
9299 placed. The default of @code{NULL} allows these to be placed in any
9303 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_PREFIX
9304 Contains the prefix to be prepended to TLS control variable names.
9305 The default of @code{NULL} uses a target-specific prefix.
9308 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_PREFIX
9309 Contains the prefix to be prepended to TLS initializer objects. The
9310 default of @code{NULL} uses a target-specific prefix.
9313 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_FIELDS (tree @var{type}, tree *@var{name})
9314 Specifies a function that generates the FIELD_DECLs for a TLS control
9315 object type. @var{type} is the RECORD_TYPE the fields are for and
9316 @var{name} should be filled with the structure tag, if the default of
9317 @code{__emutls_object} is unsuitable. The default creates a type suitable
9318 for libgcc's emulated TLS function.
9321 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_INIT (tree @var{var}, tree @var{decl}, tree @var{tmpl_addr})
9322 Specifies a function that generates the CONSTRUCTOR to initialize a
9323 TLS control object. @var{var} is the TLS control object, @var{decl}
9324 is the TLS object and @var{tmpl_addr} is the address of the
9325 initializer. The default initializes libgcc's emulated TLS control object.
9328 @deftypevr {Target Hook} {bool} TARGET_EMUTLS_VAR_ALIGN_FIXED
9329 Specifies whether the alignment of TLS control variable objects is
9330 fixed and should not be increased as some backends may do to optimize
9331 single objects. The default is false.
9334 @deftypevr {Target Hook} {bool} TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
9335 Specifies whether a DWARF @code{DW_OP_form_tls_address} location descriptor
9336 may be used to describe emulated TLS control objects.
9339 @node MIPS Coprocessors
9340 @section Defining coprocessor specifics for MIPS targets.
9341 @cindex MIPS coprocessor-definition macros
9343 The MIPS specification allows MIPS implementations to have as many as 4
9344 coprocessors, each with as many as 32 private registers. GCC supports
9345 accessing these registers and transferring values between the registers
9346 and memory using asm-ized variables. For example:
9349 register unsigned int cp0count asm ("c0r1");
9355 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
9356 names may be added as described below, or the default names may be
9357 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
9359 Coprocessor registers are assumed to be epilogue-used; sets to them will
9360 be preserved even if it does not appear that the register is used again
9361 later in the function.
9363 Another note: according to the MIPS spec, coprocessor 1 (if present) is
9364 the FPU@. One accesses COP1 registers through standard mips
9365 floating-point support; they are not included in this mechanism.
9367 There is one macro used in defining the MIPS coprocessor interface which
9368 you may want to override in subtargets; it is described below.
9370 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
9371 A comma-separated list (with leading comma) of pairs describing the
9372 alternate names of coprocessor registers. The format of each entry should be
9374 @{ @var{alternatename}, @var{register_number}@}
9380 @section Parameters for Precompiled Header Validity Checking
9381 @cindex parameters, precompiled headers
9383 @deftypefn {Target Hook} void *TARGET_GET_PCH_VALIDITY (size_t *@var{sz})
9384 This hook returns the data needed by @code{TARGET_PCH_VALID_P} and sets
9385 @samp{*@var{sz}} to the size of the data in bytes.
9388 @deftypefn {Target Hook} const char *TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz})
9389 This hook checks whether the options used to create a PCH file are
9390 compatible with the current settings. It returns @code{NULL}
9391 if so and a suitable error message if not. Error messages will
9392 be presented to the user and must be localized using @samp{_(@var{msg})}.
9394 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
9395 when the PCH file was created and @var{sz} is the size of that data in bytes.
9396 It's safe to assume that the data was created by the same version of the
9397 compiler, so no format checking is needed.
9399 The default definition of @code{default_pch_valid_p} should be
9400 suitable for most targets.
9403 @deftypefn {Target Hook} const char *TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags})
9404 If this hook is nonnull, the default implementation of
9405 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
9406 of @code{target_flags}. @var{pch_flags} specifies the value that
9407 @code{target_flags} had when the PCH file was created. The return
9408 value is the same as for @code{TARGET_PCH_VALID_P}.
9412 @section C++ ABI parameters
9413 @cindex parameters, c++ abi
9415 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
9416 Define this hook to override the integer type used for guard variables.
9417 These are used to implement one-time construction of static objects. The
9418 default is long_long_integer_type_node.
9421 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
9422 This hook determines how guard variables are used. It should return
9423 @code{false} (the default) if first byte should be used. A return value of
9424 @code{true} indicates the least significant bit should be used.
9427 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
9428 This hook returns the size of the cookie to use when allocating an array
9429 whose elements have the indicated @var{type}. Assumes that it is already
9430 known that a cookie is needed. The default is
9431 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
9432 IA64/Generic C++ ABI@.
9435 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
9436 This hook should return @code{true} if the element size should be stored in
9437 array cookies. The default is to return @code{false}.
9440 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
9441 If defined by a backend this hook allows the decision made to export
9442 class @var{type} to be overruled. Upon entry @var{import_export}
9443 will contain 1 if the class is going to be exported, @minus{}1 if it is going
9444 to be imported and 0 otherwise. This function should return the
9445 modified value and perform any other actions necessary to support the
9446 backend's targeted operating system.
9449 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
9450 This hook should return @code{true} if constructors and destructors return
9451 the address of the object created/destroyed. The default is to return
9455 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
9456 This hook returns true if the key method for a class (i.e., the method
9457 which, if defined in the current translation unit, causes the virtual
9458 table to be emitted) may be an inline function. Under the standard
9459 Itanium C++ ABI the key method may be an inline function so long as
9460 the function is not declared inline in the class definition. Under
9461 some variants of the ABI, an inline function can never be the key
9462 method. The default is to return @code{true}.
