1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,
2 @c 2002, 2003, 2004, 2005, 2006, 2007 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
7 @chapter Target Description Macros and Functions
8 @cindex machine description macros
9 @cindex target description macros
10 @cindex macros, target description
11 @cindex @file{tm.h} macros
13 In addition to the file @file{@var{machine}.md}, a machine description
14 includes a C header file conventionally given the name
15 @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}.
16 The header file defines numerous macros that convey the information
17 about the target machine that does not fit into the scheme of the
18 @file{.md} file. The file @file{tm.h} should be a link to
19 @file{@var{machine}.h}. The header file @file{config.h} includes
20 @file{tm.h} and most compiler source files include @file{config.h}. The
21 source file defines a variable @code{targetm}, which is a structure
22 containing pointers to functions and data relating to the target
23 machine. @file{@var{machine}.c} should also contain their definitions,
24 if they are not defined elsewhere in GCC, and other functions called
25 through the macros defined in the @file{.h} file.
28 * Target Structure:: The @code{targetm} variable.
29 * Driver:: Controlling how the driver runs the compilation passes.
30 * Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}.
31 * Per-Function Data:: Defining data structures for per-function information.
32 * Storage Layout:: Defining sizes and alignments of data.
33 * Type Layout:: Defining sizes and properties of basic user data types.
34 * Registers:: Naming and describing the hardware registers.
35 * Register Classes:: Defining the classes of hardware registers.
36 * Old Constraints:: The old way to define machine-specific constraints.
37 * Stack and Calling:: Defining which way the stack grows and by how much.
38 * Varargs:: Defining the varargs macros.
39 * Trampolines:: Code set up at run time to enter a nested function.
40 * Library Calls:: Controlling how library routines are implicitly called.
41 * Addressing Modes:: Defining addressing modes valid for memory operands.
42 * Anchored Addresses:: Defining how @option{-fsection-anchors} should work.
43 * Condition Code:: Defining how insns update the condition code.
44 * Costs:: Defining relative costs of different operations.
45 * Scheduling:: Adjusting the behavior of the instruction scheduler.
46 * Sections:: Dividing storage into text, data, and other sections.
47 * PIC:: Macros for position independent code.
48 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
49 * Debugging Info:: Defining the format of debugging output.
50 * Floating Point:: Handling floating point for cross-compilers.
51 * Mode Switching:: Insertion of mode-switching instructions.
52 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
53 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
54 * PCH Target:: Validity checking for precompiled headers.
55 * C++ ABI:: Controlling C++ ABI changes.
56 * Misc:: Everything else.
59 @node Target Structure
60 @section The Global @code{targetm} Variable
62 @cindex target functions
64 @deftypevar {struct gcc_target} targetm
65 The target @file{.c} file must define the global @code{targetm} variable
66 which contains pointers to functions and data relating to the target
67 machine. The variable is declared in @file{target.h};
68 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
69 used to initialize the variable, and macros for the default initializers
70 for elements of the structure. The @file{.c} file should override those
71 macros for which the default definition is inappropriate. For example:
74 #include "target-def.h"
76 /* @r{Initialize the GCC target structure.} */
78 #undef TARGET_COMP_TYPE_ATTRIBUTES
79 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
81 struct gcc_target targetm = TARGET_INITIALIZER;
85 Where a macro should be defined in the @file{.c} file in this manner to
86 form part of the @code{targetm} structure, it is documented below as a
87 ``Target Hook'' with a prototype. Many macros will change in future
88 from being defined in the @file{.h} file to being part of the
89 @code{targetm} structure.
92 @section Controlling the Compilation Driver, @file{gcc}
94 @cindex controlling the compilation driver
96 @c prevent bad page break with this line
97 You can control the compilation driver.
99 @defmac SWITCH_TAKES_ARG (@var{char})
100 A C expression which determines whether the option @option{-@var{char}}
101 takes arguments. The value should be the number of arguments that
102 option takes--zero, for many options.
104 By default, this macro is defined as
105 @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options
106 properly. You need not define @code{SWITCH_TAKES_ARG} unless you
107 wish to add additional options which take arguments. Any redefinition
108 should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for
112 @defmac WORD_SWITCH_TAKES_ARG (@var{name})
113 A C expression which determines whether the option @option{-@var{name}}
114 takes arguments. The value should be the number of arguments that
115 option takes--zero, for many options. This macro rather than
116 @code{SWITCH_TAKES_ARG} is used for multi-character option names.
118 By default, this macro is defined as
119 @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
120 properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
121 wish to add additional options which take arguments. Any redefinition
122 should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
126 @defmac SWITCH_CURTAILS_COMPILATION (@var{char})
127 A C expression which determines whether the option @option{-@var{char}}
128 stops compilation before the generation of an executable. The value is
129 boolean, nonzero if the option does stop an executable from being
130 generated, zero otherwise.
132 By default, this macro is defined as
133 @code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard
134 options properly. You need not define
135 @code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional
136 options which affect the generation of an executable. Any redefinition
137 should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check
138 for additional options.
141 @defmac SWITCHES_NEED_SPACES
142 A string-valued C expression which enumerates the options for which
143 the linker needs a space between the option and its argument.
145 If this macro is not defined, the default value is @code{""}.
148 @defmac TARGET_OPTION_TRANSLATE_TABLE
149 If defined, a list of pairs of strings, the first of which is a
150 potential command line target to the @file{gcc} driver program, and the
151 second of which is a space-separated (tabs and other whitespace are not
152 supported) list of options with which to replace the first option. The
153 target defining this list is responsible for assuring that the results
154 are valid. Replacement options may not be the @code{--opt} style, they
155 must be the @code{-opt} style. It is the intention of this macro to
156 provide a mechanism for substitution that affects the multilibs chosen,
157 such as one option that enables many options, some of which select
158 multilibs. Example nonsensical definition, where @option{-malt-abi},
159 @option{-EB}, and @option{-mspoo} cause different multilibs to be chosen:
162 #define TARGET_OPTION_TRANSLATE_TABLE \
163 @{ "-fast", "-march=fast-foo -malt-abi -I/usr/fast-foo" @}, \
164 @{ "-compat", "-EB -malign=4 -mspoo" @}
168 @defmac DRIVER_SELF_SPECS
169 A list of specs for the driver itself. It should be a suitable
170 initializer for an array of strings, with no surrounding braces.
172 The driver applies these specs to its own command line between loading
173 default @file{specs} files (but not command-line specified ones) and
174 choosing the multilib directory or running any subcommands. It
175 applies them in the order given, so each spec can depend on the
176 options added by earlier ones. It is also possible to remove options
177 using @samp{%<@var{option}} in the usual way.
179 This macro can be useful when a port has several interdependent target
180 options. It provides a way of standardizing the command line so
181 that the other specs are easier to write.
183 Do not define this macro if it does not need to do anything.
186 @defmac OPTION_DEFAULT_SPECS
187 A list of specs used to support configure-time default options (i.e.@:
188 @option{--with} options) in the driver. It should be a suitable initializer
189 for an array of structures, each containing two strings, without the
190 outermost pair of surrounding braces.
192 The first item in the pair is the name of the default. This must match
193 the code in @file{config.gcc} for the target. The second item is a spec
194 to apply if a default with this name was specified. The string
195 @samp{%(VALUE)} in the spec will be replaced by the value of the default
196 everywhere it occurs.
198 The driver will apply these specs to its own command line between loading
199 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
200 the same mechanism as @code{DRIVER_SELF_SPECS}.
202 Do not define this macro if it does not need to do anything.
206 A C string constant that tells the GCC driver program options to
207 pass to CPP@. It can also specify how to translate options you
208 give to GCC into options for GCC to pass to the CPP@.
210 Do not define this macro if it does not need to do anything.
213 @defmac CPLUSPLUS_CPP_SPEC
214 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
215 than C@. If you do not define this macro, then the value of
216 @code{CPP_SPEC} (if any) will be used instead.
220 A C string constant that tells the GCC driver program options to
221 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
223 It can also specify how to translate options you give to GCC into options
224 for GCC to pass to front ends.
226 Do not define this macro if it does not need to do anything.
230 A C string constant that tells the GCC driver program options to
231 pass to @code{cc1plus}. It can also specify how to translate options you
232 give to GCC into options for GCC to pass to the @code{cc1plus}.
234 Do not define this macro if it does not need to do anything.
235 Note that everything defined in CC1_SPEC is already passed to
236 @code{cc1plus} so there is no need to duplicate the contents of
237 CC1_SPEC in CC1PLUS_SPEC@.
241 A C string constant that tells the GCC driver program options to
242 pass to the assembler. It can also specify how to translate options
243 you give to GCC into options for GCC to pass to the assembler.
244 See the file @file{sun3.h} for an example of this.
246 Do not define this macro if it does not need to do anything.
249 @defmac ASM_FINAL_SPEC
250 A C string constant that tells the GCC driver program how to
251 run any programs which cleanup after the normal assembler.
252 Normally, this is not needed. See the file @file{mips.h} for
255 Do not define this macro if it does not need to do anything.
258 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
259 Define this macro, with no value, if the driver should give the assembler
260 an argument consisting of a single dash, @option{-}, to instruct it to
261 read from its standard input (which will be a pipe connected to the
262 output of the compiler proper). This argument is given after any
263 @option{-o} option specifying the name of the output file.
265 If you do not define this macro, the assembler is assumed to read its
266 standard input if given no non-option arguments. If your assembler
267 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
268 see @file{mips.h} for instance.
272 A C string constant that tells the GCC driver program options to
273 pass to the linker. It can also specify how to translate options you
274 give to GCC into options for GCC to pass to the linker.
276 Do not define this macro if it does not need to do anything.
280 Another C string constant used much like @code{LINK_SPEC}. The difference
281 between the two is that @code{LIB_SPEC} is used at the end of the
282 command given to the linker.
284 If this macro is not defined, a default is provided that
285 loads the standard C library from the usual place. See @file{gcc.c}.
289 Another C string constant that tells the GCC driver program
290 how and when to place a reference to @file{libgcc.a} into the
291 linker command line. This constant is placed both before and after
292 the value of @code{LIB_SPEC}.
294 If this macro is not defined, the GCC driver provides a default that
295 passes the string @option{-lgcc} to the linker.
298 @defmac REAL_LIBGCC_SPEC
299 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
300 @code{LIBGCC_SPEC} is not directly used by the driver program but is
301 instead modified to refer to different versions of @file{libgcc.a}
302 depending on the values of the command line flags @option{-static},
303 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
304 targets where these modifications are inappropriate, define
305 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
306 driver how to place a reference to @file{libgcc} on the link command
307 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
310 @defmac USE_LD_AS_NEEDED
311 A macro that controls the modifications to @code{LIBGCC_SPEC}
312 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
313 generated that uses --as-needed and the shared libgcc in place of the
314 static exception handler library, when linking without any of
315 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
319 If defined, this C string constant is added to @code{LINK_SPEC}.
320 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
321 the modifications to @code{LIBGCC_SPEC} mentioned in
322 @code{REAL_LIBGCC_SPEC}.
325 @defmac STARTFILE_SPEC
326 Another C string constant used much like @code{LINK_SPEC}. The
327 difference between the two is that @code{STARTFILE_SPEC} is used at
328 the very beginning of the command given to the linker.
330 If this macro is not defined, a default is provided that loads the
331 standard C startup file from the usual place. See @file{gcc.c}.
335 Another C string constant used much like @code{LINK_SPEC}. The
336 difference between the two is that @code{ENDFILE_SPEC} is used at
337 the very end of the command given to the linker.
339 Do not define this macro if it does not need to do anything.
342 @defmac THREAD_MODEL_SPEC
343 GCC @code{-v} will print the thread model GCC was configured to use.
344 However, this doesn't work on platforms that are multilibbed on thread
345 models, such as AIX 4.3. On such platforms, define
346 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
347 blanks that names one of the recognized thread models. @code{%*}, the
348 default value of this macro, will expand to the value of
349 @code{thread_file} set in @file{config.gcc}.
352 @defmac SYSROOT_SUFFIX_SPEC
353 Define this macro to add a suffix to the target sysroot when GCC is
354 configured with a sysroot. This will cause GCC to search for usr/lib,
355 et al, within sysroot+suffix.
358 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
359 Define this macro to add a headers_suffix to the target sysroot when
360 GCC is configured with a sysroot. This will cause GCC to pass the
361 updated sysroot+headers_suffix to CPP, causing it to search for
362 usr/include, et al, within sysroot+headers_suffix.
366 Define this macro to provide additional specifications to put in the
367 @file{specs} file that can be used in various specifications like
370 The definition should be an initializer for an array of structures,
371 containing a string constant, that defines the specification name, and a
372 string constant that provides the specification.
374 Do not define this macro if it does not need to do anything.
376 @code{EXTRA_SPECS} is useful when an architecture contains several
377 related targets, which have various @code{@dots{}_SPECS} which are similar
378 to each other, and the maintainer would like one central place to keep
381 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
382 define either @code{_CALL_SYSV} when the System V calling sequence is
383 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
386 The @file{config/rs6000/rs6000.h} target file defines:
389 #define EXTRA_SPECS \
390 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
392 #define CPP_SYS_DEFAULT ""
395 The @file{config/rs6000/sysv.h} target file defines:
399 "%@{posix: -D_POSIX_SOURCE @} \
400 %@{mcall-sysv: -D_CALL_SYSV @} \
401 %@{!mcall-sysv: %(cpp_sysv_default) @} \
402 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
404 #undef CPP_SYSV_DEFAULT
405 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
408 while the @file{config/rs6000/eabiaix.h} target file defines
409 @code{CPP_SYSV_DEFAULT} as:
412 #undef CPP_SYSV_DEFAULT
413 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
417 @defmac LINK_LIBGCC_SPECIAL_1
418 Define this macro if the driver program should find the library
419 @file{libgcc.a}. If you do not define this macro, the driver program will pass
420 the argument @option{-lgcc} to tell the linker to do the search.
423 @defmac LINK_GCC_C_SEQUENCE_SPEC
424 The sequence in which libgcc and libc are specified to the linker.
425 By default this is @code{%G %L %G}.
428 @defmac LINK_COMMAND_SPEC
429 A C string constant giving the complete command line need to execute the
430 linker. When you do this, you will need to update your port each time a
431 change is made to the link command line within @file{gcc.c}. Therefore,
432 define this macro only if you need to completely redefine the command
433 line for invoking the linker and there is no other way to accomplish
434 the effect you need. Overriding this macro may be avoidable by overriding
435 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
438 @defmac LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
439 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
440 directories from linking commands. Do not give it a nonzero value if
441 removing duplicate search directories changes the linker's semantics.
444 @defmac MULTILIB_DEFAULTS
445 Define this macro as a C expression for the initializer of an array of
446 string to tell the driver program which options are defaults for this
447 target and thus do not need to be handled specially when using
448 @code{MULTILIB_OPTIONS}.
450 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
451 the target makefile fragment or if none of the options listed in
452 @code{MULTILIB_OPTIONS} are set by default.
453 @xref{Target Fragment}.
456 @defmac RELATIVE_PREFIX_NOT_LINKDIR
457 Define this macro to tell @command{gcc} that it should only translate
458 a @option{-B} prefix into a @option{-L} linker option if the prefix
459 indicates an absolute file name.
462 @defmac MD_EXEC_PREFIX
463 If defined, this macro is an additional prefix to try after
464 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
465 when the @option{-b} option is used, or the compiler is built as a cross
466 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
467 to the list of directories used to find the assembler in @file{configure.in}.
470 @defmac STANDARD_STARTFILE_PREFIX
471 Define this macro as a C string constant if you wish to override the
472 standard choice of @code{libdir} as the default prefix to
473 try when searching for startup files such as @file{crt0.o}.
474 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
475 is built as a cross compiler.
478 @defmac STANDARD_STARTFILE_PREFIX_1
479 Define this macro as a C string constant if you wish to override the
480 standard choice of @code{/lib} as a prefix to try after the default prefix
481 when searching for startup files such as @file{crt0.o}.
482 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
483 is built as a cross compiler.
486 @defmac STANDARD_STARTFILE_PREFIX_2
487 Define this macro as a C string constant if you wish to override the
488 standard choice of @code{/lib} as yet another prefix to try after the
489 default prefix when searching for startup files such as @file{crt0.o}.
490 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
491 is built as a cross compiler.
494 @defmac MD_STARTFILE_PREFIX
495 If defined, this macro supplies an additional prefix to try after the
496 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
497 @option{-b} option is used, or when the compiler is built as a cross
501 @defmac MD_STARTFILE_PREFIX_1
502 If defined, this macro supplies yet another prefix to try after the
503 standard prefixes. It is not searched when the @option{-b} option is
504 used, or when the compiler is built as a cross compiler.
507 @defmac INIT_ENVIRONMENT
508 Define this macro as a C string constant if you wish to set environment
509 variables for programs called by the driver, such as the assembler and
510 loader. The driver passes the value of this macro to @code{putenv} to
511 initialize the necessary environment variables.
514 @defmac LOCAL_INCLUDE_DIR
515 Define this macro as a C string constant if you wish to override the
516 standard choice of @file{/usr/local/include} as the default prefix to
517 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
518 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
520 Cross compilers do not search either @file{/usr/local/include} or its
524 @defmac MODIFY_TARGET_NAME
525 Define this macro if you wish to define command-line switches that
526 modify the default target name.
528 For each switch, you can include a string to be appended to the first
529 part of the configuration name or a string to be deleted from the
530 configuration name, if present. The definition should be an initializer
531 for an array of structures. Each array element should have three
532 elements: the switch name (a string constant, including the initial
533 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
534 indicate whether the string should be inserted or deleted, and the string
535 to be inserted or deleted (a string constant).
537 For example, on a machine where @samp{64} at the end of the
538 configuration name denotes a 64-bit target and you want the @option{-32}
539 and @option{-64} switches to select between 32- and 64-bit targets, you would
543 #define MODIFY_TARGET_NAME \
544 @{ @{ "-32", DELETE, "64"@}, \
545 @{"-64", ADD, "64"@}@}
549 @defmac SYSTEM_INCLUDE_DIR
550 Define this macro as a C string constant if you wish to specify a
551 system-specific directory to search for header files before the standard
552 directory. @code{SYSTEM_INCLUDE_DIR} comes before
553 @code{STANDARD_INCLUDE_DIR} in the search order.
555 Cross compilers do not use this macro and do not search the directory
559 @defmac STANDARD_INCLUDE_DIR
560 Define this macro as a C string constant if you wish to override the
561 standard choice of @file{/usr/include} as the default prefix to
562 try when searching for header files.
564 Cross compilers ignore this macro and do not search either
565 @file{/usr/include} or its replacement.
568 @defmac STANDARD_INCLUDE_COMPONENT
569 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
570 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
571 If you do not define this macro, no component is used.
574 @defmac INCLUDE_DEFAULTS
575 Define this macro if you wish to override the entire default search path
576 for include files. For a native compiler, the default search path
577 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
578 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
579 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
580 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
581 and specify private search areas for GCC@. The directory
582 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
584 The definition should be an initializer for an array of structures.
585 Each array element should have four elements: the directory name (a
586 string constant), the component name (also a string constant), a flag
587 for C++-only directories,
588 and a flag showing that the includes in the directory don't need to be
589 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
590 the array with a null element.
592 The component name denotes what GNU package the include file is part of,
593 if any, in all uppercase letters. For example, it might be @samp{GCC}
594 or @samp{BINUTILS}. If the package is part of a vendor-supplied
595 operating system, code the component name as @samp{0}.
597 For example, here is the definition used for VAX/VMS:
600 #define INCLUDE_DEFAULTS \
602 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
603 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
604 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
611 Here is the order of prefixes tried for exec files:
615 Any prefixes specified by the user with @option{-B}.
618 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
619 is not set and the compiler has not been installed in the configure-time
620 @var{prefix}, the location in which the compiler has actually been installed.
623 The directories specified by the environment variable @code{COMPILER_PATH}.
626 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
627 in the configured-time @var{prefix}.
630 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
633 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
636 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
640 Here is the order of prefixes tried for startfiles:
644 Any prefixes specified by the user with @option{-B}.
647 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
648 value based on the installed toolchain location.
651 The directories specified by the environment variable @code{LIBRARY_PATH}
652 (or port-specific name; native only, cross compilers do not use this).
655 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
656 in the configured @var{prefix} or this is a native compiler.
659 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
662 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
666 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
667 native compiler, or we have a target system root.
670 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
671 native compiler, or we have a target system root.
674 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
675 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
676 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
679 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
680 compiler, or we have a target system root. The default for this macro is
684 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
685 compiler, or we have a target system root. The default for this macro is
689 @node Run-time Target
690 @section Run-time Target Specification
691 @cindex run-time target specification
692 @cindex predefined macros
693 @cindex target specifications
695 @c prevent bad page break with this line
696 Here are run-time target specifications.
698 @defmac TARGET_CPU_CPP_BUILTINS ()
699 This function-like macro expands to a block of code that defines
700 built-in preprocessor macros and assertions for the target cpu, using
701 the functions @code{builtin_define}, @code{builtin_define_std} and
702 @code{builtin_assert}. When the front end
703 calls this macro it provides a trailing semicolon, and since it has
704 finished command line option processing your code can use those
707 @code{builtin_assert} takes a string in the form you pass to the
708 command-line option @option{-A}, such as @code{cpu=mips}, and creates
709 the assertion. @code{builtin_define} takes a string in the form
710 accepted by option @option{-D} and unconditionally defines the macro.
712 @code{builtin_define_std} takes a string representing the name of an
713 object-like macro. If it doesn't lie in the user's namespace,
714 @code{builtin_define_std} defines it unconditionally. Otherwise, it
715 defines a version with two leading underscores, and another version
716 with two leading and trailing underscores, and defines the original
717 only if an ISO standard was not requested on the command line. For
718 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
719 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
720 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
721 defines only @code{_ABI64}.
723 You can also test for the C dialect being compiled. The variable
724 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
725 or @code{clk_objective_c}. Note that if we are preprocessing
726 assembler, this variable will be @code{clk_c} but the function-like
727 macro @code{preprocessing_asm_p()} will return true, so you might want
728 to check for that first. If you need to check for strict ANSI, the
729 variable @code{flag_iso} can be used. The function-like macro
730 @code{preprocessing_trad_p()} can be used to check for traditional
734 @defmac TARGET_OS_CPP_BUILTINS ()
735 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
736 and is used for the target operating system instead.
739 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
740 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
741 and is used for the target object format. @file{elfos.h} uses this
742 macro to define @code{__ELF__}, so you probably do not need to define
746 @deftypevar {extern int} target_flags
747 This variable is declared in @file{options.h}, which is included before
748 any target-specific headers.
751 @deftypevar {Target Hook} int TARGET_DEFAULT_TARGET_FLAGS
752 This variable specifies the initial value of @code{target_flags}.
753 Its default setting is 0.
756 @cindex optional hardware or system features
757 @cindex features, optional, in system conventions
759 @deftypefn {Target Hook} bool TARGET_HANDLE_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
760 This hook is called whenever the user specifies one of the
761 target-specific options described by the @file{.opt} definition files
762 (@pxref{Options}). It has the opportunity to do some option-specific
763 processing and should return true if the option is valid. The default
764 definition does nothing but return true.
766 @var{code} specifies the @code{OPT_@var{name}} enumeration value
767 associated with the selected option; @var{name} is just a rendering of
768 the option name in which non-alphanumeric characters are replaced by
769 underscores. @var{arg} specifies the string argument and is null if
770 no argument was given. If the option is flagged as a @code{UInteger}
771 (@pxref{Option properties}), @var{value} is the numeric value of the
772 argument. Otherwise @var{value} is 1 if the positive form of the
773 option was used and 0 if the ``no-'' form was.
776 @defmac TARGET_VERSION
777 This macro is a C statement to print on @code{stderr} a string
778 describing the particular machine description choice. Every machine
779 description should define @code{TARGET_VERSION}. For example:
783 #define TARGET_VERSION \
784 fprintf (stderr, " (68k, Motorola syntax)");
786 #define TARGET_VERSION \
787 fprintf (stderr, " (68k, MIT syntax)");
792 @defmac OVERRIDE_OPTIONS
793 Sometimes certain combinations of command options do not make sense on
794 a particular target machine. You can define a macro
795 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
796 defined, is executed once just after all the command options have been
799 Don't use this macro to turn on various extra optimizations for
800 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
803 @defmac C_COMMON_OVERRIDE_OPTIONS
804 This is similar to @code{OVERRIDE_OPTIONS} but is only used in the C
805 language frontends (C, Objective-C, C++, Objective-C++) and so can be
806 used to alter option flag variables which only exist in those
810 @defmac OPTIMIZATION_OPTIONS (@var{level}, @var{size})
811 Some machines may desire to change what optimizations are performed for
812 various optimization levels. This macro, if defined, is executed once
813 just after the optimization level is determined and before the remainder
814 of the command options have been parsed. Values set in this macro are
815 used as the default values for the other command line options.
817 @var{level} is the optimization level specified; 2 if @option{-O2} is
818 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
820 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
822 You should not use this macro to change options that are not
823 machine-specific. These should uniformly selected by the same
824 optimization level on all supported machines. Use this macro to enable
825 machine-specific optimizations.
827 @strong{Do not examine @code{write_symbols} in
828 this macro!} The debugging options are not supposed to alter the
832 @defmac CAN_DEBUG_WITHOUT_FP
833 Define this macro if debugging can be performed even without a frame
834 pointer. If this macro is defined, GCC will turn on the
835 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
838 @node Per-Function Data
839 @section Defining data structures for per-function information.
840 @cindex per-function data
841 @cindex data structures
843 If the target needs to store information on a per-function basis, GCC
844 provides a macro and a couple of variables to allow this. Note, just
845 using statics to store the information is a bad idea, since GCC supports
846 nested functions, so you can be halfway through encoding one function
847 when another one comes along.
849 GCC defines a data structure called @code{struct function} which
850 contains all of the data specific to an individual function. This
851 structure contains a field called @code{machine} whose type is
852 @code{struct machine_function *}, which can be used by targets to point
853 to their own specific data.
855 If a target needs per-function specific data it should define the type
856 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
857 This macro should be used to initialize the function pointer
858 @code{init_machine_status}. This pointer is explained below.
860 One typical use of per-function, target specific data is to create an
861 RTX to hold the register containing the function's return address. This
862 RTX can then be used to implement the @code{__builtin_return_address}
863 function, for level 0.
865 Note---earlier implementations of GCC used a single data area to hold
866 all of the per-function information. Thus when processing of a nested
867 function began the old per-function data had to be pushed onto a
868 stack, and when the processing was finished, it had to be popped off the
869 stack. GCC used to provide function pointers called
870 @code{save_machine_status} and @code{restore_machine_status} to handle
871 the saving and restoring of the target specific information. Since the
872 single data area approach is no longer used, these pointers are no
875 @defmac INIT_EXPANDERS
876 Macro called to initialize any target specific information. This macro
877 is called once per function, before generation of any RTL has begun.
878 The intention of this macro is to allow the initialization of the
879 function pointer @code{init_machine_status}.
882 @deftypevar {void (*)(struct function *)} init_machine_status
883 If this function pointer is non-@code{NULL} it will be called once per
884 function, before function compilation starts, in order to allow the
885 target to perform any target specific initialization of the
886 @code{struct function} structure. It is intended that this would be
887 used to initialize the @code{machine} of that structure.
889 @code{struct machine_function} structures are expected to be freed by GC@.
890 Generally, any memory that they reference must be allocated by using
891 @code{ggc_alloc}, including the structure itself.
895 @section Storage Layout
896 @cindex storage layout
898 Note that the definitions of the macros in this table which are sizes or
899 alignments measured in bits do not need to be constant. They can be C
900 expressions that refer to static variables, such as the @code{target_flags}.
901 @xref{Run-time Target}.
903 @defmac BITS_BIG_ENDIAN
904 Define this macro to have the value 1 if the most significant bit in a
905 byte has the lowest number; otherwise define it to have the value zero.
906 This means that bit-field instructions count from the most significant
907 bit. If the machine has no bit-field instructions, then this must still
908 be defined, but it doesn't matter which value it is defined to. This
909 macro need not be a constant.
911 This macro does not affect the way structure fields are packed into
912 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
915 @defmac BYTES_BIG_ENDIAN
916 Define this macro to have the value 1 if the most significant byte in a
917 word has the lowest number. This macro need not be a constant.
920 @defmac WORDS_BIG_ENDIAN
921 Define this macro to have the value 1 if, in a multiword object, the
922 most significant word has the lowest number. This applies to both
923 memory locations and registers; GCC fundamentally assumes that the
924 order of words in memory is the same as the order in registers. This
925 macro need not be a constant.
928 @defmac LIBGCC2_WORDS_BIG_ENDIAN
929 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
930 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
931 used only when compiling @file{libgcc2.c}. Typically the value will be set
932 based on preprocessor defines.
935 @defmac FLOAT_WORDS_BIG_ENDIAN
936 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
937 @code{TFmode} floating point numbers are stored in memory with the word
938 containing the sign bit at the lowest address; otherwise define it to
939 have the value 0. This macro need not be a constant.
941 You need not define this macro if the ordering is the same as for
945 @defmac BITS_PER_UNIT
946 Define this macro to be the number of bits in an addressable storage
947 unit (byte). If you do not define this macro the default is 8.
950 @defmac BITS_PER_WORD
951 Number of bits in a word. If you do not define this macro, the default
952 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
955 @defmac MAX_BITS_PER_WORD
956 Maximum number of bits in a word. If this is undefined, the default is
957 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
958 largest value that @code{BITS_PER_WORD} can have at run-time.
961 @defmac UNITS_PER_WORD
962 Number of storage units in a word; normally the size of a general-purpose
963 register, a power of two from 1 or 8.
966 @defmac MIN_UNITS_PER_WORD
967 Minimum number of units in a word. If this is undefined, the default is
968 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
969 smallest value that @code{UNITS_PER_WORD} can have at run-time.
972 @defmac UNITS_PER_SIMD_WORD
973 Number of units in the vectors that the vectorizer can produce.
974 The default is equal to @code{UNITS_PER_WORD}, because the vectorizer
975 can do some transformations even in absence of specialized @acronym{SIMD}
980 Width of a pointer, in bits. You must specify a value no wider than the
981 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
982 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
983 a value the default is @code{BITS_PER_WORD}.
986 @defmac POINTERS_EXTEND_UNSIGNED
987 A C expression whose value is greater than zero if pointers that need to be
988 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
989 be zero-extended and zero if they are to be sign-extended. If the value
990 is less then zero then there must be an "ptr_extend" instruction that
991 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
993 You need not define this macro if the @code{POINTER_SIZE} is equal
994 to the width of @code{Pmode}.
997 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
998 A macro to update @var{m} and @var{unsignedp} when an object whose type
999 is @var{type} and which has the specified mode and signedness is to be
1000 stored in a register. This macro is only called when @var{type} is a
1003 On most RISC machines, which only have operations that operate on a full
1004 register, define this macro to set @var{m} to @code{word_mode} if
1005 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
1006 cases, only integer modes should be widened because wider-precision
1007 floating-point operations are usually more expensive than their narrower
1010 For most machines, the macro definition does not change @var{unsignedp}.
1011 However, some machines, have instructions that preferentially handle
1012 either signed or unsigned quantities of certain modes. For example, on
1013 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
1014 sign-extend the result to 64 bits. On such machines, set
1015 @var{unsignedp} according to which kind of extension is more efficient.
1017 Do not define this macro if it would never modify @var{m}.
1020 @defmac PROMOTE_FUNCTION_MODE
1021 Like @code{PROMOTE_MODE}, but is applied to outgoing function arguments or
1022 function return values, as specified by @code{TARGET_PROMOTE_FUNCTION_ARGS}
1023 and @code{TARGET_PROMOTE_FUNCTION_RETURN}, respectively.
1025 The default is @code{PROMOTE_MODE}.
1028 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_ARGS (tree @var{fntype})
1029 This target hook should return @code{true} if the promotion described by
1030 @code{PROMOTE_FUNCTION_MODE} should be done for outgoing function
1034 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_RETURN (tree @var{fntype})
1035 This target hook should return @code{true} if the promotion described by
1036 @code{PROMOTE_FUNCTION_MODE} should be done for the return value of
1039 If this target hook returns @code{true}, @code{TARGET_FUNCTION_VALUE}
1040 must perform the same promotions done by @code{PROMOTE_FUNCTION_MODE}.
1043 @defmac PARM_BOUNDARY
1044 Normal alignment required for function parameters on the stack, in
1045 bits. All stack parameters receive at least this much alignment
1046 regardless of data type. On most machines, this is the same as the
1050 @defmac STACK_BOUNDARY
1051 Define this macro to the minimum alignment enforced by hardware for the
1052 stack pointer on this machine. The definition is a C expression for the
1053 desired alignment (measured in bits). This value is used as a default
1054 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1055 this should be the same as @code{PARM_BOUNDARY}.
1058 @defmac PREFERRED_STACK_BOUNDARY
1059 Define this macro if you wish to preserve a certain alignment for the
1060 stack pointer, greater than what the hardware enforces. The definition
1061 is a C expression for the desired alignment (measured in bits). This
1062 macro must evaluate to a value equal to or larger than
1063 @code{STACK_BOUNDARY}.
1066 @defmac FUNCTION_BOUNDARY
1067 Alignment required for a function entry point, in bits.
1070 @defmac BIGGEST_ALIGNMENT
1071 Biggest alignment that any data type can require on this machine, in bits.
1074 @defmac MINIMUM_ATOMIC_ALIGNMENT
1075 If defined, the smallest alignment, in bits, that can be given to an
1076 object that can be referenced in one operation, without disturbing any
1077 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1078 on machines that don't have byte or half-word store operations.
1081 @defmac BIGGEST_FIELD_ALIGNMENT
1082 Biggest alignment that any structure or union field can require on this
1083 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1084 structure and union fields only, unless the field alignment has been set
1085 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1088 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1089 An expression for the alignment of a structure field @var{field} if the
1090 alignment computed in the usual way (including applying of
1091 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1092 alignment) is @var{computed}. It overrides alignment only if the
1093 field alignment has not been set by the
1094 @code{__attribute__ ((aligned (@var{n})))} construct.
1097 @defmac MAX_OFILE_ALIGNMENT
1098 Biggest alignment supported by the object file format of this machine.
1099 Use this macro to limit the alignment which can be specified using the
1100 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1101 the default value is @code{BIGGEST_ALIGNMENT}.
1103 On systems that use ELF, the default (in @file{config/elfos.h}) is
1104 the largest supported 32-bit ELF section alignment representable on
1105 a 32-bit host e.g. @samp{(((unsigned HOST_WIDEST_INT) 1 << 28) * 8)}.
1106 On 32-bit ELF the largest supported section alignment in bits is
1107 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1110 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1111 If defined, a C expression to compute the alignment for a variable in
1112 the static store. @var{type} is the data type, and @var{basic-align} is
1113 the alignment that the object would ordinarily have. The value of this
1114 macro is used instead of that alignment to align the object.
1116 If this macro is not defined, then @var{basic-align} is used.
1119 One use of this macro is to increase alignment of medium-size data to
1120 make it all fit in fewer cache lines. Another is to cause character
1121 arrays to be word-aligned so that @code{strcpy} calls that copy
1122 constants to character arrays can be done inline.
1125 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1126 If defined, a C expression to compute the alignment given to a constant
1127 that is being placed in memory. @var{constant} is the constant and
1128 @var{basic-align} is the alignment that the object would ordinarily
1129 have. The value of this macro is used instead of that alignment to
1132 If this macro is not defined, then @var{basic-align} is used.
1134 The typical use of this macro is to increase alignment for string
1135 constants to be word aligned so that @code{strcpy} calls that copy
1136 constants can be done inline.
1139 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1140 If defined, a C expression to compute the alignment for a variable in
1141 the local store. @var{type} is the data type, and @var{basic-align} is
1142 the alignment that the object would ordinarily have. The value of this
1143 macro is used instead of that alignment to align the object.
1145 If this macro is not defined, then @var{basic-align} is used.
1147 One use of this macro is to increase alignment of medium-size data to
1148 make it all fit in fewer cache lines.
1151 @defmac EMPTY_FIELD_BOUNDARY
1152 Alignment in bits to be given to a structure bit-field that follows an
1153 empty field such as @code{int : 0;}.
1155 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1158 @defmac STRUCTURE_SIZE_BOUNDARY
1159 Number of bits which any structure or union's size must be a multiple of.
1160 Each structure or union's size is rounded up to a multiple of this.
1162 If you do not define this macro, the default is the same as
1163 @code{BITS_PER_UNIT}.
1166 @defmac STRICT_ALIGNMENT
1167 Define this macro to be the value 1 if instructions will fail to work
1168 if given data not on the nominal alignment. If instructions will merely
1169 go slower in that case, define this macro as 0.
1172 @defmac PCC_BITFIELD_TYPE_MATTERS
1173 Define this if you wish to imitate the way many other C compilers handle
1174 alignment of bit-fields and the structures that contain them.
1176 The behavior is that the type written for a named bit-field (@code{int},
1177 @code{short}, or other integer type) imposes an alignment for the entire
1178 structure, as if the structure really did contain an ordinary field of
1179 that type. In addition, the bit-field is placed within the structure so
1180 that it would fit within such a field, not crossing a boundary for it.
1182 Thus, on most machines, a named bit-field whose type is written as
1183 @code{int} would not cross a four-byte boundary, and would force
1184 four-byte alignment for the whole structure. (The alignment used may
1185 not be four bytes; it is controlled by the other alignment parameters.)
1187 An unnamed bit-field will not affect the alignment of the containing
1190 If the macro is defined, its definition should be a C expression;
1191 a nonzero value for the expression enables this behavior.
1193 Note that if this macro is not defined, or its value is zero, some
1194 bit-fields may cross more than one alignment boundary. The compiler can
1195 support such references if there are @samp{insv}, @samp{extv}, and
1196 @samp{extzv} insns that can directly reference memory.
1198 The other known way of making bit-fields work is to define
1199 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1200 Then every structure can be accessed with fullwords.
1202 Unless the machine has bit-field instructions or you define
1203 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1204 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1206 If your aim is to make GCC use the same conventions for laying out
1207 bit-fields as are used by another compiler, here is how to investigate
1208 what the other compiler does. Compile and run this program:
1227 printf ("Size of foo1 is %d\n",
1228 sizeof (struct foo1));
1229 printf ("Size of foo2 is %d\n",
1230 sizeof (struct foo2));
1235 If this prints 2 and 5, then the compiler's behavior is what you would
1236 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1239 @defmac BITFIELD_NBYTES_LIMITED
1240 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1241 to aligning a bit-field within the structure.
1244 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELDS (void)
1245 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1246 whether unnamed bitfields affect the alignment of the containing
1247 structure. The hook should return true if the structure should inherit
1248 the alignment requirements of an unnamed bitfield's type.
1251 @deftypefn {Target Hook} bool TARGET_NARROW_VOLATILE_BITFIELDS (void)
1252 This target hook should return @code{true} if accesses to volatile bitfields
1253 should use the narrowest mode possible. It should return @code{false} if
1254 these accesses should use the bitfield container type.
1256 The default is @code{!TARGET_STRICT_ALIGN}.
1259 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1260 Return 1 if a structure or array containing @var{field} should be accessed using
1263 If @var{field} is the only field in the structure, @var{mode} is its
1264 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1265 case where structures of one field would require the structure's mode to
1266 retain the field's mode.
1268 Normally, this is not needed. See the file @file{c4x.h} for an example
1269 of how to use this macro to prevent a structure having a floating point
1270 field from being accessed in an integer mode.
1273 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1274 Define this macro as an expression for the alignment of a type (given
1275 by @var{type} as a tree node) if the alignment computed in the usual
1276 way is @var{computed} and the alignment explicitly specified was
1279 The default is to use @var{specified} if it is larger; otherwise, use
1280 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1283 @defmac MAX_FIXED_MODE_SIZE
1284 An integer expression for the size in bits of the largest integer
1285 machine mode that should actually be used. All integer machine modes of
1286 this size or smaller can be used for structures and unions with the
1287 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1288 (DImode)} is assumed.
1291 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1292 If defined, an expression of type @code{enum machine_mode} that
1293 specifies the mode of the save area operand of a
1294 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1295 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1296 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1297 having its mode specified.
1299 You need not define this macro if it always returns @code{Pmode}. You
1300 would most commonly define this macro if the
1301 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1305 @defmac STACK_SIZE_MODE
1306 If defined, an expression of type @code{enum machine_mode} that
1307 specifies the mode of the size increment operand of an
1308 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1310 You need not define this macro if it always returns @code{word_mode}.
1311 You would most commonly define this macro if the @code{allocate_stack}
1312 pattern needs to support both a 32- and a 64-bit mode.
1315 @defmac TARGET_FLOAT_FORMAT
1316 A code distinguishing the floating point format of the target machine.
1317 There are four defined values:
1320 @item IEEE_FLOAT_FORMAT
1321 This code indicates IEEE floating point. It is the default; there is no
1322 need to define @code{TARGET_FLOAT_FORMAT} when the format is IEEE@.
