1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,
2 @c 2002, 2003, 2004 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 * Escape Sequences:: Defining the value of target character escape sequences
35 * Registers:: Naming and describing the hardware registers.
36 * Register Classes:: Defining the classes of hardware registers.
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 * Condition Code:: Defining how insns update the condition code.
43 * Costs:: Defining relative costs of different operations.
44 * Scheduling:: Adjusting the behavior of the instruction scheduler.
45 * Sections:: Dividing storage into text, data, and other sections.
46 * PIC:: Macros for position independent code.
47 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
48 * Debugging Info:: Defining the format of debugging output.
49 * Floating Point:: Handling floating point for cross-compilers.
50 * Mode Switching:: Insertion of mode-switching instructions.
51 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
52 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
53 * PCH Target:: Validity checking for precompiled headers.
54 * C++ ABI:: Controlling C++ ABI changes.
55 * Misc:: Everything else.
58 @node Target Structure
59 @section The Global @code{targetm} Variable
61 @cindex target functions
63 @deftypevar {struct gcc_target} targetm
64 The target @file{.c} file must define the global @code{targetm} variable
65 which contains pointers to functions and data relating to the target
66 machine. The variable is declared in @file{target.h};
67 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
68 used to initialize the variable, and macros for the default initializers
69 for elements of the structure. The @file{.c} file should override those
70 macros for which the default definition is inappropriate. For example:
73 #include "target-def.h"
75 /* @r{Initialize the GCC target structure.} */
77 #undef TARGET_COMP_TYPE_ATTRIBUTES
78 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
80 struct gcc_target targetm = TARGET_INITIALIZER;
84 Where a macro should be defined in the @file{.c} file in this manner to
85 form part of the @code{targetm} structure, it is documented below as a
86 ``Target Hook'' with a prototype. Many macros will change in future
87 from being defined in the @file{.h} file to being part of the
88 @code{targetm} structure.
91 @section Controlling the Compilation Driver, @file{gcc}
93 @cindex controlling the compilation driver
95 @c prevent bad page break with this line
96 You can control the compilation driver.
98 @defmac SWITCH_TAKES_ARG (@var{char})
99 A C expression which determines whether the option @option{-@var{char}}
100 takes arguments. The value should be the number of arguments that
101 option takes--zero, for many options.
103 By default, this macro is defined as
104 @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options
105 properly. You need not define @code{SWITCH_TAKES_ARG} unless you
106 wish to add additional options which take arguments. Any redefinition
107 should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for
111 @defmac WORD_SWITCH_TAKES_ARG (@var{name})
112 A C expression which determines whether the option @option{-@var{name}}
113 takes arguments. The value should be the number of arguments that
114 option takes--zero, for many options. This macro rather than
115 @code{SWITCH_TAKES_ARG} is used for multi-character option names.
117 By default, this macro is defined as
118 @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
119 properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
120 wish to add additional options which take arguments. Any redefinition
121 should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
125 @defmac SWITCH_CURTAILS_COMPILATION (@var{char})
126 A C expression which determines whether the option @option{-@var{char}}
127 stops compilation before the generation of an executable. The value is
128 boolean, nonzero if the option does stop an executable from being
129 generated, zero otherwise.
131 By default, this macro is defined as
132 @code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard
133 options properly. You need not define
134 @code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional
135 options which affect the generation of an executable. Any redefinition
136 should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check
137 for additional options.
140 @defmac SWITCHES_NEED_SPACES
141 A string-valued C expression which enumerates the options for which
142 the linker needs a space between the option and its argument.
144 If this macro is not defined, the default value is @code{""}.
147 @defmac TARGET_OPTION_TRANSLATE_TABLE
148 If defined, a list of pairs of strings, the first of which is a
149 potential command line target to the @file{gcc} driver program, and the
150 second of which is a space-separated (tabs and other whitespace are not
151 supported) list of options with which to replace the first option. The
152 target defining this list is responsible for assuring that the results
153 are valid. Replacement options may not be the @code{--opt} style, they
154 must be the @code{-opt} style. It is the intention of this macro to
155 provide a mechanism for substitution that affects the multilibs chosen,
156 such as one option that enables many options, some of which select
157 multilibs. Example nonsensical definition, where @option{-malt-abi},
158 @option{-EB}, and @option{-mspoo} cause different multilibs to be chosen:
161 #define TARGET_OPTION_TRANSLATE_TABLE \
162 @{ "-fast", "-march=fast-foo -malt-abi -I/usr/fast-foo" @}, \
163 @{ "-compat", "-EB -malign=4 -mspoo" @}
167 @defmac DRIVER_SELF_SPECS
168 A list of specs for the driver itself. It should be a suitable
169 initializer for an array of strings, with no surrounding braces.
171 The driver applies these specs to its own command line between loading
172 default @file{specs} files (but not command-line specified ones) and
173 choosing the multilib directory or running any subcommands. It
174 applies them in the order given, so each spec can depend on the
175 options added by earlier ones. It is also possible to remove options
176 using @samp{%<@var{option}} in the usual way.
178 This macro can be useful when a port has several interdependent target
179 options. It provides a way of standardizing the command line so
180 that the other specs are easier to write.
182 Do not define this macro if it does not need to do anything.
185 @defmac OPTION_DEFAULT_SPECS
186 A list of specs used to support configure-time default options (i.e.@:
187 @option{--with} options) in the driver. It should be a suitable initializer
188 for an array of structures, each containing two strings, without the
189 outermost pair of surrounding braces.
191 The first item in the pair is the name of the default. This must match
192 the code in @file{config.gcc} for the target. The second item is a spec
193 to apply if a default with this name was specified. The string
194 @samp{%(VALUE)} in the spec will be replaced by the value of the default
195 everywhere it occurs.
197 The driver will apply these specs to its own command line between loading
198 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
199 the same mechanism as @code{DRIVER_SELF_SPECS}.
201 Do not define this macro if it does not need to do anything.
205 A C string constant that tells the GCC driver program options to
206 pass to CPP@. It can also specify how to translate options you
207 give to GCC into options for GCC to pass to the CPP@.
209 Do not define this macro if it does not need to do anything.
212 @defmac CPLUSPLUS_CPP_SPEC
213 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
214 than C@. If you do not define this macro, then the value of
215 @code{CPP_SPEC} (if any) will be used instead.
219 A C string constant that tells the GCC driver program options to
220 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
222 It can also specify how to translate options you give to GCC into options
223 for GCC to pass to front ends.
225 Do not define this macro if it does not need to do anything.
229 A C string constant that tells the GCC driver program options to
230 pass to @code{cc1plus}. It can also specify how to translate options you
231 give to GCC into options for GCC to pass to the @code{cc1plus}.
233 Do not define this macro if it does not need to do anything.
234 Note that everything defined in CC1_SPEC is already passed to
235 @code{cc1plus} so there is no need to duplicate the contents of
236 CC1_SPEC in CC1PLUS_SPEC@.
240 A C string constant that tells the GCC driver program options to
241 pass to the assembler. It can also specify how to translate options
242 you give to GCC into options for GCC to pass to the assembler.
243 See the file @file{sun3.h} for an example of this.
245 Do not define this macro if it does not need to do anything.
248 @defmac ASM_FINAL_SPEC
249 A C string constant that tells the GCC driver program how to
250 run any programs which cleanup after the normal assembler.
251 Normally, this is not needed. See the file @file{mips.h} for
254 Do not define this macro if it does not need to do anything.
257 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
258 Define this macro, with no value, if the driver should give the assembler
259 an argument consisting of a single dash, @option{-}, to instruct it to
260 read from its standard input (which will be a pipe connected to the
261 output of the compiler proper). This argument is given after any
262 @option{-o} option specifying the name of the output file.
264 If you do not define this macro, the assembler is assumed to read its
265 standard input if given no non-option arguments. If your assembler
266 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
267 see @file{mips.h} for instance.
271 A C string constant that tells the GCC driver program options to
272 pass to the linker. It can also specify how to translate options you
273 give to GCC into options for GCC to pass to the linker.
275 Do not define this macro if it does not need to do anything.
279 Another C string constant used much like @code{LINK_SPEC}. The difference
280 between the two is that @code{LIB_SPEC} is used at the end of the
281 command given to the linker.
283 If this macro is not defined, a default is provided that
284 loads the standard C library from the usual place. See @file{gcc.c}.
288 Another C string constant that tells the GCC driver program
289 how and when to place a reference to @file{libgcc.a} into the
290 linker command line. This constant is placed both before and after
291 the value of @code{LIB_SPEC}.
293 If this macro is not defined, the GCC driver provides a default that
294 passes the string @option{-lgcc} to the linker.
297 @defmac REAL_LIBGCC_SPEC
298 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
299 @code{LIBGCC_SPEC} is not directly used by the driver program but is
300 instead modified to refer to different versions of @file{libgcc.a}
301 depending on the values of the command line flags @option{-static},
302 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
303 targets where these modifications are inappropriate, define
304 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
305 driver how to place a reference to @file{libgcc} on the link command
306 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
309 @defmac USE_LD_AS_NEEDED
310 A macro that controls the modifications to @code{LIBGCC_SPEC}
311 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
312 generated that uses --as-needed and the shared libgcc in place of the
313 static exception handler library, when linking without any of
314 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
318 If defined, this C string constant is added to @code{LINK_SPEC}.
319 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
320 the modifications to @code{LIBGCC_SPEC} mentioned in
321 @code{REAL_LIBGCC_SPEC}.
324 @defmac STARTFILE_SPEC
325 Another C string constant used much like @code{LINK_SPEC}. The
326 difference between the two is that @code{STARTFILE_SPEC} is used at
327 the very beginning of the command given to the linker.
329 If this macro is not defined, a default is provided that loads the
330 standard C startup file from the usual place. See @file{gcc.c}.
334 Another C string constant used much like @code{LINK_SPEC}. The
335 difference between the two is that @code{ENDFILE_SPEC} is used at
336 the very end of the command given to the linker.
338 Do not define this macro if it does not need to do anything.
341 @defmac THREAD_MODEL_SPEC
342 GCC @code{-v} will print the thread model GCC was configured to use.
343 However, this doesn't work on platforms that are multilibbed on thread
344 models, such as AIX 4.3. On such platforms, define
345 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
346 blanks that names one of the recognized thread models. @code{%*}, the
347 default value of this macro, will expand to the value of
348 @code{thread_file} set in @file{config.gcc}.
351 @defmac SYSROOT_SUFFIX_SPEC
352 Define this macro to add a suffix to the target sysroot when GCC is
353 configured with a sysroot. This will cause GCC to search for usr/lib,
354 et al, within sysroot+suffix.
357 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
358 Define this macro to add a headers_suffix to the target sysroot when
359 GCC is configured with a sysroot. This will cause GCC to pass the
360 updated sysroot+headers_suffix to CPP, causing it to search for
361 usr/include, et al, within sysroot+headers_suffix.
365 Define this macro to provide additional specifications to put in the
366 @file{specs} file that can be used in various specifications like
369 The definition should be an initializer for an array of structures,
370 containing a string constant, that defines the specification name, and a
371 string constant that provides the specification.
373 Do not define this macro if it does not need to do anything.
375 @code{EXTRA_SPECS} is useful when an architecture contains several
376 related targets, which have various @code{@dots{}_SPECS} which are similar
377 to each other, and the maintainer would like one central place to keep
380 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
381 define either @code{_CALL_SYSV} when the System V calling sequence is
382 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
385 The @file{config/rs6000/rs6000.h} target file defines:
388 #define EXTRA_SPECS \
389 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
391 #define CPP_SYS_DEFAULT ""
394 The @file{config/rs6000/sysv.h} target file defines:
398 "%@{posix: -D_POSIX_SOURCE @} \
399 %@{mcall-sysv: -D_CALL_SYSV @} \
400 %@{!mcall-sysv: %(cpp_sysv_default) @} \
401 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
403 #undef CPP_SYSV_DEFAULT
404 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
407 while the @file{config/rs6000/eabiaix.h} target file defines
408 @code{CPP_SYSV_DEFAULT} as:
411 #undef CPP_SYSV_DEFAULT
412 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
416 @defmac LINK_LIBGCC_SPECIAL
417 Define this macro if the driver program should find the library
418 @file{libgcc.a} itself and should not pass @option{-L} options to the
419 linker. 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 and will
421 pass @option{-L} options to it.
424 @defmac LINK_LIBGCC_SPECIAL_1
425 Define this macro if the driver program should find the library
426 @file{libgcc.a}. If you do not define this macro, the driver program will pass
427 the argument @option{-lgcc} to tell the linker to do the search.
428 This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does
429 not affect @option{-L} options.
432 @defmac LINK_GCC_C_SEQUENCE_SPEC
433 The sequence in which libgcc and libc are specified to the linker.
434 By default this is @code{%G %L %G}.
437 @defmac LINK_COMMAND_SPEC
438 A C string constant giving the complete command line need to execute the
439 linker. When you do this, you will need to update your port each time a
440 change is made to the link command line within @file{gcc.c}. Therefore,
441 define this macro only if you need to completely redefine the command
442 line for invoking the linker and there is no other way to accomplish
443 the effect you need. Overriding this macro may be avoidable by overriding
444 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
447 @defmac LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
448 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
449 directories from linking commands. Do not give it a nonzero value if
450 removing duplicate search directories changes the linker's semantics.
453 @defmac MULTILIB_DEFAULTS
454 Define this macro as a C expression for the initializer of an array of
455 string to tell the driver program which options are defaults for this
456 target and thus do not need to be handled specially when using
457 @code{MULTILIB_OPTIONS}.
459 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
460 the target makefile fragment or if none of the options listed in
461 @code{MULTILIB_OPTIONS} are set by default.
462 @xref{Target Fragment}.
465 @defmac RELATIVE_PREFIX_NOT_LINKDIR
466 Define this macro to tell @command{gcc} that it should only translate
467 a @option{-B} prefix into a @option{-L} linker option if the prefix
468 indicates an absolute file name.
471 @defmac MD_EXEC_PREFIX
472 If defined, this macro is an additional prefix to try after
473 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
474 when the @option{-b} option is used, or the compiler is built as a cross
475 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
476 to the list of directories used to find the assembler in @file{configure.in}.
479 @defmac STANDARD_STARTFILE_PREFIX
480 Define this macro as a C string constant if you wish to override the
481 standard choice of @code{libdir} as the default prefix to
482 try when searching for startup files such as @file{crt0.o}.
483 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
484 is built as a cross compiler.
487 @defmac STANDARD_STARTFILE_PREFIX_1
488 Define this macro as a C string constant if you wish to override the
489 standard choice of @code{/lib} as a prefix to try after the default prefix
490 when searching for startup files such as @file{crt0.o}.
491 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
492 is built as a cross compiler.
495 @defmac STANDARD_STARTFILE_PREFIX_2
496 Define this macro as a C string constant if you wish to override the
497 standard choice of @code{/lib} as yet another prefix to try after the
498 default prefix when searching for startup files such as @file{crt0.o}.
499 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
500 is built as a cross compiler.
503 @defmac MD_STARTFILE_PREFIX
504 If defined, this macro supplies an additional prefix to try after the
505 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
506 @option{-b} option is used, or when the compiler is built as a cross
510 @defmac MD_STARTFILE_PREFIX_1
511 If defined, this macro supplies yet another prefix to try after the
512 standard prefixes. It is not searched when the @option{-b} option is
513 used, or when the compiler is built as a cross compiler.
516 @defmac INIT_ENVIRONMENT
517 Define this macro as a C string constant if you wish to set environment
518 variables for programs called by the driver, such as the assembler and
519 loader. The driver passes the value of this macro to @code{putenv} to
520 initialize the necessary environment variables.
523 @defmac LOCAL_INCLUDE_DIR
524 Define this macro as a C string constant if you wish to override the
525 standard choice of @file{/usr/local/include} as the default prefix to
526 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
527 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
529 Cross compilers do not search either @file{/usr/local/include} or its
533 @defmac MODIFY_TARGET_NAME
534 Define this macro if you wish to define command-line switches that
535 modify the default target name.
537 For each switch, you can include a string to be appended to the first
538 part of the configuration name or a string to be deleted from the
539 configuration name, if present. The definition should be an initializer
540 for an array of structures. Each array element should have three
541 elements: the switch name (a string constant, including the initial
542 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
543 indicate whether the string should be inserted or deleted, and the string
544 to be inserted or deleted (a string constant).
546 For example, on a machine where @samp{64} at the end of the
547 configuration name denotes a 64-bit target and you want the @option{-32}
548 and @option{-64} switches to select between 32- and 64-bit targets, you would
552 #define MODIFY_TARGET_NAME \
553 @{ @{ "-32", DELETE, "64"@}, \
554 @{"-64", ADD, "64"@}@}
558 @defmac SYSTEM_INCLUDE_DIR
559 Define this macro as a C string constant if you wish to specify a
560 system-specific directory to search for header files before the standard
561 directory. @code{SYSTEM_INCLUDE_DIR} comes before
562 @code{STANDARD_INCLUDE_DIR} in the search order.
564 Cross compilers do not use this macro and do not search the directory
568 @defmac STANDARD_INCLUDE_DIR
569 Define this macro as a C string constant if you wish to override the
570 standard choice of @file{/usr/include} as the default prefix to
571 try when searching for header files.
573 Cross compilers ignore this macro and do not search either
574 @file{/usr/include} or its replacement.
577 @defmac STANDARD_INCLUDE_COMPONENT
578 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
579 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
580 If you do not define this macro, no component is used.
583 @defmac INCLUDE_DEFAULTS
584 Define this macro if you wish to override the entire default search path
585 for include files. For a native compiler, the default search path
586 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
587 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
588 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
589 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
590 and specify private search areas for GCC@. The directory
591 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
593 The definition should be an initializer for an array of structures.
594 Each array element should have four elements: the directory name (a
595 string constant), the component name (also a string constant), a flag
596 for C++-only directories,
597 and a flag showing that the includes in the directory don't need to be
598 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
599 the array with a null element.
601 The component name denotes what GNU package the include file is part of,
602 if any, in all uppercase letters. For example, it might be @samp{GCC}
603 or @samp{BINUTILS}. If the package is part of a vendor-supplied
604 operating system, code the component name as @samp{0}.
606 For example, here is the definition used for VAX/VMS:
609 #define INCLUDE_DEFAULTS \
611 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
612 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
613 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
620 Here is the order of prefixes tried for exec files:
624 Any prefixes specified by the user with @option{-B}.
627 The environment variable @code{GCC_EXEC_PREFIX}, if any.
630 The directories specified by the environment variable @code{COMPILER_PATH}.
633 The macro @code{STANDARD_EXEC_PREFIX}.
636 @file{/usr/lib/gcc/}.
639 The macro @code{MD_EXEC_PREFIX}, if any.
642 Here is the order of prefixes tried for startfiles:
646 Any prefixes specified by the user with @option{-B}.
649 The environment variable @code{GCC_EXEC_PREFIX}, if any.
652 The directories specified by the environment variable @code{LIBRARY_PATH}
653 (or port-specific name; native only, cross compilers do not use this).
656 The macro @code{STANDARD_EXEC_PREFIX}.
659 @file{/usr/lib/gcc/}.
662 The macro @code{MD_EXEC_PREFIX}, if any.
665 The macro @code{MD_STARTFILE_PREFIX}, if any.
668 The macro @code{STANDARD_STARTFILE_PREFIX}.
677 @node Run-time Target
678 @section Run-time Target Specification
679 @cindex run-time target specification
680 @cindex predefined macros
681 @cindex target specifications
683 @c prevent bad page break with this line
684 Here are run-time target specifications.
686 @defmac TARGET_CPU_CPP_BUILTINS ()
687 This function-like macro expands to a block of code that defines
688 built-in preprocessor macros and assertions for the target cpu, using
689 the functions @code{builtin_define}, @code{builtin_define_std} and
690 @code{builtin_assert}. When the front end
691 calls this macro it provides a trailing semicolon, and since it has
692 finished command line option processing your code can use those
695 @code{builtin_assert} takes a string in the form you pass to the
696 command-line option @option{-A}, such as @code{cpu=mips}, and creates
697 the assertion. @code{builtin_define} takes a string in the form
698 accepted by option @option{-D} and unconditionally defines the macro.
700 @code{builtin_define_std} takes a string representing the name of an
701 object-like macro. If it doesn't lie in the user's namespace,
702 @code{builtin_define_std} defines it unconditionally. Otherwise, it
703 defines a version with two leading underscores, and another version
704 with two leading and trailing underscores, and defines the original
705 only if an ISO standard was not requested on the command line. For
706 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
707 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
708 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
709 defines only @code{_ABI64}.
711 You can also test for the C dialect being compiled. The variable
712 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
713 or @code{clk_objective_c}. Note that if we are preprocessing
714 assembler, this variable will be @code{clk_c} but the function-like
715 macro @code{preprocessing_asm_p()} will return true, so you might want
716 to check for that first. If you need to check for strict ANSI, the
717 variable @code{flag_iso} can be used. The function-like macro
718 @code{preprocessing_trad_p()} can be used to check for traditional
722 @defmac TARGET_OS_CPP_BUILTINS ()
723 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
724 and is used for the target operating system instead.
727 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
728 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
729 and is used for the target object format. @file{elfos.h} uses this
730 macro to define @code{__ELF__}, so you probably do not need to define
734 @deftypevar {extern int} target_flags
735 This declaration should be present.
738 @cindex optional hardware or system features
739 @cindex features, optional, in system conventions
741 @defmac TARGET_@var{featurename}
742 This series of macros is to allow compiler command arguments to
743 enable or disable the use of optional features of the target machine.
744 For example, one machine description serves both the 68000 and
745 the 68020; a command argument tells the compiler whether it should
746 use 68020-only instructions or not. This command argument works
747 by means of a macro @code{TARGET_68020} that tests a bit in
750 Define a macro @code{TARGET_@var{featurename}} for each such option.
751 Its definition should test a bit in @code{target_flags}. It is
752 recommended that a helper macro @code{MASK_@var{featurename}}
753 is defined for each bit-value to test, and used in
754 @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For
758 #define TARGET_MASK_68020 1
759 #define TARGET_68020 (target_flags & MASK_68020)
762 One place where these macros are used is in the condition-expressions
763 of instruction patterns. Note how @code{TARGET_68020} appears
764 frequently in the 68000 machine description file, @file{m68k.md}.
765 Another place they are used is in the definitions of the other
766 macros in the @file{@var{machine}.h} file.
769 @defmac TARGET_SWITCHES
770 This macro defines names of command options to set and clear
771 bits in @code{target_flags}. Its definition is an initializer
772 with a subgrouping for each command option.
774 Each subgrouping contains a string constant, that defines the option
775 name, a number, which contains the bits to set in
776 @code{target_flags}, and a second string which is the description
777 displayed by @option{--help}. If the number is negative then the bits specified
778 by the number are cleared instead of being set. If the description
779 string is present but empty, then no help information will be displayed
780 for that option, but it will not count as an undocumented option. The
781 actual option name is made by appending @samp{-m} to the specified name.
782 Non-empty description strings should be marked with @code{N_(@dots{})} for
783 @command{xgettext}. Please do not mark empty strings because the empty
784 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
785 of the message catalog with meta information, not the empty string.
787 In addition to the description for @option{--help},
788 more detailed documentation for each option should be added to
791 One of the subgroupings should have a null string. The number in
792 this grouping is the default value for @code{target_flags}. Any
793 target options act starting with that value.
795 Here is an example which defines @option{-m68000} and @option{-m68020}
796 with opposite meanings, and picks the latter as the default:
799 #define TARGET_SWITCHES \
800 @{ @{ "68020", MASK_68020, "" @}, \
801 @{ "68000", -MASK_68020, \
802 N_("Compile for the 68000") @}, \
803 @{ "", MASK_68020, "" @}, \
808 @defmac TARGET_OPTIONS
809 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
810 options that have values. Its definition is an initializer with a
811 subgrouping for each command option.
813 Each subgrouping contains a string constant, that defines the option
814 name, the address of a variable, a description string, and a value.
815 Non-empty description strings should be marked with @code{N_(@dots{})}
816 for @command{xgettext}. Please do not mark empty strings because the
817 empty string is reserved by GNU gettext. @code{gettext("")} returns the
818 header entry of the message catalog with meta information, not the empty
821 If the value listed in the table is @code{NULL}, then the variable, type
822 @code{char *}, is set to the variable part of the given option if the
823 fixed part matches. In other words, if the first part of the option
824 matches what's in the table, the variable will be set to point to the
825 rest of the option. This allows the user to specify a value for that
826 option. The actual option name is made by appending @samp{-m} to the
827 specified name. Again, each option should also be documented in
830 If the value listed in the table is non-@code{NULL}, then the option
831 must match the option in the table exactly (with @samp{-m}), and the
832 variable is set to point to the value listed in the table.
834 Here is an example which defines @option{-mshort-data-@var{number}}. If the
835 given option is @option{-mshort-data-512}, the variable @code{m88k_short_data}
836 will be set to the string @code{"512"}.
839 extern char *m88k_short_data;
840 #define TARGET_OPTIONS \
841 @{ @{ "short-data-", &m88k_short_data, \
842 N_("Specify the size of the short data section"), 0 @} @}
845 Here is a variant of the above that allows the user to also specify
846 just @option{-mshort-data} where a default of @code{"64"} is used.
849 extern char *m88k_short_data;
850 #define TARGET_OPTIONS \
851 @{ @{ "short-data-", &m88k_short_data, \
852 N_("Specify the size of the short data section"), 0 @} \
853 @{ "short-data", &m88k_short_data, "", "64" @},
857 Here is an example which defines @option{-mno-alu}, @option{-malu1}, and
858 @option{-malu2} as a three-state switch, along with suitable macros for
859 checking the state of the option (documentation is elided for brevity).
863 char *chip_alu = ""; /* Specify default here. */
866 extern char *chip_alu;
867 #define TARGET_OPTIONS \
868 @{ @{ "no-alu", &chip_alu, "", "" @}, \
869 @{ "alu1", &chip_alu, "", "1" @}, \
870 @{ "alu2", &chip_alu, "", "2" @}, @}
871 #define TARGET_ALU (chip_alu[0] != '\0')
872 #define TARGET_ALU1 (chip_alu[0] == '1')
873 #define TARGET_ALU2 (chip_alu[0] == '2')
877 @defmac TARGET_VERSION
878 This macro is a C statement to print on @code{stderr} a string
879 describing the particular machine description choice. Every machine
880 description should define @code{TARGET_VERSION}. For example:
884 #define TARGET_VERSION \
885 fprintf (stderr, " (68k, Motorola syntax)");
887 #define TARGET_VERSION \
888 fprintf (stderr, " (68k, MIT syntax)");
893 @defmac OVERRIDE_OPTIONS
894 Sometimes certain combinations of command options do not make sense on
895 a particular target machine. You can define a macro
896 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
897 defined, is executed once just after all the command options have been
900 Don't use this macro to turn on various extra optimizations for
901 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
904 @defmac OPTIMIZATION_OPTIONS (@var{level}, @var{size})
905 Some machines may desire to change what optimizations are performed for
906 various optimization levels. This macro, if defined, is executed once
907 just after the optimization level is determined and before the remainder
908 of the command options have been parsed. Values set in this macro are
909 used as the default values for the other command line options.
911 @var{level} is the optimization level specified; 2 if @option{-O2} is
912 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
914 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
916 You should not use this macro to change options that are not
917 machine-specific. These should uniformly selected by the same
918 optimization level on all supported machines. Use this macro to enable
919 machine-specific optimizations.
921 @strong{Do not examine @code{write_symbols} in
922 this macro!} The debugging options are not supposed to alter the
926 @defmac CAN_DEBUG_WITHOUT_FP
927 Define this macro if debugging can be performed even without a frame
928 pointer. If this macro is defined, GCC will turn on the
929 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
932 @node Per-Function Data
933 @section Defining data structures for per-function information.
934 @cindex per-function data
935 @cindex data structures
937 If the target needs to store information on a per-function basis, GCC
938 provides a macro and a couple of variables to allow this. Note, just
939 using statics to store the information is a bad idea, since GCC supports
940 nested functions, so you can be halfway through encoding one function
941 when another one comes along.
943 GCC defines a data structure called @code{struct function} which
944 contains all of the data specific to an individual function. This
945 structure contains a field called @code{machine} whose type is
946 @code{struct machine_function *}, which can be used by targets to point
947 to their own specific data.
949 If a target needs per-function specific data it should define the type
950 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
951 This macro should be used to initialize the function pointer
952 @code{init_machine_status}. This pointer is explained below.
954 One typical use of per-function, target specific data is to create an
955 RTX to hold the register containing the function's return address. This
956 RTX can then be used to implement the @code{__builtin_return_address}
957 function, for level 0.
959 Note---earlier implementations of GCC used a single data area to hold
960 all of the per-function information. Thus when processing of a nested
961 function began the old per-function data had to be pushed onto a
962 stack, and when the processing was finished, it had to be popped off the
963 stack. GCC used to provide function pointers called
964 @code{save_machine_status} and @code{restore_machine_status} to handle
965 the saving and restoring of the target specific information. Since the
966 single data area approach is no longer used, these pointers are no
969 @defmac INIT_EXPANDERS
970 Macro called to initialize any target specific information. This macro
971 is called once per function, before generation of any RTL has begun.
972 The intention of this macro is to allow the initialization of the
973 function pointer @code{init_machine_status}.
976 @deftypevar {void (*)(struct function *)} init_machine_status
977 If this function pointer is non-@code{NULL} it will be called once per
978 function, before function compilation starts, in order to allow the
979 target to perform any target specific initialization of the
980 @code{struct function} structure. It is intended that this would be
981 used to initialize the @code{machine} of that structure.
983 @code{struct machine_function} structures are expected to be freed by GC@.
984 Generally, any memory that they reference must be allocated by using
985 @code{ggc_alloc}, including the structure itself.
989 @section Storage Layout
990 @cindex storage layout
992 Note that the definitions of the macros in this table which are sizes or
993 alignments measured in bits do not need to be constant. They can be C
994 expressions that refer to static variables, such as the @code{target_flags}.
995 @xref{Run-time Target}.
997 @defmac BITS_BIG_ENDIAN
998 Define this macro to have the value 1 if the most significant bit in a
999 byte has the lowest number; otherwise define it to have the value zero.
1000 This means that bit-field instructions count from the most significant
1001 bit. If the machine has no bit-field instructions, then this must still
1002 be defined, but it doesn't matter which value it is defined to. This
1003 macro need not be a constant.
1005 This macro does not affect the way structure fields are packed into
1006 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
1009 @defmac BYTES_BIG_ENDIAN
1010 Define this macro to have the value 1 if the most significant byte in a
1011 word has the lowest number. This macro need not be a constant.
1014 @defmac WORDS_BIG_ENDIAN
1015 Define this macro to have the value 1 if, in a multiword object, the
1016 most significant word has the lowest number. This applies to both
1017 memory locations and registers; GCC fundamentally assumes that the
1018 order of words in memory is the same as the order in registers. This
1019 macro need not be a constant.
1022 @defmac LIBGCC2_WORDS_BIG_ENDIAN
1023 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
1024 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
1025 used only when compiling @file{libgcc2.c}. Typically the value will be set
1026 based on preprocessor defines.
1029 @defmac FLOAT_WORDS_BIG_ENDIAN
1030 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
1031 @code{TFmode} floating point numbers are stored in memory with the word
1032 containing the sign bit at the lowest address; otherwise define it to
1033 have the value 0. This macro need not be a constant.
1035 You need not define this macro if the ordering is the same as for
1036 multi-word integers.
1039 @defmac BITS_PER_UNIT
1040 Define this macro to be the number of bits in an addressable storage
1041 unit (byte). If you do not define this macro the default is 8.
1044 @defmac BITS_PER_WORD
1045 Number of bits in a word. If you do not define this macro, the default
1046 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
1049 @defmac MAX_BITS_PER_WORD
1050 Maximum number of bits in a word. If this is undefined, the default is
1051 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
1052 largest value that @code{BITS_PER_WORD} can have at run-time.
1055 @defmac UNITS_PER_WORD
1056 Number of storage units in a word; normally 4.
1059 @defmac MIN_UNITS_PER_WORD
1060 Minimum number of units in a word. If this is undefined, the default is
1061 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
1062 smallest value that @code{UNITS_PER_WORD} can have at run-time.
1065 @defmac POINTER_SIZE
1066 Width of a pointer, in bits. You must specify a value no wider than the
1067 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
1068 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
1069 a value the default is @code{BITS_PER_WORD}.
1072 @defmac POINTERS_EXTEND_UNSIGNED
1073 A C expression whose value is greater than zero if pointers that need to be
1074 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
1075 be zero-extended and zero if they are to be sign-extended. If the value
1076 is less then zero then there must be an "ptr_extend" instruction that
1077 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
1079 You need not define this macro if the @code{POINTER_SIZE} is equal
1080 to the width of @code{Pmode}.
1083 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
1084 A macro to update @var{m} and @var{unsignedp} when an object whose type
1085 is @var{type} and which has the specified mode and signedness is to be
1086 stored in a register. This macro is only called when @var{type} is a
1089 On most RISC machines, which only have operations that operate on a full
1090 register, define this macro to set @var{m} to @code{word_mode} if
1091 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
1092 cases, only integer modes should be widened because wider-precision
1093 floating-point operations are usually more expensive than their narrower
1096 For most machines, the macro definition does not change @var{unsignedp}.
1097 However, some machines, have instructions that preferentially handle
1098 either signed or unsigned quantities of certain modes. For example, on
1099 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
1100 sign-extend the result to 64 bits. On such machines, set
1101 @var{unsignedp} according to which kind of extension is more efficient.
1103 Do not define this macro if it would never modify @var{m}.
1106 @defmac PROMOTE_FUNCTION_MODE
1107 Like @code{PROMOTE_MODE}, but is applied to outgoing function arguments or
1108 function return values, as specified by @code{TARGET_PROMOTE_FUNCTION_ARGS}
1109 and @code{TARGET_PROMOTE_FUNCTION_RETURN}, respectively.
1111 The default is @code{PROMOTE_MODE}.
1114 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_ARGS (tree @var{fntype})
1115 This target hook should return @code{true} if the promotion described by
1116 @code{PROMOTE_FUNCTION_MODE} should be done for outgoing function
1120 @deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_RETURN (tree @var{fntype})
1121 This target hook should return @code{true} if the promotion described by
1122 @code{PROMOTE_FUNCTION_MODE} should be done for the return value of
1125 If this target hook returns @code{true}, @code{FUNCTION_VALUE} must
1126 perform the same promotions done by @code{PROMOTE_FUNCTION_MODE}.
1129 @defmac PARM_BOUNDARY
1130 Normal alignment required for function parameters on the stack, in
1131 bits. All stack parameters receive at least this much alignment
1132 regardless of data type. On most machines, this is the same as the
1136 @defmac STACK_BOUNDARY
1137 Define this macro to the minimum alignment enforced by hardware for the
1138 stack pointer on this machine. The definition is a C expression for the
1139 desired alignment (measured in bits). This value is used as a default
1140 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1141 this should be the same as @code{PARM_BOUNDARY}.
1144 @defmac PREFERRED_STACK_BOUNDARY
1145 Define this macro if you wish to preserve a certain alignment for the
1146 stack pointer, greater than what the hardware enforces. The definition
1147 is a C expression for the desired alignment (measured in bits). This
1148 macro must evaluate to a value equal to or larger than
1149 @code{STACK_BOUNDARY}.
1152 @defmac FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1153 A C expression that evaluates true if @code{PREFERRED_STACK_BOUNDARY} is
1154 not guaranteed by the runtime and we should emit code to align the stack
1155 at the beginning of @code{main}.
1157 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
1158 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
1159 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
1160 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
1161 be momentarily unaligned while pushing arguments.
1164 @defmac FUNCTION_BOUNDARY
1165 Alignment required for a function entry point, in bits.
1168 @defmac BIGGEST_ALIGNMENT
1169 Biggest alignment that any data type can require on this machine, in bits.
1172 @defmac MINIMUM_ATOMIC_ALIGNMENT
1173 If defined, the smallest alignment, in bits, that can be given to an
1174 object that can be referenced in one operation, without disturbing any
1175 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1176 on machines that don't have byte or half-word store operations.
1179 @defmac BIGGEST_FIELD_ALIGNMENT
1180 Biggest alignment that any structure or union field can require on this
1181 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1182 structure and union fields only, unless the field alignment has been set
1183 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1186 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1187 An expression for the alignment of a structure field @var{field} if the
1188 alignment computed in the usual way (including applying of
1189 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1190 alignment) is @var{computed}. It overrides alignment only if the
1191 field alignment has not been set by the
1192 @code{__attribute__ ((aligned (@var{n})))} construct.
1195 @defmac MAX_OFILE_ALIGNMENT
1196 Biggest alignment supported by the object file format of this machine.
1197 Use this macro to limit the alignment which can be specified using the
1198 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1199 the default value is @code{BIGGEST_ALIGNMENT}.
1202 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1203 If defined, a C expression to compute the alignment for a variable in
1204 the static store. @var{type} is the data type, and @var{basic-align} is
1205 the alignment that the object would ordinarily have. The value of this
1206 macro is used instead of that alignment to align the object.
1208 If this macro is not defined, then @var{basic-align} is used.
1211 One use of this macro is to increase alignment of medium-size data to
1212 make it all fit in fewer cache lines. Another is to cause character
1213 arrays to be word-aligned so that @code{strcpy} calls that copy
1214 constants to character arrays can be done inline.
1217 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1218 If defined, a C expression to compute the alignment given to a constant
1219 that is being placed in memory. @var{constant} is the constant and
1220 @var{basic-align} is the alignment that the object would ordinarily
1221 have. The value of this macro is used instead of that alignment to
1224 If this macro is not defined, then @var{basic-align} is used.
1226 The typical use of this macro is to increase alignment for string
1227 constants to be word aligned so that @code{strcpy} calls that copy
1228 constants can be done inline.
1231 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1232 If defined, a C expression to compute the alignment for a variable in
1233 the local store. @var{type} is the data type, and @var{basic-align} is
1234 the alignment that the object would ordinarily have. The value of this
1235 macro is used instead of that alignment to align the object.
1237 If this macro is not defined, then @var{basic-align} is used.
1239 One use of this macro is to increase alignment of medium-size data to
1240 make it all fit in fewer cache lines.
1243 @defmac EMPTY_FIELD_BOUNDARY
1244 Alignment in bits to be given to a structure bit-field that follows an
1245 empty field such as @code{int : 0;}.
1247 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1250 @defmac STRUCTURE_SIZE_BOUNDARY
1251 Number of bits which any structure or union's size must be a multiple of.
1252 Each structure or union's size is rounded up to a multiple of this.
1254 If you do not define this macro, the default is the same as
1255 @code{BITS_PER_UNIT}.
1258 @defmac STRICT_ALIGNMENT
1259 Define this macro to be the value 1 if instructions will fail to work
1260 if given data not on the nominal alignment. If instructions will merely
1261 go slower in that case, define this macro as 0.
