1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,
2 @c 2002, 2003 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 * Misc:: Everything else.
56 @node Target Structure
57 @section The Global @code{targetm} Variable
59 @cindex target functions
61 @deftypevar {struct gcc_target} targetm
62 The target @file{.c} file must define the global @code{targetm} variable
63 which contains pointers to functions and data relating to the target
64 machine. The variable is declared in @file{target.h};
65 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
66 used to initialize the variable, and macros for the default initializers
67 for elements of the structure. The @file{.c} file should override those
68 macros for which the default definition is inappropriate. For example:
71 #include "target-def.h"
73 /* @r{Initialize the GCC target structure.} */
75 #undef TARGET_COMP_TYPE_ATTRIBUTES
76 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
78 struct gcc_target targetm = TARGET_INITIALIZER;
82 Where a macro should be defined in the @file{.c} file in this manner to
83 form part of the @code{targetm} structure, it is documented below as a
84 ``Target Hook'' with a prototype. Many macros will change in future
85 from being defined in the @file{.h} file to being part of the
86 @code{targetm} structure.
89 @section Controlling the Compilation Driver, @file{gcc}
91 @cindex controlling the compilation driver
93 @c prevent bad page break with this line
94 You can control the compilation driver.
97 @findex SWITCH_TAKES_ARG
98 @item 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
110 @findex WORD_SWITCH_TAKES_ARG
111 @item 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
124 @findex SWITCH_CURTAILS_COMPILATION
125 @item 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.
139 @findex SWITCHES_NEED_SPACES
140 @item 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{""}.
146 @findex TARGET_OPTION_TRANSLATE_TABLE
147 @item 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 @code{-malt-abi},
158 @code{-EB}, and @code{-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" @}
166 @findex DRIVER_SELF_SPECS
167 @item 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 before choosing
172 the multilib directory or running any subcommands. It applies them in
173 the order given, so each spec can depend on the options added by
174 earlier ones. It is also possible to remove options using
175 @samp{%<@var{option}} in the usual way.
177 This macro can be useful when a port has several interdependent target
178 options. It provides a way of standardizing the command line so
179 that the other specs are easier to write.
181 Do not define this macro if it does not need to do anything.
185 A C string constant that tells the GCC driver program options to
186 pass to CPP@. It can also specify how to translate options you
187 give to GCC into options for GCC to pass to the CPP@.
189 Do not define this macro if it does not need to do anything.
191 @findex CPLUSPLUS_CPP_SPEC
192 @item CPLUSPLUS_CPP_SPEC
193 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
194 than C@. If you do not define this macro, then the value of
195 @code{CPP_SPEC} (if any) will be used instead.
199 A C string constant that tells the GCC driver program options to
200 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
202 It can also specify how to translate options you give to GCC into options
203 for GCC to pass to front ends.
205 Do not define this macro if it does not need to do anything.
209 A C string constant that tells the GCC driver program options to
210 pass to @code{cc1plus}. It can also specify how to translate options you
211 give to GCC into options for GCC to pass to the @code{cc1plus}.
213 Do not define this macro if it does not need to do anything.
214 Note that everything defined in CC1_SPEC is already passed to
215 @code{cc1plus} so there is no need to duplicate the contents of
216 CC1_SPEC in CC1PLUS_SPEC@.
220 A C string constant that tells the GCC driver program options to
221 pass to the assembler. It can also specify how to translate options
222 you give to GCC into options for GCC to pass to the assembler.
223 See the file @file{sun3.h} for an example of this.
225 Do not define this macro if it does not need to do anything.
227 @findex ASM_FINAL_SPEC
229 A C string constant that tells the GCC driver program how to
230 run any programs which cleanup after the normal assembler.
231 Normally, this is not needed. See the file @file{mips.h} for
234 Do not define this macro if it does not need to do anything.
236 @findex AS_NEEDS_DASH_FOR_PIPED_INPUT
237 @item AS_NEEDS_DASH_FOR_PIPED_INPUT
238 Define this macro, with no value, if the driver should give the assembler
239 an argument consisting of a single dash, @option{-}, to instruct it to
240 read from its standard input (which will be a pipe connected to the
241 output of the compiler proper). This argument is given after any
242 @option{-o} option specifying the name of the output file.