9465 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
9466 @var{decl} is a virtual table, virtual table table, typeinfo object,
9467 or other similar implicit class data object that will be emitted with
9468 external linkage in this translation unit. No ELF visibility has been
9469 explicitly specified. If the target needs to specify a visibility
9470 other than that of the containing class, use this hook to set
9471 @code{DECL_VISIBILITY} and @code{DECL_VISIBILITY_SPECIFIED}.
9474 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
9475 This hook returns true (the default) if virtual tables and other
9476 similar implicit class data objects are always COMDAT if they have
9477 external linkage. If this hook returns false, then class data for
9478 classes whose virtual table will be emitted in only one translation
9479 unit will not be COMDAT.
9482 @deftypefn {Target Hook} bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void)
9483 This hook returns true (the default) if the RTTI information for
9484 the basic types which is defined in the C++ runtime should always
9485 be COMDAT, false if it should not be COMDAT.
9488 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
9489 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
9490 should be used to register static destructors when @option{-fuse-cxa-atexit}
9491 is in effect. The default is to return false to use @code{__cxa_atexit}.
9494 @deftypefn {Target Hook} bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void)
9495 This hook returns true if the target @code{atexit} function can be used
9496 in the same manner as @code{__cxa_atexit} to register C++ static
9497 destructors. This requires that @code{atexit}-registered functions in
9498 shared libraries are run in the correct order when the libraries are
9499 unloaded. The default is to return false.
9502 @deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type})
9503 @var{type} is a C++ class (i.e., RECORD_TYPE or UNION_TYPE) that has just been
9504 defined. Use this hook to make adjustments to the class (eg, tweak
9505 visibility or perform any other required target modifications).
9509 @section Miscellaneous Parameters
9510 @cindex parameters, miscellaneous
9512 @c prevent bad page break with this line
9513 Here are several miscellaneous parameters.
9515 @defmac HAS_LONG_COND_BRANCH
9516 Define this boolean macro to indicate whether or not your architecture
9517 has conditional branches that can span all of memory. It is used in
9518 conjunction with an optimization that partitions hot and cold basic
9519 blocks into separate sections of the executable. If this macro is
9520 set to false, gcc will convert any conditional branches that attempt
9521 to cross between sections into unconditional branches or indirect jumps.
9524 @defmac HAS_LONG_UNCOND_BRANCH
9525 Define this boolean macro to indicate whether or not your architecture
9526 has unconditional branches that can span all of memory. It is used in
9527 conjunction with an optimization that partitions hot and cold basic
9528 blocks into separate sections of the executable. If this macro is
9529 set to false, gcc will convert any unconditional branches that attempt
9530 to cross between sections into indirect jumps.
9533 @defmac CASE_VECTOR_MODE
9534 An alias for a machine mode name. This is the machine mode that
9535 elements of a jump-table should have.
9538 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
9539 Optional: return the preferred mode for an @code{addr_diff_vec}
9540 when the minimum and maximum offset are known. If you define this,
9541 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
9542 To make this work, you also have to define @code{INSN_ALIGN} and
9543 make the alignment for @code{addr_diff_vec} explicit.
9544 The @var{body} argument is provided so that the offset_unsigned and scale
9545 flags can be updated.
9548 @defmac CASE_VECTOR_PC_RELATIVE
9549 Define this macro to be a C expression to indicate when jump-tables
9550 should contain relative addresses. You need not define this macro if
9551 jump-tables never contain relative addresses, or jump-tables should
9552 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
9556 @defmac CASE_VALUES_THRESHOLD
9557 Define this to be the smallest number of different values for which it
9558 is best to use a jump-table instead of a tree of conditional branches.
9559 The default is four for machines with a @code{casesi} instruction and
9560 five otherwise. This is best for most machines.
9563 @defmac CASE_USE_BIT_TESTS
9564 Define this macro to be a C expression to indicate whether C switch
9565 statements may be implemented by a sequence of bit tests. This is
9566 advantageous on processors that can efficiently implement left shift
9567 of 1 by the number of bits held in a register, but inappropriate on
9568 targets that would require a loop. By default, this macro returns
9569 @code{true} if the target defines an @code{ashlsi3} pattern, and
9570 @code{false} otherwise.
9573 @defmac WORD_REGISTER_OPERATIONS
9574 Define this macro if operations between registers with integral mode
9575 smaller than a word are always performed on the entire register.
9576 Most RISC machines have this property and most CISC machines do not.
9579 @defmac LOAD_EXTEND_OP (@var{mem_mode})
9580 Define this macro to be a C expression indicating when insns that read
9581 memory in @var{mem_mode}, an integral mode narrower than a word, set the
9582 bits outside of @var{mem_mode} to be either the sign-extension or the
9583 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
9584 of @var{mem_mode} for which the
9585 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
9586 @code{UNKNOWN} for other modes.
9588 This macro is not called with @var{mem_mode} non-integral or with a width
9589 greater than or equal to @code{BITS_PER_WORD}, so you may return any
9590 value in this case. Do not define this macro if it would always return
9591 @code{UNKNOWN}. On machines where this macro is defined, you will normally
9592 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
9594 You may return a non-@code{UNKNOWN} value even if for some hard registers
9595 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
9596 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
9597 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
9598 integral mode larger than this but not larger than @code{word_mode}.