1264 @defmac PCC_BITFIELD_TYPE_MATTERS
1265 Define this if you wish to imitate the way many other C compilers handle
1266 alignment of bit-fields and the structures that contain them.
1268 The behavior is that the type written for a named bit-field (@code{int},
1269 @code{short}, or other integer type) imposes an alignment for the entire
1270 structure, as if the structure really did contain an ordinary field of
1271 that type. In addition, the bit-field is placed within the structure so
1272 that it would fit within such a field, not crossing a boundary for it.
1274 Thus, on most machines, a named bit-field whose type is written as
1275 @code{int} would not cross a four-byte boundary, and would force
1276 four-byte alignment for the whole structure. (The alignment used may
1277 not be four bytes; it is controlled by the other alignment parameters.)
1279 An unnamed bit-field will not affect the alignment of the containing
1282 If the macro is defined, its definition should be a C expression;
1283 a nonzero value for the expression enables this behavior.
1285 Note that if this macro is not defined, or its value is zero, some
1286 bit-fields may cross more than one alignment boundary. The compiler can
1287 support such references if there are @samp{insv}, @samp{extv}, and
1288 @samp{extzv} insns that can directly reference memory.
1290 The other known way of making bit-fields work is to define
1291 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1292 Then every structure can be accessed with fullwords.
1294 Unless the machine has bit-field instructions or you define
1295 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1296 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1298 If your aim is to make GCC use the same conventions for laying out
1299 bit-fields as are used by another compiler, here is how to investigate
1300 what the other compiler does. Compile and run this program:
1319 printf ("Size of foo1 is %d\n",
1320 sizeof (struct foo1));
1321 printf ("Size of foo2 is %d\n",
1322 sizeof (struct foo2));
1327 If this prints 2 and 5, then the compiler's behavior is what you would
1328 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1331 @defmac BITFIELD_NBYTES_LIMITED
1332 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1333 to aligning a bit-field within the structure.
1336 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELDS (void)
1337 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1338 whether unnamed bitfields affect the alignment of the containing
1339 structure. The hook should return true if the structure should inherit
1340 the alignment requirements of an unnamed bitfield's type.
1343 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1344 Return 1 if a structure or array containing @var{field} should be accessed using
1347 If @var{field} is the only field in the structure, @var{mode} is its
1348 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1349 case where structures of one field would require the structure's mode to
1350 retain the field's mode.
1352 Normally, this is not needed. See the file @file{c4x.h} for an example
1353 of how to use this macro to prevent a structure having a floating point
1354 field from being accessed in an integer mode.
1357 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1358 Define this macro as an expression for the alignment of a type (given
1359 by @var{type} as a tree node) if the alignment computed in the usual
1360 way is @var{computed} and the alignment explicitly specified was
1363 The default is to use @var{specified} if it is larger; otherwise, use
1364 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1367 @defmac MAX_FIXED_MODE_SIZE
1368 An integer expression for the size in bits of the largest integer
1369 machine mode that should actually be used. All integer machine modes of
1370 this size or smaller can be used for structures and unions with the
1371 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1372 (DImode)} is assumed.
1375 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1376 If defined, an expression of type @code{enum machine_mode} that
1377 specifies the mode of the save area operand of a
1378 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1379 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1380 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1381 having its mode specified.
1383 You need not define this macro if it always returns @code{Pmode}. You
1384 would most commonly define this macro if the
1385 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1389 @defmac STACK_SIZE_MODE
1390 If defined, an expression of type @code{enum machine_mode} that
1391 specifies the mode of the size increment operand of an
1392 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1394 You need not define this macro if it always returns @code{word_mode}.
1395 You would most commonly define this macro if the @code{allocate_stack}
1396 pattern needs to support both a 32- and a 64-bit mode.
1399 @defmac TARGET_FLOAT_FORMAT
1400 A code distinguishing the floating point format of the target machine.
1401 There are four defined values:
1404 @item IEEE_FLOAT_FORMAT
1405 This code indicates IEEE floating point. It is the default; there is no
1406 need to define @code{TARGET_FLOAT_FORMAT} when the format is IEEE@.
1408 @item VAX_FLOAT_FORMAT
1409 This code indicates the ``F float'' (for @code{float}) and ``D float''
1410 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
1412 @item IBM_FLOAT_FORMAT
1413 This code indicates the format used on the IBM System/370.
1415 @item C4X_FLOAT_FORMAT
1416 This code indicates the format used on the TMS320C3x/C4x.
1419 If your target uses a floating point format other than these, you must
1420 define a new @var{name}_FLOAT_FORMAT code for it, and add support for
1421 it to @file{real.c}.
1423 The ordering of the component words of floating point values stored in
1424 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1427 @defmac MODE_HAS_NANS (@var{mode})
1428 When defined, this macro should be true if @var{mode} has a NaN
1429 representation. The compiler assumes that NaNs are not equal to
1430 anything (including themselves) and that addition, subtraction,
1431 multiplication and division all return NaNs when one operand is
1434 By default, this macro is true if @var{mode} is a floating-point
1435 mode and the target floating-point format is IEEE@.
1438 @defmac MODE_HAS_INFINITIES (@var{mode})
1439 This macro should be true if @var{mode} can represent infinity. At
1440 present, the compiler uses this macro to decide whether @samp{x - x}
1441 is always defined. By default, the macro is true when @var{mode}
1442 is a floating-point mode and the target format is IEEE@.
1445 @defmac MODE_HAS_SIGNED_ZEROS (@var{mode})
1446 True if @var{mode} distinguishes between positive and negative zero.
1447 The rules are expected to follow the IEEE standard:
1451 @samp{x + x} has the same sign as @samp{x}.
1454 If the sum of two values with opposite sign is zero, the result is
1455 positive for all rounding modes expect towards @minus{}infinity, for
1456 which it is negative.
1459 The sign of a product or quotient is negative when exactly one
1460 of the operands is negative.
1463 The default definition is true if @var{mode} is a floating-point
1464 mode and the target format is IEEE@.
1467 @defmac MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1468 If defined, this macro should be true for @var{mode} if it has at
1469 least one rounding mode in which @samp{x} and @samp{-x} can be
1470 rounded to numbers of different magnitude. Two such modes are
1471 towards @minus{}infinity and towards +infinity.
1473 The default definition of this macro is true if @var{mode} is
1474 a floating-point mode and the target format is IEEE@.
1477 @defmac ROUND_TOWARDS_ZERO
1478 If defined, this macro should be true if the prevailing rounding
1479 mode is towards zero. A true value has the following effects:
1483 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1486 @file{libgcc.a}'s floating-point emulator will round towards zero
1487 rather than towards nearest.
1490 The compiler's floating-point emulator will round towards zero after
1491 doing arithmetic, and when converting from the internal float format to
1495 The macro does not affect the parsing of string literals. When the
1496 primary rounding mode is towards zero, library functions like
1497 @code{strtod} might still round towards nearest, and the compiler's
1498 parser should behave like the target's @code{strtod} where possible.
1500 Not defining this macro is equivalent to returning zero.
1503 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1504 This macro should return true if floats with @var{size}
1505 bits do not have a NaN or infinity representation, but use the largest
1506 exponent for normal numbers instead.
1508 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1509 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1510 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1511 floating-point arithmetic.
1513 The default definition of this macro returns false for all sizes.
1516 @deftypefn {Target Hook} bool TARGET_VECTOR_OPAQUE_P (tree @var{type})
1517 This target hook should return @code{true} a vector is opaque. That
1518 is, if no cast is needed when copying a vector value of type
1519 @var{type} into another vector lvalue of the same size. Vector opaque
1520 types cannot be initialized. The default is that there are no such
1524 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1525 This target hook returns @code{true} if bit-fields in the given
1526 @var{record_type} are to be laid out following the rules of Microsoft
1527 Visual C/C++, namely: (i) a bit-field won't share the same storage
1528 unit with the previous bit-field if their underlying types have
1529 different sizes, and the bit-field will be aligned to the highest
1530 alignment of the underlying types of itself and of the previous
1531 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1532 the whole enclosing structure, even if it is unnamed; except that
1533 (iii) a zero-sized bit-field will be disregarded unless it follows
1534 another bit-field of nonzero size. If this hook returns @code{true},
1535 other macros that control bit-field layout are ignored.
1537 When a bit-field is inserted into a packed record, the whole size
1538 of the underlying type is used by one or more same-size adjacent
1539 bit-fields (that is, if its long:3, 32 bits is used in the record,
1540 and any additional adjacent long bit-fields are packed into the same
1541 chunk of 32 bits. However, if the size changes, a new field of that
1542 size is allocated). In an unpacked record, this is the same as using
1543 alignment, but not equivalent when packing.
1545 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1546 the latter will take precedence. If @samp{__attribute__((packed))} is
1547 used on a single field when MS bit-fields are in use, it will take
1548 precedence for that field, but the alignment of the rest of the structure
1549 may affect its placement.
1552 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_FUNDAMENTAL_TYPE (tree @var{type})
1553 If your target defines any fundamental types, define this hook to
1554 return the appropriate encoding for these types as part of a C++
1555 mangled name. The @var{type} argument is the tree structure
1556 representing the type to be mangled. The hook may be applied to trees
1557 which are not target-specific fundamental types; it should return
1558 @code{NULL} for all such types, as well as arguments it does not
1559 recognize. If the return value is not @code{NULL}, it must point to
1560 a statically-allocated string constant.
1562 Target-specific fundamental types might be new fundamental types or
1563 qualified versions of ordinary fundamental types. Encode new
1564 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1565 is the name used for the type in source code, and @var{n} is the
1566 length of @var{name} in decimal. Encode qualified versions of
1567 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1568 @var{name} is the name used for the type qualifier in source code,
1569 @var{n} is the length of @var{name} as above, and @var{code} is the
1570 code used to represent the unqualified version of this type. (See
1571 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1572 codes.) In both cases the spaces are for clarity; do not include any
1573 spaces in your string.
1575 The default version of this hook always returns @code{NULL}, which is
1576 appropriate for a target that does not define any new fundamental
1581 @section Layout of Source Language Data Types
1583 These macros define the sizes and other characteristics of the standard
1584 basic data types used in programs being compiled. Unlike the macros in
1585 the previous section, these apply to specific features of C and related
1586 languages, rather than to fundamental aspects of storage layout.
1588 @defmac INT_TYPE_SIZE
1589 A C expression for the size in bits of the type @code{int} on the
1590 target machine. If you don't define this, the default is one word.
1593 @defmac SHORT_TYPE_SIZE
1594 A C expression for the size in bits of the type @code{short} on the
1595 target machine. If you don't define this, the default is half a word.
1596 (If this would be less than one storage unit, it is rounded up to one
1600 @defmac LONG_TYPE_SIZE
1601 A C expression for the size in bits of the type @code{long} on the
1602 target machine. If you don't define this, the default is one word.
1605 @defmac ADA_LONG_TYPE_SIZE
1606 On some machines, the size used for the Ada equivalent of the type
1607 @code{long} by a native Ada compiler differs from that used by C@. In
1608 that situation, define this macro to be a C expression to be used for
1609 the size of that type. If you don't define this, the default is the
1610 value of @code{LONG_TYPE_SIZE}.
1613 @defmac LONG_LONG_TYPE_SIZE
1614 A C expression for the size in bits of the type @code{long long} on the
1615 target machine. If you don't define this, the default is two
1616 words. If you want to support GNU Ada on your machine, the value of this
1617 macro must be at least 64.
1620 @defmac CHAR_TYPE_SIZE
1621 A C expression for the size in bits of the type @code{char} on the
1622 target machine. If you don't define this, the default is
1623 @code{BITS_PER_UNIT}.
1626 @defmac BOOL_TYPE_SIZE
1627 A C expression for the size in bits of the C++ type @code{bool} and
1628 C99 type @code{_Bool} on the target machine. If you don't define
1629 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1632 @defmac FLOAT_TYPE_SIZE
1633 A C expression for the size in bits of the type @code{float} on the
1634 target machine. If you don't define this, the default is one word.
1637 @defmac DOUBLE_TYPE_SIZE
1638 A C expression for the size in bits of the type @code{double} on the
1639 target machine. If you don't define this, the default is two
1643 @defmac LONG_DOUBLE_TYPE_SIZE
1644 A C expression for the size in bits of the type @code{long double} on
1645 the target machine. If you don't define this, the default is two
1649 @defmac TARGET_FLT_EVAL_METHOD
1650 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1651 assuming, if applicable, that the floating-point control word is in its
1652 default state. If you do not define this macro the value of
1653 @code{FLT_EVAL_METHOD} will be zero.
1656 @defmac WIDEST_HARDWARE_FP_SIZE
1657 A C expression for the size in bits of the widest floating-point format
1658 supported by the hardware. If you define this macro, you must specify a
1659 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1660 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1664 @defmac DEFAULT_SIGNED_CHAR
1665 An expression whose value is 1 or 0, according to whether the type
1666 @code{char} should be signed or unsigned by default. The user can
1667 always override this default with the options @option{-fsigned-char}
1668 and @option{-funsigned-char}.
1671 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1672 This target hook should return true if the compiler should give an
1673 @code{enum} type only as many bytes as it takes to represent the range
1674 of possible values of that type. It should return false if all
1675 @code{enum} types should be allocated like @code{int}.
1677 The default is to return false.
1681 A C expression for a string describing the name of the data type to use
1682 for size values. The typedef name @code{size_t} is defined using the
1683 contents of the string.
1685 The string can contain more than one keyword. If so, separate them with
1686 spaces, and write first any length keyword, then @code{unsigned} if
1687 appropriate, and finally @code{int}. The string must exactly match one
1688 of the data type names defined in the function
1689 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1690 omit @code{int} or change the order---that would cause the compiler to
1693 If you don't define this macro, the default is @code{"long unsigned
1697 @defmac PTRDIFF_TYPE
1698 A C expression for a string describing the name of the data type to use
1699 for the result of subtracting two pointers. The typedef name
1700 @code{ptrdiff_t} is defined using the contents of the string. See
1701 @code{SIZE_TYPE} above for more information.
1703 If you don't define this macro, the default is @code{"long int"}.
1707 A C expression for a string describing the name of the data type to use
1708 for wide characters. The typedef name @code{wchar_t} is defined using
1709 the contents of the string. See @code{SIZE_TYPE} above for more
1712 If you don't define this macro, the default is @code{"int"}.
1715 @defmac WCHAR_TYPE_SIZE
1716 A C expression for the size in bits of the data type for wide
1717 characters. This is used in @code{cpp}, which cannot make use of
1722 A C expression for a string describing the name of the data type to
1723 use for wide characters passed to @code{printf} and returned from
1724 @code{getwc}. The typedef name @code{wint_t} is defined using the
1725 contents of the string. See @code{SIZE_TYPE} above for more
1728 If you don't define this macro, the default is @code{"unsigned int"}.
1732 A C expression for a string describing the name of the data type that
1733 can represent any value of any standard or extended signed integer type.
1734 The typedef name @code{intmax_t} is defined using the contents of the
1735 string. See @code{SIZE_TYPE} above for more information.
1737 If you don't define this macro, the default is the first of
1738 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1739 much precision as @code{long long int}.
1742 @defmac UINTMAX_TYPE
1743 A C expression for a string describing the name of the data type that
1744 can represent any value of any standard or extended unsigned integer
1745 type. The typedef name @code{uintmax_t} is defined using the contents
1746 of the string. See @code{SIZE_TYPE} above for more information.
1748 If you don't define this macro, the default is the first of
1749 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1750 unsigned int"} that has as much precision as @code{long long unsigned
1754 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1755 The C++ compiler represents a pointer-to-member-function with a struct
1762 ptrdiff_t vtable_index;
1769 The C++ compiler must use one bit to indicate whether the function that
1770 will be called through a pointer-to-member-function is virtual.
1771 Normally, we assume that the low-order bit of a function pointer must
1772 always be zero. Then, by ensuring that the vtable_index is odd, we can
1773 distinguish which variant of the union is in use. But, on some
1774 platforms function pointers can be odd, and so this doesn't work. In
1775 that case, we use the low-order bit of the @code{delta} field, and shift
1776 the remainder of the @code{delta} field to the left.
1778 GCC will automatically make the right selection about where to store
1779 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1780 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1781 set such that functions always start at even addresses, but the lowest
1782 bit of pointers to functions indicate whether the function at that
1783 address is in ARM or Thumb mode. If this is the case of your
1784 architecture, you should define this macro to
1785 @code{ptrmemfunc_vbit_in_delta}.
1787 In general, you should not have to define this macro. On architectures
1788 in which function addresses are always even, according to
1789 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1790 @code{ptrmemfunc_vbit_in_pfn}.
1793 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1794 Normally, the C++ compiler uses function pointers in vtables. This
1795 macro allows the target to change to use ``function descriptors''
1796 instead. Function descriptors are found on targets for whom a
1797 function pointer is actually a small data structure. Normally the
1798 data structure consists of the actual code address plus a data
1799 pointer to which the function's data is relative.
1801 If vtables are used, the value of this macro should be the number
1802 of words that the function descriptor occupies.
1805 @defmac TARGET_VTABLE_ENTRY_ALIGN
1806 By default, the vtable entries are void pointers, the so the alignment
1807 is the same as pointer alignment. The value of this macro specifies
1808 the alignment of the vtable entry in bits. It should be defined only
1809 when special alignment is necessary. */
1812 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1813 There are a few non-descriptor entries in the vtable at offsets below
1814 zero. If these entries must be padded (say, to preserve the alignment
1815 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1816 of words in each data entry.
1819 @node Escape Sequences
1820 @section Target Character Escape Sequences
1821 @cindex escape sequences
1823 By default, GCC assumes that the C character escape sequences and other
1824 characters take on their ASCII values for the target. If this is not
1825 correct, you must explicitly define all of the macros below. All of
1826 them must evaluate to constants; they are used in @code{case}
1832 @findex TARGET_DIGIT0
1835 @findex TARGET_NEWLINE
1838 @multitable {@code{TARGET_NEWLINE}} {Escape} {ASCII character}
1839 @item Macro @tab Escape @tab ASCII character
1840 @item @code{TARGET_BELL} @tab @kbd{\a} @tab @code{07}, @code{BEL}
1841 @item @code{TARGET_BS} @tab @kbd{\b} @tab @code{08}, @code{BS}
1842 @item @code{TARGET_CR} @tab @kbd{\r} @tab @code{0D}, @code{CR}
1843 @item @code{TARGET_DIGIT0} @tab @kbd{0} @tab @code{30}, @code{ZERO}
1844 @item @code{TARGET_ESC} @tab @kbd{\e}, @kbd{\E} @tab @code{1B}, @code{ESC}
1845 @item @code{TARGET_FF} @tab @kbd{\f} @tab @code{0C}, @code{FF}
1846 @item @code{TARGET_NEWLINE} @tab @kbd{\n} @tab @code{0A}, @code{LF}
1847 @item @code{TARGET_TAB} @tab @kbd{\t} @tab @code{09}, @code{HT}
1848 @item @code{TARGET_VT} @tab @kbd{\v} @tab @code{0B}, @code{VT}
1852 Note that the @kbd{\e} and @kbd{\E} escapes are GNU extensions, not
1853 part of the C standard.
1856 @section Register Usage
1857 @cindex register usage
1859 This section explains how to describe what registers the target machine
1860 has, and how (in general) they can be used.
1862 The description of which registers a specific instruction can use is
1863 done with register classes; see @ref{Register Classes}. For information
1864 on using registers to access a stack frame, see @ref{Frame Registers}.
1865 For passing values in registers, see @ref{Register Arguments}.
1866 For returning values in registers, see @ref{Scalar Return}.
1869 * Register Basics:: Number and kinds of registers.
1870 * Allocation Order:: Order in which registers are allocated.
1871 * Values in Registers:: What kinds of values each reg can hold.
1872 * Leaf Functions:: Renumbering registers for leaf functions.
1873 * Stack Registers:: Handling a register stack such as 80387.
1876 @node Register Basics
1877 @subsection Basic Characteristics of Registers
1879 @c prevent bad page break with this line
1880 Registers have various characteristics.
1882 @defmac FIRST_PSEUDO_REGISTER
1883 Number of hardware registers known to the compiler. They receive
1884 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1885 pseudo register's number really is assigned the number
1886 @code{FIRST_PSEUDO_REGISTER}.
1889 @defmac FIXED_REGISTERS
1890 @cindex fixed register
1891 An initializer that says which registers are used for fixed purposes
1892 all throughout the compiled code and are therefore not available for
1893 general allocation. These would include the stack pointer, the frame
1894 pointer (except on machines where that can be used as a general
1895 register when no frame pointer is needed), the program counter on
1896 machines where that is considered one of the addressable registers,
1897 and any other numbered register with a standard use.
1899 This information is expressed as a sequence of numbers, separated by
1900 commas and surrounded by braces. The @var{n}th number is 1 if
1901 register @var{n} is fixed, 0 otherwise.
1903 The table initialized from this macro, and the table initialized by
1904 the following one, may be overridden at run time either automatically,
1905 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1906 the user with the command options @option{-ffixed-@var{reg}},
1907 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1910 @defmac CALL_USED_REGISTERS
1911 @cindex call-used register
1912 @cindex call-clobbered register
1913 @cindex call-saved register
1914 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1915 clobbered (in general) by function calls as well as for fixed
1916 registers. This macro therefore identifies the registers that are not
1917 available for general allocation of values that must live across
1920 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1921 automatically saves it on function entry and restores it on function
1922 exit, if the register is used within the function.
1925 @defmac CALL_REALLY_USED_REGISTERS
1926 @cindex call-used register
1927 @cindex call-clobbered register
1928 @cindex call-saved register
1929 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1930 that the entire set of @code{FIXED_REGISTERS} be included.
1931 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1932 This macro is optional. If not specified, it defaults to the value
1933 of @code{CALL_USED_REGISTERS}.
1936 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1937 @cindex call-used register
1938 @cindex call-clobbered register
1939 @cindex call-saved register
1940 A C expression that is nonzero if it is not permissible to store a
1941 value of mode @var{mode} in hard register number @var{regno} across a
1942 call without some part of it being clobbered. For most machines this
1943 macro need not be defined. It is only required for machines that do not
1944 preserve the entire contents of a register across a call.
1948 @findex call_used_regs
1951 @findex reg_class_contents
1952 @defmac CONDITIONAL_REGISTER_USAGE
1953 Zero or more C statements that may conditionally modify five variables
1954 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1955 @code{reg_names}, and @code{reg_class_contents}, to take into account
1956 any dependence of these register sets on target flags. The first three
1957 of these are of type @code{char []} (interpreted as Boolean vectors).
1958 @code{global_regs} is a @code{const char *[]}, and
1959 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1960 called, @code{fixed_regs}, @code{call_used_regs},
1961 @code{reg_class_contents}, and @code{reg_names} have been initialized
1962 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1963 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1964 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1965 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1966 command options have been applied.
1968 You need not define this macro if it has no work to do.
1970 @cindex disabling certain registers
1971 @cindex controlling register usage
1972 If the usage of an entire class of registers depends on the target
1973 flags, you may indicate this to GCC by using this macro to modify
1974 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1975 registers in the classes which should not be used by GCC@. Also define
1976 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1977 to return @code{NO_REGS} if it
1978 is called with a letter for a class that shouldn't be used.
1980 (However, if this class is not included in @code{GENERAL_REGS} and all
1981 of the insn patterns whose constraints permit this class are
1982 controlled by target switches, then GCC will automatically avoid using
1983 these registers when the target switches are opposed to them.)
1986 @defmac INCOMING_REGNO (@var{out})
1987 Define this macro if the target machine has register windows. This C
1988 expression returns the register number as seen by the called function
1989 corresponding to the register number @var{out} as seen by the calling
1990 function. Return @var{out} if register number @var{out} is not an
1994 @defmac OUTGOING_REGNO (@var{in})
1995 Define this macro if the target machine has register windows. This C
1996 expression returns the register number as seen by the calling function
1997 corresponding to the register number @var{in} as seen by the called
1998 function. Return @var{in} if register number @var{in} is not an inbound
2002 @defmac LOCAL_REGNO (@var{regno})
2003 Define this macro if the target machine has register windows. This C
2004 expression returns true if the register is call-saved but is in the
2005 register window. Unlike most call-saved registers, such registers
2006 need not be explicitly restored on function exit or during non-local
2011 If the program counter has a register number, define this as that
2012 register number. Otherwise, do not define it.
2015 @node Allocation Order
2016 @subsection Order of Allocation of Registers
2017 @cindex order of register allocation
2018 @cindex register allocation order
2020 @c prevent bad page break with this line
2021 Registers are allocated in order.
2023 @defmac REG_ALLOC_ORDER
2024 If defined, an initializer for a vector of integers, containing the
2025 numbers of hard registers in the order in which GCC should prefer
2026 to use them (from most preferred to least).
2028 If this macro is not defined, registers are used lowest numbered first
2029 (all else being equal).
2031 One use of this macro is on machines where the highest numbered
2032 registers must always be saved and the save-multiple-registers
2033 instruction supports only sequences of consecutive registers. On such
2034 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
2035 the highest numbered allocable register first.
2038 @defmac ORDER_REGS_FOR_LOCAL_ALLOC
2039 A C statement (sans semicolon) to choose the order in which to allocate
2040 hard registers for pseudo-registers local to a basic block.
2042 Store the desired register order in the array @code{reg_alloc_order}.
2043 Element 0 should be the register to allocate first; element 1, the next
2044 register; and so on.
2046 The macro body should not assume anything about the contents of
2047 @code{reg_alloc_order} before execution of the macro.
2049 On most machines, it is not necessary to define this macro.
2052 @node Values in Registers
2053 @subsection How Values Fit in Registers
2055 This section discusses the macros that describe which kinds of values
2056 (specifically, which machine modes) each register can hold, and how many
2057 consecutive registers are needed for a given mode.
2059 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2060 A C expression for the number of consecutive hard registers, starting
2061 at register number @var{regno}, required to hold a value of mode
2064 On a machine where all registers are exactly one word, a suitable
2065 definition of this macro is
2068 #define HARD_REGNO_NREGS(REGNO, MODE) \
2069 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2074 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2075 Define this macro if the natural size of registers that hold values
2076 of mode @var{mode} is not the word size. It is a C expression that
2077 should give the natural size in bytes for the specified mode. It is
2078 used by the register allocator to try to optimize its results. This
2079 happens for example on SPARC 64-bit where the natural size of
2080 floating-point registers is still 32-bit.
2083 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2084 A C expression that is nonzero if it is permissible to store a value
2085 of mode @var{mode} in hard register number @var{regno} (or in several
2086 registers starting with that one). For a machine where all registers
2087 are equivalent, a suitable definition is
2090 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2093 You need not include code to check for the numbers of fixed registers,
2094 because the allocation mechanism considers them to be always occupied.
2096 @cindex register pairs
2097 On some machines, double-precision values must be kept in even/odd
2098 register pairs. You can implement that by defining this macro to reject
2099 odd register numbers for such modes.
2101 The minimum requirement for a mode to be OK in a register is that the
2102 @samp{mov@var{mode}} instruction pattern support moves between the
2103 register and other hard register in the same class and that moving a
2104 value into the register and back out not alter it.
2106 Since the same instruction used to move @code{word_mode} will work for
2107 all narrower integer modes, it is not necessary on any machine for
2108 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2109 you define patterns @samp{movhi}, etc., to take advantage of this. This
2110 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2111 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2114 Many machines have special registers for floating point arithmetic.
2115 Often people assume that floating point machine modes are allowed only
2116 in floating point registers. This is not true. Any registers that
2117 can hold integers can safely @emph{hold} a floating point machine
2118 mode, whether or not floating arithmetic can be done on it in those
2119 registers. Integer move instructions can be used to move the values.
2121 On some machines, though, the converse is true: fixed-point machine
2122 modes may not go in floating registers. This is true if the floating
2123 registers normalize any value stored in them, because storing a
2124 non-floating value there would garble it. In this case,
2125 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2126 floating registers. But if the floating registers do not automatically
2127 normalize, if you can store any bit pattern in one and retrieve it
2128 unchanged without a trap, then any machine mode may go in a floating
2129 register, so you can define this macro to say so.
2131 The primary significance of special floating registers is rather that
2132 they are the registers acceptable in floating point arithmetic
2133 instructions. However, this is of no concern to
2134 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2135 constraints for those instructions.
2137 On some machines, the floating registers are especially slow to access,
2138 so that it is better to store a value in a stack frame than in such a
2139 register if floating point arithmetic is not being done. As long as the
2140 floating registers are not in class @code{GENERAL_REGS}, they will not
2141 be used unless some pattern's constraint asks for one.
2144 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2145 A C expression that is nonzero if it is OK to rename a hard register
2146 @var{from} to another hard register @var{to}.
2148 One common use of this macro is to prevent renaming of a register to
2149 another register that is not saved by a prologue in an interrupt
2152 The default is always nonzero.
2155 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2156 A C expression that is nonzero if a value of mode
2157 @var{mode1} is accessible in mode @var{mode2} without copying.
2159 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2160 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2161 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2162 should be nonzero. If they differ for any @var{r}, you should define
2163 this macro to return zero unless some other mechanism ensures the
2164 accessibility of the value in a narrower mode.
2166 You should define this macro to return nonzero in as many cases as
2167 possible since doing so will allow GCC to perform better register
2171 @defmac AVOID_CCMODE_COPIES
2172 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2173 registers. You should only define this macro if support for copying to/from
2174 @code{CCmode} is incomplete.
2177 @node Leaf Functions
2178 @subsection Handling Leaf Functions
2180 @cindex leaf functions
2181 @cindex functions, leaf
2182 On some machines, a leaf function (i.e., one which makes no calls) can run
2183 more efficiently if it does not make its own register window. Often this
2184 means it is required to receive its arguments in the registers where they
2185 are passed by the caller, instead of the registers where they would
2188 The special treatment for leaf functions generally applies only when
2189 other conditions are met; for example, often they may use only those
2190 registers for its own variables and temporaries. We use the term ``leaf
2191 function'' to mean a function that is suitable for this special
2192 handling, so that functions with no calls are not necessarily ``leaf
2195 GCC assigns register numbers before it knows whether the function is
2196 suitable for leaf function treatment. So it needs to renumber the
2197 registers in order to output a leaf function. The following macros
2200 @defmac LEAF_REGISTERS
2201 Name of a char vector, indexed by hard register number, which
2202 contains 1 for a register that is allowable in a candidate for leaf
2205 If leaf function treatment involves renumbering the registers, then the
2206 registers marked here should be the ones before renumbering---those that
2207 GCC would ordinarily allocate. The registers which will actually be
2208 used in the assembler code, after renumbering, should not be marked with 1
2211 Define this macro only if the target machine offers a way to optimize
2212 the treatment of leaf functions.
2215 @defmac LEAF_REG_REMAP (@var{regno})
2216 A C expression whose value is the register number to which @var{regno}
2217 should be renumbered, when a function is treated as a leaf function.
2219 If @var{regno} is a register number which should not appear in a leaf
2220 function before renumbering, then the expression should yield @minus{}1, which
2221 will cause the compiler to abort.
2223 Define this macro only if the target machine offers a way to optimize the
2224 treatment of leaf functions, and registers need to be renumbered to do
2228 @findex current_function_is_leaf
2229 @findex current_function_uses_only_leaf_regs
2230 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2231 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2232 specially. They can test the C variable @code{current_function_is_leaf}
2233 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2234 set prior to local register allocation and is valid for the remaining
2235 compiler passes. They can also test the C variable
2236 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2237 functions which only use leaf registers.
2238 @code{current_function_uses_only_leaf_regs} is valid after all passes
2239 that modify the instructions have been run and is only useful if
2240 @code{LEAF_REGISTERS} is defined.
2241 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2242 @c of the next paragraph?! --mew 2feb93
2244 @node Stack Registers
2245 @subsection Registers That Form a Stack
2247 There are special features to handle computers where some of the
2248 ``registers'' form a stack. Stack registers are normally written by
2249 pushing onto the stack, and are numbered relative to the top of the
2252 Currently, GCC can only handle one group of stack-like registers, and
2253 they must be consecutively numbered. Furthermore, the existing
2254 support for stack-like registers is specific to the 80387 floating
2255 point coprocessor. If you have a new architecture that uses
2256 stack-like registers, you will need to do substantial work on
2257 @file{reg-stack.c} and write your machine description to cooperate
2258 with it, as well as defining these macros.
2261 Define this if the machine has any stack-like registers.
2264 @defmac FIRST_STACK_REG
2265 The number of the first stack-like register. This one is the top
2269 @defmac LAST_STACK_REG
2270 The number of the last stack-like register. This one is the bottom of
2274 @node Register Classes
2275 @section Register Classes
2276 @cindex register class definitions
2277 @cindex class definitions, register
2279 On many machines, the numbered registers are not all equivalent.
2280 For example, certain registers may not be allowed for indexed addressing;
2281 certain registers may not be allowed in some instructions. These machine
2282 restrictions are described to the compiler using @dfn{register classes}.
2284 You define a number of register classes, giving each one a name and saying
2285 which of the registers belong to it. Then you can specify register classes
2286 that are allowed as operands to particular instruction patterns.
2290 In general, each register will belong to several classes. In fact, one
2291 class must be named @code{ALL_REGS} and contain all the registers. Another
2292 class must be named @code{NO_REGS} and contain no registers. Often the
2293 union of two classes will be another class; however, this is not required.
2295 @findex GENERAL_REGS
2296 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2297 terribly special about the name, but the operand constraint letters
2298 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2299 the same as @code{ALL_REGS}, just define it as a macro which expands
2302 Order the classes so that if class @var{x} is contained in class @var{y}
2303 then @var{x} has a lower class number than @var{y}.
2305 The way classes other than @code{GENERAL_REGS} are specified in operand
2306 constraints is through machine-dependent operand constraint letters.
2307 You can define such letters to correspond to various classes, then use
2308 them in operand constraints.
2310 You should define a class for the union of two classes whenever some
2311 instruction allows both classes. For example, if an instruction allows
2312 either a floating point (coprocessor) register or a general register for a
2313 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2314 which includes both of them. Otherwise you will get suboptimal code.
2316 You must also specify certain redundant information about the register
2317 classes: for each class, which classes contain it and which ones are
2318 contained in it; for each pair of classes, the largest class contained
2321 When a value occupying several consecutive registers is expected in a
2322 certain class, all the registers used must belong to that class.
2323 Therefore, register classes cannot be used to enforce a requirement for
2324 a register pair to start with an even-numbered register. The way to
2325 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2327 Register classes used for input-operands of bitwise-and or shift
2328 instructions have a special requirement: each such class must have, for
2329 each fixed-point machine mode, a subclass whose registers can transfer that
2330 mode to or from memory. For example, on some machines, the operations for
2331 single-byte values (@code{QImode}) are limited to certain registers. When
2332 this is so, each register class that is used in a bitwise-and or shift
2333 instruction must have a subclass consisting of registers from which
2334 single-byte values can be loaded or stored. This is so that
2335 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2337 @deftp {Data type} {enum reg_class}
2338 An enumerated type that must be defined with all the register class names
2339 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2340 must be the last register class, followed by one more enumerated value,
2341 @code{LIM_REG_CLASSES}, which is not a register class but rather
2342 tells how many classes there are.
2344 Each register class has a number, which is the value of casting
2345 the class name to type @code{int}. The number serves as an index
2346 in many of the tables described below.
2349 @defmac N_REG_CLASSES
2350 The number of distinct register classes, defined as follows:
2353 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2357 @defmac REG_CLASS_NAMES
2358 An initializer containing the names of the register classes as C string
2359 constants. These names are used in writing some of the debugging dumps.
2362 @defmac REG_CLASS_CONTENTS
2363 An initializer containing the contents of the register classes, as integers
2364 which are bit masks. The @var{n}th integer specifies the contents of class
2365 @var{n}. The way the integer @var{mask} is interpreted is that
2366 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2368 When the machine has more than 32 registers, an integer does not suffice.
2369 Then the integers are replaced by sub-initializers, braced groupings containing
2370 several integers. Each sub-initializer must be suitable as an initializer
2371 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2372 In this situation, the first integer in each sub-initializer corresponds to
2373 registers 0 through 31, the second integer to registers 32 through 63, and
2377 @defmac REGNO_REG_CLASS (@var{regno})
2378 A C expression whose value is a register class containing hard register
2379 @var{regno}. In general there is more than one such class; choose a class
2380 which is @dfn{minimal}, meaning that no smaller class also contains the
2384 @defmac BASE_REG_CLASS
2385 A macro whose definition is the name of the class to which a valid
2386 base register must belong. A base register is one used in an address
2387 which is the register value plus a displacement.
2390 @defmac MODE_BASE_REG_CLASS (@var{mode})
2391 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2392 the selection of a base register in a mode dependent manner. If
2393 @var{mode} is VOIDmode then it should return the same value as
2394 @code{BASE_REG_CLASS}.
2397 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2398 A C expression whose value is the register class to which a valid
2399 base register must belong in order to be used in a base plus index
2400 register address. You should define this macro if base plus index
2401 addresses have different requirements than other base register uses.
2404 @defmac INDEX_REG_CLASS
2405 A macro whose definition is the name of the class to which a valid
2406 index register must belong. An index register is one used in an
2407 address where its value is either multiplied by a scale factor or
2408 added to another register (as well as added to a displacement).
2411 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2412 For the constraint at the start of @var{str}, which starts with the letter
2413 @var{c}, return the length. This allows you to have register class /
2414 constant / extra constraints that are longer than a single letter;
2415 you don't need to define this macro if you can do with single-letter
2416 constraints only. The definition of this macro should use
2417 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2418 to handle specially.
2419 There are some sanity checks in genoutput.c that check the constraint lengths
2420 for the md file, so you can also use this macro to help you while you are
2421 transitioning from a byzantine single-letter-constraint scheme: when you
2422 return a negative length for a constraint you want to re-use, genoutput
2423 will complain about every instance where it is used in the md file.
2426 @defmac REG_CLASS_FROM_LETTER (@var{char})
2427 A C expression which defines the machine-dependent operand constraint
2428 letters for register classes. If @var{char} is such a letter, the
2429 value should be the register class corresponding to it. Otherwise,
2430 the value should be @code{NO_REGS}. The register letter @samp{r},
2431 corresponding to class @code{GENERAL_REGS}, will not be passed
2432 to this macro; you do not need to handle it.
2435 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2436 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2437 passed in @var{str}, so that you can use suffixes to distinguish between
2441 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2442 A C expression which is nonzero if register number @var{num} is
2443 suitable for use as a base register in operand addresses. It may be
2444 either a suitable hard register or a pseudo register that has been
2445 allocated such a hard register.
2448 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2449 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2450 that expression may examine the mode of the memory reference in
2451 @var{mode}. You should define this macro if the mode of the memory
2452 reference affects whether a register may be used as a base register. If
2453 you define this macro, the compiler will use it instead of
2454 @code{REGNO_OK_FOR_BASE_P}.