244 If you do not define this macro, the assembler is assumed to read its
245 standard input if given no non-option arguments. If your assembler
246 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
247 see @file{mips.h} for instance.
251 A C string constant that tells the GCC driver program options to
252 pass to the linker. It can also specify how to translate options you
253 give to GCC into options for GCC to pass to the linker.
255 Do not define this macro if it does not need to do anything.
259 Another C string constant used much like @code{LINK_SPEC}. The difference
260 between the two is that @code{LIB_SPEC} is used at the end of the
261 command given to the linker.
263 If this macro is not defined, a default is provided that
264 loads the standard C library from the usual place. See @file{gcc.c}.
268 Another C string constant that tells the GCC driver program
269 how and when to place a reference to @file{libgcc.a} into the
270 linker command line. This constant is placed both before and after
271 the value of @code{LIB_SPEC}.
273 If this macro is not defined, the GCC driver provides a default that
274 passes the string @option{-lgcc} to the linker.
276 @findex STARTFILE_SPEC
278 Another C string constant used much like @code{LINK_SPEC}. The
279 difference between the two is that @code{STARTFILE_SPEC} is used at
280 the very beginning of the command given to the linker.
282 If this macro is not defined, a default is provided that loads the
283 standard C startup file from the usual place. See @file{gcc.c}.
287 Another C string constant used much like @code{LINK_SPEC}. The
288 difference between the two is that @code{ENDFILE_SPEC} is used at
289 the very end of the command given to the linker.
291 Do not define this macro if it does not need to do anything.
293 @findex THREAD_MODEL_SPEC
294 @item THREAD_MODEL_SPEC
295 GCC @code{-v} will print the thread model GCC was configured to use.
296 However, this doesn't work on platforms that are multilibbed on thread
297 models, such as AIX 4.3. On such platforms, define
298 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
299 blanks that names one of the recognized thread models. @code{%*}, the
300 default value of this macro, will expand to the value of
301 @code{thread_file} set in @file{config.gcc}.
305 Define this macro to provide additional specifications to put in the
306 @file{specs} file that can be used in various specifications like
309 The definition should be an initializer for an array of structures,
310 containing a string constant, that defines the specification name, and a
311 string constant that provides the specification.
313 Do not define this macro if it does not need to do anything.
315 @code{EXTRA_SPECS} is useful when an architecture contains several
316 related targets, which have various @code{@dots{}_SPECS} which are similar
317 to each other, and the maintainer would like one central place to keep
320 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
321 define either @code{_CALL_SYSV} when the System V calling sequence is
322 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
325 The @file{config/rs6000/rs6000.h} target file defines:
328 #define EXTRA_SPECS \
329 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
331 #define CPP_SYS_DEFAULT ""
334 The @file{config/rs6000/sysv.h} target file defines:
338 "%@{posix: -D_POSIX_SOURCE @} \
339 %@{mcall-sysv: -D_CALL_SYSV @} %@{mcall-aix: -D_CALL_AIX @} \
340 %@{!mcall-sysv: %@{!mcall-aix: %(cpp_sysv_default) @}@} \
341 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
343 #undef CPP_SYSV_DEFAULT
344 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
347 while the @file{config/rs6000/eabiaix.h} target file defines
348 @code{CPP_SYSV_DEFAULT} as:
351 #undef CPP_SYSV_DEFAULT
352 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
355 @findex LINK_LIBGCC_SPECIAL
356 @item LINK_LIBGCC_SPECIAL
357 Define this macro if the driver program should find the library
358 @file{libgcc.a} itself and should not pass @option{-L} options to the
359 linker. If you do not define this macro, the driver program will pass
360 the argument @option{-lgcc} to tell the linker to do the search and will
361 pass @option{-L} options to it.