1324 @item VAX_FLOAT_FORMAT
1325 This code indicates the ``F float'' (for @code{float}) and ``D float''
1326 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
1328 @item IBM_FLOAT_FORMAT
1329 This code indicates the format used on the IBM System/370.
1331 @item C4X_FLOAT_FORMAT
1332 This code indicates the format used on the TMS320C3x/C4x.
1335 If your target uses a floating point format other than these, you must
1336 define a new @var{name}_FLOAT_FORMAT code for it, and add support for
1337 it to @file{real.c}.
1339 The ordering of the component words of floating point values stored in
1340 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1343 @defmac MODE_HAS_NANS (@var{mode})
1344 When defined, this macro should be true if @var{mode} has a NaN
1345 representation. The compiler assumes that NaNs are not equal to
1346 anything (including themselves) and that addition, subtraction,
1347 multiplication and division all return NaNs when one operand is
1350 By default, this macro is true if @var{mode} is a floating-point
1351 mode and the target floating-point format is IEEE@.
1354 @defmac MODE_HAS_INFINITIES (@var{mode})
1355 This macro should be true if @var{mode} can represent infinity. At
1356 present, the compiler uses this macro to decide whether @samp{x - x}
1357 is always defined. By default, the macro is true when @var{mode}
1358 is a floating-point mode and the target format is IEEE@.
1361 @defmac MODE_HAS_SIGNED_ZEROS (@var{mode})
1362 True if @var{mode} distinguishes between positive and negative zero.
1363 The rules are expected to follow the IEEE standard:
1367 @samp{x + x} has the same sign as @samp{x}.
1370 If the sum of two values with opposite sign is zero, the result is
1371 positive for all rounding modes expect towards @minus{}infinity, for
1372 which it is negative.
1375 The sign of a product or quotient is negative when exactly one
1376 of the operands is negative.
1379 The default definition is true if @var{mode} is a floating-point
1380 mode and the target format is IEEE@.
1383 @defmac MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1384 If defined, this macro should be true for @var{mode} if it has at
1385 least one rounding mode in which @samp{x} and @samp{-x} can be
1386 rounded to numbers of different magnitude. Two such modes are
1387 towards @minus{}infinity and towards +infinity.
1389 The default definition of this macro is true if @var{mode} is
1390 a floating-point mode and the target format is IEEE@.
1393 @defmac ROUND_TOWARDS_ZERO
1394 If defined, this macro should be true if the prevailing rounding
1395 mode is towards zero. A true value has the following effects:
1399 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1402 @file{libgcc.a}'s floating-point emulator will round towards zero
1403 rather than towards nearest.
1406 The compiler's floating-point emulator will round towards zero after
1407 doing arithmetic, and when converting from the internal float format to
1411 The macro does not affect the parsing of string literals. When the
1412 primary rounding mode is towards zero, library functions like
1413 @code{strtod} might still round towards nearest, and the compiler's
1414 parser should behave like the target's @code{strtod} where possible.
1416 Not defining this macro is equivalent to returning zero.
1419 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1420 This macro should return true if floats with @var{size}
1421 bits do not have a NaN or infinity representation, but use the largest
1422 exponent for normal numbers instead.
1424 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1425 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1426 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1427 floating-point arithmetic.
1429 The default definition of this macro returns false for all sizes.
1432 @deftypefn {Target Hook} bool TARGET_VECTOR_OPAQUE_P (tree @var{type})
1433 This target hook should return @code{true} a vector is opaque. That
1434 is, if no cast is needed when copying a vector value of type
1435 @var{type} into another vector lvalue of the same size. Vector opaque
1436 types cannot be initialized. The default is that there are no such
1440 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1441 This target hook returns @code{true} if bit-fields in the given
1442 @var{record_type} are to be laid out following the rules of Microsoft
1443 Visual C/C++, namely: (i) a bit-field won't share the same storage
1444 unit with the previous bit-field if their underlying types have
1445 different sizes, and the bit-field will be aligned to the highest
1446 alignment of the underlying types of itself and of the previous
1447 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1448 the whole enclosing structure, even if it is unnamed; except that
1449 (iii) a zero-sized bit-field will be disregarded unless it follows
1450 another bit-field of nonzero size. If this hook returns @code{true},
1451 other macros that control bit-field layout are ignored.
1453 When a bit-field is inserted into a packed record, the whole size
1454 of the underlying type is used by one or more same-size adjacent
1455 bit-fields (that is, if its long:3, 32 bits is used in the record,
1456 and any additional adjacent long bit-fields are packed into the same
1457 chunk of 32 bits. However, if the size changes, a new field of that
1458 size is allocated). In an unpacked record, this is the same as using
1459 alignment, but not equivalent when packing.
1461 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1462 the latter will take precedence. If @samp{__attribute__((packed))} is
1463 used on a single field when MS bit-fields are in use, it will take
1464 precedence for that field, but the alignment of the rest of the structure
1465 may affect its placement.
1468 @deftypefn {Target Hook} {bool} TARGET_DECIMAL_FLOAT_SUPPORTED_P (void)
1469 Returns true if the target supports decimal floating point.
1472 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_FUNDAMENTAL_TYPE (tree @var{type})
1473 If your target defines any fundamental types, define this hook to
1474 return the appropriate encoding for these types as part of a C++
1475 mangled name. The @var{type} argument is the tree structure
1476 representing the type to be mangled. The hook may be applied to trees
1477 which are not target-specific fundamental types; it should return
1478 @code{NULL} for all such types, as well as arguments it does not
1479 recognize. If the return value is not @code{NULL}, it must point to
1480 a statically-allocated string constant.
1482 Target-specific fundamental types might be new fundamental types or
1483 qualified versions of ordinary fundamental types. Encode new
1484 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1485 is the name used for the type in source code, and @var{n} is the
1486 length of @var{name} in decimal. Encode qualified versions of
1487 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1488 @var{name} is the name used for the type qualifier in source code,
1489 @var{n} is the length of @var{name} as above, and @var{code} is the
1490 code used to represent the unqualified version of this type. (See
1491 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1492 codes.) In both cases the spaces are for clarity; do not include any
1493 spaces in your string.
1495 The default version of this hook always returns @code{NULL}, which is
1496 appropriate for a target that does not define any new fundamental
1501 @section Layout of Source Language Data Types
1503 These macros define the sizes and other characteristics of the standard
1504 basic data types used in programs being compiled. Unlike the macros in
1505 the previous section, these apply to specific features of C and related
1506 languages, rather than to fundamental aspects of storage layout.
1508 @defmac INT_TYPE_SIZE
1509 A C expression for the size in bits of the type @code{int} on the
1510 target machine. If you don't define this, the default is one word.
1513 @defmac SHORT_TYPE_SIZE
1514 A C expression for the size in bits of the type @code{short} on the
1515 target machine. If you don't define this, the default is half a word.
1516 (If this would be less than one storage unit, it is rounded up to one
1520 @defmac LONG_TYPE_SIZE
1521 A C expression for the size in bits of the type @code{long} on the
1522 target machine. If you don't define this, the default is one word.
1525 @defmac ADA_LONG_TYPE_SIZE
1526 On some machines, the size used for the Ada equivalent of the type
1527 @code{long} by a native Ada compiler differs from that used by C@. In
1528 that situation, define this macro to be a C expression to be used for
1529 the size of that type. If you don't define this, the default is the
1530 value of @code{LONG_TYPE_SIZE}.
1533 @defmac LONG_LONG_TYPE_SIZE
1534 A C expression for the size in bits of the type @code{long long} on the
1535 target machine. If you don't define this, the default is two
1536 words. If you want to support GNU Ada on your machine, the value of this
1537 macro must be at least 64.
1540 @defmac CHAR_TYPE_SIZE
1541 A C expression for the size in bits of the type @code{char} on the
1542 target machine. If you don't define this, the default is
1543 @code{BITS_PER_UNIT}.
1546 @defmac BOOL_TYPE_SIZE
1547 A C expression for the size in bits of the C++ type @code{bool} and
1548 C99 type @code{_Bool} on the target machine. If you don't define
1549 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1552 @defmac FLOAT_TYPE_SIZE
1553 A C expression for the size in bits of the type @code{float} on the
1554 target machine. If you don't define this, the default is one word.
1557 @defmac DOUBLE_TYPE_SIZE
1558 A C expression for the size in bits of the type @code{double} on the
1559 target machine. If you don't define this, the default is two
1563 @defmac LONG_DOUBLE_TYPE_SIZE
1564 A C expression for the size in bits of the type @code{long double} on
1565 the target machine. If you don't define this, the default is two
1569 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1570 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1571 if you want routines in @file{libgcc2.a} for a size other than
1572 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1573 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1576 @defmac LIBGCC2_HAS_DF_MODE
1577 Define this macro if neither @code{LIBGCC2_DOUBLE_TYPE_SIZE} nor
1578 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1579 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1580 anyway. If you don't define this and either @code{LIBGCC2_DOUBLE_TYPE_SIZE}
1581 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1585 @defmac LIBGCC2_HAS_XF_MODE
1586 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1587 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1588 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1589 is 80 then the default is 1, otherwise it is 0.
1592 @defmac LIBGCC2_HAS_TF_MODE
1593 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1594 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1595 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1596 is 128 then the default is 1, otherwise it is 0.
1603 Define these macros to be the size in bits of the mantissa of
1604 @code{SFmode}, @code{DFmode}, @code{XFmode} and @code{TFmode} values,
1605 if the defaults in @file{libgcc2.h} are inappropriate. By default,
1606 @code{FLT_MANT_DIG} is used for @code{SF_SIZE}, @code{LDBL_MANT_DIG}
1607 for @code{XF_SIZE} and @code{TF_SIZE}, and @code{DBL_MANT_DIG} or
1608 @code{LDBL_MANT_DIG} for @code{DF_SIZE} according to whether
1609 @code{LIBGCC2_DOUBLE_TYPE_SIZE} or
1610 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64.
1613 @defmac TARGET_FLT_EVAL_METHOD
1614 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1615 assuming, if applicable, that the floating-point control word is in its
1616 default state. If you do not define this macro the value of
1617 @code{FLT_EVAL_METHOD} will be zero.
1620 @defmac WIDEST_HARDWARE_FP_SIZE
1621 A C expression for the size in bits of the widest floating-point format
1622 supported by the hardware. If you define this macro, you must specify a
1623 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1624 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1628 @defmac DEFAULT_SIGNED_CHAR
1629 An expression whose value is 1 or 0, according to whether the type
1630 @code{char} should be signed or unsigned by default. The user can
1631 always override this default with the options @option{-fsigned-char}
1632 and @option{-funsigned-char}.
1635 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1636 This target hook should return true if the compiler should give an
1637 @code{enum} type only as many bytes as it takes to represent the range
1638 of possible values of that type. It should return false if all
1639 @code{enum} types should be allocated like @code{int}.
1641 The default is to return false.
1645 A C expression for a string describing the name of the data type to use
1646 for size values. The typedef name @code{size_t} is defined using the
1647 contents of the string.
1649 The string can contain more than one keyword. If so, separate them with
1650 spaces, and write first any length keyword, then @code{unsigned} if
1651 appropriate, and finally @code{int}. The string must exactly match one
1652 of the data type names defined in the function
1653 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1654 omit @code{int} or change the order---that would cause the compiler to
1657 If you don't define this macro, the default is @code{"long unsigned
1661 @defmac PTRDIFF_TYPE
1662 A C expression for a string describing the name of the data type to use
1663 for the result of subtracting two pointers. The typedef name
1664 @code{ptrdiff_t} is defined using the contents of the string. See
1665 @code{SIZE_TYPE} above for more information.
1667 If you don't define this macro, the default is @code{"long int"}.
1671 A C expression for a string describing the name of the data type to use
1672 for wide characters. The typedef name @code{wchar_t} is defined using
1673 the contents of the string. See @code{SIZE_TYPE} above for more
1676 If you don't define this macro, the default is @code{"int"}.
1679 @defmac WCHAR_TYPE_SIZE
1680 A C expression for the size in bits of the data type for wide
1681 characters. This is used in @code{cpp}, which cannot make use of
1686 A C expression for a string describing the name of the data type to
1687 use for wide characters passed to @code{printf} and returned from
1688 @code{getwc}. The typedef name @code{wint_t} is defined using the
1689 contents of the string. See @code{SIZE_TYPE} above for more
1692 If you don't define this macro, the default is @code{"unsigned int"}.
1696 A C expression for a string describing the name of the data type that
1697 can represent any value of any standard or extended signed integer type.
1698 The typedef name @code{intmax_t} is defined using the contents of the
1699 string. See @code{SIZE_TYPE} above for more information.
1701 If you don't define this macro, the default is the first of
1702 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1703 much precision as @code{long long int}.
1706 @defmac UINTMAX_TYPE
1707 A C expression for a string describing the name of the data type that
1708 can represent any value of any standard or extended unsigned integer
1709 type. The typedef name @code{uintmax_t} is defined using the contents
1710 of the string. See @code{SIZE_TYPE} above for more information.
1712 If you don't define this macro, the default is the first of
1713 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1714 unsigned int"} that has as much precision as @code{long long unsigned
1718 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1719 The C++ compiler represents a pointer-to-member-function with a struct
1726 ptrdiff_t vtable_index;
1733 The C++ compiler must use one bit to indicate whether the function that
1734 will be called through a pointer-to-member-function is virtual.
1735 Normally, we assume that the low-order bit of a function pointer must
1736 always be zero. Then, by ensuring that the vtable_index is odd, we can
1737 distinguish which variant of the union is in use. But, on some
1738 platforms function pointers can be odd, and so this doesn't work. In
1739 that case, we use the low-order bit of the @code{delta} field, and shift
1740 the remainder of the @code{delta} field to the left.
1742 GCC will automatically make the right selection about where to store
1743 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1744 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1745 set such that functions always start at even addresses, but the lowest
1746 bit of pointers to functions indicate whether the function at that
1747 address is in ARM or Thumb mode. If this is the case of your
1748 architecture, you should define this macro to
1749 @code{ptrmemfunc_vbit_in_delta}.
1751 In general, you should not have to define this macro. On architectures
1752 in which function addresses are always even, according to
1753 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1754 @code{ptrmemfunc_vbit_in_pfn}.
1757 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1758 Normally, the C++ compiler uses function pointers in vtables. This
1759 macro allows the target to change to use ``function descriptors''
1760 instead. Function descriptors are found on targets for whom a
1761 function pointer is actually a small data structure. Normally the
1762 data structure consists of the actual code address plus a data
1763 pointer to which the function's data is relative.
1765 If vtables are used, the value of this macro should be the number
1766 of words that the function descriptor occupies.
1769 @defmac TARGET_VTABLE_ENTRY_ALIGN
1770 By default, the vtable entries are void pointers, the so the alignment
1771 is the same as pointer alignment. The value of this macro specifies
1772 the alignment of the vtable entry in bits. It should be defined only
1773 when special alignment is necessary. */
1776 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1777 There are a few non-descriptor entries in the vtable at offsets below
1778 zero. If these entries must be padded (say, to preserve the alignment
1779 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1780 of words in each data entry.
1784 @section Register Usage
1785 @cindex register usage
1787 This section explains how to describe what registers the target machine
1788 has, and how (in general) they can be used.
1790 The description of which registers a specific instruction can use is
1791 done with register classes; see @ref{Register Classes}. For information
1792 on using registers to access a stack frame, see @ref{Frame Registers}.
1793 For passing values in registers, see @ref{Register Arguments}.
1794 For returning values in registers, see @ref{Scalar Return}.
1797 * Register Basics:: Number and kinds of registers.
1798 * Allocation Order:: Order in which registers are allocated.
1799 * Values in Registers:: What kinds of values each reg can hold.
1800 * Leaf Functions:: Renumbering registers for leaf functions.
1801 * Stack Registers:: Handling a register stack such as 80387.
1804 @node Register Basics
1805 @subsection Basic Characteristics of Registers
1807 @c prevent bad page break with this line
1808 Registers have various characteristics.
1810 @defmac FIRST_PSEUDO_REGISTER
1811 Number of hardware registers known to the compiler. They receive
1812 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1813 pseudo register's number really is assigned the number
1814 @code{FIRST_PSEUDO_REGISTER}.
1817 @defmac FIXED_REGISTERS
1818 @cindex fixed register
1819 An initializer that says which registers are used for fixed purposes
1820 all throughout the compiled code and are therefore not available for
1821 general allocation. These would include the stack pointer, the frame
1822 pointer (except on machines where that can be used as a general
1823 register when no frame pointer is needed), the program counter on
1824 machines where that is considered one of the addressable registers,
1825 and any other numbered register with a standard use.
1827 This information is expressed as a sequence of numbers, separated by
1828 commas and surrounded by braces. The @var{n}th number is 1 if
1829 register @var{n} is fixed, 0 otherwise.
1831 The table initialized from this macro, and the table initialized by
1832 the following one, may be overridden at run time either automatically,
1833 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1834 the user with the command options @option{-ffixed-@var{reg}},
1835 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1838 @defmac CALL_USED_REGISTERS
1839 @cindex call-used register
1840 @cindex call-clobbered register
1841 @cindex call-saved register
1842 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1843 clobbered (in general) by function calls as well as for fixed
1844 registers. This macro therefore identifies the registers that are not
1845 available for general allocation of values that must live across
1848 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1849 automatically saves it on function entry and restores it on function
1850 exit, if the register is used within the function.
1853 @defmac CALL_REALLY_USED_REGISTERS
1854 @cindex call-used register
1855 @cindex call-clobbered register
1856 @cindex call-saved register
1857 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1858 that the entire set of @code{FIXED_REGISTERS} be included.
1859 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1860 This macro is optional. If not specified, it defaults to the value
1861 of @code{CALL_USED_REGISTERS}.
1864 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1865 @cindex call-used register
1866 @cindex call-clobbered register
1867 @cindex call-saved register
1868 A C expression that is nonzero if it is not permissible to store a
1869 value of mode @var{mode} in hard register number @var{regno} across a
1870 call without some part of it being clobbered. For most machines this
1871 macro need not be defined. It is only required for machines that do not
1872 preserve the entire contents of a register across a call.
1876 @findex call_used_regs
1879 @findex reg_class_contents
1880 @defmac CONDITIONAL_REGISTER_USAGE
1881 Zero or more C statements that may conditionally modify five variables
1882 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1883 @code{reg_names}, and @code{reg_class_contents}, to take into account
1884 any dependence of these register sets on target flags. The first three
1885 of these are of type @code{char []} (interpreted as Boolean vectors).
1886 @code{global_regs} is a @code{const char *[]}, and
1887 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1888 called, @code{fixed_regs}, @code{call_used_regs},
1889 @code{reg_class_contents}, and @code{reg_names} have been initialized
1890 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1891 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1892 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1893 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1894 command options have been applied.
1896 You need not define this macro if it has no work to do.
1898 @cindex disabling certain registers
1899 @cindex controlling register usage
1900 If the usage of an entire class of registers depends on the target
1901 flags, you may indicate this to GCC by using this macro to modify
1902 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1903 registers in the classes which should not be used by GCC@. Also define
1904 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1905 to return @code{NO_REGS} if it
1906 is called with a letter for a class that shouldn't be used.
1908 (However, if this class is not included in @code{GENERAL_REGS} and all
1909 of the insn patterns whose constraints permit this class are
1910 controlled by target switches, then GCC will automatically avoid using
1911 these registers when the target switches are opposed to them.)
1914 @defmac INCOMING_REGNO (@var{out})
1915 Define this macro if the target machine has register windows. This C
1916 expression returns the register number as seen by the called function
1917 corresponding to the register number @var{out} as seen by the calling
1918 function. Return @var{out} if register number @var{out} is not an
1922 @defmac OUTGOING_REGNO (@var{in})
1923 Define this macro if the target machine has register windows. This C
1924 expression returns the register number as seen by the calling function
1925 corresponding to the register number @var{in} as seen by the called
1926 function. Return @var{in} if register number @var{in} is not an inbound
1930 @defmac LOCAL_REGNO (@var{regno})
1931 Define this macro if the target machine has register windows. This C
1932 expression returns true if the register is call-saved but is in the
1933 register window. Unlike most call-saved registers, such registers
1934 need not be explicitly restored on function exit or during non-local
1939 If the program counter has a register number, define this as that
1940 register number. Otherwise, do not define it.
1943 @node Allocation Order
1944 @subsection Order of Allocation of Registers
1945 @cindex order of register allocation
1946 @cindex register allocation order
1948 @c prevent bad page break with this line
1949 Registers are allocated in order.
1951 @defmac REG_ALLOC_ORDER
1952 If defined, an initializer for a vector of integers, containing the
1953 numbers of hard registers in the order in which GCC should prefer
1954 to use them (from most preferred to least).
1956 If this macro is not defined, registers are used lowest numbered first
1957 (all else being equal).
1959 One use of this macro is on machines where the highest numbered
1960 registers must always be saved and the save-multiple-registers
1961 instruction supports only sequences of consecutive registers. On such
1962 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1963 the highest numbered allocable register first.
1966 @defmac ORDER_REGS_FOR_LOCAL_ALLOC
1967 A C statement (sans semicolon) to choose the order in which to allocate
1968 hard registers for pseudo-registers local to a basic block.
1970 Store the desired register order in the array @code{reg_alloc_order}.
1971 Element 0 should be the register to allocate first; element 1, the next
1972 register; and so on.
1974 The macro body should not assume anything about the contents of
1975 @code{reg_alloc_order} before execution of the macro.
1977 On most machines, it is not necessary to define this macro.
1980 @node Values in Registers
1981 @subsection How Values Fit in Registers
1983 This section discusses the macros that describe which kinds of values
1984 (specifically, which machine modes) each register can hold, and how many
1985 consecutive registers are needed for a given mode.
1987 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
1988 A C expression for the number of consecutive hard registers, starting
1989 at register number @var{regno}, required to hold a value of mode
1992 On a machine where all registers are exactly one word, a suitable
1993 definition of this macro is
1996 #define HARD_REGNO_NREGS(REGNO, MODE) \
1997 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2002 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2003 A C expression that is nonzero if a value of mode @var{mode}, stored
2004 in memory, ends with padding that causes it to take up more space than
2005 in registers starting at register number @var{regno} (as determined by
2006 multiplying GCC's notion of the size of the register when containing
2007 this mode by the number of registers returned by
2008 @code{HARD_REGNO_NREGS}). By default this is zero.
2010 For example, if a floating-point value is stored in three 32-bit
2011 registers but takes up 128 bits in memory, then this would be
2014 This macros only needs to be defined if there are cases where
2015 @code{subreg_get_info}
2016 would otherwise wrongly determine that a @code{subreg} can be
2017 represented by an offset to the register number, when in fact such a
2018 @code{subreg} would contain some of the padding not stored in
2019 registers and so not be representable.
2022 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2023 For values of @var{regno} and @var{mode} for which
2024 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2025 returning the greater number of registers required to hold the value
2026 including any padding. In the example above, the value would be four.
2029 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2030 Define this macro if the natural size of registers that hold values
2031 of mode @var{mode} is not the word size. It is a C expression that
2032 should give the natural size in bytes for the specified mode. It is
2033 used by the register allocator to try to optimize its results. This
2034 happens for example on SPARC 64-bit where the natural size of
2035 floating-point registers is still 32-bit.
2038 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2039 A C expression that is nonzero if it is permissible to store a value
2040 of mode @var{mode} in hard register number @var{regno} (or in several
2041 registers starting with that one). For a machine where all registers
2042 are equivalent, a suitable definition is
2045 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2048 You need not include code to check for the numbers of fixed registers,
2049 because the allocation mechanism considers them to be always occupied.
2051 @cindex register pairs
2052 On some machines, double-precision values must be kept in even/odd
2053 register pairs. You can implement that by defining this macro to reject
2054 odd register numbers for such modes.
2056 The minimum requirement for a mode to be OK in a register is that the
2057 @samp{mov@var{mode}} instruction pattern support moves between the
2058 register and other hard register in the same class and that moving a
2059 value into the register and back out not alter it.
2061 Since the same instruction used to move @code{word_mode} will work for
2062 all narrower integer modes, it is not necessary on any machine for
2063 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2064 you define patterns @samp{movhi}, etc., to take advantage of this. This
2065 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2066 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2069 Many machines have special registers for floating point arithmetic.
2070 Often people assume that floating point machine modes are allowed only
2071 in floating point registers. This is not true. Any registers that
2072 can hold integers can safely @emph{hold} a floating point machine
2073 mode, whether or not floating arithmetic can be done on it in those
2074 registers. Integer move instructions can be used to move the values.
2076 On some machines, though, the converse is true: fixed-point machine
2077 modes may not go in floating registers. This is true if the floating
2078 registers normalize any value stored in them, because storing a
2079 non-floating value there would garble it. In this case,
2080 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2081 floating registers. But if the floating registers do not automatically
2082 normalize, if you can store any bit pattern in one and retrieve it
2083 unchanged without a trap, then any machine mode may go in a floating
2084 register, so you can define this macro to say so.
2086 The primary significance of special floating registers is rather that
2087 they are the registers acceptable in floating point arithmetic
2088 instructions. However, this is of no concern to
2089 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2090 constraints for those instructions.
2092 On some machines, the floating registers are especially slow to access,
2093 so that it is better to store a value in a stack frame than in such a
2094 register if floating point arithmetic is not being done. As long as the
2095 floating registers are not in class @code{GENERAL_REGS}, they will not
2096 be used unless some pattern's constraint asks for one.
2099 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2100 A C expression that is nonzero if it is OK to rename a hard register
2101 @var{from} to another hard register @var{to}.
2103 One common use of this macro is to prevent renaming of a register to
2104 another register that is not saved by a prologue in an interrupt
2107 The default is always nonzero.
2110 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2111 A C expression that is nonzero if a value of mode
2112 @var{mode1} is accessible in mode @var{mode2} without copying.
2114 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2115 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2116 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2117 should be nonzero. If they differ for any @var{r}, you should define
2118 this macro to return zero unless some other mechanism ensures the
2119 accessibility of the value in a narrower mode.
2121 You should define this macro to return nonzero in as many cases as
2122 possible since doing so will allow GCC to perform better register
2126 @defmac AVOID_CCMODE_COPIES
2127 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2128 registers. You should only define this macro if support for copying to/from
2129 @code{CCmode} is incomplete.
2132 @node Leaf Functions
2133 @subsection Handling Leaf Functions
2135 @cindex leaf functions
2136 @cindex functions, leaf
2137 On some machines, a leaf function (i.e., one which makes no calls) can run
2138 more efficiently if it does not make its own register window. Often this
2139 means it is required to receive its arguments in the registers where they
2140 are passed by the caller, instead of the registers where they would
2143 The special treatment for leaf functions generally applies only when
2144 other conditions are met; for example, often they may use only those
2145 registers for its own variables and temporaries. We use the term ``leaf
2146 function'' to mean a function that is suitable for this special
2147 handling, so that functions with no calls are not necessarily ``leaf
2150 GCC assigns register numbers before it knows whether the function is
2151 suitable for leaf function treatment. So it needs to renumber the
2152 registers in order to output a leaf function. The following macros
2155 @defmac LEAF_REGISTERS
2156 Name of a char vector, indexed by hard register number, which
2157 contains 1 for a register that is allowable in a candidate for leaf
2160 If leaf function treatment involves renumbering the registers, then the
2161 registers marked here should be the ones before renumbering---those that
2162 GCC would ordinarily allocate. The registers which will actually be
2163 used in the assembler code, after renumbering, should not be marked with 1
2166 Define this macro only if the target machine offers a way to optimize
2167 the treatment of leaf functions.
2170 @defmac LEAF_REG_REMAP (@var{regno})
2171 A C expression whose value is the register number to which @var{regno}
2172 should be renumbered, when a function is treated as a leaf function.
2174 If @var{regno} is a register number which should not appear in a leaf
2175 function before renumbering, then the expression should yield @minus{}1, which
2176 will cause the compiler to abort.
2178 Define this macro only if the target machine offers a way to optimize the
2179 treatment of leaf functions, and registers need to be renumbered to do
2183 @findex current_function_is_leaf
2184 @findex current_function_uses_only_leaf_regs
2185 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2186 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2187 specially. They can test the C variable @code{current_function_is_leaf}
2188 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2189 set prior to local register allocation and is valid for the remaining
2190 compiler passes. They can also test the C variable
2191 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2192 functions which only use leaf registers.
2193 @code{current_function_uses_only_leaf_regs} is valid after all passes
2194 that modify the instructions have been run and is only useful if
2195 @code{LEAF_REGISTERS} is defined.
2196 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2197 @c of the next paragraph?! --mew 2feb93
2199 @node Stack Registers
2200 @subsection Registers That Form a Stack
2202 There are special features to handle computers where some of the
2203 ``registers'' form a stack. Stack registers are normally written by
2204 pushing onto the stack, and are numbered relative to the top of the
2207 Currently, GCC can only handle one group of stack-like registers, and
2208 they must be consecutively numbered. Furthermore, the existing
2209 support for stack-like registers is specific to the 80387 floating
2210 point coprocessor. If you have a new architecture that uses
2211 stack-like registers, you will need to do substantial work on
2212 @file{reg-stack.c} and write your machine description to cooperate
2213 with it, as well as defining these macros.
2216 Define this if the machine has any stack-like registers.
2219 @defmac FIRST_STACK_REG
2220 The number of the first stack-like register. This one is the top
2224 @defmac LAST_STACK_REG
2225 The number of the last stack-like register. This one is the bottom of
2229 @node Register Classes
2230 @section Register Classes
2231 @cindex register class definitions
2232 @cindex class definitions, register
2234 On many machines, the numbered registers are not all equivalent.
2235 For example, certain registers may not be allowed for indexed addressing;
2236 certain registers may not be allowed in some instructions. These machine
2237 restrictions are described to the compiler using @dfn{register classes}.
2239 You define a number of register classes, giving each one a name and saying
2240 which of the registers belong to it. Then you can specify register classes
2241 that are allowed as operands to particular instruction patterns.
2245 In general, each register will belong to several classes. In fact, one
2246 class must be named @code{ALL_REGS} and contain all the registers. Another
2247 class must be named @code{NO_REGS} and contain no registers. Often the
2248 union of two classes will be another class; however, this is not required.
2250 @findex GENERAL_REGS
2251 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2252 terribly special about the name, but the operand constraint letters
2253 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2254 the same as @code{ALL_REGS}, just define it as a macro which expands
2257 Order the classes so that if class @var{x} is contained in class @var{y}
2258 then @var{x} has a lower class number than @var{y}.
2260 The way classes other than @code{GENERAL_REGS} are specified in operand
2261 constraints is through machine-dependent operand constraint letters.
2262 You can define such letters to correspond to various classes, then use
2263 them in operand constraints.
2265 You should define a class for the union of two classes whenever some
2266 instruction allows both classes. For example, if an instruction allows
2267 either a floating point (coprocessor) register or a general register for a
2268 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2269 which includes both of them. Otherwise you will get suboptimal code.
2271 You must also specify certain redundant information about the register
2272 classes: for each class, which classes contain it and which ones are
2273 contained in it; for each pair of classes, the largest class contained
2276 When a value occupying several consecutive registers is expected in a
2277 certain class, all the registers used must belong to that class.
2278 Therefore, register classes cannot be used to enforce a requirement for
2279 a register pair to start with an even-numbered register. The way to
2280 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2282 Register classes used for input-operands of bitwise-and or shift
2283 instructions have a special requirement: each such class must have, for
2284 each fixed-point machine mode, a subclass whose registers can transfer that
2285 mode to or from memory. For example, on some machines, the operations for
2286 single-byte values (@code{QImode}) are limited to certain registers. When
2287 this is so, each register class that is used in a bitwise-and or shift
2288 instruction must have a subclass consisting of registers from which
2289 single-byte values can be loaded or stored. This is so that
2290 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2292 @deftp {Data type} {enum reg_class}
2293 An enumerated type that must be defined with all the register class names
2294 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2295 must be the last register class, followed by one more enumerated value,
2296 @code{LIM_REG_CLASSES}, which is not a register class but rather
2297 tells how many classes there are.
2299 Each register class has a number, which is the value of casting
2300 the class name to type @code{int}. The number serves as an index
2301 in many of the tables described below.
2304 @defmac N_REG_CLASSES
2305 The number of distinct register classes, defined as follows:
2308 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2312 @defmac REG_CLASS_NAMES
2313 An initializer containing the names of the register classes as C string
2314 constants. These names are used in writing some of the debugging dumps.
2317 @defmac REG_CLASS_CONTENTS
2318 An initializer containing the contents of the register classes, as integers
2319 which are bit masks. The @var{n}th integer specifies the contents of class
2320 @var{n}. The way the integer @var{mask} is interpreted is that
2321 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2323 When the machine has more than 32 registers, an integer does not suffice.
2324 Then the integers are replaced by sub-initializers, braced groupings containing
2325 several integers. Each sub-initializer must be suitable as an initializer
2326 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2327 In this situation, the first integer in each sub-initializer corresponds to
2328 registers 0 through 31, the second integer to registers 32 through 63, and
2332 @defmac REGNO_REG_CLASS (@var{regno})
2333 A C expression whose value is a register class containing hard register
2334 @var{regno}. In general there is more than one such class; choose a class
2335 which is @dfn{minimal}, meaning that no smaller class also contains the
2339 @defmac BASE_REG_CLASS
2340 A macro whose definition is the name of the class to which a valid
2341 base register must belong. A base register is one used in an address
2342 which is the register value plus a displacement.
2345 @defmac MODE_BASE_REG_CLASS (@var{mode})
2346 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2347 the selection of a base register in a mode dependent manner. If
2348 @var{mode} is VOIDmode then it should return the same value as
2349 @code{BASE_REG_CLASS}.
2352 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2353 A C expression whose value is the register class to which a valid
2354 base register must belong in order to be used in a base plus index
2355 register address. You should define this macro if base plus index
2356 addresses have different requirements than other base register uses.
2359 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{outer_code}, @var{index_code})
2360 A C expression whose value is the register class to which a valid
2361 base register must belong. @var{outer_code} and @var{index_code} define the
2362 context in which the base register occurs. @var{outer_code} is the code of
2363 the immediately enclosing expression (@code{MEM} for the top level of an
2364 address, @code{ADDRESS} for something that occurs in an
2365 @code{address_operand}). @var{index_code} is the code of the corresponding
2366 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2369 @defmac INDEX_REG_CLASS
2370 A macro whose definition is the name of the class to which a valid
2371 index register must belong. An index register is one used in an
2372 address where its value is either multiplied by a scale factor or
2373 added to another register (as well as added to a displacement).
2376 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2377 A C expression which is nonzero if register number @var{num} is
2378 suitable for use as a base register in operand addresses. It may be
2379 either a suitable hard register or a pseudo register that has been
2380 allocated such a hard register.
2383 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2384 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2385 that expression may examine the mode of the memory reference in
2386 @var{mode}. You should define this macro if the mode of the memory
2387 reference affects whether a register may be used as a base register. If
2388 you define this macro, the compiler will use it instead of
2389 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2390 addresses that appear outside a @code{MEM}, i.e., as an
2391 @code{address_operand}.
2395 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2396 A C expression which is nonzero if register number @var{num} is suitable for
2397 use as a base register in base plus index operand addresses, accessing
2398 memory in mode @var{mode}. It may be either a suitable hard register or a
2399 pseudo register that has been allocated such a hard register. You should
2400 define this macro if base plus index addresses have different requirements
2401 than other base register uses.
2403 Use of this macro is deprecated; please use the more general
2404 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2407 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{outer_code}, @var{index_code})
2408 A C expression that is just like @code{REGNO_MODE_OK_FOR_BASE_P}, except
2409 that that expression may examine the context in which the register
2410 appears in the memory reference. @var{outer_code} is the code of the
2411 immediately enclosing expression (@code{MEM} if at the top level of the
2412 address, @code{ADDRESS} for something that occurs in an
2413 @code{address_operand}). @var{index_code} is the code of the
2414 corresponding index expression if @var{outer_code} is @code{PLUS};
2415 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2416 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2419 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2420 A C expression which is nonzero if register number @var{num} is
2421 suitable for use as an index register in operand addresses. It may be
2422 either a suitable hard register or a pseudo register that has been
2423 allocated such a hard register.
2425 The difference between an index register and a base register is that
2426 the index register may be scaled. If an address involves the sum of
2427 two registers, neither one of them scaled, then either one may be
2428 labeled the ``base'' and the other the ``index''; but whichever
2429 labeling is used must fit the machine's constraints of which registers
2430 may serve in each capacity. The compiler will try both labelings,
2431 looking for one that is valid, and will reload one or both registers
2432 only if neither labeling works.
2435 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2436 A C expression that places additional restrictions on the register class
2437 to use when it is necessary to copy value @var{x} into a register in class
2438 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2439 another, smaller class. On many machines, the following definition is
2443 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2446 Sometimes returning a more restrictive class makes better code. For
2447 example, on the 68000, when @var{x} is an integer constant that is in range
2448 for a @samp{moveq} instruction, the value of this macro is always
2449 @code{DATA_REGS} as long as @var{class} includes the data registers.
2450 Requiring a data register guarantees that a @samp{moveq} will be used.
2452 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2453 @var{class} is if @var{x} is a legitimate constant which cannot be
2454 loaded into some register class. By returning @code{NO_REGS} you can
2455 force @var{x} into a memory location. For example, rs6000 can load
2456 immediate values into general-purpose registers, but does not have an
2457 instruction for loading an immediate value into a floating-point
2458 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2459 @var{x} is a floating-point constant. If the constant can't be loaded
2460 into any kind of register, code generation will be better if
2461 @code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2462 of using @code{PREFERRED_RELOAD_CLASS}.
2464 If an insn has pseudos in it after register allocation, reload will go
2465 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2466 to find the best one. Returning @code{NO_REGS}, in this case, makes
2467 reload add a @code{!} in front of the constraint: the x86 back-end uses
2468 this feature to discourage usage of 387 registers when math is done in
2469 the SSE registers (and vice versa).
2472 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2473 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2474 input reloads. If you don't define this macro, the default is to use
2475 @var{class}, unchanged.
2477 You can also use @code{PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2478 reload from using some alternatives, like @code{PREFERRED_RELOAD_CLASS}.
2481 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2482 A C expression that places additional restrictions on the register class
2483 to use when it is necessary to be able to hold a value of mode
2484 @var{mode} in a reload register for which class @var{class} would
2487 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2488 there are certain modes that simply can't go in certain reload classes.
2490 The value is a register class; perhaps @var{class}, or perhaps another,
2493 Don't define this macro unless the target machine has limitations which
2494 require the macro to do something nontrivial.
2497 @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})
2498 Many machines have some registers that cannot be copied directly to or
2499 from memory or even from other types of registers. An example is the
2500 @samp{MQ} register, which on most machines, can only be copied to or
2501 from general registers, but not memory. Below, we shall be using the
2502 term 'intermediate register' when a move operation cannot be performed
2503 directly, but has to be done by copying the source into the intermediate
2504 register first, and then copying the intermediate register to the
2505 destination. An intermediate register always has the same mode as
2506 source and destination. Since it holds the actual value being copied,
2507 reload might apply optimizations to re-use an intermediate register
2508 and eliding the copy from the source when it can determine that the
2509 intermediate register still holds the required value.
2511 Another kind of secondary reload is required on some machines which
2512 allow copying all registers to and from memory, but require a scratch
2513 register for stores to some memory locations (e.g., those with symbolic
2514 address on the RT, and those with certain symbolic address on the SPARC
2515 when compiling PIC)@. Scratch registers need not have the same mode
2516 as the value being copied, and usually hold a different value that
2517 that being copied. Special patterns in the md file are needed to
2518 describe how the copy is performed with the help of the scratch register;
2519 these patterns also describe the number, register class(es) and mode(s)
2520 of the scratch register(s).
2522 In some cases, both an intermediate and a scratch register are required.
2524 For input reloads, this target hook is called with nonzero @var{in_p},
2525 and @var{x} is an rtx that needs to be copied to a register in of class
2526 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2527 hook is called with zero @var{in_p}, and a register of class @var{reload_mode}
2528 needs to be copied to rtx @var{x} in @var{reload_mode}.
2530 If copying a register of @var{reload_class} from/to @var{x} requires
2531 an intermediate register, the hook @code{secondary_reload} should
2532 return the register class required for this intermediate register.
2533 If no intermediate register is required, it should return NO_REGS.
2534 If more than one intermediate register is required, describe the one
2535 that is closest in the copy chain to the reload register.
2537 If scratch registers are needed, you also have to describe how to
2538 perform the copy from/to the reload register to/from this
2539 closest intermediate register. Or if no intermediate register is
2540 required, but still a scratch register is needed, describe the
2541 copy from/to the reload register to/from the reload operand @var{x}.
2543 You do this by setting @code{sri->icode} to the instruction code of a pattern
2544 in the md file which performs the move. Operands 0 and 1 are the output
2545 and input of this copy, respectively. Operands from operand 2 onward are
2546 for scratch operands. These scratch operands must have a mode, and a
2547 single-register-class
2548 @c [later: or memory]
2551 When an intermediate register is used, the @code{secondary_reload}
2552 hook will be called again to determine how to copy the intermediate
2553 register to/from the reload operand @var{x}, so your hook must also
2554 have code to handle the register class of the intermediate operand.
2556 @c [For later: maybe we'll allow multi-alternative reload patterns -
2557 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2558 @c and match the constraints of input and output to determine the required
2559 @c alternative. A restriction would be that constraints used to match
2560 @c against reloads registers would have to be written as register class
2561 @c constraints, or we need a new target macro / hook that tells us if an
2562 @c arbitrary constraint can match an unknown register of a given class.