2457 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2458 A C expression which is nonzero if register number @var{num} is suitable for
2459 use as a base register in base plus index operand addresses, accessing
2460 memory in mode @var{mode}. It may be either a suitable hard register or a
2461 pseudo register that has been allocated such a hard register. You should
2462 define this macro if base plus index addresses have different requirements
2463 than other base register uses.
2466 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2467 A C expression which is nonzero if register number @var{num} is
2468 suitable for use as an index register in operand addresses. It may be
2469 either a suitable hard register or a pseudo register that has been
2470 allocated such a hard register.
2472 The difference between an index register and a base register is that
2473 the index register may be scaled. If an address involves the sum of
2474 two registers, neither one of them scaled, then either one may be
2475 labeled the ``base'' and the other the ``index''; but whichever
2476 labeling is used must fit the machine's constraints of which registers
2477 may serve in each capacity. The compiler will try both labelings,
2478 looking for one that is valid, and will reload one or both registers
2479 only if neither labeling works.
2482 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2483 A C expression that places additional restrictions on the register class
2484 to use when it is necessary to copy value @var{x} into a register in class
2485 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2486 another, smaller class. On many machines, the following definition is
2490 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2493 Sometimes returning a more restrictive class makes better code. For
2494 example, on the 68000, when @var{x} is an integer constant that is in range
2495 for a @samp{moveq} instruction, the value of this macro is always
2496 @code{DATA_REGS} as long as @var{class} includes the data registers.
2497 Requiring a data register guarantees that a @samp{moveq} will be used.
2499 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2500 @var{class} is if @var{x} is a legitimate constant which cannot be
2501 loaded into some register class. By returning @code{NO_REGS} you can
2502 force @var{x} into a memory location. For example, rs6000 can load
2503 immediate values into general-purpose registers, but does not have an
2504 instruction for loading an immediate value into a floating-point
2505 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2506 @var{x} is a floating-point constant. If the constant can't be loaded
2507 into any kind of register, code generation will be better if
2508 @code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2509 of using @code{PREFERRED_RELOAD_CLASS}.
2512 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2513 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2514 input reloads. If you don't define this macro, the default is to use
2515 @var{class}, unchanged.
2518 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2519 A C expression that places additional restrictions on the register class
2520 to use when it is necessary to be able to hold a value of mode
2521 @var{mode} in a reload register for which class @var{class} would
2524 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2525 there are certain modes that simply can't go in certain reload classes.
2527 The value is a register class; perhaps @var{class}, or perhaps another,
2530 Don't define this macro unless the target machine has limitations which
2531 require the macro to do something nontrivial.
2534 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2535 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2536 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2537 Many machines have some registers that cannot be copied directly to or
2538 from memory or even from other types of registers. An example is the
2539 @samp{MQ} register, which on most machines, can only be copied to or
2540 from general registers, but not memory. Some machines allow copying all
2541 registers to and from memory, but require a scratch register for stores
2542 to some memory locations (e.g., those with symbolic address on the RT,
2543 and those with certain symbolic address on the SPARC when compiling
2544 PIC)@. In some cases, both an intermediate and a scratch register are
2547 You should define these macros to indicate to the reload phase that it may
2548 need to allocate at least one register for a reload in addition to the
2549 register to contain the data. Specifically, if copying @var{x} to a
2550 register @var{class} in @var{mode} requires an intermediate register,
2551 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2552 largest register class all of whose registers can be used as
2553 intermediate registers or scratch registers.
2555 If copying a register @var{class} in @var{mode} to @var{x} requires an
2556 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2557 should be defined to return the largest register class required. If the
2558 requirements for input and output reloads are the same, the macro
2559 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2562 The values returned by these macros are often @code{GENERAL_REGS}.
2563 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2564 can be directly copied to or from a register of @var{class} in
2565 @var{mode} without requiring a scratch register. Do not define this
2566 macro if it would always return @code{NO_REGS}.
2568 If a scratch register is required (either with or without an
2569 intermediate register), you should define patterns for
2570 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2571 (@pxref{Standard Names}. These patterns, which will normally be
2572 implemented with a @code{define_expand}, should be similar to the
2573 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2576 Define constraints for the reload register and scratch register that
2577 contain a single register class. If the original reload register (whose
2578 class is @var{class}) can meet the constraint given in the pattern, the
2579 value returned by these macros is used for the class of the scratch
2580 register. Otherwise, two additional reload registers are required.
2581 Their classes are obtained from the constraints in the insn pattern.
2583 @var{x} might be a pseudo-register or a @code{subreg} of a
2584 pseudo-register, which could either be in a hard register or in memory.
2585 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2586 in memory and the hard register number if it is in a register.
2588 These macros should not be used in the case where a particular class of
2589 registers can only be copied to memory and not to another class of
2590 registers. In that case, secondary reload registers are not needed and
2591 would not be helpful. Instead, a stack location must be used to perform
2592 the copy and the @code{mov@var{m}} pattern should use memory as an
2593 intermediate storage. This case often occurs between floating-point and
2597 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2598 Certain machines have the property that some registers cannot be copied
2599 to some other registers without using memory. Define this macro on
2600 those machines to be a C expression that is nonzero if objects of mode
2601 @var{m} in registers of @var{class1} can only be copied to registers of
2602 class @var{class2} by storing a register of @var{class1} into memory
2603 and loading that memory location into a register of @var{class2}.
2605 Do not define this macro if its value would always be zero.
2608 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2609 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2610 allocates a stack slot for a memory location needed for register copies.
2611 If this macro is defined, the compiler instead uses the memory location
2612 defined by this macro.
2614 Do not define this macro if you do not define
2615 @code{SECONDARY_MEMORY_NEEDED}.
2618 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2619 When the compiler needs a secondary memory location to copy between two
2620 registers of mode @var{mode}, it normally allocates sufficient memory to
2621 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2622 load operations in a mode that many bits wide and whose class is the
2623 same as that of @var{mode}.
2625 This is right thing to do on most machines because it ensures that all
2626 bits of the register are copied and prevents accesses to the registers
2627 in a narrower mode, which some machines prohibit for floating-point
2630 However, this default behavior is not correct on some machines, such as
2631 the DEC Alpha, that store short integers in floating-point registers
2632 differently than in integer registers. On those machines, the default
2633 widening will not work correctly and you must define this macro to
2634 suppress that widening in some cases. See the file @file{alpha.h} for
2637 Do not define this macro if you do not define
2638 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2639 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2642 @defmac SMALL_REGISTER_CLASSES
2643 On some machines, it is risky to let hard registers live across arbitrary
2644 insns. Typically, these machines have instructions that require values
2645 to be in specific registers (like an accumulator), and reload will fail
2646 if the required hard register is used for another purpose across such an
2649 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2650 value on these machines. When this macro has a nonzero value, the
2651 compiler will try to minimize the lifetime of hard registers.
2653 It is always safe to define this macro with a nonzero value, but if you
2654 unnecessarily define it, you will reduce the amount of optimizations
2655 that can be performed in some cases. If you do not define this macro
2656 with a nonzero value when it is required, the compiler will run out of
2657 spill registers and print a fatal error message. For most machines, you
2658 should not define this macro at all.
2661 @defmac CLASS_LIKELY_SPILLED_P (@var{class})
2662 A C expression whose value is nonzero if pseudos that have been assigned
2663 to registers of class @var{class} would likely be spilled because
2664 registers of @var{class} are needed for spill registers.
2666 The default value of this macro returns 1 if @var{class} has exactly one
2667 register and zero otherwise. On most machines, this default should be
2668 used. Only define this macro to some other expression if pseudos
2669 allocated by @file{local-alloc.c} end up in memory because their hard
2670 registers were needed for spill registers. If this macro returns nonzero
2671 for those classes, those pseudos will only be allocated by
2672 @file{global.c}, which knows how to reallocate the pseudo to another
2673 register. If there would not be another register available for
2674 reallocation, you should not change the definition of this macro since
2675 the only effect of such a definition would be to slow down register
2679 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2680 A C expression for the maximum number of consecutive registers
2681 of class @var{class} needed to hold a value of mode @var{mode}.
2683 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2684 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2685 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2686 @var{mode})} for all @var{regno} values in the class @var{class}.
2688 This macro helps control the handling of multiple-word values
2692 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2693 If defined, a C expression that returns nonzero for a @var{class} for which
2694 a change from mode @var{from} to mode @var{to} is invalid.
2696 For the example, loading 32-bit integer or floating-point objects into
2697 floating-point registers on the Alpha extends them to 64 bits.
2698 Therefore loading a 64-bit object and then storing it as a 32-bit object
2699 does not store the low-order 32 bits, as would be the case for a normal
2700 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2704 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2705 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2706 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2710 Three other special macros describe which operands fit which constraint
2713 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2714 A C expression that defines the machine-dependent operand constraint
2715 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2716 particular ranges of integer values. If @var{c} is one of those
2717 letters, the expression should check that @var{value}, an integer, is in
2718 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2719 not one of those letters, the value should be 0 regardless of
2723 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2724 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2725 string passed in @var{str}, so that you can use suffixes to distinguish
2726 between different variants.
2729 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2730 A C expression that defines the machine-dependent operand constraint
2731 letters that specify particular ranges of @code{const_double} values
2732 (@samp{G} or @samp{H}).
2734 If @var{c} is one of those letters, the expression should check that
2735 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2736 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2737 letters, the value should be 0 regardless of @var{value}.
2739 @code{const_double} is used for all floating-point constants and for
2740 @code{DImode} fixed-point constants. A given letter can accept either
2741 or both kinds of values. It can use @code{GET_MODE} to distinguish
2742 between these kinds.
2745 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2746 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2747 string passed in @var{str}, so that you can use suffixes to distinguish
2748 between different variants.
2751 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
2752 A C expression that defines the optional machine-dependent constraint
2753 letters that can be used to segregate specific types of operands, usually
2754 memory references, for the target machine. Any letter that is not
2755 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2756 @code{REG_CLASS_FROM_CONSTRAINT}
2757 may be used. Normally this macro will not be defined.
2759 If it is required for a particular target machine, it should return 1
2760 if @var{value} corresponds to the operand type represented by the
2761 constraint letter @var{c}. If @var{c} is not defined as an extra
2762 constraint, the value returned should be 0 regardless of @var{value}.
2764 For example, on the ROMP, load instructions cannot have their output
2765 in r0 if the memory reference contains a symbolic address. Constraint
2766 letter @samp{Q} is defined as representing a memory address that does
2767 @emph{not} contain a symbolic address. An alternative is specified with
2768 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2769 alternative specifies @samp{m} on the input and a register class that
2770 does not include r0 on the output.
2773 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2774 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2775 in @var{str}, so that you can use suffixes to distinguish between different
2779 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2780 A C expression that defines the optional machine-dependent constraint
2781 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2782 be treated like memory constraints by the reload pass.
2784 It should return 1 if the operand type represented by the constraint
2785 at the start of @var{str}, the first letter of which is the letter @var{c},
2786 comprises a subset of all memory references including
2787 all those whose address is simply a base register. This allows the reload
2788 pass to reload an operand, if it does not directly correspond to the operand
2789 type of @var{c}, by copying its address into a base register.
2791 For example, on the S/390, some instructions do not accept arbitrary
2792 memory references, but only those that do not make use of an index
2793 register. The constraint letter @samp{Q} is defined via
2794 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2795 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2796 a @samp{Q} constraint can handle any memory operand, because the
2797 reload pass knows it can be reloaded by copying the memory address
2798 into a base register if required. This is analogous to the way
2799 a @samp{o} constraint can handle any memory operand.
2802 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2803 A C expression that defines the optional machine-dependent constraint
2804 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2805 @code{EXTRA_CONSTRAINT_STR}, that should
2806 be treated like address constraints by the reload pass.
2808 It should return 1 if the operand type represented by the constraint
2809 at the start of @var{str}, which starts with the letter @var{c}, comprises
2810 a subset of all memory addresses including
2811 all those that consist of just a base register. This allows the reload
2812 pass to reload an operand, if it does not directly correspond to the operand
2813 type of @var{str}, by copying it into a base register.
2815 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2816 be used with the @code{address_operand} predicate. It is treated
2817 analogously to the @samp{p} constraint.
2820 @node Stack and Calling
2821 @section Stack Layout and Calling Conventions
2822 @cindex calling conventions
2824 @c prevent bad page break with this line
2825 This describes the stack layout and calling conventions.
2829 * Exception Handling::
2834 * Register Arguments::
2836 * Aggregate Return::
2844 @subsection Basic Stack Layout
2845 @cindex stack frame layout
2846 @cindex frame layout
2848 @c prevent bad page break with this line
2849 Here is the basic stack layout.
2851 @defmac STACK_GROWS_DOWNWARD
2852 Define this macro if pushing a word onto the stack moves the stack
2853 pointer to a smaller address.
2855 When we say, ``define this macro if @dots{}'', it means that the
2856 compiler checks this macro only with @code{#ifdef} so the precise
2857 definition used does not matter.
2860 @defmac STACK_PUSH_CODE
2861 This macro defines the operation used when something is pushed
2862 on the stack. In RTL, a push operation will be
2863 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2865 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2866 and @code{POST_INC}. Which of these is correct depends on
2867 the stack direction and on whether the stack pointer points
2868 to the last item on the stack or whether it points to the
2869 space for the next item on the stack.
2871 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2872 defined, which is almost always right, and @code{PRE_INC} otherwise,
2873 which is often wrong.
2876 @defmac FRAME_GROWS_DOWNWARD
2877 Define this macro if the addresses of local variable slots are at negative
2878 offsets from the frame pointer.
2881 @defmac ARGS_GROW_DOWNWARD
2882 Define this macro if successive arguments to a function occupy decreasing
2883 addresses on the stack.
2886 @defmac STARTING_FRAME_OFFSET
2887 Offset from the frame pointer to the first local variable slot to be allocated.
2889 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2890 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2891 Otherwise, it is found by adding the length of the first slot to the
2892 value @code{STARTING_FRAME_OFFSET}.
2893 @c i'm not sure if the above is still correct.. had to change it to get
2894 @c rid of an overfull. --mew 2feb93
2897 @defmac STACK_ALIGNMENT_NEEDED
2898 Define to zero to disable final alignment of the stack during reload.
2899 The nonzero default for this macro is suitable for most ports.
2901 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
2902 is a register save block following the local block that doesn't require
2903 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2904 stack alignment and do it in the backend.
2907 @defmac STACK_POINTER_OFFSET
2908 Offset from the stack pointer register to the first location at which
2909 outgoing arguments are placed. If not specified, the default value of
2910 zero is used. This is the proper value for most machines.
2912 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2913 the first location at which outgoing arguments are placed.
2916 @defmac FIRST_PARM_OFFSET (@var{fundecl})
2917 Offset from the argument pointer register to the first argument's
2918 address. On some machines it may depend on the data type of the
2921 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2922 the first argument's address.
2925 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
2926 Offset from the stack pointer register to an item dynamically allocated
2927 on the stack, e.g., by @code{alloca}.
2929 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2930 length of the outgoing arguments. The default is correct for most
2931 machines. See @file{function.c} for details.
2934 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2935 A C expression whose value is RTL representing the address in a stack
2936 frame where the pointer to the caller's frame is stored. Assume that
2937 @var{frameaddr} is an RTL expression for the address of the stack frame
2940 If you don't define this macro, the default is to return the value
2941 of @var{frameaddr}---that is, the stack frame address is also the
2942 address of the stack word that points to the previous frame.
2945 @defmac SETUP_FRAME_ADDRESSES
2946 If defined, a C expression that produces the machine-specific code to
2947 setup the stack so that arbitrary frames can be accessed. For example,
2948 on the SPARC, we must flush all of the register windows to the stack
2949 before we can access arbitrary stack frames. You will seldom need to
2953 @deftypefn {Target Hook} bool TARGET_BUILTIN_SETJMP_FRAME_VALUE ()
2954 This target hook should return an rtx that is used to store
2955 the address of the current frame into the built in @code{setjmp} buffer.
2956 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2957 machines. One reason you may need to define this target hook is if
2958 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2961 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2962 A C expression whose value is RTL representing the value of the return
2963 address for the frame @var{count} steps up from the current frame, after
2964 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2965 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2966 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2968 The value of the expression must always be the correct address when
2969 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2970 determine the return address of other frames.
2973 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
2974 Define this if the return address of a particular stack frame is accessed
2975 from the frame pointer of the previous stack frame.
2978 @defmac INCOMING_RETURN_ADDR_RTX
2979 A C expression whose value is RTL representing the location of the
2980 incoming return address at the beginning of any function, before the
2981 prologue. This RTL is either a @code{REG}, indicating that the return
2982 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2985 You only need to define this macro if you want to support call frame
2986 debugging information like that provided by DWARF 2.
2988 If this RTL is a @code{REG}, you should also define
2989 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2992 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
2993 A C expression whose value is an integer giving a DWARF 2 column
2994 number that may be used as an alternate return column. This should
2995 be defined only if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
2996 general register, but an alternate column needs to be used for
3000 @defmac DWARF_ZERO_REG
3001 A C expression whose value is an integer giving a DWARF 2 register
3002 number that is considered to always have the value zero. This should
3003 only be defined if the target has an architected zero register, and
3004 someone decided it was a good idea to use that register number to
3005 terminate the stack backtrace. New ports should avoid this.
3008 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3009 This target hook allows the backend to emit frame-related insns that
3010 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3011 info engine will invoke it on insns of the form
3013 (set (reg) (unspec [...] UNSPEC_INDEX))
3017 (set (reg) (unspec_volatile [...] UNSPECV_INDEX)).
3019 to let the backend emit the call frame instructions. @var{label} is
3020 the CFI label attached to the insn, @var{pattern} is the pattern of
3021 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3024 @defmac INCOMING_FRAME_SP_OFFSET
3025 A C expression whose value is an integer giving the offset, in bytes,
3026 from the value of the stack pointer register to the top of the stack
3027 frame at the beginning of any function, before the prologue. The top of
3028 the frame is defined to be the value of the stack pointer in the
3029 previous frame, just before the call instruction.
3031 You only need to define this macro if you want to support call frame
3032 debugging information like that provided by DWARF 2.
3035 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3036 A C expression whose value is an integer giving the offset, in bytes,
3037 from the argument pointer to the canonical frame address (cfa). The
3038 final value should coincide with that calculated by
3039 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3040 during virtual register instantiation.
3042 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
3043 which is correct for most machines; in general, the arguments are found
3044 immediately before the stack frame. Note that this is not the case on
3045 some targets that save registers into the caller's frame, such as SPARC
3046 and rs6000, and so such targets need to define this macro.
3048 You only need to define this macro if the default is incorrect, and you
3049 want to support call frame debugging information like that provided by
3053 @node Exception Handling
3054 @subsection Exception Handling Support
3055 @cindex exception handling
3057 @defmac EH_RETURN_DATA_REGNO (@var{N})
3058 A C expression whose value is the @var{N}th register number used for
3059 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3060 @var{N} registers are usable.
3062 The exception handling library routines communicate with the exception
3063 handlers via a set of agreed upon registers. Ideally these registers
3064 should be call-clobbered; it is possible to use call-saved registers,
3065 but may negatively impact code size. The target must support at least
3066 2 data registers, but should define 4 if there are enough free registers.
3068 You must define this macro if you want to support call frame exception
3069 handling like that provided by DWARF 2.
3072 @defmac EH_RETURN_STACKADJ_RTX
3073 A C expression whose value is RTL representing a location in which
3074 to store a stack adjustment to be applied before function return.
3075 This is used to unwind the stack to an exception handler's call frame.
3076 It will be assigned zero on code paths that return normally.
3078 Typically this is a call-clobbered hard register that is otherwise
3079 untouched by the epilogue, but could also be a stack slot.
3081 Do not define this macro if the stack pointer is saved and restored
3082 by the regular prolog and epilog code in the call frame itself; in
3083 this case, the exception handling library routines will update the
3084 stack location to be restored in place. Otherwise, you must define
3085 this macro if you want to support call frame exception handling like
3086 that provided by DWARF 2.
3089 @defmac EH_RETURN_HANDLER_RTX
3090 A C expression whose value is RTL representing a location in which
3091 to store the address of an exception handler to which we should
3092 return. It will not be assigned on code paths that return normally.
3094 Typically this is the location in the call frame at which the normal
3095 return address is stored. For targets that return by popping an
3096 address off the stack, this might be a memory address just below
3097 the @emph{target} call frame rather than inside the current call
3098 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3099 been assigned, so it may be used to calculate the location of the
3102 Some targets have more complex requirements than storing to an
3103 address calculable during initial code generation. In that case
3104 the @code{eh_return} instruction pattern should be used instead.
3106 If you want to support call frame exception handling, you must
3107 define either this macro or the @code{eh_return} instruction pattern.
3110 @defmac RETURN_ADDR_OFFSET
3111 If defined, an integer-valued C expression for which rtl will be generated
3112 to add it to the exception handler address before it is searched in the
3113 exception handling tables, and to subtract it again from the address before
3114 using it to return to the exception handler.
3117 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3118 This macro chooses the encoding of pointers embedded in the exception
3119 handling sections. If at all possible, this should be defined such
3120 that the exception handling section will not require dynamic relocations,
3121 and so may be read-only.
3123 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3124 @var{global} is true if the symbol may be affected by dynamic relocations.
3125 The macro should return a combination of the @code{DW_EH_PE_*} defines
3126 as found in @file{dwarf2.h}.
3128 If this macro is not defined, pointers will not be encoded but
3129 represented directly.
3132 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3133 This macro allows the target to emit whatever special magic is required
3134 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3135 Generic code takes care of pc-relative and indirect encodings; this must
3136 be defined if the target uses text-relative or data-relative encodings.
3138 This is a C statement that branches to @var{done} if the format was
3139 handled. @var{encoding} is the format chosen, @var{size} is the number
3140 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3144 @defmac MD_UNWIND_SUPPORT
3145 A string specifying a file to be #include'd in unwind-dw2.c. The file
3146 so included typically defines @code{MD_FALLBACK_FRAME_STATE_FOR}.
3149 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3150 This macro allows the target to add cpu and operating system specific
3151 code to the call-frame unwinder for use when there is no unwind data
3152 available. The most common reason to implement this macro is to unwind
3153 through signal frames.
3155 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
3156 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3157 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3158 for the address of the code being executed and @code{context->cfa} for
3159 the stack pointer value. If the frame can be decoded, the register save
3160 addresses should be updated in @var{fs} and the macro should evaluate to
3161 @code{_URC_NO_REASON}. If the frame cannot be decoded, the macro should
3162 evaluate to @code{_URC_END_OF_STACK}.
3164 For proper signal handling in Java this macro is accompanied by
3165 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3168 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3169 This macro allows the target to add operating system specific code to the
3170 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3171 usually used for signal or interrupt frames.
3173 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3174 @var{context} is an @code{_Unwind_Context};
3175 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3176 for the abi and context in the @code{.unwabi} directive. If the
3177 @code{.unwabi} directive can be handled, the register save addresses should
3178 be updated in @var{fs}.
3181 @defmac TARGET_USES_WEAK_UNWIND_INFO
3182 A C expression that evaluates to true if the target requires unwind
3183 info to be given comdat linkage. Define it to be @code{1} if comdat
3184 linkage is necessary. The default is @code{0}.
3187 @node Stack Checking
3188 @subsection Specifying How Stack Checking is Done
3190 GCC will check that stack references are within the boundaries of
3191 the stack, if the @option{-fstack-check} is specified, in one of three ways:
3195 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3196 will assume that you have arranged for stack checking to be done at
3197 appropriate places in the configuration files, e.g., in
3198 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
3202 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
3203 called @code{check_stack} in your @file{md} file, GCC will call that
3204 pattern with one argument which is the address to compare the stack
3205 value against. You must arrange for this pattern to report an error if
3206 the stack pointer is out of range.
3209 If neither of the above are true, GCC will generate code to periodically
3210 ``probe'' the stack pointer using the values of the macros defined below.
3213 Normally, you will use the default values of these macros, so GCC
3214 will use the third approach.
3216 @defmac STACK_CHECK_BUILTIN
3217 A nonzero value if stack checking is done by the configuration files in a
3218 machine-dependent manner. You should define this macro if stack checking
3219 is require by the ABI of your machine or if you would like to have to stack
3220 checking in some more efficient way than GCC's portable approach.
3221 The default value of this macro is zero.
3224 @defmac STACK_CHECK_PROBE_INTERVAL
3225 An integer representing the interval at which GCC must generate stack
3226 probe instructions. You will normally define this macro to be no larger
3227 than the size of the ``guard pages'' at the end of a stack area. The
3228 default value of 4096 is suitable for most systems.
3231 @defmac STACK_CHECK_PROBE_LOAD
3232 A integer which is nonzero if GCC should perform the stack probe
3233 as a load instruction and zero if GCC should use a store instruction.
3234 The default is zero, which is the most efficient choice on most systems.
3237 @defmac STACK_CHECK_PROTECT
3238 The number of bytes of stack needed to recover from a stack overflow,
3239 for languages where such a recovery is supported. The default value of
3240 75 words should be adequate for most machines.
3243 @defmac STACK_CHECK_MAX_FRAME_SIZE
3244 The maximum size of a stack frame, in bytes. GCC will generate probe
3245 instructions in non-leaf functions to ensure at least this many bytes of
3246 stack are available. If a stack frame is larger than this size, stack
3247 checking will not be reliable and GCC will issue a warning. The
3248 default is chosen so that GCC only generates one instruction on most
3249 systems. You should normally not change the default value of this macro.
3252 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3253 GCC uses this value to generate the above warning message. It
3254 represents the amount of fixed frame used by a function, not including
3255 space for any callee-saved registers, temporaries and user variables.
3256 You need only specify an upper bound for this amount and will normally
3257 use the default of four words.
3260 @defmac STACK_CHECK_MAX_VAR_SIZE
3261 The maximum size, in bytes, of an object that GCC will place in the
3262 fixed area of the stack frame when the user specifies
3263 @option{-fstack-check}.
3264 GCC computed the default from the values of the above macros and you will
3265 normally not need to override that default.
3269 @node Frame Registers
3270 @subsection Registers That Address the Stack Frame
3272 @c prevent bad page break with this line
3273 This discusses registers that address the stack frame.
3275 @defmac STACK_POINTER_REGNUM
3276 The register number of the stack pointer register, which must also be a
3277 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3278 the hardware determines which register this is.
3281 @defmac FRAME_POINTER_REGNUM
3282 The register number of the frame pointer register, which is used to
3283 access automatic variables in the stack frame. On some machines, the
3284 hardware determines which register this is. On other machines, you can
3285 choose any register you wish for this purpose.
3288 @defmac HARD_FRAME_POINTER_REGNUM
3289 On some machines the offset between the frame pointer and starting
3290 offset of the automatic variables is not known until after register
3291 allocation has been done (for example, because the saved registers are
3292 between these two locations). On those machines, define
3293 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3294 be used internally until the offset is known, and define
3295 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3296 used for the frame pointer.
3298 You should define this macro only in the very rare circumstances when it
3299 is not possible to calculate the offset between the frame pointer and
3300 the automatic variables until after register allocation has been
3301 completed. When this macro is defined, you must also indicate in your
3302 definition of @code{ELIMINABLE_REGS} how to eliminate
3303 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3304 or @code{STACK_POINTER_REGNUM}.
3306 Do not define this macro if it would be the same as
3307 @code{FRAME_POINTER_REGNUM}.
3310 @defmac ARG_POINTER_REGNUM
3311 The register number of the arg pointer register, which is used to access
3312 the function's argument list. On some machines, this is the same as the
3313 frame pointer register. On some machines, the hardware determines which
3314 register this is. On other machines, you can choose any register you
3315 wish for this purpose. If this is not the same register as the frame
3316 pointer register, then you must mark it as a fixed register according to
3317 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3318 (@pxref{Elimination}).
3321 @defmac RETURN_ADDRESS_POINTER_REGNUM
3322 The register number of the return address pointer register, which is used to
3323 access the current function's return address from the stack. On some
3324 machines, the return address is not at a fixed offset from the frame
3325 pointer or stack pointer or argument pointer. This register can be defined
3326 to point to the return address on the stack, and then be converted by
3327 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3329 Do not define this macro unless there is no other way to get the return
3330 address from the stack.
3333 @defmac STATIC_CHAIN_REGNUM
3334 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3335 Register numbers used for passing a function's static chain pointer. If
3336 register windows are used, the register number as seen by the called
3337 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3338 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3339 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3342 The static chain register need not be a fixed register.
3344 If the static chain is passed in memory, these macros should not be
3345 defined; instead, the next two macros should be defined.
3348 @defmac STATIC_CHAIN
3349 @defmacx STATIC_CHAIN_INCOMING
3350 If the static chain is passed in memory, these macros provide rtx giving
3351 @code{mem} expressions that denote where they are stored.
3352 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3353 as seen by the calling and called functions, respectively. Often the former
3354 will be at an offset from the stack pointer and the latter at an offset from
3357 @findex stack_pointer_rtx
3358 @findex frame_pointer_rtx
3359 @findex arg_pointer_rtx
3360 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3361 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3362 macros and should be used to refer to those items.
3364 If the static chain is passed in a register, the two previous macros should
3368 @defmac DWARF_FRAME_REGISTERS
3369 This macro specifies the maximum number of hard registers that can be
3370 saved in a call frame. This is used to size data structures used in
3371 DWARF2 exception handling.
3373 Prior to GCC 3.0, this macro was needed in order to establish a stable
3374 exception handling ABI in the face of adding new hard registers for ISA
3375 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3376 in the number of hard registers. Nevertheless, this macro can still be
3377 used to reduce the runtime memory requirements of the exception handling
3378 routines, which can be substantial if the ISA contains a lot of
3379 registers that are not call-saved.
3381 If this macro is not defined, it defaults to
3382 @code{FIRST_PSEUDO_REGISTER}.
3385 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3387 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3388 for backward compatibility in pre GCC 3.0 compiled code.
3390 If this macro is not defined, it defaults to
3391 @code{DWARF_FRAME_REGISTERS}.
3394 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3396 Define this macro if the target's representation for dwarf registers
3397 is different than the internal representation for unwind column.
3398 Given a dwarf register, this macro should return the internal unwind
3399 column number to use instead.
3401 See the PowerPC's SPE target for an example.
3404 @defmac DWARF_FRAME_REGNUM (@var{regno})
3406 Define this macro if the target's representation for dwarf registers
3407 used in .eh_frame or .debug_frame is different from that used in other
3408 debug info sections. Given a GCC hard register number, this macro
3409 should return the .eh_frame register number. The default is
3410 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3414 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3416 Define this macro to map register numbers held in the call frame info
3417 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3418 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3419 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3420 return @code{@var{regno}}.
3425 @subsection Eliminating Frame Pointer and Arg Pointer
3427 @c prevent bad page break with this line
3428 This is about eliminating the frame pointer and arg pointer.
3430 @defmac FRAME_POINTER_REQUIRED
3431 A C expression which is nonzero if a function must have and use a frame
3432 pointer. This expression is evaluated in the reload pass. If its value is
3433 nonzero the function will have a frame pointer.
3435 The expression can in principle examine the current function and decide
3436 according to the facts, but on most machines the constant 0 or the
3437 constant 1 suffices. Use 0 when the machine allows code to be generated
3438 with no frame pointer, and doing so saves some time or space. Use 1
3439 when there is no possible advantage to avoiding a frame pointer.
3441 In certain cases, the compiler does not know how to produce valid code
3442 without a frame pointer. The compiler recognizes those cases and
3443 automatically gives the function a frame pointer regardless of what
3444 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3447 In a function that does not require a frame pointer, the frame pointer
3448 register can be allocated for ordinary usage, unless you mark it as a
3449 fixed register. See @code{FIXED_REGISTERS} for more information.
3452 @findex get_frame_size
3453 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3454 A C statement to store in the variable @var{depth-var} the difference
3455 between the frame pointer and the stack pointer values immediately after
3456 the function prologue. The value would be computed from information
3457 such as the result of @code{get_frame_size ()} and the tables of
3458 registers @code{regs_ever_live} and @code{call_used_regs}.
3460 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3461 need not be defined. Otherwise, it must be defined even if
3462 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3463 case, you may set @var{depth-var} to anything.
3466 @defmac ELIMINABLE_REGS
3467 If defined, this macro specifies a table of register pairs used to
3468 eliminate unneeded registers that point into the stack frame. If it is not
3469 defined, the only elimination attempted by the compiler is to replace
3470 references to the frame pointer with references to the stack pointer.
3472 The definition of this macro is a list of structure initializations, each
3473 of which specifies an original and replacement register.
3475 On some machines, the position of the argument pointer is not known until
3476 the compilation is completed. In such a case, a separate hard register
3477 must be used for the argument pointer. This register can be eliminated by
3478 replacing it with either the frame pointer or the argument pointer,
3479 depending on whether or not the frame pointer has been eliminated.
3481 In this case, you might specify:
3483 #define ELIMINABLE_REGS \
3484 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3485 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3486 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3489 Note that the elimination of the argument pointer with the stack pointer is
3490 specified first since that is the preferred elimination.
3493 @defmac CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3494 A C expression that returns nonzero if the compiler is allowed to try
3495 to replace register number @var{from-reg} with register number
3496 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3497 is defined, and will usually be the constant 1, since most of the cases
3498 preventing register elimination are things that the compiler already
3502 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3503 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3504 specifies the initial difference between the specified pair of
3505 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3509 @node Stack Arguments
3510 @subsection Passing Function Arguments on the Stack
3511 @cindex arguments on stack
3512 @cindex stack arguments
3514 The macros in this section control how arguments are passed
3515 on the stack. See the following section for other macros that
3516 control passing certain arguments in registers.
3518 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (tree @var{fntype})
3519 This target hook returns @code{true} if an argument declared in a
3520 prototype as an integral type smaller than @code{int} should actually be
3521 passed as an @code{int}. In addition to avoiding errors in certain
3522 cases of mismatch, it also makes for better code on certain machines.
3523 The default is to not promote prototypes.
3527 A C expression. If nonzero, push insns will be used to pass
3529 If the target machine does not have a push instruction, set it to zero.
3530 That directs GCC to use an alternate strategy: to
3531 allocate the entire argument block and then store the arguments into
3532 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3535 @defmac PUSH_ARGS_REVERSED
3536 A C expression. If nonzero, function arguments will be evaluated from
3537 last to first, rather than from first to last. If this macro is not
3538 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3539 and args grow in opposite directions, and 0 otherwise.
3542 @defmac PUSH_ROUNDING (@var{npushed})
3543 A C expression that is the number of bytes actually pushed onto the
3544 stack when an instruction attempts to push @var{npushed} bytes.
3546 On some machines, the definition
3549 #define PUSH_ROUNDING(BYTES) (BYTES)
3553 will suffice. But on other machines, instructions that appear
3554 to push one byte actually push two bytes in an attempt to maintain
3555 alignment. Then the definition should be
3558 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3562 @findex current_function_outgoing_args_size
3563 @defmac ACCUMULATE_OUTGOING_ARGS
3564 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3565 will be computed and placed into the variable
3566 @code{current_function_outgoing_args_size}. No space will be pushed
3567 onto the stack for each call; instead, the function prologue should
3568 increase the stack frame size by this amount.
3570 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3574 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3575 Define this macro if functions should assume that stack space has been
3576 allocated for arguments even when their values are passed in
3579 The value of this macro is the size, in bytes, of the area reserved for
3580 arguments passed in registers for the function represented by @var{fndecl},
3581 which can be zero if GCC is calling a library function.
3583 This space can be allocated by the caller, or be a part of the
3584 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3587 @c above is overfull. not sure what to do. --mew 5feb93 did
3588 @c something, not sure if it looks good. --mew 10feb93
3590 @defmac OUTGOING_REG_PARM_STACK_SPACE
3591 Define this if it is the responsibility of the caller to allocate the area
3592 reserved for arguments passed in registers.
3594 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3595 whether the space for these arguments counts in the value of
3596 @code{current_function_outgoing_args_size}.
3599 @defmac STACK_PARMS_IN_REG_PARM_AREA
3600 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3601 stack parameters don't skip the area specified by it.
3602 @c i changed this, makes more sens and it should have taken care of the
3603 @c overfull.. not as specific, tho. --mew 5feb93
3605 Normally, when a parameter is not passed in registers, it is placed on the
3606 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3607 suppresses this behavior and causes the parameter to be passed on the
3608 stack in its natural location.
3611 @defmac RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3612 A C expression that should indicate the number of bytes of its own
3613 arguments that a function pops on returning, or 0 if the
3614 function pops no arguments and the caller must therefore pop them all
3615 after the function returns.
3617 @var{fundecl} is a C variable whose value is a tree node that describes
3618 the function in question. Normally it is a node of type
3619 @code{FUNCTION_DECL} that describes the declaration of the function.
3620 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3622 @var{funtype} is a C variable whose value is a tree node that
3623 describes the function in question. Normally it is a node of type
3624 @code{FUNCTION_TYPE} that describes the data type of the function.
3625 From this it is possible to obtain the data types of the value and
3626 arguments (if known).
3628 When a call to a library function is being considered, @var{fundecl}
3629 will contain an identifier node for the library function. Thus, if
3630 you need to distinguish among various library functions, you can do so
3631 by their names. Note that ``library function'' in this context means
3632 a function used to perform arithmetic, whose name is known specially
3633 in the compiler and was not mentioned in the C code being compiled.
3635 @var{stack-size} is the number of bytes of arguments passed on the
3636 stack. If a variable number of bytes is passed, it is zero, and
3637 argument popping will always be the responsibility of the calling function.
3639 On the VAX, all functions always pop their arguments, so the definition
3640 of this macro is @var{stack-size}. On the 68000, using the standard
3641 calling convention, no functions pop their arguments, so the value of
3642 the macro is always 0 in this case. But an alternative calling
3643 convention is available in which functions that take a fixed number of
3644 arguments pop them but other functions (such as @code{printf}) pop
3645 nothing (the caller pops all). When this convention is in use,
3646 @var{funtype} is examined to determine whether a function takes a fixed
3647 number of arguments.
3650 @defmac CALL_POPS_ARGS (@var{cum})
3651 A C expression that should indicate the number of bytes a call sequence
3652 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3653 when compiling a function call.
3655 @var{cum} is the variable in which all arguments to the called function
3656 have been accumulated.
3658 On certain architectures, such as the SH5, a call trampoline is used
3659 that pops certain registers off the stack, depending on the arguments
3660 that have been passed to the function. Since this is a property of the
3661 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3665 @node Register Arguments
3666 @subsection Passing Arguments in Registers
3667 @cindex arguments in registers
3668 @cindex registers arguments
3670 This section describes the macros which let you control how various
3671 types of arguments are passed in registers or how they are arranged in
3674 @defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3675 A C expression that controls whether a function argument is passed
3676 in a register, and which register.
3678 The arguments are @var{cum}, which summarizes all the previous
3679 arguments; @var{mode}, the machine mode of the argument; @var{type},
3680 the data type of the argument as a tree node or 0 if that is not known
3681 (which happens for C support library functions); and @var{named},
3682 which is 1 for an ordinary argument and 0 for nameless arguments that
3683 correspond to @samp{@dots{}} in the called function's prototype.