9600 You must return @code{UNKNOWN} if for some hard registers that allow this
9601 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
9602 @code{word_mode}, but that they can change to another integral mode that
9603 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
9606 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
9607 Define this macro if loading short immediate values into registers sign
9611 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
9612 Define this macro if the same instructions that convert a floating
9613 point number to a signed fixed point number also convert validly to an
9617 @deftypefn {Target Hook} int TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (enum machine_mode @var{mode})
9618 When @option{-ffast-math} is in effect, GCC tries to optimize
9619 divisions by the same divisor, by turning them into multiplications by
9620 the reciprocal. This target hook specifies the minimum number of divisions
9621 that should be there for GCC to perform the optimization for a variable
9622 of mode @var{mode}. The default implementation returns 3 if the machine
9623 has an instruction for the division, and 2 if it does not.
9627 The maximum number of bytes that a single instruction can move quickly
9628 between memory and registers or between two memory locations.
9631 @defmac MAX_MOVE_MAX
9632 The maximum number of bytes that a single instruction can move quickly
9633 between memory and registers or between two memory locations. If this
9634 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
9635 constant value that is the largest value that @code{MOVE_MAX} can have
9639 @defmac SHIFT_COUNT_TRUNCATED
9640 A C expression that is nonzero if on this machine the number of bits
9641 actually used for the count of a shift operation is equal to the number
9642 of bits needed to represent the size of the object being shifted. When
9643 this macro is nonzero, the compiler will assume that it is safe to omit
9644 a sign-extend, zero-extend, and certain bitwise `and' instructions that
9645 truncates the count of a shift operation. On machines that have
9646 instructions that act on bit-fields at variable positions, which may
9647 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
9648 also enables deletion of truncations of the values that serve as
9649 arguments to bit-field instructions.
9651 If both types of instructions truncate the count (for shifts) and
9652 position (for bit-field operations), or if no variable-position bit-field
9653 instructions exist, you should define this macro.
9655 However, on some machines, such as the 80386 and the 680x0, truncation
9656 only applies to shift operations and not the (real or pretended)
9657 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
9658 such machines. Instead, add patterns to the @file{md} file that include
9659 the implied truncation of the shift instructions.
9661 You need not define this macro if it would always have the value of zero.
9664 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
9665 @deftypefn {Target Hook} int TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode})
9666 This function describes how the standard shift patterns for @var{mode}
9667 deal with shifts by negative amounts or by more than the width of the mode.
9668 @xref{shift patterns}.
9670 On many machines, the shift patterns will apply a mask @var{m} to the
9671 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
9672 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
9673 this is true for mode @var{mode}, the function should return @var{m},
9674 otherwise it should return 0. A return value of 0 indicates that no
9675 particular behavior is guaranteed.
9677 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
9678 @emph{not} apply to general shift rtxes; it applies only to instructions
9679 that are generated by the named shift patterns.
9681 The default implementation of this function returns
9682 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
9683 and 0 otherwise. This definition is always safe, but if
9684 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
9685 nevertheless truncate the shift count, you may get better code
9689 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
9690 A C expression which is nonzero if on this machine it is safe to
9691 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
9692 bits (where @var{outprec} is smaller than @var{inprec}) by merely
9693 operating on it as if it had only @var{outprec} bits.
9695 On many machines, this expression can be 1.
9697 @c rearranged this, removed the phrase "it is reported that". this was
9698 @c to fix an overfull hbox. --mew 10feb93
9699 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
9700 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
9701 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
9702 such cases may improve things.
9705 @deftypefn {Target Hook} int TARGET_MODE_REP_EXTENDED (enum machine_mode @var{mode}, enum machine_mode @var{rep_mode})
9706 The representation of an integral mode can be such that the values
9707 are always extended to a wider integral mode. Return
9708 @code{SIGN_EXTEND} if values of @var{mode} are represented in
9709 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
9710 otherwise. (Currently, none of the targets use zero-extended
9711 representation this way so unlike @code{LOAD_EXTEND_OP},
9712 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
9713 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
9714 @var{mode} to @var{mode_rep} so that @var{mode_rep} is not the next
9715 widest integral mode and currently we take advantage of this fact.)
9717 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
9718 value even if the extension is not performed on certain hard registers
9719 as long as for the @code{REGNO_REG_CLASS} of these hard registers
9720 @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero.
9722 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
9723 describe two related properties. If you define
9724 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
9725 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
9728 In order to enforce the representation of @code{mode},
9729 @code{TRULY_NOOP_TRUNCATION} should return false when truncating to
9733 @defmac STORE_FLAG_VALUE
9734 A C expression describing the value returned by a comparison operator
9735 with an integral mode and stored by a store-flag instruction
9736 (@samp{s@var{cond}}) when the condition is true. This description must
9737 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
9738 comparison operators whose results have a @code{MODE_INT} mode.
9740 A value of 1 or @minus{}1 means that the instruction implementing the
9741 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
9742 and 0 when the comparison is false. Otherwise, the value indicates
9743 which bits of the result are guaranteed to be 1 when the comparison is
9744 true. This value is interpreted in the mode of the comparison
9745 operation, which is given by the mode of the first operand in the
9746 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
9747 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
9750 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
9751 generate code that depends only on the specified bits. It can also
9752 replace comparison operators with equivalent operations if they cause
9753 the required bits to be set, even if the remaining bits are undefined.
9754 For example, on a machine whose comparison operators return an
9755 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
9756 @samp{0x80000000}, saying that just the sign bit is relevant, the
9760 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
9767 (ashift:SI @var{x} (const_int @var{n}))
9771 where @var{n} is the appropriate shift count to move the bit being
9772 tested into the sign bit.