2563 @c Such a macro / hook would also be useful in other places.]
2566 @var{x} might be a pseudo-register or a @code{subreg} of a
2567 pseudo-register, which could either be in a hard register or in memory.
2568 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2569 in memory and the hard register number if it is in a register.
2571 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2572 currently not supported. For the time being, you will have to continue
2573 to use @code{SECONDARY_MEMORY_NEEDED} for that purpose.
2575 @code{copy_cost} also uses this target hook to find out how values are
2576 copied. If you want it to include some extra cost for the need to allocate
2577 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2578 Or if two dependent moves are supposed to have a lower cost than the sum
2579 of the individual moves due to expected fortuitous scheduling and/or special
2580 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2583 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2584 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2585 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2586 These macros are obsolete, new ports should use the target hook
2587 @code{TARGET_SECONDARY_RELOAD} instead.
2589 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2590 target hook. Older ports still define these macros to indicate to the
2591 reload phase that it may
2592 need to allocate at least one register for a reload in addition to the
2593 register to contain the data. Specifically, if copying @var{x} to a
2594 register @var{class} in @var{mode} requires an intermediate register,
2595 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2596 largest register class all of whose registers can be used as
2597 intermediate registers or scratch registers.
2599 If copying a register @var{class} in @var{mode} to @var{x} requires an
2600 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2601 was supposed to be defined be defined to return the largest register
2602 class required. If the
2603 requirements for input and output reloads were the same, the macro
2604 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2607 The values returned by these macros are often @code{GENERAL_REGS}.
2608 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2609 can be directly copied to or from a register of @var{class} in
2610 @var{mode} without requiring a scratch register. Do not define this
2611 macro if it would always return @code{NO_REGS}.
2613 If a scratch register is required (either with or without an
2614 intermediate register), you were supposed to define patterns for
2615 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2616 (@pxref{Standard Names}. These patterns, which were normally
2617 implemented with a @code{define_expand}, should be similar to the
2618 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2621 These patterns need constraints for the reload register and scratch
2623 contain a single register class. If the original reload register (whose
2624 class is @var{class}) can meet the constraint given in the pattern, the
2625 value returned by these macros is used for the class of the scratch
2626 register. Otherwise, two additional reload registers are required.
2627 Their classes are obtained from the constraints in the insn pattern.
2629 @var{x} might be a pseudo-register or a @code{subreg} of a
2630 pseudo-register, which could either be in a hard register or in memory.
2631 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2632 in memory and the hard register number if it is in a register.
2634 These macros should not be used in the case where a particular class of
2635 registers can only be copied to memory and not to another class of
2636 registers. In that case, secondary reload registers are not needed and
2637 would not be helpful. Instead, a stack location must be used to perform
2638 the copy and the @code{mov@var{m}} pattern should use memory as an
2639 intermediate storage. This case often occurs between floating-point and
2643 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2644 Certain machines have the property that some registers cannot be copied
2645 to some other registers without using memory. Define this macro on
2646 those machines to be a C expression that is nonzero if objects of mode
2647 @var{m} in registers of @var{class1} can only be copied to registers of
2648 class @var{class2} by storing a register of @var{class1} into memory
2649 and loading that memory location into a register of @var{class2}.
2651 Do not define this macro if its value would always be zero.
2654 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2655 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2656 allocates a stack slot for a memory location needed for register copies.
2657 If this macro is defined, the compiler instead uses the memory location
2658 defined by this macro.
2660 Do not define this macro if you do not define
2661 @code{SECONDARY_MEMORY_NEEDED}.
2664 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2665 When the compiler needs a secondary memory location to copy between two
2666 registers of mode @var{mode}, it normally allocates sufficient memory to
2667 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2668 load operations in a mode that many bits wide and whose class is the
2669 same as that of @var{mode}.
2671 This is right thing to do on most machines because it ensures that all
2672 bits of the register are copied and prevents accesses to the registers
2673 in a narrower mode, which some machines prohibit for floating-point
2676 However, this default behavior is not correct on some machines, such as
2677 the DEC Alpha, that store short integers in floating-point registers
2678 differently than in integer registers. On those machines, the default
2679 widening will not work correctly and you must define this macro to
2680 suppress that widening in some cases. See the file @file{alpha.h} for
2683 Do not define this macro if you do not define
2684 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2685 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2688 @defmac SMALL_REGISTER_CLASSES
2689 On some machines, it is risky to let hard registers live across arbitrary
2690 insns. Typically, these machines have instructions that require values
2691 to be in specific registers (like an accumulator), and reload will fail
2692 if the required hard register is used for another purpose across such an
2695 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2696 value on these machines. When this macro has a nonzero value, the
2697 compiler will try to minimize the lifetime of hard registers.
2699 It is always safe to define this macro with a nonzero value, but if you
2700 unnecessarily define it, you will reduce the amount of optimizations
2701 that can be performed in some cases. If you do not define this macro
2702 with a nonzero value when it is required, the compiler will run out of
2703 spill registers and print a fatal error message. For most machines, you
2704 should not define this macro at all.
2707 @defmac CLASS_LIKELY_SPILLED_P (@var{class})
2708 A C expression whose value is nonzero if pseudos that have been assigned
2709 to registers of class @var{class} would likely be spilled because
2710 registers of @var{class} are needed for spill registers.
2712 The default value of this macro returns 1 if @var{class} has exactly one
2713 register and zero otherwise. On most machines, this default should be
2714 used. Only define this macro to some other expression if pseudos
2715 allocated by @file{local-alloc.c} end up in memory because their hard
2716 registers were needed for spill registers. If this macro returns nonzero
2717 for those classes, those pseudos will only be allocated by
2718 @file{global.c}, which knows how to reallocate the pseudo to another
2719 register. If there would not be another register available for
2720 reallocation, you should not change the definition of this macro since
2721 the only effect of such a definition would be to slow down register
2725 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2726 A C expression for the maximum number of consecutive registers
2727 of class @var{class} needed to hold a value of mode @var{mode}.
2729 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2730 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2731 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2732 @var{mode})} for all @var{regno} values in the class @var{class}.
2734 This macro helps control the handling of multiple-word values
2738 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2739 If defined, a C expression that returns nonzero for a @var{class} for which
2740 a change from mode @var{from} to mode @var{to} is invalid.
2742 For the example, loading 32-bit integer or floating-point objects into
2743 floating-point registers on the Alpha extends them to 64 bits.
2744 Therefore loading a 64-bit object and then storing it as a 32-bit object
2745 does not store the low-order 32 bits, as would be the case for a normal
2746 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2750 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2751 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2752 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2756 @node Old Constraints
2757 @section Obsolete Macros for Defining Constraints
2758 @cindex defining constraints, obsolete method
2759 @cindex constraints, defining, obsolete method
2761 Machine-specific constraints can be defined with these macros instead
2762 of the machine description constructs described in @ref{Define
2763 Constraints}. This mechanism is obsolete. New ports should not use
2764 it; old ports should convert to the new mechanism.
2766 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2767 For the constraint at the start of @var{str}, which starts with the letter
2768 @var{c}, return the length. This allows you to have register class /
2769 constant / extra constraints that are longer than a single letter;
2770 you don't need to define this macro if you can do with single-letter
2771 constraints only. The definition of this macro should use
2772 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2773 to handle specially.
2774 There are some sanity checks in genoutput.c that check the constraint lengths
2775 for the md file, so you can also use this macro to help you while you are
2776 transitioning from a byzantine single-letter-constraint scheme: when you
2777 return a negative length for a constraint you want to re-use, genoutput
2778 will complain about every instance where it is used in the md file.
2781 @defmac REG_CLASS_FROM_LETTER (@var{char})
2782 A C expression which defines the machine-dependent operand constraint
2783 letters for register classes. If @var{char} is such a letter, the
2784 value should be the register class corresponding to it. Otherwise,
2785 the value should be @code{NO_REGS}. The register letter @samp{r},
2786 corresponding to class @code{GENERAL_REGS}, will not be passed
2787 to this macro; you do not need to handle it.
2790 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2791 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2792 passed in @var{str}, so that you can use suffixes to distinguish between
2796 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2797 A C expression that defines the machine-dependent operand constraint
2798 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2799 particular ranges of integer values. If @var{c} is one of those
2800 letters, the expression should check that @var{value}, an integer, is in
2801 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2802 not one of those letters, the value should be 0 regardless of
2806 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2807 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2808 string passed in @var{str}, so that you can use suffixes to distinguish
2809 between different variants.
2812 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2813 A C expression that defines the machine-dependent operand constraint
2814 letters that specify particular ranges of @code{const_double} values
2815 (@samp{G} or @samp{H}).
2817 If @var{c} is one of those letters, the expression should check that
2818 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2819 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2820 letters, the value should be 0 regardless of @var{value}.
2822 @code{const_double} is used for all floating-point constants and for
2823 @code{DImode} fixed-point constants. A given letter can accept either
2824 or both kinds of values. It can use @code{GET_MODE} to distinguish
2825 between these kinds.
2828 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2829 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2830 string passed in @var{str}, so that you can use suffixes to distinguish
2831 between different variants.
2834 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2835 A C expression that defines the optional machine-dependent constraint
2836 letters that can be used to segregate specific types of operands, usually
2837 memory references, for the target machine. Any letter that is not
2838 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2839 @code{REG_CLASS_FROM_CONSTRAINT}
2840 may be used. Normally this macro will not be defined.
2842 If it is required for a particular target machine, it should return 1
2843 if @var{value} corresponds to the operand type represented by the
2844 constraint letter @var{c}. If @var{c} is not defined as an extra
2845 constraint, the value returned should be 0 regardless of @var{value}.
2847 For example, on the ROMP, load instructions cannot have their output
2848 in r0 if the memory reference contains a symbolic address. Constraint
2849 letter @samp{Q} is defined as representing a memory address that does
2850 @emph{not} contain a symbolic address. An alternative is specified with
2851 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2852 alternative specifies @samp{m} on the input and a register class that
2853 does not include r0 on the output.
2856 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2857 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2858 in @var{str}, so that you can use suffixes to distinguish between different
2862 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2863 A C expression that defines the optional machine-dependent constraint
2864 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2865 be treated like memory constraints by the reload pass.
2867 It should return 1 if the operand type represented by the constraint
2868 at the start of @var{str}, the first letter of which is the letter @var{c},
2869 comprises a subset of all memory references including
2870 all those whose address is simply a base register. This allows the reload
2871 pass to reload an operand, if it does not directly correspond to the operand
2872 type of @var{c}, by copying its address into a base register.
2874 For example, on the S/390, some instructions do not accept arbitrary
2875 memory references, but only those that do not make use of an index
2876 register. The constraint letter @samp{Q} is defined via
2877 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2878 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2879 a @samp{Q} constraint can handle any memory operand, because the
2880 reload pass knows it can be reloaded by copying the memory address
2881 into a base register if required. This is analogous to the way
2882 a @samp{o} constraint can handle any memory operand.
2885 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2886 A C expression that defines the optional machine-dependent constraint
2887 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2888 @code{EXTRA_CONSTRAINT_STR}, that should
2889 be treated like address constraints by the reload pass.
2891 It should return 1 if the operand type represented by the constraint
2892 at the start of @var{str}, which starts with the letter @var{c}, comprises
2893 a subset of all memory addresses including
2894 all those that consist of just a base register. This allows the reload
2895 pass to reload an operand, if it does not directly correspond to the operand
2896 type of @var{str}, by copying it into a base register.
2898 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2899 be used with the @code{address_operand} predicate. It is treated
2900 analogously to the @samp{p} constraint.
2903 @node Stack and Calling
2904 @section Stack Layout and Calling Conventions
2905 @cindex calling conventions
2907 @c prevent bad page break with this line
2908 This describes the stack layout and calling conventions.
2912 * Exception Handling::
2917 * Register Arguments::
2919 * Aggregate Return::
2924 * Stack Smashing Protection::
2928 @subsection Basic Stack Layout
2929 @cindex stack frame layout
2930 @cindex frame layout
2932 @c prevent bad page break with this line
2933 Here is the basic stack layout.
2935 @defmac STACK_GROWS_DOWNWARD
2936 Define this macro if pushing a word onto the stack moves the stack
2937 pointer to a smaller address.
2939 When we say, ``define this macro if @dots{}'', it means that the
2940 compiler checks this macro only with @code{#ifdef} so the precise
2941 definition used does not matter.
2944 @defmac STACK_PUSH_CODE
2945 This macro defines the operation used when something is pushed
2946 on the stack. In RTL, a push operation will be
2947 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2949 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2950 and @code{POST_INC}. Which of these is correct depends on
2951 the stack direction and on whether the stack pointer points
2952 to the last item on the stack or whether it points to the
2953 space for the next item on the stack.
2955 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2956 defined, which is almost always right, and @code{PRE_INC} otherwise,
2957 which is often wrong.
2960 @defmac FRAME_GROWS_DOWNWARD
2961 Define this macro to nonzero value if the addresses of local variable slots
2962 are at negative offsets from the frame pointer.
2965 @defmac ARGS_GROW_DOWNWARD
2966 Define this macro if successive arguments to a function occupy decreasing
2967 addresses on the stack.
2970 @defmac STARTING_FRAME_OFFSET
2971 Offset from the frame pointer to the first local variable slot to be allocated.
2973 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2974 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2975 Otherwise, it is found by adding the length of the first slot to the
2976 value @code{STARTING_FRAME_OFFSET}.
2977 @c i'm not sure if the above is still correct.. had to change it to get
2978 @c rid of an overfull. --mew 2feb93
2981 @defmac STACK_ALIGNMENT_NEEDED
2982 Define to zero to disable final alignment of the stack during reload.
2983 The nonzero default for this macro is suitable for most ports.
2985 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
2986 is a register save block following the local block that doesn't require
2987 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2988 stack alignment and do it in the backend.
2991 @defmac STACK_POINTER_OFFSET
2992 Offset from the stack pointer register to the first location at which
2993 outgoing arguments are placed. If not specified, the default value of
2994 zero is used. This is the proper value for most machines.
2996 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2997 the first location at which outgoing arguments are placed.
3000 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3001 Offset from the argument pointer register to the first argument's
3002 address. On some machines it may depend on the data type of the
3005 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3006 the first argument's address.
3009 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3010 Offset from the stack pointer register to an item dynamically allocated
3011 on the stack, e.g., by @code{alloca}.
3013 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3014 length of the outgoing arguments. The default is correct for most
3015 machines. See @file{function.c} for details.
3018 @defmac INITIAL_FRAME_ADDRESS_RTX
3019 A C expression whose value is RTL representing the address of the initial
3020 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3021 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3022 default value will be used. Define this macro in order to make frame pointer
3023 elimination work in the presence of @code{__builtin_frame_address (count)} and
3024 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3027 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3028 A C expression whose value is RTL representing the address in a stack
3029 frame where the pointer to the caller's frame is stored. Assume that
3030 @var{frameaddr} is an RTL expression for the address of the stack frame
3033 If you don't define this macro, the default is to return the value
3034 of @var{frameaddr}---that is, the stack frame address is also the
3035 address of the stack word that points to the previous frame.
3038 @defmac SETUP_FRAME_ADDRESSES
3039 If defined, a C expression that produces the machine-specific code to
3040 setup the stack so that arbitrary frames can be accessed. For example,
3041 on the SPARC, we must flush all of the register windows to the stack
3042 before we can access arbitrary stack frames. You will seldom need to
3046 @deftypefn {Target Hook} bool TARGET_BUILTIN_SETJMP_FRAME_VALUE ()
3047 This target hook should return an rtx that is used to store
3048 the address of the current frame into the built in @code{setjmp} buffer.
3049 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3050 machines. One reason you may need to define this target hook is if
3051 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3054 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3055 A C expression whose value is RTL representing the value of the frame
3056 address for the current frame. @var{frameaddr} is the frame pointer
3057 of the current frame. This is used for __builtin_frame_address.
3058 You need only define this macro if the frame address is not the same
3059 as the frame pointer. Most machines do not need to define it.
3062 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3063 A C expression whose value is RTL representing the value of the return
3064 address for the frame @var{count} steps up from the current frame, after
3065 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3066 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3067 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
3069 The value of the expression must always be the correct address when
3070 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
3071 determine the return address of other frames.
3074 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3075 Define this if the return address of a particular stack frame is accessed
3076 from the frame pointer of the previous stack frame.
3079 @defmac INCOMING_RETURN_ADDR_RTX
3080 A C expression whose value is RTL representing the location of the
3081 incoming return address at the beginning of any function, before the
3082 prologue. This RTL is either a @code{REG}, indicating that the return
3083 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3086 You only need to define this macro if you want to support call frame
3087 debugging information like that provided by DWARF 2.
3089 If this RTL is a @code{REG}, you should also define
3090 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3093 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3094 A C expression whose value is an integer giving a DWARF 2 column
3095 number that may be used as an alternative return column. The column
3096 must not correspond to any gcc hard register (that is, it must not
3097 be in the range of @code{DWARF_FRAME_REGNUM}).
3099 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3100 general register, but an alternative column needs to be used for signal
3101 frames. Some targets have also used different frame return columns
3105 @defmac DWARF_ZERO_REG
3106 A C expression whose value is an integer giving a DWARF 2 register
3107 number that is considered to always have the value zero. This should
3108 only be defined if the target has an architected zero register, and
3109 someone decided it was a good idea to use that register number to
3110 terminate the stack backtrace. New ports should avoid this.
3113 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3114 This target hook allows the backend to emit frame-related insns that
3115 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3116 info engine will invoke it on insns of the form
3118 (set (reg) (unspec [...] UNSPEC_INDEX))
3122 (set (reg) (unspec_volatile [...] UNSPECV_INDEX)).
3124 to let the backend emit the call frame instructions. @var{label} is
3125 the CFI label attached to the insn, @var{pattern} is the pattern of
3126 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3129 @defmac INCOMING_FRAME_SP_OFFSET
3130 A C expression whose value is an integer giving the offset, in bytes,
3131 from the value of the stack pointer register to the top of the stack
3132 frame at the beginning of any function, before the prologue. The top of
3133 the frame is defined to be the value of the stack pointer in the
3134 previous frame, just before the call instruction.
3136 You only need to define this macro if you want to support call frame
3137 debugging information like that provided by DWARF 2.
3140 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3141 A C expression whose value is an integer giving the offset, in bytes,
3142 from the argument pointer to the canonical frame address (cfa). The
3143 final value should coincide with that calculated by
3144 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3145 during virtual register instantiation.
3147 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
3148 which is correct for most machines; in general, the arguments are found
3149 immediately before the stack frame. Note that this is not the case on
3150 some targets that save registers into the caller's frame, such as SPARC
3151 and rs6000, and so such targets need to define this macro.
3153 You only need to define this macro if the default is incorrect, and you
3154 want to support call frame debugging information like that provided by
3158 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3159 If defined, a C expression whose value is an integer giving the offset
3160 in bytes from the frame pointer to the canonical frame address (cfa).
3161 The final value should coincide with that calculated by
3162 @code{INCOMING_FRAME_SP_OFFSET}.
3164 Normally the CFA is calculated as an offset from the argument pointer,
3165 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3166 variable due to the ABI, this may not be possible. If this macro is
3167 defined, it implies that the virtual register instantiation should be
3168 based on the frame pointer instead of the argument pointer. Only one
3169 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3173 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3174 If defined, a C expression whose value is an integer giving the offset
3175 in bytes from the canonical frame address (cfa) to the frame base used
3176 in DWARF 2 debug information. The default is zero. A different value
3177 may reduce the size of debug information on some ports.
3180 @node Exception Handling
3181 @subsection Exception Handling Support
3182 @cindex exception handling
3184 @defmac EH_RETURN_DATA_REGNO (@var{N})
3185 A C expression whose value is the @var{N}th register number used for
3186 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3187 @var{N} registers are usable.
3189 The exception handling library routines communicate with the exception
3190 handlers via a set of agreed upon registers. Ideally these registers
3191 should be call-clobbered; it is possible to use call-saved registers,
3192 but may negatively impact code size. The target must support at least
3193 2 data registers, but should define 4 if there are enough free registers.
3195 You must define this macro if you want to support call frame exception
3196 handling like that provided by DWARF 2.
3199 @defmac EH_RETURN_STACKADJ_RTX
3200 A C expression whose value is RTL representing a location in which
3201 to store a stack adjustment to be applied before function return.
3202 This is used to unwind the stack to an exception handler's call frame.
3203 It will be assigned zero on code paths that return normally.
3205 Typically this is a call-clobbered hard register that is otherwise
3206 untouched by the epilogue, but could also be a stack slot.
3208 Do not define this macro if the stack pointer is saved and restored
3209 by the regular prolog and epilog code in the call frame itself; in
3210 this case, the exception handling library routines will update the
3211 stack location to be restored in place. Otherwise, you must define
3212 this macro if you want to support call frame exception handling like
3213 that provided by DWARF 2.
3216 @defmac EH_RETURN_HANDLER_RTX
3217 A C expression whose value is RTL representing a location in which
3218 to store the address of an exception handler to which we should
3219 return. It will not be assigned on code paths that return normally.
3221 Typically this is the location in the call frame at which the normal
3222 return address is stored. For targets that return by popping an
3223 address off the stack, this might be a memory address just below
3224 the @emph{target} call frame rather than inside the current call
3225 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3226 been assigned, so it may be used to calculate the location of the
3229 Some targets have more complex requirements than storing to an
3230 address calculable during initial code generation. In that case
3231 the @code{eh_return} instruction pattern should be used instead.
3233 If you want to support call frame exception handling, you must
3234 define either this macro or the @code{eh_return} instruction pattern.
3237 @defmac RETURN_ADDR_OFFSET
3238 If defined, an integer-valued C expression for which rtl will be generated
3239 to add it to the exception handler address before it is searched in the
3240 exception handling tables, and to subtract it again from the address before
3241 using it to return to the exception handler.
3244 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3245 This macro chooses the encoding of pointers embedded in the exception
3246 handling sections. If at all possible, this should be defined such
3247 that the exception handling section will not require dynamic relocations,
3248 and so may be read-only.
3250 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3251 @var{global} is true if the symbol may be affected by dynamic relocations.
3252 The macro should return a combination of the @code{DW_EH_PE_*} defines
3253 as found in @file{dwarf2.h}.
3255 If this macro is not defined, pointers will not be encoded but
3256 represented directly.
3259 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3260 This macro allows the target to emit whatever special magic is required
3261 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3262 Generic code takes care of pc-relative and indirect encodings; this must
3263 be defined if the target uses text-relative or data-relative encodings.
3265 This is a C statement that branches to @var{done} if the format was
3266 handled. @var{encoding} is the format chosen, @var{size} is the number
3267 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3271 @defmac MD_UNWIND_SUPPORT
3272 A string specifying a file to be #include'd in unwind-dw2.c. The file
3273 so included typically defines @code{MD_FALLBACK_FRAME_STATE_FOR}.
3276 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3277 This macro allows the target to add cpu and operating system specific
3278 code to the call-frame unwinder for use when there is no unwind data
3279 available. The most common reason to implement this macro is to unwind
3280 through signal frames.
3282 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
3283 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3284 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3285 for the address of the code being executed and @code{context->cfa} for
3286 the stack pointer value. If the frame can be decoded, the register save
3287 addresses should be updated in @var{fs} and the macro should evaluate to
3288 @code{_URC_NO_REASON}. If the frame cannot be decoded, the macro should
3289 evaluate to @code{_URC_END_OF_STACK}.
3291 For proper signal handling in Java this macro is accompanied by
3292 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3295 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3296 This macro allows the target to add operating system specific code to the
3297 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3298 usually used for signal or interrupt frames.
3300 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3301 @var{context} is an @code{_Unwind_Context};
3302 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3303 for the abi and context in the @code{.unwabi} directive. If the
3304 @code{.unwabi} directive can be handled, the register save addresses should
3305 be updated in @var{fs}.
3308 @defmac TARGET_USES_WEAK_UNWIND_INFO
3309 A C expression that evaluates to true if the target requires unwind
3310 info to be given comdat linkage. Define it to be @code{1} if comdat
3311 linkage is necessary. The default is @code{0}.
3314 @node Stack Checking
3315 @subsection Specifying How Stack Checking is Done
3317 GCC will check that stack references are within the boundaries of
3318 the stack, if the @option{-fstack-check} is specified, in one of three ways:
3322 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3323 will assume that you have arranged for stack checking to be done at
3324 appropriate places in the configuration files, e.g., in
3325 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
3329 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
3330 called @code{check_stack} in your @file{md} file, GCC will call that
3331 pattern with one argument which is the address to compare the stack
3332 value against. You must arrange for this pattern to report an error if
3333 the stack pointer is out of range.
3336 If neither of the above are true, GCC will generate code to periodically
3337 ``probe'' the stack pointer using the values of the macros defined below.
3340 Normally, you will use the default values of these macros, so GCC
3341 will use the third approach.
3343 @defmac STACK_CHECK_BUILTIN
3344 A nonzero value if stack checking is done by the configuration files in a
3345 machine-dependent manner. You should define this macro if stack checking
3346 is require by the ABI of your machine or if you would like to have to stack
3347 checking in some more efficient way than GCC's portable approach.
3348 The default value of this macro is zero.
3351 @defmac STACK_CHECK_PROBE_INTERVAL
3352 An integer representing the interval at which GCC must generate stack
3353 probe instructions. You will normally define this macro to be no larger
3354 than the size of the ``guard pages'' at the end of a stack area. The
3355 default value of 4096 is suitable for most systems.
3358 @defmac STACK_CHECK_PROBE_LOAD
3359 A integer which is nonzero if GCC should perform the stack probe
3360 as a load instruction and zero if GCC should use a store instruction.
3361 The default is zero, which is the most efficient choice on most systems.
3364 @defmac STACK_CHECK_PROTECT
3365 The number of bytes of stack needed to recover from a stack overflow,
3366 for languages where such a recovery is supported. The default value of
3367 75 words should be adequate for most machines.
3370 @defmac STACK_CHECK_MAX_FRAME_SIZE
3371 The maximum size of a stack frame, in bytes. GCC will generate probe
3372 instructions in non-leaf functions to ensure at least this many bytes of
3373 stack are available. If a stack frame is larger than this size, stack
3374 checking will not be reliable and GCC will issue a warning. The
3375 default is chosen so that GCC only generates one instruction on most
3376 systems. You should normally not change the default value of this macro.
3379 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3380 GCC uses this value to generate the above warning message. It
3381 represents the amount of fixed frame used by a function, not including
3382 space for any callee-saved registers, temporaries and user variables.
3383 You need only specify an upper bound for this amount and will normally
3384 use the default of four words.
3387 @defmac STACK_CHECK_MAX_VAR_SIZE
3388 The maximum size, in bytes, of an object that GCC will place in the
3389 fixed area of the stack frame when the user specifies
3390 @option{-fstack-check}.
3391 GCC computed the default from the values of the above macros and you will
3392 normally not need to override that default.
3396 @node Frame Registers
3397 @subsection Registers That Address the Stack Frame
3399 @c prevent bad page break with this line
3400 This discusses registers that address the stack frame.
3402 @defmac STACK_POINTER_REGNUM
3403 The register number of the stack pointer register, which must also be a
3404 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3405 the hardware determines which register this is.
3408 @defmac FRAME_POINTER_REGNUM
3409 The register number of the frame pointer register, which is used to
3410 access automatic variables in the stack frame. On some machines, the
3411 hardware determines which register this is. On other machines, you can
3412 choose any register you wish for this purpose.
3415 @defmac HARD_FRAME_POINTER_REGNUM
3416 On some machines the offset between the frame pointer and starting
3417 offset of the automatic variables is not known until after register
3418 allocation has been done (for example, because the saved registers are
3419 between these two locations). On those machines, define
3420 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3421 be used internally until the offset is known, and define
3422 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3423 used for the frame pointer.
3425 You should define this macro only in the very rare circumstances when it
3426 is not possible to calculate the offset between the frame pointer and
3427 the automatic variables until after register allocation has been
3428 completed. When this macro is defined, you must also indicate in your
3429 definition of @code{ELIMINABLE_REGS} how to eliminate
3430 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3431 or @code{STACK_POINTER_REGNUM}.
3433 Do not define this macro if it would be the same as
3434 @code{FRAME_POINTER_REGNUM}.
3437 @defmac ARG_POINTER_REGNUM
3438 The register number of the arg pointer register, which is used to access
3439 the function's argument list. On some machines, this is the same as the
3440 frame pointer register. On some machines, the hardware determines which
3441 register this is. On other machines, you can choose any register you
3442 wish for this purpose. If this is not the same register as the frame
3443 pointer register, then you must mark it as a fixed register according to
3444 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3445 (@pxref{Elimination}).
3448 @defmac RETURN_ADDRESS_POINTER_REGNUM
3449 The register number of the return address pointer register, which is used to
3450 access the current function's return address from the stack. On some
3451 machines, the return address is not at a fixed offset from the frame
3452 pointer or stack pointer or argument pointer. This register can be defined
3453 to point to the return address on the stack, and then be converted by
3454 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3456 Do not define this macro unless there is no other way to get the return
3457 address from the stack.
3460 @defmac STATIC_CHAIN_REGNUM
3461 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3462 Register numbers used for passing a function's static chain pointer. If
3463 register windows are used, the register number as seen by the called
3464 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3465 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3466 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3469 The static chain register need not be a fixed register.
3471 If the static chain is passed in memory, these macros should not be
3472 defined; instead, the next two macros should be defined.
3475 @defmac STATIC_CHAIN
3476 @defmacx STATIC_CHAIN_INCOMING
3477 If the static chain is passed in memory, these macros provide rtx giving
3478 @code{mem} expressions that denote where they are stored.
3479 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3480 as seen by the calling and called functions, respectively. Often the former
3481 will be at an offset from the stack pointer and the latter at an offset from
3484 @findex stack_pointer_rtx
3485 @findex frame_pointer_rtx
3486 @findex arg_pointer_rtx
3487 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3488 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3489 macros and should be used to refer to those items.
3491 If the static chain is passed in a register, the two previous macros should
3495 @defmac DWARF_FRAME_REGISTERS
3496 This macro specifies the maximum number of hard registers that can be
3497 saved in a call frame. This is used to size data structures used in
3498 DWARF2 exception handling.
3500 Prior to GCC 3.0, this macro was needed in order to establish a stable
3501 exception handling ABI in the face of adding new hard registers for ISA
3502 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3503 in the number of hard registers. Nevertheless, this macro can still be
3504 used to reduce the runtime memory requirements of the exception handling
3505 routines, which can be substantial if the ISA contains a lot of
3506 registers that are not call-saved.
3508 If this macro is not defined, it defaults to
3509 @code{FIRST_PSEUDO_REGISTER}.
3512 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3514 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3515 for backward compatibility in pre GCC 3.0 compiled code.
3517 If this macro is not defined, it defaults to
3518 @code{DWARF_FRAME_REGISTERS}.
3521 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3523 Define this macro if the target's representation for dwarf registers
3524 is different than the internal representation for unwind column.
3525 Given a dwarf register, this macro should return the internal unwind
3526 column number to use instead.
3528 See the PowerPC's SPE target for an example.
3531 @defmac DWARF_FRAME_REGNUM (@var{regno})
3533 Define this macro if the target's representation for dwarf registers
3534 used in .eh_frame or .debug_frame is different from that used in other
3535 debug info sections. Given a GCC hard register number, this macro
3536 should return the .eh_frame register number. The default is
3537 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3541 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3543 Define this macro to map register numbers held in the call frame info
3544 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3545 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3546 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3547 return @code{@var{regno}}.
3552 @subsection Eliminating Frame Pointer and Arg Pointer
3554 @c prevent bad page break with this line
3555 This is about eliminating the frame pointer and arg pointer.
3557 @defmac FRAME_POINTER_REQUIRED
3558 A C expression which is nonzero if a function must have and use a frame
3559 pointer. This expression is evaluated in the reload pass. If its value is
3560 nonzero the function will have a frame pointer.
3562 The expression can in principle examine the current function and decide
3563 according to the facts, but on most machines the constant 0 or the
3564 constant 1 suffices. Use 0 when the machine allows code to be generated
3565 with no frame pointer, and doing so saves some time or space. Use 1
3566 when there is no possible advantage to avoiding a frame pointer.
3568 In certain cases, the compiler does not know how to produce valid code
3569 without a frame pointer. The compiler recognizes those cases and
3570 automatically gives the function a frame pointer regardless of what
3571 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3574 In a function that does not require a frame pointer, the frame pointer
3575 register can be allocated for ordinary usage, unless you mark it as a
3576 fixed register. See @code{FIXED_REGISTERS} for more information.
3579 @findex get_frame_size
3580 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3581 A C statement to store in the variable @var{depth-var} the difference
3582 between the frame pointer and the stack pointer values immediately after
3583 the function prologue. The value would be computed from information
3584 such as the result of @code{get_frame_size ()} and the tables of
3585 registers @code{regs_ever_live} and @code{call_used_regs}.
3587 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3588 need not be defined. Otherwise, it must be defined even if
3589 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3590 case, you may set @var{depth-var} to anything.
3593 @defmac ELIMINABLE_REGS
3594 If defined, this macro specifies a table of register pairs used to
3595 eliminate unneeded registers that point into the stack frame. If it is not
3596 defined, the only elimination attempted by the compiler is to replace
3597 references to the frame pointer with references to the stack pointer.
3599 The definition of this macro is a list of structure initializations, each
3600 of which specifies an original and replacement register.
3602 On some machines, the position of the argument pointer is not known until
3603 the compilation is completed. In such a case, a separate hard register
3604 must be used for the argument pointer. This register can be eliminated by
3605 replacing it with either the frame pointer or the argument pointer,
3606 depending on whether or not the frame pointer has been eliminated.
3608 In this case, you might specify:
3610 #define ELIMINABLE_REGS \
3611 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3612 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3613 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3616 Note that the elimination of the argument pointer with the stack pointer is
3617 specified first since that is the preferred elimination.
3620 @defmac CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3621 A C expression that returns nonzero if the compiler is allowed to try
3622 to replace register number @var{from-reg} with register number
3623 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3624 is defined, and will usually be the constant 1, since most of the cases
3625 preventing register elimination are things that the compiler already
3629 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3630 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3631 specifies the initial difference between the specified pair of
3632 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3636 @node Stack Arguments
3637 @subsection Passing Function Arguments on the Stack
3638 @cindex arguments on stack
3639 @cindex stack arguments
3641 The macros in this section control how arguments are passed
3642 on the stack. See the following section for other macros that
3643 control passing certain arguments in registers.
3645 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (tree @var{fntype})
3646 This target hook returns @code{true} if an argument declared in a
3647 prototype as an integral type smaller than @code{int} should actually be
3648 passed as an @code{int}. In addition to avoiding errors in certain
3649 cases of mismatch, it also makes for better code on certain machines.
3650 The default is to not promote prototypes.
3654 A C expression. If nonzero, push insns will be used to pass
3656 If the target machine does not have a push instruction, set it to zero.
3657 That directs GCC to use an alternate strategy: to
3658 allocate the entire argument block and then store the arguments into
3659 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3662 @defmac PUSH_ARGS_REVERSED
3663 A C expression. If nonzero, function arguments will be evaluated from
3664 last to first, rather than from first to last. If this macro is not
3665 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3666 and args grow in opposite directions, and 0 otherwise.
3669 @defmac PUSH_ROUNDING (@var{npushed})
3670 A C expression that is the number of bytes actually pushed onto the
3671 stack when an instruction attempts to push @var{npushed} bytes.
3673 On some machines, the definition
3676 #define PUSH_ROUNDING(BYTES) (BYTES)
3680 will suffice. But on other machines, instructions that appear
3681 to push one byte actually push two bytes in an attempt to maintain
3682 alignment. Then the definition should be
3685 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3689 @findex current_function_outgoing_args_size
3690 @defmac ACCUMULATE_OUTGOING_ARGS
3691 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3692 will be computed and placed into the variable
3693 @code{current_function_outgoing_args_size}. No space will be pushed
3694 onto the stack for each call; instead, the function prologue should
3695 increase the stack frame size by this amount.
3697 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3701 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3702 Define this macro if functions should assume that stack space has been
3703 allocated for arguments even when their values are passed in
3706 The value of this macro is the size, in bytes, of the area reserved for
3707 arguments passed in registers for the function represented by @var{fndecl},
3708 which can be zero if GCC is calling a library function.
3710 This space can be allocated by the caller, or be a part of the
3711 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3714 @c above is overfull. not sure what to do. --mew 5feb93 did
3715 @c something, not sure if it looks good. --mew 10feb93
3717 @defmac OUTGOING_REG_PARM_STACK_SPACE
3718 Define this if it is the responsibility of the caller to allocate the area
3719 reserved for arguments passed in registers.
3721 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3722 whether the space for these arguments counts in the value of
3723 @code{current_function_outgoing_args_size}.
3726 @defmac STACK_PARMS_IN_REG_PARM_AREA
3727 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3728 stack parameters don't skip the area specified by it.
3729 @c i changed this, makes more sens and it should have taken care of the
3730 @c overfull.. not as specific, tho. --mew 5feb93
3732 Normally, when a parameter is not passed in registers, it is placed on the
3733 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3734 suppresses this behavior and causes the parameter to be passed on the
3735 stack in its natural location.
3738 @defmac RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3739 A C expression that should indicate the number of bytes of its own
3740 arguments that a function pops on returning, or 0 if the
3741 function pops no arguments and the caller must therefore pop them all
3742 after the function returns.
3744 @var{fundecl} is a C variable whose value is a tree node that describes
3745 the function in question. Normally it is a node of type
3746 @code{FUNCTION_DECL} that describes the declaration of the function.
3747 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3749 @var{funtype} is a C variable whose value is a tree node that
3750 describes the function in question. Normally it is a node of type
3751 @code{FUNCTION_TYPE} that describes the data type of the function.
3752 From this it is possible to obtain the data types of the value and
3753 arguments (if known).
3755 When a call to a library function is being considered, @var{fundecl}
3756 will contain an identifier node for the library function. Thus, if
3757 you need to distinguish among various library functions, you can do so
3758 by their names. Note that ``library function'' in this context means
3759 a function used to perform arithmetic, whose name is known specially
3760 in the compiler and was not mentioned in the C code being compiled.
3762 @var{stack-size} is the number of bytes of arguments passed on the
3763 stack. If a variable number of bytes is passed, it is zero, and
3764 argument popping will always be the responsibility of the calling function.
3766 On the VAX, all functions always pop their arguments, so the definition
3767 of this macro is @var{stack-size}. On the 68000, using the standard
3768 calling convention, no functions pop their arguments, so the value of
3769 the macro is always 0 in this case. But an alternative calling
3770 convention is available in which functions that take a fixed number of
3771 arguments pop them but other functions (such as @code{printf}) pop
3772 nothing (the caller pops all). When this convention is in use,
3773 @var{funtype} is examined to determine whether a function takes a fixed
3774 number of arguments.
3777 @defmac CALL_POPS_ARGS (@var{cum})
3778 A C expression that should indicate the number of bytes a call sequence
3779 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3780 when compiling a function call.
3782 @var{cum} is the variable in which all arguments to the called function
3783 have been accumulated.
3785 On certain architectures, such as the SH5, a call trampoline is used
3786 that pops certain registers off the stack, depending on the arguments
3787 that have been passed to the function. Since this is a property of the
3788 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3792 @node Register Arguments
3793 @subsection Passing Arguments in Registers
3794 @cindex arguments in registers
3795 @cindex registers arguments
3797 This section describes the macros which let you control how various
3798 types of arguments are passed in registers or how they are arranged in
3801 @defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3802 A C expression that controls whether a function argument is passed
3803 in a register, and which register.
3805 The arguments are @var{cum}, which summarizes all the previous
3806 arguments; @var{mode}, the machine mode of the argument; @var{type},
3807 the data type of the argument as a tree node or 0 if that is not known
3808 (which happens for C support library functions); and @var{named},
3809 which is 1 for an ordinary argument and 0 for nameless arguments that
3810 correspond to @samp{@dots{}} in the called function's prototype.
3811 @var{type} can be an incomplete type if a syntax error has previously
3814 The value of the expression is usually either a @code{reg} RTX for the
3815 hard register in which to pass the argument, or zero to pass the
3816 argument on the stack.
3818 For machines like the VAX and 68000, where normally all arguments are
3819 pushed, zero suffices as a definition.
3821 The value of the expression can also be a @code{parallel} RTX@. This is
3822 used when an argument is passed in multiple locations. The mode of the
3823 @code{parallel} should be the mode of the entire argument. The
3824 @code{parallel} holds any number of @code{expr_list} pairs; each one
3825 describes where part of the argument is passed. In each
3826 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3827 register in which to pass this part of the argument, and the mode of the
3828 register RTX indicates how large this part of the argument is. The
3829 second operand of the @code{expr_list} is a @code{const_int} which gives
3830 the offset in bytes into the entire argument of where this part starts.
3831 As a special exception the first @code{expr_list} in the @code{parallel}
3832 RTX may have a first operand of zero. This indicates that the entire
3833 argument is also stored on the stack.
3835 The last time this macro is called, it is called with @code{MODE ==
3836 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3837 pattern as operands 2 and 3 respectively.