3684 @var{type} can be an incomplete type if a syntax error has previously
3687 The value of the expression is usually either a @code{reg} RTX for the
3688 hard register in which to pass the argument, or zero to pass the
3689 argument on the stack.
3691 For machines like the VAX and 68000, where normally all arguments are
3692 pushed, zero suffices as a definition.
3694 The value of the expression can also be a @code{parallel} RTX@. This is
3695 used when an argument is passed in multiple locations. The mode of the
3696 @code{parallel} should be the mode of the entire argument. The
3697 @code{parallel} holds any number of @code{expr_list} pairs; each one
3698 describes where part of the argument is passed. In each
3699 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3700 register in which to pass this part of the argument, and the mode of the
3701 register RTX indicates how large this part of the argument is. The
3702 second operand of the @code{expr_list} is a @code{const_int} which gives
3703 the offset in bytes into the entire argument of where this part starts.
3704 As a special exception the first @code{expr_list} in the @code{parallel}
3705 RTX may have a first operand of zero. This indicates that the entire
3706 argument is also stored on the stack.
3708 The last time this macro is called, it is called with @code{MODE ==
3709 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3710 pattern as operands 2 and 3 respectively.
3712 @cindex @file{stdarg.h} and register arguments
3713 The usual way to make the ISO library @file{stdarg.h} work on a machine
3714 where some arguments are usually passed in registers, is to cause
3715 nameless arguments to be passed on the stack instead. This is done
3716 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3718 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3719 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3720 You may use the hook @code{targetm.calls.must_pass_in_stack}
3721 in the definition of this macro to determine if this argument is of a
3722 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3723 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3724 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3725 defined, the argument will be computed in the stack and then loaded into
3729 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, tree @var{type})
3730 This target hook should return @code{true} if we should not pass @var{type}
3731 solely in registers. The file @file{expr.h} defines a
3732 definition that is usually appropriate, refer to @file{expr.h} for additional
3736 @defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3737 Define this macro if the target machine has ``register windows'', so
3738 that the register in which a function sees an arguments is not
3739 necessarily the same as the one in which the caller passed the
3742 For such machines, @code{FUNCTION_ARG} computes the register in which
3743 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3744 be defined in a similar fashion to tell the function being called
3745 where the arguments will arrive.
3747 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3748 serves both purposes.
3751 @defmac FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
3752 A C expression for the number of words, at the beginning of an
3753 argument, that must be put in registers. The value must be zero for
3754 arguments that are passed entirely in registers or that are entirely
3755 pushed on the stack.
3757 On some machines, certain arguments must be passed partially in
3758 registers and partially in memory. On these machines, typically the
3759 first @var{n} words of arguments are passed in registers, and the rest
3760 on the stack. If a multi-word argument (a @code{double} or a
3761 structure) crosses that boundary, its first few words must be passed
3762 in registers and the rest must be pushed. This macro tells the
3763 compiler when this occurs, and how many of the words should go in
3766 @code{FUNCTION_ARG} for these arguments should return the first
3767 register to be used by the caller for this argument; likewise
3768 @code{FUNCTION_INCOMING_ARG}, for the called function.
3771 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3772 This target hook should return @code{true} if an argument at the
3773 position indicated by @var{cum} should be passed by reference. This
3774 predicate is queried after target independent reasons for being
3775 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
3777 If the hook returns true, a copy of that argument is made in memory and a
3778 pointer to the argument is passed instead of the argument itself.
3779 The pointer is passed in whatever way is appropriate for passing a pointer
3783 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
3784 The function argument described by the parameters to this hook is
3785 known to be passed by reference. The hook should return true if the
3786 function argument should be copied by the callee instead of copied
3789 For any argument for which the hook returns true, if it can be
3790 determined that the argument is not modified, then a copy need
3793 The default version of this hook always returns false.
3796 @defmac CUMULATIVE_ARGS
3797 A C type for declaring a variable that is used as the first argument of
3798 @code{FUNCTION_ARG} and other related values. For some target machines,
3799 the type @code{int} suffices and can hold the number of bytes of
3802 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3803 arguments that have been passed on the stack. The compiler has other
3804 variables to keep track of that. For target machines on which all
3805 arguments are passed on the stack, there is no need to store anything in
3806 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3807 should not be empty, so use @code{int}.
3810 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
3811 A C statement (sans semicolon) for initializing the variable
3812 @var{cum} for the state at the beginning of the argument list. The
3813 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
3814 is the tree node for the data type of the function which will receive
3815 the args, or 0 if the args are to a compiler support library function.
3816 For direct calls that are not libcalls, @var{fndecl} contain the
3817 declaration node of the function. @var{fndecl} is also set when
3818 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3819 being compiled. @var{n_named_args} is set to the number of named
3820 arguments, including a structure return address if it is passed as a
3821 parameter, when making a call. When processing incoming arguments,
3822 @var{n_named_args} is set to @minus{}1.
3824 When processing a call to a compiler support library function,
3825 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3826 contains the name of the function, as a string. @var{libname} is 0 when
3827 an ordinary C function call is being processed. Thus, each time this
3828 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3829 never both of them at once.
3832 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3833 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3834 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3835 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3836 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3837 0)} is used instead.
3840 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3841 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3842 finding the arguments for the function being compiled. If this macro is
3843 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3845 The value passed for @var{libname} is always 0, since library routines
3846 with special calling conventions are never compiled with GCC@. The
3847 argument @var{libname} exists for symmetry with
3848 @code{INIT_CUMULATIVE_ARGS}.
3849 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3850 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3853 @defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3854 A C statement (sans semicolon) to update the summarizer variable
3855 @var{cum} to advance past an argument in the argument list. The
3856 values @var{mode}, @var{type} and @var{named} describe that argument.
3857 Once this is done, the variable @var{cum} is suitable for analyzing
3858 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
3860 This macro need not do anything if the argument in question was passed
3861 on the stack. The compiler knows how to track the amount of stack space
3862 used for arguments without any special help.
3865 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3866 If defined, a C expression which determines whether, and in which direction,
3867 to pad out an argument with extra space. The value should be of type
3868 @code{enum direction}: either @code{upward} to pad above the argument,
3869 @code{downward} to pad below, or @code{none} to inhibit padding.
3871 The @emph{amount} of padding is always just enough to reach the next
3872 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3875 This macro has a default definition which is right for most systems.
3876 For little-endian machines, the default is to pad upward. For
3877 big-endian machines, the default is to pad downward for an argument of
3878 constant size shorter than an @code{int}, and upward otherwise.
3881 @defmac PAD_VARARGS_DOWN
3882 If defined, a C expression which determines whether the default
3883 implementation of va_arg will attempt to pad down before reading the
3884 next argument, if that argument is smaller than its aligned space as
3885 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3886 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3889 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
3890 Specify padding for the last element of a block move between registers and
3891 memory. @var{first} is nonzero if this is the only element. Defining this
3892 macro allows better control of register function parameters on big-endian
3893 machines, without using @code{PARALLEL} rtl. In particular,
3894 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
3895 registers, as there is no longer a "wrong" part of a register; For example,
3896 a three byte aggregate may be passed in the high part of a register if so
3900 @defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3901 If defined, a C expression that gives the alignment boundary, in bits,
3902 of an argument with the specified mode and type. If it is not defined,
3903 @code{PARM_BOUNDARY} is used for all arguments.
3906 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
3907 A C expression that is nonzero if @var{regno} is the number of a hard
3908 register in which function arguments are sometimes passed. This does
3909 @emph{not} include implicit arguments such as the static chain and
3910 the structure-value address. On many machines, no registers can be
3911 used for this purpose since all function arguments are pushed on the
3915 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (tree @var{type})
3916 This hook should return true if parameter of type @var{type} are passed
3917 as two scalar parameters. By default, GCC will attempt to pack complex
3918 arguments into the target's word size. Some ABIs require complex arguments
3919 to be split and treated as their individual components. For example, on
3920 AIX64, complex floats should be passed in a pair of floating point
3921 registers, even though a complex float would fit in one 64-bit floating
3924 The default value of this hook is @code{NULL}, which is treated as always
3928 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
3929 This hook returns a type node for @code{va_list} for the target.
3930 The default version of the hook returns @code{void*}.
3933 @deftypefn {Target Hook} tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree @var{valist}, tree @var{type}, tree *@var{pre_p}, tree *@var{post_p})
3934 This hook performs target-specific gimplification of
3935 @code{VA_ARG_EXPR}. The first two parameters correspond to the
3936 arguments to @code{va_arg}; the latter two are as in
3937 @code{gimplify.c:gimplify_expr}.
3940 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (enum machine_mode @var{mode})
3941 Define this to return nonzero if the port can handle pointers
3942 with machine mode @var{mode}. The default version of this
3943 hook returns true for both @code{ptr_mode} and @code{Pmode}.
3946 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
3947 Define this to return nonzero if the port is prepared to handle
3948 insns involving scalar mode @var{mode}. For a scalar mode to be
3949 considered supported, all the basic arithmetic and comparisons
3952 The default version of this hook returns true for any mode
3953 required to handle the basic C types (as defined by the port).
3954 Included here are the double-word arithmetic supported by the
3955 code in @file{optabs.c}.
3958 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
3959 Define this to return nonzero if the port is prepared to handle
3960 insns involving vector mode @var{mode}. At the very least, it
3961 must have move patterns for this mode.
3965 @subsection How Scalar Function Values Are Returned
3966 @cindex return values in registers
3967 @cindex values, returned by functions
3968 @cindex scalars, returned as values
3970 This section discusses the macros that control returning scalars as
3971 values---values that can fit in registers.
3973 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
3974 A C expression to create an RTX representing the place where a
3975 function returns a value of data type @var{valtype}. @var{valtype} is
3976 a tree node representing a data type. Write @code{TYPE_MODE
3977 (@var{valtype})} to get the machine mode used to represent that type.
3978 On many machines, only the mode is relevant. (Actually, on most
3979 machines, scalar values are returned in the same place regardless of
3982 The value of the expression is usually a @code{reg} RTX for the hard
3983 register where the return value is stored. The value can also be a
3984 @code{parallel} RTX, if the return value is in multiple places. See
3985 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3987 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply the same
3988 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3991 If the precise function being called is known, @var{func} is a tree
3992 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3993 pointer. This makes it possible to use a different value-returning
3994 convention for specific functions when all their calls are
3997 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3998 types, because these are returned in another way. See
3999 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4002 @defmac FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
4003 Define this macro if the target machine has ``register windows''
4004 so that the register in which a function returns its value is not
4005 the same as the one in which the caller sees the value.
4007 For such machines, @code{FUNCTION_VALUE} computes the register in which
4008 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
4009 defined in a similar fashion to tell the function where to put the
4012 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
4013 @code{FUNCTION_VALUE} serves both purposes.
4015 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
4016 aggregate data types, because these are returned in another way. See
4017 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4020 @defmac LIBCALL_VALUE (@var{mode})
4021 A C expression to create an RTX representing the place where a library
4022 function returns a value of mode @var{mode}. If the precise function
4023 being called is known, @var{func} is a tree node
4024 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4025 pointer. This makes it possible to use a different value-returning
4026 convention for specific functions when all their calls are
4029 Note that ``library function'' in this context means a compiler
4030 support routine, used to perform arithmetic, whose name is known
4031 specially by the compiler and was not mentioned in the C code being
4034 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
4035 data types, because none of the library functions returns such types.
4038 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4039 A C expression that is nonzero if @var{regno} is the number of a hard
4040 register in which the values of called function may come back.
4042 A register whose use for returning values is limited to serving as the
4043 second of a pair (for a value of type @code{double}, say) need not be
4044 recognized by this macro. So for most machines, this definition
4048 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4051 If the machine has register windows, so that the caller and the called
4052 function use different registers for the return value, this macro
4053 should recognize only the caller's register numbers.
4056 @defmac APPLY_RESULT_SIZE
4057 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4058 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4059 saving and restoring an arbitrary return value.
4062 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (tree @var{type})
4063 This hook should return true if values of type @var{type} are returned
4064 at the most significant end of a register (in other words, if they are
4065 padded at the least significant end). You can assume that @var{type}
4066 is returned in a register; the caller is required to check this.
4068 Note that the register provided by @code{FUNCTION_VALUE} must be able
4069 to hold the complete return value. For example, if a 1-, 2- or 3-byte
4070 structure is returned at the most significant end of a 4-byte register,
4071 @code{FUNCTION_VALUE} should provide an @code{SImode} rtx.
4074 @node Aggregate Return
4075 @subsection How Large Values Are Returned
4076 @cindex aggregates as return values
4077 @cindex large return values
4078 @cindex returning aggregate values
4079 @cindex structure value address
4081 When a function value's mode is @code{BLKmode} (and in some other
4082 cases), the value is not returned according to @code{FUNCTION_VALUE}
4083 (@pxref{Scalar Return}). Instead, the caller passes the address of a
4084 block of memory in which the value should be stored. This address
4085 is called the @dfn{structure value address}.
4087 This section describes how to control returning structure values in
4090 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (tree @var{type}, tree @var{fntype})
4091 This target hook should return a nonzero value to say to return the
4092 function value in memory, just as large structures are always returned.
4093 Here @var{type} will be the data type of the value, and @var{fntype}
4094 will be the type of the function doing the returning, or @code{NULL} for
4097 Note that values of mode @code{BLKmode} must be explicitly handled
4098 by this function. Also, the option @option{-fpcc-struct-return}
4099 takes effect regardless of this macro. On most systems, it is
4100 possible to leave the hook undefined; this causes a default
4101 definition to be used, whose value is the constant 1 for @code{BLKmode}
4102 values, and 0 otherwise.
4104 Do not use this hook to indicate that structures and unions should always
4105 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4109 @defmac DEFAULT_PCC_STRUCT_RETURN
4110 Define this macro to be 1 if all structure and union return values must be
4111 in memory. Since this results in slower code, this should be defined
4112 only if needed for compatibility with other compilers or with an ABI@.
4113 If you define this macro to be 0, then the conventions used for structure
4114 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4117 If not defined, this defaults to the value 1.
4120 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4121 This target hook should return the location of the structure value
4122 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4123 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4124 be @code{NULL}, for libcalls. You do not need to define this target
4125 hook if the address is always passed as an ``invisible'' first
4128 On some architectures the place where the structure value address
4129 is found by the called function is not the same place that the
4130 caller put it. This can be due to register windows, or it could
4131 be because the function prologue moves it to a different place.
4132 @var{incoming} is @code{true} when the location is needed in
4133 the context of the called function, and @code{false} in the context of
4136 If @var{incoming} is @code{true} and the address is to be found on the
4137 stack, return a @code{mem} which refers to the frame pointer.
4140 @defmac PCC_STATIC_STRUCT_RETURN
4141 Define this macro if the usual system convention on the target machine
4142 for returning structures and unions is for the called function to return
4143 the address of a static variable containing the value.
4145 Do not define this if the usual system convention is for the caller to
4146 pass an address to the subroutine.
4148 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4149 nothing when you use @option{-freg-struct-return} mode.
4153 @subsection Caller-Saves Register Allocation
4155 If you enable it, GCC can save registers around function calls. This
4156 makes it possible to use call-clobbered registers to hold variables that
4157 must live across calls.
4159 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4160 A C expression to determine whether it is worthwhile to consider placing
4161 a pseudo-register in a call-clobbered hard register and saving and
4162 restoring it around each function call. The expression should be 1 when
4163 this is worth doing, and 0 otherwise.
4165 If you don't define this macro, a default is used which is good on most
4166 machines: @code{4 * @var{calls} < @var{refs}}.
4169 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4170 A C expression specifying which mode is required for saving @var{nregs}
4171 of a pseudo-register in call-clobbered hard register @var{regno}. If
4172 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4173 returned. For most machines this macro need not be defined since GCC
4174 will select the smallest suitable mode.
4177 @node Function Entry
4178 @subsection Function Entry and Exit
4179 @cindex function entry and exit
4183 This section describes the macros that output function entry
4184 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4186 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4187 If defined, a function that outputs the assembler code for entry to a
4188 function. The prologue is responsible for setting up the stack frame,
4189 initializing the frame pointer register, saving registers that must be
4190 saved, and allocating @var{size} additional bytes of storage for the
4191 local variables. @var{size} is an integer. @var{file} is a stdio
4192 stream to which the assembler code should be output.
4194 The label for the beginning of the function need not be output by this
4195 macro. That has already been done when the macro is run.
4197 @findex regs_ever_live
4198 To determine which registers to save, the macro can refer to the array
4199 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4200 @var{r} is used anywhere within the function. This implies the function
4201 prologue should save register @var{r}, provided it is not one of the
4202 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4203 @code{regs_ever_live}.)
4205 On machines that have ``register windows'', the function entry code does
4206 not save on the stack the registers that are in the windows, even if
4207 they are supposed to be preserved by function calls; instead it takes
4208 appropriate steps to ``push'' the register stack, if any non-call-used
4209 registers are used in the function.
4211 @findex frame_pointer_needed
4212 On machines where functions may or may not have frame-pointers, the
4213 function entry code must vary accordingly; it must set up the frame
4214 pointer if one is wanted, and not otherwise. To determine whether a
4215 frame pointer is in wanted, the macro can refer to the variable
4216 @code{frame_pointer_needed}. The variable's value will be 1 at run
4217 time in a function that needs a frame pointer. @xref{Elimination}.
4219 The function entry code is responsible for allocating any stack space
4220 required for the function. This stack space consists of the regions
4221 listed below. In most cases, these regions are allocated in the
4222 order listed, with the last listed region closest to the top of the
4223 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4224 the highest address if it is not defined). You can use a different order
4225 for a machine if doing so is more convenient or required for
4226 compatibility reasons. Except in cases where required by standard
4227 or by a debugger, there is no reason why the stack layout used by GCC
4228 need agree with that used by other compilers for a machine.
4231 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4232 If defined, a function that outputs assembler code at the end of a
4233 prologue. This should be used when the function prologue is being
4234 emitted as RTL, and you have some extra assembler that needs to be
4235 emitted. @xref{prologue instruction pattern}.
4238 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4239 If defined, a function that outputs assembler code at the start of an
4240 epilogue. This should be used when the function epilogue is being
4241 emitted as RTL, and you have some extra assembler that needs to be
4242 emitted. @xref{epilogue instruction pattern}.
4245 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4246 If defined, a function that outputs the assembler code for exit from a
4247 function. The epilogue is responsible for restoring the saved
4248 registers and stack pointer to their values when the function was
4249 called, and returning control to the caller. This macro takes the
4250 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4251 registers to restore are determined from @code{regs_ever_live} and
4252 @code{CALL_USED_REGISTERS} in the same way.
4254 On some machines, there is a single instruction that does all the work
4255 of returning from the function. On these machines, give that
4256 instruction the name @samp{return} and do not define the macro
4257 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4259 Do not define a pattern named @samp{return} if you want the
4260 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4261 switches to control whether return instructions or epilogues are used,
4262 define a @samp{return} pattern with a validity condition that tests the
4263 target switches appropriately. If the @samp{return} pattern's validity
4264 condition is false, epilogues will be used.
4266 On machines where functions may or may not have frame-pointers, the
4267 function exit code must vary accordingly. Sometimes the code for these
4268 two cases is completely different. To determine whether a frame pointer
4269 is wanted, the macro can refer to the variable
4270 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4271 a function that needs a frame pointer.
4273 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4274 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4275 The C variable @code{current_function_is_leaf} is nonzero for such a
4276 function. @xref{Leaf Functions}.
4278 On some machines, some functions pop their arguments on exit while
4279 others leave that for the caller to do. For example, the 68020 when
4280 given @option{-mrtd} pops arguments in functions that take a fixed
4281 number of arguments.
4283 @findex current_function_pops_args
4284 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4285 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4286 needs to know what was decided. The variable that is called
4287 @code{current_function_pops_args} is the number of bytes of its
4288 arguments that a function should pop. @xref{Scalar Return}.
4289 @c what is the "its arguments" in the above sentence referring to, pray
4290 @c tell? --mew 5feb93
4295 @findex current_function_pretend_args_size
4296 A region of @code{current_function_pretend_args_size} bytes of
4297 uninitialized space just underneath the first argument arriving on the
4298 stack. (This may not be at the very start of the allocated stack region
4299 if the calling sequence has pushed anything else since pushing the stack
4300 arguments. But usually, on such machines, nothing else has been pushed
4301 yet, because the function prologue itself does all the pushing.) This
4302 region is used on machines where an argument may be passed partly in
4303 registers and partly in memory, and, in some cases to support the
4304 features in @code{<stdarg.h>}.
4307 An area of memory used to save certain registers used by the function.
4308 The size of this area, which may also include space for such things as
4309 the return address and pointers to previous stack frames, is
4310 machine-specific and usually depends on which registers have been used
4311 in the function. Machines with register windows often do not require
4315 A region of at least @var{size} bytes, possibly rounded up to an allocation
4316 boundary, to contain the local variables of the function. On some machines,
4317 this region and the save area may occur in the opposite order, with the
4318 save area closer to the top of the stack.
4321 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4322 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4323 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4324 argument lists of the function. @xref{Stack Arguments}.
4327 @defmac EXIT_IGNORE_STACK
4328 Define this macro as a C expression that is nonzero if the return
4329 instruction or the function epilogue ignores the value of the stack
4330 pointer; in other words, if it is safe to delete an instruction to
4331 adjust the stack pointer before a return from the function. The
4334 Note that this macro's value is relevant only for functions for which
4335 frame pointers are maintained. It is never safe to delete a final
4336 stack adjustment in a function that has no frame pointer, and the
4337 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4340 @defmac EPILOGUE_USES (@var{regno})
4341 Define this macro as a C expression that is nonzero for registers that are
4342 used by the epilogue or the @samp{return} pattern. The stack and frame
4343 pointer registers are already be assumed to be used as needed.
4346 @defmac EH_USES (@var{regno})
4347 Define this macro as a C expression that is nonzero for registers that are
4348 used by the exception handling mechanism, and so should be considered live
4349 on entry to an exception edge.
4352 @defmac DELAY_SLOTS_FOR_EPILOGUE
4353 Define this macro if the function epilogue contains delay slots to which
4354 instructions from the rest of the function can be ``moved''. The
4355 definition should be a C expression whose value is an integer
4356 representing the number of delay slots there.
4359 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4360 A C expression that returns 1 if @var{insn} can be placed in delay
4361 slot number @var{n} of the epilogue.
4363 The argument @var{n} is an integer which identifies the delay slot now
4364 being considered (since different slots may have different rules of
4365 eligibility). It is never negative and is always less than the number
4366 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4367 If you reject a particular insn for a given delay slot, in principle, it
4368 may be reconsidered for a subsequent delay slot. Also, other insns may
4369 (at least in principle) be considered for the so far unfilled delay
4372 @findex current_function_epilogue_delay_list
4373 @findex final_scan_insn
4374 The insns accepted to fill the epilogue delay slots are put in an RTL
4375 list made with @code{insn_list} objects, stored in the variable
4376 @code{current_function_epilogue_delay_list}. The insn for the first
4377 delay slot comes first in the list. Your definition of the macro
4378 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4379 outputting the insns in this list, usually by calling
4380 @code{final_scan_insn}.
4382 You need not define this macro if you did not define
4383 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4386 @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})
4387 A function that outputs the assembler code for a thunk
4388 function, used to implement C++ virtual function calls with multiple
4389 inheritance. The thunk acts as a wrapper around a virtual function,
4390 adjusting the implicit object parameter before handing control off to
4393 First, emit code to add the integer @var{delta} to the location that
4394 contains the incoming first argument. Assume that this argument
4395 contains a pointer, and is the one used to pass the @code{this} pointer
4396 in C++. This is the incoming argument @emph{before} the function prologue,
4397 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4398 all other incoming arguments.
4400 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4401 made after adding @code{delta}. In particular, if @var{p} is the
4402 adjusted pointer, the following adjustment should be made:
4405 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4408 After the additions, emit code to jump to @var{function}, which is a
4409 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4410 not touch the return address. Hence returning from @var{FUNCTION} will
4411 return to whoever called the current @samp{thunk}.
4413 The effect must be as if @var{function} had been called directly with
4414 the adjusted first argument. This macro is responsible for emitting all
4415 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4416 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4418 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4419 have already been extracted from it.) It might possibly be useful on
4420 some targets, but probably not.
4422 If you do not define this macro, the target-independent code in the C++
4423 front end will generate a less efficient heavyweight thunk that calls
4424 @var{function} instead of jumping to it. The generic approach does
4425 not support varargs.
4428 @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})
4429 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4430 to output the assembler code for the thunk function specified by the
4431 arguments it is passed, and false otherwise. In the latter case, the
4432 generic approach will be used by the C++ front end, with the limitations
4437 @subsection Generating Code for Profiling
4438 @cindex profiling, code generation
4440 These macros will help you generate code for profiling.
4442 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4443 A C statement or compound statement to output to @var{file} some
4444 assembler code to call the profiling subroutine @code{mcount}.
4447 The details of how @code{mcount} expects to be called are determined by
4448 your operating system environment, not by GCC@. To figure them out,
4449 compile a small program for profiling using the system's installed C
4450 compiler and look at the assembler code that results.
4452 Older implementations of @code{mcount} expect the address of a counter
4453 variable to be loaded into some register. The name of this variable is
4454 @samp{LP} followed by the number @var{labelno}, so you would generate
4455 the name using @samp{LP%d} in a @code{fprintf}.
4458 @defmac PROFILE_HOOK
4459 A C statement or compound statement to output to @var{file} some assembly
4460 code to call the profiling subroutine @code{mcount} even the target does
4461 not support profiling.
4464 @defmac NO_PROFILE_COUNTERS
4465 Define this macro if the @code{mcount} subroutine on your system does
4466 not need a counter variable allocated for each function. This is true
4467 for almost all modern implementations. If you define this macro, you
4468 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
4471 @defmac PROFILE_BEFORE_PROLOGUE
4472 Define this macro if the code for function profiling should come before
4473 the function prologue. Normally, the profiling code comes after.
4477 @subsection Permitting tail calls
4480 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4481 True if it is ok to do sibling call optimization for the specified
4482 call expression @var{exp}. @var{decl} will be the called function,
4483 or @code{NULL} if this is an indirect call.
4485 It is not uncommon for limitations of calling conventions to prevent
4486 tail calls to functions outside the current unit of translation, or
4487 during PIC compilation. The hook is used to enforce these restrictions,
4488 as the @code{sibcall} md pattern can not fail, or fall over to a
4489 ``normal'' call. The criteria for successful sibling call optimization
4490 may vary greatly between different architectures.
4494 @section Implementing the Varargs Macros
4495 @cindex varargs implementation
4497 GCC comes with an implementation of @code{<varargs.h>} and
4498 @code{<stdarg.h>} that work without change on machines that pass arguments
4499 on the stack. Other machines require their own implementations of
4500 varargs, and the two machine independent header files must have
4501 conditionals to include it.
4503 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4504 the calling convention for @code{va_start}. The traditional
4505 implementation takes just one argument, which is the variable in which
4506 to store the argument pointer. The ISO implementation of
4507 @code{va_start} takes an additional second argument. The user is
4508 supposed to write the last named argument of the function here.
4510 However, @code{va_start} should not use this argument. The way to find
4511 the end of the named arguments is with the built-in functions described
4514 @defmac __builtin_saveregs ()
4515 Use this built-in function to save the argument registers in memory so
4516 that the varargs mechanism can access them. Both ISO and traditional
4517 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4518 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4520 On some machines, @code{__builtin_saveregs} is open-coded under the
4521 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4522 other machines, it calls a routine written in assembler language,
4523 found in @file{libgcc2.c}.
4525 Code generated for the call to @code{__builtin_saveregs} appears at the
4526 beginning of the function, as opposed to where the call to
4527 @code{__builtin_saveregs} is written, regardless of what the code is.
4528 This is because the registers must be saved before the function starts
4529 to use them for its own purposes.
4530 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4534 @defmac __builtin_args_info (@var{category})
4535 Use this built-in function to find the first anonymous arguments in
4538 In general, a machine may have several categories of registers used for
4539 arguments, each for a particular category of data types. (For example,
4540 on some machines, floating-point registers are used for floating-point
4541 arguments while other arguments are passed in the general registers.)
4542 To make non-varargs functions use the proper calling convention, you
4543 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4544 registers in each category have been used so far
4546 @code{__builtin_args_info} accesses the same data structure of type
4547 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4548 with it, with @var{category} specifying which word to access. Thus, the
4549 value indicates the first unused register in a given category.
4551 Normally, you would use @code{__builtin_args_info} in the implementation
4552 of @code{va_start}, accessing each category just once and storing the
4553 value in the @code{va_list} object. This is because @code{va_list} will
4554 have to update the values, and there is no way to alter the
4555 values accessed by @code{__builtin_args_info}.
4558 @defmac __builtin_next_arg (@var{lastarg})
4559 This is the equivalent of @code{__builtin_args_info}, for stack
4560 arguments. It returns the address of the first anonymous stack
4561 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4562 returns the address of the location above the first anonymous stack
4563 argument. Use it in @code{va_start} to initialize the pointer for
4564 fetching arguments from the stack. Also use it in @code{va_start} to
4565 verify that the second parameter @var{lastarg} is the last named argument
4566 of the current function.
4569 @defmac __builtin_classify_type (@var{object})
4570 Since each machine has its own conventions for which data types are
4571 passed in which kind of register, your implementation of @code{va_arg}
4572 has to embody these conventions. The easiest way to categorize the
4573 specified data type is to use @code{__builtin_classify_type} together
4574 with @code{sizeof} and @code{__alignof__}.
4576 @code{__builtin_classify_type} ignores the value of @var{object},
4577 considering only its data type. It returns an integer describing what
4578 kind of type that is---integer, floating, pointer, structure, and so on.
4580 The file @file{typeclass.h} defines an enumeration that you can use to
4581 interpret the values of @code{__builtin_classify_type}.
4584 These machine description macros help implement varargs:
4586 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
4587 If defined, this hook produces the machine-specific code for a call to
4588 @code{__builtin_saveregs}. This code will be moved to the very
4589 beginning of the function, before any parameter access are made. The
4590 return value of this function should be an RTX that contains the value
4591 to use as the return of @code{__builtin_saveregs}.
4594 @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})
4595 This target hook offers an alternative to using
4596 @code{__builtin_saveregs} and defining the hook
4597 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
4598 register arguments into the stack so that all the arguments appear to
4599 have been passed consecutively on the stack. Once this is done, you can
4600 use the standard implementation of varargs that works for machines that
4601 pass all their arguments on the stack.
4603 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
4604 structure, containing the values that are obtained after processing the
4605 named arguments. The arguments @var{mode} and @var{type} describe the
4606 last named argument---its machine mode and its data type as a tree node.
4608 The target hook should do two things: first, push onto the stack all the
4609 argument registers @emph{not} used for the named arguments, and second,
4610 store the size of the data thus pushed into the @code{int}-valued
4611 variable pointed to by @var{pretend_args_size}. The value that you
4612 store here will serve as additional offset for setting up the stack
4615 Because you must generate code to push the anonymous arguments at
4616 compile time without knowing their data types,
4617 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
4618 have just a single category of argument register and use it uniformly
4621 If the argument @var{second_time} is nonzero, it means that the
4622 arguments of the function are being analyzed for the second time. This
4623 happens for an inline function, which is not actually compiled until the
4624 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
4625 not generate any instructions in this case.
4628 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca})
4629 Define this hook to return @code{true} if the location where a function
4630 argument is passed depends on whether or not it is a named argument.
4632 This hook controls how the @var{named} argument to @code{FUNCTION_ARG}
4633 is set for varargs and stdarg functions. If this hook returns
4634 @code{true}, the @var{named} argument is always true for named
4635 arguments, and false for unnamed arguments. If it returns @code{false},
4636 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
4637 then all arguments are treated as named. Otherwise, all named arguments
4638 except the last are treated as named.
4640 You need not define this hook if it always returns zero.
4643 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED
4644 If you need to conditionally change ABIs so that one works with
4645 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
4646 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
4647 defined, then define this hook to return @code{true} if
4648 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
4649 Otherwise, you should not define this hook.
4653 @section Trampolines for Nested Functions
4654 @cindex trampolines for nested functions
4655 @cindex nested functions, trampolines for
4657 A @dfn{trampoline} is a small piece of code that is created at run time
4658 when the address of a nested function is taken. It normally resides on
4659 the stack, in the stack frame of the containing function. These macros
4660 tell GCC how to generate code to allocate and initialize a
4663 The instructions in the trampoline must do two things: load a constant
4664 address into the static chain register, and jump to the real address of
4665 the nested function. On CISC machines such as the m68k, this requires
4666 two instructions, a move immediate and a jump. Then the two addresses
4667 exist in the trampoline as word-long immediate operands. On RISC
4668 machines, it is often necessary to load each address into a register in
4669 two parts. Then pieces of each address form separate immediate
4672 The code generated to initialize the trampoline must store the variable
4673 parts---the static chain value and the function address---into the
4674 immediate operands of the instructions. On a CISC machine, this is
4675 simply a matter of copying each address to a memory reference at the
4676 proper offset from the start of the trampoline. On a RISC machine, it
4677 may be necessary to take out pieces of the address and store them
4680 @defmac TRAMPOLINE_TEMPLATE (@var{file})
4681 A C statement to output, on the stream @var{file}, assembler code for a
4682 block of data that contains the constant parts of a trampoline. This
4683 code should not include a label---the label is taken care of
4686 If you do not define this macro, it means no template is needed
4687 for the target. Do not define this macro on systems where the block move
4688 code to copy the trampoline into place would be larger than the code
4689 to generate it on the spot.
4692 @defmac TRAMPOLINE_SECTION
4693 The name of a subroutine to switch to the section in which the
4694 trampoline template is to be placed (@pxref{Sections}). The default is
4695 a value of @samp{readonly_data_section}, which places the trampoline in
4696 the section containing read-only data.
4699 @defmac TRAMPOLINE_SIZE
4700 A C expression for the size in bytes of the trampoline, as an integer.
4703 @defmac TRAMPOLINE_ALIGNMENT
4704 Alignment required for trampolines, in bits.
4706 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4707 is used for aligning trampolines.
4710 @defmac INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4711 A C statement to initialize the variable parts of a trampoline.
4712 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4713 an RTX for the address of the nested function; @var{static_chain} is an
4714 RTX for the static chain value that should be passed to the function
4718 @defmac TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4719 A C statement that should perform any machine-specific adjustment in
4720 the address of the trampoline. Its argument contains the address that
4721 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4722 used for a function call should be different from the address in which
4723 the template was stored, the different address should be assigned to
4724 @var{addr}. If this macro is not defined, @var{addr} will be used for
4727 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4728 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4729 If this macro is not defined, by default the trampoline is allocated as
4730 a stack slot. This default is right for most machines. The exceptions
4731 are machines where it is impossible to execute instructions in the stack
4732 area. On such machines, you may have to implement a separate stack,
4733 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4734 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4736 @var{fp} points to a data structure, a @code{struct function}, which
4737 describes the compilation status of the immediate containing function of
4738 the function which the trampoline is for. The stack slot for the
4739 trampoline is in the stack frame of this containing function. Other
4740 allocation strategies probably must do something analogous with this
4744 Implementing trampolines is difficult on many machines because they have
4745 separate instruction and data caches. Writing into a stack location
4746 fails to clear the memory in the instruction cache, so when the program
4747 jumps to that location, it executes the old contents.
4749 Here are two possible solutions. One is to clear the relevant parts of
4750 the instruction cache whenever a trampoline is set up. The other is to
4751 make all trampolines identical, by having them jump to a standard
4752 subroutine. The former technique makes trampoline execution faster; the
4753 latter makes initialization faster.
4755 To clear the instruction cache when a trampoline is initialized, define
4756 the following macro.
4758 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
4759 If defined, expands to a C expression clearing the @emph{instruction
4760 cache} in the specified interval. The definition of this macro would
4761 typically be a series of @code{asm} statements. Both @var{beg} and
4762 @var{end} are both pointer expressions.
4765 The operating system may also require the stack to be made executable
4766 before calling the trampoline. To implement this requirement, define
4767 the following macro.
4769 @defmac ENABLE_EXECUTE_STACK
4770 Define this macro if certain operations must be performed before executing
4771 code located on the stack. The macro should expand to a series of C
4772 file-scope constructs (e.g.@: functions) and provide a unique entry point
4773 named @code{__enable_execute_stack}. The target is responsible for
4774 emitting calls to the entry point in the code, for example from the
4775 @code{INITIALIZE_TRAMPOLINE} macro.
4778 To use a standard subroutine, define the following macro. In addition,
4779 you must make sure that the instructions in a trampoline fill an entire
4780 cache line with identical instructions, or else ensure that the
4781 beginning of the trampoline code is always aligned at the same point in
4782 its cache line. Look in @file{m68k.h} as a guide.
4784 @defmac TRANSFER_FROM_TRAMPOLINE
4785 Define this macro if trampolines need a special subroutine to do their
4786 work. The macro should expand to a series of @code{asm} statements
4787 which will be compiled with GCC@. They go in a library function named
4788 @code{__transfer_from_trampoline}.
4790 If you need to avoid executing the ordinary prologue code of a compiled
4791 C function when you jump to the subroutine, you can do so by placing a
4792 special label of your own in the assembler code. Use one @code{asm}
4793 statement to generate an assembler label, and another to make the label
4794 global. Then trampolines can use that label to jump directly to your
4795 special assembler code.
4799 @section Implicit Calls to Library Routines
4800 @cindex library subroutine names
4801 @cindex @file{libgcc.a}
4803 @c prevent bad page break with this line
4804 Here is an explanation of implicit calls to library routines.
4806 @defmac DECLARE_LIBRARY_RENAMES
4807 This macro, if defined, should expand to a piece of C code that will get
4808 expanded when compiling functions for libgcc.a. It can be used to
4809 provide alternate names for GCC's internal library functions if there
4810 are ABI-mandated names that the compiler should provide.
4813 @findex init_one_libfunc
4814 @findex set_optab_libfunc
4815 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
4816 This hook should declare additional library routines or rename
4817 existing ones, using the functions @code{set_optab_libfunc} and
4818 @code{init_one_libfunc} defined in @file{optabs.c}.
4819 @code{init_optabs} calls this macro after initializing all the normal
4822 The default is to do nothing. Most ports don't need to define this hook.
4825 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4826 This macro should return @code{true} if the library routine that
4827 implements the floating point comparison operator @var{comparison} in
4828 mode @var{mode} will return a boolean, and @var{false} if it will
4831 GCC's own floating point libraries return tristates from the
4832 comparison operators, so the default returns false always. Most ports
4833 don't need to define this macro.
4836 @defmac TARGET_LIB_INT_CMP_BIASED
4837 This macro should evaluate to @code{true} if the integer comparison
4838 functions (like @code{__cmpdi2}) return 0 to indicate that the first
4839 operand is smaller than the second, 1 to indicate that they are equal,
4840 and 2 to indicate that the first operand is greater than the second.