9774 There is no way to describe a machine that always sets the low-order bit
9775 for a true value, but does not guarantee the value of any other bits,
9776 but we do not know of any machine that has such an instruction. If you
9777 are trying to port GCC to such a machine, include an instruction to
9778 perform a logical-and of the result with 1 in the pattern for the
9779 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
9781 Often, a machine will have multiple instructions that obtain a value
9782 from a comparison (or the condition codes). Here are rules to guide the
9783 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
9788 Use the shortest sequence that yields a valid definition for
9789 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
9790 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
9791 comparison operators to do so because there may be opportunities to
9792 combine the normalization with other operations.
9795 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
9796 slightly preferred on machines with expensive jumps and 1 preferred on
9800 As a second choice, choose a value of @samp{0x80000001} if instructions
9801 exist that set both the sign and low-order bits but do not define the
9805 Otherwise, use a value of @samp{0x80000000}.
9808 Many machines can produce both the value chosen for
9809 @code{STORE_FLAG_VALUE} and its negation in the same number of
9810 instructions. On those machines, you should also define a pattern for
9811 those cases, e.g., one matching
9814 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
9817 Some machines can also perform @code{and} or @code{plus} operations on
9818 condition code values with less instructions than the corresponding
9819 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
9820 machines, define the appropriate patterns. Use the names @code{incscc}
9821 and @code{decscc}, respectively, for the patterns which perform
9822 @code{plus} or @code{minus} operations on condition code values. See
9823 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
9824 find such instruction sequences on other machines.
9826 If this macro is not defined, the default value, 1, is used. You need
9827 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
9828 instructions, or if the value generated by these instructions is 1.
9831 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
9832 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
9833 returned when comparison operators with floating-point results are true.
9834 Define this macro on machines that have comparison operations that return
9835 floating-point values. If there are no such operations, do not define
9839 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
9840 A C expression that gives a rtx representing the nonzero true element
9841 for vector comparisons. The returned rtx should be valid for the inner
9842 mode of @var{mode} which is guaranteed to be a vector mode. Define
9843 this macro on machines that have vector comparison operations that
9844 return a vector result. If there are no such operations, do not define
9845 this macro. Typically, this macro is defined as @code{const1_rtx} or
9846 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
9847 the compiler optimizing such vector comparison operations for the
9851 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
9852 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
9853 A C expression that indicates whether the architecture defines a value
9854 for @code{clz} or @code{ctz} with a zero operand.
9855 A result of @code{0} indicates the value is undefined.
9856 If the value is defined for only the RTL expression, the macro should
9857 evaluate to @code{1}; if the value applies also to the corresponding optab
9858 entry (which is normally the case if it expands directly into
9859 the corresponding RTL), then the macro should evaluate to @code{2}.
9860 In the cases where the value is defined, @var{value} should be set to
9863 If this macro is not defined, the value of @code{clz} or
9864 @code{ctz} at zero is assumed to be undefined.
9866 This macro must be defined if the target's expansion for @code{ffs}
9867 relies on a particular value to get correct results. Otherwise it
9868 is not necessary, though it may be used to optimize some corner cases, and
9869 to provide a default expansion for the @code{ffs} optab.
9871 Note that regardless of this macro the ``definedness'' of @code{clz}
9872 and @code{ctz} at zero do @emph{not} extend to the builtin functions
9873 visible to the user. Thus one may be free to adjust the value at will
9874 to match the target expansion of these operations without fear of
9879 An alias for the machine mode for pointers. On most machines, define
9880 this to be the integer mode corresponding to the width of a hardware
9881 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
9882 On some machines you must define this to be one of the partial integer
9883 modes, such as @code{PSImode}.
9885 The width of @code{Pmode} must be at least as large as the value of
9886 @code{POINTER_SIZE}. If it is not equal, you must define the macro
9887 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
9891 @defmac FUNCTION_MODE
9892 An alias for the machine mode used for memory references to functions
9893 being called, in @code{call} RTL expressions. On most CISC machines,
9894 where an instruction can begin at any byte address, this should be
9895 @code{QImode}. On most RISC machines, where all instructions have fixed
9896 size and alignment, this should be a mode with the same size and alignment
9897 as the machine instruction words - typically @code{SImode} or @code{HImode}.
9900 @defmac STDC_0_IN_SYSTEM_HEADERS
9901 In normal operation, the preprocessor expands @code{__STDC__} to the
9902 constant 1, to signify that GCC conforms to ISO Standard C@. On some
9903 hosts, like Solaris, the system compiler uses a different convention,
9904 where @code{__STDC__} is normally 0, but is 1 if the user specifies
9905 strict conformance to the C Standard.
9907 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
9908 convention when processing system header files, but when processing user
9909 files @code{__STDC__} will always expand to 1.
9912 @defmac NO_IMPLICIT_EXTERN_C
9913 Define this macro if the system header files support C++ as well as C@.
9914 This macro inhibits the usual method of using system header files in
9915 C++, which is to pretend that the file's contents are enclosed in
9916 @samp{extern "C" @{@dots{}@}}.
9921 @defmac REGISTER_TARGET_PRAGMAS ()
9922 Define this macro if you want to implement any target-specific pragmas.
9923 If defined, it is a C expression which makes a series of calls to
9924 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
9925 for each pragma. The macro may also do any
9926 setup required for the pragmas.
9928 The primary reason to define this macro is to provide compatibility with
9929 other compilers for the same target. In general, we discourage
9930 definition of target-specific pragmas for GCC@.