363 @findex LINK_LIBGCC_SPECIAL_1
364 @item LINK_LIBGCC_SPECIAL_1
365 Define this macro if the driver program should find the library
366 @file{libgcc.a}. If you do not define this macro, the driver program will pass
367 the argument @option{-lgcc} to tell the linker to do the search.
368 This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does
369 not affect @option{-L} options.
371 @findex LINK_GCC_C_SEQUENCE_SPEC
372 @item LINK_GCC_C_SEQUENCE_SPEC
373 The sequence in which libgcc and libc are specified to the linker.
374 By default this is @code{%G %L %G}.
376 @findex LINK_COMMAND_SPEC
377 @item LINK_COMMAND_SPEC
378 A C string constant giving the complete command line need to execute the
379 linker. When you do this, you will need to update your port each time a
380 change is made to the link command line within @file{gcc.c}. Therefore,
381 define this macro only if you need to completely redefine the command
382 line for invoking the linker and there is no other way to accomplish
383 the effect you need. Overriding this macro may be avoidable by overriding
384 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
386 @findex LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
387 @item LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
388 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
389 directories from linking commands. Do not give it a nonzero value if
390 removing duplicate search directories changes the linker's semantics.
392 @findex MULTILIB_DEFAULTS
393 @item MULTILIB_DEFAULTS
394 Define this macro as a C expression for the initializer of an array of
395 string to tell the driver program which options are defaults for this
396 target and thus do not need to be handled specially when using
397 @code{MULTILIB_OPTIONS}.
399 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
400 the target makefile fragment or if none of the options listed in
401 @code{MULTILIB_OPTIONS} are set by default.
402 @xref{Target Fragment}.
404 @findex RELATIVE_PREFIX_NOT_LINKDIR
405 @item RELATIVE_PREFIX_NOT_LINKDIR
406 Define this macro to tell @code{gcc} that it should only translate
407 a @option{-B} prefix into a @option{-L} linker option if the prefix
408 indicates an absolute file name.
410 @findex STANDARD_EXEC_PREFIX
411 @item STANDARD_EXEC_PREFIX
412 Define this macro as a C string constant if you wish to override the
413 standard choice of @file{/usr/local/lib/gcc-lib/} as the default prefix to
414 try when searching for the executable files of the compiler.
416 @findex MD_EXEC_PREFIX
418 If defined, this macro is an additional prefix to try after
419 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
420 when the @option{-b} option is used, or the compiler is built as a cross
421 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
422 to the list of directories used to find the assembler in @file{configure.in}.
424 @findex STANDARD_STARTFILE_PREFIX
425 @item STANDARD_STARTFILE_PREFIX
426 Define this macro as a C string constant if you wish to override the
427 standard choice of @file{/usr/local/lib/} as the default prefix to
428 try when searching for startup files such as @file{crt0.o}.
430 @findex MD_STARTFILE_PREFIX
431 @item MD_STARTFILE_PREFIX
432 If defined, this macro supplies an additional prefix to try after the
433 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
434 @option{-b} option is used, or when the compiler is built as a cross
437 @findex MD_STARTFILE_PREFIX_1
438 @item MD_STARTFILE_PREFIX_1
439 If defined, this macro supplies yet another prefix to try after the
440 standard prefixes. It is not searched when the @option{-b} option is
441 used, or when the compiler is built as a cross compiler.
443 @findex INIT_ENVIRONMENT
444 @item INIT_ENVIRONMENT
445 Define this macro as a C string constant if you wish to set environment
446 variables for programs called by the driver, such as the assembler and
447 loader. The driver passes the value of this macro to @code{putenv} to
448 initialize the necessary environment variables.
450 @findex LOCAL_INCLUDE_DIR
451 @item LOCAL_INCLUDE_DIR
452 Define this macro as a C string constant if you wish to override the
453 standard choice of @file{/usr/local/include} as the default prefix to
454 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
455 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
457 Cross compilers do not search either @file{/usr/local/include} or its
460 @findex MODIFY_TARGET_NAME
461 @item MODIFY_TARGET_NAME
462 Define this macro if you with to define command-line switches that modify the
465 For each switch, you can include a string to be appended to the first
466 part of the configuration name or a string to be deleted from the
467 configuration name, if present. The definition should be an initializer
468 for an array of structures. Each array element should have three
469 elements: the switch name (a string constant, including the initial
470 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
471 indicate whether the string should be inserted or deleted, and the string
472 to be inserted or deleted (a string constant).