4841 If this macro evalutes to @code{false} the comparison functions return
4842 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
4843 in @file{libgcc.a}, you do not need to define this macro.
4846 @cindex US Software GOFAST, floating point emulation library
4847 @cindex floating point emulation library, US Software GOFAST
4848 @cindex GOFAST, floating point emulation library
4849 @findex gofast_maybe_init_libfuncs
4850 @defmac US_SOFTWARE_GOFAST
4851 Define this macro if your system C library uses the US Software GOFAST
4852 library to provide floating point emulation.
4854 In addition to defining this macro, your architecture must set
4855 @code{TARGET_INIT_LIBFUNCS} to @code{gofast_maybe_init_libfuncs}, or
4856 else call that function from its version of that hook. It is defined
4857 in @file{config/gofast.h}, which must be included by your
4858 architecture's @file{@var{cpu}.c} file. See @file{sparc/sparc.c} for
4861 If this macro is defined, the
4862 @code{TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL} target hook must return
4863 false for @code{SFmode} and @code{DFmode} comparisons.
4866 @cindex @code{EDOM}, implicit usage
4869 The value of @code{EDOM} on the target machine, as a C integer constant
4870 expression. If you don't define this macro, GCC does not attempt to
4871 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4872 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4875 If you do not define @code{TARGET_EDOM}, then compiled code reports
4876 domain errors by calling the library function and letting it report the
4877 error. If mathematical functions on your system use @code{matherr} when
4878 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4879 that @code{matherr} is used normally.
4882 @cindex @code{errno}, implicit usage
4883 @defmac GEN_ERRNO_RTX
4884 Define this macro as a C expression to create an rtl expression that
4885 refers to the global ``variable'' @code{errno}. (On certain systems,
4886 @code{errno} may not actually be a variable.) If you don't define this
4887 macro, a reasonable default is used.
4890 @cindex C99 math functions, implicit usage
4891 @defmac TARGET_C99_FUNCTIONS
4892 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
4893 @code{sinf} and similarly for other functions defined by C99 standard. The
4894 default is nonzero that should be proper value for most modern systems, however
4895 number of existing systems lacks support for these functions in the runtime so
4896 they needs this macro to be redefined to 0.
4899 @defmac NEXT_OBJC_RUNTIME
4900 Define this macro to generate code for Objective-C message sending using
4901 the calling convention of the NeXT system. This calling convention
4902 involves passing the object, the selector and the method arguments all
4903 at once to the method-lookup library function.
4905 The default calling convention passes just the object and the selector
4906 to the lookup function, which returns a pointer to the method.
4909 @node Addressing Modes
4910 @section Addressing Modes
4911 @cindex addressing modes
4913 @c prevent bad page break with this line
4914 This is about addressing modes.
4916 @defmac HAVE_PRE_INCREMENT
4917 @defmacx HAVE_PRE_DECREMENT
4918 @defmacx HAVE_POST_INCREMENT
4919 @defmacx HAVE_POST_DECREMENT
4920 A C expression that is nonzero if the machine supports pre-increment,
4921 pre-decrement, post-increment, or post-decrement addressing respectively.
4924 @defmac HAVE_PRE_MODIFY_DISP
4925 @defmacx HAVE_POST_MODIFY_DISP
4926 A C expression that is nonzero if the machine supports pre- or
4927 post-address side-effect generation involving constants other than
4928 the size of the memory operand.
4931 @defmac HAVE_PRE_MODIFY_REG
4932 @defmacx HAVE_POST_MODIFY_REG
4933 A C expression that is nonzero if the machine supports pre- or
4934 post-address side-effect generation involving a register displacement.
4937 @defmac CONSTANT_ADDRESS_P (@var{x})
4938 A C expression that is 1 if the RTX @var{x} is a constant which
4939 is a valid address. On most machines, this can be defined as
4940 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4941 in which constant addresses are supported.
4944 @defmac CONSTANT_P (@var{x})
4945 @code{CONSTANT_P}, which is defined by target-independent code,
4946 accepts integer-values expressions whose values are not explicitly
4947 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
4948 expressions and @code{const} arithmetic expressions, in addition to
4949 @code{const_int} and @code{const_double} expressions.
4952 @defmac MAX_REGS_PER_ADDRESS
4953 A number, the maximum number of registers that can appear in a valid
4954 memory address. Note that it is up to you to specify a value equal to
4955 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4959 @defmac GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4960 A C compound statement with a conditional @code{goto @var{label};}
4961 executed if @var{x} (an RTX) is a legitimate memory address on the
4962 target machine for a memory operand of mode @var{mode}.
4964 It usually pays to define several simpler macros to serve as
4965 subroutines for this one. Otherwise it may be too complicated to
4968 This macro must exist in two variants: a strict variant and a
4969 non-strict one. The strict variant is used in the reload pass. It
4970 must be defined so that any pseudo-register that has not been
4971 allocated a hard register is considered a memory reference. In
4972 contexts where some kind of register is required, a pseudo-register
4973 with no hard register must be rejected.
4975 The non-strict variant is used in other passes. It must be defined to
4976 accept all pseudo-registers in every context where some kind of
4977 register is required.
4979 @findex REG_OK_STRICT
4980 Compiler source files that want to use the strict variant of this
4981 macro define the macro @code{REG_OK_STRICT}. You should use an
4982 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4983 in that case and the non-strict variant otherwise.
4985 Subroutines to check for acceptable registers for various purposes (one
4986 for base registers, one for index registers, and so on) are typically
4987 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4988 Then only these subroutine macros need have two variants; the higher
4989 levels of macros may be the same whether strict or not.
4991 Normally, constant addresses which are the sum of a @code{symbol_ref}
4992 and an integer are stored inside a @code{const} RTX to mark them as
4993 constant. Therefore, there is no need to recognize such sums
4994 specifically as legitimate addresses. Normally you would simply
4995 recognize any @code{const} as legitimate.
4997 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4998 sums that are not marked with @code{const}. It assumes that a naked
4999 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5000 naked constant sums as illegitimate addresses, so that none of them will
5001 be given to @code{PRINT_OPERAND_ADDRESS}.
5003 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5004 On some machines, whether a symbolic address is legitimate depends on
5005 the section that the address refers to. On these machines, define the
5006 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5007 into the @code{symbol_ref}, and then check for it here. When you see a
5008 @code{const}, you will have to look inside it to find the
5009 @code{symbol_ref} in order to determine the section. @xref{Assembler
5013 @defmac REG_OK_FOR_BASE_P (@var{x})
5014 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
5015 RTX) is valid for use as a base register. For hard registers, it
5016 should always accept those which the hardware permits and reject the
5017 others. Whether the macro accepts or rejects pseudo registers must be
5018 controlled by @code{REG_OK_STRICT} as described above. This usually
5019 requires two variant definitions, of which @code{REG_OK_STRICT}
5020 controls the one actually used.
5023 @defmac REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
5024 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
5025 that expression may examine the mode of the memory reference in
5026 @var{mode}. You should define this macro if the mode of the memory
5027 reference affects whether a register may be used as a base register. If
5028 you define this macro, the compiler will use it instead of
5029 @code{REG_OK_FOR_BASE_P}.
5032 @defmac REG_MODE_OK_FOR_REG_BASE_P (@var{x}, @var{mode})
5033 A C expression which is nonzero if @var{x} (assumed to be a @code{reg} RTX)
5034 is suitable for use as a base register in base plus index operand addresses,
5035 accessing memory in mode @var{mode}. It may be either a suitable hard
5036 register or a pseudo register that has been allocated such a hard register.
5037 You should define this macro if base plus index addresses have different
5038 requirements than other base register uses.
5041 @defmac REG_OK_FOR_INDEX_P (@var{x})
5042 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
5043 RTX) is valid for use as an index register.
5045 The difference between an index register and a base register is that
5046 the index register may be scaled. If an address involves the sum of
5047 two registers, neither one of them scaled, then either one may be
5048 labeled the ``base'' and the other the ``index''; but whichever
5049 labeling is used must fit the machine's constraints of which registers
5050 may serve in each capacity. The compiler will try both labelings,
5051 looking for one that is valid, and will reload one or both registers
5052 only if neither labeling works.
5055 @defmac FIND_BASE_TERM (@var{x})
5056 A C expression to determine the base term of address @var{x}.
5057 This macro is used in only one place: `find_base_term' in alias.c.
5059 It is always safe for this macro to not be defined. It exists so
5060 that alias analysis can understand machine-dependent addresses.
5062 The typical use of this macro is to handle addresses containing
5063 a label_ref or symbol_ref within an UNSPEC@.
5066 @defmac LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
5067 A C compound statement that attempts to replace @var{x} with a valid
5068 memory address for an operand of mode @var{mode}. @var{win} will be a
5069 C statement label elsewhere in the code; the macro definition may use
5072 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
5076 to avoid further processing if the address has become legitimate.
5078 @findex break_out_memory_refs
5079 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5080 and @var{oldx} will be the operand that was given to that function to produce
5083 The code generated by this macro should not alter the substructure of
5084 @var{x}. If it transforms @var{x} into a more legitimate form, it
5085 should assign @var{x} (which will always be a C variable) a new value.
5087 It is not necessary for this macro to come up with a legitimate
5088 address. The compiler has standard ways of doing so in all cases. In
5089 fact, it is safe to omit this macro. But often a
5090 machine-dependent strategy can generate better code.
5093 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5094 A C compound statement that attempts to replace @var{x}, which is an address
5095 that needs reloading, with a valid memory address for an operand of mode
5096 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5097 It is not necessary to define this macro, but it might be useful for
5098 performance reasons.
5100 For example, on the i386, it is sometimes possible to use a single
5101 reload register instead of two by reloading a sum of two pseudo
5102 registers into a register. On the other hand, for number of RISC
5103 processors offsets are limited so that often an intermediate address
5104 needs to be generated in order to address a stack slot. By defining
5105 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5106 generated for adjacent some stack slots can be made identical, and thus
5109 @emph{Note}: This macro should be used with caution. It is necessary
5110 to know something of how reload works in order to effectively use this,
5111 and it is quite easy to produce macros that build in too much knowledge
5112 of reload internals.
5114 @emph{Note}: This macro must be able to reload an address created by a
5115 previous invocation of this macro. If it fails to handle such addresses
5116 then the compiler may generate incorrect code or abort.
5119 The macro definition should use @code{push_reload} to indicate parts that
5120 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5121 suitable to be passed unaltered to @code{push_reload}.
5123 The code generated by this macro must not alter the substructure of
5124 @var{x}. If it transforms @var{x} into a more legitimate form, it
5125 should assign @var{x} (which will always be a C variable) a new value.
5126 This also applies to parts that you change indirectly by calling
5129 @findex strict_memory_address_p
5130 The macro definition may use @code{strict_memory_address_p} to test if
5131 the address has become legitimate.
5134 If you want to change only a part of @var{x}, one standard way of doing
5135 this is to use @code{copy_rtx}. Note, however, that is unshares only a
5136 single level of rtl. Thus, if the part to be changed is not at the
5137 top level, you'll need to replace first the top level.
5138 It is not necessary for this macro to come up with a legitimate
5139 address; but often a machine-dependent strategy can generate better code.
5142 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5143 A C statement or compound statement with a conditional @code{goto
5144 @var{label};} executed if memory address @var{x} (an RTX) can have
5145 different meanings depending on the machine mode of the memory
5146 reference it is used for or if the address is valid for some modes
5149 Autoincrement and autodecrement addresses typically have mode-dependent
5150 effects because the amount of the increment or decrement is the size
5151 of the operand being addressed. Some machines have other mode-dependent
5152 addresses. Many RISC machines have no mode-dependent addresses.
5154 You may assume that @var{addr} is a valid address for the machine.
5157 @defmac LEGITIMATE_CONSTANT_P (@var{x})
5158 A C expression that is nonzero if @var{x} is a legitimate constant for
5159 an immediate operand on the target machine. You can assume that
5160 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5161 @samp{1} is a suitable definition for this macro on machines where
5162 anything @code{CONSTANT_P} is valid.
5165 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5166 This hook is used to undo the possibly obfuscating effects of the
5167 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5168 macros. Some backend implementations of these macros wrap symbol
5169 references inside an @code{UNSPEC} rtx to represent PIC or similar
5170 addressing modes. This target hook allows GCC's optimizers to understand
5171 the semantics of these opaque @code{UNSPEC}s by converting them back
5172 into their original form.
5175 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (rtx @var{x})
5176 This hook should return true if @var{x} is of a form that cannot (or
5177 should not) be spilled to the constant pool. The default version of
5178 this hook returns false.
5180 The primary reason to define this hook is to prevent reload from
5181 deciding that a non-legitimate constant would be better reloaded
5182 from the constant pool instead of spilling and reloading a register
5183 holding the constant. This restriction is often true of addresses
5184 of TLS symbols for various targets.
5187 @deftypefn {Target Hook} bool TARGET_VECTORIZE_MISALIGNED_MEM_OK (@var{mode})
5188 This hook should return true if a move* pattern to/from memory
5189 can be generated for machine_mode @var{mode} even if the memory location
5191 If a move* of data to/from unaligned memory locations is not supported for
5192 machine_mode @var{mode}, the hook should return false.
5193 This hook is used by the autovectorizer, and when expanding a
5194 @code{MISALIGNED_INDIRECT_REF} expression.
5197 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5198 This hook should return the DECL of a function @var{f} that given an
5199 address @var{addr} as an argument returns a mask @var{m} that can be
5200 used to extract from two vectors the relevant data that resides in
5201 @var{addr} in case @var{addr} is not properly aligned.
5203 The autovectrizer, when vectorizing a load operation from an address
5204 @var{addr} that may be unaligned, will generate two vector loads from
5205 the two aligned addresses around @var{addr}. It then generates a
5206 @code{REALIGN_LOAD} operation to extract the relevant data from the
5207 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5208 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5209 the third argument, @var{OFF}, defines how the data will be extracted
5210 from these two vectors: if @var{OFF} is 0, then the returned vector is
5211 @var{v2}; otherwise, the returned vector is composed from the last
5212 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5213 @var{OFF} elements of @var{v2}.
5215 If this hook is defined, the autovectorizer will generate a call
5216 to @var{f} (using the DECL tree that this hook returns) and will
5217 use the return value of @var{f} as the argument @var{OFF} to
5218 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5219 should comply with the semantics expected by @code{REALIGN_LOAD}
5221 If this hook is not defined, then @var{addr} will be used as
5222 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5223 log2(@var{VS})-1 bits of @var{addr} will be considered.
5226 @node Condition Code
5227 @section Condition Code Status
5228 @cindex condition code status
5230 @c prevent bad page break with this line
5231 This describes the condition code status.
5234 The file @file{conditions.h} defines a variable @code{cc_status} to
5235 describe how the condition code was computed (in case the interpretation of
5236 the condition code depends on the instruction that it was set by). This
5237 variable contains the RTL expressions on which the condition code is
5238 currently based, and several standard flags.
5240 Sometimes additional machine-specific flags must be defined in the machine
5241 description header file. It can also add additional machine-specific
5242 information by defining @code{CC_STATUS_MDEP}.
5244 @defmac CC_STATUS_MDEP
5245 C code for a data type which is used for declaring the @code{mdep}
5246 component of @code{cc_status}. It defaults to @code{int}.
5248 This macro is not used on machines that do not use @code{cc0}.
5251 @defmac CC_STATUS_MDEP_INIT
5252 A C expression to initialize the @code{mdep} field to ``empty''.
5253 The default definition does nothing, since most machines don't use
5254 the field anyway. If you want to use the field, you should probably
5255 define this macro to initialize it.
5257 This macro is not used on machines that do not use @code{cc0}.
5260 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5261 A C compound statement to set the components of @code{cc_status}
5262 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5263 this macro's responsibility to recognize insns that set the condition
5264 code as a byproduct of other activity as well as those that explicitly
5267 This macro is not used on machines that do not use @code{cc0}.
5269 If there are insns that do not set the condition code but do alter
5270 other machine registers, this macro must check to see whether they
5271 invalidate the expressions that the condition code is recorded as
5272 reflecting. For example, on the 68000, insns that store in address
5273 registers do not set the condition code, which means that usually
5274 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5275 insns. But suppose that the previous insn set the condition code
5276 based on location @samp{a4@@(102)} and the current insn stores a new
5277 value in @samp{a4}. Although the condition code is not changed by
5278 this, it will no longer be true that it reflects the contents of
5279 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5280 @code{cc_status} in this case to say that nothing is known about the
5281 condition code value.
5283 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5284 with the results of peephole optimization: insns whose patterns are
5285 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5286 constants which are just the operands. The RTL structure of these
5287 insns is not sufficient to indicate what the insns actually do. What
5288 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5289 @code{CC_STATUS_INIT}.
5291 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5292 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5293 @samp{cc}. This avoids having detailed information about patterns in
5294 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5297 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5298 Returns a mode from class @code{MODE_CC} to be used when comparison
5299 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5300 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5301 @pxref{Jump Patterns} for a description of the reason for this
5305 #define SELECT_CC_MODE(OP,X,Y) \
5306 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5307 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5308 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5309 || GET_CODE (X) == NEG) \
5310 ? CC_NOOVmode : CCmode))
5313 You should define this macro if and only if you define extra CC modes
5314 in @file{@var{machine}-modes.def}.
5317 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5318 On some machines not all possible comparisons are defined, but you can
5319 convert an invalid comparison into a valid one. For example, the Alpha
5320 does not have a @code{GT} comparison, but you can use an @code{LT}
5321 comparison instead and swap the order of the operands.
5323 On such machines, define this macro to be a C statement to do any
5324 required conversions. @var{code} is the initial comparison code
5325 and @var{op0} and @var{op1} are the left and right operands of the
5326 comparison, respectively. You should modify @var{code}, @var{op0}, and
5327 @var{op1} as required.
5329 GCC will not assume that the comparison resulting from this macro is
5330 valid but will see if the resulting insn matches a pattern in the
5333 You need not define this macro if it would never change the comparison
5337 @defmac REVERSIBLE_CC_MODE (@var{mode})
5338 A C expression whose value is one if it is always safe to reverse a
5339 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5340 can ever return @var{mode} for a floating-point inequality comparison,
5341 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5343 You need not define this macro if it would always returns zero or if the
5344 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5345 For example, here is the definition used on the SPARC, where floating-point
5346 inequality comparisons are always given @code{CCFPEmode}:
5349 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5353 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
5354 A C expression whose value is reversed condition code of the @var{code} for
5355 comparison done in CC_MODE @var{mode}. The macro is used only in case
5356 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5357 machine has some non-standard way how to reverse certain conditionals. For
5358 instance in case all floating point conditions are non-trapping, compiler may
5359 freely convert unordered compares to ordered one. Then definition may look
5363 #define REVERSE_CONDITION(CODE, MODE) \
5364 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5365 : reverse_condition_maybe_unordered (CODE))
5369 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
5370 A C expression that returns true if the conditional execution predicate
5371 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
5372 versa. Define this to return 0 if the target has conditional execution
5373 predicates that cannot be reversed safely. There is no need to validate
5374 that the arguments of op1 and op2 are the same, this is done separately.
5375 If no expansion is specified, this macro is defined as follows:
5378 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5379 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
5383 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *, unsigned int *)
5384 On targets which do not use @code{(cc0)}, and which use a hard
5385 register rather than a pseudo-register to hold condition codes, the
5386 regular CSE passes are often not able to identify cases in which the
5387 hard register is set to a common value. Use this hook to enable a
5388 small pass which optimizes such cases. This hook should return true
5389 to enable this pass, and it should set the integers to which its
5390 arguments point to the hard register numbers used for condition codes.
5391 When there is only one such register, as is true on most systems, the
5392 integer pointed to by the second argument should be set to
5393 @code{INVALID_REGNUM}.
5395 The default version of this hook returns false.
5398 @deftypefn {Target Hook} enum machine_mode TARGET_CC_MODES_COMPATIBLE (enum machine_mode, enum machine_mode)
5399 On targets which use multiple condition code modes in class
5400 @code{MODE_CC}, it is sometimes the case that a comparison can be
5401 validly done in more than one mode. On such a system, define this
5402 target hook to take two mode arguments and to return a mode in which
5403 both comparisons may be validly done. If there is no such mode,
5404 return @code{VOIDmode}.
5406 The default version of this hook checks whether the modes are the
5407 same. If they are, it returns that mode. If they are different, it
5408 returns @code{VOIDmode}.
5412 @section Describing Relative Costs of Operations
5413 @cindex costs of instructions
5414 @cindex relative costs
5415 @cindex speed of instructions
5417 These macros let you describe the relative speed of various operations
5418 on the target machine.
5420 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5421 A C expression for the cost of moving data of mode @var{mode} from a
5422 register in class @var{from} to one in class @var{to}. The classes are
5423 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5424 value of 2 is the default; other values are interpreted relative to
5427 It is not required that the cost always equal 2 when @var{from} is the
5428 same as @var{to}; on some machines it is expensive to move between
5429 registers if they are not general registers.
5431 If reload sees an insn consisting of a single @code{set} between two
5432 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5433 classes returns a value of 2, reload does not check to ensure that the
5434 constraints of the insn are met. Setting a cost of other than 2 will
5435 allow reload to verify that the constraints are met. You should do this
5436 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5439 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5440 A C expression for the cost of moving data of mode @var{mode} between a
5441 register of class @var{class} and memory; @var{in} is zero if the value
5442 is to be written to memory, nonzero if it is to be read in. This cost
5443 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5444 registers and memory is more expensive than between two registers, you
5445 should define this macro to express the relative cost.
5447 If you do not define this macro, GCC uses a default cost of 4 plus
5448 the cost of copying via a secondary reload register, if one is
5449 needed. If your machine requires a secondary reload register to copy
5450 between memory and a register of @var{class} but the reload mechanism is
5451 more complex than copying via an intermediate, define this macro to
5452 reflect the actual cost of the move.
5454 GCC defines the function @code{memory_move_secondary_cost} if
5455 secondary reloads are needed. It computes the costs due to copying via
5456 a secondary register. If your machine copies from memory using a
5457 secondary register in the conventional way but the default base value of
5458 4 is not correct for your machine, define this macro to add some other
5459 value to the result of that function. The arguments to that function
5460 are the same as to this macro.
5464 A C expression for the cost of a branch instruction. A value of 1 is
5465 the default; other values are interpreted relative to that.
5468 Here are additional macros which do not specify precise relative costs,
5469 but only that certain actions are more expensive than GCC would
5472 @defmac SLOW_BYTE_ACCESS
5473 Define this macro as a C expression which is nonzero if accessing less
5474 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5475 faster than accessing a word of memory, i.e., if such access
5476 require more than one instruction or if there is no difference in cost
5477 between byte and (aligned) word loads.
5479 When this macro is not defined, the compiler will access a field by
5480 finding the smallest containing object; when it is defined, a fullword
5481 load will be used if alignment permits. Unless bytes accesses are
5482 faster than word accesses, using word accesses is preferable since it
5483 may eliminate subsequent memory access if subsequent accesses occur to
5484 other fields in the same word of the structure, but to different bytes.
5487 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5488 Define this macro to be the value 1 if memory accesses described by the
5489 @var{mode} and @var{alignment} parameters have a cost many times greater
5490 than aligned accesses, for example if they are emulated in a trap
5493 When this macro is nonzero, the compiler will act as if
5494 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5495 moves. This can cause significantly more instructions to be produced.
5496 Therefore, do not set this macro nonzero if unaligned accesses only add a
5497 cycle or two to the time for a memory access.
5499 If the value of this macro is always zero, it need not be defined. If
5500 this macro is defined, it should produce a nonzero value when
5501 @code{STRICT_ALIGNMENT} is nonzero.
5505 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5506 which a sequence of insns should be generated instead of a
5507 string move insn or a library call. Increasing the value will always
5508 make code faster, but eventually incurs high cost in increased code size.
5510 Note that on machines where the corresponding move insn is a
5511 @code{define_expand} that emits a sequence of insns, this macro counts
5512 the number of such sequences.
5514 If you don't define this, a reasonable default is used.
5517 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5518 A C expression used to determine whether @code{move_by_pieces} will be used to
5519 copy a chunk of memory, or whether some other block move mechanism
5520 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5521 than @code{MOVE_RATIO}.
5524 @defmac MOVE_MAX_PIECES
5525 A C expression used by @code{move_by_pieces} to determine the largest unit
5526 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5530 The threshold of number of scalar move insns, @emph{below} which a sequence
5531 of insns should be generated to clear memory instead of a string clear insn
5532 or a library call. Increasing the value will always make code faster, but
5533 eventually incurs high cost in increased code size.
5535 If you don't define this, a reasonable default is used.
5538 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5539 A C expression used to determine whether @code{clear_by_pieces} will be used
5540 to clear a chunk of memory, or whether some other block clear mechanism
5541 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5542 than @code{CLEAR_RATIO}.
5545 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
5546 A C expression used to determine whether @code{store_by_pieces} will be
5547 used to set a chunk of memory to a constant value, or whether some other
5548 mechanism will be used. Used by @code{__builtin_memset} when storing
5549 values other than constant zero and by @code{__builtin_strcpy} when
5550 when called with a constant source string.
5551 Defaults to to 1 if @code{move_by_pieces_ninsns} returns less
5552 than @code{MOVE_RATIO}.
5555 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
5556 A C expression used to determine whether a load postincrement is a good
5557 thing to use for a given mode. Defaults to the value of
5558 @code{HAVE_POST_INCREMENT}.
5561 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
5562 A C expression used to determine whether a load postdecrement is a good
5563 thing to use for a given mode. Defaults to the value of
5564 @code{HAVE_POST_DECREMENT}.
5567 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
5568 A C expression used to determine whether a load preincrement is a good
5569 thing to use for a given mode. Defaults to the value of
5570 @code{HAVE_PRE_INCREMENT}.
5573 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
5574 A C expression used to determine whether a load predecrement is a good
5575 thing to use for a given mode. Defaults to the value of
5576 @code{HAVE_PRE_DECREMENT}.
5579 @defmac USE_STORE_POST_INCREMENT (@var{mode})
5580 A C expression used to determine whether a store postincrement is a good
5581 thing to use for a given mode. Defaults to the value of
5582 @code{HAVE_POST_INCREMENT}.
5585 @defmac USE_STORE_POST_DECREMENT (@var{mode})
5586 A C expression used to determine whether a store postdecrement is a good
5587 thing to use for a given mode. Defaults to the value of
5588 @code{HAVE_POST_DECREMENT}.
5591 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
5592 This macro is used to determine whether a store preincrement is a good
5593 thing to use for a given mode. Defaults to the value of
5594 @code{HAVE_PRE_INCREMENT}.
5597 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
5598 This macro is used to determine whether a store predecrement is a good
5599 thing to use for a given mode. Defaults to the value of
5600 @code{HAVE_PRE_DECREMENT}.
5603 @defmac NO_FUNCTION_CSE
5604 Define this macro if it is as good or better to call a constant
5605 function address than to call an address kept in a register.
5608 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
5609 Define this macro if a non-short-circuit operation produced by
5610 @samp{fold_range_test ()} is optimal. This macro defaults to true if
5611 @code{BRANCH_COST} is greater than or equal to the value 2.
5614 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total})
5615 This target hook describes the relative costs of RTL expressions.
5617 The cost may depend on the precise form of the expression, which is
5618 available for examination in @var{x}, and the rtx code of the expression
5619 in which it is contained, found in @var{outer_code}. @var{code} is the
5620 expression code---redundant, since it can be obtained with
5621 @code{GET_CODE (@var{x})}.
5623 In implementing this hook, you can use the construct
5624 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
5627 On entry to the hook, @code{*@var{total}} contains a default estimate
5628 for the cost of the expression. The hook should modify this value as
5629 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
5630 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
5631 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
5633 When optimizing for code size, i.e.@: when @code{optimize_size} is
5634 nonzero, this target hook should be used to estimate the relative
5635 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
5637 The hook returns true when all subexpressions of @var{x} have been
5638 processed, and false when @code{rtx_cost} should recurse.
5641 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address})
5642 This hook computes the cost of an addressing mode that contains
5643 @var{address}. If not defined, the cost is computed from
5644 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
5646 For most CISC machines, the default cost is a good approximation of the
5647 true cost of the addressing mode. However, on RISC machines, all
5648 instructions normally have the same length and execution time. Hence
5649 all addresses will have equal costs.
5651 In cases where more than one form of an address is known, the form with
5652 the lowest cost will be used. If multiple forms have the same, lowest,
5653 cost, the one that is the most complex will be used.
5655 For example, suppose an address that is equal to the sum of a register
5656 and a constant is used twice in the same basic block. When this macro
5657 is not defined, the address will be computed in a register and memory
5658 references will be indirect through that register. On machines where
5659 the cost of the addressing mode containing the sum is no higher than
5660 that of a simple indirect reference, this will produce an additional
5661 instruction and possibly require an additional register. Proper
5662 specification of this macro eliminates this overhead for such machines.
5664 This hook is never called with an invalid address.
5666 On machines where an address involving more than one register is as
5667 cheap as an address computation involving only one register, defining
5668 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
5669 be live over a region of code where only one would have been if
5670 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
5671 should be considered in the definition of this macro. Equivalent costs
5672 should probably only be given to addresses with different numbers of
5673 registers on machines with lots of registers.
5677 @section Adjusting the Instruction Scheduler
5679 The instruction scheduler may need a fair amount of machine-specific
5680 adjustment in order to produce good code. GCC provides several target
5681 hooks for this purpose. It is usually enough to define just a few of
5682 them: try the first ones in this list first.
5684 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
5685 This hook returns the maximum number of instructions that can ever
5686 issue at the same time on the target machine. The default is one.
5687 Although the insn scheduler can define itself the possibility of issue
5688 an insn on the same cycle, the value can serve as an additional
5689 constraint to issue insns on the same simulated processor cycle (see
5690 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
5691 This value must be constant over the entire compilation. If you need
5692 it to vary depending on what the instructions are, you must use
5693 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
5695 You could define this hook to return the value of the macro
5696 @code{MAX_DFA_ISSUE_RATE}.
5699 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
5700 This hook is executed by the scheduler after it has scheduled an insn
5701 from the ready list. It should return the number of insns which can
5702 still be issued in the current cycle. The default is
5703 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
5704 @code{USE}, which normally are not counted against the issue rate.
5705 You should define this hook if some insns take more machine resources
5706 than others, so that fewer insns can follow them in the same cycle.
5707 @var{file} is either a null pointer, or a stdio stream to write any
5708 debug output to. @var{verbose} is the verbose level provided by
5709 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
5713 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
5714 This function corrects the value of @var{cost} based on the
5715 relationship between @var{insn} and @var{dep_insn} through the
5716 dependence @var{link}. It should return the new value. The default
5717 is to make no adjustment to @var{cost}. This can be used for example
5718 to specify to the scheduler using the traditional pipeline description
5719 that an output- or anti-dependence does not incur the same cost as a
5720 data-dependence. If the scheduler using the automaton based pipeline
5721 description, the cost of anti-dependence is zero and the cost of
5722 output-dependence is maximum of one and the difference of latency
5723 times of the first and the second insns. If these values are not
5724 acceptable, you could use the hook to modify them too. See also
5725 @pxref{Processor pipeline description}.
5728 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
5729 This hook adjusts the integer scheduling priority @var{priority} of
5730 @var{insn}. It should return the new priority. Reduce the priority to
5731 execute @var{insn} earlier, increase the priority to execute @var{insn}
5732 later. Do not define this hook if you do not need to adjust the
5733 scheduling priorities of insns.
5736 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
5737 This hook is executed by the scheduler after it has scheduled the ready
5738 list, to allow the machine description to reorder it (for example to
5739 combine two small instructions together on @samp{VLIW} machines).
5740 @var{file} is either a null pointer, or a stdio stream to write any
5741 debug output to. @var{verbose} is the verbose level provided by
5742 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
5743 list of instructions that are ready to be scheduled. @var{n_readyp} is
5744 a pointer to the number of elements in the ready list. The scheduler
5745 reads the ready list in reverse order, starting with
5746 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
5747 is the timer tick of the scheduler. You may modify the ready list and
5748 the number of ready insns. The return value is the number of insns that
5749 can issue this cycle; normally this is just @code{issue_rate}. See also
5750 @samp{TARGET_SCHED_REORDER2}.
5753 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
5754 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
5755 function is called whenever the scheduler starts a new cycle. This one
5756 is called once per iteration over a cycle, immediately after
5757 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
5758 return the number of insns to be scheduled in the same cycle. Defining
5759 this hook can be useful if there are frequent situations where
5760 scheduling one insn causes other insns to become ready in the same
5761 cycle. These other insns can then be taken into account properly.
5764 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
5765 This hook is called after evaluation forward dependencies of insns in
5766 chain given by two parameter values (@var{head} and @var{tail}
5767 correspondingly) but before insns scheduling of the insn chain. For
5768 example, it can be used for better insn classification if it requires
5769 analysis of dependencies. This hook can use backward and forward
5770 dependencies of the insn scheduler because they are already
5774 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
5775 This hook is executed by the scheduler at the beginning of each block of
5776 instructions that are to be scheduled. @var{file} is either a null
5777 pointer, or a stdio stream to write any debug output to. @var{verbose}
5778 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5779 @var{max_ready} is the maximum number of insns in the current scheduling
5780 region that can be live at the same time. This can be used to allocate
5781 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
5784 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
5785 This hook is executed by the scheduler at the end of each block of
5786 instructions that are to be scheduled. It can be used to perform
5787 cleanup of any actions done by the other scheduling hooks. @var{file}
5788 is either a null pointer, or a stdio stream to write any debug output
5789 to. @var{verbose} is the verbose level provided by
5790 @option{-fsched-verbose-@var{n}}.
5793 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
5794 This hook is executed by the scheduler after function level initializations.
5795 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
5796 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5797 @var{old_max_uid} is the maximum insn uid when scheduling begins.
5800 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
5801 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
5802 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
5803 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5806 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
5807 The hook returns an RTL insn. The automaton state used in the
5808 pipeline hazard recognizer is changed as if the insn were scheduled
5809 when the new simulated processor cycle starts. Usage of the hook may
5810 simplify the automaton pipeline description for some @acronym{VLIW}
5811 processors. If the hook is defined, it is used only for the automaton
5812 based pipeline description. The default is not to change the state
5813 when the new simulated processor cycle starts.
5816 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
5817 The hook can be used to initialize data used by the previous hook.
5820 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
5821 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
5822 to changed the state as if the insn were scheduled when the new
5823 simulated processor cycle finishes.
5826 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
5827 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
5828 used to initialize data used by the previous hook.
5831 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
5832 This hook controls better choosing an insn from the ready insn queue
5833 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
5834 chooses the first insn from the queue. If the hook returns a positive
5835 value, an additional scheduler code tries all permutations of
5836 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
5837 subsequent ready insns to choose an insn whose issue will result in
5838 maximal number of issued insns on the same cycle. For the
5839 @acronym{VLIW} processor, the code could actually solve the problem of
5840 packing simple insns into the @acronym{VLIW} insn. Of course, if the
5841 rules of @acronym{VLIW} packing are described in the automaton.
5843 This code also could be used for superscalar @acronym{RISC}
5844 processors. Let us consider a superscalar @acronym{RISC} processor
5845 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
5846 @var{B}, some insns can be executed only in pipelines @var{B} or
5847 @var{C}, and one insn can be executed in pipeline @var{B}. The
5848 processor may issue the 1st insn into @var{A} and the 2nd one into
5849 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
5850 until the next cycle. If the scheduler issues the 3rd insn the first,
5851 the processor could issue all 3 insns per cycle.
5853 Actually this code demonstrates advantages of the automaton based
5854 pipeline hazard recognizer. We try quickly and easy many insn
5855 schedules to choose the best one.
5857 The default is no multipass scheduling.
5860 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
5862 This hook controls what insns from the ready insn queue will be
5863 considered for the multipass insn scheduling. If the hook returns
5864 zero for insn passed as the parameter, the insn will be not chosen to
5867 The default is that any ready insns can be chosen to be issued.
5870 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
5872 This hook is called by the insn scheduler before issuing insn passed
5873 as the third parameter on given cycle. If the hook returns nonzero,
5874 the insn is not issued on given processors cycle. Instead of that,
5875 the processor cycle is advanced. If the value passed through the last
5876 parameter is zero, the insn ready queue is not sorted on the new cycle
5877 start as usually. The first parameter passes file for debugging
5878 output. The second one passes the scheduler verbose level of the
5879 debugging output. The forth and the fifth parameter values are
5880 correspondingly processor cycle on which the previous insn has been
5881 issued and the current processor cycle.
5884 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (rtx @var{insn1}, rtx @var{insn2}, rtx @var{dep_link}, int @var{dep_cost}, int @var{distance})
5885 This hook is used to define which dependences are considered costly by
5886 the target, so costly that it is not advisable to schedule the insns that
5887 are involved in the dependence too close to one another. The parameters
5888 to this hook are as follows: The second parameter @var{insn2} is dependent
5889 upon the first parameter @var{insn1}. The dependence between @var{insn1}
5890 and @var{insn2} is represented by the third parameter @var{dep_link}. The
5891 fourth parameter @var{cost} is the cost of the dependence, and the fifth
5892 parameter @var{distance} is the distance in cycles between the two insns.
5893 The hook returns @code{true} if considering the distance between the two
5894 insns the dependence between them is considered costly by the target,
5895 and @code{false} otherwise.
5897 Defining this hook can be useful in multiple-issue out-of-order machines,
5898 where (a) it's practically hopeless to predict the actual data/resource
5899 delays, however: (b) there's a better chance to predict the actual grouping
5900 that will be formed, and (c) correctly emulating the grouping can be very
5901 important. In such targets one may want to allow issuing dependent insns
5902 closer to one another---i.e., closer than the dependence distance; however,
5903 not in cases of "costly dependences", which this hooks allows to define.
5906 Macros in the following table are generated by the program
5907 @file{genattr} and can be useful for writing the hooks.
5909 @defmac MAX_DFA_ISSUE_RATE
5910 The macro definition is generated in the automaton based pipeline
5911 description interface. Its value is calculated from the automaton
5912 based pipeline description and is equal to maximal number of all insns
5913 described in constructions @samp{define_insn_reservation} which can be
5914 issued on the same processor cycle.
5918 @section Dividing the Output into Sections (Texts, Data, @dots{})
5919 @c the above section title is WAY too long. maybe cut the part between
5920 @c the (...)? --mew 10feb93
5922 An object file is divided into sections containing different types of
5923 data. In the most common case, there are three sections: the @dfn{text
5924 section}, which holds instructions and read-only data; the @dfn{data
5925 section}, which holds initialized writable data; and the @dfn{bss
5926 section}, which holds uninitialized data. Some systems have other kinds
5929 The compiler must tell the assembler when to switch sections. These
5930 macros control what commands to output to tell the assembler this. You
5931 can also define additional sections.
5933 @defmac TEXT_SECTION_ASM_OP
5934 A C expression whose value is a string, including spacing, containing the
5935 assembler operation that should precede instructions and read-only data.