9932 If the pragma can be implemented by attributes then you should consider
9933 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
9935 Preprocessor macros that appear on pragma lines are not expanded. All
9936 @samp{#pragma} directives that do not match any registered pragma are
9937 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
9940 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9941 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9943 Each call to @code{c_register_pragma} or
9944 @code{c_register_pragma_with_expansion} establishes one pragma. The
9945 @var{callback} routine will be called when the preprocessor encounters a
9949 #pragma [@var{space}] @var{name} @dots{}
9952 @var{space} is the case-sensitive namespace of the pragma, or
9953 @code{NULL} to put the pragma in the global namespace. The callback
9954 routine receives @var{pfile} as its first argument, which can be passed
9955 on to cpplib's functions if necessary. You can lex tokens after the
9956 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
9957 callback will be silently ignored. The end of the line is indicated by
9958 a token of type @code{CPP_EOF}. Macro expansion occurs on the
9959 arguments of pragmas registered with
9960 @code{c_register_pragma_with_expansion} but not on the arguments of
9961 pragmas registered with @code{c_register_pragma}.
9963 Note that the use of @code{pragma_lex} is specific to the C and C++
9964 compilers. It will not work in the Java or Fortran compilers, or any
9965 other language compilers for that matter. Thus if @code{pragma_lex} is going
9966 to be called from target-specific code, it must only be done so when
9967 building the C and C++ compilers. This can be done by defining the
9968 variables @code{c_target_objs} and @code{cxx_target_objs} in the
9969 target entry in the @file{config.gcc} file. These variables should name
9970 the target-specific, language-specific object file which contains the
9971 code that uses @code{pragma_lex}. Note it will also be necessary to add a
9972 rule to the makefile fragment pointed to by @code{tmake_file} that shows
9973 how to build this object file.
9978 @defmac HANDLE_SYSV_PRAGMA
9979 Define this macro (to a value of 1) if you want the System V style
9980 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
9981 [=<value>]} to be supported by gcc.
9983 The pack pragma specifies the maximum alignment (in bytes) of fields
9984 within a structure, in much the same way as the @samp{__aligned__} and
9985 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
9986 the behavior to the default.
9988 A subtlety for Microsoft Visual C/C++ style bit-field packing
9989 (e.g.@: -mms-bitfields) for targets that support it:
9990 When a bit-field is inserted into a packed record, the whole size
9991 of the underlying type is used by one or more same-size adjacent
9992 bit-fields (that is, if its long:3, 32 bits is used in the record,
9993 and any additional adjacent long bit-fields are packed into the same
9994 chunk of 32 bits. However, if the size changes, a new field of that
9997 If both MS bit-fields and @samp{__attribute__((packed))} are used,
9998 the latter will take precedence. If @samp{__attribute__((packed))} is
9999 used on a single field when MS bit-fields are in use, it will take
10000 precedence for that field, but the alignment of the rest of the structure
10001 may affect its placement.
10003 The weak pragma only works if @code{SUPPORTS_WEAK} and
10004 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
10005 of specifically named weak labels, optionally with a value.
10010 @defmac HANDLE_PRAGMA_PACK_PUSH_POP
10011 Define this macro (to a value of 1) if you want to support the Win32
10012 style pragmas @samp{#pragma pack(push[,@var{n}])} and @samp{#pragma
10013 pack(pop)}. The @samp{pack(push,[@var{n}])} pragma specifies the maximum
10014 alignment (in bytes) of fields within a structure, in much the same way as
10015 the @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
10016 pack value of zero resets the behavior to the default. Successive
10017 invocations of this pragma cause the previous values to be stacked, so
10018 that invocations of @samp{#pragma pack(pop)} will return to the previous
10022 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
10023 Define this macro, as well as
10024 @code{HANDLE_SYSV_PRAGMA}, if macros should be expanded in the
10025 arguments of @samp{#pragma pack}.
10028 @defmac TARGET_DEFAULT_PACK_STRUCT
10029 If your target requires a structure packing default other than 0 (meaning
10030 the machine default), define this macro to the necessary value (in bytes).
10031 This must be a value that would also be valid to use with
10032 @samp{#pragma pack()} (that is, a small power of two).
10037 @defmac HANDLE_PRAGMA_PUSH_POP_MACRO
10038 Define this macro if you want to support the Win32 style pragmas
10039 @samp{#pragma push_macro(macro-name-as-string)} and @samp{#pragma
10040 pop_macro(macro-name-as-string)}. The @samp{#pragma push_macro(
10041 macro-name-as-string)} pragma saves the named macro and via
10042 @samp{#pragma pop_macro(macro-name-as-string)} it will return to the
10047 @defmac DOLLARS_IN_IDENTIFIERS
10048 Define this macro to control use of the character @samp{$} in
10049 identifier names for the C family of languages. 0 means @samp{$} is
10050 not allowed by default; 1 means it is allowed. 1 is the default;
10051 there is no need to define this macro in that case.
10054 @defmac NO_DOLLAR_IN_LABEL
10055 Define this macro if the assembler does not accept the character
10056 @samp{$} in label names. By default constructors and destructors in
10057 G++ have @samp{$} in the identifiers. If this macro is defined,
10058 @samp{.} is used instead.
10061 @defmac NO_DOT_IN_LABEL
10062 Define this macro if the assembler does not accept the character
10063 @samp{.} in label names. By default constructors and destructors in G++
10064 have names that use @samp{.}. If this macro is defined, these names
10065 are rewritten to avoid @samp{.}.
10068 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
10069 Define this macro as a C expression that is nonzero if it is safe for the
10070 delay slot scheduler to place instructions in the delay slot of @var{insn},
10071 even if they appear to use a resource set or clobbered in @var{insn}.
10072 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
10073 every @code{call_insn} has this behavior. On machines where some @code{insn}
10074 or @code{jump_insn} is really a function call and hence has this behavior,
10075 you should define this macro.