3839 @cindex @file{stdarg.h} and register arguments
3840 The usual way to make the ISO library @file{stdarg.h} work on a machine
3841 where some arguments are usually passed in registers, is to cause
3842 nameless arguments to be passed on the stack instead. This is done
3843 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3845 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3846 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3847 You may use the hook @code{targetm.calls.must_pass_in_stack}
3848 in the definition of this macro to determine if this argument is of a
3849 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3850 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3851 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3852 defined, the argument will be computed in the stack and then loaded into
3856 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, tree @var{type})
3857 This target hook should return @code{true} if we should not pass @var{type}
3858 solely in registers. The file @file{expr.h} defines a
3859 definition that is usually appropriate, refer to @file{expr.h} for additional
3863 @defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3864 Define this macro if the target machine has ``register windows'', so
3865 that the register in which a function sees an arguments is not
3866 necessarily the same as the one in which the caller passed the
3869 For such machines, @code{FUNCTION_ARG} computes the register in which
3870 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3871 be defined in a similar fashion to tell the function being called
3872 where the arguments will arrive.
3874 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3875 serves both purposes.
3878 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3879 This target hook returns the number of bytes at the beginning of an
3880 argument that must be put in registers. The value must be zero for
3881 arguments that are passed entirely in registers or that are entirely
3882 pushed on the stack.
3884 On some machines, certain arguments must be passed partially in
3885 registers and partially in memory. On these machines, typically the
3886 first few words of arguments are passed in registers, and the rest
3887 on the stack. If a multi-word argument (a @code{double} or a
3888 structure) crosses that boundary, its first few words must be passed
3889 in registers and the rest must be pushed. This macro tells the
3890 compiler when this occurs, and how many bytes should go in registers.
3892 @code{FUNCTION_ARG} for these arguments should return the first
3893 register to be used by the caller for this argument; likewise
3894 @code{FUNCTION_INCOMING_ARG}, for the called function.
3897 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3898 This target hook should return @code{true} if an argument at the
3899 position indicated by @var{cum} should be passed by reference. This
3900 predicate is queried after target independent reasons for being
3901 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
3903 If the hook returns true, a copy of that argument is made in memory and a
3904 pointer to the argument is passed instead of the argument itself.
3905 The pointer is passed in whatever way is appropriate for passing a pointer
3909 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3910 The function argument described by the parameters to this hook is
3911 known to be passed by reference. The hook should return true if the
3912 function argument should be copied by the callee instead of copied
3915 For any argument for which the hook returns true, if it can be
3916 determined that the argument is not modified, then a copy need
3919 The default version of this hook always returns false.
3922 @defmac CUMULATIVE_ARGS
3923 A C type for declaring a variable that is used as the first argument of
3924 @code{FUNCTION_ARG} and other related values. For some target machines,
3925 the type @code{int} suffices and can hold the number of bytes of
3928 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3929 arguments that have been passed on the stack. The compiler has other
3930 variables to keep track of that. For target machines on which all
3931 arguments are passed on the stack, there is no need to store anything in
3932 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3933 should not be empty, so use @code{int}.
3936 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
3937 A C statement (sans semicolon) for initializing the variable
3938 @var{cum} for the state at the beginning of the argument list. The
3939 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
3940 is the tree node for the data type of the function which will receive
3941 the args, or 0 if the args are to a compiler support library function.
3942 For direct calls that are not libcalls, @var{fndecl} contain the
3943 declaration node of the function. @var{fndecl} is also set when
3944 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3945 being compiled. @var{n_named_args} is set to the number of named
3946 arguments, including a structure return address if it is passed as a
3947 parameter, when making a call. When processing incoming arguments,
3948 @var{n_named_args} is set to @minus{}1.
3950 When processing a call to a compiler support library function,
3951 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3952 contains the name of the function, as a string. @var{libname} is 0 when
3953 an ordinary C function call is being processed. Thus, each time this
3954 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3955 never both of them at once.
3958 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3959 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3960 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3961 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3962 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3963 0)} is used instead.
3966 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3967 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3968 finding the arguments for the function being compiled. If this macro is
3969 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3971 The value passed for @var{libname} is always 0, since library routines
3972 with special calling conventions are never compiled with GCC@. The
3973 argument @var{libname} exists for symmetry with
3974 @code{INIT_CUMULATIVE_ARGS}.
3975 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3976 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3979 @defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3980 A C statement (sans semicolon) to update the summarizer variable
3981 @var{cum} to advance past an argument in the argument list. The
3982 values @var{mode}, @var{type} and @var{named} describe that argument.
3983 Once this is done, the variable @var{cum} is suitable for analyzing
3984 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
3986 This macro need not do anything if the argument in question was passed
3987 on the stack. The compiler knows how to track the amount of stack space
3988 used for arguments without any special help.
3991 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3992 If defined, a C expression which determines whether, and in which direction,
3993 to pad out an argument with extra space. The value should be of type
3994 @code{enum direction}: either @code{upward} to pad above the argument,
3995 @code{downward} to pad below, or @code{none} to inhibit padding.
3997 The @emph{amount} of padding is always just enough to reach the next
3998 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
4001 This macro has a default definition which is right for most systems.
4002 For little-endian machines, the default is to pad upward. For
4003 big-endian machines, the default is to pad downward for an argument of
4004 constant size shorter than an @code{int}, and upward otherwise.
4007 @defmac PAD_VARARGS_DOWN
4008 If defined, a C expression which determines whether the default
4009 implementation of va_arg will attempt to pad down before reading the
4010 next argument, if that argument is smaller than its aligned space as
4011 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4012 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4015 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4016 Specify padding for the last element of a block move between registers and
4017 memory. @var{first} is nonzero if this is the only element. Defining this
4018 macro allows better control of register function parameters on big-endian
4019 machines, without using @code{PARALLEL} rtl. In particular,
4020 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4021 registers, as there is no longer a "wrong" part of a register; For example,
4022 a three byte aggregate may be passed in the high part of a register if so
4026 @defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
4027 If defined, a C expression that gives the alignment boundary, in bits,
4028 of an argument with the specified mode and type. If it is not defined,
4029 @code{PARM_BOUNDARY} is used for all arguments.
4032 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4033 A C expression that is nonzero if @var{regno} is the number of a hard
4034 register in which function arguments are sometimes passed. This does
4035 @emph{not} include implicit arguments such as the static chain and
4036 the structure-value address. On many machines, no registers can be
4037 used for this purpose since all function arguments are pushed on the
4041 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (tree @var{type})
4042 This hook should return true if parameter of type @var{type} are passed
4043 as two scalar parameters. By default, GCC will attempt to pack complex
4044 arguments into the target's word size. Some ABIs require complex arguments
4045 to be split and treated as their individual components. For example, on
4046 AIX64, complex floats should be passed in a pair of floating point
4047 registers, even though a complex float would fit in one 64-bit floating
4050 The default value of this hook is @code{NULL}, which is treated as always
4054 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
4055 This hook returns a type node for @code{va_list} for the target.
4056 The default version of the hook returns @code{void*}.
4059 @deftypefn {Target Hook} tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree @var{valist}, tree @var{type}, tree *@var{pre_p}, tree *@var{post_p})
4060 This hook performs target-specific gimplification of
4061 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4062 arguments to @code{va_arg}; the latter two are as in
4063 @code{gimplify.c:gimplify_expr}.
4066 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (enum machine_mode @var{mode})
4067 Define this to return nonzero if the port can handle pointers
4068 with machine mode @var{mode}. The default version of this
4069 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4072 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4073 Define this to return nonzero if the port is prepared to handle
4074 insns involving scalar mode @var{mode}. For a scalar mode to be
4075 considered supported, all the basic arithmetic and comparisons
4078 The default version of this hook returns true for any mode
4079 required to handle the basic C types (as defined by the port).
4080 Included here are the double-word arithmetic supported by the
4081 code in @file{optabs.c}.
4084 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4085 Define this to return nonzero if the port is prepared to handle
4086 insns involving vector mode @var{mode}. At the very least, it
4087 must have move patterns for this mode.
4091 @subsection How Scalar Function Values Are Returned
4092 @cindex return values in registers
4093 @cindex values, returned by functions
4094 @cindex scalars, returned as values
4096 This section discusses the macros that control returning scalars as
4097 values---values that can fit in registers.
4099 @deftypefn {Target Hook} rtx TARGET_FUNCTION_VALUE (tree @var{ret_type}, tree @var{fn_decl_or_type}, bool @var{outgoing})
4101 Define this to return an RTX representing the place where a function
4102 returns or receives a value of data type @var{ret_type}, a tree node
4103 node representing a data type. @var{fn_decl_or_type} is a tree node
4104 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4105 function being called. If @var{outgoing} is false, the hook should
4106 compute the register in which the caller will see the return value.
4107 Otherwise, the hook should return an RTX representing the place where
4108 a function returns a value.
4110 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4111 (Actually, on most machines, scalar values are returned in the same
4112 place regardless of mode.) The value of the expression is usually a
4113 @code{reg} RTX for the hard register where the return value is stored.
4114 The value can also be a @code{parallel} RTX, if the return value is in
4115 multiple places. See @code{FUNCTION_ARG} for an explanation of the
4116 @code{parallel} form. Note that the callee will populate every
4117 location specified in the @code{parallel}, but if the first element of
4118 the @code{parallel} contains the whole return value, callers will use
4119 that element as the canonical location and ignore the others. The m68k
4120 port uses this type of @code{parallel} to return pointers in both
4121 @samp{%a0} (the canonical location) and @samp{%d0}.
4123 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4124 the same promotion rules specified in @code{PROMOTE_MODE} if
4125 @var{valtype} is a scalar type.
4127 If the precise function being called is known, @var{func} is a tree
4128 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4129 pointer. This makes it possible to use a different value-returning
4130 convention for specific functions when all their calls are
4133 Some target machines have ``register windows'' so that the register in
4134 which a function returns its value is not the same as the one in which
4135 the caller sees the value. For such machines, you should return
4136 different RTX depending on @var{outgoing}.
4138 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4139 aggregate data types, because these are returned in another way. See
4140 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4143 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4144 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4145 a new target instead.
4148 @defmac FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
4149 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4150 a new target instead.
4153 @defmac LIBCALL_VALUE (@var{mode})
4154 A C expression to create an RTX representing the place where a library
4155 function returns a value of mode @var{mode}. If the precise function
4156 being called is known, @var{func} is a tree node
4157 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4158 pointer. This makes it possible to use a different value-returning
4159 convention for specific functions when all their calls are
4162 Note that ``library function'' in this context means a compiler
4163 support routine, used to perform arithmetic, whose name is known
4164 specially by the compiler and was not mentioned in the C code being
4167 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
4168 data types, because none of the library functions returns such types.
4171 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4172 A C expression that is nonzero if @var{regno} is the number of a hard
4173 register in which the values of called function may come back.
4175 A register whose use for returning values is limited to serving as the
4176 second of a pair (for a value of type @code{double}, say) need not be
4177 recognized by this macro. So for most machines, this definition
4181 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4184 If the machine has register windows, so that the caller and the called
4185 function use different registers for the return value, this macro
4186 should recognize only the caller's register numbers.
4189 @defmac APPLY_RESULT_SIZE
4190 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4191 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4192 saving and restoring an arbitrary return value.
4195 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (tree @var{type})
4196 This hook should return true if values of type @var{type} are returned
4197 at the most significant end of a register (in other words, if they are
4198 padded at the least significant end). You can assume that @var{type}
4199 is returned in a register; the caller is required to check this.
4201 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4202 be able to hold the complete return value. For example, if a 1-, 2-
4203 or 3-byte structure is returned at the most significant end of a
4204 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4208 @node Aggregate Return
4209 @subsection How Large Values Are Returned
4210 @cindex aggregates as return values
4211 @cindex large return values
4212 @cindex returning aggregate values
4213 @cindex structure value address
4215 When a function value's mode is @code{BLKmode} (and in some other
4216 cases), the value is not returned according to
4217 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4218 caller passes the address of a block of memory in which the value
4219 should be stored. This address is called the @dfn{structure value
4222 This section describes how to control returning structure values in
4225 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (tree @var{type}, tree @var{fntype})
4226 This target hook should return a nonzero value to say to return the
4227 function value in memory, just as large structures are always returned.
4228 Here @var{type} will be the data type of the value, and @var{fntype}
4229 will be the type of the function doing the returning, or @code{NULL} for
4232 Note that values of mode @code{BLKmode} must be explicitly handled
4233 by this function. Also, the option @option{-fpcc-struct-return}
4234 takes effect regardless of this macro. On most systems, it is
4235 possible to leave the hook undefined; this causes a default
4236 definition to be used, whose value is the constant 1 for @code{BLKmode}
4237 values, and 0 otherwise.
4239 Do not use this hook to indicate that structures and unions should always
4240 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4244 @defmac DEFAULT_PCC_STRUCT_RETURN
4245 Define this macro to be 1 if all structure and union return values must be
4246 in memory. Since this results in slower code, this should be defined
4247 only if needed for compatibility with other compilers or with an ABI@.
4248 If you define this macro to be 0, then the conventions used for structure
4249 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4252 If not defined, this defaults to the value 1.
4255 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4256 This target hook should return the location of the structure value
4257 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4258 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4259 be @code{NULL}, for libcalls. You do not need to define this target
4260 hook if the address is always passed as an ``invisible'' first
4263 On some architectures the place where the structure value address
4264 is found by the called function is not the same place that the
4265 caller put it. This can be due to register windows, or it could
4266 be because the function prologue moves it to a different place.
4267 @var{incoming} is @code{1} or @code{2} when the location is needed in
4268 the context of the called function, and @code{0} in the context of
4271 If @var{incoming} is nonzero and the address is to be found on the
4272 stack, return a @code{mem} which refers to the frame pointer. If
4273 @var{incoming} is @code{2}, the result is being used to fetch the
4274 structure value address at the beginning of a function. If you need
4275 to emit adjusting code, you should do it at this point.
4278 @defmac PCC_STATIC_STRUCT_RETURN
4279 Define this macro if the usual system convention on the target machine
4280 for returning structures and unions is for the called function to return
4281 the address of a static variable containing the value.
4283 Do not define this if the usual system convention is for the caller to
4284 pass an address to the subroutine.
4286 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4287 nothing when you use @option{-freg-struct-return} mode.
4291 @subsection Caller-Saves Register Allocation
4293 If you enable it, GCC can save registers around function calls. This
4294 makes it possible to use call-clobbered registers to hold variables that
4295 must live across calls.
4297 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4298 A C expression to determine whether it is worthwhile to consider placing
4299 a pseudo-register in a call-clobbered hard register and saving and
4300 restoring it around each function call. The expression should be 1 when
4301 this is worth doing, and 0 otherwise.
4303 If you don't define this macro, a default is used which is good on most
4304 machines: @code{4 * @var{calls} < @var{refs}}.
4307 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4308 A C expression specifying which mode is required for saving @var{nregs}
4309 of a pseudo-register in call-clobbered hard register @var{regno}. If
4310 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4311 returned. For most machines this macro need not be defined since GCC
4312 will select the smallest suitable mode.
4315 @node Function Entry
4316 @subsection Function Entry and Exit
4317 @cindex function entry and exit
4321 This section describes the macros that output function entry
4322 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4324 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4325 If defined, a function that outputs the assembler code for entry to a
4326 function. The prologue is responsible for setting up the stack frame,
4327 initializing the frame pointer register, saving registers that must be
4328 saved, and allocating @var{size} additional bytes of storage for the
4329 local variables. @var{size} is an integer. @var{file} is a stdio
4330 stream to which the assembler code should be output.
4332 The label for the beginning of the function need not be output by this
4333 macro. That has already been done when the macro is run.
4335 @findex regs_ever_live
4336 To determine which registers to save, the macro can refer to the array
4337 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4338 @var{r} is used anywhere within the function. This implies the function
4339 prologue should save register @var{r}, provided it is not one of the
4340 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4341 @code{regs_ever_live}.)
4343 On machines that have ``register windows'', the function entry code does
4344 not save on the stack the registers that are in the windows, even if
4345 they are supposed to be preserved by function calls; instead it takes
4346 appropriate steps to ``push'' the register stack, if any non-call-used
4347 registers are used in the function.
4349 @findex frame_pointer_needed
4350 On machines where functions may or may not have frame-pointers, the
4351 function entry code must vary accordingly; it must set up the frame
4352 pointer if one is wanted, and not otherwise. To determine whether a
4353 frame pointer is in wanted, the macro can refer to the variable
4354 @code{frame_pointer_needed}. The variable's value will be 1 at run
4355 time in a function that needs a frame pointer. @xref{Elimination}.
4357 The function entry code is responsible for allocating any stack space
4358 required for the function. This stack space consists of the regions
4359 listed below. In most cases, these regions are allocated in the
4360 order listed, with the last listed region closest to the top of the
4361 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4362 the highest address if it is not defined). You can use a different order
4363 for a machine if doing so is more convenient or required for
4364 compatibility reasons. Except in cases where required by standard
4365 or by a debugger, there is no reason why the stack layout used by GCC
4366 need agree with that used by other compilers for a machine.
4369 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4370 If defined, a function that outputs assembler code at the end of a
4371 prologue. This should be used when the function prologue is being
4372 emitted as RTL, and you have some extra assembler that needs to be
4373 emitted. @xref{prologue instruction pattern}.
4376 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4377 If defined, a function that outputs assembler code at the start of an
4378 epilogue. This should be used when the function epilogue is being
4379 emitted as RTL, and you have some extra assembler that needs to be
4380 emitted. @xref{epilogue instruction pattern}.
4383 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4384 If defined, a function that outputs the assembler code for exit from a
4385 function. The epilogue is responsible for restoring the saved
4386 registers and stack pointer to their values when the function was
4387 called, and returning control to the caller. This macro takes the
4388 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4389 registers to restore are determined from @code{regs_ever_live} and
4390 @code{CALL_USED_REGISTERS} in the same way.
4392 On some machines, there is a single instruction that does all the work
4393 of returning from the function. On these machines, give that
4394 instruction the name @samp{return} and do not define the macro
4395 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4397 Do not define a pattern named @samp{return} if you want the
4398 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4399 switches to control whether return instructions or epilogues are used,
4400 define a @samp{return} pattern with a validity condition that tests the
4401 target switches appropriately. If the @samp{return} pattern's validity
4402 condition is false, epilogues will be used.
4404 On machines where functions may or may not have frame-pointers, the
4405 function exit code must vary accordingly. Sometimes the code for these
4406 two cases is completely different. To determine whether a frame pointer
4407 is wanted, the macro can refer to the variable
4408 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4409 a function that needs a frame pointer.
4411 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4412 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4413 The C variable @code{current_function_is_leaf} is nonzero for such a
4414 function. @xref{Leaf Functions}.
4416 On some machines, some functions pop their arguments on exit while
4417 others leave that for the caller to do. For example, the 68020 when
4418 given @option{-mrtd} pops arguments in functions that take a fixed
4419 number of arguments.
4421 @findex current_function_pops_args
4422 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4423 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4424 needs to know what was decided. The variable that is called
4425 @code{current_function_pops_args} is the number of bytes of its
4426 arguments that a function should pop. @xref{Scalar Return}.
4427 @c what is the "its arguments" in the above sentence referring to, pray
4428 @c tell? --mew 5feb93
4433 @findex current_function_pretend_args_size
4434 A region of @code{current_function_pretend_args_size} bytes of
4435 uninitialized space just underneath the first argument arriving on the
4436 stack. (This may not be at the very start of the allocated stack region
4437 if the calling sequence has pushed anything else since pushing the stack
4438 arguments. But usually, on such machines, nothing else has been pushed
4439 yet, because the function prologue itself does all the pushing.) This
4440 region is used on machines where an argument may be passed partly in
4441 registers and partly in memory, and, in some cases to support the
4442 features in @code{<stdarg.h>}.
4445 An area of memory used to save certain registers used by the function.
4446 The size of this area, which may also include space for such things as
4447 the return address and pointers to previous stack frames, is
4448 machine-specific and usually depends on which registers have been used
4449 in the function. Machines with register windows often do not require
4453 A region of at least @var{size} bytes, possibly rounded up to an allocation
4454 boundary, to contain the local variables of the function. On some machines,
4455 this region and the save area may occur in the opposite order, with the
4456 save area closer to the top of the stack.
4459 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4460 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4461 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4462 argument lists of the function. @xref{Stack Arguments}.
4465 @defmac EXIT_IGNORE_STACK
4466 Define this macro as a C expression that is nonzero if the return
4467 instruction or the function epilogue ignores the value of the stack
4468 pointer; in other words, if it is safe to delete an instruction to
4469 adjust the stack pointer before a return from the function. The
4472 Note that this macro's value is relevant only for functions for which
4473 frame pointers are maintained. It is never safe to delete a final
4474 stack adjustment in a function that has no frame pointer, and the
4475 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4478 @defmac EPILOGUE_USES (@var{regno})
4479 Define this macro as a C expression that is nonzero for registers that are
4480 used by the epilogue or the @samp{return} pattern. The stack and frame
4481 pointer registers are already assumed to be used as needed.
4484 @defmac EH_USES (@var{regno})
4485 Define this macro as a C expression that is nonzero for registers that are
4486 used by the exception handling mechanism, and so should be considered live
4487 on entry to an exception edge.
4490 @defmac DELAY_SLOTS_FOR_EPILOGUE
4491 Define this macro if the function epilogue contains delay slots to which
4492 instructions from the rest of the function can be ``moved''. The
4493 definition should be a C expression whose value is an integer
4494 representing the number of delay slots there.
4497 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4498 A C expression that returns 1 if @var{insn} can be placed in delay
4499 slot number @var{n} of the epilogue.
4501 The argument @var{n} is an integer which identifies the delay slot now
4502 being considered (since different slots may have different rules of
4503 eligibility). It is never negative and is always less than the number
4504 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4505 If you reject a particular insn for a given delay slot, in principle, it
4506 may be reconsidered for a subsequent delay slot. Also, other insns may
4507 (at least in principle) be considered for the so far unfilled delay
4510 @findex current_function_epilogue_delay_list
4511 @findex final_scan_insn
4512 The insns accepted to fill the epilogue delay slots are put in an RTL
4513 list made with @code{insn_list} objects, stored in the variable
4514 @code{current_function_epilogue_delay_list}. The insn for the first
4515 delay slot comes first in the list. Your definition of the macro
4516 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4517 outputting the insns in this list, usually by calling
4518 @code{final_scan_insn}.
4520 You need not define this macro if you did not define
4521 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4524 @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})
4525 A function that outputs the assembler code for a thunk
4526 function, used to implement C++ virtual function calls with multiple
4527 inheritance. The thunk acts as a wrapper around a virtual function,
4528 adjusting the implicit object parameter before handing control off to
4531 First, emit code to add the integer @var{delta} to the location that
4532 contains the incoming first argument. Assume that this argument
4533 contains a pointer, and is the one used to pass the @code{this} pointer
4534 in C++. This is the incoming argument @emph{before} the function prologue,
4535 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4536 all other incoming arguments.
4538 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4539 made after adding @code{delta}. In particular, if @var{p} is the
4540 adjusted pointer, the following adjustment should be made:
4543 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4546 After the additions, emit code to jump to @var{function}, which is a
4547 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4548 not touch the return address. Hence returning from @var{FUNCTION} will
4549 return to whoever called the current @samp{thunk}.
4551 The effect must be as if @var{function} had been called directly with
4552 the adjusted first argument. This macro is responsible for emitting all
4553 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4554 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4556 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4557 have already been extracted from it.) It might possibly be useful on
4558 some targets, but probably not.
4560 If you do not define this macro, the target-independent code in the C++
4561 front end will generate a less efficient heavyweight thunk that calls
4562 @var{function} instead of jumping to it. The generic approach does
4563 not support varargs.
4566 @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})
4567 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4568 to output the assembler code for the thunk function specified by the
4569 arguments it is passed, and false otherwise. In the latter case, the
4570 generic approach will be used by the C++ front end, with the limitations
4575 @subsection Generating Code for Profiling
4576 @cindex profiling, code generation
4578 These macros will help you generate code for profiling.
4580 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4581 A C statement or compound statement to output to @var{file} some
4582 assembler code to call the profiling subroutine @code{mcount}.
4585 The details of how @code{mcount} expects to be called are determined by
4586 your operating system environment, not by GCC@. To figure them out,
4587 compile a small program for profiling using the system's installed C
4588 compiler and look at the assembler code that results.
4590 Older implementations of @code{mcount} expect the address of a counter
4591 variable to be loaded into some register. The name of this variable is
4592 @samp{LP} followed by the number @var{labelno}, so you would generate
4593 the name using @samp{LP%d} in a @code{fprintf}.
4596 @defmac PROFILE_HOOK
4597 A C statement or compound statement to output to @var{file} some assembly
4598 code to call the profiling subroutine @code{mcount} even the target does
4599 not support profiling.
4602 @defmac NO_PROFILE_COUNTERS
4603 Define this macro to be an expression with a nonzero value if the
4604 @code{mcount} subroutine on your system does not need a counter variable
4605 allocated for each function. This is true for almost all modern
4606 implementations. If you define this macro, you must not use the
4607 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4610 @defmac PROFILE_BEFORE_PROLOGUE
4611 Define this macro if the code for function profiling should come before
4612 the function prologue. Normally, the profiling code comes after.
4616 @subsection Permitting tail calls
4619 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4620 True if it is ok to do sibling call optimization for the specified
4621 call expression @var{exp}. @var{decl} will be the called function,
4622 or @code{NULL} if this is an indirect call.
4624 It is not uncommon for limitations of calling conventions to prevent
4625 tail calls to functions outside the current unit of translation, or
4626 during PIC compilation. The hook is used to enforce these restrictions,
4627 as the @code{sibcall} md pattern can not fail, or fall over to a
4628 ``normal'' call. The criteria for successful sibling call optimization
4629 may vary greatly between different architectures.
4632 @deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap *@var{regs})
4633 Add any hard registers to @var{regs} that are live on entry to the
4634 function. This hook only needs to be defined to provide registers that
4635 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4636 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4637 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4638 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4641 @node Stack Smashing Protection
4642 @subsection Stack smashing protection
4643 @cindex stack smashing protection
4645 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void)
4646 This hook returns a @code{DECL} node for the external variable to use
4647 for the stack protection guard. This variable is initialized by the
4648 runtime to some random value and is used to initialize the guard value
4649 that is placed at the top of the local stack frame. The type of this
4650 variable must be @code{ptr_type_node}.
4652 The default version of this hook creates a variable called
4653 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
4656 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void)
4657 This hook returns a tree expression that alerts the runtime that the
4658 stack protect guard variable has been modified. This expression should
4659 involve a call to a @code{noreturn} function.
4661 The default version of this hook invokes a function called
4662 @samp{__stack_chk_fail}, taking no arguments. This function is
4663 normally defined in @file{libgcc2.c}.
4667 @section Implementing the Varargs Macros
4668 @cindex varargs implementation
4670 GCC comes with an implementation of @code{<varargs.h>} and
4671 @code{<stdarg.h>} that work without change on machines that pass arguments
4672 on the stack. Other machines require their own implementations of
4673 varargs, and the two machine independent header files must have
4674 conditionals to include it.
4676 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4677 the calling convention for @code{va_start}. The traditional
4678 implementation takes just one argument, which is the variable in which
4679 to store the argument pointer. The ISO implementation of
4680 @code{va_start} takes an additional second argument. The user is
4681 supposed to write the last named argument of the function here.
4683 However, @code{va_start} should not use this argument. The way to find
4684 the end of the named arguments is with the built-in functions described
4687 @defmac __builtin_saveregs ()
4688 Use this built-in function to save the argument registers in memory so
4689 that the varargs mechanism can access them. Both ISO and traditional
4690 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4691 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4693 On some machines, @code{__builtin_saveregs} is open-coded under the
4694 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4695 other machines, it calls a routine written in assembler language,
4696 found in @file{libgcc2.c}.
4698 Code generated for the call to @code{__builtin_saveregs} appears at the
4699 beginning of the function, as opposed to where the call to
4700 @code{__builtin_saveregs} is written, regardless of what the code is.
4701 This is because the registers must be saved before the function starts
4702 to use them for its own purposes.
4703 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4707 @defmac __builtin_args_info (@var{category})
4708 Use this built-in function to find the first anonymous arguments in
4711 In general, a machine may have several categories of registers used for
4712 arguments, each for a particular category of data types. (For example,
4713 on some machines, floating-point registers are used for floating-point
4714 arguments while other arguments are passed in the general registers.)
4715 To make non-varargs functions use the proper calling convention, you
4716 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4717 registers in each category have been used so far
4719 @code{__builtin_args_info} accesses the same data structure of type
4720 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4721 with it, with @var{category} specifying which word to access. Thus, the
4722 value indicates the first unused register in a given category.
4724 Normally, you would use @code{__builtin_args_info} in the implementation
4725 of @code{va_start}, accessing each category just once and storing the
4726 value in the @code{va_list} object. This is because @code{va_list} will
4727 have to update the values, and there is no way to alter the
4728 values accessed by @code{__builtin_args_info}.
4731 @defmac __builtin_next_arg (@var{lastarg})
4732 This is the equivalent of @code{__builtin_args_info}, for stack
4733 arguments. It returns the address of the first anonymous stack
4734 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4735 returns the address of the location above the first anonymous stack
4736 argument. Use it in @code{va_start} to initialize the pointer for
4737 fetching arguments from the stack. Also use it in @code{va_start} to
4738 verify that the second parameter @var{lastarg} is the last named argument
4739 of the current function.
4742 @defmac __builtin_classify_type (@var{object})
4743 Since each machine has its own conventions for which data types are
4744 passed in which kind of register, your implementation of @code{va_arg}
4745 has to embody these conventions. The easiest way to categorize the
4746 specified data type is to use @code{__builtin_classify_type} together
4747 with @code{sizeof} and @code{__alignof__}.
4749 @code{__builtin_classify_type} ignores the value of @var{object},
4750 considering only its data type. It returns an integer describing what
4751 kind of type that is---integer, floating, pointer, structure, and so on.
4753 The file @file{typeclass.h} defines an enumeration that you can use to
4754 interpret the values of @code{__builtin_classify_type}.
4757 These machine description macros help implement varargs:
4759 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
4760 If defined, this hook produces the machine-specific code for a call to
4761 @code{__builtin_saveregs}. This code will be moved to the very
4762 beginning of the function, before any parameter access are made. The
4763 return value of this function should be an RTX that contains the value
4764 to use as the return of @code{__builtin_saveregs}.
4767 @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})
4768 This target hook offers an alternative to using
4769 @code{__builtin_saveregs} and defining the hook
4770 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
4771 register arguments into the stack so that all the arguments appear to
4772 have been passed consecutively on the stack. Once this is done, you can
4773 use the standard implementation of varargs that works for machines that
4774 pass all their arguments on the stack.
4776 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
4777 structure, containing the values that are obtained after processing the
4778 named arguments. The arguments @var{mode} and @var{type} describe the
4779 last named argument---its machine mode and its data type as a tree node.
4781 The target hook should do two things: first, push onto the stack all the
4782 argument registers @emph{not} used for the named arguments, and second,
4783 store the size of the data thus pushed into the @code{int}-valued
4784 variable pointed to by @var{pretend_args_size}. The value that you
4785 store here will serve as additional offset for setting up the stack
4788 Because you must generate code to push the anonymous arguments at
4789 compile time without knowing their data types,
4790 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
4791 have just a single category of argument register and use it uniformly
4794 If the argument @var{second_time} is nonzero, it means that the
4795 arguments of the function are being analyzed for the second time. This
4796 happens for an inline function, which is not actually compiled until the
4797 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
4798 not generate any instructions in this case.
4801 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca})
4802 Define this hook to return @code{true} if the location where a function
4803 argument is passed depends on whether or not it is a named argument.
4805 This hook controls how the @var{named} argument to @code{FUNCTION_ARG}
4806 is set for varargs and stdarg functions. If this hook returns
4807 @code{true}, the @var{named} argument is always true for named
4808 arguments, and false for unnamed arguments. If it returns @code{false},
4809 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
4810 then all arguments are treated as named. Otherwise, all named arguments
4811 except the last are treated as named.
4813 You need not define this hook if it always returns zero.
4816 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED
4817 If you need to conditionally change ABIs so that one works with
4818 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
4819 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
4820 defined, then define this hook to return @code{true} if
4821 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
4822 Otherwise, you should not define this hook.
4826 @section Trampolines for Nested Functions
4827 @cindex trampolines for nested functions
4828 @cindex nested functions, trampolines for
4830 A @dfn{trampoline} is a small piece of code that is created at run time
4831 when the address of a nested function is taken. It normally resides on
4832 the stack, in the stack frame of the containing function. These macros
4833 tell GCC how to generate code to allocate and initialize a
4836 The instructions in the trampoline must do two things: load a constant
4837 address into the static chain register, and jump to the real address of
4838 the nested function. On CISC machines such as the m68k, this requires
4839 two instructions, a move immediate and a jump. Then the two addresses
4840 exist in the trampoline as word-long immediate operands. On RISC
4841 machines, it is often necessary to load each address into a register in
4842 two parts. Then pieces of each address form separate immediate
4845 The code generated to initialize the trampoline must store the variable
4846 parts---the static chain value and the function address---into the
4847 immediate operands of the instructions. On a CISC machine, this is
4848 simply a matter of copying each address to a memory reference at the
4849 proper offset from the start of the trampoline. On a RISC machine, it
4850 may be necessary to take out pieces of the address and store them
4853 @defmac TRAMPOLINE_TEMPLATE (@var{file})
4854 A C statement to output, on the stream @var{file}, assembler code for a
4855 block of data that contains the constant parts of a trampoline. This
4856 code should not include a label---the label is taken care of
4859 If you do not define this macro, it means no template is needed
4860 for the target. Do not define this macro on systems where the block move
4861 code to copy the trampoline into place would be larger than the code
4862 to generate it on the spot.
4865 @defmac TRAMPOLINE_SECTION
4866 Return the section into which the trampoline template is to be placed
4867 (@pxref{Sections}). The default value is @code{readonly_data_section}.
4870 @defmac TRAMPOLINE_SIZE
4871 A C expression for the size in bytes of the trampoline, as an integer.
4874 @defmac TRAMPOLINE_ALIGNMENT
4875 Alignment required for trampolines, in bits.
4877 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4878 is used for aligning trampolines.
4881 @defmac INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4882 A C statement to initialize the variable parts of a trampoline.
4883 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4884 an RTX for the address of the nested function; @var{static_chain} is an
4885 RTX for the static chain value that should be passed to the function
4889 @defmac TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4890 A C statement that should perform any machine-specific adjustment in
4891 the address of the trampoline. Its argument contains the address that
4892 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4893 used for a function call should be different from the address in which
4894 the template was stored, the different address should be assigned to
4895 @var{addr}. If this macro is not defined, @var{addr} will be used for
4898 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4899 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4900 If this macro is not defined, by default the trampoline is allocated as
4901 a stack slot. This default is right for most machines. The exceptions
4902 are machines where it is impossible to execute instructions in the stack
4903 area. On such machines, you may have to implement a separate stack,
4904 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4905 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4907 @var{fp} points to a data structure, a @code{struct function}, which
4908 describes the compilation status of the immediate containing function of
4909 the function which the trampoline is for. The stack slot for the
4910 trampoline is in the stack frame of this containing function. Other
4911 allocation strategies probably must do something analogous with this
4915 Implementing trampolines is difficult on many machines because they have
4916 separate instruction and data caches. Writing into a stack location
4917 fails to clear the memory in the instruction cache, so when the program
4918 jumps to that location, it executes the old contents.
4920 Here are two possible solutions. One is to clear the relevant parts of
4921 the instruction cache whenever a trampoline is set up. The other is to
4922 make all trampolines identical, by having them jump to a standard
4923 subroutine. The former technique makes trampoline execution faster; the
4924 latter makes initialization faster.
4926 To clear the instruction cache when a trampoline is initialized, define
4927 the following macro.
4929 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
4930 If defined, expands to a C expression clearing the @emph{instruction
4931 cache} in the specified interval. The definition of this macro would
4932 typically be a series of @code{asm} statements. Both @var{beg} and
4933 @var{end} are both pointer expressions.
4936 The operating system may also require the stack to be made executable
4937 before calling the trampoline. To implement this requirement, define
4938 the following macro.
4940 @defmac ENABLE_EXECUTE_STACK
4941 Define this macro if certain operations must be performed before executing
4942 code located on the stack. The macro should expand to a series of C
4943 file-scope constructs (e.g.@: functions) and provide a unique entry point
4944 named @code{__enable_execute_stack}. The target is responsible for
4945 emitting calls to the entry point in the code, for example from the
4946 @code{INITIALIZE_TRAMPOLINE} macro.
4949 To use a standard subroutine, define the following macro. In addition,
4950 you must make sure that the instructions in a trampoline fill an entire
4951 cache line with identical instructions, or else ensure that the
4952 beginning of the trampoline code is always aligned at the same point in
4953 its cache line. Look in @file{m68k.h} as a guide.
4955 @defmac TRANSFER_FROM_TRAMPOLINE
4956 Define this macro if trampolines need a special subroutine to do their
4957 work. The macro should expand to a series of @code{asm} statements
4958 which will be compiled with GCC@. They go in a library function named
4959 @code{__transfer_from_trampoline}.
4961 If you need to avoid executing the ordinary prologue code of a compiled
4962 C function when you jump to the subroutine, you can do so by placing a
4963 special label of your own in the assembler code. Use one @code{asm}
4964 statement to generate an assembler label, and another to make the label
4965 global. Then trampolines can use that label to jump directly to your
4966 special assembler code.
4970 @section Implicit Calls to Library Routines
4971 @cindex library subroutine names
4972 @cindex @file{libgcc.a}
4974 @c prevent bad page break with this line
4975 Here is an explanation of implicit calls to library routines.
4977 @defmac DECLARE_LIBRARY_RENAMES
4978 This macro, if defined, should expand to a piece of C code that will get
4979 expanded when compiling functions for libgcc.a. It can be used to
4980 provide alternate names for GCC's internal library functions if there
4981 are ABI-mandated names that the compiler should provide.
4984 @findex init_one_libfunc
4985 @findex set_optab_libfunc
4986 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
4987 This hook should declare additional library routines or rename
4988 existing ones, using the functions @code{set_optab_libfunc} and
4989 @code{init_one_libfunc} defined in @file{optabs.c}.
4990 @code{init_optabs} calls this macro after initializing all the normal
4993 The default is to do nothing. Most ports don't need to define this hook.
4996 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4997 This macro should return @code{true} if the library routine that
4998 implements the floating point comparison operator @var{comparison} in
4999 mode @var{mode} will return a boolean, and @var{false} if it will
5002 GCC's own floating point libraries return tristates from the
5003 comparison operators, so the default returns false always. Most ports
5004 don't need to define this macro.
5007 @defmac TARGET_LIB_INT_CMP_BIASED
5008 This macro should evaluate to @code{true} if the integer comparison
5009 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5010 operand is smaller than the second, 1 to indicate that they are equal,
5011 and 2 to indicate that the first operand is greater than the second.
5012 If this macro evaluates to @code{false} the comparison functions return
5013 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5014 in @file{libgcc.a}, you do not need to define this macro.
5017 @cindex US Software GOFAST, floating point emulation library
5018 @cindex floating point emulation library, US Software GOFAST
5019 @cindex GOFAST, floating point emulation library
5020 @findex gofast_maybe_init_libfuncs
5021 @defmac US_SOFTWARE_GOFAST
5022 Define this macro if your system C library uses the US Software GOFAST
5023 library to provide floating point emulation.
5025 In addition to defining this macro, your architecture must set
5026 @code{TARGET_INIT_LIBFUNCS} to @code{gofast_maybe_init_libfuncs}, or
5027 else call that function from its version of that hook. It is defined
5028 in @file{config/gofast.h}, which must be included by your
5029 architecture's @file{@var{cpu}.c} file. See @file{sparc/sparc.c} for
5032 If this macro is defined, the
5033 @code{TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL} target hook must return
5034 false for @code{SFmode} and @code{DFmode} comparisons.
5037 @cindex @code{EDOM}, implicit usage
5040 The value of @code{EDOM} on the target machine, as a C integer constant
5041 expression. If you don't define this macro, GCC does not attempt to
5042 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5043 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5046 If you do not define @code{TARGET_EDOM}, then compiled code reports
5047 domain errors by calling the library function and letting it report the
5048 error. If mathematical functions on your system use @code{matherr} when
5049 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5050 that @code{matherr} is used normally.
5053 @cindex @code{errno}, implicit usage
5054 @defmac GEN_ERRNO_RTX
5055 Define this macro as a C expression to create an rtl expression that
5056 refers to the global ``variable'' @code{errno}. (On certain systems,
5057 @code{errno} may not actually be a variable.) If you don't define this
5058 macro, a reasonable default is used.
5061 @cindex C99 math functions, implicit usage
5062 @defmac TARGET_C99_FUNCTIONS
5063 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
5064 @code{sinf} and similarly for other functions defined by C99 standard. The
5065 default is nonzero that should be proper value for most modern systems, however
5066 number of existing systems lacks support for these functions in the runtime so
5067 they needs this macro to be redefined to 0.