5936 Normally @code{"\t.text"} is right.
5939 @defmac HOT_TEXT_SECTION_NAME
5940 If defined, a C string constant for the name of the section containing most
5941 frequently executed functions of the program. If not defined, GCC will provide
5942 a default definition if the target supports named sections.
5945 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5946 If defined, a C string constant for the name of the section containing unlikely
5947 executed functions in the program.
5950 @defmac DATA_SECTION_ASM_OP
5951 A C expression whose value is a string, including spacing, containing the
5952 assembler operation to identify the following data as writable initialized
5953 data. Normally @code{"\t.data"} is right.
5956 @defmac READONLY_DATA_SECTION_ASM_OP
5957 A C expression whose value is a string, including spacing, containing the
5958 assembler operation to identify the following data as read-only initialized
5962 @defmac READONLY_DATA_SECTION
5963 A macro naming a function to call to switch to the proper section for
5964 read-only data. The default is to use @code{READONLY_DATA_SECTION_ASM_OP}
5965 if defined, else fall back to @code{text_section}.
5967 The most common definition will be @code{data_section}, if the target
5968 does not have a special read-only data section, and does not put data
5969 in the text section.
5972 @defmac BSS_SECTION_ASM_OP
5973 If defined, a C expression whose value is a string, including spacing,
5974 containing the assembler operation to identify the following data as
5975 uninitialized global data. If not defined, and neither
5976 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5977 uninitialized global data will be output in the data section if
5978 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5982 @defmac INIT_SECTION_ASM_OP
5983 If defined, a C expression whose value is a string, including spacing,
5984 containing the assembler operation to identify the following data as
5985 initialization code. If not defined, GCC will assume such a section does
5989 @defmac FINI_SECTION_ASM_OP
5990 If defined, a C expression whose value is a string, including spacing,
5991 containing the assembler operation to identify the following data as
5992 finalization code. If not defined, GCC will assume such a section does
5996 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
5997 If defined, an ASM statement that switches to a different section
5998 via @var{section_op}, calls @var{function}, and switches back to
5999 the text section. This is used in @file{crtstuff.c} if
6000 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
6001 to initialization and finalization functions from the init and fini
6002 sections. By default, this macro uses a simple function call. Some
6003 ports need hand-crafted assembly code to avoid dependencies on
6004 registers initialized in the function prologue or to ensure that
6005 constant pools don't end up too far way in the text section.
6008 @defmac FORCE_CODE_SECTION_ALIGN
6009 If defined, an ASM statement that aligns a code section to some
6010 arbitrary boundary. This is used to force all fragments of the
6011 @code{.init} and @code{.fini} sections to have to same alignment
6012 and thus prevent the linker from having to add any padding.
6017 @defmac EXTRA_SECTIONS
6018 A list of names for sections other than the standard two, which are
6019 @code{in_text} and @code{in_data}. You need not define this macro
6020 on a system with no other sections (that GCC needs to use).
6023 @findex text_section
6024 @findex data_section
6025 @defmac EXTRA_SECTION_FUNCTIONS
6026 One or more functions to be defined in @file{varasm.c}. These
6027 functions should do jobs analogous to those of @code{text_section} and
6028 @code{data_section}, for your additional sections. Do not define this
6029 macro if you do not define @code{EXTRA_SECTIONS}.
6032 @defmac JUMP_TABLES_IN_TEXT_SECTION
6033 Define this macro to be an expression with a nonzero value if jump
6034 tables (for @code{tablejump} insns) should be output in the text
6035 section, along with the assembler instructions. Otherwise, the
6036 readonly data section is used.
6038 This macro is irrelevant if there is no separate readonly data section.
6041 @deftypefn {Target Hook} void TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
6042 Switches to the appropriate section for output of @var{exp}. You can
6043 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
6044 some sort. @var{reloc} indicates whether the initial value of @var{exp}
6045 requires link-time relocations. Bit 0 is set when variable contains
6046 local relocations only, while bit 1 is set for global relocations.
6047 Select the section by calling @code{data_section} or one of the
6048 alternatives for other sections. @var{align} is the constant alignment
6051 The default version of this function takes care of putting read-only
6052 variables in @code{readonly_data_section}.
6054 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
6057 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
6058 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
6059 for @code{FUNCTION_DECL}s as well as for variables and constants.
6061 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
6062 function has been determined to be likely to be called, and nonzero if
6063 it is unlikely to be called.
6066 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
6067 Build up a unique section name, expressed as a @code{STRING_CST} node,
6068 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
6069 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
6070 the initial value of @var{exp} requires link-time relocations.
6072 The default version of this function appends the symbol name to the
6073 ELF section name that would normally be used for the symbol. For
6074 example, the function @code{foo} would be placed in @code{.text.foo}.
6075 Whatever the actual target object format, this is often good enough.
6078 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
6079 Switches to a readonly data section associated with
6080 @samp{DECL_SECTION_NAME (@var{decl})}.
6081 The default version of this function switches to @code{.gnu.linkonce.r.name}
6082 section if function's section is @code{.gnu.linkonce.t.name}, to
6083 @code{.rodata.name} if function is in @code{.text.name} section
6084 and otherwise switches to the normal readonly data section.
6087 @deftypefn {Target Hook} void TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
6088 Switches to the appropriate section for output of constant pool entry
6089 @var{x} in @var{mode}. You can assume that @var{x} is some kind of
6090 constant in RTL@. The argument @var{mode} is redundant except in the
6091 case of a @code{const_int} rtx. Select the section by calling
6092 @code{readonly_data_section} or one of the alternatives for other
6093 sections. @var{align} is the constant alignment in bits.
6095 The default version of this function takes care of putting symbolic
6096 constants in @code{flag_pic} mode in @code{data_section} and everything
6097 else in @code{readonly_data_section}.
6100 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
6101 Define this hook if references to a symbol or a constant must be
6102 treated differently depending on something about the variable or
6103 function named by the symbol (such as what section it is in).
6105 The hook is executed immediately after rtl has been created for
6106 @var{decl}, which may be a variable or function declaration or
6107 an entry in the constant pool. In either case, @var{rtl} is the
6108 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
6109 in this hook; that field may not have been initialized yet.
6111 In the case of a constant, it is safe to assume that the rtl is
6112 a @code{mem} whose address is a @code{symbol_ref}. Most decls
6113 will also have this form, but that is not guaranteed. Global
6114 register variables, for instance, will have a @code{reg} for their
6115 rtl. (Normally the right thing to do with such unusual rtl is
6118 The @var{new_decl_p} argument will be true if this is the first time
6119 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
6120 be false for subsequent invocations, which will happen for duplicate
6121 declarations. Whether or not anything must be done for the duplicate
6122 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
6123 @var{new_decl_p} is always true when the hook is called for a constant.
6125 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
6126 The usual thing for this hook to do is to record flags in the
6127 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
6128 Historically, the name string was modified if it was necessary to
6129 encode more than one bit of information, but this practice is now
6130 discouraged; use @code{SYMBOL_REF_FLAGS}.
6132 The default definition of this hook, @code{default_encode_section_info}
6133 in @file{varasm.c}, sets a number of commonly-useful bits in
6134 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
6135 before overriding it.
6138 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
6139 Decode @var{name} and return the real name part, sans
6140 the characters that @code{TARGET_ENCODE_SECTION_INFO}
6144 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
6145 Returns true if @var{exp} should be placed into a ``small data'' section.
6146 The default version of this hook always returns false.
6149 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
6150 Contains the value true if the target places read-only
6151 ``small data'' into a separate section. The default value is false.
6154 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
6155 Returns true if @var{exp} names an object for which name resolution
6156 rules must resolve to the current ``module'' (dynamic shared library
6157 or executable image).
6159 The default version of this hook implements the name resolution rules
6160 for ELF, which has a looser model of global name binding than other
6161 currently supported object file formats.
6164 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
6165 Contains the value true if the target supports thread-local storage.
6166 The default value is false.
6171 @section Position Independent Code
6172 @cindex position independent code
6175 This section describes macros that help implement generation of position
6176 independent code. Simply defining these macros is not enough to
6177 generate valid PIC; you must also add support to the macros
6178 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
6179 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
6180 @samp{movsi} to do something appropriate when the source operand
6181 contains a symbolic address. You may also need to alter the handling of
6182 switch statements so that they use relative addresses.
6183 @c i rearranged the order of the macros above to try to force one of
6184 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
6186 @defmac PIC_OFFSET_TABLE_REGNUM
6187 The register number of the register used to address a table of static
6188 data addresses in memory. In some cases this register is defined by a
6189 processor's ``application binary interface'' (ABI)@. When this macro
6190 is defined, RTL is generated for this register once, as with the stack
6191 pointer and frame pointer registers. If this macro is not defined, it
6192 is up to the machine-dependent files to allocate such a register (if
6193 necessary). Note that this register must be fixed when in use (e.g.@:
6194 when @code{flag_pic} is true).
6197 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6198 Define this macro if the register defined by
6199 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
6200 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
6203 @defmac FINALIZE_PIC
6204 By generating position-independent code, when two different programs (A
6205 and B) share a common library (libC.a), the text of the library can be
6206 shared whether or not the library is linked at the same address for both
6207 programs. In some of these environments, position-independent code
6208 requires not only the use of different addressing modes, but also
6209 special code to enable the use of these addressing modes.
6211 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
6212 codes once the function is being compiled into assembly code, but not
6213 before. (It is not done before, because in the case of compiling an
6214 inline function, it would lead to multiple PIC prologues being
6215 included in functions which used inline functions and were compiled to
6219 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
6220 A C expression that is nonzero if @var{x} is a legitimate immediate
6221 operand on the target machine when generating position independent code.
6222 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6223 check this. You can also assume @var{flag_pic} is true, so you need not
6224 check it either. You need not define this macro if all constants
6225 (including @code{SYMBOL_REF}) can be immediate operands when generating
6226 position independent code.
6229 @node Assembler Format
6230 @section Defining the Output Assembler Language
6232 This section describes macros whose principal purpose is to describe how
6233 to write instructions in assembler language---rather than what the
6237 * File Framework:: Structural information for the assembler file.
6238 * Data Output:: Output of constants (numbers, strings, addresses).
6239 * Uninitialized Data:: Output of uninitialized variables.
6240 * Label Output:: Output and generation of labels.
6241 * Initialization:: General principles of initialization
6242 and termination routines.
6243 * Macros for Initialization::
6244 Specific macros that control the handling of
6245 initialization and termination routines.
6246 * Instruction Output:: Output of actual instructions.
6247 * Dispatch Tables:: Output of jump tables.
6248 * Exception Region Output:: Output of exception region code.
6249 * Alignment Output:: Pseudo ops for alignment and skipping data.
6252 @node File Framework
6253 @subsection The Overall Framework of an Assembler File
6254 @cindex assembler format
6255 @cindex output of assembler code
6257 @c prevent bad page break with this line
6258 This describes the overall framework of an assembly file.
6260 @deftypefn {Target Hook} void TARGET_ASM_FILE_START ()
6261 @findex default_file_start
6262 Output to @code{asm_out_file} any text which the assembler expects to
6263 find at the beginning of a file. The default behavior is controlled
6264 by two flags, documented below. Unless your target's assembler is
6265 quite unusual, if you override the default, you should call
6266 @code{default_file_start} at some point in your target hook. This
6267 lets other target files rely on these variables.
6270 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
6271 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
6272 printed as the very first line in the assembly file, unless
6273 @option{-fverbose-asm} is in effect. (If that macro has been defined
6274 to the empty string, this variable has no effect.) With the normal
6275 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
6276 assembler that it need not bother stripping comments or extra
6277 whitespace from its input. This allows it to work a bit faster.
6279 The default is false. You should not set it to true unless you have
6280 verified that your port does not generate any extra whitespace or
6281 comments that will cause GAS to issue errors in NO_APP mode.
6284 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
6285 If this flag is true, @code{output_file_directive} will be called
6286 for the primary source file, immediately after printing
6287 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
6288 this to be done. The default is false.
6291 @deftypefn {Target Hook} void TARGET_ASM_FILE_END ()
6292 Output to @code{asm_out_file} any text which the assembler expects
6293 to find at the end of a file. The default is to output nothing.
6296 @deftypefun void file_end_indicate_exec_stack ()
6297 Some systems use a common convention, the @samp{.note.GNU-stack}
6298 special section, to indicate whether or not an object file relies on
6299 the stack being executable. If your system uses this convention, you
6300 should define @code{TARGET_ASM_FILE_END} to this function. If you
6301 need to do other things in that hook, have your hook function call
6305 @defmac ASM_COMMENT_START
6306 A C string constant describing how to begin a comment in the target
6307 assembler language. The compiler assumes that the comment will end at
6308 the end of the line.
6312 A C string constant for text to be output before each @code{asm}
6313 statement or group of consecutive ones. Normally this is
6314 @code{"#APP"}, which is a comment that has no effect on most
6315 assemblers but tells the GNU assembler that it must check the lines
6316 that follow for all valid assembler constructs.
6320 A C string constant for text to be output after each @code{asm}
6321 statement or group of consecutive ones. Normally this is
6322 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6323 time-saving assumptions that are valid for ordinary compiler output.
6326 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6327 A C statement to output COFF information or DWARF debugging information
6328 which indicates that filename @var{name} is the current source file to
6329 the stdio stream @var{stream}.
6331 This macro need not be defined if the standard form of output
6332 for the file format in use is appropriate.
6335 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6336 A C statement to output the string @var{string} to the stdio stream
6337 @var{stream}. If you do not call the function @code{output_quoted_string}
6338 in your config files, GCC will only call it to output filenames to
6339 the assembler source. So you can use it to canonicalize the format
6340 of the filename using this macro.
6343 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6344 A C statement to output something to the assembler file to handle a
6345 @samp{#ident} directive containing the text @var{string}. If this
6346 macro is not defined, nothing is output for a @samp{#ident} directive.
6349 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6350 Output assembly directives to switch to section @var{name}. The section
6351 should have attributes as specified by @var{flags}, which is a bit mask
6352 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6353 is nonzero, it contains an alignment in bytes to be used for the section,
6354 otherwise some target default should be used. Only targets that must
6355 specify an alignment within the section directive need pay attention to
6356 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6359 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6360 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6363 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6364 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6365 based on a variable or function decl, a section name, and whether or not the
6366 declaration's initializer may contain runtime relocations. @var{decl} may be
6367 null, in which case read-write data should be assumed.
6369 The default version if this function handles choosing code vs data,
6370 read-only vs read-write data, and @code{flag_pic}. You should only
6371 need to override this if your target has special flags that might be
6372 set via @code{__attribute__}.
6377 @subsection Output of Data
6380 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6381 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6382 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6383 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6384 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6385 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6386 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6387 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6388 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6389 These hooks specify assembly directives for creating certain kinds
6390 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6391 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6392 aligned two-byte object, and so on. Any of the hooks may be
6393 @code{NULL}, indicating that no suitable directive is available.
6395 The compiler will print these strings at the start of a new line,
6396 followed immediately by the object's initial value. In most cases,
6397 the string should contain a tab, a pseudo-op, and then another tab.
6400 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6401 The @code{assemble_integer} function uses this hook to output an
6402 integer object. @var{x} is the object's value, @var{size} is its size
6403 in bytes and @var{aligned_p} indicates whether it is aligned. The
6404 function should return @code{true} if it was able to output the
6405 object. If it returns false, @code{assemble_integer} will try to
6406 split the object into smaller parts.
6408 The default implementation of this hook will use the
6409 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6410 when the relevant string is @code{NULL}.
6413 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6414 A C statement to recognize @var{rtx} patterns that
6415 @code{output_addr_const} can't deal with, and output assembly code to
6416 @var{stream} corresponding to the pattern @var{x}. This may be used to
6417 allow machine-dependent @code{UNSPEC}s to appear within constants.
6419 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6420 @code{goto fail}, so that a standard error message is printed. If it
6421 prints an error message itself, by calling, for example,
6422 @code{output_operand_lossage}, it may just complete normally.
6425 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6426 A C statement to output to the stdio stream @var{stream} an assembler
6427 instruction to assemble a string constant containing the @var{len}
6428 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6429 @code{char *} and @var{len} a C expression of type @code{int}.
6431 If the assembler has a @code{.ascii} pseudo-op as found in the
6432 Berkeley Unix assembler, do not define the macro
6433 @code{ASM_OUTPUT_ASCII}.
6436 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6437 A C statement to output word @var{n} of a function descriptor for
6438 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6439 is defined, and is otherwise unused.
6442 @defmac CONSTANT_POOL_BEFORE_FUNCTION
6443 You may define this macro as a C expression. You should define the
6444 expression to have a nonzero value if GCC should output the constant
6445 pool for a function before the code for the function, or a zero value if
6446 GCC should output the constant pool after the function. If you do
6447 not define this macro, the usual case, GCC will output the constant
6448 pool before the function.
6451 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
6452 A C statement to output assembler commands to define the start of the
6453 constant pool for a function. @var{funname} is a string giving
6454 the name of the function. Should the return type of the function
6455 be required, it can be obtained via @var{fundecl}. @var{size}
6456 is the size, in bytes, of the constant pool that will be written
6457 immediately after this call.
6459 If no constant-pool prefix is required, the usual case, this macro need
6463 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
6464 A C statement (with or without semicolon) to output a constant in the
6465 constant pool, if it needs special treatment. (This macro need not do
6466 anything for RTL expressions that can be output normally.)
6468 The argument @var{file} is the standard I/O stream to output the
6469 assembler code on. @var{x} is the RTL expression for the constant to
6470 output, and @var{mode} is the machine mode (in case @var{x} is a
6471 @samp{const_int}). @var{align} is the required alignment for the value
6472 @var{x}; you should output an assembler directive to force this much
6475 The argument @var{labelno} is a number to use in an internal label for
6476 the address of this pool entry. The definition of this macro is
6477 responsible for outputting the label definition at the proper place.
6478 Here is how to do this:
6481 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
6484 When you output a pool entry specially, you should end with a
6485 @code{goto} to the label @var{jumpto}. This will prevent the same pool
6486 entry from being output a second time in the usual manner.
6488 You need not define this macro if it would do nothing.
6491 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
6492 A C statement to output assembler commands to at the end of the constant
6493 pool for a function. @var{funname} is a string giving the name of the
6494 function. Should the return type of the function be required, you can
6495 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
6496 constant pool that GCC wrote immediately before this call.
6498 If no constant-pool epilogue is required, the usual case, you need not
6502 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
6503 Define this macro as a C expression which is nonzero if @var{C} is
6504 used as a logical line separator by the assembler.
6506 If you do not define this macro, the default is that only
6507 the character @samp{;} is treated as a logical line separator.
6510 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
6511 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
6512 These target hooks are C string constants, describing the syntax in the
6513 assembler for grouping arithmetic expressions. If not overridden, they
6514 default to normal parentheses, which is correct for most assemblers.
6517 These macros are provided by @file{real.h} for writing the definitions
6518 of @code{ASM_OUTPUT_DOUBLE} and the like:
6520 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
6521 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
6522 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
6523 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
6524 floating point representation, and store its bit pattern in the variable
6525 @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE}, this variable should
6526 be a simple @code{long int}. For the others, it should be an array of
6527 @code{long int}. The number of elements in this array is determined by
6528 the size of the desired target floating point data type: 32 bits of it
6529 go in each @code{long int} array element. Each array element holds 32
6530 bits of the result, even if @code{long int} is wider than 32 bits on the
6533 The array element values are designed so that you can print them out
6534 using @code{fprintf} in the order they should appear in the target
6538 @node Uninitialized Data
6539 @subsection Output of Uninitialized Variables
6541 Each of the macros in this section is used to do the whole job of
6542 outputting a single uninitialized variable.
6544 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6545 A C statement (sans semicolon) to output to the stdio stream
6546 @var{stream} the assembler definition of a common-label named
6547 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6548 is the size rounded up to whatever alignment the caller wants.
6550 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6551 output the name itself; before and after that, output the additional
6552 assembler syntax for defining the name, and a newline.
6554 This macro controls how the assembler definitions of uninitialized
6555 common global variables are output.
6558 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
6559 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
6560 separate, explicit argument. If you define this macro, it is used in
6561 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
6562 handling the required alignment of the variable. The alignment is specified
6563 as the number of bits.
6566 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6567 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
6568 variable to be output, if there is one, or @code{NULL_TREE} if there
6569 is no corresponding variable. If you define this macro, GCC will use it
6570 in place of both @code{ASM_OUTPUT_COMMON} and
6571 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
6572 the variable's decl in order to chose what to output.
6575 @defmac ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6576 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
6577 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
6581 @defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6582 A C statement (sans semicolon) to output to the stdio stream
6583 @var{stream} the assembler definition of uninitialized global @var{decl} named
6584 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6585 is the size rounded up to whatever alignment the caller wants.
6587 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
6588 defining this macro. If unable, use the expression
6589 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
6590 before and after that, output the additional assembler syntax for defining
6591 the name, and a newline.
6593 This macro controls how the assembler definitions of uninitialized global
6594 variables are output. This macro exists to properly support languages like
6595 C++ which do not have @code{common} data. However, this macro currently
6596 is not defined for all targets. If this macro and
6597 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
6598 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
6599 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
6602 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6603 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
6604 separate, explicit argument. If you define this macro, it is used in
6605 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
6606 handling the required alignment of the variable. The alignment is specified
6607 as the number of bits.
6609 Try to use function @code{asm_output_aligned_bss} defined in file
6610 @file{varasm.c} when defining this macro.
6613 @defmac ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6614 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
6615 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
6619 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6620 A C statement (sans semicolon) to output to the stdio stream
6621 @var{stream} the assembler definition of a local-common-label named
6622 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6623 is the size rounded up to whatever alignment the caller wants.
6625 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6626 output the name itself; before and after that, output the additional
6627 assembler syntax for defining the name, and a newline.
6629 This macro controls how the assembler definitions of uninitialized
6630 static variables are output.
6633 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
6634 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
6635 separate, explicit argument. If you define this macro, it is used in
6636 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
6637 handling the required alignment of the variable. The alignment is specified
6638 as the number of bits.
6641 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6642 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
6643 variable to be output, if there is one, or @code{NULL_TREE} if there
6644 is no corresponding variable. If you define this macro, GCC will use it
6645 in place of both @code{ASM_OUTPUT_DECL} and
6646 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
6647 the variable's decl in order to chose what to output.
6650 @defmac ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6651 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
6652 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
6657 @subsection Output and Generation of Labels
6659 @c prevent bad page break with this line
6660 This is about outputting labels.
6662 @findex assemble_name
6663 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
6664 A C statement (sans semicolon) to output to the stdio stream
6665 @var{stream} the assembler definition of a label named @var{name}.
6666 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6667 output the name itself; before and after that, output the additional
6668 assembler syntax for defining the name, and a newline. A default
6669 definition of this macro is provided which is correct for most systems.
6673 A C string containing the appropriate assembler directive to specify the
6674 size of a symbol, without any arguments. On systems that use ELF, the
6675 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
6676 systems, the default is not to define this macro.
6678 Define this macro only if it is correct to use the default definitions
6679 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
6680 for your system. If you need your own custom definitions of those
6681 macros, or if you do not need explicit symbol sizes at all, do not
6685 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
6686 A C statement (sans semicolon) to output to the stdio stream
6687 @var{stream} a directive telling the assembler that the size of the
6688 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
6689 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6693 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
6694 A C statement (sans semicolon) to output to the stdio stream
6695 @var{stream} a directive telling the assembler to calculate the size of
6696 the symbol @var{name} by subtracting its address from the current
6699 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6700 provided. The default assumes that the assembler recognizes a special
6701 @samp{.} symbol as referring to the current address, and can calculate
6702 the difference between this and another symbol. If your assembler does
6703 not recognize @samp{.} or cannot do calculations with it, you will need
6704 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
6708 A C string containing the appropriate assembler directive to specify the
6709 type of a symbol, without any arguments. On systems that use ELF, the
6710 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
6711 systems, the default is not to define this macro.
6713 Define this macro only if it is correct to use the default definition of
6714 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6715 custom definition of this macro, or if you do not need explicit symbol
6716 types at all, do not define this macro.
6719 @defmac TYPE_OPERAND_FMT
6720 A C string which specifies (using @code{printf} syntax) the format of
6721 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
6722 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
6723 the default is not to define this macro.
6725 Define this macro only if it is correct to use the default definition of
6726 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6727 custom definition of this macro, or if you do not need explicit symbol
6728 types at all, do not define this macro.
6731 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
6732 A C statement (sans semicolon) to output to the stdio stream
6733 @var{stream} a directive telling the assembler that the type of the
6734 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
6735 that string is always either @samp{"function"} or @samp{"object"}, but
6736 you should not count on this.
6738 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
6739 definition of this macro is provided.
6742 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
6743 A C statement (sans semicolon) to output to the stdio stream
6744 @var{stream} any text necessary for declaring the name @var{name} of a
6745 function which is being defined. This macro is responsible for
6746 outputting the label definition (perhaps using
6747 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
6748 @code{FUNCTION_DECL} tree node representing the function.
6750 If this macro is not defined, then the function name is defined in the
6751 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6753 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6757 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
6758 A C statement (sans semicolon) to output to the stdio stream
6759 @var{stream} any text necessary for declaring the size of a function
6760 which is being defined. The argument @var{name} is the name of the
6761 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
6762 representing the function.
6764 If this macro is not defined, then the function size is not defined.
6766 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
6770 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6771 A C statement (sans semicolon) to output to the stdio stream
6772 @var{stream} any text necessary for declaring the name @var{name} of an
6773 initialized variable which is being defined. This macro must output the
6774 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6775 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6777 If this macro is not defined, then the variable name is defined in the
6778 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6780 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
6781 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
6784 @defmac ASM_DECLARE_CONSTANT_NAME (@var{stream}, @var{name}, @var{exp}, @var{size})
6785 A C statement (sans semicolon) to output to the stdio stream
6786 @var{stream} any text necessary for declaring the name @var{name} of a
6787 constant which is being defined. This macro is responsible for
6788 outputting the label definition (perhaps using
6789 @code{ASM_OUTPUT_LABEL}). The argument @var{exp} is the
6790 value of the constant, and @var{size} is the size of the constant
6791 in bytes. @var{name} will be an internal label.
6793 If this macro is not defined, then the @var{name} is defined in the
6794 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6796 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6800 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6801 A C statement (sans semicolon) to output to the stdio stream
6802 @var{stream} any text necessary for claiming a register @var{regno}
6803 for a global variable @var{decl} with name @var{name}.
6805 If you don't define this macro, that is equivalent to defining it to do
6809 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6810 A C statement (sans semicolon) to finish up declaring a variable name
6811 once the compiler has processed its initializer fully and thus has had a
6812 chance to determine the size of an array when controlled by an
6813 initializer. This is used on systems where it's necessary to declare
6814 something about the size of the object.
6816 If you don't define this macro, that is equivalent to defining it to do
6819 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
6820 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
6823 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
6824 This target hook is a function to output to the stdio stream
6825 @var{stream} some commands that will make the label @var{name} global;
6826 that is, available for reference from other files.
6828 The default implementation relies on a proper definition of
6829 @code{GLOBAL_ASM_OP}.
6832 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
6833 A C statement (sans semicolon) to output to the stdio stream
6834 @var{stream} some commands that will make the label @var{name} weak;
6835 that is, available for reference from other files but only used if
6836 no other definition is available. Use the expression
6837 @code{assemble_name (@var{stream}, @var{name})} to output the name
6838 itself; before and after that, output the additional assembler syntax
6839 for making that name weak, and a newline.
6841 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
6842 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
6846 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
6847 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
6848 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
6849 or variable decl. If @var{value} is not @code{NULL}, this C statement
6850 should output to the stdio stream @var{stream} assembler code which
6851 defines (equates) the weak symbol @var{name} to have the value
6852 @var{value}. If @var{value} is @code{NULL}, it should output commands
6853 to make @var{name} weak.
6856 @defmac SUPPORTS_WEAK
6857 A C expression which evaluates to true if the target supports weak symbols.
6859 If you don't define this macro, @file{defaults.h} provides a default
6860 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
6861 is defined, the default definition is @samp{1}; otherwise, it is
6862 @samp{0}. Define this macro if you want to control weak symbol support
6863 with a compiler flag such as @option{-melf}.
6866 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
6867 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
6868 public symbol such that extra copies in multiple translation units will
6869 be discarded by the linker. Define this macro if your object file
6870 format provides support for this concept, such as the @samp{COMDAT}
6871 section flags in the Microsoft Windows PE/COFF format, and this support
6872 requires changes to @var{decl}, such as putting it in a separate section.
6875 @defmac SUPPORTS_ONE_ONLY
6876 A C expression which evaluates to true if the target supports one-only
6879 If you don't define this macro, @file{varasm.c} provides a default
6880 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6881 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6882 you want to control one-only symbol support with a compiler flag, or if
6883 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6884 be emitted as one-only.
6887 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
6888 This target hook is a function to output to @var{asm_out_file} some
6889 commands that will make the symbol(s) associated with @var{decl} have
6890 hidden, protected or internal visibility as specified by @var{visibility}.
6893 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
6894 A C expression that evaluates to true if the target's linker expects
6895 that weak symbols do not appear in a static archive's table of contents.
6896 The default is @code{0}.
6898 Leaving weak symbols out of an archive's table of contents means that,
6899 if a symbol will only have a definition in one translation unit and
6900 will have undefined references from other translation units, that
6901 symbol should not be weak. Defining this macro to be nonzero will
6902 thus have the effect that certain symbols that would normally be weak
6903 (explicit template instantiations, and vtables for polymorphic classes
6904 with noninline key methods) will instead be nonweak.
6906 The C++ ABI requires this macro to be zero. Define this macro for
6907 targets where full C++ ABI compliance is impossible and where linker
6908 restrictions require weak symbols to be left out of a static archive's
6912 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6913 A C statement (sans semicolon) to output to the stdio stream
6914 @var{stream} any text necessary for declaring the name of an external
6915 symbol named @var{name} which is referenced in this compilation but
6916 not defined. The value of @var{decl} is the tree node for the
6919 This macro need not be defined if it does not need to output anything.
6920 The GNU assembler and most Unix assemblers don't require anything.
6923 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
6924 This target hook is a function to output to @var{asm_out_file} an assembler
6925 pseudo-op to declare a library function name external. The name of the
6926 library function is given by @var{symref}, which is a @code{symbol_ref}.
6929 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (tree @var{decl})
6930 This target hook is a function to output to @var{asm_out_file} an assembler
6931 directive to annotate used symbol. Darwin target use .no_dead_code_strip
6935 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6936 A C statement (sans semicolon) to output to the stdio stream
6937 @var{stream} a reference in assembler syntax to a label named
6938 @var{name}. This should add @samp{_} to the front of the name, if that
6939 is customary on your operating system, as it is in most Berkeley Unix
6940 systems. This macro is used in @code{assemble_name}.
6943 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6944 A C statement (sans semicolon) to output a reference to
6945 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
6946 will be used to output the name of the symbol. This macro may be used
6947 to modify the way a symbol is referenced depending on information
6948 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
6951 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
6952 A C statement (sans semicolon) to output a reference to @var{buf}, the
6953 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
6954 @code{assemble_name} will be used to output the name of the symbol.
6955 This macro is not used by @code{output_asm_label}, or the @code{%l}
6956 specifier that calls it; the intention is that this macro should be set
6957 when it is necessary to output a label differently when its address is
6961 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
6962 A function to output to the stdio stream @var{stream} a label whose
6963 name is made from the string @var{prefix} and the number @var{labelno}.
6965 It is absolutely essential that these labels be distinct from the labels
6966 used for user-level functions and variables. Otherwise, certain programs
6967 will have name conflicts with internal labels.
6969 It is desirable to exclude internal labels from the symbol table of the
6970 object file. Most assemblers have a naming convention for labels that
6971 should be excluded; on many systems, the letter @samp{L} at the
6972 beginning of a label has this effect. You should find out what
6973 convention your system uses, and follow it.
6975 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
6978 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
6979 A C statement to output to the stdio stream @var{stream} a debug info
6980 label whose name is made from the string @var{prefix} and the number
6981 @var{num}. This is useful for VLIW targets, where debug info labels
6982 may need to be treated differently than branch target labels. On some
6983 systems, branch target labels must be at the beginning of instruction
6984 bundles, but debug info labels can occur in the middle of instruction
6987 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
6991 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
6992 A C statement to store into the string @var{string} a label whose name
6993 is made from the string @var{prefix} and the number @var{num}.
6995 This string, when output subsequently by @code{assemble_name}, should
6996 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
6997 with the same @var{prefix} and @var{num}.
6999 If the string begins with @samp{*}, then @code{assemble_name} will
7000 output the rest of the string unchanged. It is often convenient for
7001 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
7002 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
7003 to output the string, and may change it. (Of course,
7004 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
7005 you should know what it does on your machine.)
7008 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
7009 A C expression to assign to @var{outvar} (which is a variable of type
7010 @code{char *}) a newly allocated string made from the string
7011 @var{name} and the number @var{number}, with some suitable punctuation
7012 added. Use @code{alloca} to get space for the string.
7014 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
7015 produce an assembler label for an internal static variable whose name is
7016 @var{name}. Therefore, the string must be such as to result in valid
7017 assembler code. The argument @var{number} is different each time this
7018 macro is executed; it prevents conflicts between similarly-named
7019 internal static variables in different scopes.
7021 Ideally this string should not be a valid C identifier, to prevent any
7022 conflict with the user's own symbols. Most assemblers allow periods
7023 or percent signs in assembler symbols; putting at least one of these
7024 between the name and the number will suffice.
7026 If this macro is not defined, a default definition will be provided
7027 which is correct for most systems.
7030 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
7031 A C statement to output to the stdio stream @var{stream} assembler code
7032 which defines (equates) the symbol @var{name} to have the value @var{value}.
7035 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7036 correct for most systems.
7039 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
7040 A C statement to output to the stdio stream @var{stream} assembler code
7041 which defines (equates) the symbol whose tree node is @var{decl_of_name}
7042 to have the value of the tree node @var{decl_of_value}. This macro will
7043 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
7044 the tree nodes are available.
7047 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7048 correct for most systems.
7051 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
7052 A C statement to output to the stdio stream @var{stream} assembler code
7053 which defines (equates) the weak symbol @var{name} to have the value
7054 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
7055 an undefined weak symbol.
7057 Define this macro if the target only supports weak aliases; define
7058 @code{ASM_OUTPUT_DEF} instead if possible.
7061 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
7062 Define this macro to override the default assembler names used for
7063 Objective-C methods.
7065 The default name is a unique method number followed by the name of the
7066 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
7067 the category is also included in the assembler name (e.g.@:
7070 These names are safe on most systems, but make debugging difficult since
7071 the method's selector is not present in the name. Therefore, particular
7072 systems define other ways of computing names.
7074 @var{buf} is an expression of type @code{char *} which gives you a
7075 buffer in which to store the name; its length is as long as
7076 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
7077 50 characters extra.
7079 The argument @var{is_inst} specifies whether the method is an instance
7080 method or a class method; @var{class_name} is the name of the class;
7081 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
7082 in a category); and @var{sel_name} is the name of the selector.
7084 On systems where the assembler can handle quoted names, you can use this
7085 macro to provide more human-readable names.
7088 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
7089 A C statement (sans semicolon) to output to the stdio stream
7090 @var{stream} commands to declare that the label @var{name} is an
7091 Objective-C class reference. This is only needed for targets whose
7092 linkers have special support for NeXT-style runtimes.
7095 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
7096 A C statement (sans semicolon) to output to the stdio stream
7097 @var{stream} commands to declare that the label @var{name} is an
7098 unresolved Objective-C class reference. This is only needed for targets
7099 whose linkers have special support for NeXT-style runtimes.
7102 @node Initialization
7103 @subsection How Initialization Functions Are Handled
7104 @cindex initialization routines
7105 @cindex termination routines
7106 @cindex constructors, output of
7107 @cindex destructors, output of
7109 The compiled code for certain languages includes @dfn{constructors}
7110 (also called @dfn{initialization routines})---functions to initialize
7111 data in the program when the program is started. These functions need
7112 to be called before the program is ``started''---that is to say, before
7113 @code{main} is called.
7115 Compiling some languages generates @dfn{destructors} (also called
7116 @dfn{termination routines}) that should be called when the program
7119 To make the initialization and termination functions work, the compiler
7120 must output something in the assembler code to cause those functions to
7121 be called at the appropriate time. When you port the compiler to a new
7122 system, you need to specify how to do this.
7124 There are two major ways that GCC currently supports the execution of
7125 initialization and termination functions. Each way has two variants.
7126 Much of the structure is common to all four variations.
7128 @findex __CTOR_LIST__
7129 @findex __DTOR_LIST__
7130 The linker must build two lists of these functions---a list of
7131 initialization functions, called @code{__CTOR_LIST__}, and a list of
7132 termination functions, called @code{__DTOR_LIST__}.
7134 Each list always begins with an ignored function pointer (which may hold
7135 0, @minus{}1, or a count of the function pointers after it, depending on
7136 the environment). This is followed by a series of zero or more function
7137 pointers to constructors (or destructors), followed by a function
7138 pointer containing zero.
7140 Depending on the operating system and its executable file format, either
7141 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
7142 time and exit time. Constructors are called in reverse order of the
7143 list; destructors in forward order.
7145 The best way to handle static constructors works only for object file
7146 formats which provide arbitrarily-named sections. A section is set
7147 aside for a list of constructors, and another for a list of destructors.
7148 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
7149 object file that defines an initialization function also puts a word in
7150 the constructor section to point to that function. The linker
7151 accumulates all these words into one contiguous @samp{.ctors} section.
7152 Termination functions are handled similarly.
7154 This method will be chosen as the default by @file{target-def.h} if
7155 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
7156 support arbitrary sections, but does support special designated
7157 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
7158 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
7160 When arbitrary sections are available, there are two variants, depending
7161 upon how the code in @file{crtstuff.c} is called. On systems that
7162 support a @dfn{.init} section which is executed at program startup,
7163 parts of @file{crtstuff.c} are compiled into that section. The
7164 program is linked by the @command{gcc} driver like this:
7167 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
7170 The prologue of a function (@code{__init}) appears in the @code{.init}
7171 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
7172 for the function @code{__fini} in the @dfn{.fini} section. Normally these
7173 files are provided by the operating system or by the GNU C library, but
7174 are provided by GCC for a few targets.
7176 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
7177 compiled from @file{crtstuff.c}. They contain, among other things, code
7178 fragments within the @code{.init} and @code{.fini} sections that branch
7179 to routines in the @code{.text} section. The linker will pull all parts
7180 of a section together, which results in a complete @code{__init} function
7181 that invokes the routines we need at startup.
7183 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
7186 If no init section is available, when GCC compiles any function called
7187 @code{main} (or more accurately, any function designated as a program
7188 entry point by the language front end calling @code{expand_main_function}),
7189 it inserts a procedure call to @code{__main} as the first executable code
7190 after the function prologue. The @code{__main} function is defined
7191 in @file{libgcc2.c} and runs the global constructors.