10077 You need not define this macro if it would always return zero.
10080 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
10081 Define this macro as a C expression that is nonzero if it is safe for the
10082 delay slot scheduler to place instructions in the delay slot of @var{insn},
10083 even if they appear to set or clobber a resource referenced in @var{insn}.
10084 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
10085 some @code{insn} or @code{jump_insn} is really a function call and its operands
10086 are registers whose use is actually in the subroutine it calls, you should
10087 define this macro. Doing so allows the delay slot scheduler to move
10088 instructions which copy arguments into the argument registers into the delay
10089 slot of @var{insn}.
10091 You need not define this macro if it would always return zero.
10094 @defmac MULTIPLE_SYMBOL_SPACES
10095 Define this macro as a C expression that is nonzero if, in some cases,
10096 global symbols from one translation unit may not be bound to undefined
10097 symbols in another translation unit without user intervention. For
10098 instance, under Microsoft Windows symbols must be explicitly imported
10099 from shared libraries (DLLs).
10101 You need not define this macro if it would always evaluate to zero.
10104 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{outputs}, tree @var{inputs}, tree @var{clobbers})
10105 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
10106 any hard regs the port wishes to automatically clobber for an asm.
10107 It should return the result of the last @code{tree_cons} used to add a
10108 clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the
10109 corresponding parameters to the asm and may be inspected to avoid
10110 clobbering a register that is an input or output of the asm. You can use
10111 @code{tree_overlaps_hard_reg_set}, declared in @file{tree.h}, to test
10112 for overlap with regards to asm-declared registers.
10115 @defmac MATH_LIBRARY
10116 Define this macro as a C string constant for the linker argument to link
10117 in the system math library, or @samp{""} if the target does not have a
10118 separate math library.
10120 You need only define this macro if the default of @samp{"-lm"} is wrong.
10123 @defmac LIBRARY_PATH_ENV
10124 Define this macro as a C string constant for the environment variable that
10125 specifies where the linker should look for libraries.
10127 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
10131 @defmac TARGET_POSIX_IO
10132 Define this macro if the target supports the following POSIX@ file
10133 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
10134 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
10135 to use file locking when exiting a program, which avoids race conditions
10136 if the program has forked. It will also create directories at run-time
10137 for cross-profiling.
10140 @defmac MAX_CONDITIONAL_EXECUTE
10142 A C expression for the maximum number of instructions to execute via
10143 conditional execution instructions instead of a branch. A value of
10144 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
10145 1 if it does use cc0.
10148 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
10149 Used if the target needs to perform machine-dependent modifications on the
10150 conditionals used for turning basic blocks into conditionally executed code.
10151 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
10152 contains information about the currently processed blocks. @var{true_expr}
10153 and @var{false_expr} are the tests that are used for converting the
10154 then-block and the else-block, respectively. Set either @var{true_expr} or
10155 @var{false_expr} to a null pointer if the tests cannot be converted.
10158 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
10159 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
10160 if-statements into conditions combined by @code{and} and @code{or} operations.
10161 @var{bb} contains the basic block that contains the test that is currently
10162 being processed and about to be turned into a condition.
10165 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
10166 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
10167 be converted to conditional execution format. @var{ce_info} points to
10168 a data structure, @code{struct ce_if_block}, which contains information
10169 about the currently processed blocks.
10172 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
10173 A C expression to perform any final machine dependent modifications in
10174 converting code to conditional execution. The involved basic blocks
10175 can be found in the @code{struct ce_if_block} structure that is pointed
10176 to by @var{ce_info}.
10179 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
10180 A C expression to cancel any machine dependent modifications in
10181 converting code to conditional execution. The involved basic blocks
10182 can be found in the @code{struct ce_if_block} structure that is pointed
10183 to by @var{ce_info}.
10186 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
10187 A C expression to initialize any extra fields in a @code{struct ce_if_block}
10188 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
10191 @defmac IFCVT_EXTRA_FIELDS
10192 If defined, it should expand to a set of field declarations that will be
10193 added to the @code{struct ce_if_block} structure. These should be initialized
10194 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
10197 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG ()
10198 If non-null, this hook performs a target-specific pass over the
10199 instruction stream. The compiler will run it at all optimization levels,
10200 just before the point at which it normally does delayed-branch scheduling.
10202 The exact purpose of the hook varies from target to target. Some use
10203 it to do transformations that are necessary for correctness, such as
10204 laying out in-function constant pools or avoiding hardware hazards.
10205 Others use it as an opportunity to do some machine-dependent optimizations.
10207 You need not implement the hook if it has nothing to do. The default
10208 definition is null.
10211 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
10212 Define this hook if you have any machine-specific built-in functions
10213 that need to be defined. It should be a function that performs the
10216 Machine specific built-in functions can be useful to expand special machine
10217 instructions that would otherwise not normally be generated because
10218 they have no equivalent in the source language (for example, SIMD vector
10219 instructions or prefetch instructions).
10221 To create a built-in function, call the function
10222 @code{lang_hooks.builtin_function}
10223 which is defined by the language front end. You can use any type nodes set
10224 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
10225 only language front ends that use those two functions will call
10226 @samp{TARGET_INIT_BUILTINS}.
10229 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
10231 Expand a call to a machine specific built-in function that was set up by
10232 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
10233 function call; the result should go to @var{target} if that is
10234 convenient, and have mode @var{mode} if that is convenient.
10235 @var{subtarget} may be used as the target for computing one of
10236 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
10237 ignored. This function should return the result of the call to the
10241 @deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (tree @var{fndecl}, tree @var{arglist})
10243 Select a replacement for a machine specific built-in function that
10244 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
10245 @emph{before} regular type checking, and so allows the target to
10246 implement a crude form of function overloading. @var{fndecl} is the
10247 declaration of the built-in function. @var{arglist} is the list of
10248 arguments passed to the built-in function. The result is a
10249 complete expression that implements the operation, usually
10250 another @code{CALL_EXPR}.