5070 @cindex sincos math function, implicit usage
5071 @defmac TARGET_HAS_SINCOS
5072 When this macro is nonzero, GCC will implicitly optimize calls to @code{sin}
5073 and @code{cos} with the same argument to a call to @code{sincos}. The
5074 default is zero. The target has to provide the following functions:
5076 void sincos(double x, double *sin, double *cos);
5077 void sincosf(float x, float *sin, float *cos);
5078 void sincosl(long double x, long double *sin, long double *cos);
5082 @defmac NEXT_OBJC_RUNTIME
5083 Define this macro to generate code for Objective-C message sending using
5084 the calling convention of the NeXT system. This calling convention
5085 involves passing the object, the selector and the method arguments all
5086 at once to the method-lookup library function.
5088 The default calling convention passes just the object and the selector
5089 to the lookup function, which returns a pointer to the method.
5092 @node Addressing Modes
5093 @section Addressing Modes
5094 @cindex addressing modes
5096 @c prevent bad page break with this line
5097 This is about addressing modes.
5099 @defmac HAVE_PRE_INCREMENT
5100 @defmacx HAVE_PRE_DECREMENT
5101 @defmacx HAVE_POST_INCREMENT
5102 @defmacx HAVE_POST_DECREMENT
5103 A C expression that is nonzero if the machine supports pre-increment,
5104 pre-decrement, post-increment, or post-decrement addressing respectively.
5107 @defmac HAVE_PRE_MODIFY_DISP
5108 @defmacx HAVE_POST_MODIFY_DISP
5109 A C expression that is nonzero if the machine supports pre- or
5110 post-address side-effect generation involving constants other than
5111 the size of the memory operand.
5114 @defmac HAVE_PRE_MODIFY_REG
5115 @defmacx HAVE_POST_MODIFY_REG
5116 A C expression that is nonzero if the machine supports pre- or
5117 post-address side-effect generation involving a register displacement.
5120 @defmac CONSTANT_ADDRESS_P (@var{x})
5121 A C expression that is 1 if the RTX @var{x} is a constant which
5122 is a valid address. On most machines, this can be defined as
5123 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
5124 in which constant addresses are supported.
5127 @defmac CONSTANT_P (@var{x})
5128 @code{CONSTANT_P}, which is defined by target-independent code,
5129 accepts integer-values expressions whose values are not explicitly
5130 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5131 expressions and @code{const} arithmetic expressions, in addition to
5132 @code{const_int} and @code{const_double} expressions.
5135 @defmac MAX_REGS_PER_ADDRESS
5136 A number, the maximum number of registers that can appear in a valid
5137 memory address. Note that it is up to you to specify a value equal to
5138 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
5142 @defmac GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5143 A C compound statement with a conditional @code{goto @var{label};}
5144 executed if @var{x} (an RTX) is a legitimate memory address on the
5145 target machine for a memory operand of mode @var{mode}.
5147 It usually pays to define several simpler macros to serve as
5148 subroutines for this one. Otherwise it may be too complicated to
5151 This macro must exist in two variants: a strict variant and a
5152 non-strict one. The strict variant is used in the reload pass. It
5153 must be defined so that any pseudo-register that has not been
5154 allocated a hard register is considered a memory reference. In
5155 contexts where some kind of register is required, a pseudo-register
5156 with no hard register must be rejected.
5158 The non-strict variant is used in other passes. It must be defined to
5159 accept all pseudo-registers in every context where some kind of
5160 register is required.
5162 @findex REG_OK_STRICT
5163 Compiler source files that want to use the strict variant of this
5164 macro define the macro @code{REG_OK_STRICT}. You should use an
5165 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
5166 in that case and the non-strict variant otherwise.
5168 Subroutines to check for acceptable registers for various purposes (one
5169 for base registers, one for index registers, and so on) are typically
5170 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
5171 Then only these subroutine macros need have two variants; the higher
5172 levels of macros may be the same whether strict or not.
5174 Normally, constant addresses which are the sum of a @code{symbol_ref}
5175 and an integer are stored inside a @code{const} RTX to mark them as
5176 constant. Therefore, there is no need to recognize such sums
5177 specifically as legitimate addresses. Normally you would simply
5178 recognize any @code{const} as legitimate.
5180 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5181 sums that are not marked with @code{const}. It assumes that a naked
5182 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5183 naked constant sums as illegitimate addresses, so that none of them will
5184 be given to @code{PRINT_OPERAND_ADDRESS}.
5186 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5187 On some machines, whether a symbolic address is legitimate depends on
5188 the section that the address refers to. On these machines, define the
5189 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5190 into the @code{symbol_ref}, and then check for it here. When you see a
5191 @code{const}, you will have to look inside it to find the
5192 @code{symbol_ref} in order to determine the section. @xref{Assembler
5196 @defmac FIND_BASE_TERM (@var{x})
5197 A C expression to determine the base term of address @var{x}.
5198 This macro is used in only one place: `find_base_term' in alias.c.
5200 It is always safe for this macro to not be defined. It exists so
5201 that alias analysis can understand machine-dependent addresses.
5203 The typical use of this macro is to handle addresses containing
5204 a label_ref or symbol_ref within an UNSPEC@.
5207 @defmac LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
5208 A C compound statement that attempts to replace @var{x} with a valid
5209 memory address for an operand of mode @var{mode}. @var{win} will be a
5210 C statement label elsewhere in the code; the macro definition may use
5213 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
5217 to avoid further processing if the address has become legitimate.
5219 @findex break_out_memory_refs
5220 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5221 and @var{oldx} will be the operand that was given to that function to produce
5224 The code generated by this macro should not alter the substructure of
5225 @var{x}. If it transforms @var{x} into a more legitimate form, it
5226 should assign @var{x} (which will always be a C variable) a new value.
5228 It is not necessary for this macro to come up with a legitimate
5229 address. The compiler has standard ways of doing so in all cases. In
5230 fact, it is safe to omit this macro. But often a
5231 machine-dependent strategy can generate better code.
5234 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5235 A C compound statement that attempts to replace @var{x}, which is an address
5236 that needs reloading, with a valid memory address for an operand of mode
5237 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5238 It is not necessary to define this macro, but it might be useful for
5239 performance reasons.
5241 For example, on the i386, it is sometimes possible to use a single
5242 reload register instead of two by reloading a sum of two pseudo
5243 registers into a register. On the other hand, for number of RISC
5244 processors offsets are limited so that often an intermediate address
5245 needs to be generated in order to address a stack slot. By defining
5246 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5247 generated for adjacent some stack slots can be made identical, and thus
5250 @emph{Note}: This macro should be used with caution. It is necessary
5251 to know something of how reload works in order to effectively use this,
5252 and it is quite easy to produce macros that build in too much knowledge
5253 of reload internals.
5255 @emph{Note}: This macro must be able to reload an address created by a
5256 previous invocation of this macro. If it fails to handle such addresses
5257 then the compiler may generate incorrect code or abort.
5260 The macro definition should use @code{push_reload} to indicate parts that
5261 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5262 suitable to be passed unaltered to @code{push_reload}.
5264 The code generated by this macro must not alter the substructure of
5265 @var{x}. If it transforms @var{x} into a more legitimate form, it
5266 should assign @var{x} (which will always be a C variable) a new value.
5267 This also applies to parts that you change indirectly by calling
5270 @findex strict_memory_address_p
5271 The macro definition may use @code{strict_memory_address_p} to test if
5272 the address has become legitimate.
5275 If you want to change only a part of @var{x}, one standard way of doing
5276 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5277 single level of rtl. Thus, if the part to be changed is not at the
5278 top level, you'll need to replace first the top level.
5279 It is not necessary for this macro to come up with a legitimate
5280 address; but often a machine-dependent strategy can generate better code.
5283 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5284 A C statement or compound statement with a conditional @code{goto
5285 @var{label};} executed if memory address @var{x} (an RTX) can have
5286 different meanings depending on the machine mode of the memory
5287 reference it is used for or if the address is valid for some modes
5290 Autoincrement and autodecrement addresses typically have mode-dependent
5291 effects because the amount of the increment or decrement is the size
5292 of the operand being addressed. Some machines have other mode-dependent
5293 addresses. Many RISC machines have no mode-dependent addresses.
5295 You may assume that @var{addr} is a valid address for the machine.
5298 @defmac LEGITIMATE_CONSTANT_P (@var{x})
5299 A C expression that is nonzero if @var{x} is a legitimate constant for
5300 an immediate operand on the target machine. You can assume that
5301 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5302 @samp{1} is a suitable definition for this macro on machines where
5303 anything @code{CONSTANT_P} is valid.
5306 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5307 This hook is used to undo the possibly obfuscating effects of the
5308 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5309 macros. Some backend implementations of these macros wrap symbol
5310 references inside an @code{UNSPEC} rtx to represent PIC or similar
5311 addressing modes. This target hook allows GCC's optimizers to understand
5312 the semantics of these opaque @code{UNSPEC}s by converting them back
5313 into their original form.
5316 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (rtx @var{x})
5317 This hook should return true if @var{x} is of a form that cannot (or
5318 should not) be spilled to the constant pool. The default version of
5319 this hook returns false.
5321 The primary reason to define this hook is to prevent reload from
5322 deciding that a non-legitimate constant would be better reloaded
5323 from the constant pool instead of spilling and reloading a register
5324 holding the constant. This restriction is often true of addresses
5325 of TLS symbols for various targets.
5328 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (enum machine_mode @var{mode}, rtx @var{x})
5329 This hook should return true if pool entries for constant @var{x} can
5330 be placed in an @code{object_block} structure. @var{mode} is the mode
5333 The default version returns false for all constants.
5336 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5337 This hook should return the DECL of a function @var{f} that given an
5338 address @var{addr} as an argument returns a mask @var{m} that can be
5339 used to extract from two vectors the relevant data that resides in
5340 @var{addr} in case @var{addr} is not properly aligned.
5342 The autovectorizer, when vectorizing a load operation from an address
5343 @var{addr} that may be unaligned, will generate two vector loads from
5344 the two aligned addresses around @var{addr}. It then generates a
5345 @code{REALIGN_LOAD} operation to extract the relevant data from the
5346 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5347 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5348 the third argument, @var{OFF}, defines how the data will be extracted
5349 from these two vectors: if @var{OFF} is 0, then the returned vector is
5350 @var{v2}; otherwise, the returned vector is composed from the last
5351 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5352 @var{OFF} elements of @var{v2}.
5354 If this hook is defined, the autovectorizer will generate a call
5355 to @var{f} (using the DECL tree that this hook returns) and will
5356 use the return value of @var{f} as the argument @var{OFF} to
5357 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5358 should comply with the semantics expected by @code{REALIGN_LOAD}
5360 If this hook is not defined, then @var{addr} will be used as
5361 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5362 log2(@var{VS})-1 bits of @var{addr} will be considered.
5365 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN (tree @var{x})
5366 This hook should return the DECL of a function @var{f} that implements
5367 widening multiplication of the even elements of two input vectors of type @var{x}.
5369 If this hook is defined, the autovectorizer will use it along with the
5370 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD} target hook when vectorizing
5371 widening multiplication in cases that the order of the results does not have to be
5372 preserved (e.g. used only by a reduction computation). Otherwise, the
5373 @code{widen_mult_hi/lo} idioms will be used.
5376 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD (tree @var{x})
5377 This hook should return the DECL of a function @var{f} that implements
5378 widening multiplication of the odd elements of two input vectors of type @var{x}.
5380 If this hook is defined, the autovectorizer will use it along with the
5381 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN} target hook when vectorizing
5382 widening multiplication in cases that the order of the results does not have to be
5383 preserved (e.g. used only by a reduction computation). Otherwise, the
5384 @code{widen_mult_hi/lo} idioms will be used.
5387 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION (enum built_in_function @var{code}, tree @var{vec_type})
5388 This hook should return the decl of a function that implements the vectorized
5389 variant of the builtin function with builtin function code @var{code} or
5390 @code{NULL_TREE} if such a function is not available. The return type of
5391 the vectorized function shall be of vector type @var{vec_type}.
5394 @node Anchored Addresses
5395 @section Anchored Addresses
5396 @cindex anchored addresses
5397 @cindex @option{-fsection-anchors}
5399 GCC usually addresses every static object as a separate entity.
5400 For example, if we have:
5404 int foo (void) @{ return a + b + c; @}
5407 the code for @code{foo} will usually calculate three separate symbolic
5408 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
5409 it would be better to calculate just one symbolic address and access
5410 the three variables relative to it. The equivalent pseudocode would
5416 register int *xr = &x;
5417 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5421 (which isn't valid C). We refer to shared addresses like @code{x} as
5422 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
5424 The hooks below describe the target properties that GCC needs to know
5425 in order to make effective use of section anchors. It won't use
5426 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
5427 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
5429 @deftypevar {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET
5430 The minimum offset that should be applied to a section anchor.
5431 On most targets, it should be the smallest offset that can be
5432 applied to a base register while still giving a legitimate address
5433 for every mode. The default value is 0.
5436 @deftypevar {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET
5437 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
5438 offset that should be applied to section anchors. The default
5442 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x})
5443 Write the assembly code to define section anchor @var{x}, which is a
5444 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
5445 The hook is called with the assembly output position set to the beginning
5446 of @code{SYMBOL_REF_BLOCK (@var{x})}.
5448 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
5449 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
5450 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
5451 is @code{NULL}, which disables the use of section anchors altogether.
5454 @deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (rtx @var{x})
5455 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
5456 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
5457 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
5459 The default version is correct for most targets, but you might need to
5460 intercept this hook to handle things like target-specific attributes
5461 or target-specific sections.
5464 @node Condition Code
5465 @section Condition Code Status
5466 @cindex condition code status
5468 @c prevent bad page break with this line
5469 This describes the condition code status.
5472 The file @file{conditions.h} defines a variable @code{cc_status} to
5473 describe how the condition code was computed (in case the interpretation of
5474 the condition code depends on the instruction that it was set by). This
5475 variable contains the RTL expressions on which the condition code is
5476 currently based, and several standard flags.
5478 Sometimes additional machine-specific flags must be defined in the machine
5479 description header file. It can also add additional machine-specific
5480 information by defining @code{CC_STATUS_MDEP}.
5482 @defmac CC_STATUS_MDEP
5483 C code for a data type which is used for declaring the @code{mdep}
5484 component of @code{cc_status}. It defaults to @code{int}.
5486 This macro is not used on machines that do not use @code{cc0}.
5489 @defmac CC_STATUS_MDEP_INIT
5490 A C expression to initialize the @code{mdep} field to ``empty''.
5491 The default definition does nothing, since most machines don't use
5492 the field anyway. If you want to use the field, you should probably
5493 define this macro to initialize it.
5495 This macro is not used on machines that do not use @code{cc0}.
5498 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5499 A C compound statement to set the components of @code{cc_status}
5500 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5501 this macro's responsibility to recognize insns that set the condition
5502 code as a byproduct of other activity as well as those that explicitly
5505 This macro is not used on machines that do not use @code{cc0}.
5507 If there are insns that do not set the condition code but do alter
5508 other machine registers, this macro must check to see whether they
5509 invalidate the expressions that the condition code is recorded as
5510 reflecting. For example, on the 68000, insns that store in address
5511 registers do not set the condition code, which means that usually
5512 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5513 insns. But suppose that the previous insn set the condition code
5514 based on location @samp{a4@@(102)} and the current insn stores a new
5515 value in @samp{a4}. Although the condition code is not changed by
5516 this, it will no longer be true that it reflects the contents of
5517 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5518 @code{cc_status} in this case to say that nothing is known about the
5519 condition code value.
5521 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5522 with the results of peephole optimization: insns whose patterns are
5523 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5524 constants which are just the operands. The RTL structure of these
5525 insns is not sufficient to indicate what the insns actually do. What
5526 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5527 @code{CC_STATUS_INIT}.
5529 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5530 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5531 @samp{cc}. This avoids having detailed information about patterns in
5532 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5535 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5536 Returns a mode from class @code{MODE_CC} to be used when comparison
5537 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5538 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5539 @pxref{Jump Patterns} for a description of the reason for this
5543 #define SELECT_CC_MODE(OP,X,Y) \
5544 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5545 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5546 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5547 || GET_CODE (X) == NEG) \
5548 ? CC_NOOVmode : CCmode))
5551 You should define this macro if and only if you define extra CC modes
5552 in @file{@var{machine}-modes.def}.
5555 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5556 On some machines not all possible comparisons are defined, but you can
5557 convert an invalid comparison into a valid one. For example, the Alpha
5558 does not have a @code{GT} comparison, but you can use an @code{LT}
5559 comparison instead and swap the order of the operands.
5561 On such machines, define this macro to be a C statement to do any
5562 required conversions. @var{code} is the initial comparison code
5563 and @var{op0} and @var{op1} are the left and right operands of the
5564 comparison, respectively. You should modify @var{code}, @var{op0}, and
5565 @var{op1} as required.
5567 GCC will not assume that the comparison resulting from this macro is
5568 valid but will see if the resulting insn matches a pattern in the
5571 You need not define this macro if it would never change the comparison
5575 @defmac REVERSIBLE_CC_MODE (@var{mode})
5576 A C expression whose value is one if it is always safe to reverse a
5577 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5578 can ever return @var{mode} for a floating-point inequality comparison,
5579 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5581 You need not define this macro if it would always returns zero or if the
5582 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5583 For example, here is the definition used on the SPARC, where floating-point
5584 inequality comparisons are always given @code{CCFPEmode}:
5587 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5591 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
5592 A C expression whose value is reversed condition code of the @var{code} for
5593 comparison done in CC_MODE @var{mode}. The macro is used only in case
5594 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5595 machine has some non-standard way how to reverse certain conditionals. For
5596 instance in case all floating point conditions are non-trapping, compiler may
5597 freely convert unordered compares to ordered one. Then definition may look
5601 #define REVERSE_CONDITION(CODE, MODE) \
5602 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5603 : reverse_condition_maybe_unordered (CODE))
5607 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
5608 A C expression that returns true if the conditional execution predicate
5609 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
5610 versa. Define this to return 0 if the target has conditional execution
5611 predicates that cannot be reversed safely. There is no need to validate
5612 that the arguments of op1 and op2 are the same, this is done separately.
5613 If no expansion is specified, this macro is defined as follows:
5616 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5617 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
5621 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *, unsigned int *)
5622 On targets which do not use @code{(cc0)}, and which use a hard
5623 register rather than a pseudo-register to hold condition codes, the
5624 regular CSE passes are often not able to identify cases in which the
5625 hard register is set to a common value. Use this hook to enable a
5626 small pass which optimizes such cases. This hook should return true
5627 to enable this pass, and it should set the integers to which its
5628 arguments point to the hard register numbers used for condition codes.
5629 When there is only one such register, as is true on most systems, the
5630 integer pointed to by the second argument should be set to
5631 @code{INVALID_REGNUM}.
5633 The default version of this hook returns false.
5636 @deftypefn {Target Hook} enum machine_mode TARGET_CC_MODES_COMPATIBLE (enum machine_mode, enum machine_mode)
5637 On targets which use multiple condition code modes in class
5638 @code{MODE_CC}, it is sometimes the case that a comparison can be
5639 validly done in more than one mode. On such a system, define this
5640 target hook to take two mode arguments and to return a mode in which
5641 both comparisons may be validly done. If there is no such mode,
5642 return @code{VOIDmode}.
5644 The default version of this hook checks whether the modes are the
5645 same. If they are, it returns that mode. If they are different, it
5646 returns @code{VOIDmode}.
5650 @section Describing Relative Costs of Operations
5651 @cindex costs of instructions
5652 @cindex relative costs
5653 @cindex speed of instructions
5655 These macros let you describe the relative speed of various operations
5656 on the target machine.
5658 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5659 A C expression for the cost of moving data of mode @var{mode} from a
5660 register in class @var{from} to one in class @var{to}. The classes are
5661 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5662 value of 2 is the default; other values are interpreted relative to
5665 It is not required that the cost always equal 2 when @var{from} is the
5666 same as @var{to}; on some machines it is expensive to move between
5667 registers if they are not general registers.
5669 If reload sees an insn consisting of a single @code{set} between two
5670 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5671 classes returns a value of 2, reload does not check to ensure that the
5672 constraints of the insn are met. Setting a cost of other than 2 will
5673 allow reload to verify that the constraints are met. You should do this
5674 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5677 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5678 A C expression for the cost of moving data of mode @var{mode} between a
5679 register of class @var{class} and memory; @var{in} is zero if the value
5680 is to be written to memory, nonzero if it is to be read in. This cost
5681 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5682 registers and memory is more expensive than between two registers, you
5683 should define this macro to express the relative cost.
5685 If you do not define this macro, GCC uses a default cost of 4 plus
5686 the cost of copying via a secondary reload register, if one is
5687 needed. If your machine requires a secondary reload register to copy
5688 between memory and a register of @var{class} but the reload mechanism is
5689 more complex than copying via an intermediate, define this macro to
5690 reflect the actual cost of the move.
5692 GCC defines the function @code{memory_move_secondary_cost} if
5693 secondary reloads are needed. It computes the costs due to copying via
5694 a secondary register. If your machine copies from memory using a
5695 secondary register in the conventional way but the default base value of
5696 4 is not correct for your machine, define this macro to add some other
5697 value to the result of that function. The arguments to that function
5698 are the same as to this macro.
5702 A C expression for the cost of a branch instruction. A value of 1 is
5703 the default; other values are interpreted relative to that.
5706 Here are additional macros which do not specify precise relative costs,
5707 but only that certain actions are more expensive than GCC would
5710 @defmac SLOW_BYTE_ACCESS
5711 Define this macro as a C expression which is nonzero if accessing less
5712 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5713 faster than accessing a word of memory, i.e., if such access
5714 require more than one instruction or if there is no difference in cost
5715 between byte and (aligned) word loads.
5717 When this macro is not defined, the compiler will access a field by
5718 finding the smallest containing object; when it is defined, a fullword
5719 load will be used if alignment permits. Unless bytes accesses are
5720 faster than word accesses, using word accesses is preferable since it
5721 may eliminate subsequent memory access if subsequent accesses occur to
5722 other fields in the same word of the structure, but to different bytes.
5725 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5726 Define this macro to be the value 1 if memory accesses described by the
5727 @var{mode} and @var{alignment} parameters have a cost many times greater
5728 than aligned accesses, for example if they are emulated in a trap
5731 When this macro is nonzero, the compiler will act as if
5732 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5733 moves. This can cause significantly more instructions to be produced.
5734 Therefore, do not set this macro nonzero if unaligned accesses only add a
5735 cycle or two to the time for a memory access.
5737 If the value of this macro is always zero, it need not be defined. If
5738 this macro is defined, it should produce a nonzero value when
5739 @code{STRICT_ALIGNMENT} is nonzero.
5743 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5744 which a sequence of insns should be generated instead of a
5745 string move insn or a library call. Increasing the value will always
5746 make code faster, but eventually incurs high cost in increased code size.
5748 Note that on machines where the corresponding move insn is a
5749 @code{define_expand} that emits a sequence of insns, this macro counts
5750 the number of such sequences.
5752 If you don't define this, a reasonable default is used.
5755 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5756 A C expression used to determine whether @code{move_by_pieces} will be used to
5757 copy a chunk of memory, or whether some other block move mechanism
5758 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5759 than @code{MOVE_RATIO}.
5762 @defmac MOVE_MAX_PIECES
5763 A C expression used by @code{move_by_pieces} to determine the largest unit
5764 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5768 The threshold of number of scalar move insns, @emph{below} which a sequence
5769 of insns should be generated to clear memory instead of a string clear insn
5770 or a library call. Increasing the value will always make code faster, but
5771 eventually incurs high cost in increased code size.
5773 If you don't define this, a reasonable default is used.
5776 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5777 A C expression used to determine whether @code{clear_by_pieces} will be used
5778 to clear a chunk of memory, or whether some other block clear mechanism
5779 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5780 than @code{CLEAR_RATIO}.
5783 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
5784 A C expression used to determine whether @code{store_by_pieces} will be
5785 used to set a chunk of memory to a constant value, or whether some other
5786 mechanism will be used. Used by @code{__builtin_memset} when storing
5787 values other than constant zero and by @code{__builtin_strcpy} when
5788 when called with a constant source string.
5789 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5790 than @code{MOVE_RATIO}.
5793 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
5794 A C expression used to determine whether a load postincrement is a good
5795 thing to use for a given mode. Defaults to the value of
5796 @code{HAVE_POST_INCREMENT}.
5799 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
5800 A C expression used to determine whether a load postdecrement is a good
5801 thing to use for a given mode. Defaults to the value of
5802 @code{HAVE_POST_DECREMENT}.
5805 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
5806 A C expression used to determine whether a load preincrement is a good
5807 thing to use for a given mode. Defaults to the value of
5808 @code{HAVE_PRE_INCREMENT}.
5811 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
5812 A C expression used to determine whether a load predecrement is a good
5813 thing to use for a given mode. Defaults to the value of
5814 @code{HAVE_PRE_DECREMENT}.
5817 @defmac USE_STORE_POST_INCREMENT (@var{mode})
5818 A C expression used to determine whether a store postincrement is a good
5819 thing to use for a given mode. Defaults to the value of
5820 @code{HAVE_POST_INCREMENT}.
5823 @defmac USE_STORE_POST_DECREMENT (@var{mode})
5824 A C expression used to determine whether a store postdecrement is a good
5825 thing to use for a given mode. Defaults to the value of
5826 @code{HAVE_POST_DECREMENT}.
5829 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
5830 This macro is used to determine whether a store preincrement is a good
5831 thing to use for a given mode. Defaults to the value of
5832 @code{HAVE_PRE_INCREMENT}.
5835 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
5836 This macro is used to determine whether a store predecrement is a good
5837 thing to use for a given mode. Defaults to the value of
5838 @code{HAVE_PRE_DECREMENT}.
5841 @defmac NO_FUNCTION_CSE
5842 Define this macro if it is as good or better to call a constant
5843 function address than to call an address kept in a register.
5846 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
5847 Define this macro if a non-short-circuit operation produced by
5848 @samp{fold_range_test ()} is optimal. This macro defaults to true if
5849 @code{BRANCH_COST} is greater than or equal to the value 2.
5852 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total})
5853 This target hook describes the relative costs of RTL expressions.
5855 The cost may depend on the precise form of the expression, which is
5856 available for examination in @var{x}, and the rtx code of the expression
5857 in which it is contained, found in @var{outer_code}. @var{code} is the
5858 expression code---redundant, since it can be obtained with
5859 @code{GET_CODE (@var{x})}.
5861 In implementing this hook, you can use the construct
5862 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
5865 On entry to the hook, @code{*@var{total}} contains a default estimate
5866 for the cost of the expression. The hook should modify this value as
5867 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
5868 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
5869 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
5871 When optimizing for code size, i.e.@: when @code{optimize_size} is
5872 nonzero, this target hook should be used to estimate the relative
5873 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
5875 The hook returns true when all subexpressions of @var{x} have been
5876 processed, and false when @code{rtx_cost} should recurse.
5879 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address})
5880 This hook computes the cost of an addressing mode that contains
5881 @var{address}. If not defined, the cost is computed from
5882 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
5884 For most CISC machines, the default cost is a good approximation of the
5885 true cost of the addressing mode. However, on RISC machines, all
5886 instructions normally have the same length and execution time. Hence
5887 all addresses will have equal costs.
5889 In cases where more than one form of an address is known, the form with
5890 the lowest cost will be used. If multiple forms have the same, lowest,
5891 cost, the one that is the most complex will be used.
5893 For example, suppose an address that is equal to the sum of a register
5894 and a constant is used twice in the same basic block. When this macro
5895 is not defined, the address will be computed in a register and memory
5896 references will be indirect through that register. On machines where
5897 the cost of the addressing mode containing the sum is no higher than
5898 that of a simple indirect reference, this will produce an additional
5899 instruction and possibly require an additional register. Proper
5900 specification of this macro eliminates this overhead for such machines.
5902 This hook is never called with an invalid address.
5904 On machines where an address involving more than one register is as
5905 cheap as an address computation involving only one register, defining
5906 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
5907 be live over a region of code where only one would have been if
5908 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
5909 should be considered in the definition of this macro. Equivalent costs
5910 should probably only be given to addresses with different numbers of
5911 registers on machines with lots of registers.
5915 @section Adjusting the Instruction Scheduler
5917 The instruction scheduler may need a fair amount of machine-specific
5918 adjustment in order to produce good code. GCC provides several target
5919 hooks for this purpose. It is usually enough to define just a few of
5920 them: try the first ones in this list first.
5922 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
5923 This hook returns the maximum number of instructions that can ever
5924 issue at the same time on the target machine. The default is one.
5925 Although the insn scheduler can define itself the possibility of issue
5926 an insn on the same cycle, the value can serve as an additional
5927 constraint to issue insns on the same simulated processor cycle (see
5928 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
5929 This value must be constant over the entire compilation. If you need
5930 it to vary depending on what the instructions are, you must use
5931 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
5934 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
5935 This hook is executed by the scheduler after it has scheduled an insn
5936 from the ready list. It should return the number of insns which can
5937 still be issued in the current cycle. The default is
5938 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
5939 @code{USE}, which normally are not counted against the issue rate.
5940 You should define this hook if some insns take more machine resources
5941 than others, so that fewer insns can follow them in the same cycle.
5942 @var{file} is either a null pointer, or a stdio stream to write any
5943 debug output to. @var{verbose} is the verbose level provided by
5944 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
5948 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
5949 This function corrects the value of @var{cost} based on the
5950 relationship between @var{insn} and @var{dep_insn} through the
5951 dependence @var{link}. It should return the new value. The default
5952 is to make no adjustment to @var{cost}. This can be used for example
5953 to specify to the scheduler using the traditional pipeline description
5954 that an output- or anti-dependence does not incur the same cost as a
5955 data-dependence. If the scheduler using the automaton based pipeline
5956 description, the cost of anti-dependence is zero and the cost of
5957 output-dependence is maximum of one and the difference of latency
5958 times of the first and the second insns. If these values are not
5959 acceptable, you could use the hook to modify them too. See also
5960 @pxref{Processor pipeline description}.
5963 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
5964 This hook adjusts the integer scheduling priority @var{priority} of
5965 @var{insn}. It should return the new priority. Increase the priority to
5966 execute @var{insn} earlier, reduce the priority to execute @var{insn}
5967 later. Do not define this hook if you do not need to adjust the
5968 scheduling priorities of insns.
5971 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
5972 This hook is executed by the scheduler after it has scheduled the ready
5973 list, to allow the machine description to reorder it (for example to
5974 combine two small instructions together on @samp{VLIW} machines).
5975 @var{file} is either a null pointer, or a stdio stream to write any
5976 debug output to. @var{verbose} is the verbose level provided by
5977 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
5978 list of instructions that are ready to be scheduled. @var{n_readyp} is
5979 a pointer to the number of elements in the ready list. The scheduler
5980 reads the ready list in reverse order, starting with
5981 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
5982 is the timer tick of the scheduler. You may modify the ready list and
5983 the number of ready insns. The return value is the number of insns that
5984 can issue this cycle; normally this is just @code{issue_rate}. See also
5985 @samp{TARGET_SCHED_REORDER2}.
5988 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
5989 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
5990 function is called whenever the scheduler starts a new cycle. This one
5991 is called once per iteration over a cycle, immediately after
5992 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
5993 return the number of insns to be scheduled in the same cycle. Defining
5994 this hook can be useful if there are frequent situations where
5995 scheduling one insn causes other insns to become ready in the same
5996 cycle. These other insns can then be taken into account properly.
5999 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
6000 This hook is called after evaluation forward dependencies of insns in
6001 chain given by two parameter values (@var{head} and @var{tail}
6002 correspondingly) but before insns scheduling of the insn chain. For
6003 example, it can be used for better insn classification if it requires
6004 analysis of dependencies. This hook can use backward and forward
6005 dependencies of the insn scheduler because they are already
6009 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
6010 This hook is executed by the scheduler at the beginning of each block of
6011 instructions that are to be scheduled. @var{file} is either a null
6012 pointer, or a stdio stream to write any debug output to. @var{verbose}
6013 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6014 @var{max_ready} is the maximum number of insns in the current scheduling
6015 region that can be live at the same time. This can be used to allocate
6016 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
6019 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
6020 This hook is executed by the scheduler at the end of each block of
6021 instructions that are to be scheduled. It can be used to perform
6022 cleanup of any actions done by the other scheduling hooks. @var{file}
6023 is either a null pointer, or a stdio stream to write any debug output
6024 to. @var{verbose} is the verbose level provided by
6025 @option{-fsched-verbose-@var{n}}.
6028 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
6029 This hook is executed by the scheduler after function level initializations.
6030 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6031 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6032 @var{old_max_uid} is the maximum insn uid when scheduling begins.
6035 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
6036 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
6037 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6038 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6041 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
6042 The hook returns an RTL insn. The automaton state used in the
6043 pipeline hazard recognizer is changed as if the insn were scheduled
6044 when the new simulated processor cycle starts. Usage of the hook may
6045 simplify the automaton pipeline description for some @acronym{VLIW}
6046 processors. If the hook is defined, it is used only for the automaton
6047 based pipeline description. The default is not to change the state
6048 when the new simulated processor cycle starts.
6051 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
6052 The hook can be used to initialize data used by the previous hook.
6055 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
6056 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6057 to changed the state as if the insn were scheduled when the new
6058 simulated processor cycle finishes.
6061 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
6062 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
6063 used to initialize data used by the previous hook.
6066 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
6067 This hook controls better choosing an insn from the ready insn queue
6068 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
6069 chooses the first insn from the queue. If the hook returns a positive
6070 value, an additional scheduler code tries all permutations of
6071 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
6072 subsequent ready insns to choose an insn whose issue will result in
6073 maximal number of issued insns on the same cycle. For the
6074 @acronym{VLIW} processor, the code could actually solve the problem of
6075 packing simple insns into the @acronym{VLIW} insn. Of course, if the
6076 rules of @acronym{VLIW} packing are described in the automaton.
6078 This code also could be used for superscalar @acronym{RISC}
6079 processors. Let us consider a superscalar @acronym{RISC} processor
6080 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
6081 @var{B}, some insns can be executed only in pipelines @var{B} or
6082 @var{C}, and one insn can be executed in pipeline @var{B}. The
6083 processor may issue the 1st insn into @var{A} and the 2nd one into
6084 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
6085 until the next cycle. If the scheduler issues the 3rd insn the first,
6086 the processor could issue all 3 insns per cycle.
6088 Actually this code demonstrates advantages of the automaton based
6089 pipeline hazard recognizer. We try quickly and easy many insn
6090 schedules to choose the best one.
6092 The default is no multipass scheduling.
6095 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
6097 This hook controls what insns from the ready insn queue will be
6098 considered for the multipass insn scheduling. If the hook returns
6099 zero for insn passed as the parameter, the insn will be not chosen to
6102 The default is that any ready insns can be chosen to be issued.
6105 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
6107 This hook is called by the insn scheduler before issuing insn passed
6108 as the third parameter on given cycle. If the hook returns nonzero,
6109 the insn is not issued on given processors cycle. Instead of that,
6110 the processor cycle is advanced. If the value passed through the last
6111 parameter is zero, the insn ready queue is not sorted on the new cycle
6112 start as usually. The first parameter passes file for debugging
6113 output. The second one passes the scheduler verbose level of the
6114 debugging output. The forth and the fifth parameter values are
6115 correspondingly processor cycle on which the previous insn has been
6116 issued and the current processor cycle.
6119 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct dep_def *@var{_dep}, int @var{cost}, int @var{distance})
6120 This hook is used to define which dependences are considered costly by
6121 the target, so costly that it is not advisable to schedule the insns that
6122 are involved in the dependence too close to one another. The parameters
6123 to this hook are as follows: The first parameter @var{_dep} is the dependence
6124 being evaluated. The second parameter @var{cost} is the cost of the
6125 dependence, and the third
6126 parameter @var{distance} is the distance in cycles between the two insns.
6127 The hook returns @code{true} if considering the distance between the two
6128 insns the dependence between them is considered costly by the target,
6129 and @code{false} otherwise.
6131 Defining this hook can be useful in multiple-issue out-of-order machines,
6132 where (a) it's practically hopeless to predict the actual data/resource
6133 delays, however: (b) there's a better chance to predict the actual grouping
6134 that will be formed, and (c) correctly emulating the grouping can be very
6135 important. In such targets one may want to allow issuing dependent insns
6136 closer to one another---i.e., closer than the dependence distance; however,
6137 not in cases of "costly dependences", which this hooks allows to define.
6140 @deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void)
6141 This hook is called by the insn scheduler after emitting a new instruction to
6142 the instruction stream. The hook notifies a target backend to extend its
6143 per instruction data structures.
6146 @deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx @var{insn}, int @var{request}, rtx *@var{new_pat})
6147 This hook is called by the insn scheduler when @var{insn} has only
6148 speculative dependencies and therefore can be scheduled speculatively.
6149 The hook is used to check if the pattern of @var{insn} has a speculative
6150 version and, in case of successful check, to generate that speculative
6151 pattern. The hook should return 1, if the instruction has a speculative form,
6152 or -1, if it doesn't. @var{request} describes the type of requested
6153 speculation. If the return value equals 1 then @var{new_pat} is assigned
6154 the generated speculative pattern.
6157 @deftypefn {Target Hook} int TARGET_SCHED_NEEDS_BLOCK_P (rtx @var{insn})
6158 This hook is called by the insn scheduler during generation of recovery code
6159 for @var{insn}. It should return nonzero, if the corresponding check
6160 instruction should branch to recovery code, or zero otherwise.
6163 @deftypefn {Target Hook} rtx TARGET_SCHED_GEN_CHECK (rtx @var{insn}, rtx @var{label}, int @var{mutate_p})
6164 This hook is called by the insn scheduler to generate a pattern for recovery
6165 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
6166 speculative instruction for which the check should be generated.
6167 @var{label} is either a label of a basic block, where recovery code should
6168 be emitted, or a null pointer, when requested check doesn't branch to
6169 recovery code (a simple check). If @var{mutate_p} is nonzero, then
6170 a pattern for a branchy check corresponding to a simple check denoted by
6171 @var{insn} should be generated. In this case @var{label} can't be null.
6174 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC (rtx @var{insn})
6175 This hook is used as a workaround for
6176 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD} not being
6177 called on the first instruction of the ready list. The hook is used to
6178 discard speculative instruction that stand first in the ready list from
6179 being scheduled on the current cycle. For non-speculative instructions,
6180 the hook should always return nonzero. For example, in the ia64 backend
6181 the hook is used to cancel data speculative insns when the ALAT table
6185 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (unsigned int *@var{flags}, spec_info_t @var{spec_info})
6186 This hook is used by the insn scheduler to find out what features should be
6187 enabled/used. @var{flags} initially may have either the SCHED_RGN or SCHED_EBB
6188 bit set. This denotes the scheduler pass for which the data should be
6189 provided. The target backend should modify @var{flags} by modifying
6190 the bits corresponding to the following features: USE_DEPS_LIST, USE_GLAT,
6191 DETACH_LIFE_INFO, and DO_SPECULATION. For the DO_SPECULATION feature
6192 an additional structure @var{spec_info} should be filled by the target.
6193 The structure describes speculation types that can be used in the scheduler.
6197 @section Dividing the Output into Sections (Texts, Data, @dots{})
6198 @c the above section title is WAY too long. maybe cut the part between
6199 @c the (...)? --mew 10feb93
6201 An object file is divided into sections containing different types of
6202 data. In the most common case, there are three sections: the @dfn{text
6203 section}, which holds instructions and read-only data; the @dfn{data
6204 section}, which holds initialized writable data; and the @dfn{bss
6205 section}, which holds uninitialized data. Some systems have other kinds
6208 @file{varasm.c} provides several well-known sections, such as
6209 @code{text_section}, @code{data_section} and @code{bss_section}.
6210 The normal way of controlling a @code{@var{foo}_section} variable
6211 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
6212 as described below. The macros are only read once, when @file{varasm.c}
6213 initializes itself, so their values must be run-time constants.
6214 They may however depend on command-line flags.
6216 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
6217 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
6218 to be string literals.
6220 Some assemblers require a different string to be written every time a
6221 section is selected. If your assembler falls into this category, you
6222 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
6223 @code{get_unnamed_section} to set up the sections.
6225 You must always create a @code{text_section}, either by defining
6226 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
6227 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
6228 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
6229 create a distinct @code{readonly_data_section}, the default is to
6230 reuse @code{text_section}.
6232 All the other @file{varasm.c} sections are optional, and are null
6233 if the target does not provide them.
6235 @defmac TEXT_SECTION_ASM_OP
6236 A C expression whose value is a string, including spacing, containing the
6237 assembler operation that should precede instructions and read-only data.
6238 Normally @code{"\t.text"} is right.
6241 @defmac HOT_TEXT_SECTION_NAME
6242 If defined, a C string constant for the name of the section containing most
6243 frequently executed functions of the program. If not defined, GCC will provide
6244 a default definition if the target supports named sections.
6247 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
6248 If defined, a C string constant for the name of the section containing unlikely
6249 executed functions in the program.
6252 @defmac DATA_SECTION_ASM_OP
6253 A C expression whose value is a string, including spacing, containing the
6254 assembler operation to identify the following data as writable initialized
6255 data. Normally @code{"\t.data"} is right.
6258 @defmac SDATA_SECTION_ASM_OP
6259 If defined, a C expression whose value is a string, including spacing,
6260 containing the assembler operation to identify the following data as
6261 initialized, writable small data.