7193 In file formats that don't support arbitrary sections, there are again
7194 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
7195 and an `a.out' format must be used. In this case,
7196 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
7197 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
7198 and with the address of the void function containing the initialization
7199 code as its value. The GNU linker recognizes this as a request to add
7200 the value to a @dfn{set}; the values are accumulated, and are eventually
7201 placed in the executable as a vector in the format described above, with
7202 a leading (ignored) count and a trailing zero element.
7203 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
7204 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
7205 the compilation of @code{main} to call @code{__main} as above, starting
7206 the initialization process.
7208 The last variant uses neither arbitrary sections nor the GNU linker.
7209 This is preferable when you want to do dynamic linking and when using
7210 file formats which the GNU linker does not support, such as `ECOFF'@. In
7211 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
7212 termination functions are recognized simply by their names. This requires
7213 an extra program in the linkage step, called @command{collect2}. This program
7214 pretends to be the linker, for use with GCC; it does its job by running
7215 the ordinary linker, but also arranges to include the vectors of
7216 initialization and termination functions. These functions are called
7217 via @code{__main} as described above. In order to use this method,
7218 @code{use_collect2} must be defined in the target in @file{config.gcc}.
7221 The following section describes the specific macros that control and
7222 customize the handling of initialization and termination functions.
7225 @node Macros for Initialization
7226 @subsection Macros Controlling Initialization Routines
7228 Here are the macros that control how the compiler handles initialization
7229 and termination functions:
7231 @defmac INIT_SECTION_ASM_OP
7232 If defined, a C string constant, including spacing, for the assembler
7233 operation to identify the following data as initialization code. If not
7234 defined, GCC will assume such a section does not exist. When you are
7235 using special sections for initialization and termination functions, this
7236 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7237 run the initialization functions.
7240 @defmac HAS_INIT_SECTION
7241 If defined, @code{main} will not call @code{__main} as described above.
7242 This macro should be defined for systems that control start-up code
7243 on a symbol-by-symbol basis, such as OSF/1, and should not
7244 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7247 @defmac LD_INIT_SWITCH
7248 If defined, a C string constant for a switch that tells the linker that
7249 the following symbol is an initialization routine.
7252 @defmac LD_FINI_SWITCH
7253 If defined, a C string constant for a switch that tells the linker that
7254 the following symbol is a finalization routine.
7257 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7258 If defined, a C statement that will write a function that can be
7259 automatically called when a shared library is loaded. The function
7260 should call @var{func}, which takes no arguments. If not defined, and
7261 the object format requires an explicit initialization function, then a
7262 function called @code{_GLOBAL__DI} will be generated.
7264 This function and the following one are used by collect2 when linking a
7265 shared library that needs constructors or destructors, or has DWARF2
7266 exception tables embedded in the code.
7269 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7270 If defined, a C statement that will write a function that can be
7271 automatically called when a shared library is unloaded. The function
7272 should call @var{func}, which takes no arguments. If not defined, and
7273 the object format requires an explicit finalization function, then a
7274 function called @code{_GLOBAL__DD} will be generated.
7277 @defmac INVOKE__main
7278 If defined, @code{main} will call @code{__main} despite the presence of
7279 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7280 where the init section is not actually run automatically, but is still
7281 useful for collecting the lists of constructors and destructors.
7284 @defmac SUPPORTS_INIT_PRIORITY
7285 If nonzero, the C++ @code{init_priority} attribute is supported and the
7286 compiler should emit instructions to control the order of initialization
7287 of objects. If zero, the compiler will issue an error message upon
7288 encountering an @code{init_priority} attribute.
7291 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7292 This value is true if the target supports some ``native'' method of
7293 collecting constructors and destructors to be run at startup and exit.
7294 It is false if we must use @command{collect2}.
7297 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7298 If defined, a function that outputs assembler code to arrange to call
7299 the function referenced by @var{symbol} at initialization time.
7301 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7302 no arguments and with no return value. If the target supports initialization
7303 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7304 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7306 If this macro is not defined by the target, a suitable default will
7307 be chosen if (1) the target supports arbitrary section names, (2) the
7308 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7312 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7313 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7314 functions rather than initialization functions.
7317 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7318 generated for the generated object file will have static linkage.
7320 If your system uses @command{collect2} as the means of processing
7321 constructors, then that program normally uses @command{nm} to scan
7322 an object file for constructor functions to be called.
7324 On certain kinds of systems, you can define this macro to make
7325 @command{collect2} work faster (and, in some cases, make it work at all):
7327 @defmac OBJECT_FORMAT_COFF
7328 Define this macro if the system uses COFF (Common Object File Format)
7329 object files, so that @command{collect2} can assume this format and scan
7330 object files directly for dynamic constructor/destructor functions.
7332 This macro is effective only in a native compiler; @command{collect2} as
7333 part of a cross compiler always uses @command{nm} for the target machine.
7336 @defmac REAL_NM_FILE_NAME
7337 Define this macro as a C string constant containing the file name to use
7338 to execute @command{nm}. The default is to search the path normally for
7341 If your system supports shared libraries and has a program to list the
7342 dynamic dependencies of a given library or executable, you can define
7343 these macros to enable support for running initialization and
7344 termination functions in shared libraries:
7348 Define this macro to a C string constant containing the name of the program
7349 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7352 @defmac PARSE_LDD_OUTPUT (@var{ptr})
7353 Define this macro to be C code that extracts filenames from the output
7354 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7355 of type @code{char *} that points to the beginning of a line of output
7356 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7357 code must advance @var{ptr} to the beginning of the filename on that
7358 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7361 @node Instruction Output
7362 @subsection Output of Assembler Instructions
7364 @c prevent bad page break with this line
7365 This describes assembler instruction output.
7367 @defmac REGISTER_NAMES
7368 A C initializer containing the assembler's names for the machine
7369 registers, each one as a C string constant. This is what translates
7370 register numbers in the compiler into assembler language.
7373 @defmac ADDITIONAL_REGISTER_NAMES
7374 If defined, a C initializer for an array of structures containing a name
7375 and a register number. This macro defines additional names for hard
7376 registers, thus allowing the @code{asm} option in declarations to refer
7377 to registers using alternate names.
7380 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7381 Define this macro if you are using an unusual assembler that
7382 requires different names for the machine instructions.
7384 The definition is a C statement or statements which output an
7385 assembler instruction opcode to the stdio stream @var{stream}. The
7386 macro-operand @var{ptr} is a variable of type @code{char *} which
7387 points to the opcode name in its ``internal'' form---the form that is
7388 written in the machine description. The definition should output the
7389 opcode name to @var{stream}, performing any translation you desire, and
7390 increment the variable @var{ptr} to point at the end of the opcode
7391 so that it will not be output twice.
7393 In fact, your macro definition may process less than the entire opcode
7394 name, or more than the opcode name; but if you want to process text
7395 that includes @samp{%}-sequences to substitute operands, you must take
7396 care of the substitution yourself. Just be sure to increment
7397 @var{ptr} over whatever text should not be output normally.
7399 @findex recog_data.operand
7400 If you need to look at the operand values, they can be found as the
7401 elements of @code{recog_data.operand}.
7403 If the macro definition does nothing, the instruction is output
7407 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7408 If defined, a C statement to be executed just prior to the output of
7409 assembler code for @var{insn}, to modify the extracted operands so
7410 they will be output differently.
7412 Here the argument @var{opvec} is the vector containing the operands
7413 extracted from @var{insn}, and @var{noperands} is the number of
7414 elements of the vector which contain meaningful data for this insn.
7415 The contents of this vector are what will be used to convert the insn
7416 template into assembler code, so you can change the assembler output
7417 by changing the contents of the vector.
7419 This macro is useful when various assembler syntaxes share a single
7420 file of instruction patterns; by defining this macro differently, you
7421 can cause a large class of instructions to be output differently (such
7422 as with rearranged operands). Naturally, variations in assembler
7423 syntax affecting individual insn patterns ought to be handled by
7424 writing conditional output routines in those patterns.
7426 If this macro is not defined, it is equivalent to a null statement.
7429 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
7430 A C compound statement to output to stdio stream @var{stream} the
7431 assembler syntax for an instruction operand @var{x}. @var{x} is an
7434 @var{code} is a value that can be used to specify one of several ways
7435 of printing the operand. It is used when identical operands must be
7436 printed differently depending on the context. @var{code} comes from
7437 the @samp{%} specification that was used to request printing of the
7438 operand. If the specification was just @samp{%@var{digit}} then
7439 @var{code} is 0; if the specification was @samp{%@var{ltr}
7440 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
7443 If @var{x} is a register, this macro should print the register's name.
7444 The names can be found in an array @code{reg_names} whose type is
7445 @code{char *[]}. @code{reg_names} is initialized from
7446 @code{REGISTER_NAMES}.
7448 When the machine description has a specification @samp{%@var{punct}}
7449 (a @samp{%} followed by a punctuation character), this macro is called
7450 with a null pointer for @var{x} and the punctuation character for
7454 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
7455 A C expression which evaluates to true if @var{code} is a valid
7456 punctuation character for use in the @code{PRINT_OPERAND} macro. If
7457 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
7458 punctuation characters (except for the standard one, @samp{%}) are used
7462 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
7463 A C compound statement to output to stdio stream @var{stream} the
7464 assembler syntax for an instruction operand that is a memory reference
7465 whose address is @var{x}. @var{x} is an RTL expression.
7467 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
7468 On some machines, the syntax for a symbolic address depends on the
7469 section that the address refers to. On these machines, define the hook
7470 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
7471 @code{symbol_ref}, and then check for it here. @xref{Assembler
7475 @findex dbr_sequence_length
7476 @defmac DBR_OUTPUT_SEQEND (@var{file})
7477 A C statement, to be executed after all slot-filler instructions have
7478 been output. If necessary, call @code{dbr_sequence_length} to
7479 determine the number of slots filled in a sequence (zero if not
7480 currently outputting a sequence), to decide how many no-ops to output,
7483 Don't define this macro if it has nothing to do, but it is helpful in
7484 reading assembly output if the extent of the delay sequence is made
7485 explicit (e.g.@: with white space).
7488 @findex final_sequence
7489 Note that output routines for instructions with delay slots must be
7490 prepared to deal with not being output as part of a sequence
7491 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
7492 found.) The variable @code{final_sequence} is null when not
7493 processing a sequence, otherwise it contains the @code{sequence} rtx
7497 @defmac REGISTER_PREFIX
7498 @defmacx LOCAL_LABEL_PREFIX
7499 @defmacx USER_LABEL_PREFIX
7500 @defmacx IMMEDIATE_PREFIX
7501 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
7502 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
7503 @file{final.c}). These are useful when a single @file{md} file must
7504 support multiple assembler formats. In that case, the various @file{tm.h}
7505 files can define these macros differently.
7508 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
7509 If defined this macro should expand to a series of @code{case}
7510 statements which will be parsed inside the @code{switch} statement of
7511 the @code{asm_fprintf} function. This allows targets to define extra
7512 printf formats which may useful when generating their assembler
7513 statements. Note that uppercase letters are reserved for future
7514 generic extensions to asm_fprintf, and so are not available to target
7515 specific code. The output file is given by the parameter @var{file}.
7516 The varargs input pointer is @var{argptr} and the rest of the format
7517 string, starting the character after the one that is being switched
7518 upon, is pointed to by @var{format}.
7521 @defmac ASSEMBLER_DIALECT
7522 If your target supports multiple dialects of assembler language (such as
7523 different opcodes), define this macro as a C expression that gives the
7524 numeric index of the assembler language dialect to use, with zero as the
7527 If this macro is defined, you may use constructs of the form
7529 @samp{@{option0|option1|option2@dots{}@}}
7532 in the output templates of patterns (@pxref{Output Template}) or in the
7533 first argument of @code{asm_fprintf}. This construct outputs
7534 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
7535 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
7536 within these strings retain their usual meaning. If there are fewer
7537 alternatives within the braces than the value of
7538 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
7540 If you do not define this macro, the characters @samp{@{}, @samp{|} and
7541 @samp{@}} do not have any special meaning when used in templates or
7542 operands to @code{asm_fprintf}.
7544 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
7545 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
7546 the variations in assembler language syntax with that mechanism. Define
7547 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
7548 if the syntax variant are larger and involve such things as different
7549 opcodes or operand order.
7552 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
7553 A C expression to output to @var{stream} some assembler code
7554 which will push hard register number @var{regno} onto the stack.
7555 The code need not be optimal, since this macro is used only when
7559 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
7560 A C expression to output to @var{stream} some assembler code
7561 which will pop hard register number @var{regno} off of the stack.
7562 The code need not be optimal, since this macro is used only when
7566 @node Dispatch Tables
7567 @subsection Output of Dispatch Tables
7569 @c prevent bad page break with this line
7570 This concerns dispatch tables.
7572 @cindex dispatch table
7573 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
7574 A C statement to output to the stdio stream @var{stream} an assembler
7575 pseudo-instruction to generate a difference between two labels.
7576 @var{value} and @var{rel} are the numbers of two internal labels. The
7577 definitions of these labels are output using
7578 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
7579 way here. For example,
7582 fprintf (@var{stream}, "\t.word L%d-L%d\n",
7583 @var{value}, @var{rel})
7586 You must provide this macro on machines where the addresses in a
7587 dispatch table are relative to the table's own address. If defined, GCC
7588 will also use this macro on all machines when producing PIC@.
7589 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
7590 mode and flags can be read.
7593 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
7594 This macro should be provided on machines where the addresses
7595 in a dispatch table are absolute.
7597 The definition should be a C statement to output to the stdio stream
7598 @var{stream} an assembler pseudo-instruction to generate a reference to
7599 a label. @var{value} is the number of an internal label whose
7600 definition is output using @code{(*targetm.asm_out.internal_label)}.
7604 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
7608 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
7609 Define this if the label before a jump-table needs to be output
7610 specially. The first three arguments are the same as for
7611 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
7612 jump-table which follows (a @code{jump_insn} containing an
7613 @code{addr_vec} or @code{addr_diff_vec}).
7615 This feature is used on system V to output a @code{swbeg} statement
7618 If this macro is not defined, these labels are output with
7619 @code{(*targetm.asm_out.internal_label)}.
7622 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
7623 Define this if something special must be output at the end of a
7624 jump-table. The definition should be a C statement to be executed
7625 after the assembler code for the table is written. It should write
7626 the appropriate code to stdio stream @var{stream}. The argument
7627 @var{table} is the jump-table insn, and @var{num} is the label-number
7628 of the preceding label.
7630 If this macro is not defined, nothing special is output at the end of
7634 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (@var{stream}, @var{decl}, @var{for_eh}, @var{empty})
7635 This target hook emits a label at the beginning of each FDE@. It
7636 should be defined on targets where FDEs need special labels, and it
7637 should write the appropriate label, for the FDE associated with the
7638 function declaration @var{decl}, to the stdio stream @var{stream}.
7639 The third argument, @var{for_eh}, is a boolean: true if this is for an
7640 exception table. The fourth argument, @var{empty}, is a boolean:
7641 true if this is a placeholder label for an omitted FDE@.
7643 The default is that FDEs are not given nonlocal labels.
7646 @deftypefn {Taget Hook} void TARGET_UNWIND_EMIT (FILE * @var{stream}, rtx @var{insn})
7647 This target hook emits and assembly directives required to unwind the
7648 given instruction. This is only used when TARGET_UNWIND_INFO is set.
7651 @node Exception Region Output
7652 @subsection Assembler Commands for Exception Regions
7654 @c prevent bad page break with this line
7656 This describes commands marking the start and the end of an exception
7659 @defmac EH_FRAME_SECTION_NAME
7660 If defined, a C string constant for the name of the section containing
7661 exception handling frame unwind information. If not defined, GCC will
7662 provide a default definition if the target supports named sections.
7663 @file{crtstuff.c} uses this macro to switch to the appropriate section.
7665 You should define this symbol if your target supports DWARF 2 frame
7666 unwind information and the default definition does not work.
7669 @defmac EH_FRAME_IN_DATA_SECTION
7670 If defined, DWARF 2 frame unwind information will be placed in the
7671 data section even though the target supports named sections. This
7672 might be necessary, for instance, if the system linker does garbage
7673 collection and sections cannot be marked as not to be collected.
7675 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
7679 @defmac EH_TABLES_CAN_BE_READ_ONLY
7680 Define this macro to 1 if your target is such that no frame unwind
7681 information encoding used with non-PIC code will ever require a
7682 runtime relocation, but the linker may not support merging read-only
7683 and read-write sections into a single read-write section.
7686 @defmac MASK_RETURN_ADDR
7687 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
7688 that it does not contain any extraneous set bits in it.
7691 @defmac DWARF2_UNWIND_INFO
7692 Define this macro to 0 if your target supports DWARF 2 frame unwind
7693 information, but it does not yet work with exception handling.
7694 Otherwise, if your target supports this information (if it defines
7695 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
7696 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
7699 If @code{TARGET_UNWIND_INFO} is defined, the target specific unwinder
7700 will be used in all cases. Defining this macro will enable the generation
7701 of DWARF 2 frame debugging information.
7703 If @code{TARGET_UNWIND_INFO} is not defined, and this macro is defined to 1,
7704 the DWARF 2 unwinder will be the default exception handling mechanism;
7705 otherwise, @code{setjmp}/@code{longjmp} will be used by default.
7708 @defmac TARGET_UNWIND_INFO
7709 Define this macro if your target has ABI specified unwind tables. Usually
7710 these will be output by @code{TARGET_UNWIND_EMIT}.
7713 @defmac MUST_USE_SJLJ_EXCEPTIONS
7714 This macro need only be defined if @code{DWARF2_UNWIND_INFO} is
7715 runtime-variable. In that case, @file{except.h} cannot correctly
7716 determine the corresponding definition of
7717 @code{MUST_USE_SJLJ_EXCEPTIONS}, so the target must provide it directly.
7720 @defmac DWARF_CIE_DATA_ALIGNMENT
7721 This macro need only be defined if the target might save registers in the
7722 function prologue at an offset to the stack pointer that is not aligned to
7723 @code{UNITS_PER_WORD}. The definition should be the negative minimum
7724 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
7725 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
7726 the target supports DWARF 2 frame unwind information.
7729 @deftypefn {Target Hook} void TARGET_ASM_EXCEPTION_SECTION ()
7730 If defined, a function that switches to the section in which the main
7731 exception table is to be placed (@pxref{Sections}). The default is a
7732 function that switches to a section named @code{.gcc_except_table} on
7733 machines that support named sections via
7734 @code{TARGET_ASM_NAMED_SECTION}, otherwise if @option{-fpic} or
7735 @option{-fPIC} is in effect, the @code{data_section}, otherwise the
7736 @code{readonly_data_section}.
7739 @deftypefn {Target Hook} void TARGET_ASM_EH_FRAME_SECTION ()
7740 If defined, a function that switches to the section in which the DWARF 2
7741 frame unwind information to be placed (@pxref{Sections}). The default
7742 is a function that outputs a standard GAS section directive, if
7743 @code{EH_FRAME_SECTION_NAME} is defined, or else a data section
7744 directive followed by a synthetic label.
7747 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
7748 Contains the value true if the target should add a zero word onto the
7749 end of a Dwarf-2 frame info section when used for exception handling.
7750 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
7754 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
7755 Given a register, this hook should return a parallel of registers to
7756 represent where to find the register pieces. Define this hook if the
7757 register and its mode are represented in Dwarf in non-contiguous
7758 locations, or if the register should be represented in more than one
7759 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
7760 If not defined, the default is to return @code{NULL_RTX}.
7763 @node Alignment Output
7764 @subsection Assembler Commands for Alignment
7766 @c prevent bad page break with this line
7767 This describes commands for alignment.
7769 @defmac JUMP_ALIGN (@var{label})
7770 The alignment (log base 2) to put in front of @var{label}, which is
7771 a common destination of jumps and has no fallthru incoming edge.
7773 This macro need not be defined if you don't want any special alignment
7774 to be done at such a time. Most machine descriptions do not currently
7777 Unless it's necessary to inspect the @var{label} parameter, it is better
7778 to set the variable @var{align_jumps} in the target's
7779 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7780 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
7783 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
7784 The alignment (log base 2) to put in front of @var{label}, which follows
7787 This macro need not be defined if you don't want any special alignment
7788 to be done at such a time. Most machine descriptions do not currently
7792 @defmac LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7793 The maximum number of bytes to skip when applying
7794 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
7795 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7798 @defmac LOOP_ALIGN (@var{label})
7799 The alignment (log base 2) to put in front of @var{label}, which follows
7800 a @code{NOTE_INSN_LOOP_BEG} note.
7802 This macro need not be defined if you don't want any special alignment
7803 to be done at such a time. Most machine descriptions do not currently
7806 Unless it's necessary to inspect the @var{label} parameter, it is better
7807 to set the variable @code{align_loops} in the target's
7808 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7809 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
7812 @defmac LOOP_ALIGN_MAX_SKIP
7813 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
7814 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7817 @defmac LABEL_ALIGN (@var{label})
7818 The alignment (log base 2) to put in front of @var{label}.
7819 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
7820 the maximum of the specified values is used.
7822 Unless it's necessary to inspect the @var{label} parameter, it is better
7823 to set the variable @code{align_labels} in the target's
7824 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7825 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
7828 @defmac LABEL_ALIGN_MAX_SKIP
7829 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
7830 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7833 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
7834 A C statement to output to the stdio stream @var{stream} an assembler
7835 instruction to advance the location counter by @var{nbytes} bytes.
7836 Those bytes should be zero when loaded. @var{nbytes} will be a C
7837 expression of type @code{int}.
7840 @defmac ASM_NO_SKIP_IN_TEXT
7841 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
7842 text section because it fails to put zeros in the bytes that are skipped.
7843 This is true on many Unix systems, where the pseudo--op to skip bytes
7844 produces no-op instructions rather than zeros when used in the text
7848 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
7849 A C statement to output to the stdio stream @var{stream} an assembler
7850 command to advance the location counter to a multiple of 2 to the
7851 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
7854 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
7855 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
7856 for padding, if necessary.
7859 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
7860 A C statement to output to the stdio stream @var{stream} an assembler
7861 command to advance the location counter to a multiple of 2 to the
7862 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
7863 satisfy the alignment request. @var{power} and @var{max_skip} will be
7864 a C expression of type @code{int}.
7868 @node Debugging Info
7869 @section Controlling Debugging Information Format
7871 @c prevent bad page break with this line
7872 This describes how to specify debugging information.
7875 * All Debuggers:: Macros that affect all debugging formats uniformly.
7876 * DBX Options:: Macros enabling specific options in DBX format.
7877 * DBX Hooks:: Hook macros for varying DBX format.
7878 * File Names and DBX:: Macros controlling output of file names in DBX format.
7879 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
7880 * VMS Debug:: Macros for VMS debug format.
7884 @subsection Macros Affecting All Debugging Formats
7886 @c prevent bad page break with this line
7887 These macros affect all debugging formats.
7889 @defmac DBX_REGISTER_NUMBER (@var{regno})
7890 A C expression that returns the DBX register number for the compiler
7891 register number @var{regno}. In the default macro provided, the value
7892 of this expression will be @var{regno} itself. But sometimes there are
7893 some registers that the compiler knows about and DBX does not, or vice
7894 versa. In such cases, some register may need to have one number in the
7895 compiler and another for DBX@.
7897 If two registers have consecutive numbers inside GCC, and they can be
7898 used as a pair to hold a multiword value, then they @emph{must} have
7899 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
7900 Otherwise, debuggers will be unable to access such a pair, because they
7901 expect register pairs to be consecutive in their own numbering scheme.
7903 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
7904 does not preserve register pairs, then what you must do instead is
7905 redefine the actual register numbering scheme.
7908 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
7909 A C expression that returns the integer offset value for an automatic
7910 variable having address @var{x} (an RTL expression). The default
7911 computation assumes that @var{x} is based on the frame-pointer and
7912 gives the offset from the frame-pointer. This is required for targets
7913 that produce debugging output for DBX or COFF-style debugging output
7914 for SDB and allow the frame-pointer to be eliminated when the
7915 @option{-g} options is used.
7918 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
7919 A C expression that returns the integer offset value for an argument
7920 having address @var{x} (an RTL expression). The nominal offset is
7924 @defmac PREFERRED_DEBUGGING_TYPE
7925 A C expression that returns the type of debugging output GCC should
7926 produce when the user specifies just @option{-g}. Define
7927 this if you have arranged for GCC to support more than one format of
7928 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7929 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
7930 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
7932 When the user specifies @option{-ggdb}, GCC normally also uses the
7933 value of this macro to select the debugging output format, but with two
7934 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
7935 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7936 defined, GCC uses @code{DBX_DEBUG}.
7938 The value of this macro only affects the default debugging output; the
7939 user can always get a specific type of output by using @option{-gstabs},
7940 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
7944 @subsection Specific Options for DBX Output
7946 @c prevent bad page break with this line
7947 These are specific options for DBX output.
7949 @defmac DBX_DEBUGGING_INFO
7950 Define this macro if GCC should produce debugging output for DBX
7951 in response to the @option{-g} option.
7954 @defmac XCOFF_DEBUGGING_INFO
7955 Define this macro if GCC should produce XCOFF format debugging output
7956 in response to the @option{-g} option. This is a variant of DBX format.
7959 @defmac DEFAULT_GDB_EXTENSIONS
7960 Define this macro to control whether GCC should by default generate
7961 GDB's extended version of DBX debugging information (assuming DBX-format
7962 debugging information is enabled at all). If you don't define the
7963 macro, the default is 1: always generate the extended information
7964 if there is any occasion to.
7967 @defmac DEBUG_SYMS_TEXT
7968 Define this macro if all @code{.stabs} commands should be output while
7969 in the text section.
7972 @defmac ASM_STABS_OP
7973 A C string constant, including spacing, naming the assembler pseudo op to
7974 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
7975 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
7976 applies only to DBX debugging information format.
7979 @defmac ASM_STABD_OP
7980 A C string constant, including spacing, naming the assembler pseudo op to
7981 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
7982 value is the current location. If you don't define this macro,
7983 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
7987 @defmac ASM_STABN_OP
7988 A C string constant, including spacing, naming the assembler pseudo op to
7989 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
7990 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
7991 macro applies only to DBX debugging information format.
7994 @defmac DBX_NO_XREFS
7995 Define this macro if DBX on your system does not support the construct
7996 @samp{xs@var{tagname}}. On some systems, this construct is used to
7997 describe a forward reference to a structure named @var{tagname}.
7998 On other systems, this construct is not supported at all.
8001 @defmac DBX_CONTIN_LENGTH
8002 A symbol name in DBX-format debugging information is normally
8003 continued (split into two separate @code{.stabs} directives) when it
8004 exceeds a certain length (by default, 80 characters). On some
8005 operating systems, DBX requires this splitting; on others, splitting
8006 must not be done. You can inhibit splitting by defining this macro
8007 with the value zero. You can override the default splitting-length by
8008 defining this macro as an expression for the length you desire.
8011 @defmac DBX_CONTIN_CHAR
8012 Normally continuation is indicated by adding a @samp{\} character to
8013 the end of a @code{.stabs} string when a continuation follows. To use
8014 a different character instead, define this macro as a character
8015 constant for the character you want to use. Do not define this macro
8016 if backslash is correct for your system.
8019 @defmac DBX_STATIC_STAB_DATA_SECTION
8020 Define this macro if it is necessary to go to the data section before
8021 outputting the @samp{.stabs} pseudo-op for a non-global static
8025 @defmac DBX_TYPE_DECL_STABS_CODE
8026 The value to use in the ``code'' field of the @code{.stabs} directive
8027 for a typedef. The default is @code{N_LSYM}.
8030 @defmac DBX_STATIC_CONST_VAR_CODE
8031 The value to use in the ``code'' field of the @code{.stabs} directive
8032 for a static variable located in the text section. DBX format does not
8033 provide any ``right'' way to do this. The default is @code{N_FUN}.
8036 @defmac DBX_REGPARM_STABS_CODE
8037 The value to use in the ``code'' field of the @code{.stabs} directive
8038 for a parameter passed in registers. DBX format does not provide any
8039 ``right'' way to do this. The default is @code{N_RSYM}.
8042 @defmac DBX_REGPARM_STABS_LETTER
8043 The letter to use in DBX symbol data to identify a symbol as a parameter
8044 passed in registers. DBX format does not customarily provide any way to
8045 do this. The default is @code{'P'}.
8048 @defmac DBX_FUNCTION_FIRST
8049 Define this macro if the DBX information for a function and its
8050 arguments should precede the assembler code for the function. Normally,
8051 in DBX format, the debugging information entirely follows the assembler
8055 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
8056 Define this macro, with value 1, if the value of a symbol describing
8057 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
8058 relative to the start of the enclosing function. Normally, GCC uses
8059 an absolute address.
8062 @defmac DBX_LINES_FUNCTION_RELATIVE
8063 Define this macro, with value 1, if the value of a symbol indicating
8064 the current line number (@code{N_SLINE}) should be relative to the
8065 start of the enclosing function. Normally, GCC uses an absolute address.
8068 @defmac DBX_USE_BINCL
8069 Define this macro if GCC should generate @code{N_BINCL} and
8070 @code{N_EINCL} stabs for included header files, as on Sun systems. This
8071 macro also directs GCC to output a type number as a pair of a file
8072 number and a type number within the file. Normally, GCC does not
8073 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
8074 number for a type number.
8078 @subsection Open-Ended Hooks for DBX Format
8080 @c prevent bad page break with this line
8081 These are hooks for DBX format.
8083 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
8084 Define this macro to say how to output to @var{stream} the debugging
8085 information for the start of a scope level for variable names. The
8086 argument @var{name} is the name of an assembler symbol (for use with
8087 @code{assemble_name}) whose value is the address where the scope begins.
8090 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
8091 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
8094 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
8095 Define this macro if the target machine requires special handling to
8096 output an @code{N_FUN} entry for the function @var{decl}.
8099 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
8100 A C statement to output DBX debugging information before code for line
8101 number @var{line} of the current source file to the stdio stream
8102 @var{stream}. @var{counter} is the number of time the macro was
8103 invoked, including the current invocation; it is intended to generate
8104 unique labels in the assembly output.
8106 This macro should not be defined if the default output is correct, or
8107 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
8110 @defmac NO_DBX_FUNCTION_END
8111 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
8112 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
8113 On those machines, define this macro to turn this feature off without
8114 disturbing the rest of the gdb extensions.
8117 @defmac NO_DBX_BNSYM_ENSYM
8118 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
8119 extension construct. On those machines, define this macro to turn this
8120 feature off without disturbing the rest of the gdb extensions.
8123 @node File Names and DBX
8124 @subsection File Names in DBX Format
8126 @c prevent bad page break with this line
8127 This describes file names in DBX format.
8129 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
8130 A C statement to output DBX debugging information to the stdio stream
8131 @var{stream}, which indicates that file @var{name} is the main source
8132 file---the file specified as the input file for compilation.
8133 This macro is called only once, at the beginning of compilation.
8135 This macro need not be defined if the standard form of output
8136 for DBX debugging information is appropriate.
8138 It may be necessary to refer to a label equal to the beginning of the
8139 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
8140 to do so. If you do this, you must also set the variable
8141 @var{used_ltext_label_name} to @code{true}.
8144 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
8145 Define this macro, with value 1, if GCC should not emit an indication
8146 of the current directory for compilation and current source language at
8147 the beginning of the file.
8150 @defmac NO_DBX_GCC_MARKER
8151 Define this macro, with value 1, if GCC should not emit an indication
8152 that this object file was compiled by GCC@. The default is to emit
8153 an @code{N_OPT} stab at the beginning of every source file, with
8154 @samp{gcc2_compiled.} for the string and value 0.
8157 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
8158 A C statement to output DBX debugging information at the end of
8159 compilation of the main source file @var{name}. Output should be
8160 written to the stdio stream @var{stream}.
8162 If you don't define this macro, nothing special is output at the end
8163 of compilation, which is correct for most machines.
8166 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
8167 Define this macro @emph{instead of} defining
8168 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
8169 the end of compilation is a @code{N_SO} stab with an empty string,
8170 whose value is the highest absolute text address in the file.
8175 @subsection Macros for SDB and DWARF Output
8177 @c prevent bad page break with this line
8178 Here are macros for SDB and DWARF output.
8180 @defmac SDB_DEBUGGING_INFO
8181 Define this macro if GCC should produce COFF-style debugging output
8182 for SDB in response to the @option{-g} option.
8185 @defmac DWARF2_DEBUGGING_INFO
8186 Define this macro if GCC should produce dwarf version 2 format
8187 debugging output in response to the @option{-g} option.
8189 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (tree @var{function})
8190 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
8191 be emitted for each function. Instead of an integer return the enum
8192 value for the @code{DW_CC_} tag.
8195 To support optional call frame debugging information, you must also
8196 define @code{INCOMING_RETURN_ADDR_RTX} and either set
8197 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
8198 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
8199 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
8202 @defmac DWARF2_FRAME_INFO
8203 Define this macro to a nonzero value if GCC should always output
8204 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
8205 (@pxref{Exception Region Output} is nonzero, GCC will output this
8206 information not matter how you define @code{DWARF2_FRAME_INFO}.
8209 @defmac DWARF2_ASM_LINE_DEBUG_INFO
8210 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8211 line debug info sections. This will result in much more compact line number
8212 tables, and hence is desirable if it works.
8215 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8216 A C statement to issue assembly directives that create a difference
8217 between the two given labels, using an integer of the given size.
8220 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label})
8221 A C statement to issue assembly directives that create a
8222 section-relative reference to the given label, using an integer of the
8226 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8227 A C statement to issue assembly directives that create a self-relative
8228 reference to the given label, using an integer of the given size.
8231 @defmac PUT_SDB_@dots{}
8232 Define these macros to override the assembler syntax for the special
8233 SDB assembler directives. See @file{sdbout.c} for a list of these
8234 macros and their arguments. If the standard syntax is used, you need
8235 not define them yourself.
8239 Some assemblers do not support a semicolon as a delimiter, even between
8240 SDB assembler directives. In that case, define this macro to be the
8241 delimiter to use (usually @samp{\n}). It is not necessary to define
8242 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8246 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
8247 Define this macro to allow references to unknown structure,
8248 union, or enumeration tags to be emitted. Standard COFF does not
8249 allow handling of unknown references, MIPS ECOFF has support for
8253 @defmac SDB_ALLOW_FORWARD_REFERENCES
8254 Define this macro to allow references to structure, union, or
8255 enumeration tags that have not yet been seen to be handled. Some
8256 assemblers choke if forward tags are used, while some require it.
8259 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
8260 A C statement to output SDB debugging information before code for line
8261 number @var{line} of the current source file to the stdio stream
8262 @var{stream}. The default is to emit an @code{.ln} directive.
8267 @subsection Macros for VMS Debug Format
8269 @c prevent bad page break with this line
8270 Here are macros for VMS debug format.
8272 @defmac VMS_DEBUGGING_INFO
8273 Define this macro if GCC should produce debugging output for VMS
8274 in response to the @option{-g} option. The default behavior for VMS
8275 is to generate minimal debug info for a traceback in the absence of
8276 @option{-g} unless explicitly overridden with @option{-g0}. This
8277 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8278 @code{OVERRIDE_OPTIONS}.
8281 @node Floating Point
8282 @section Cross Compilation and Floating Point
8283 @cindex cross compilation and floating point
8284 @cindex floating point and cross compilation
8286 While all modern machines use twos-complement representation for integers,
8287 there are a variety of representations for floating point numbers. This
8288 means that in a cross-compiler the representation of floating point numbers
8289 in the compiled program may be different from that used in the machine
8290 doing the compilation.
8292 Because different representation systems may offer different amounts of
8293 range and precision, all floating point constants must be represented in
8294 the target machine's format. Therefore, the cross compiler cannot
8295 safely use the host machine's floating point arithmetic; it must emulate
8296 the target's arithmetic. To ensure consistency, GCC always uses
8297 emulation to work with floating point values, even when the host and
8298 target floating point formats are identical.
8300 The following macros are provided by @file{real.h} for the compiler to
8301 use. All parts of the compiler which generate or optimize
8302 floating-point calculations must use these macros. They may evaluate
8303 their operands more than once, so operands must not have side effects.
8305 @defmac REAL_VALUE_TYPE
8306 The C data type to be used to hold a floating point value in the target
8307 machine's format. Typically this is a @code{struct} containing an
8308 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8312 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8313 Compares for equality the two values, @var{x} and @var{y}. If the target
8314 floating point format supports negative zeroes and/or NaNs,
8315 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8316 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8319 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8320 Tests whether @var{x} is less than @var{y}.
8323 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8324 Truncates @var{x} to a signed integer, rounding toward zero.
8327 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8328 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8329 @var{x} is negative, returns zero.
8332 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8333 Converts @var{string} into a floating point number in the target machine's
8334 representation for mode @var{mode}. This routine can handle both
8335 decimal and hexadecimal floating point constants, using the syntax
8336 defined by the C language for both.
8339 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8340 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8343 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8344 Determines whether @var{x} represents infinity (positive or negative).
8347 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8348 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8351 @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})
8352 Calculates an arithmetic operation on the two floating point values
8353 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8356 The operation to be performed is specified by @var{code}. Only the
8357 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8358 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8360 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8361 target's floating point format cannot represent infinity, it will call
8362 @code{abort}. Callers should check for this situation first, using
8363 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8366 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8367 Returns the negative of the floating point value @var{x}.
8370 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8371 Returns the absolute value of @var{x}.
8374 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8375 Truncates the floating point value @var{x} to fit in @var{mode}. The
8376 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8377 appropriate bit pattern to be output asa floating constant whose
8378 precision accords with mode @var{mode}.
8381 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8382 Converts a floating point value @var{x} into a double-precision integer
8383 which is then stored into @var{low} and @var{high}. If the value is not
8384 integral, it is truncated.
8387 @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})
8388 Converts a double-precision integer found in @var{low} and @var{high},
8389 into a floating point value which is then stored into @var{x}. The
8390 value is truncated to fit in mode @var{mode}.
8393 @node Mode Switching
8394 @section Mode Switching Instructions
8395 @cindex mode switching
8396 The following macros control mode switching optimizations:
8398 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
8399 Define this macro if the port needs extra instructions inserted for mode
8400 switching in an optimizing compilation.
8402 For an example, the SH4 can perform both single and double precision
8403 floating point operations, but to perform a single precision operation,
8404 the FPSCR PR bit has to be cleared, while for a double precision
8405 operation, this bit has to be set. Changing the PR bit requires a general
8406 purpose register as a scratch register, hence these FPSCR sets have to
8407 be inserted before reload, i.e.@: you can't put this into instruction emitting
8408 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
8410 You can have multiple entities that are mode-switched, and select at run time
8411 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
8412 return nonzero for any @var{entity} that needs mode-switching.
8413 If you define this macro, you also have to define
8414 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
8415 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
8416 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
8420 @defmac NUM_MODES_FOR_MODE_SWITCHING
8421 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
8422 initializer for an array of integers. Each initializer element
8423 N refers to an entity that needs mode switching, and specifies the number
8424 of different modes that might need to be set for this entity.