10253 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, tree @var{arglist}, bool @var{ignore})
10255 Fold a call to a machine specific built-in function that was set up by
10256 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
10257 built-in function. @var{arglist} is the list of arguments passed to
10258 the built-in function. The result is another tree containing a
10259 simplified expression for the call's result. If @var{ignore} is true
10260 the value will be ignored.
10263 @deftypefn {Target Hook} const char * TARGET_INVALID_WITHIN_DOLOOP (rtx @var{insn})
10265 Take an instruction in @var{insn} and return NULL if it is valid within a
10266 low-overhead loop, otherwise return a string why doloop could not be applied.
10268 Many targets use special registers for low-overhead looping. For any
10269 instruction that clobbers these this function should return a string indicating
10270 the reason why the doloop could not be applied.
10271 By default, the RTL loop optimizer does not use a present doloop pattern for
10272 loops containing function calls or branch on table instructions.
10275 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
10277 Take a branch insn in @var{branch1} and another in @var{branch2}.
10278 Return true if redirecting @var{branch1} to the destination of
10279 @var{branch2} is possible.
10281 On some targets, branches may have a limited range. Optimizing the
10282 filling of delay slots can result in branches being redirected, and this
10283 may in turn cause a branch offset to overflow.
10286 @deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (rtx @var{x}, @var{outer_code})
10287 This target hook returns @code{true} if @var{x} is considered to be commutative.
10288 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
10289 PLUS to be commutative inside a MEM@. @var{outer_code} is the rtx code
10290 of the enclosing rtl, if known, otherwise it is UNKNOWN.
10293 @deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg})
10295 When the initial value of a hard register has been copied in a pseudo
10296 register, it is often not necessary to actually allocate another register
10297 to this pseudo register, because the original hard register or a stack slot
10298 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
10299 is called at the start of register allocation once for each hard register
10300 that had its initial value copied by using
10301 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
10302 Possible values are @code{NULL_RTX}, if you don't want
10303 to do any special allocation, a @code{REG} rtx---that would typically be
10304 the hard register itself, if it is known not to be clobbered---or a
10306 If you are returning a @code{MEM}, this is only a hint for the allocator;
10307 it might decide to use another register anyways.
10308 You may use @code{current_function_leaf_function} in the hook, functions
10309 that use @code{REG_N_SETS}, to determine if the hard
10310 register in question will not be clobbered.
10311 The default value of this hook is @code{NULL}, which disables any special
10315 @deftypefn {Target Hook} int TARGET_UNSPEC_MAY_TRAP_P (const_rtx @var{x}, unsigned @var{flags})
10316 This target hook returns nonzero if @var{x}, an @code{unspec} or
10317 @code{unspec_volatile} operation, might cause a trap. Targets can use
10318 this hook to enhance precision of analysis for @code{unspec} and
10319 @code{unspec_volatile} operations. You may call @code{may_trap_p_1}
10320 to analyze inner elements of @var{x} in which case @var{flags} should be
10324 @deftypefn {Target Hook} void TARGET_SET_CURRENT_FUNCTION (tree @var{decl})
10325 The compiler invokes this hook whenever it changes its current function
10326 context (@code{cfun}). You can define this function if
10327 the back end needs to perform any initialization or reset actions on a
10328 per-function basis. For example, it may be used to implement function
10329 attributes that affect register usage or code generation patterns.
10330 The argument @var{decl} is the declaration for the new function context,
10331 and may be null to indicate that the compiler has left a function context
10332 and is returning to processing at the top level.
10333 The default hook function does nothing.
10335 GCC sets @code{cfun} to a dummy function context during initialization of
10336 some parts of the back end. The hook function is not invoked in this
10337 situation; you need not worry about the hook being invoked recursively,
10338 or when the back end is in a partially-initialized state.
10341 @defmac TARGET_OBJECT_SUFFIX
10342 Define this macro to be a C string representing the suffix for object
10343 files on your target machine. If you do not define this macro, GCC will
10344 use @samp{.o} as the suffix for object files.
10347 @defmac TARGET_EXECUTABLE_SUFFIX
10348 Define this macro to be a C string representing the suffix to be
10349 automatically added to executable files on your target machine. If you
10350 do not define this macro, GCC will use the null string as the suffix for
10354 @defmac COLLECT_EXPORT_LIST
10355 If defined, @code{collect2} will scan the individual object files
10356 specified on its command line and create an export list for the linker.
10357 Define this macro for systems like AIX, where the linker discards
10358 object files that are not referenced from @code{main} and uses export
10362 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
10363 Define this macro to a C expression representing a variant of the
10364 method call @var{mdecl}, if Java Native Interface (JNI) methods
10365 must be invoked differently from other methods on your target.
10366 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
10367 the @code{stdcall} calling convention and this macro is then
10368 defined as this expression:
10371 build_type_attribute_variant (@var{mdecl},
10373 (get_identifier ("stdcall"),
10378 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
10379 This target hook returns @code{true} past the point in which new jump
10380 instructions could be created. On machines that require a register for
10381 every jump such as the SHmedia ISA of SH5, this point would typically be
10382 reload, so this target hook should be defined to a function such as:
10386 cannot_modify_jumps_past_reload_p ()
10388 return (reload_completed || reload_in_progress);
10393 @deftypefn {Target Hook} int TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
10394 This target hook returns a register class for which branch target register
10395 optimizations should be applied. All registers in this class should be
10396 usable interchangeably. After reload, registers in this class will be
10397 re-allocated and loads will be hoisted out of loops and be subjected
10398 to inter-block scheduling.