6264 @defmac READONLY_DATA_SECTION_ASM_OP
6265 A C expression whose value is a string, including spacing, containing the
6266 assembler operation to identify the following data as read-only initialized
6270 @defmac BSS_SECTION_ASM_OP
6271 If defined, a C expression whose value is a string, including spacing,
6272 containing the assembler operation to identify the following data as
6273 uninitialized global data. If not defined, and neither
6274 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
6275 uninitialized global data will be output in the data section if
6276 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
6280 @defmac SBSS_SECTION_ASM_OP
6281 If defined, a C expression whose value is a string, including spacing,
6282 containing the assembler operation to identify the following data as
6283 uninitialized, writable small data.
6286 @defmac INIT_SECTION_ASM_OP
6287 If defined, a C expression whose value is a string, including spacing,
6288 containing the assembler operation to identify the following data as
6289 initialization code. If not defined, GCC will assume such a section does
6290 not exist. This section has no corresponding @code{init_section}
6291 variable; it is used entirely in runtime code.
6294 @defmac FINI_SECTION_ASM_OP
6295 If defined, a C expression whose value is a string, including spacing,
6296 containing the assembler operation to identify the following data as
6297 finalization code. If not defined, GCC will assume such a section does
6298 not exist. This section has no corresponding @code{fini_section}
6299 variable; it is used entirely in runtime code.
6302 @defmac INIT_ARRAY_SECTION_ASM_OP
6303 If defined, a C expression whose value is a string, including spacing,
6304 containing the assembler operation to identify the following data as
6305 part of the @code{.init_array} (or equivalent) section. If not
6306 defined, GCC will assume such a section does not exist. Do not define
6307 both this macro and @code{INIT_SECTION_ASM_OP}.
6310 @defmac FINI_ARRAY_SECTION_ASM_OP
6311 If defined, a C expression whose value is a string, including spacing,
6312 containing the assembler operation to identify the following data as
6313 part of the @code{.fini_array} (or equivalent) section. If not
6314 defined, GCC will assume such a section does not exist. Do not define
6315 both this macro and @code{FINI_SECTION_ASM_OP}.
6318 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
6319 If defined, an ASM statement that switches to a different section
6320 via @var{section_op}, calls @var{function}, and switches back to
6321 the text section. This is used in @file{crtstuff.c} if
6322 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
6323 to initialization and finalization functions from the init and fini
6324 sections. By default, this macro uses a simple function call. Some
6325 ports need hand-crafted assembly code to avoid dependencies on
6326 registers initialized in the function prologue or to ensure that
6327 constant pools don't end up too far way in the text section.
6330 @defmac TARGET_LIBGCC_SDATA_SECTION
6331 If defined, a string which names the section into which small
6332 variables defined in crtstuff and libgcc should go. This is useful
6333 when the target has options for optimizing access to small data, and
6334 you want the crtstuff and libgcc routines to be conservative in what
6335 they expect of your application yet liberal in what your application
6336 expects. For example, for targets with a @code{.sdata} section (like
6337 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
6338 require small data support from your application, but use this macro
6339 to put small data into @code{.sdata} so that your application can
6340 access these variables whether it uses small data or not.
6343 @defmac FORCE_CODE_SECTION_ALIGN
6344 If defined, an ASM statement that aligns a code section to some
6345 arbitrary boundary. This is used to force all fragments of the
6346 @code{.init} and @code{.fini} sections to have to same alignment
6347 and thus prevent the linker from having to add any padding.
6350 @defmac JUMP_TABLES_IN_TEXT_SECTION
6351 Define this macro to be an expression with a nonzero value if jump
6352 tables (for @code{tablejump} insns) should be output in the text
6353 section, along with the assembler instructions. Otherwise, the
6354 readonly data section is used.
6356 This macro is irrelevant if there is no separate readonly data section.
6359 @deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void)
6360 Define this hook if you need to do something special to set up the
6361 @file{varasm.c} sections, or if your target has some special sections
6362 of its own that you need to create.
6364 GCC calls this hook after processing the command line, but before writing
6365 any assembly code, and before calling any of the section-returning hooks
6369 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
6370 Return the section into which @var{exp} should be placed. You can
6371 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
6372 some sort. @var{reloc} indicates whether the initial value of @var{exp}
6373 requires link-time relocations. Bit 0 is set when variable contains
6374 local relocations only, while bit 1 is set for global relocations.
6375 @var{align} is the constant alignment in bits.
6377 The default version of this function takes care of putting read-only
6378 variables in @code{readonly_data_section}.
6380 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
6383 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
6384 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
6385 for @code{FUNCTION_DECL}s as well as for variables and constants.
6387 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
6388 function has been determined to be likely to be called, and nonzero if
6389 it is unlikely to be called.
6392 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
6393 Build up a unique section name, expressed as a @code{STRING_CST} node,
6394 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
6395 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
6396 the initial value of @var{exp} requires link-time relocations.
6398 The default version of this function appends the symbol name to the
6399 ELF section name that would normally be used for the symbol. For
6400 example, the function @code{foo} would be placed in @code{.text.foo}.
6401 Whatever the actual target object format, this is often good enough.
6404 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
6405 Return the readonly data section associated with
6406 @samp{DECL_SECTION_NAME (@var{decl})}.
6407 The default version of this function selects @code{.gnu.linkonce.r.name} if
6408 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
6409 if function is in @code{.text.name}, and the normal readonly-data section
6413 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
6414 Return the section into which a constant @var{x}, of mode @var{mode},
6415 should be placed. You can assume that @var{x} is some kind of
6416 constant in RTL@. The argument @var{mode} is redundant except in the
6417 case of a @code{const_int} rtx. @var{align} is the constant alignment
6420 The default version of this function takes care of putting symbolic
6421 constants in @code{flag_pic} mode in @code{data_section} and everything
6422 else in @code{readonly_data_section}.
6425 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
6426 Define this hook if references to a symbol or a constant must be
6427 treated differently depending on something about the variable or
6428 function named by the symbol (such as what section it is in).
6430 The hook is executed immediately after rtl has been created for
6431 @var{decl}, which may be a variable or function declaration or
6432 an entry in the constant pool. In either case, @var{rtl} is the
6433 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
6434 in this hook; that field may not have been initialized yet.
6436 In the case of a constant, it is safe to assume that the rtl is
6437 a @code{mem} whose address is a @code{symbol_ref}. Most decls
6438 will also have this form, but that is not guaranteed. Global
6439 register variables, for instance, will have a @code{reg} for their
6440 rtl. (Normally the right thing to do with such unusual rtl is
6443 The @var{new_decl_p} argument will be true if this is the first time
6444 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
6445 be false for subsequent invocations, which will happen for duplicate
6446 declarations. Whether or not anything must be done for the duplicate
6447 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
6448 @var{new_decl_p} is always true when the hook is called for a constant.
6450 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
6451 The usual thing for this hook to do is to record flags in the
6452 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
6453 Historically, the name string was modified if it was necessary to
6454 encode more than one bit of information, but this practice is now
6455 discouraged; use @code{SYMBOL_REF_FLAGS}.
6457 The default definition of this hook, @code{default_encode_section_info}
6458 in @file{varasm.c}, sets a number of commonly-useful bits in
6459 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
6460 before overriding it.
6463 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
6464 Decode @var{name} and return the real name part, sans
6465 the characters that @code{TARGET_ENCODE_SECTION_INFO}
6469 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
6470 Returns true if @var{exp} should be placed into a ``small data'' section.
6471 The default version of this hook always returns false.
6474 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
6475 Contains the value true if the target places read-only
6476 ``small data'' into a separate section. The default value is false.
6479 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
6480 Returns true if @var{exp} names an object for which name resolution
6481 rules must resolve to the current ``module'' (dynamic shared library
6482 or executable image).
6484 The default version of this hook implements the name resolution rules
6485 for ELF, which has a looser model of global name binding than other
6486 currently supported object file formats.
6489 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
6490 Contains the value true if the target supports thread-local storage.
6491 The default value is false.
6496 @section Position Independent Code
6497 @cindex position independent code
6500 This section describes macros that help implement generation of position
6501 independent code. Simply defining these macros is not enough to
6502 generate valid PIC; you must also add support to the macros
6503 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
6504 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
6505 @samp{movsi} to do something appropriate when the source operand
6506 contains a symbolic address. You may also need to alter the handling of
6507 switch statements so that they use relative addresses.
6508 @c i rearranged the order of the macros above to try to force one of
6509 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
6511 @defmac PIC_OFFSET_TABLE_REGNUM
6512 The register number of the register used to address a table of static
6513 data addresses in memory. In some cases this register is defined by a
6514 processor's ``application binary interface'' (ABI)@. When this macro
6515 is defined, RTL is generated for this register once, as with the stack
6516 pointer and frame pointer registers. If this macro is not defined, it
6517 is up to the machine-dependent files to allocate such a register (if
6518 necessary). Note that this register must be fixed when in use (e.g.@:
6519 when @code{flag_pic} is true).
6522 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6523 Define this macro if the register defined by
6524 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
6525 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
6528 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
6529 A C expression that is nonzero if @var{x} is a legitimate immediate
6530 operand on the target machine when generating position independent code.
6531 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6532 check this. You can also assume @var{flag_pic} is true, so you need not
6533 check it either. You need not define this macro if all constants
6534 (including @code{SYMBOL_REF}) can be immediate operands when generating
6535 position independent code.
6538 @node Assembler Format
6539 @section Defining the Output Assembler Language
6541 This section describes macros whose principal purpose is to describe how
6542 to write instructions in assembler language---rather than what the
6546 * File Framework:: Structural information for the assembler file.
6547 * Data Output:: Output of constants (numbers, strings, addresses).
6548 * Uninitialized Data:: Output of uninitialized variables.
6549 * Label Output:: Output and generation of labels.
6550 * Initialization:: General principles of initialization
6551 and termination routines.
6552 * Macros for Initialization::
6553 Specific macros that control the handling of
6554 initialization and termination routines.
6555 * Instruction Output:: Output of actual instructions.
6556 * Dispatch Tables:: Output of jump tables.
6557 * Exception Region Output:: Output of exception region code.
6558 * Alignment Output:: Pseudo ops for alignment and skipping data.
6561 @node File Framework
6562 @subsection The Overall Framework of an Assembler File
6563 @cindex assembler format
6564 @cindex output of assembler code
6566 @c prevent bad page break with this line
6567 This describes the overall framework of an assembly file.
6569 @deftypefn {Target Hook} void TARGET_ASM_FILE_START ()
6570 @findex default_file_start
6571 Output to @code{asm_out_file} any text which the assembler expects to
6572 find at the beginning of a file. The default behavior is controlled
6573 by two flags, documented below. Unless your target's assembler is
6574 quite unusual, if you override the default, you should call
6575 @code{default_file_start} at some point in your target hook. This
6576 lets other target files rely on these variables.
6579 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
6580 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
6581 printed as the very first line in the assembly file, unless
6582 @option{-fverbose-asm} is in effect. (If that macro has been defined
6583 to the empty string, this variable has no effect.) With the normal
6584 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
6585 assembler that it need not bother stripping comments or extra
6586 whitespace from its input. This allows it to work a bit faster.
6588 The default is false. You should not set it to true unless you have
6589 verified that your port does not generate any extra whitespace or
6590 comments that will cause GAS to issue errors in NO_APP mode.
6593 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
6594 If this flag is true, @code{output_file_directive} will be called
6595 for the primary source file, immediately after printing
6596 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
6597 this to be done. The default is false.
6600 @deftypefn {Target Hook} void TARGET_ASM_FILE_END ()
6601 Output to @code{asm_out_file} any text which the assembler expects
6602 to find at the end of a file. The default is to output nothing.
6605 @deftypefun void file_end_indicate_exec_stack ()
6606 Some systems use a common convention, the @samp{.note.GNU-stack}
6607 special section, to indicate whether or not an object file relies on
6608 the stack being executable. If your system uses this convention, you
6609 should define @code{TARGET_ASM_FILE_END} to this function. If you
6610 need to do other things in that hook, have your hook function call
6614 @defmac ASM_COMMENT_START
6615 A C string constant describing how to begin a comment in the target
6616 assembler language. The compiler assumes that the comment will end at
6617 the end of the line.
6621 A C string constant for text to be output before each @code{asm}
6622 statement or group of consecutive ones. Normally this is
6623 @code{"#APP"}, which is a comment that has no effect on most
6624 assemblers but tells the GNU assembler that it must check the lines
6625 that follow for all valid assembler constructs.
6629 A C string constant for text to be output after each @code{asm}
6630 statement or group of consecutive ones. Normally this is
6631 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6632 time-saving assumptions that are valid for ordinary compiler output.
6635 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6636 A C statement to output COFF information or DWARF debugging information
6637 which indicates that filename @var{name} is the current source file to
6638 the stdio stream @var{stream}.
6640 This macro need not be defined if the standard form of output
6641 for the file format in use is appropriate.
6644 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6645 A C statement to output the string @var{string} to the stdio stream
6646 @var{stream}. If you do not call the function @code{output_quoted_string}
6647 in your config files, GCC will only call it to output filenames to
6648 the assembler source. So you can use it to canonicalize the format
6649 of the filename using this macro.
6652 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6653 A C statement to output something to the assembler file to handle a
6654 @samp{#ident} directive containing the text @var{string}. If this
6655 macro is not defined, nothing is output for a @samp{#ident} directive.
6658 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6659 Output assembly directives to switch to section @var{name}. The section
6660 should have attributes as specified by @var{flags}, which is a bit mask
6661 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6662 is nonzero, it contains an alignment in bytes to be used for the section,
6663 otherwise some target default should be used. Only targets that must
6664 specify an alignment within the section directive need pay attention to
6665 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6668 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6669 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6672 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
6673 @deftypefn {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
6674 This flag is true if we can create zeroed data by switching to a BSS
6675 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
6676 This is true on most ELF targets.
6679 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6680 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6681 based on a variable or function decl, a section name, and whether or not the
6682 declaration's initializer may contain runtime relocations. @var{decl} may be
6683 null, in which case read-write data should be assumed.
6685 The default version of this function handles choosing code vs data,
6686 read-only vs read-write data, and @code{flag_pic}. You should only
6687 need to override this if your target has special flags that might be
6688 set via @code{__attribute__}.
6691 @deftypefn {Target Hook} {int} TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type @var{type}, const char * @var{text})
6692 Provides the target with the ability to record the gcc command line
6693 switches that have been passed to the compiler, and options that are
6694 enabled. The @var{type} argument specifies what is being recorded.
6695 It can take the following values:
6698 @item SWITCH_TYPE_PASSED
6699 @var{text} is a command line switch that has been set by the user.
6701 @item SWITCH_TYPE_ENABLED
6702 @var{text} is an option which has been enabled. This might be as a
6703 direct result of a command line switch, or because it is enabled by
6704 default or because it has been enabled as a side effect of a different
6705 command line switch. For example, the @option{-O2} switch enables
6706 various different individual optimization passes.
6708 @item SWITCH_TYPE_DESCRIPTIVE
6709 @var{text} is either NULL or some descriptive text which should be
6710 ignored. If @var{text} is NULL then it is being used to warn the
6711 target hook that either recording is starting or ending. The first
6712 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
6713 warning is for start up and the second time the warning is for
6714 wind down. This feature is to allow the target hook to make any
6715 necessary preparations before it starts to record switches and to
6716 perform any necessary tidying up after it has finished recording
6719 @item SWITCH_TYPE_LINE_START
6720 This option can be ignored by this target hook.
6722 @item SWITCH_TYPE_LINE_END
6723 This option can be ignored by this target hook.
6726 The hook's return value must be zero. Other return values may be
6727 supported in the future.
6729 By default this hook is set to NULL, but an example implementation is
6730 provided for ELF based targets. Called @var{elf_record_gcc_switches},
6731 it records the switches as ASCII text inside a new, string mergeable
6732 section in the assembler output file. The name of the new section is
6733 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
6737 @deftypefn {Target Hook} {const char *} TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
6738 This is the name of the section that will be created by the example
6739 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
6745 @subsection Output of Data
6748 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6749 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6750 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6751 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6752 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6753 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6754 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6755 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6756 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6757 These hooks specify assembly directives for creating certain kinds
6758 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6759 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6760 aligned two-byte object, and so on. Any of the hooks may be
6761 @code{NULL}, indicating that no suitable directive is available.
6763 The compiler will print these strings at the start of a new line,
6764 followed immediately by the object's initial value. In most cases,
6765 the string should contain a tab, a pseudo-op, and then another tab.
6768 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6769 The @code{assemble_integer} function uses this hook to output an
6770 integer object. @var{x} is the object's value, @var{size} is its size
6771 in bytes and @var{aligned_p} indicates whether it is aligned. The
6772 function should return @code{true} if it was able to output the
6773 object. If it returns false, @code{assemble_integer} will try to
6774 split the object into smaller parts.
6776 The default implementation of this hook will use the
6777 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6778 when the relevant string is @code{NULL}.
6781 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6782 A C statement to recognize @var{rtx} patterns that
6783 @code{output_addr_const} can't deal with, and output assembly code to
6784 @var{stream} corresponding to the pattern @var{x}. This may be used to
6785 allow machine-dependent @code{UNSPEC}s to appear within constants.
6787 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6788 @code{goto fail}, so that a standard error message is printed. If it
6789 prints an error message itself, by calling, for example,
6790 @code{output_operand_lossage}, it may just complete normally.
6793 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6794 A C statement to output to the stdio stream @var{stream} an assembler
6795 instruction to assemble a string constant containing the @var{len}
6796 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6797 @code{char *} and @var{len} a C expression of type @code{int}.
6799 If the assembler has a @code{.ascii} pseudo-op as found in the
6800 Berkeley Unix assembler, do not define the macro
6801 @code{ASM_OUTPUT_ASCII}.
6804 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6805 A C statement to output word @var{n} of a function descriptor for
6806 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6807 is defined, and is otherwise unused.
6810 @defmac CONSTANT_POOL_BEFORE_FUNCTION
6811 You may define this macro as a C expression. You should define the
6812 expression to have a nonzero value if GCC should output the constant
6813 pool for a function before the code for the function, or a zero value if
6814 GCC should output the constant pool after the function. If you do
6815 not define this macro, the usual case, GCC will output the constant
6816 pool before the function.
6819 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
6820 A C statement to output assembler commands to define the start of the
6821 constant pool for a function. @var{funname} is a string giving
6822 the name of the function. Should the return type of the function
6823 be required, it can be obtained via @var{fundecl}. @var{size}
6824 is the size, in bytes, of the constant pool that will be written
6825 immediately after this call.
6827 If no constant-pool prefix is required, the usual case, this macro need
6831 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
6832 A C statement (with or without semicolon) to output a constant in the
6833 constant pool, if it needs special treatment. (This macro need not do
6834 anything for RTL expressions that can be output normally.)
6836 The argument @var{file} is the standard I/O stream to output the
6837 assembler code on. @var{x} is the RTL expression for the constant to
6838 output, and @var{mode} is the machine mode (in case @var{x} is a
6839 @samp{const_int}). @var{align} is the required alignment for the value
6840 @var{x}; you should output an assembler directive to force this much
6843 The argument @var{labelno} is a number to use in an internal label for
6844 the address of this pool entry. The definition of this macro is
6845 responsible for outputting the label definition at the proper place.
6846 Here is how to do this:
6849 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
6852 When you output a pool entry specially, you should end with a
6853 @code{goto} to the label @var{jumpto}. This will prevent the same pool
6854 entry from being output a second time in the usual manner.
6856 You need not define this macro if it would do nothing.
6859 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
6860 A C statement to output assembler commands to at the end of the constant
6861 pool for a function. @var{funname} is a string giving the name of the
6862 function. Should the return type of the function be required, you can
6863 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
6864 constant pool that GCC wrote immediately before this call.
6866 If no constant-pool epilogue is required, the usual case, you need not
6870 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
6871 Define this macro as a C expression which is nonzero if @var{C} is
6872 used as a logical line separator by the assembler.
6874 If you do not define this macro, the default is that only
6875 the character @samp{;} is treated as a logical line separator.
6878 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
6879 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
6880 These target hooks are C string constants, describing the syntax in the
6881 assembler for grouping arithmetic expressions. If not overridden, they
6882 default to normal parentheses, which is correct for most assemblers.
6885 These macros are provided by @file{real.h} for writing the definitions
6886 of @code{ASM_OUTPUT_DOUBLE} and the like:
6888 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
6889 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
6890 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
6891 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
6892 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
6893 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
6894 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
6895 target's floating point representation, and store its bit pattern in
6896 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
6897 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
6898 simple @code{long int}. For the others, it should be an array of
6899 @code{long int}. The number of elements in this array is determined
6900 by the size of the desired target floating point data type: 32 bits of
6901 it go in each @code{long int} array element. Each array element holds
6902 32 bits of the result, even if @code{long int} is wider than 32 bits
6903 on the host machine.
6905 The array element values are designed so that you can print them out
6906 using @code{fprintf} in the order they should appear in the target
6910 @node Uninitialized Data
6911 @subsection Output of Uninitialized Variables
6913 Each of the macros in this section is used to do the whole job of
6914 outputting a single uninitialized variable.
6916 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6917 A C statement (sans semicolon) to output to the stdio stream
6918 @var{stream} the assembler definition of a common-label named
6919 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6920 is the size rounded up to whatever alignment the caller wants.
6922 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6923 output the name itself; before and after that, output the additional
6924 assembler syntax for defining the name, and a newline.
6926 This macro controls how the assembler definitions of uninitialized
6927 common global variables are output.
6930 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
6931 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
6932 separate, explicit argument. If you define this macro, it is used in
6933 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
6934 handling the required alignment of the variable. The alignment is specified
6935 as the number of bits.
6938 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6939 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
6940 variable to be output, if there is one, or @code{NULL_TREE} if there
6941 is no corresponding variable. If you define this macro, GCC will use it
6942 in place of both @code{ASM_OUTPUT_COMMON} and
6943 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
6944 the variable's decl in order to chose what to output.
6947 @defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6948 A C statement (sans semicolon) to output to the stdio stream
6949 @var{stream} the assembler definition of uninitialized global @var{decl} named
6950 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6951 is the size rounded up to whatever alignment the caller wants.
6953 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
6954 defining this macro. If unable, use the expression
6955 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
6956 before and after that, output the additional assembler syntax for defining
6957 the name, and a newline.
6959 There are two ways of handling global BSS. One is to define either
6960 this macro or its aligned counterpart, @code{ASM_OUTPUT_ALIGNED_BSS}.
6961 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
6962 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
6963 You do not need to do both.
6965 Some languages do not have @code{common} data, and require a
6966 non-common form of global BSS in order to handle uninitialized globals
6967 efficiently. C++ is one example of this. However, if the target does
6968 not support global BSS, the front end may choose to make globals
6969 common in order to save space in the object file.
6972 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6973 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
6974 separate, explicit argument. If you define this macro, it is used in
6975 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
6976 handling the required alignment of the variable. The alignment is specified
6977 as the number of bits.
6979 Try to use function @code{asm_output_aligned_bss} defined in file
6980 @file{varasm.c} when defining this macro.
6983 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6984 A C statement (sans semicolon) to output to the stdio stream
6985 @var{stream} the assembler definition of a local-common-label named
6986 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6987 is the size rounded up to whatever alignment the caller wants.
6989 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6990 output the name itself; before and after that, output the additional
6991 assembler syntax for defining the name, and a newline.
6993 This macro controls how the assembler definitions of uninitialized
6994 static variables are output.
6997 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
6998 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
6999 separate, explicit argument. If you define this macro, it is used in
7000 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
7001 handling the required alignment of the variable. The alignment is specified
7002 as the number of bits.
7005 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7006 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
7007 variable to be output, if there is one, or @code{NULL_TREE} if there
7008 is no corresponding variable. If you define this macro, GCC will use it
7009 in place of both @code{ASM_OUTPUT_DECL} and
7010 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
7011 the variable's decl in order to chose what to output.
7015 @subsection Output and Generation of Labels
7017 @c prevent bad page break with this line
7018 This is about outputting labels.
7020 @findex assemble_name
7021 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
7022 A C statement (sans semicolon) to output to the stdio stream
7023 @var{stream} the assembler definition of a label named @var{name}.
7024 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7025 output the name itself; before and after that, output the additional
7026 assembler syntax for defining the name, and a newline. A default
7027 definition of this macro is provided which is correct for most systems.
7030 @findex assemble_name_raw
7031 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
7032 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
7033 to refer to a compiler-generated label. The default definition uses
7034 @code{assemble_name_raw}, which is like @code{assemble_name} except
7035 that it is more efficient.
7039 A C string containing the appropriate assembler directive to specify the
7040 size of a symbol, without any arguments. On systems that use ELF, the
7041 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
7042 systems, the default is not to define this macro.
7044 Define this macro only if it is correct to use the default definitions
7045 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
7046 for your system. If you need your own custom definitions of those
7047 macros, or if you do not need explicit symbol sizes at all, do not
7051 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
7052 A C statement (sans semicolon) to output to the stdio stream
7053 @var{stream} a directive telling the assembler that the size of the
7054 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
7055 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7059 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
7060 A C statement (sans semicolon) to output to the stdio stream
7061 @var{stream} a directive telling the assembler to calculate the size of
7062 the symbol @var{name} by subtracting its address from the current
7065 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7066 provided. The default assumes that the assembler recognizes a special
7067 @samp{.} symbol as referring to the current address, and can calculate
7068 the difference between this and another symbol. If your assembler does
7069 not recognize @samp{.} or cannot do calculations with it, you will need
7070 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
7074 A C string containing the appropriate assembler directive to specify the
7075 type of a symbol, without any arguments. On systems that use ELF, the
7076 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
7077 systems, the default is not to define this macro.
7079 Define this macro only if it is correct to use the default definition of
7080 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7081 custom definition of this macro, or if you do not need explicit symbol
7082 types at all, do not define this macro.
7085 @defmac TYPE_OPERAND_FMT
7086 A C string which specifies (using @code{printf} syntax) the format of
7087 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
7088 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
7089 the default is not to define this macro.
7091 Define this macro only if it is correct to use the default definition of
7092 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7093 custom definition of this macro, or if you do not need explicit symbol
7094 types at all, do not define this macro.
7097 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
7098 A C statement (sans semicolon) to output to the stdio stream
7099 @var{stream} a directive telling the assembler that the type of the
7100 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
7101 that string is always either @samp{"function"} or @samp{"object"}, but
7102 you should not count on this.
7104 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
7105 definition of this macro is provided.
7108 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
7109 A C statement (sans semicolon) to output to the stdio stream
7110 @var{stream} any text necessary for declaring the name @var{name} of a
7111 function which is being defined. This macro is responsible for
7112 outputting the label definition (perhaps using
7113 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
7114 @code{FUNCTION_DECL} tree node representing the function.
7116 If this macro is not defined, then the function name is defined in the
7117 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7119 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7123 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
7124 A C statement (sans semicolon) to output to the stdio stream
7125 @var{stream} any text necessary for declaring the size of a function
7126 which is being defined. The argument @var{name} is the name of the
7127 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
7128 representing the function.
7130 If this macro is not defined, then the function size is not defined.
7132 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
7136 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
7137 A C statement (sans semicolon) to output to the stdio stream
7138 @var{stream} any text necessary for declaring the name @var{name} of an
7139 initialized variable which is being defined. This macro must output the
7140 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
7141 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
7143 If this macro is not defined, then the variable name is defined in the
7144 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7146 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
7147 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
7150 @defmac ASM_DECLARE_CONSTANT_NAME (@var{stream}, @var{name}, @var{exp}, @var{size})
7151 A C statement (sans semicolon) to output to the stdio stream
7152 @var{stream} any text necessary for declaring the name @var{name} of a
7153 constant which is being defined. This macro is responsible for
7154 outputting the label definition (perhaps using
7155 @code{ASM_OUTPUT_LABEL}). The argument @var{exp} is the
7156 value of the constant, and @var{size} is the size of the constant
7157 in bytes. @var{name} will be an internal label.
7159 If this macro is not defined, then the @var{name} is defined in the
7160 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7162 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7166 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
7167 A C statement (sans semicolon) to output to the stdio stream
7168 @var{stream} any text necessary for claiming a register @var{regno}
7169 for a global variable @var{decl} with name @var{name}.
7171 If you don't define this macro, that is equivalent to defining it to do
7175 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
7176 A C statement (sans semicolon) to finish up declaring a variable name
7177 once the compiler has processed its initializer fully and thus has had a
7178 chance to determine the size of an array when controlled by an
7179 initializer. This is used on systems where it's necessary to declare
7180 something about the size of the object.
7182 If you don't define this macro, that is equivalent to defining it to do
7185 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
7186 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
7189 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
7190 This target hook is a function to output to the stdio stream
7191 @var{stream} some commands that will make the label @var{name} global;
7192 that is, available for reference from other files.
7194 The default implementation relies on a proper definition of
7195 @code{GLOBAL_ASM_OP}.
7198 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *@var{stream}, tree @var{decl})
7199 This target hook is a function to output to the stdio stream
7200 @var{stream} some commands that will make the name associated with @var{decl}
7201 global; that is, available for reference from other files.
7203 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
7206 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
7207 A C statement (sans semicolon) to output to the stdio stream
7208 @var{stream} some commands that will make the label @var{name} weak;
7209 that is, available for reference from other files but only used if
7210 no other definition is available. Use the expression
7211 @code{assemble_name (@var{stream}, @var{name})} to output the name
7212 itself; before and after that, output the additional assembler syntax
7213 for making that name weak, and a newline.
7215 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
7216 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
7220 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
7221 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
7222 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
7223 or variable decl. If @var{value} is not @code{NULL}, this C statement
7224 should output to the stdio stream @var{stream} assembler code which
7225 defines (equates) the weak symbol @var{name} to have the value
7226 @var{value}. If @var{value} is @code{NULL}, it should output commands
7227 to make @var{name} weak.
7230 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
7231 Outputs a directive that enables @var{name} to be used to refer to
7232 symbol @var{value} with weak-symbol semantics. @code{decl} is the
7233 declaration of @code{name}.
7236 @defmac SUPPORTS_WEAK
7237 A C expression which evaluates to true if the target supports weak symbols.
7239 If you don't define this macro, @file{defaults.h} provides a default
7240 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
7241 is defined, the default definition is @samp{1}; otherwise, it is
7242 @samp{0}. Define this macro if you want to control weak symbol support
7243 with a compiler flag such as @option{-melf}.
7246 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
7247 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
7248 public symbol such that extra copies in multiple translation units will
7249 be discarded by the linker. Define this macro if your object file
7250 format provides support for this concept, such as the @samp{COMDAT}
7251 section flags in the Microsoft Windows PE/COFF format, and this support
7252 requires changes to @var{decl}, such as putting it in a separate section.
7255 @defmac SUPPORTS_ONE_ONLY
7256 A C expression which evaluates to true if the target supports one-only
7259 If you don't define this macro, @file{varasm.c} provides a default
7260 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
7261 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
7262 you want to control one-only symbol support with a compiler flag, or if
7263 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
7264 be emitted as one-only.
7267 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
7268 This target hook is a function to output to @var{asm_out_file} some
7269 commands that will make the symbol(s) associated with @var{decl} have
7270 hidden, protected or internal visibility as specified by @var{visibility}.
7273 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
7274 A C expression that evaluates to true if the target's linker expects
7275 that weak symbols do not appear in a static archive's table of contents.
7276 The default is @code{0}.
7278 Leaving weak symbols out of an archive's table of contents means that,
7279 if a symbol will only have a definition in one translation unit and
7280 will have undefined references from other translation units, that
7281 symbol should not be weak. Defining this macro to be nonzero will
7282 thus have the effect that certain symbols that would normally be weak
7283 (explicit template instantiations, and vtables for polymorphic classes
7284 with noninline key methods) will instead be nonweak.
7286 The C++ ABI requires this macro to be zero. Define this macro for
7287 targets where full C++ ABI compliance is impossible and where linker
7288 restrictions require weak symbols to be left out of a static archive's
7292 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
7293 A C statement (sans semicolon) to output to the stdio stream
7294 @var{stream} any text necessary for declaring the name of an external
7295 symbol named @var{name} which is referenced in this compilation but
7296 not defined. The value of @var{decl} is the tree node for the
7299 This macro need not be defined if it does not need to output anything.
7300 The GNU assembler and most Unix assemblers don't require anything.
7303 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
7304 This target hook is a function to output to @var{asm_out_file} an assembler
7305 pseudo-op to declare a library function name external. The name of the
7306 library function is given by @var{symref}, which is a @code{symbol_ref}.
7309 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (tree @var{decl})
7310 This target hook is a function to output to @var{asm_out_file} an assembler
7311 directive to annotate used symbol. Darwin target use .no_dead_code_strip
7315 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
7316 A C statement (sans semicolon) to output to the stdio stream
7317 @var{stream} a reference in assembler syntax to a label named
7318 @var{name}. This should add @samp{_} to the front of the name, if that
7319 is customary on your operating system, as it is in most Berkeley Unix
7320 systems. This macro is used in @code{assemble_name}.
7323 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
7324 A C statement (sans semicolon) to output a reference to
7325 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
7326 will be used to output the name of the symbol. This macro may be used
7327 to modify the way a symbol is referenced depending on information
7328 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
7331 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
7332 A C statement (sans semicolon) to output a reference to @var{buf}, the
7333 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
7334 @code{assemble_name} will be used to output the name of the symbol.
7335 This macro is not used by @code{output_asm_label}, or the @code{%l}
7336 specifier that calls it; the intention is that this macro should be set
7337 when it is necessary to output a label differently when its address is
7341 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
7342 A function to output to the stdio stream @var{stream} a label whose
7343 name is made from the string @var{prefix} and the number @var{labelno}.
7345 It is absolutely essential that these labels be distinct from the labels
7346 used for user-level functions and variables. Otherwise, certain programs
7347 will have name conflicts with internal labels.
7349 It is desirable to exclude internal labels from the symbol table of the
7350 object file. Most assemblers have a naming convention for labels that
7351 should be excluded; on many systems, the letter @samp{L} at the
7352 beginning of a label has this effect. You should find out what
7353 convention your system uses, and follow it.
7355 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
7358 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
7359 A C statement to output to the stdio stream @var{stream} a debug info
7360 label whose name is made from the string @var{prefix} and the number
7361 @var{num}. This is useful for VLIW targets, where debug info labels
7362 may need to be treated differently than branch target labels. On some
7363 systems, branch target labels must be at the beginning of instruction
7364 bundles, but debug info labels can occur in the middle of instruction
7367 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
7371 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
7372 A C statement to store into the string @var{string} a label whose name
7373 is made from the string @var{prefix} and the number @var{num}.
7375 This string, when output subsequently by @code{assemble_name}, should
7376 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
7377 with the same @var{prefix} and @var{num}.
7379 If the string begins with @samp{*}, then @code{assemble_name} will
7380 output the rest of the string unchanged. It is often convenient for
7381 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
7382 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
7383 to output the string, and may change it. (Of course,
7384 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
7385 you should know what it does on your machine.)
7388 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
7389 A C expression to assign to @var{outvar} (which is a variable of type
7390 @code{char *}) a newly allocated string made from the string
7391 @var{name} and the number @var{number}, with some suitable punctuation
7392 added. Use @code{alloca} to get space for the string.
7394 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
7395 produce an assembler label for an internal static variable whose name is
7396 @var{name}. Therefore, the string must be such as to result in valid
7397 assembler code. The argument @var{number} is different each time this
7398 macro is executed; it prevents conflicts between similarly-named
7399 internal static variables in different scopes.
7401 Ideally this string should not be a valid C identifier, to prevent any
7402 conflict with the user's own symbols. Most assemblers allow periods
7403 or percent signs in assembler symbols; putting at least one of these
7404 between the name and the number will suffice.
7406 If this macro is not defined, a default definition will be provided
7407 which is correct for most systems.
7410 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
7411 A C statement to output to the stdio stream @var{stream} assembler code
7412 which defines (equates) the symbol @var{name} to have the value @var{value}.
7415 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7416 correct for most systems.
7419 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
7420 A C statement to output to the stdio stream @var{stream} assembler code
7421 which defines (equates) the symbol whose tree node is @var{decl_of_name}
7422 to have the value of the tree node @var{decl_of_value}. This macro will
7423 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
7424 the tree nodes are available.
7427 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7428 correct for most systems.
7431 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
7432 A C statement that evaluates to true if the assembler code which defines
7433 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
7434 of the tree node @var{decl_of_value} should be emitted near the end of the
7435 current compilation unit. The default is to not defer output of defines.
7436 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
7437 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
7440 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
7441 A C statement to output to the stdio stream @var{stream} assembler code
7442 which defines (equates) the weak symbol @var{name} to have the value
7443 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
7444 an undefined weak symbol.
7446 Define this macro if the target only supports weak aliases; define
7447 @code{ASM_OUTPUT_DEF} instead if possible.
7450 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
7451 Define this macro to override the default assembler names used for
7452 Objective-C methods.
7454 The default name is a unique method number followed by the name of the
7455 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
7456 the category is also included in the assembler name (e.g.@:
7459 These names are safe on most systems, but make debugging difficult since
7460 the method's selector is not present in the name. Therefore, particular
7461 systems define other ways of computing names.
7463 @var{buf} is an expression of type @code{char *} which gives you a
7464 buffer in which to store the name; its length is as long as
7465 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
7466 50 characters extra.
7468 The argument @var{is_inst} specifies whether the method is an instance
7469 method or a class method; @var{class_name} is the name of the class;
7470 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
7471 in a category); and @var{sel_name} is the name of the selector.
7473 On systems where the assembler can handle quoted names, you can use this
7474 macro to provide more human-readable names.
7477 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
7478 A C statement (sans semicolon) to output to the stdio stream
7479 @var{stream} commands to declare that the label @var{name} is an
7480 Objective-C class reference. This is only needed for targets whose
7481 linkers have special support for NeXT-style runtimes.
7484 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
7485 A C statement (sans semicolon) to output to the stdio stream
7486 @var{stream} commands to declare that the label @var{name} is an
7487 unresolved Objective-C class reference. This is only needed for targets
7488 whose linkers have special support for NeXT-style runtimes.
7491 @node Initialization
7492 @subsection How Initialization Functions Are Handled
7493 @cindex initialization routines
7494 @cindex termination routines
7495 @cindex constructors, output of
7496 @cindex destructors, output of
7498 The compiled code for certain languages includes @dfn{constructors}
7499 (also called @dfn{initialization routines})---functions to initialize
7500 data in the program when the program is started. These functions need
7501 to be called before the program is ``started''---that is to say, before
7502 @code{main} is called.
7504 Compiling some languages generates @dfn{destructors} (also called
7505 @dfn{termination routines}) that should be called when the program
7508 To make the initialization and termination functions work, the compiler
7509 must output something in the assembler code to cause those functions to
7510 be called at the appropriate time. When you port the compiler to a new
7511 system, you need to specify how to do this.
7513 There are two major ways that GCC currently supports the execution of
7514 initialization and termination functions. Each way has two variants.
7515 Much of the structure is common to all four variations.
7517 @findex __CTOR_LIST__
7518 @findex __DTOR_LIST__
7519 The linker must build two lists of these functions---a list of
7520 initialization functions, called @code{__CTOR_LIST__}, and a list of
7521 termination functions, called @code{__DTOR_LIST__}.
7523 Each list always begins with an ignored function pointer (which may hold
7524 0, @minus{}1, or a count of the function pointers after it, depending on
7525 the environment). This is followed by a series of zero or more function
7526 pointers to constructors (or destructors), followed by a function
7527 pointer containing zero.
7529 Depending on the operating system and its executable file format, either
7530 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
7531 time and exit time. Constructors are called in reverse order of the
7532 list; destructors in forward order.
7534 The best way to handle static constructors works only for object file
7535 formats which provide arbitrarily-named sections. A section is set
7536 aside for a list of constructors, and another for a list of destructors.
7537 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
7538 object file that defines an initialization function also puts a word in
7539 the constructor section to point to that function. The linker
7540 accumulates all these words into one contiguous @samp{.ctors} section.
7541 Termination functions are handled similarly.
7543 This method will be chosen as the default by @file{target-def.h} if
7544 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
7545 support arbitrary sections, but does support special designated
7546 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
7547 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
7549 When arbitrary sections are available, there are two variants, depending
7550 upon how the code in @file{crtstuff.c} is called. On systems that
7551 support a @dfn{.init} section which is executed at program startup,
7552 parts of @file{crtstuff.c} are compiled into that section. The
7553 program is linked by the @command{gcc} driver like this:
7556 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
7559 The prologue of a function (@code{__init}) appears in the @code{.init}
7560 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
7561 for the function @code{__fini} in the @dfn{.fini} section. Normally these
7562 files are provided by the operating system or by the GNU C library, but
7563 are provided by GCC for a few targets.
7565 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
7566 compiled from @file{crtstuff.c}. They contain, among other things, code
7567 fragments within the @code{.init} and @code{.fini} sections that branch
7568 to routines in the @code{.text} section. The linker will pull all parts
7569 of a section together, which results in a complete @code{__init} function
7570 that invokes the routines we need at startup.
7572 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
7575 If no init section is available, when GCC compiles any function called
7576 @code{main} (or more accurately, any function designated as a program
7577 entry point by the language front end calling @code{expand_main_function}),
7578 it inserts a procedure call to @code{__main} as the first executable code
7579 after the function prologue. The @code{__main} function is defined
7580 in @file{libgcc2.c} and runs the global constructors.