8425 The position of the initializer in the initializer---starting counting at
8426 zero---determines the integer that is used to refer to the mode-switched
8428 In macros that take mode arguments / yield a mode result, modes are
8429 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
8430 switch is needed / supplied.
8433 @defmac MODE_NEEDED (@var{entity}, @var{insn})
8434 @var{entity} is an integer specifying a mode-switched entity. If
8435 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
8436 return an integer value not larger than the corresponding element in
8437 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
8438 be switched into prior to the execution of @var{insn}.
8441 @defmac MODE_AFTER (@var{mode}, @var{insn})
8442 If this macro is defined, it is evaluated for every @var{insn} during
8443 mode switching. It determines the mode that an insn results in (if
8444 different from the incoming mode).
8447 @defmac MODE_ENTRY (@var{entity})
8448 If this macro is defined, it is evaluated for every @var{entity} that needs
8449 mode switching. It should evaluate to an integer, which is a mode that
8450 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
8451 is defined then @code{MODE_EXIT} must be defined.
8454 @defmac MODE_EXIT (@var{entity})
8455 If this macro is defined, it is evaluated for every @var{entity} that needs
8456 mode switching. It should evaluate to an integer, which is a mode that
8457 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
8458 is defined then @code{MODE_ENTRY} must be defined.
8461 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
8462 This macro specifies the order in which modes for @var{entity} are processed.
8463 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
8464 lowest. The value of the macro should be an integer designating a mode
8465 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
8466 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
8467 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
8470 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
8471 Generate one or more insns to set @var{entity} to @var{mode}.
8472 @var{hard_reg_live} is the set of hard registers live at the point where
8473 the insn(s) are to be inserted.
8476 @node Target Attributes
8477 @section Defining target-specific uses of @code{__attribute__}
8478 @cindex target attributes
8479 @cindex machine attributes
8480 @cindex attributes, target-specific
8482 Target-specific attributes may be defined for functions, data and types.
8483 These are described using the following target hooks; they also need to
8484 be documented in @file{extend.texi}.
8486 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
8487 If defined, this target hook points to an array of @samp{struct
8488 attribute_spec} (defined in @file{tree.h}) specifying the machine
8489 specific attributes for this target and some of the restrictions on the
8490 entities to which these attributes are applied and the arguments they
8494 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8495 If defined, this target hook is a function which returns zero if the attributes on
8496 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8497 and two if they are nearly compatible (which causes a warning to be
8498 generated). If this is not defined, machine-specific attributes are
8499 supposed always to be compatible.
8502 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
8503 If defined, this target hook is a function which assigns default attributes to
8504 newly defined @var{type}.
8507 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8508 Define this target hook if the merging of type attributes needs special
8509 handling. If defined, the result is a list of the combined
8510 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
8511 that @code{comptypes} has already been called and returned 1. This
8512 function may call @code{merge_attributes} to handle machine-independent
8516 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
8517 Define this target hook if the merging of decl attributes needs special
8518 handling. If defined, the result is a list of the combined
8519 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
8520 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
8521 when this is needed are when one attribute overrides another, or when an
8522 attribute is nullified by a subsequent definition. This function may
8523 call @code{merge_attributes} to handle machine-independent merging.
8525 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
8526 If the only target-specific handling you require is @samp{dllimport}
8527 for Microsoft Windows targets, you should define the macro
8528 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
8529 will then define a function called
8530 @code{merge_dllimport_decl_attributes} which can then be defined as
8531 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
8532 add @code{handle_dll_attribute} in the attribute table for your port
8533 to perform initial processing of the @samp{dllimport} and
8534 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
8535 @file{i386/i386.c}, for example.
8538 @defmac TARGET_DECLSPEC
8539 Define this macro to a nonzero value if you want to treat
8540 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
8541 default, this behavior is enabled only for targets that define
8542 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
8543 of @code{__declspec} is via a built-in macro, but you should not rely
8544 on this implementation detail.
8547 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
8548 Define this target hook if you want to be able to add attributes to a decl
8549 when it is being created. This is normally useful for back ends which
8550 wish to implement a pragma by using the attributes which correspond to
8551 the pragma's effect. The @var{node} argument is the decl which is being
8552 created. The @var{attr_ptr} argument is a pointer to the attribute list
8553 for this decl. The list itself should not be modified, since it may be
8554 shared with other decls, but attributes may be chained on the head of
8555 the list and @code{*@var{attr_ptr}} modified to point to the new
8556 attributes, or a copy of the list may be made if further changes are
8560 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
8562 This target hook returns @code{true} if it is ok to inline @var{fndecl}
8563 into the current function, despite its having target-specific
8564 attributes, @code{false} otherwise. By default, if a function has a
8565 target specific attribute attached to it, it will not be inlined.
8568 @node MIPS Coprocessors
8569 @section Defining coprocessor specifics for MIPS targets.
8570 @cindex MIPS coprocessor-definition macros
8572 The MIPS specification allows MIPS implementations to have as many as 4
8573 coprocessors, each with as many as 32 private registers. GCC supports
8574 accessing these registers and transferring values between the registers
8575 and memory using asm-ized variables. For example:
8578 register unsigned int cp0count asm ("c0r1");
8584 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
8585 names may be added as described below, or the default names may be
8586 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
8588 Coprocessor registers are assumed to be epilogue-used; sets to them will
8589 be preserved even if it does not appear that the register is used again
8590 later in the function.
8592 Another note: according to the MIPS spec, coprocessor 1 (if present) is
8593 the FPU@. One accesses COP1 registers through standard mips
8594 floating-point support; they are not included in this mechanism.
8596 There is one macro used in defining the MIPS coprocessor interface which
8597 you may want to override in subtargets; it is described below.
8599 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
8600 A comma-separated list (with leading comma) of pairs describing the
8601 alternate names of coprocessor registers. The format of each entry should be
8603 @{ @var{alternatename}, @var{register_number}@}
8609 @section Parameters for Precompiled Header Validity Checking
8610 @cindex parameters, precompiled headers
8612 @deftypefn {Target Hook} void * TARGET_GET_PCH_VALIDITY (size_t * @var{sz})
8613 Define this hook if your target needs to check a different collection
8614 of flags than the default, which is every flag defined by
8615 @code{TARGET_SWITCHES} and @code{TARGET_OPTIONS}. It should return
8616 some data which will be saved in the PCH file and presented to
8617 @code{TARGET_PCH_VALID_P} later; it should set @code{SZ} to the size
8621 @deftypefn {Target Hook} const char * TARGET_PCH_VALID_P (const void * @var{data}, size_t @var{sz})
8622 Define this hook if your target needs to check a different collection of
8623 flags than the default, which is every flag defined by @code{TARGET_SWITCHES}
8624 and @code{TARGET_OPTIONS}. It is given data which came from
8625 @code{TARGET_GET_PCH_VALIDITY} (in this version of this compiler, so there
8626 is no need for extensive validity checking). It returns @code{NULL} if
8627 it is safe to load a PCH file with this data, or a suitable error message
8628 if not. The error message will be presented to the user, so it should
8633 @section C++ ABI parameters
8634 @cindex parameters, c++ abi
8636 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
8637 Define this hook to override the integer type used for guard variables.
8638 These are used to implement one-time construction of static objects. The
8639 default is long_long_integer_type_node.
8642 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
8643 This hook determines how guard variables are used. It should return
8644 @code{false} (the default) if first byte should be used. A return value of
8645 @code{true} indicates the least significant bit should be used.
8648 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
8649 This hook returns the size of the cookie to use when allocating an array
8650 whose elements have the indicated @var{type}. Assumes that it is already
8651 known that a cookie is needed. The default is
8652 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
8653 IA64/Generic C++ ABI@.
8656 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
8657 This hook should return @code{true} if the element size should be stored in
8658 array cookies. The default is to return @code{false}.
8661 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
8662 If defined by a backend this hook allows the decision made to export
8663 class @var{type} to be overruled. Upon entry @var{import_export}
8664 will contain 1 if the class is going to be exported, @minus{}1 if it is going
8665 to be imported and 0 otherwise. This function should return the
8666 modified value and perform any other actions necessary to support the
8667 backend's targeted operating system.
8670 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
8671 This hook should return @code{true} if constructors and destructors return
8672 the address of the object created/destroyed. The default is to return
8676 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
8677 This hook returns true if the key method for a class (i.e., the method
8678 which, if defined in the current translation unit, causes the virtual
8679 table to be emitted) may be an inline function. Under the standard
8680 Itanium C++ ABI the key method may be an inline function so long as
8681 the function is not declared inline in the class definition. Under
8682 some variants of the ABI, an inline function can never be the key
8683 method. The default is to return @code{true}.
8686 @deftypefn {Target Hook} bool TARGET_CXX_EXPORT_CLASS_DATA (void)
8687 If this hook returns false (the default), then virtual tables and RTTI
8688 data structures will have the ELF visibility of their containing
8689 class. If this hook returns true, then these data structures will
8690 have ELF ``default'' visibility, independently of the visibility of
8691 the containing class.
8695 @section Miscellaneous Parameters
8696 @cindex parameters, miscellaneous
8698 @c prevent bad page break with this line
8699 Here are several miscellaneous parameters.
8701 @defmac PREDICATE_CODES
8702 Define this if you have defined special-purpose predicates in the file
8703 @file{@var{machine}.c}. This macro is called within an initializer of an
8704 array of structures. The first field in the structure is the name of a
8705 predicate and the second field is an array of rtl codes. For each
8706 predicate, list all rtl codes that can be in expressions matched by the
8707 predicate. The list should have a trailing comma. Here is an example
8708 of two entries in the list for a typical RISC machine:
8711 #define PREDICATE_CODES \
8712 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
8713 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
8716 Defining this macro does not affect the generated code (however,
8717 incorrect definitions that omit an rtl code that may be matched by the
8718 predicate can cause the compiler to malfunction). Instead, it allows
8719 the table built by @file{genrecog} to be more compact and efficient,
8720 thus speeding up the compiler. The most important predicates to include
8721 in the list specified by this macro are those used in the most insn
8724 For each predicate function named in @code{PREDICATE_CODES}, a
8725 declaration will be generated in @file{insn-codes.h}.
8727 Use of this macro is deprecated; use @code{define_predicate} instead.
8728 @xref{Defining Predicates}.
8731 @defmac SPECIAL_MODE_PREDICATES
8732 Define this if you have special predicates that know special things
8733 about modes. Genrecog will warn about certain forms of
8734 @code{match_operand} without a mode; if the operand predicate is
8735 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
8738 Here is an example from the IA-32 port (@code{ext_register_operand}
8739 specially checks for @code{HImode} or @code{SImode} in preparation
8740 for a byte extraction from @code{%ah} etc.).
8743 #define SPECIAL_MODE_PREDICATES \
8744 "ext_register_operand",
8747 Use of this macro is deprecated; use @code{define_special_predicate}
8748 instead. @xref{Defining Predicates}.
8751 @defmac HAS_LONG_COND_BRANCH
8752 Define this boolean macro to indicate whether or not your architecture
8753 has conditional branches that can span all of memory. It is used in
8754 conjunction with an optimization that partitions hot and cold basic
8755 blocks into separate sections of the executable. If this macro is
8756 set to false, gcc will convert any conditional branches that attempt
8757 to cross between sections into unconditional branches or indirect jumps.
8760 @defmac HAS_LONG_UNCOND_BRANCH
8761 Define this boolean macro to indicate whether or not your architecture
8762 has unconditional branches that can span all of memory. It is used in
8763 conjunction with an optimization that partitions hot and cold basic
8764 blocks into separate sections of the executable. If this macro is
8765 set to false, gcc will convert any unconditional branches that attempt
8766 to cross between sections into indirect jumps.
8769 @defmac CASE_VECTOR_MODE
8770 An alias for a machine mode name. This is the machine mode that
8771 elements of a jump-table should have.
8774 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
8775 Optional: return the preferred mode for an @code{addr_diff_vec}
8776 when the minimum and maximum offset are known. If you define this,
8777 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
8778 To make this work, you also have to define @code{INSN_ALIGN} and
8779 make the alignment for @code{addr_diff_vec} explicit.
8780 The @var{body} argument is provided so that the offset_unsigned and scale
8781 flags can be updated.
8784 @defmac CASE_VECTOR_PC_RELATIVE
8785 Define this macro to be a C expression to indicate when jump-tables
8786 should contain relative addresses. You need not define this macro if
8787 jump-tables never contain relative addresses, or jump-tables should
8788 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
8792 @defmac CASE_VALUES_THRESHOLD
8793 Define this to be the smallest number of different values for which it
8794 is best to use a jump-table instead of a tree of conditional branches.
8795 The default is four for machines with a @code{casesi} instruction and
8796 five otherwise. This is best for most machines.
8799 @defmac CASE_USE_BIT_TESTS
8800 Define this macro to be a C expression to indicate whether C switch
8801 statements may be implemented by a sequence of bit tests. This is
8802 advantageous on processors that can efficiently implement left shift
8803 of 1 by the number of bits held in a register, but inappropriate on
8804 targets that would require a loop. By default, this macro returns
8805 @code{true} if the target defines an @code{ashlsi3} pattern, and
8806 @code{false} otherwise.
8809 @defmac WORD_REGISTER_OPERATIONS
8810 Define this macro if operations between registers with integral mode
8811 smaller than a word are always performed on the entire register.
8812 Most RISC machines have this property and most CISC machines do not.
8815 @defmac LOAD_EXTEND_OP (@var{mem_mode})
8816 Define this macro to be a C expression indicating when insns that read
8817 memory in @var{mem_mode}, an integral mode narrower than a word, set the
8818 bits outside of @var{mem_mode} to be either the sign-extension or the
8819 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
8820 of @var{mem_mode} for which the
8821 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
8822 @code{UNKNOWN} for other modes.
8824 This macro is not called with @var{mem_mode} non-integral or with a width
8825 greater than or equal to @code{BITS_PER_WORD}, so you may return any
8826 value in this case. Do not define this macro if it would always return
8827 @code{UNKNOWN}. On machines where this macro is defined, you will normally
8828 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
8830 You may return a non-@code{UNKNOWN} value even if for some hard registers
8831 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
8832 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
8833 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
8834 integral mode larger than this but not larger than @code{word_mode}.
8836 You must return @code{UNKNOWN} if for some hard registers that allow this
8837 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
8838 @code{word_mode}, but that they can change to another integral mode that
8839 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
8842 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
8843 Define this macro if loading short immediate values into registers sign
8847 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
8848 Define this macro if the same instructions that convert a floating
8849 point number to a signed fixed point number also convert validly to an
8854 The maximum number of bytes that a single instruction can move quickly
8855 between memory and registers or between two memory locations.
8858 @defmac MAX_MOVE_MAX
8859 The maximum number of bytes that a single instruction can move quickly
8860 between memory and registers or between two memory locations. If this
8861 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
8862 constant value that is the largest value that @code{MOVE_MAX} can have
8866 @defmac SHIFT_COUNT_TRUNCATED
8867 A C expression that is nonzero if on this machine the number of bits
8868 actually used for the count of a shift operation is equal to the number
8869 of bits needed to represent the size of the object being shifted. When
8870 this macro is nonzero, the compiler will assume that it is safe to omit
8871 a sign-extend, zero-extend, and certain bitwise `and' instructions that
8872 truncates the count of a shift operation. On machines that have
8873 instructions that act on bit-fields at variable positions, which may
8874 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
8875 also enables deletion of truncations of the values that serve as
8876 arguments to bit-field instructions.
8878 If both types of instructions truncate the count (for shifts) and
8879 position (for bit-field operations), or if no variable-position bit-field
8880 instructions exist, you should define this macro.
8882 However, on some machines, such as the 80386 and the 680x0, truncation
8883 only applies to shift operations and not the (real or pretended)
8884 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
8885 such machines. Instead, add patterns to the @file{md} file that include
8886 the implied truncation of the shift instructions.
8888 You need not define this macro if it would always have the value of zero.
8891 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
8892 @deftypefn {Target Hook} int TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode})
8893 This function describes how the standard shift patterns for @var{mode}
8894 deal with shifts by negative amounts or by more than the width of the mode.
8895 @xref{shift patterns}.
8897 On many machines, the shift patterns will apply a mask @var{m} to the
8898 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
8899 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
8900 this is true for mode @var{mode}, the function should return @var{m},
8901 otherwise it should return 0. A return value of 0 indicates that no
8902 particular behavior is guaranteed.
8904 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
8905 @emph{not} apply to general shift rtxes; it applies only to instructions
8906 that are generated by the named shift patterns.
8908 The default implementation of this function returns
8909 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
8910 and 0 otherwise. This definition is always safe, but if
8911 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
8912 nevertheless truncate the shift count, you may get better code
8916 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
8917 A C expression which is nonzero if on this machine it is safe to
8918 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
8919 bits (where @var{outprec} is smaller than @var{inprec}) by merely
8920 operating on it as if it had only @var{outprec} bits.
8922 On many machines, this expression can be 1.
8924 @c rearranged this, removed the phrase "it is reported that". this was
8925 @c to fix an overfull hbox. --mew 10feb93
8926 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
8927 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
8928 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
8929 such cases may improve things.
8932 @defmac STORE_FLAG_VALUE
8933 A C expression describing the value returned by a comparison operator
8934 with an integral mode and stored by a store-flag instruction
8935 (@samp{s@var{cond}}) when the condition is true. This description must
8936 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
8937 comparison operators whose results have a @code{MODE_INT} mode.
8939 A value of 1 or @minus{}1 means that the instruction implementing the
8940 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
8941 and 0 when the comparison is false. Otherwise, the value indicates
8942 which bits of the result are guaranteed to be 1 when the comparison is
8943 true. This value is interpreted in the mode of the comparison
8944 operation, which is given by the mode of the first operand in the
8945 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
8946 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
8949 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
8950 generate code that depends only on the specified bits. It can also
8951 replace comparison operators with equivalent operations if they cause
8952 the required bits to be set, even if the remaining bits are undefined.
8953 For example, on a machine whose comparison operators return an
8954 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
8955 @samp{0x80000000}, saying that just the sign bit is relevant, the
8959 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
8966 (ashift:SI @var{x} (const_int @var{n}))
8970 where @var{n} is the appropriate shift count to move the bit being
8971 tested into the sign bit.
8973 There is no way to describe a machine that always sets the low-order bit
8974 for a true value, but does not guarantee the value of any other bits,
8975 but we do not know of any machine that has such an instruction. If you
8976 are trying to port GCC to such a machine, include an instruction to
8977 perform a logical-and of the result with 1 in the pattern for the
8978 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
8980 Often, a machine will have multiple instructions that obtain a value
8981 from a comparison (or the condition codes). Here are rules to guide the
8982 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
8987 Use the shortest sequence that yields a valid definition for
8988 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
8989 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
8990 comparison operators to do so because there may be opportunities to
8991 combine the normalization with other operations.
8994 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
8995 slightly preferred on machines with expensive jumps and 1 preferred on
8999 As a second choice, choose a value of @samp{0x80000001} if instructions
9000 exist that set both the sign and low-order bits but do not define the
9004 Otherwise, use a value of @samp{0x80000000}.
9007 Many machines can produce both the value chosen for
9008 @code{STORE_FLAG_VALUE} and its negation in the same number of
9009 instructions. On those machines, you should also define a pattern for
9010 those cases, e.g., one matching
9013 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
9016 Some machines can also perform @code{and} or @code{plus} operations on
9017 condition code values with less instructions than the corresponding
9018 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
9019 machines, define the appropriate patterns. Use the names @code{incscc}
9020 and @code{decscc}, respectively, for the patterns which perform
9021 @code{plus} or @code{minus} operations on condition code values. See
9022 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
9023 find such instruction sequences on other machines.
9025 If this macro is not defined, the default value, 1, is used. You need
9026 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
9027 instructions, or if the value generated by these instructions is 1.
9030 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
9031 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
9032 returned when comparison operators with floating-point results are true.
9033 Define this macro on machines that have comparison operations that return
9034 floating-point values. If there are no such operations, do not define
9038 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
9039 A C expression that gives a rtx representing the non-zero true element
9040 for vector comparisons. The returned rtx should be valid for the inner
9041 mode of @var{mode} which is guaranteed to be a vector mode. Define
9042 this macro on machines that have vector comparison operations that
9043 return a vector result. If there are no such operations, do not define
9044 this macro. Typically, this macro is defined as @code{const1_rtx} or
9045 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
9046 the compiler optimizing such vector comparison operations for the
9050 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
9051 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
9052 A C expression that evaluates to true if the architecture defines a value
9053 for @code{clz} or @code{ctz} with a zero operand. If so, @var{value}
9054 should be set to this value. If this macro is not defined, the value of
9055 @code{clz} or @code{ctz} is assumed to be undefined.
9057 This macro must be defined if the target's expansion for @code{ffs}
9058 relies on a particular value to get correct results. Otherwise it
9059 is not necessary, though it may be used to optimize some corner cases.
9061 Note that regardless of this macro the ``definedness'' of @code{clz}
9062 and @code{ctz} at zero do @emph{not} extend to the builtin functions
9063 visible to the user. Thus one may be free to adjust the value at will
9064 to match the target expansion of these operations without fear of
9069 An alias for the machine mode for pointers. On most machines, define
9070 this to be the integer mode corresponding to the width of a hardware
9071 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
9072 On some machines you must define this to be one of the partial integer
9073 modes, such as @code{PSImode}.
9075 The width of @code{Pmode} must be at least as large as the value of
9076 @code{POINTER_SIZE}. If it is not equal, you must define the macro
9077 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
9081 @defmac FUNCTION_MODE
9082 An alias for the machine mode used for memory references to functions
9083 being called, in @code{call} RTL expressions. On most machines this
9084 should be @code{QImode}.
9087 @defmac STDC_0_IN_SYSTEM_HEADERS
9088 In normal operation, the preprocessor expands @code{__STDC__} to the
9089 constant 1, to signify that GCC conforms to ISO Standard C@. On some
9090 hosts, like Solaris, the system compiler uses a different convention,
9091 where @code{__STDC__} is normally 0, but is 1 if the user specifies
9092 strict conformance to the C Standard.
9094 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
9095 convention when processing system header files, but when processing user
9096 files @code{__STDC__} will always expand to 1.
9099 @defmac NO_IMPLICIT_EXTERN_C
9100 Define this macro if the system header files support C++ as well as C@.
9101 This macro inhibits the usual method of using system header files in
9102 C++, which is to pretend that the file's contents are enclosed in
9103 @samp{extern "C" @{@dots{}@}}.
9108 @defmac REGISTER_TARGET_PRAGMAS ()
9109 Define this macro if you want to implement any target-specific pragmas.
9110 If defined, it is a C expression which makes a series of calls to
9111 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
9112 for each pragma. The macro may also do any
9113 setup required for the pragmas.
9115 The primary reason to define this macro is to provide compatibility with
9116 other compilers for the same target. In general, we discourage
9117 definition of target-specific pragmas for GCC@.
9119 If the pragma can be implemented by attributes then you should consider
9120 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
9122 Preprocessor macros that appear on pragma lines are not expanded. All
9123 @samp{#pragma} directives that do not match any registered pragma are
9124 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
9127 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9128 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
9130 Each call to @code{c_register_pragma} or
9131 @code{c_register_pragma_with_expansion} establishes one pragma. The
9132 @var{callback} routine will be called when the preprocessor encounters a
9136 #pragma [@var{space}] @var{name} @dots{}
9139 @var{space} is the case-sensitive namespace of the pragma, or
9140 @code{NULL} to put the pragma in the global namespace. The callback
9141 routine receives @var{pfile} as its first argument, which can be passed
9142 on to cpplib's functions if necessary. You can lex tokens after the
9143 @var{name} by calling @code{c_lex}. Tokens that are not read by the
9144 callback will be silently ignored. The end of the line is indicated by
9145 a token of type @code{CPP_EOF}. Macro expansion occurs on the
9146 arguments of pragmas registered with
9147 @code{c_register_pragma_with_expansion} but not on the arguments of
9148 pragmas registered with @code{c_register_pragma}.
9150 For an example use of this routine, see @file{c4x.h} and the callback
9151 routines defined in @file{c4x-c.c}.
9153 Note that the use of @code{c_lex} is specific to the C and C++
9154 compilers. It will not work in the Java or Fortran compilers, or any
9155 other language compilers for that matter. Thus if @code{c_lex} is going
9156 to be called from target-specific code, it must only be done so when
9157 building the C and C++ compilers. This can be done by defining the
9158 variables @code{c_target_objs} and @code{cxx_target_objs} in the
9159 target entry in the @file{config.gcc} file. These variables should name
9160 the target-specific, language-specific object file which contains the
9161 code that uses @code{c_lex}. Note it will also be necessary to add a
9162 rule to the makefile fragment pointed to by @code{tmake_file} that shows
9163 how to build this object file.
9168 @defmac HANDLE_SYSV_PRAGMA
9169 Define this macro (to a value of 1) if you want the System V style
9170 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
9171 [=<value>]} to be supported by gcc.
9173 The pack pragma specifies the maximum alignment (in bytes) of fields
9174 within a structure, in much the same way as the @samp{__aligned__} and
9175 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
9176 the behavior to the default.
9178 A subtlety for Microsoft Visual C/C++ style bit-field packing
9179 (e.g.@: -mms-bitfields) for targets that support it:
9180 When a bit-field is inserted into a packed record, the whole size
9181 of the underlying type is used by one or more same-size adjacent
9182 bit-fields (that is, if its long:3, 32 bits is used in the record,
9183 and any additional adjacent long bit-fields are packed into the same
9184 chunk of 32 bits. However, if the size changes, a new field of that
9187 If both MS bit-fields and @samp{__attribute__((packed))} are used,
9188 the latter will take precedence. If @samp{__attribute__((packed))} is
9189 used on a single field when MS bit-fields are in use, it will take
9190 precedence for that field, but the alignment of the rest of the structure
9191 may affect its placement.
9193 The weak pragma only works if @code{SUPPORTS_WEAK} and
9194 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
9195 of specifically named weak labels, optionally with a value.
9200 @defmac HANDLE_PRAGMA_PACK_PUSH_POP
9201 Define this macro (to a value of 1) if you want to support the Win32
9202 style pragmas @samp{#pragma pack(push[,@var{n}])} and @samp{#pragma
9203 pack(pop)}. The @samp{pack(push,[@var{n}])} pragma specifies the maximum
9204 alignment (in bytes) of fields within a structure, in much the same way as
9205 the @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
9206 pack value of zero resets the behavior to the default. Successive
9207 invocations of this pragma cause the previous values to be stacked, so
9208 that invocations of @samp{#pragma pack(pop)} will return to the previous
9212 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
9213 Define this macro, as well as
9214 @code{HANDLE_SYSV_PRAGMA}, if macros should be expanded in the
9215 arguments of @samp{#pragma pack}.
9218 @defmac TARGET_DEFAULT_PACK_STRUCT
9219 If your target requires a structure packing default other than 0 (meaning
9220 the machine default), define this macro the the necessary value (in bytes).
9221 This must be a value that would also valid to be used with
9222 @samp{#pragma pack()} (that is, a small power of two).
9225 @defmac DOLLARS_IN_IDENTIFIERS
9226 Define this macro to control use of the character @samp{$} in
9227 identifier names for the C family of languages. 0 means @samp{$} is
9228 not allowed by default; 1 means it is allowed. 1 is the default;
9229 there is no need to define this macro in that case.
9232 @defmac NO_DOLLAR_IN_LABEL
9233 Define this macro if the assembler does not accept the character
9234 @samp{$} in label names. By default constructors and destructors in
9235 G++ have @samp{$} in the identifiers. If this macro is defined,
9236 @samp{.} is used instead.
9239 @defmac NO_DOT_IN_LABEL
9240 Define this macro if the assembler does not accept the character
9241 @samp{.} in label names. By default constructors and destructors in G++
9242 have names that use @samp{.}. If this macro is defined, these names
9243 are rewritten to avoid @samp{.}.
9246 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
9247 Define this macro as a C expression that is nonzero if it is safe for the
9248 delay slot scheduler to place instructions in the delay slot of @var{insn},
9249 even if they appear to use a resource set or clobbered in @var{insn}.
9250 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
9251 every @code{call_insn} has this behavior. On machines where some @code{insn}
9252 or @code{jump_insn} is really a function call and hence has this behavior,
9253 you should define this macro.
9255 You need not define this macro if it would always return zero.
9258 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
9259 Define this macro as a C expression that is nonzero if it is safe for the
9260 delay slot scheduler to place instructions in the delay slot of @var{insn},
9261 even if they appear to set or clobber a resource referenced in @var{insn}.
9262 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
9263 some @code{insn} or @code{jump_insn} is really a function call and its operands
9264 are registers whose use is actually in the subroutine it calls, you should
9265 define this macro. Doing so allows the delay slot scheduler to move
9266 instructions which copy arguments into the argument registers into the delay
9269 You need not define this macro if it would always return zero.
9272 @defmac MULTIPLE_SYMBOL_SPACES
9273 Define this macro as a C expression that is nonzero if, in some cases,
9274 global symbols from one translation unit may not be bound to undefined
9275 symbols in another translation unit without user intervention. For
9276 instance, under Microsoft Windows symbols must be explicitly imported
9277 from shared libraries (DLLs).
9279 You need not define this macro if it would always evaluate to zero.
9282 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{clobbers})
9283 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
9284 any hard regs the port wishes to automatically clobber for all asms.
9285 It should return the result of the last @code{tree_cons} used to add a
9289 @defmac MATH_LIBRARY
9290 Define this macro as a C string constant for the linker argument to link
9291 in the system math library, or @samp{""} if the target does not have a
9292 separate math library.
9294 You need only define this macro if the default of @samp{"-lm"} is wrong.
9297 @defmac LIBRARY_PATH_ENV
9298 Define this macro as a C string constant for the environment variable that
9299 specifies where the linker should look for libraries.
9301 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
9305 @defmac TARGET_HAS_F_SETLKW
9306 Define this macro if the target supports file locking with fcntl / F_SETLKW@.
9307 Note that this functionality is part of POSIX@.
9308 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
9309 to use file locking when exiting a program, which avoids race conditions
9310 if the program has forked.
9313 @defmac MAX_CONDITIONAL_EXECUTE
9315 A C expression for the maximum number of instructions to execute via
9316 conditional execution instructions instead of a branch. A value of
9317 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
9318 1 if it does use cc0.
9321 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
9322 Used if the target needs to perform machine-dependent modifications on the
9323 conditionals used for turning basic blocks into conditionally executed code.
9324 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
9325 contains information about the currently processed blocks. @var{true_expr}
9326 and @var{false_expr} are the tests that are used for converting the
9327 then-block and the else-block, respectively. Set either @var{true_expr} or
9328 @var{false_expr} to a null pointer if the tests cannot be converted.
9331 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
9332 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
9333 if-statements into conditions combined by @code{and} and @code{or} operations.
9334 @var{bb} contains the basic block that contains the test that is currently
9335 being processed and about to be turned into a condition.
9338 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
9339 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
9340 be converted to conditional execution format. @var{ce_info} points to
9341 a data structure, @code{struct ce_if_block}, which contains information
9342 about the currently processed blocks.
9345 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
9346 A C expression to perform any final machine dependent modifications in
9347 converting code to conditional execution. The involved basic blocks
9348 can be found in the @code{struct ce_if_block} structure that is pointed
9349 to by @var{ce_info}.
9352 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
9353 A C expression to cancel any machine dependent modifications in
9354 converting code to conditional execution. The involved basic blocks
9355 can be found in the @code{struct ce_if_block} structure that is pointed
9356 to by @var{ce_info}.
9359 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
9360 A C expression to initialize any extra fields in a @code{struct ce_if_block}
9361 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
9364 @defmac IFCVT_EXTRA_FIELDS
9365 If defined, it should expand to a set of field declarations that will be
9366 added to the @code{struct ce_if_block} structure. These should be initialized
9367 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
9370 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG ()
9371 If non-null, this hook performs a target-specific pass over the
9372 instruction stream. The compiler will run it at all optimization levels,
9373 just before the point at which it normally does delayed-branch scheduling.
9375 The exact purpose of the hook varies from target to target. Some use
9376 it to do transformations that are necessary for correctness, such as
9377 laying out in-function constant pools or avoiding hardware hazards.
9378 Others use it as an opportunity to do some machine-dependent optimizations.
9380 You need not implement the hook if it has nothing to do. The default
9384 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
9385 Define this hook if you have any machine-specific built-in functions
9386 that need to be defined. It should be a function that performs the
9389 Machine specific built-in functions can be useful to expand special machine
9390 instructions that would otherwise not normally be generated because
9391 they have no equivalent in the source language (for example, SIMD vector
9392 instructions or prefetch instructions).
9394 To create a built-in function, call the function
9395 @code{lang_hooks.builtin_function}
9396 which is defined by the language front end. You can use any type nodes set
9397 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
9398 only language front ends that use those two functions will call
9399 @samp{TARGET_INIT_BUILTINS}.
9402 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
9404 Expand a call to a machine specific built-in function that was set up by
9405 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
9406 function call; the result should go to @var{target} if that is
9407 convenient, and have mode @var{mode} if that is convenient.
9408 @var{subtarget} may be used as the target for computing one of
9409 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
9410 ignored. This function should return the result of the call to the
9414 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{exp}, bool @var{ignore})
9416 Expand a call to a machine specific built-in function that was set up by
9417 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
9418 function call; the result is another tree containing a simplified
9419 expression for the call's result. If @var{ignore} is true the
9420 value will be ignored.
9423 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
9425 Take a branch insn in @var{branch1} and another in @var{branch2}.
9426 Return true if redirecting @var{branch1} to the destination of
9427 @var{branch2} is possible.
9429 On some targets, branches may have a limited range. Optimizing the
9430 filling of delay slots can result in branches being redirected, and this
9431 may in turn cause a branch offset to overflow.
9434 @defmac ALLOCATE_INITIAL_VALUE (@var{hard_reg})
9436 When the initial value of a hard register has been copied in a pseudo
9437 register, it is often not necessary to actually allocate another register
9438 to this pseudo register, because the original hard register or a stack slot
9439 it has been saved into can be used. @code{ALLOCATE_INITIAL_VALUE}, if
9440 defined, is called at the start of register allocation once for each
9441 hard register that had its initial value copied by using
9442 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
9443 Possible values are @code{NULL_RTX}, if you don't want
9444 to do any special allocation, a @code{REG} rtx---that would typically be
9445 the hard register itself, if it is known not to be clobbered---or a
9447 If you are returning a @code{MEM}, this is only a hint for the allocator;
9448 it might decide to use another register anyways.
9449 You may use @code{current_function_leaf_function} in the definition of the
9450 macro, functions that use @code{REG_N_SETS}, to determine if the hard
9451 register in question will not be clobbered.
9454 @defmac TARGET_OBJECT_SUFFIX
9455 Define this macro to be a C string representing the suffix for object
9456 files on your target machine. If you do not define this macro, GCC will
9457 use @samp{.o} as the suffix for object files.
9460 @defmac TARGET_EXECUTABLE_SUFFIX
9461 Define this macro to be a C string representing the suffix to be
9462 automatically added to executable files on your target machine. If you
9463 do not define this macro, GCC will use the null string as the suffix for
9467 @defmac COLLECT_EXPORT_LIST
9468 If defined, @code{collect2} will scan the individual object files
9469 specified on its command line and create an export list for the linker.
9470 Define this macro for systems like AIX, where the linker discards
9471 object files that are not referenced from @code{main} and uses export
9475 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
9476 Define this macro to a C expression representing a variant of the
9477 method call @var{mdecl}, if Java Native Interface (JNI) methods
9478 must be invoked differently from other methods on your target.
9479 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
9480 the @code{stdcall} calling convention and this macro is then
9481 defined as this expression:
9484 build_type_attribute_variant (@var{mdecl},
9486 (get_identifier ("stdcall"),
9491 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
9492 This target hook returns @code{true} past the point in which new jump
9493 instructions could be created. On machines that require a register for
9494 every jump such as the SHmedia ISA of SH5, this point would typically be
9495 reload, so this target hook should be defined to a function such as:
9499 cannot_modify_jumps_past_reload_p ()
9501 return (reload_completed || reload_in_progress);
9506 @deftypefn {Target Hook} int TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
9507 This target hook returns a register class for which branch target register
9508 optimizations should be applied. All registers in this class should be
9509 usable interchangeably. After reload, registers in this class will be
9510 re-allocated and loads will be hoisted out of loops and be subjected
9511 to inter-block scheduling.
9514 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
9515 Branch target register optimization will by default exclude callee-saved
9517 that are not already live during the current function; if this target hook
9518 returns true, they will be included. The target code must than make sure
9519 that all target registers in the class returned by
9520 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
9521 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
9522 epilogues have already been generated. Note, even if you only return
9523 true when @var{after_prologue_epilogue_gen} is false, you still are likely
9524 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
9525 to reserve space for caller-saved target registers.
9528 @defmac POWI_MAX_MULTS
9529 If defined, this macro is interpreted as a signed integer C expression
9530 that specifies the maximum number of floating point multiplications
9531 that should be emitted when expanding exponentiation by an integer
9532 constant inline. When this value is defined, exponentiation requiring
9533 more than this number of multiplications is implemented by calling the
9534 system library's @code{pow}, @code{powf} or @code{powl} routines.
9535 The default value places no upper bound on the multiplication count.
9538 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
9539 This target hook should register any extra include files for the
9540 target. The parameter @var{stdinc} indicates if normal include files
9541 are present. The parameter @var{sysroot} is the system root directory.
9542 The parameter @var{iprefix} is the prefix for the gcc directory.
9545 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
9546 This target hook should register any extra include files for the
9547 target before any standard headers. The parameter @var{stdinc}
9548 indicates if normal include files are present. The parameter
9549 @var{sysroot} is the system root directory. The parameter
9550 @var{iprefix} is the prefix for the gcc directory.
9553 @deftypefn Macro void TARGET_OPTF (char *@var{path})
9554 This target hook should register special include paths for the target.
9555 The parameter @var{path} is the include to register. On Darwin
9556 systems, this is used for Framework includes, which have semantics
9557 that are different from @option{-I}.
9560 @deftypefn {Target Hook} bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
9561 This target hook returns @code{true} if it is safe to use a local alias
9562 for a virtual function @var{fndecl} when constructing thunks,
9563 @code{false} otherwise. By default, the hook returns @code{true} for all
9564 functions, if a target supports aliases (i.e.@: defines
9565 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
9568 @defmac TARGET_FORMAT_TYPES
9569 If defined, this macro is the name of a global variable containing
9570 target-specific format checking information for the @option{-Wformat}
9571 option. The default is to have no target-specific format checks.
9574 @defmac TARGET_N_FORMAT_TYPES
9575 If defined, this macro is the number of entries in
9576 @code{TARGET_FORMAT_TYPES}.
9579 @defmac TARGET_USE_JCR_SECTION
9580 This macro determines whether to use the JCR section to register Java
9581 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
9582 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.