10401 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
10402 Branch target register optimization will by default exclude callee-saved
10404 that are not already live during the current function; if this target hook
10405 returns true, they will be included. The target code must than make sure
10406 that all target registers in the class returned by
10407 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
10408 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
10409 epilogues have already been generated. Note, even if you only return
10410 true when @var{after_prologue_epilogue_gen} is false, you still are likely
10411 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
10412 to reserve space for caller-saved target registers.
10415 @defmac POWI_MAX_MULTS
10416 If defined, this macro is interpreted as a signed integer C expression
10417 that specifies the maximum number of floating point multiplications
10418 that should be emitted when expanding exponentiation by an integer
10419 constant inline. When this value is defined, exponentiation requiring
10420 more than this number of multiplications is implemented by calling the
10421 system library's @code{pow}, @code{powf} or @code{powl} routines.
10422 The default value places no upper bound on the multiplication count.
10425 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
10426 This target hook should register any extra include files for the
10427 target. The parameter @var{stdinc} indicates if normal include files
10428 are present. The parameter @var{sysroot} is the system root directory.
10429 The parameter @var{iprefix} is the prefix for the gcc directory.
10432 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
10433 This target hook should register any extra include files for the
10434 target before any standard headers. The parameter @var{stdinc}
10435 indicates if normal include files are present. The parameter
10436 @var{sysroot} is the system root directory. The parameter
10437 @var{iprefix} is the prefix for the gcc directory.
10440 @deftypefn Macro void TARGET_OPTF (char *@var{path})
10441 This target hook should register special include paths for the target.
10442 The parameter @var{path} is the include to register. On Darwin
10443 systems, this is used for Framework includes, which have semantics
10444 that are different from @option{-I}.
10447 @deftypefn {Target Hook} bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
10448 This target hook returns @code{true} if it is safe to use a local alias
10449 for a virtual function @var{fndecl} when constructing thunks,
10450 @code{false} otherwise. By default, the hook returns @code{true} for all
10451 functions, if a target supports aliases (i.e.@: defines
10452 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
10455 @defmac TARGET_FORMAT_TYPES
10456 If defined, this macro is the name of a global variable containing
10457 target-specific format checking information for the @option{-Wformat}
10458 option. The default is to have no target-specific format checks.
10461 @defmac TARGET_N_FORMAT_TYPES
10462 If defined, this macro is the number of entries in
10463 @code{TARGET_FORMAT_TYPES}.
10466 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
10467 If defined, this macro is the name of a global variable containing
10468 target-specific format overrides for the @option{-Wformat} option. The
10469 default is to have no target-specific format overrides. If defined,
10470 @code{TARGET_FORMAT_TYPES} must be defined, too.
10473 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
10474 If defined, this macro specifies the number of entries in
10475 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
10478 @deftypefn {Target Hook} bool TARGET_RELAXED_ORDERING
10479 If set to @code{true}, means that the target's memory model does not
10480 guarantee that loads which do not depend on one another will access
10481 main memory in the order of the instruction stream; if ordering is
10482 important, an explicit memory barrier must be used. This is true of
10483 many recent processors which implement a policy of ``relaxed,''
10484 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
10485 and ia64. The default is @code{false}.
10488 @deftypefn {Target Hook} const char *TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (tree @var{typelist}, tree @var{funcdecl}, tree @var{val})
10489 If defined, this macro returns the diagnostic message when it is
10490 illegal to pass argument @var{val} to function @var{funcdecl}
10491 with prototype @var{typelist}.
10494 @deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (tree @var{fromtype}, tree @var{totype})
10495 If defined, this macro returns the diagnostic message when it is
10496 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
10497 if validity should be determined by the front end.
10500 @deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, tree @var{type})
10501 If defined, this macro returns the diagnostic message when it is
10502 invalid to apply operation @var{op} (where unary plus is denoted by
10503 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
10504 if validity should be determined by the front end.
10507 @deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, tree @var{type1}, tree @var{type2})
10508 If defined, this macro returns the diagnostic message when it is
10509 invalid to apply operation @var{op} to operands of types @var{type1}
10510 and @var{type2}, or @code{NULL} if validity should be determined by
10514 @defmac TARGET_USE_JCR_SECTION
10515 This macro determines whether to use the JCR section to register Java
10516 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
10517 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.
10521 This macro determines the size of the objective C jump buffer for the
10522 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
10525 @defmac LIBGCC2_UNWIND_ATTRIBUTE
10526 Define this macro if any target-specific attributes need to be attached
10527 to the functions in @file{libgcc} that provide low-level support for
10528 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
10529 and the associated definitions of those functions.
10532 @deftypefn {Target Hook} void TARGET_UPDATE_STACK_BOUNDARY (void)
10533 Define this macro to update the current function stack boundary if
10537 @deftypefn {Target Hook} rtx TARGET_GET_DRAP_RTX (void)
10538 Define this macro to an rtx for Dynamic Realign Argument Pointer if a
10539 different argument pointer register is needed to access the function's
10540 argument list when stack is aligned.
10543 @deftypefn {Target Hook} {bool} TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS (void)
10544 When optimization is disabled, this hook indicates whether or not
10545 arguments should be allocated to stack slots. Normally, GCC allocates
10546 stacks slots for arguments when not optimizing in order to make
10547 debugging easier. However, when a function is declared with
10548 @code{__attribute__((naked))}, there is no stack frame, and the compiler
10549 cannot safely move arguments from the registers in which they are passed
10550 to the stack. Therefore, this hook should return true in general, but
10551 false for naked functions. The default implementation always returns true.