7582 In file formats that don't support arbitrary sections, there are again
7583 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
7584 and an `a.out' format must be used. In this case,
7585 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
7586 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
7587 and with the address of the void function containing the initialization
7588 code as its value. The GNU linker recognizes this as a request to add
7589 the value to a @dfn{set}; the values are accumulated, and are eventually
7590 placed in the executable as a vector in the format described above, with
7591 a leading (ignored) count and a trailing zero element.
7592 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
7593 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
7594 the compilation of @code{main} to call @code{__main} as above, starting
7595 the initialization process.
7597 The last variant uses neither arbitrary sections nor the GNU linker.
7598 This is preferable when you want to do dynamic linking and when using
7599 file formats which the GNU linker does not support, such as `ECOFF'@. In
7600 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
7601 termination functions are recognized simply by their names. This requires
7602 an extra program in the linkage step, called @command{collect2}. This program
7603 pretends to be the linker, for use with GCC; it does its job by running
7604 the ordinary linker, but also arranges to include the vectors of
7605 initialization and termination functions. These functions are called
7606 via @code{__main} as described above. In order to use this method,
7607 @code{use_collect2} must be defined in the target in @file{config.gcc}.
7610 The following section describes the specific macros that control and
7611 customize the handling of initialization and termination functions.
7614 @node Macros for Initialization
7615 @subsection Macros Controlling Initialization Routines
7617 Here are the macros that control how the compiler handles initialization
7618 and termination functions:
7620 @defmac INIT_SECTION_ASM_OP
7621 If defined, a C string constant, including spacing, for the assembler
7622 operation to identify the following data as initialization code. If not
7623 defined, GCC will assume such a section does not exist. When you are
7624 using special sections for initialization and termination functions, this
7625 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7626 run the initialization functions.
7629 @defmac HAS_INIT_SECTION
7630 If defined, @code{main} will not call @code{__main} as described above.
7631 This macro should be defined for systems that control start-up code
7632 on a symbol-by-symbol basis, such as OSF/1, and should not
7633 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7636 @defmac LD_INIT_SWITCH
7637 If defined, a C string constant for a switch that tells the linker that
7638 the following symbol is an initialization routine.
7641 @defmac LD_FINI_SWITCH
7642 If defined, a C string constant for a switch that tells the linker that
7643 the following symbol is a finalization routine.
7646 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7647 If defined, a C statement that will write a function that can be
7648 automatically called when a shared library is loaded. The function
7649 should call @var{func}, which takes no arguments. If not defined, and
7650 the object format requires an explicit initialization function, then a
7651 function called @code{_GLOBAL__DI} will be generated.
7653 This function and the following one are used by collect2 when linking a
7654 shared library that needs constructors or destructors, or has DWARF2
7655 exception tables embedded in the code.
7658 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7659 If defined, a C statement that will write a function that can be
7660 automatically called when a shared library is unloaded. The function
7661 should call @var{func}, which takes no arguments. If not defined, and
7662 the object format requires an explicit finalization function, then a
7663 function called @code{_GLOBAL__DD} will be generated.
7666 @defmac INVOKE__main
7667 If defined, @code{main} will call @code{__main} despite the presence of
7668 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7669 where the init section is not actually run automatically, but is still
7670 useful for collecting the lists of constructors and destructors.
7673 @defmac SUPPORTS_INIT_PRIORITY
7674 If nonzero, the C++ @code{init_priority} attribute is supported and the
7675 compiler should emit instructions to control the order of initialization
7676 of objects. If zero, the compiler will issue an error message upon
7677 encountering an @code{init_priority} attribute.
7680 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7681 This value is true if the target supports some ``native'' method of
7682 collecting constructors and destructors to be run at startup and exit.
7683 It is false if we must use @command{collect2}.
7686 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7687 If defined, a function that outputs assembler code to arrange to call
7688 the function referenced by @var{symbol} at initialization time.
7690 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7691 no arguments and with no return value. If the target supports initialization
7692 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7693 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7695 If this macro is not defined by the target, a suitable default will
7696 be chosen if (1) the target supports arbitrary section names, (2) the
7697 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7701 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7702 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7703 functions rather than initialization functions.
7706 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7707 generated for the generated object file will have static linkage.
7709 If your system uses @command{collect2} as the means of processing
7710 constructors, then that program normally uses @command{nm} to scan
7711 an object file for constructor functions to be called.
7713 On certain kinds of systems, you can define this macro to make
7714 @command{collect2} work faster (and, in some cases, make it work at all):
7716 @defmac OBJECT_FORMAT_COFF
7717 Define this macro if the system uses COFF (Common Object File Format)
7718 object files, so that @command{collect2} can assume this format and scan
7719 object files directly for dynamic constructor/destructor functions.
7721 This macro is effective only in a native compiler; @command{collect2} as
7722 part of a cross compiler always uses @command{nm} for the target machine.
7725 @defmac REAL_NM_FILE_NAME
7726 Define this macro as a C string constant containing the file name to use
7727 to execute @command{nm}. The default is to search the path normally for
7730 If your system supports shared libraries and has a program to list the
7731 dynamic dependencies of a given library or executable, you can define
7732 these macros to enable support for running initialization and
7733 termination functions in shared libraries:
7737 Define this macro to a C string constant containing the name of the program
7738 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7741 @defmac PARSE_LDD_OUTPUT (@var{ptr})
7742 Define this macro to be C code that extracts filenames from the output
7743 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7744 of type @code{char *} that points to the beginning of a line of output
7745 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7746 code must advance @var{ptr} to the beginning of the filename on that
7747 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7750 @node Instruction Output
7751 @subsection Output of Assembler Instructions
7753 @c prevent bad page break with this line
7754 This describes assembler instruction output.
7756 @defmac REGISTER_NAMES
7757 A C initializer containing the assembler's names for the machine
7758 registers, each one as a C string constant. This is what translates
7759 register numbers in the compiler into assembler language.
7762 @defmac ADDITIONAL_REGISTER_NAMES
7763 If defined, a C initializer for an array of structures containing a name
7764 and a register number. This macro defines additional names for hard
7765 registers, thus allowing the @code{asm} option in declarations to refer
7766 to registers using alternate names.
7769 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7770 Define this macro if you are using an unusual assembler that
7771 requires different names for the machine instructions.
7773 The definition is a C statement or statements which output an
7774 assembler instruction opcode to the stdio stream @var{stream}. The
7775 macro-operand @var{ptr} is a variable of type @code{char *} which
7776 points to the opcode name in its ``internal'' form---the form that is
7777 written in the machine description. The definition should output the
7778 opcode name to @var{stream}, performing any translation you desire, and
7779 increment the variable @var{ptr} to point at the end of the opcode
7780 so that it will not be output twice.
7782 In fact, your macro definition may process less than the entire opcode
7783 name, or more than the opcode name; but if you want to process text
7784 that includes @samp{%}-sequences to substitute operands, you must take
7785 care of the substitution yourself. Just be sure to increment
7786 @var{ptr} over whatever text should not be output normally.
7788 @findex recog_data.operand
7789 If you need to look at the operand values, they can be found as the
7790 elements of @code{recog_data.operand}.
7792 If the macro definition does nothing, the instruction is output
7796 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7797 If defined, a C statement to be executed just prior to the output of
7798 assembler code for @var{insn}, to modify the extracted operands so
7799 they will be output differently.
7801 Here the argument @var{opvec} is the vector containing the operands
7802 extracted from @var{insn}, and @var{noperands} is the number of
7803 elements of the vector which contain meaningful data for this insn.
7804 The contents of this vector are what will be used to convert the insn
7805 template into assembler code, so you can change the assembler output
7806 by changing the contents of the vector.
7808 This macro is useful when various assembler syntaxes share a single
7809 file of instruction patterns; by defining this macro differently, you
7810 can cause a large class of instructions to be output differently (such
7811 as with rearranged operands). Naturally, variations in assembler
7812 syntax affecting individual insn patterns ought to be handled by
7813 writing conditional output routines in those patterns.
7815 If this macro is not defined, it is equivalent to a null statement.
7818 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
7819 A C compound statement to output to stdio stream @var{stream} the
7820 assembler syntax for an instruction operand @var{x}. @var{x} is an
7823 @var{code} is a value that can be used to specify one of several ways
7824 of printing the operand. It is used when identical operands must be
7825 printed differently depending on the context. @var{code} comes from
7826 the @samp{%} specification that was used to request printing of the
7827 operand. If the specification was just @samp{%@var{digit}} then
7828 @var{code} is 0; if the specification was @samp{%@var{ltr}
7829 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
7832 If @var{x} is a register, this macro should print the register's name.
7833 The names can be found in an array @code{reg_names} whose type is
7834 @code{char *[]}. @code{reg_names} is initialized from
7835 @code{REGISTER_NAMES}.
7837 When the machine description has a specification @samp{%@var{punct}}
7838 (a @samp{%} followed by a punctuation character), this macro is called
7839 with a null pointer for @var{x} and the punctuation character for
7843 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
7844 A C expression which evaluates to true if @var{code} is a valid
7845 punctuation character for use in the @code{PRINT_OPERAND} macro. If
7846 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
7847 punctuation characters (except for the standard one, @samp{%}) are used
7851 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
7852 A C compound statement to output to stdio stream @var{stream} the
7853 assembler syntax for an instruction operand that is a memory reference
7854 whose address is @var{x}. @var{x} is an RTL expression.
7856 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
7857 On some machines, the syntax for a symbolic address depends on the
7858 section that the address refers to. On these machines, define the hook
7859 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
7860 @code{symbol_ref}, and then check for it here. @xref{Assembler
7864 @findex dbr_sequence_length
7865 @defmac DBR_OUTPUT_SEQEND (@var{file})
7866 A C statement, to be executed after all slot-filler instructions have
7867 been output. If necessary, call @code{dbr_sequence_length} to
7868 determine the number of slots filled in a sequence (zero if not
7869 currently outputting a sequence), to decide how many no-ops to output,
7872 Don't define this macro if it has nothing to do, but it is helpful in
7873 reading assembly output if the extent of the delay sequence is made
7874 explicit (e.g.@: with white space).
7877 @findex final_sequence
7878 Note that output routines for instructions with delay slots must be
7879 prepared to deal with not being output as part of a sequence
7880 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
7881 found.) The variable @code{final_sequence} is null when not
7882 processing a sequence, otherwise it contains the @code{sequence} rtx
7886 @defmac REGISTER_PREFIX
7887 @defmacx LOCAL_LABEL_PREFIX
7888 @defmacx USER_LABEL_PREFIX
7889 @defmacx IMMEDIATE_PREFIX
7890 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
7891 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
7892 @file{final.c}). These are useful when a single @file{md} file must
7893 support multiple assembler formats. In that case, the various @file{tm.h}
7894 files can define these macros differently.
7897 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
7898 If defined this macro should expand to a series of @code{case}
7899 statements which will be parsed inside the @code{switch} statement of
7900 the @code{asm_fprintf} function. This allows targets to define extra
7901 printf formats which may useful when generating their assembler
7902 statements. Note that uppercase letters are reserved for future
7903 generic extensions to asm_fprintf, and so are not available to target
7904 specific code. The output file is given by the parameter @var{file}.
7905 The varargs input pointer is @var{argptr} and the rest of the format
7906 string, starting the character after the one that is being switched
7907 upon, is pointed to by @var{format}.
7910 @defmac ASSEMBLER_DIALECT
7911 If your target supports multiple dialects of assembler language (such as
7912 different opcodes), define this macro as a C expression that gives the
7913 numeric index of the assembler language dialect to use, with zero as the
7916 If this macro is defined, you may use constructs of the form
7918 @samp{@{option0|option1|option2@dots{}@}}
7921 in the output templates of patterns (@pxref{Output Template}) or in the
7922 first argument of @code{asm_fprintf}. This construct outputs
7923 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
7924 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
7925 within these strings retain their usual meaning. If there are fewer
7926 alternatives within the braces than the value of
7927 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
7929 If you do not define this macro, the characters @samp{@{}, @samp{|} and
7930 @samp{@}} do not have any special meaning when used in templates or
7931 operands to @code{asm_fprintf}.
7933 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
7934 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
7935 the variations in assembler language syntax with that mechanism. Define
7936 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
7937 if the syntax variant are larger and involve such things as different
7938 opcodes or operand order.
7941 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
7942 A C expression to output to @var{stream} some assembler code
7943 which will push hard register number @var{regno} onto the stack.
7944 The code need not be optimal, since this macro is used only when
7948 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
7949 A C expression to output to @var{stream} some assembler code
7950 which will pop hard register number @var{regno} off of the stack.
7951 The code need not be optimal, since this macro is used only when
7955 @node Dispatch Tables
7956 @subsection Output of Dispatch Tables
7958 @c prevent bad page break with this line
7959 This concerns dispatch tables.
7961 @cindex dispatch table
7962 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
7963 A C statement to output to the stdio stream @var{stream} an assembler
7964 pseudo-instruction to generate a difference between two labels.
7965 @var{value} and @var{rel} are the numbers of two internal labels. The
7966 definitions of these labels are output using
7967 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
7968 way here. For example,
7971 fprintf (@var{stream}, "\t.word L%d-L%d\n",
7972 @var{value}, @var{rel})
7975 You must provide this macro on machines where the addresses in a
7976 dispatch table are relative to the table's own address. If defined, GCC
7977 will also use this macro on all machines when producing PIC@.
7978 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
7979 mode and flags can be read.
7982 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
7983 This macro should be provided on machines where the addresses
7984 in a dispatch table are absolute.
7986 The definition should be a C statement to output to the stdio stream
7987 @var{stream} an assembler pseudo-instruction to generate a reference to
7988 a label. @var{value} is the number of an internal label whose
7989 definition is output using @code{(*targetm.asm_out.internal_label)}.
7993 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
7997 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
7998 Define this if the label before a jump-table needs to be output
7999 specially. The first three arguments are the same as for
8000 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
8001 jump-table which follows (a @code{jump_insn} containing an
8002 @code{addr_vec} or @code{addr_diff_vec}).
8004 This feature is used on system V to output a @code{swbeg} statement
8007 If this macro is not defined, these labels are output with
8008 @code{(*targetm.asm_out.internal_label)}.
8011 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
8012 Define this if something special must be output at the end of a
8013 jump-table. The definition should be a C statement to be executed
8014 after the assembler code for the table is written. It should write
8015 the appropriate code to stdio stream @var{stream}. The argument
8016 @var{table} is the jump-table insn, and @var{num} is the label-number
8017 of the preceding label.
8019 If this macro is not defined, nothing special is output at the end of
8023 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (@var{stream}, @var{decl}, @var{for_eh}, @var{empty})
8024 This target hook emits a label at the beginning of each FDE@. It
8025 should be defined on targets where FDEs need special labels, and it
8026 should write the appropriate label, for the FDE associated with the
8027 function declaration @var{decl}, to the stdio stream @var{stream}.
8028 The third argument, @var{for_eh}, is a boolean: true if this is for an
8029 exception table. The fourth argument, @var{empty}, is a boolean:
8030 true if this is a placeholder label for an omitted FDE@.
8032 The default is that FDEs are not given nonlocal labels.
8035 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (@var{stream})
8036 This target hook emits a label at the beginning of the exception table.
8037 It should be defined on targets where it is desirable for the table
8038 to be broken up according to function.
8040 The default is that no label is emitted.
8043 @deftypefn {Target Hook} void TARGET_UNWIND_EMIT (FILE * @var{stream}, rtx @var{insn})
8044 This target hook emits and assembly directives required to unwind the
8045 given instruction. This is only used when TARGET_UNWIND_INFO is set.
8048 @node Exception Region Output
8049 @subsection Assembler Commands for Exception Regions
8051 @c prevent bad page break with this line
8053 This describes commands marking the start and the end of an exception
8056 @defmac EH_FRAME_SECTION_NAME
8057 If defined, a C string constant for the name of the section containing
8058 exception handling frame unwind information. If not defined, GCC will
8059 provide a default definition if the target supports named sections.
8060 @file{crtstuff.c} uses this macro to switch to the appropriate section.
8062 You should define this symbol if your target supports DWARF 2 frame
8063 unwind information and the default definition does not work.
8066 @defmac EH_FRAME_IN_DATA_SECTION
8067 If defined, DWARF 2 frame unwind information will be placed in the
8068 data section even though the target supports named sections. This
8069 might be necessary, for instance, if the system linker does garbage
8070 collection and sections cannot be marked as not to be collected.
8072 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
8076 @defmac EH_TABLES_CAN_BE_READ_ONLY
8077 Define this macro to 1 if your target is such that no frame unwind
8078 information encoding used with non-PIC code will ever require a
8079 runtime relocation, but the linker may not support merging read-only
8080 and read-write sections into a single read-write section.
8083 @defmac MASK_RETURN_ADDR
8084 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
8085 that it does not contain any extraneous set bits in it.
8088 @defmac DWARF2_UNWIND_INFO
8089 Define this macro to 0 if your target supports DWARF 2 frame unwind
8090 information, but it does not yet work with exception handling.
8091 Otherwise, if your target supports this information (if it defines
8092 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
8093 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of 1.
8095 If @code{TARGET_UNWIND_INFO} is defined, the target specific unwinder
8096 will be used in all cases. Defining this macro will enable the generation
8097 of DWARF 2 frame debugging information.
8099 If @code{TARGET_UNWIND_INFO} is not defined, and this macro is defined to 1,
8100 the DWARF 2 unwinder will be the default exception handling mechanism;
8101 otherwise, the @code{setjmp}/@code{longjmp}-based scheme will be used by
8105 @defmac TARGET_UNWIND_INFO
8106 Define this macro if your target has ABI specified unwind tables. Usually
8107 these will be output by @code{TARGET_UNWIND_EMIT}.
8110 @deftypevar {Target Hook} bool TARGET_UNWIND_TABLES_DEFAULT
8111 This variable should be set to @code{true} if the target ABI requires unwinding
8112 tables even when exceptions are not used.
8115 @defmac MUST_USE_SJLJ_EXCEPTIONS
8116 This macro need only be defined if @code{DWARF2_UNWIND_INFO} is
8117 runtime-variable. In that case, @file{except.h} cannot correctly
8118 determine the corresponding definition of @code{MUST_USE_SJLJ_EXCEPTIONS},
8119 so the target must provide it directly.
8122 @defmac DONT_USE_BUILTIN_SETJMP
8123 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
8124 should use the @code{setjmp}/@code{longjmp} functions from the C library
8125 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
8128 @defmac DWARF_CIE_DATA_ALIGNMENT
8129 This macro need only be defined if the target might save registers in the
8130 function prologue at an offset to the stack pointer that is not aligned to
8131 @code{UNITS_PER_WORD}. The definition should be the negative minimum
8132 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
8133 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
8134 the target supports DWARF 2 frame unwind information.
8137 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
8138 Contains the value true if the target should add a zero word onto the
8139 end of a Dwarf-2 frame info section when used for exception handling.
8140 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
8144 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
8145 Given a register, this hook should return a parallel of registers to
8146 represent where to find the register pieces. Define this hook if the
8147 register and its mode are represented in Dwarf in non-contiguous
8148 locations, or if the register should be represented in more than one
8149 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
8150 If not defined, the default is to return @code{NULL_RTX}.
8153 @deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym})
8154 This hook is used to output a reference from a frame unwinding table to
8155 the type_info object identified by @var{sym}. It should return @code{true}
8156 if the reference was output. Returning @code{false} will cause the
8157 reference to be output using the normal Dwarf2 routines.
8160 @deftypefn {Target Hook} bool TARGET_ARM_EABI_UNWINDER
8161 This hook should be set to @code{true} on targets that use an ARM EABI
8162 based unwinding library, and @code{false} on other targets. This effects
8163 the format of unwinding tables, and how the unwinder in entered after
8164 running a cleanup. The default is @code{false}.
8167 @node Alignment Output
8168 @subsection Assembler Commands for Alignment
8170 @c prevent bad page break with this line
8171 This describes commands for alignment.
8173 @defmac JUMP_ALIGN (@var{label})
8174 The alignment (log base 2) to put in front of @var{label}, which is
8175 a common destination of jumps and has no fallthru incoming edge.
8177 This macro need not be defined if you don't want any special alignment
8178 to be done at such a time. Most machine descriptions do not currently
8181 Unless it's necessary to inspect the @var{label} parameter, it is better
8182 to set the variable @var{align_jumps} in the target's
8183 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8184 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
8187 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
8188 The alignment (log base 2) to put in front of @var{label}, which follows
8191 This macro need not be defined if you don't want any special alignment
8192 to be done at such a time. Most machine descriptions do not currently
8196 @defmac LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
8197 The maximum number of bytes to skip when applying
8198 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
8199 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8202 @defmac LOOP_ALIGN (@var{label})
8203 The alignment (log base 2) to put in front of @var{label}, which follows
8204 a @code{NOTE_INSN_LOOP_BEG} note.
8206 This macro need not be defined if you don't want any special alignment
8207 to be done at such a time. Most machine descriptions do not currently
8210 Unless it's necessary to inspect the @var{label} parameter, it is better
8211 to set the variable @code{align_loops} in the target's
8212 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8213 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
8216 @defmac LOOP_ALIGN_MAX_SKIP
8217 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
8218 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8221 @defmac LABEL_ALIGN (@var{label})
8222 The alignment (log base 2) to put in front of @var{label}.
8223 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
8224 the maximum of the specified values is used.
8226 Unless it's necessary to inspect the @var{label} parameter, it is better
8227 to set the variable @code{align_labels} in the target's
8228 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8229 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
8232 @defmac LABEL_ALIGN_MAX_SKIP
8233 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
8234 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8237 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
8238 A C statement to output to the stdio stream @var{stream} an assembler
8239 instruction to advance the location counter by @var{nbytes} bytes.
8240 Those bytes should be zero when loaded. @var{nbytes} will be a C
8241 expression of type @code{unsigned HOST_WIDE_INT}.
8244 @defmac ASM_NO_SKIP_IN_TEXT
8245 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
8246 text section because it fails to put zeros in the bytes that are skipped.
8247 This is true on many Unix systems, where the pseudo--op to skip bytes
8248 produces no-op instructions rather than zeros when used in the text
8252 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
8253 A C statement to output to the stdio stream @var{stream} an assembler
8254 command to advance the location counter to a multiple of 2 to the
8255 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
8258 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
8259 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
8260 for padding, if necessary.
8263 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
8264 A C statement to output to the stdio stream @var{stream} an assembler
8265 command to advance the location counter to a multiple of 2 to the
8266 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
8267 satisfy the alignment request. @var{power} and @var{max_skip} will be
8268 a C expression of type @code{int}.
8272 @node Debugging Info
8273 @section Controlling Debugging Information Format
8275 @c prevent bad page break with this line
8276 This describes how to specify debugging information.
8279 * All Debuggers:: Macros that affect all debugging formats uniformly.
8280 * DBX Options:: Macros enabling specific options in DBX format.
8281 * DBX Hooks:: Hook macros for varying DBX format.
8282 * File Names and DBX:: Macros controlling output of file names in DBX format.
8283 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
8284 * VMS Debug:: Macros for VMS debug format.
8288 @subsection Macros Affecting All Debugging Formats
8290 @c prevent bad page break with this line
8291 These macros affect all debugging formats.
8293 @defmac DBX_REGISTER_NUMBER (@var{regno})
8294 A C expression that returns the DBX register number for the compiler
8295 register number @var{regno}. In the default macro provided, the value
8296 of this expression will be @var{regno} itself. But sometimes there are
8297 some registers that the compiler knows about and DBX does not, or vice
8298 versa. In such cases, some register may need to have one number in the
8299 compiler and another for DBX@.
8301 If two registers have consecutive numbers inside GCC, and they can be
8302 used as a pair to hold a multiword value, then they @emph{must} have
8303 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
8304 Otherwise, debuggers will be unable to access such a pair, because they
8305 expect register pairs to be consecutive in their own numbering scheme.
8307 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
8308 does not preserve register pairs, then what you must do instead is
8309 redefine the actual register numbering scheme.
8312 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
8313 A C expression that returns the integer offset value for an automatic
8314 variable having address @var{x} (an RTL expression). The default
8315 computation assumes that @var{x} is based on the frame-pointer and
8316 gives the offset from the frame-pointer. This is required for targets
8317 that produce debugging output for DBX or COFF-style debugging output
8318 for SDB and allow the frame-pointer to be eliminated when the
8319 @option{-g} options is used.
8322 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
8323 A C expression that returns the integer offset value for an argument
8324 having address @var{x} (an RTL expression). The nominal offset is
8328 @defmac PREFERRED_DEBUGGING_TYPE
8329 A C expression that returns the type of debugging output GCC should
8330 produce when the user specifies just @option{-g}. Define
8331 this if you have arranged for GCC to support more than one format of
8332 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
8333 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
8334 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
8336 When the user specifies @option{-ggdb}, GCC normally also uses the
8337 value of this macro to select the debugging output format, but with two
8338 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
8339 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
8340 defined, GCC uses @code{DBX_DEBUG}.
8342 The value of this macro only affects the default debugging output; the
8343 user can always get a specific type of output by using @option{-gstabs},
8344 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
8348 @subsection Specific Options for DBX Output
8350 @c prevent bad page break with this line
8351 These are specific options for DBX output.
8353 @defmac DBX_DEBUGGING_INFO
8354 Define this macro if GCC should produce debugging output for DBX
8355 in response to the @option{-g} option.
8358 @defmac XCOFF_DEBUGGING_INFO
8359 Define this macro if GCC should produce XCOFF format debugging output
8360 in response to the @option{-g} option. This is a variant of DBX format.
8363 @defmac DEFAULT_GDB_EXTENSIONS
8364 Define this macro to control whether GCC should by default generate
8365 GDB's extended version of DBX debugging information (assuming DBX-format
8366 debugging information is enabled at all). If you don't define the
8367 macro, the default is 1: always generate the extended information
8368 if there is any occasion to.
8371 @defmac DEBUG_SYMS_TEXT
8372 Define this macro if all @code{.stabs} commands should be output while
8373 in the text section.
8376 @defmac ASM_STABS_OP
8377 A C string constant, including spacing, naming the assembler pseudo op to
8378 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
8379 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
8380 applies only to DBX debugging information format.
8383 @defmac ASM_STABD_OP
8384 A C string constant, including spacing, naming the assembler pseudo op to
8385 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
8386 value is the current location. If you don't define this macro,
8387 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
8391 @defmac ASM_STABN_OP
8392 A C string constant, including spacing, naming the assembler pseudo op to
8393 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
8394 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
8395 macro applies only to DBX debugging information format.
8398 @defmac DBX_NO_XREFS
8399 Define this macro if DBX on your system does not support the construct
8400 @samp{xs@var{tagname}}. On some systems, this construct is used to
8401 describe a forward reference to a structure named @var{tagname}.
8402 On other systems, this construct is not supported at all.
8405 @defmac DBX_CONTIN_LENGTH
8406 A symbol name in DBX-format debugging information is normally
8407 continued (split into two separate @code{.stabs} directives) when it
8408 exceeds a certain length (by default, 80 characters). On some
8409 operating systems, DBX requires this splitting; on others, splitting
8410 must not be done. You can inhibit splitting by defining this macro
8411 with the value zero. You can override the default splitting-length by
8412 defining this macro as an expression for the length you desire.
8415 @defmac DBX_CONTIN_CHAR
8416 Normally continuation is indicated by adding a @samp{\} character to
8417 the end of a @code{.stabs} string when a continuation follows. To use
8418 a different character instead, define this macro as a character
8419 constant for the character you want to use. Do not define this macro
8420 if backslash is correct for your system.
8423 @defmac DBX_STATIC_STAB_DATA_SECTION
8424 Define this macro if it is necessary to go to the data section before
8425 outputting the @samp{.stabs} pseudo-op for a non-global static
8429 @defmac DBX_TYPE_DECL_STABS_CODE
8430 The value to use in the ``code'' field of the @code{.stabs} directive
8431 for a typedef. The default is @code{N_LSYM}.
8434 @defmac DBX_STATIC_CONST_VAR_CODE
8435 The value to use in the ``code'' field of the @code{.stabs} directive
8436 for a static variable located in the text section. DBX format does not
8437 provide any ``right'' way to do this. The default is @code{N_FUN}.
8440 @defmac DBX_REGPARM_STABS_CODE
8441 The value to use in the ``code'' field of the @code{.stabs} directive
8442 for a parameter passed in registers. DBX format does not provide any
8443 ``right'' way to do this. The default is @code{N_RSYM}.
8446 @defmac DBX_REGPARM_STABS_LETTER
8447 The letter to use in DBX symbol data to identify a symbol as a parameter
8448 passed in registers. DBX format does not customarily provide any way to
8449 do this. The default is @code{'P'}.
8452 @defmac DBX_FUNCTION_FIRST
8453 Define this macro if the DBX information for a function and its
8454 arguments should precede the assembler code for the function. Normally,
8455 in DBX format, the debugging information entirely follows the assembler
8459 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
8460 Define this macro, with value 1, if the value of a symbol describing
8461 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
8462 relative to the start of the enclosing function. Normally, GCC uses
8463 an absolute address.
8466 @defmac DBX_LINES_FUNCTION_RELATIVE
8467 Define this macro, with value 1, if the value of a symbol indicating
8468 the current line number (@code{N_SLINE}) should be relative to the
8469 start of the enclosing function. Normally, GCC uses an absolute address.
8472 @defmac DBX_USE_BINCL
8473 Define this macro if GCC should generate @code{N_BINCL} and
8474 @code{N_EINCL} stabs for included header files, as on Sun systems. This
8475 macro also directs GCC to output a type number as a pair of a file
8476 number and a type number within the file. Normally, GCC does not
8477 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
8478 number for a type number.
8482 @subsection Open-Ended Hooks for DBX Format
8484 @c prevent bad page break with this line
8485 These are hooks for DBX format.
8487 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
8488 Define this macro to say how to output to @var{stream} the debugging
8489 information for the start of a scope level for variable names. The
8490 argument @var{name} is the name of an assembler symbol (for use with
8491 @code{assemble_name}) whose value is the address where the scope begins.
8494 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
8495 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
8498 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
8499 Define this macro if the target machine requires special handling to
8500 output an @code{N_FUN} entry for the function @var{decl}.
8503 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
8504 A C statement to output DBX debugging information before code for line
8505 number @var{line} of the current source file to the stdio stream
8506 @var{stream}. @var{counter} is the number of time the macro was
8507 invoked, including the current invocation; it is intended to generate
8508 unique labels in the assembly output.
8510 This macro should not be defined if the default output is correct, or
8511 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
8514 @defmac NO_DBX_FUNCTION_END
8515 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
8516 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
8517 On those machines, define this macro to turn this feature off without
8518 disturbing the rest of the gdb extensions.
8521 @defmac NO_DBX_BNSYM_ENSYM
8522 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
8523 extension construct. On those machines, define this macro to turn this
8524 feature off without disturbing the rest of the gdb extensions.
8527 @node File Names and DBX
8528 @subsection File Names in DBX Format
8530 @c prevent bad page break with this line
8531 This describes file names in DBX format.
8533 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
8534 A C statement to output DBX debugging information to the stdio stream
8535 @var{stream}, which indicates that file @var{name} is the main source
8536 file---the file specified as the input file for compilation.
8537 This macro is called only once, at the beginning of compilation.
8539 This macro need not be defined if the standard form of output
8540 for DBX debugging information is appropriate.
8542 It may be necessary to refer to a label equal to the beginning of the
8543 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
8544 to do so. If you do this, you must also set the variable
8545 @var{used_ltext_label_name} to @code{true}.
8548 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
8549 Define this macro, with value 1, if GCC should not emit an indication
8550 of the current directory for compilation and current source language at
8551 the beginning of the file.
8554 @defmac NO_DBX_GCC_MARKER
8555 Define this macro, with value 1, if GCC should not emit an indication
8556 that this object file was compiled by GCC@. The default is to emit
8557 an @code{N_OPT} stab at the beginning of every source file, with
8558 @samp{gcc2_compiled.} for the string and value 0.
8561 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
8562 A C statement to output DBX debugging information at the end of
8563 compilation of the main source file @var{name}. Output should be
8564 written to the stdio stream @var{stream}.
8566 If you don't define this macro, nothing special is output at the end
8567 of compilation, which is correct for most machines.
8570 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
8571 Define this macro @emph{instead of} defining
8572 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
8573 the end of compilation is a @code{N_SO} stab with an empty string,
8574 whose value is the highest absolute text address in the file.
8579 @subsection Macros for SDB and DWARF Output
8581 @c prevent bad page break with this line
8582 Here are macros for SDB and DWARF output.
8584 @defmac SDB_DEBUGGING_INFO
8585 Define this macro if GCC should produce COFF-style debugging output
8586 for SDB in response to the @option{-g} option.
8589 @defmac DWARF2_DEBUGGING_INFO
8590 Define this macro if GCC should produce dwarf version 2 format
8591 debugging output in response to the @option{-g} option.
8593 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (tree @var{function})
8594 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
8595 be emitted for each function. Instead of an integer return the enum
8596 value for the @code{DW_CC_} tag.
8599 To support optional call frame debugging information, you must also
8600 define @code{INCOMING_RETURN_ADDR_RTX} and either set
8601 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
8602 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
8603 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
8606 @defmac DWARF2_FRAME_INFO
8607 Define this macro to a nonzero value if GCC should always output
8608 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
8609 (@pxref{Exception Region Output} is nonzero, GCC will output this
8610 information not matter how you define @code{DWARF2_FRAME_INFO}.
8613 @defmac DWARF2_ASM_LINE_DEBUG_INFO
8614 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8615 line debug info sections. This will result in much more compact line number
8616 tables, and hence is desirable if it works.
8619 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8620 A C statement to issue assembly directives that create a difference
8621 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
8624 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
8625 A C statement to issue assembly directives that create a
8626 section-relative reference to the given @var{label}, using an integer of the
8627 given @var{size}. The label is known to be defined in the given @var{section}.
8630 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8631 A C statement to issue assembly directives that create a self-relative
8632 reference to the given @var{label}, using an integer of the given @var{size}.
8635 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{FILE}, int @var{size}, rtx @var{x})
8636 If defined, this target hook is a function which outputs a DTP-relative
8637 reference to the given TLS symbol of the specified size.
8640 @defmac PUT_SDB_@dots{}
8641 Define these macros to override the assembler syntax for the special
8642 SDB assembler directives. See @file{sdbout.c} for a list of these
8643 macros and their arguments. If the standard syntax is used, you need
8644 not define them yourself.
8648 Some assemblers do not support a semicolon as a delimiter, even between
8649 SDB assembler directives. In that case, define this macro to be the
8650 delimiter to use (usually @samp{\n}). It is not necessary to define
8651 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8655 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
8656 Define this macro to allow references to unknown structure,
8657 union, or enumeration tags to be emitted. Standard COFF does not
8658 allow handling of unknown references, MIPS ECOFF has support for
8662 @defmac SDB_ALLOW_FORWARD_REFERENCES
8663 Define this macro to allow references to structure, union, or
8664 enumeration tags that have not yet been seen to be handled. Some
8665 assemblers choke if forward tags are used, while some require it.
8668 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
8669 A C statement to output SDB debugging information before code for line
8670 number @var{line} of the current source file to the stdio stream
8671 @var{stream}. The default is to emit an @code{.ln} directive.
8676 @subsection Macros for VMS Debug Format
8678 @c prevent bad page break with this line
8679 Here are macros for VMS debug format.
8681 @defmac VMS_DEBUGGING_INFO
8682 Define this macro if GCC should produce debugging output for VMS
8683 in response to the @option{-g} option. The default behavior for VMS
8684 is to generate minimal debug info for a traceback in the absence of
8685 @option{-g} unless explicitly overridden with @option{-g0}. This
8686 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8687 @code{OVERRIDE_OPTIONS}.
8690 @node Floating Point
8691 @section Cross Compilation and Floating Point
8692 @cindex cross compilation and floating point
8693 @cindex floating point and cross compilation
8695 While all modern machines use twos-complement representation for integers,
8696 there are a variety of representations for floating point numbers. This
8697 means that in a cross-compiler the representation of floating point numbers
8698 in the compiled program may be different from that used in the machine
8699 doing the compilation.
8701 Because different representation systems may offer different amounts of
8702 range and precision, all floating point constants must be represented in
8703 the target machine's format. Therefore, the cross compiler cannot
8704 safely use the host machine's floating point arithmetic; it must emulate
8705 the target's arithmetic. To ensure consistency, GCC always uses
8706 emulation to work with floating point values, even when the host and
8707 target floating point formats are identical.
8709 The following macros are provided by @file{real.h} for the compiler to
8710 use. All parts of the compiler which generate or optimize
8711 floating-point calculations must use these macros. They may evaluate
8712 their operands more than once, so operands must not have side effects.
8714 @defmac REAL_VALUE_TYPE
8715 The C data type to be used to hold a floating point value in the target
8716 machine's format. Typically this is a @code{struct} containing an
8717 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8721 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8722 Compares for equality the two values, @var{x} and @var{y}. If the target
8723 floating point format supports negative zeroes and/or NaNs,
8724 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8725 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8728 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8729 Tests whether @var{x} is less than @var{y}.
8732 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8733 Truncates @var{x} to a signed integer, rounding toward zero.
8736 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8737 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8738 @var{x} is negative, returns zero.
8741 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8742 Converts @var{string} into a floating point number in the target machine's
8743 representation for mode @var{mode}. This routine can handle both
8744 decimal and hexadecimal floating point constants, using the syntax
8745 defined by the C language for both.
8748 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8749 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8752 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8753 Determines whether @var{x} represents infinity (positive or negative).
8756 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8757 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8760 @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})
8761 Calculates an arithmetic operation on the two floating point values
8762 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8765 The operation to be performed is specified by @var{code}. Only the
8766 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8767 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8769 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8770 target's floating point format cannot represent infinity, it will call
8771 @code{abort}. Callers should check for this situation first, using
8772 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8775 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8776 Returns the negative of the floating point value @var{x}.
8779 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8780 Returns the absolute value of @var{x}.
8783 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8784 Truncates the floating point value @var{x} to fit in @var{mode}. The
8785 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8786 appropriate bit pattern to be output as a floating constant whose
8787 precision accords with mode @var{mode}.
8790 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8791 Converts a floating point value @var{x} into a double-precision integer
8792 which is then stored into @var{low} and @var{high}. If the value is not
8793 integral, it is truncated.
8796 @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})
8797 Converts a double-precision integer found in @var{low} and @var{high},
8798 into a floating point value which is then stored into @var{x}. The
8799 value is truncated to fit in mode @var{mode}.
8802 @node Mode Switching
8803 @section Mode Switching Instructions
8804 @cindex mode switching
8805 The following macros control mode switching optimizations:
8807 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
8808 Define this macro if the port needs extra instructions inserted for mode
8809 switching in an optimizing compilation.
8811 For an example, the SH4 can perform both single and double precision
8812 floating point operations, but to perform a single precision operation,
8813 the FPSCR PR bit has to be cleared, while for a double precision
8814 operation, this bit has to be set. Changing the PR bit requires a general
8815 purpose register as a scratch register, hence these FPSCR sets have to
8816 be inserted before reload, i.e.@: you can't put this into instruction emitting
8817 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
8819 You can have multiple entities that are mode-switched, and select at run time
8820 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
8821 return nonzero for any @var{entity} that needs mode-switching.
8822 If you define this macro, you also have to define
8823 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
8824 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
8825 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
8829 @defmac NUM_MODES_FOR_MODE_SWITCHING
8830 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
8831 initializer for an array of integers. Each initializer element
8832 N refers to an entity that needs mode switching, and specifies the number
8833 of different modes that might need to be set for this entity.
8834 The position of the initializer in the initializer---starting counting at
8835 zero---determines the integer that is used to refer to the mode-switched
8837 In macros that take mode arguments / yield a mode result, modes are
8838 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
8839 switch is needed / supplied.
8842 @defmac MODE_NEEDED (@var{entity}, @var{insn})
8843 @var{entity} is an integer specifying a mode-switched entity. If
8844 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
8845 return an integer value not larger than the corresponding element in
8846 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
8847 be switched into prior to the execution of @var{insn}.
8850 @defmac MODE_AFTER (@var{mode}, @var{insn})
8851 If this macro is defined, it is evaluated for every @var{insn} during
8852 mode switching. It determines the mode that an insn results in (if
8853 different from the incoming mode).
8856 @defmac MODE_ENTRY (@var{entity})
8857 If this macro is defined, it is evaluated for every @var{entity} that needs
8858 mode switching. It should evaluate to an integer, which is a mode that
8859 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
8860 is defined then @code{MODE_EXIT} must be defined.
8863 @defmac MODE_EXIT (@var{entity})
8864 If this macro is defined, it is evaluated for every @var{entity} that needs
8865 mode switching. It should evaluate to an integer, which is a mode that
8866 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
8867 is defined then @code{MODE_ENTRY} must be defined.
8870 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
8871 This macro specifies the order in which modes for @var{entity} are processed.
8872 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
8873 lowest. The value of the macro should be an integer designating a mode
8874 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
8875 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
8876 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
8879 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
8880 Generate one or more insns to set @var{entity} to @var{mode}.
8881 @var{hard_reg_live} is the set of hard registers live at the point where
8882 the insn(s) are to be inserted.
8885 @node Target Attributes
8886 @section Defining target-specific uses of @code{__attribute__}
8887 @cindex target attributes
8888 @cindex machine attributes
8889 @cindex attributes, target-specific
8891 Target-specific attributes may be defined for functions, data and types.
8892 These are described using the following target hooks; they also need to
8893 be documented in @file{extend.texi}.
8895 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
8896 If defined, this target hook points to an array of @samp{struct
8897 attribute_spec} (defined in @file{tree.h}) specifying the machine
8898 specific attributes for this target and some of the restrictions on the
8899 entities to which these attributes are applied and the arguments they
8903 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8904 If defined, this target hook is a function which returns zero if the attributes on
8905 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8906 and two if they are nearly compatible (which causes a warning to be
8907 generated). If this is not defined, machine-specific attributes are
8908 supposed always to be compatible.
8911 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
8912 If defined, this target hook is a function which assigns default attributes to
8913 newly defined @var{type}.
8916 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8917 Define this target hook if the merging of type attributes needs special
8918 handling. If defined, the result is a list of the combined
8919 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
8920 that @code{comptypes} has already been called and returned 1. This
8921 function may call @code{merge_attributes} to handle machine-independent
8925 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
8926 Define this target hook if the merging of decl attributes needs special
8927 handling. If defined, the result is a list of the combined
8928 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
8929 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
8930 when this is needed are when one attribute overrides another, or when an
8931 attribute is nullified by a subsequent definition. This function may
8932 call @code{merge_attributes} to handle machine-independent merging.
8934 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
8935 If the only target-specific handling you require is @samp{dllimport}
8936 for Microsoft Windows targets, you should define the macro
8937 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
8938 will then define a function called
8939 @code{merge_dllimport_decl_attributes} which can then be defined as
8940 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
8941 add @code{handle_dll_attribute} in the attribute table for your port
8942 to perform initial processing of the @samp{dllimport} and
8943 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
8944 @file{i386/i386.c}, for example.
8947 @deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (tree @var{decl})
8948 @var{decl} is a variable or function with @code{__attribute__((dllimport))}
8949 specified. Use this hook if the target needs to add extra validation
8950 checks to @code{handle_dll_attribute}.
8953 @defmac TARGET_DECLSPEC
8954 Define this macro to a nonzero value if you want to treat
8955 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
8956 default, this behavior is enabled only for targets that define
8957 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
8958 of @code{__declspec} is via a built-in macro, but you should not rely
8959 on this implementation detail.
8962 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
8963 Define this target hook if you want to be able to add attributes to a decl
8964 when it is being created. This is normally useful for back ends which
8965 wish to implement a pragma by using the attributes which correspond to
8966 the pragma's effect. The @var{node} argument is the decl which is being
8967 created. The @var{attr_ptr} argument is a pointer to the attribute list
8968 for this decl. The list itself should not be modified, since it may be
8969 shared with other decls, but attributes may be chained on the head of
8970 the list and @code{*@var{attr_ptr}} modified to point to the new
8971 attributes, or a copy of the list may be made if further changes are
8975 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
8977 This target hook returns @code{true} if it is ok to inline @var{fndecl}
8978 into the current function, despite its having target-specific
8979 attributes, @code{false} otherwise. By default, if a function has a
8980 target specific attribute attached to it, it will not be inlined.
8983 @node MIPS Coprocessors
8984 @section Defining coprocessor specifics for MIPS targets.
8985 @cindex MIPS coprocessor-definition macros
8987 The MIPS specification allows MIPS implementations to have as many as 4
8988 coprocessors, each with as many as 32 private registers. GCC supports
8989 accessing these registers and transferring values between the registers
8990 and memory using asm-ized variables. For example:
8993 register unsigned int cp0count asm ("c0r1");
8999 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
9000 names may be added as described below, or the default names may be
9001 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
9003 Coprocessor registers are assumed to be epilogue-used; sets to them will
9004 be preserved even if it does not appear that the register is used again
9005 later in the function.
9007 Another note: according to the MIPS spec, coprocessor 1 (if present) is
9008 the FPU@. One accesses COP1 registers through standard mips
9009 floating-point support; they are not included in this mechanism.
9011 There is one macro used in defining the MIPS coprocessor interface which
9012 you may want to override in subtargets; it is described below.
9014 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
9015 A comma-separated list (with leading comma) of pairs describing the
9016 alternate names of coprocessor registers. The format of each entry should be
9018 @{ @var{alternatename}, @var{register_number}@}
9024 @section Parameters for Precompiled Header Validity Checking
9025 @cindex parameters, precompiled headers
9027 @deftypefn {Target Hook} void *TARGET_GET_PCH_VALIDITY (size_t *@var{sz})
9028 This hook returns the data needed by @code{TARGET_PCH_VALID_P} and sets
9029 @samp{*@var{sz}} to the size of the data in bytes.
9032 @deftypefn {Target Hook} const char *TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz})
9033 This hook checks whether the options used to create a PCH file are
9034 compatible with the current settings. It returns @code{NULL}
9035 if so and a suitable error message if not. Error messages will
9036 be presented to the user and must be localized using @samp{_(@var{msg})}.
9038 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
9039 when the PCH file was created and @var{sz} is the size of that data in bytes.
9040 It's safe to assume that the data was created by the same version of the
9041 compiler, so no format checking is needed.
9043 The default definition of @code{default_pch_valid_p} should be
9044 suitable for most targets.
9047 @deftypefn {Target Hook} const char *TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags})
9048 If this hook is nonnull, the default implementation of
9049 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
9050 of @code{target_flags}. @var{pch_flags} specifies the value that
9051 @code{target_flags} had when the PCH file was created. The return
9052 value is the same as for @code{TARGET_PCH_VALID_P}.
9056 @section C++ ABI parameters
9057 @cindex parameters, c++ abi
9059 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
9060 Define this hook to override the integer type used for guard variables.
9061 These are used to implement one-time construction of static objects. The
9062 default is long_long_integer_type_node.
9065 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
9066 This hook determines how guard variables are used. It should return
9067 @code{false} (the default) if first byte should be used. A return value of
9068 @code{true} indicates the least significant bit should be used.
9071 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
9072 This hook returns the size of the cookie to use when allocating an array
9073 whose elements have the indicated @var{type}. Assumes that it is already
9074 known that a cookie is needed. The default is
9075 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
9076 IA64/Generic C++ ABI@.
9079 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
9080 This hook should return @code{true} if the element size should be stored in
9081 array cookies. The default is to return @code{false}.
9084 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
9085 If defined by a backend this hook allows the decision made to export
9086 class @var{type} to be overruled. Upon entry @var{import_export}
9087 will contain 1 if the class is going to be exported, @minus{}1 if it is going
9088 to be imported and 0 otherwise. This function should return the
9089 modified value and perform any other actions necessary to support the
9090 backend's targeted operating system.
9093 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
9094 This hook should return @code{true} if constructors and destructors return
9095 the address of the object created/destroyed. The default is to return
9099 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
9100 This hook returns true if the key method for a class (i.e., the method
9101 which, if defined in the current translation unit, causes the virtual
9102 table to be emitted) may be an inline function. Under the standard
9103 Itanium C++ ABI the key method may be an inline function so long as
9104 the function is not declared inline in the class definition. Under
9105 some variants of the ABI, an inline function can never be the key
9106 method. The default is to return @code{true}.
9109 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
9110 @var{decl} is a virtual table, virtual table table, typeinfo object,
9111 or other similar implicit class data object that will be emitted with
9112 external linkage in this translation unit. No ELF visibility has been
9113 explicitly specified. If the target needs to specify a visibility
9114 other than that of the containing class, use this hook to set
9115 @code{DECL_VISIBILITY} and @code{DECL_VISIBILITY_SPECIFIED}.
9118 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
9119 This hook returns true (the default) if virtual tables and other
9120 similar implicit class data objects are always COMDAT if they have
9121 external linkage. If this hook returns false, then class data for
9122 classes whose virtual table will be emitted in only one translation
9123 unit will not be COMDAT.
9126 @deftypefn {Target Hook} bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void)
9127 This hook returns true (the default) if the RTTI information for
9128 the basic types which is defined in the C++ runtime should always
9129 be COMDAT, false if it should not be COMDAT.
9132 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
9133 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
9134 should be used to register static destructors when @option{-fuse-cxa-atexit}
9135 is in effect. The default is to return false to use @code{__cxa_atexit}.
9138 @deftypefn {Target Hook} bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void)
9139 This hook returns true if the target @code{atexit} function can be used
9140 in the same manner as @code{__cxa_atexit} to register C++ static
9141 destructors. This requires that @code{atexit}-registered functions in
9142 shared libraries are run in the correct order when the libraries are
9143 unloaded. The default is to return false.
9146 @deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type})
9147 @var{type} is a C++ class (i.e., RECORD_TYPE or UNION_TYPE) that has just been
9148 defined. Use this hook to make adjustments to the class (eg, tweak
9149 visibility or perform any other required target modifications).
9153 @section Miscellaneous Parameters
9154 @cindex parameters, miscellaneous
9156 @c prevent bad page break with this line
9157 Here are several miscellaneous parameters.
9159 @defmac HAS_LONG_COND_BRANCH
9160 Define this boolean macro to indicate whether or not your architecture
9161 has conditional branches that can span all of memory. It is used in
9162 conjunction with an optimization that partitions hot and cold basic
9163 blocks into separate sections of the executable. If this macro is
9164 set to false, gcc will convert any conditional branches that attempt
9165 to cross between sections into unconditional branches or indirect jumps.
9168 @defmac HAS_LONG_UNCOND_BRANCH
9169 Define this boolean macro to indicate whether or not your architecture
9170 has unconditional branches that can span all of memory. It is used in
9171 conjunction with an optimization that partitions hot and cold basic
9172 blocks into separate sections of the executable. If this macro is
9173 set to false, gcc will convert any unconditional branches that attempt
9174 to cross between sections into indirect jumps.
9177 @defmac CASE_VECTOR_MODE
9178 An alias for a machine mode name. This is the machine mode that
9179 elements of a jump-table should have.
9182 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
9183 Optional: return the preferred mode for an @code{addr_diff_vec}
9184 when the minimum and maximum offset are known. If you define this,
9185 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
9186 To make this work, you also have to define @code{INSN_ALIGN} and
9187 make the alignment for @code{addr_diff_vec} explicit.
9188 The @var{body} argument is provided so that the offset_unsigned and scale
9189 flags can be updated.
9192 @defmac CASE_VECTOR_PC_RELATIVE
9193 Define this macro to be a C expression to indicate when jump-tables
9194 should contain relative addresses. You need not define this macro if
9195 jump-tables never contain relative addresses, or jump-tables should
9196 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
9200 @defmac CASE_VALUES_THRESHOLD
9201 Define this to be the smallest number of different values for which it
9202 is best to use a jump-table instead of a tree of conditional branches.
9203 The default is four for machines with a @code{casesi} instruction and
9204 five otherwise. This is best for most machines.
9207 @defmac CASE_USE_BIT_TESTS
9208 Define this macro to be a C expression to indicate whether C switch
9209 statements may be implemented by a sequence of bit tests. This is
9210 advantageous on processors that can efficiently implement left shift
9211 of 1 by the number of bits held in a register, but inappropriate on
9212 targets that would require a loop. By default, this macro returns
9213 @code{true} if the target defines an @code{ashlsi3} pattern, and
9214 @code{false} otherwise.
9217 @defmac WORD_REGISTER_OPERATIONS
9218 Define this macro if operations between registers with integral mode
9219 smaller than a word are always performed on the entire register.
9220 Most RISC machines have this property and most CISC machines do not.
9223 @defmac LOAD_EXTEND_OP (@var{mem_mode})
9224 Define this macro to be a C expression indicating when insns that read
9225 memory in @var{mem_mode}, an integral mode narrower than a word, set the
9226 bits outside of @var{mem_mode} to be either the sign-extension or the
9227 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
9228 of @var{mem_mode} for which the
9229 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
9230 @code{UNKNOWN} for other modes.
9232 This macro is not called with @var{mem_mode} non-integral or with a width
9233 greater than or equal to @code{BITS_PER_WORD}, so you may return any
9234 value in this case. Do not define this macro if it would always return
9235 @code{UNKNOWN}. On machines where this macro is defined, you will normally
9236 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
9238 You may return a non-@code{UNKNOWN} value even if for some hard registers
9239 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
9240 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
9241 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
9242 integral mode larger than this but not larger than @code{word_mode}.
9244 You must return @code{UNKNOWN} if for some hard registers that allow this
9245 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
9246 @code{word_mode}, but that they can change to another integral mode that
9247 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
9250 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
9251 Define this macro if loading short immediate values into registers sign
9255 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
9256 Define this macro if the same instructions that convert a floating
9257 point number to a signed fixed point number also convert validly to an
9261 @deftypefn {Target Hook} int TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (enum machine_mode @var{mode})
9262 When @option{-ffast-math} is in effect, GCC tries to optimize
9263 divisions by the same divisor, by turning them into multiplications by
9264 the reciprocal. This target hook specifies the minimum number of divisions
9265 that should be there for GCC to perform the optimization for a variable
9266 of mode @var{mode}. The default implementation returns 3 if the machine
9267 has an instruction for the division, and 2 if it does not.
9271 The maximum number of bytes that a single instruction can move quickly
9272 between memory and registers or between two memory locations.
9275 @defmac MAX_MOVE_MAX
9276 The maximum number of bytes that a single instruction can move quickly
9277 between memory and registers or between two memory locations. If this
9278 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
9279 constant value that is the largest value that @code{MOVE_MAX} can have
9283 @defmac SHIFT_COUNT_TRUNCATED
9284 A C expression that is nonzero if on this machine the number of bits
9285 actually used for the count of a shift operation is equal to the number
9286 of bits needed to represent the size of the object being shifted. When
9287 this macro is nonzero, the compiler will assume that it is safe to omit
9288 a sign-extend, zero-extend, and certain bitwise `and' instructions that
9289 truncates the count of a shift operation. On machines that have
9290 instructions that act on bit-fields at variable positions, which may
9291 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
9292 also enables deletion of truncations of the values that serve as
9293 arguments to bit-field instructions.
9295 If both types of instructions truncate the count (for shifts) and
9296 position (for bit-field operations), or if no variable-position bit-field
9297 instructions exist, you should define this macro.
9299 However, on some machines, such as the 80386 and the 680x0, truncation
9300 only applies to shift operations and not the (real or pretended)
9301 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
9302 such machines. Instead, add patterns to the @file{md} file that include
9303 the implied truncation of the shift instructions.
9305 You need not define this macro if it would always have the value of zero.
9308 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
9309 @deftypefn {Target Hook} int TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode})
9310 This function describes how the standard shift patterns for @var{mode}
9311 deal with shifts by negative amounts or by more than the width of the mode.
9312 @xref{shift patterns}.
9314 On many machines, the shift patterns will apply a mask @var{m} to the
9315 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
9316 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
9317 this is true for mode @var{mode}, the function should return @var{m},
9318 otherwise it should return 0. A return value of 0 indicates that no
9319 particular behavior is guaranteed.
9321 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
9322 @emph{not} apply to general shift rtxes; it applies only to instructions
9323 that are generated by the named shift patterns.
9325 The default implementation of this function returns
9326 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
9327 and 0 otherwise. This definition is always safe, but if
9328 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
9329 nevertheless truncate the shift count, you may get better code
9333 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
9334 A C expression which is nonzero if on this machine it is safe to
9335 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
9336 bits (where @var{outprec} is smaller than @var{inprec}) by merely
9337 operating on it as if it had only @var{outprec} bits.
9339 On many machines, this expression can be 1.
9341 @c rearranged this, removed the phrase "it is reported that". this was
9342 @c to fix an overfull hbox. --mew 10feb93
9343 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
9344 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
9345 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
9346 such cases may improve things.
9349 @deftypefn {Target Hook} int TARGET_MODE_REP_EXTENDED (enum machine_mode @var{mode}, enum machine_mode @var{rep_mode})
9350 The representation of an integral mode can be such that the values
9351 are always extended to a wider integral mode. Return
9352 @code{SIGN_EXTEND} if values of @var{mode} are represented in
9353 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
9354 otherwise. (Currently, none of the targets use zero-extended
9355 representation this way so unlike @code{LOAD_EXTEND_OP},
9356 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
9357 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
9358 @var{mode} to @var{mode_rep} so that @var{mode_rep} is not the next
9359 widest integral mode and currently we take advantage of this fact.)
9361 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
9362 value even if the extension is not performed on certain hard registers
9363 as long as for the @code{REGNO_REG_CLASS} of these hard registers
9364 @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero.
9366 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
9367 describe two related properties. If you define
9368 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
9369 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
9372 In order to enforce the representation of @code{mode},
9373 @code{TRULY_NOOP_TRUNCATION} should return false when truncating to
9377 @defmac STORE_FLAG_VALUE
9378 A C expression describing the value returned by a comparison operator
9379 with an integral mode and stored by a store-flag instruction
9380 (@samp{s@var{cond}}) when the condition is true. This description must
9381 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
9382 comparison operators whose results have a @code{MODE_INT} mode.
9384 A value of 1 or @minus{}1 means that the instruction implementing the
9385 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
9386 and 0 when the comparison is false. Otherwise, the value indicates
9387 which bits of the result are guaranteed to be 1 when the comparison is
9388 true. This value is interpreted in the mode of the comparison
9389 operation, which is given by the mode of the first operand in the
9390 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
9391 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
9394 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
9395 generate code that depends only on the specified bits. It can also
9396 replace comparison operators with equivalent operations if they cause
9397 the required bits to be set, even if the remaining bits are undefined.
9398 For example, on a machine whose comparison operators return an
9399 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
9400 @samp{0x80000000}, saying that just the sign bit is relevant, the
9404 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
9411 (ashift:SI @var{x} (const_int @var{n}))
9415 where @var{n} is the appropriate shift count to move the bit being
9416 tested into the sign bit.
9418 There is no way to describe a machine that always sets the low-order bit
9419 for a true value, but does not guarantee the value of any other bits,
9420 but we do not know of any machine that has such an instruction. If you
9421 are trying to port GCC to such a machine, include an instruction to
9422 perform a logical-and of the result with 1 in the pattern for the
9423 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
9425 Often, a machine will have multiple instructions that obtain a value
9426 from a comparison (or the condition codes). Here are rules to guide the
9427 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
9432 Use the shortest sequence that yields a valid definition for
9433 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
9434 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
9435 comparison operators to do so because there may be opportunities to
9436 combine the normalization with other operations.
9439 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
9440 slightly preferred on machines with expensive jumps and 1 preferred on
9444 As a second choice, choose a value of @samp{0x80000001} if instructions
9445 exist that set both the sign and low-order bits but do not define the
9449 Otherwise, use a value of @samp{0x80000000}.
9452 Many machines can produce both the value chosen for
9453 @code{STORE_FLAG_VALUE} and its negation in the same number of
9454 instructions. On those machines, you should also define a pattern for
9455 those cases, e.g., one matching
9458 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
9461 Some machines can also perform @code{and} or @code{plus} operations on
9462 condition code values with less instructions than the corresponding
9463 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
9464 machines, define the appropriate patterns. Use the names @code{incscc}
9465 and @code{decscc}, respectively, for the patterns which perform
9466 @code{plus} or @code{minus} operations on condition code values. See
9467 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
9468 find such instruction sequences on other machines.
9470 If this macro is not defined, the default value, 1, is used. You need
9471 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
9472 instructions, or if the value generated by these instructions is 1.
9475 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
9476 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
9477 returned when comparison operators with floating-point results are true.
9478 Define this macro on machines that have comparison operations that return
9479 floating-point values. If there are no such operations, do not define
9483 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
9484 A C expression that gives a rtx representing the nonzero true element
9485 for vector comparisons. The returned rtx should be valid for the inner
9486 mode of @var{mode} which is guaranteed to be a vector mode. Define
9487 this macro on machines that have vector comparison operations that
9488 return a vector result. If there are no such operations, do not define
9489 this macro. Typically, this macro is defined as @code{const1_rtx} or
9490 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
9491 the compiler optimizing such vector comparison operations for the
9495 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
9496 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
9497 A C expression that evaluates to true if the architecture defines a value
9498 for @code{clz} or @code{ctz} with a zero operand. If so, @var{value}
9499 should be set to this value. If this macro is not defined, the value of
9500 @code{clz} or @code{ctz} is assumed to be undefined.
9502 This macro must be defined if the target's expansion for @code{ffs}
9503 relies on a particular value to get correct results. Otherwise it
9504 is not necessary, though it may be used to optimize some corner cases.
9506 Note that regardless of this macro the ``definedness'' of @code{clz}
9507 and @code{ctz} at zero do @emph{not} extend to the builtin functions
9508 visible to the user. Thus one may be free to adjust the value at will
9509 to match the target expansion of these operations without fear of
9514 An alias for the machine mode for pointers. On most machines, define
9515 this to be the integer mode corresponding to the width of a hardware
9516 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
9517 On some machines you must define this to be one of the partial integer
9518 modes, such as @code{PSImode}.
9520 The width of @code{Pmode} must be at least as large as the value of
9521 @code{POINTER_SIZE}. If it is not equal, you must define the macro
9522 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
9526 @defmac FUNCTION_MODE
9527 An alias for the machine mode used for memory references to functions
9528 being called, in @code{call} RTL expressions. On most machines this
9529 should be @code{QImode}.
9532 @defmac STDC_0_IN_SYSTEM_HEADERS
9533 In normal operation, the preprocessor expands @code{__STDC__} to the
9534 constant 1, to signify that GCC conforms to ISO Standard C@. On some
9535 hosts, like Solaris, the system compiler uses a different convention,
9536 where @code{__STDC__} is normally 0, but is 1 if the user specifies
9537 strict conformance to the C Standard.
9539 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
9540 convention when processing system header files, but when processing user
9541 files @code{__STDC__} will always expand to 1.
9544 @defmac NO_IMPLICIT_EXTERN_C
9545 Define this macro if the system header files support C++ as well as C@.
9546 This macro inhibits the usual method of using system header files in
9547 C++, which is to pretend that the file's contents are enclosed in
9548 @samp{extern "C" @{@dots{}@}}.
9553 @defmac REGISTER_TARGET_PRAGMAS ()
9554 Define this macro if you want to implement any target-specific pragmas.
9555 If defined, it is a C expression which makes a series of calls to
9556 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
9557 for each pragma. The macro may also do any
9558 setup required for the pragmas.
9560 The primary reason to define this macro is to provide compatibility with
9561 other compilers for the same target. In general, we discourage
9562 definition of target-specific pragmas for GCC@.
9564 If the pragma can be implemented by attributes then you should consider
9565 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
9567 Preprocessor macros that appear on pragma lines are not expanded. All
9568 @samp{#pragma} directives that do not match any registered pragma are
9569 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
9572 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9573 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9575 Each call to @code{c_register_pragma} or
9576 @code{c_register_pragma_with_expansion} establishes one pragma. The
9577 @var{callback} routine will be called when the preprocessor encounters a
9581 #pragma [@var{space}] @var{name} @dots{}
9584 @var{space} is the case-sensitive namespace of the pragma, or
9585 @code{NULL} to put the pragma in the global namespace. The callback
9586 routine receives @var{pfile} as its first argument, which can be passed
9587 on to cpplib's functions if necessary. You can lex tokens after the
9588 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
9589 callback will be silently ignored. The end of the line is indicated by
9590 a token of type @code{CPP_EOF}. Macro expansion occurs on the
9591 arguments of pragmas registered with
9592 @code{c_register_pragma_with_expansion} but not on the arguments of
9593 pragmas registered with @code{c_register_pragma}.
9595 For an example use of this routine, see @file{c4x.h} and the callback
9596 routines defined in @file{c4x-c.c}.
9598 Note that the use of @code{pragma_lex} is specific to the C and C++
9599 compilers. It will not work in the Java or Fortran compilers, or any
9600 other language compilers for that matter. Thus if @code{pragma_lex} is going
9601 to be called from target-specific code, it must only be done so when
9602 building the C and C++ compilers. This can be done by defining the
9603 variables @code{c_target_objs} and @code{cxx_target_objs} in the
9604 target entry in the @file{config.gcc} file. These variables should name
9605 the target-specific, language-specific object file which contains the
9606 code that uses @code{pragma_lex}. Note it will also be necessary to add a
9607 rule to the makefile fragment pointed to by @code{tmake_file} that shows
9608 how to build this object file.
9613 @defmac HANDLE_SYSV_PRAGMA
9614 Define this macro (to a value of 1) if you want the System V style
9615 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
9616 [=<value>]} to be supported by gcc.
9618 The pack pragma specifies the maximum alignment (in bytes) of fields
9619 within a structure, in much the same way as the @samp{__aligned__} and
9620 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
9621 the behavior to the default.
9623 A subtlety for Microsoft Visual C/C++ style bit-field packing
9624 (e.g.@: -mms-bitfields) for targets that support it:
9625 When a bit-field is inserted into a packed record, the whole size
9626 of the underlying type is used by one or more same-size adjacent
9627 bit-fields (that is, if its long:3, 32 bits is used in the record,
9628 and any additional adjacent long bit-fields are packed into the same
9629 chunk of 32 bits. However, if the size changes, a new field of that
9632 If both MS bit-fields and @samp{__attribute__((packed))} are used,
9633 the latter will take precedence. If @samp{__attribute__((packed))} is
9634 used on a single field when MS bit-fields are in use, it will take
9635 precedence for that field, but the alignment of the rest of the structure
9636 may affect its placement.
9638 The weak pragma only works if @code{SUPPORTS_WEAK} and
9639 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
9640 of specifically named weak labels, optionally with a value.
9645 @defmac HANDLE_PRAGMA_PACK_PUSH_POP
9646 Define this macro (to a value of 1) if you want to support the Win32
9647 style pragmas @samp{#pragma pack(push[,@var{n}])} and @samp{#pragma
9648 pack(pop)}. The @samp{pack(push,[@var{n}])} pragma specifies the maximum
9649 alignment (in bytes) of fields within a structure, in much the same way as
9650 the @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
9651 pack value of zero resets the behavior to the default. Successive
9652 invocations of this pragma cause the previous values to be stacked, so
9653 that invocations of @samp{#pragma pack(pop)} will return to the previous
9657 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
9658 Define this macro, as well as
9659 @code{HANDLE_SYSV_PRAGMA}, if macros should be expanded in the
9660 arguments of @samp{#pragma pack}.
9663 @defmac TARGET_DEFAULT_PACK_STRUCT
9664 If your target requires a structure packing default other than 0 (meaning
9665 the machine default), define this macro to the necessary value (in bytes).
9666 This must be a value that would also be valid to use with
9667 @samp{#pragma pack()} (that is, a small power of two).
9670 @defmac DOLLARS_IN_IDENTIFIERS
9671 Define this macro to control use of the character @samp{$} in
9672 identifier names for the C family of languages. 0 means @samp{$} is
9673 not allowed by default; 1 means it is allowed. 1 is the default;
9674 there is no need to define this macro in that case.
9677 @defmac NO_DOLLAR_IN_LABEL
9678 Define this macro if the assembler does not accept the character
9679 @samp{$} in label names. By default constructors and destructors in
9680 G++ have @samp{$} in the identifiers. If this macro is defined,
9681 @samp{.} is used instead.
9684 @defmac NO_DOT_IN_LABEL
9685 Define this macro if the assembler does not accept the character
9686 @samp{.} in label names. By default constructors and destructors in G++
9687 have names that use @samp{.}. If this macro is defined, these names
9688 are rewritten to avoid @samp{.}.
9691 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
9692 Define this macro as a C expression that is nonzero if it is safe for the
9693 delay slot scheduler to place instructions in the delay slot of @var{insn},
9694 even if they appear to use a resource set or clobbered in @var{insn}.
9695 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
9696 every @code{call_insn} has this behavior. On machines where some @code{insn}
9697 or @code{jump_insn} is really a function call and hence has this behavior,
9698 you should define this macro.
9700 You need not define this macro if it would always return zero.
9703 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
9704 Define this macro as a C expression that is nonzero if it is safe for the
9705 delay slot scheduler to place instructions in the delay slot of @var{insn},
9706 even if they appear to set or clobber a resource referenced in @var{insn}.
9707 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
9708 some @code{insn} or @code{jump_insn} is really a function call and its operands
9709 are registers whose use is actually in the subroutine it calls, you should
9710 define this macro. Doing so allows the delay slot scheduler to move
9711 instructions which copy arguments into the argument registers into the delay
9714 You need not define this macro if it would always return zero.
9717 @defmac MULTIPLE_SYMBOL_SPACES
9718 Define this macro as a C expression that is nonzero if, in some cases,
9719 global symbols from one translation unit may not be bound to undefined
9720 symbols in another translation unit without user intervention. For
9721 instance, under Microsoft Windows symbols must be explicitly imported
9722 from shared libraries (DLLs).
9724 You need not define this macro if it would always evaluate to zero.
9727 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{outputs}, tree @var{inputs}, tree @var{clobbers})
9728 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
9729 any hard regs the port wishes to automatically clobber for an asm.
9730 It should return the result of the last @code{tree_cons} used to add a
9731 clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the
9732 corresponding parameters to the asm and may be inspected to avoid
9733 clobbering a register that is an input or output of the asm. You can use
9734 @code{tree_overlaps_hard_reg_set}, declared in @file{tree.h}, to test
9735 for overlap with regards to asm-declared registers.
9738 @defmac MATH_LIBRARY
9739 Define this macro as a C string constant for the linker argument to link
9740 in the system math library, or @samp{""} if the target does not have a
9741 separate math library.
9743 You need only define this macro if the default of @samp{"-lm"} is wrong.
9746 @defmac LIBRARY_PATH_ENV
9747 Define this macro as a C string constant for the environment variable that
9748 specifies where the linker should look for libraries.
9750 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
9754 @defmac TARGET_POSIX_IO
9755 Define this macro if the target supports the following POSIX@ file
9756 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
9757 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
9758 to use file locking when exiting a program, which avoids race conditions
9759 if the program has forked. It will also create directories at run-time
9760 for cross-profiling.
9763 @defmac MAX_CONDITIONAL_EXECUTE
9765 A C expression for the maximum number of instructions to execute via
9766 conditional execution instructions instead of a branch. A value of
9767 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
9768 1 if it does use cc0.
9771 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
9772 Used if the target needs to perform machine-dependent modifications on the
9773 conditionals used for turning basic blocks into conditionally executed code.
9774 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
9775 contains information about the currently processed blocks. @var{true_expr}
9776 and @var{false_expr} are the tests that are used for converting the
9777 then-block and the else-block, respectively. Set either @var{true_expr} or
9778 @var{false_expr} to a null pointer if the tests cannot be converted.
9781 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
9782 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
9783 if-statements into conditions combined by @code{and} and @code{or} operations.
9784 @var{bb} contains the basic block that contains the test that is currently
9785 being processed and about to be turned into a condition.
9788 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
9789 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
9790 be converted to conditional execution format. @var{ce_info} points to
9791 a data structure, @code{struct ce_if_block}, which contains information
9792 about the currently processed blocks.
9795 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
9796 A C expression to perform any final machine dependent modifications in
9797 converting code to conditional execution. The involved basic blocks
9798 can be found in the @code{struct ce_if_block} structure that is pointed
9799 to by @var{ce_info}.
9802 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
9803 A C expression to cancel any machine dependent modifications in
9804 converting code to conditional execution. The involved basic blocks
9805 can be found in the @code{struct ce_if_block} structure that is pointed
9806 to by @var{ce_info}.
9809 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
9810 A C expression to initialize any extra fields in a @code{struct ce_if_block}
9811 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
9814 @defmac IFCVT_EXTRA_FIELDS
9815 If defined, it should expand to a set of field declarations that will be
9816 added to the @code{struct ce_if_block} structure. These should be initialized
9817 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
9820 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG ()
9821 If non-null, this hook performs a target-specific pass over the
9822 instruction stream. The compiler will run it at all optimization levels,
9823 just before the point at which it normally does delayed-branch scheduling.
9825 The exact purpose of the hook varies from target to target. Some use
9826 it to do transformations that are necessary for correctness, such as
9827 laying out in-function constant pools or avoiding hardware hazards.
9828 Others use it as an opportunity to do some machine-dependent optimizations.
9830 You need not implement the hook if it has nothing to do. The default
9834 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
9835 Define this hook if you have any machine-specific built-in functions
9836 that need to be defined. It should be a function that performs the
9839 Machine specific built-in functions can be useful to expand special machine
9840 instructions that would otherwise not normally be generated because
9841 they have no equivalent in the source language (for example, SIMD vector
9842 instructions or prefetch instructions).
9844 To create a built-in function, call the function
9845 @code{lang_hooks.builtin_function}
9846 which is defined by the language front end. You can use any type nodes set
9847 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
9848 only language front ends that use those two functions will call
9849 @samp{TARGET_INIT_BUILTINS}.
9852 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
9854 Expand a call to a machine specific built-in function that was set up by
9855 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
9856 function call; the result should go to @var{target} if that is
9857 convenient, and have mode @var{mode} if that is convenient.
9858 @var{subtarget} may be used as the target for computing one of
9859 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
9860 ignored. This function should return the result of the call to the
9864 @deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (tree @var{fndecl}, tree @var{arglist})
9866 Select a replacement for a machine specific built-in function that
9867 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
9868 @emph{before} regular type checking, and so allows the target to
9869 implement a crude form of function overloading. @var{fndecl} is the
9870 declaration of the built-in function. @var{arglist} is the list of
9871 arguments passed to the built-in function. The result is a
9872 complete expression that implements the operation, usually
9873 another @code{CALL_EXPR}.
9876 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, tree @var{arglist}, bool @var{ignore})
9878 Fold a call to a machine specific built-in function that was set up by
9879 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
9880 built-in function. @var{arglist} is the list of arguments passed to
9881 the built-in function. The result is another tree containing a
9882 simplified expression for the call's result. If @var{ignore} is true
9883 the value will be ignored.
9886 @deftypefn {Target Hook} const char * TARGET_INVALID_WITHIN_DOLOOP (rtx @var{insn})
9888 Take an instruction in @var{insn} and return NULL if it is valid within a
9889 low-overhead loop, otherwise return a string why doloop could not be applied.
9891 Many targets use special registers for low-overhead looping. For any
9892 instruction that clobbers these this function should return a string indicating
9893 the reason why the doloop could not be applied.
9894 By default, the RTL loop optimizer does not use a present doloop pattern for
9895 loops containing function calls or branch on table instructions.
9898 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
9900 Take a branch insn in @var{branch1} and another in @var{branch2}.
9901 Return true if redirecting @var{branch1} to the destination of
9902 @var{branch2} is possible.
9904 On some targets, branches may have a limited range. Optimizing the
9905 filling of delay slots can result in branches being redirected, and this
9906 may in turn cause a branch offset to overflow.
9909 @deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (rtx @var{x}, @var{outer_code})
9910 This target hook returns @code{true} if @var{x} is considered to be commutative.
9911 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
9912 PLUS to be commutative inside a MEM. @var{outer_code} is the rtx code
9913 of the enclosing rtl, if known, otherwise it is UNKNOWN.
9916 @deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg})
9918 When the initial value of a hard register has been copied in a pseudo
9919 register, it is often not necessary to actually allocate another register
9920 to this pseudo register, because the original hard register or a stack slot
9921 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
9922 is called at the start of register allocation once for each hard register
9923 that had its initial value copied by using
9924 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
9925 Possible values are @code{NULL_RTX}, if you don't want
9926 to do any special allocation, a @code{REG} rtx---that would typically be
9927 the hard register itself, if it is known not to be clobbered---or a
9929 If you are returning a @code{MEM}, this is only a hint for the allocator;
9930 it might decide to use another register anyways.
9931 You may use @code{current_function_leaf_function} in the hook, functions
9932 that use @code{REG_N_SETS}, to determine if the hard
9933 register in question will not be clobbered.
9934 The default value of this hook is @code{NULL}, which disables any special
9938 @defmac TARGET_OBJECT_SUFFIX
9939 Define this macro to be a C string representing the suffix for object
9940 files on your target machine. If you do not define this macro, GCC will
9941 use @samp{.o} as the suffix for object files.
9944 @defmac TARGET_EXECUTABLE_SUFFIX
9945 Define this macro to be a C string representing the suffix to be
9946 automatically added to executable files on your target machine. If you
9947 do not define this macro, GCC will use the null string as the suffix for
9951 @defmac COLLECT_EXPORT_LIST
9952 If defined, @code{collect2} will scan the individual object files
9953 specified on its command line and create an export list for the linker.
9954 Define this macro for systems like AIX, where the linker discards
9955 object files that are not referenced from @code{main} and uses export
9959 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
9960 Define this macro to a C expression representing a variant of the
9961 method call @var{mdecl}, if Java Native Interface (JNI) methods
9962 must be invoked differently from other methods on your target.
9963 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
9964 the @code{stdcall} calling convention and this macro is then
9965 defined as this expression:
9968 build_type_attribute_variant (@var{mdecl},
9970 (get_identifier ("stdcall"),
9975 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
9976 This target hook returns @code{true} past the point in which new jump
9977 instructions could be created. On machines that require a register for
9978 every jump such as the SHmedia ISA of SH5, this point would typically be
9979 reload, so this target hook should be defined to a function such as:
9983 cannot_modify_jumps_past_reload_p ()
9985 return (reload_completed || reload_in_progress);
9990 @deftypefn {Target Hook} int TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
9991 This target hook returns a register class for which branch target register
9992 optimizations should be applied. All registers in this class should be
9993 usable interchangeably. After reload, registers in this class will be
9994 re-allocated and loads will be hoisted out of loops and be subjected
9995 to inter-block scheduling.
9998 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
9999 Branch target register optimization will by default exclude callee-saved
10001 that are not already live during the current function; if this target hook
10002 returns true, they will be included. The target code must than make sure
10003 that all target registers in the class returned by
10004 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
10005 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
10006 epilogues have already been generated. Note, even if you only return
10007 true when @var{after_prologue_epilogue_gen} is false, you still are likely
10008 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
10009 to reserve space for caller-saved target registers.
10012 @defmac POWI_MAX_MULTS
10013 If defined, this macro is interpreted as a signed integer C expression
10014 that specifies the maximum number of floating point multiplications
10015 that should be emitted when expanding exponentiation by an integer
10016 constant inline. When this value is defined, exponentiation requiring
10017 more than this number of multiplications is implemented by calling the
10018 system library's @code{pow}, @code{powf} or @code{powl} routines.
10019 The default value places no upper bound on the multiplication count.
10022 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
10023 This target hook should register any extra include files for the
10024 target. The parameter @var{stdinc} indicates if normal include files
10025 are present. The parameter @var{sysroot} is the system root directory.
10026 The parameter @var{iprefix} is the prefix for the gcc directory.
10029 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
10030 This target hook should register any extra include files for the
10031 target before any standard headers. The parameter @var{stdinc}
10032 indicates if normal include files are present. The parameter
10033 @var{sysroot} is the system root directory. The parameter
10034 @var{iprefix} is the prefix for the gcc directory.
10037 @deftypefn Macro void TARGET_OPTF (char *@var{path})
10038 This target hook should register special include paths for the target.
10039 The parameter @var{path} is the include to register. On Darwin
10040 systems, this is used for Framework includes, which have semantics
10041 that are different from @option{-I}.
10044 @deftypefn {Target Hook} bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
10045 This target hook returns @code{true} if it is safe to use a local alias
10046 for a virtual function @var{fndecl} when constructing thunks,
10047 @code{false} otherwise. By default, the hook returns @code{true} for all
10048 functions, if a target supports aliases (i.e.@: defines
10049 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
10052 @defmac TARGET_FORMAT_TYPES
10053 If defined, this macro is the name of a global variable containing
10054 target-specific format checking information for the @option{-Wformat}
10055 option. The default is to have no target-specific format checks.
10058 @defmac TARGET_N_FORMAT_TYPES
10059 If defined, this macro is the number of entries in
10060 @code{TARGET_FORMAT_TYPES}.
10063 @deftypefn {Target Hook} bool TARGET_RELAXED_ORDERING
10064 If set to @code{true}, means that the target's memory model does not
10065 guarantee that loads which do not depend on one another will access
10066 main memory in the order of the instruction stream; if ordering is
10067 important, an explicit memory barrier must be used. This is true of
10068 many recent processors which implement a policy of ``relaxed,''
10069 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
10070 and ia64. The default is @code{false}.
10073 @deftypefn {Target Hook} const char *TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (tree @var{typelist}, tree @var{funcdecl}, tree @var{val})
10074 If defined, this macro returns the diagnostic message when it is
10075 illegal to pass argument @var{val} to function @var{funcdecl}
10076 with prototype @var{typelist}.
10079 @deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (tree @var{fromtype}, tree @var{totype})
10080 If defined, this macro returns the diagnostic message when it is
10081 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
10082 if validity should be determined by the front end.
10085 @deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, tree @var{type})
10086 If defined, this macro returns the diagnostic message when it is
10087 invalid to apply operation @var{op} (where unary plus is denoted by
10088 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
10089 if validity should be determined by the front end.
10092 @deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, tree @var{type1}, tree @var{type2})
10093 If defined, this macro returns the diagnostic message when it is
10094 invalid to apply operation @var{op} to operands of types @var{type1}
10095 and @var{type2}, or @code{NULL} if validity should be determined by
10099 @defmac TARGET_USE_JCR_SECTION
10100 This macro determines whether to use the JCR section to register Java
10101 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
10102 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.
10106 This macro determines the size of the objective C jump buffer for the
10107 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.