474 For example, on a machine where @samp{64} at the end of the
475 configuration name denotes a 64-bit target and you want the @option{-32}
476 and @option{-64} switches to select between 32- and 64-bit targets, you would
480 #define MODIFY_TARGET_NAME \
481 @{ @{ "-32", DELETE, "64"@}, \
482 @{"-64", ADD, "64"@}@}
486 @findex SYSTEM_INCLUDE_DIR
487 @item SYSTEM_INCLUDE_DIR
488 Define this macro as a C string constant if you wish to specify a
489 system-specific directory to search for header files before the standard
490 directory. @code{SYSTEM_INCLUDE_DIR} comes before
491 @code{STANDARD_INCLUDE_DIR} in the search order.
493 Cross compilers do not use this macro and do not search the directory
496 @findex STANDARD_INCLUDE_DIR
497 @item STANDARD_INCLUDE_DIR
498 Define this macro as a C string constant if you wish to override the
499 standard choice of @file{/usr/include} as the default prefix to
500 try when searching for header files.
502 Cross compilers do not use this macro and do not search either
503 @file{/usr/include} or its replacement.
505 @findex STANDARD_INCLUDE_COMPONENT
506 @item STANDARD_INCLUDE_COMPONENT
507 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
508 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
509 If you do not define this macro, no component is used.
511 @findex INCLUDE_DEFAULTS
512 @item INCLUDE_DEFAULTS
513 Define this macro if you wish to override the entire default search path
514 for include files. For a native compiler, the default search path
515 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
516 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
517 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
518 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
519 and specify private search areas for GCC@. The directory
520 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
522 The definition should be an initializer for an array of structures.
523 Each array element should have four elements: the directory name (a
524 string constant), the component name (also a string constant), a flag
525 for C++-only directories,
526 and a flag showing that the includes in the directory don't need to be
527 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
528 the array with a null element.
530 The component name denotes what GNU package the include file is part of,
531 if any, in all upper-case letters. For example, it might be @samp{GCC}
532 or @samp{BINUTILS}. If the package is part of a vendor-supplied
533 operating system, code the component name as @samp{0}.
535 For example, here is the definition used for VAX/VMS:
538 #define INCLUDE_DEFAULTS \
540 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
541 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
542 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
549 Here is the order of prefixes tried for exec files:
553 Any prefixes specified by the user with @option{-B}.
556 The environment variable @code{GCC_EXEC_PREFIX}, if any.
559 The directories specified by the environment variable @code{COMPILER_PATH}.
562 The macro @code{STANDARD_EXEC_PREFIX}.
565 @file{/usr/lib/gcc/}.
568 The macro @code{MD_EXEC_PREFIX}, if any.
571 Here is the order of prefixes tried for startfiles:
575 Any prefixes specified by the user with @option{-B}.
578 The environment variable @code{GCC_EXEC_PREFIX}, if any.
581 The directories specified by the environment variable @code{LIBRARY_PATH}
582 (or port-specific name; native only, cross compilers do not use this).
585 The macro @code{STANDARD_EXEC_PREFIX}.
588 @file{/usr/lib/gcc/}.
591 The macro @code{MD_EXEC_PREFIX}, if any.
594 The macro @code{MD_STARTFILE_PREFIX}, if any.
597 The macro @code{STANDARD_STARTFILE_PREFIX}.
606 @node Run-time Target
607 @section Run-time Target Specification
608 @cindex run-time target specification
609 @cindex predefined macros
610 @cindex target specifications
612 @c prevent bad page break with this line
613 Here are run-time target specifications.
616 @findex TARGET_CPU_CPP_BUILTINS
617 @item TARGET_CPU_CPP_BUILTINS()
618 This function-like macro expands to a block of code that defines
619 built-in preprocessor macros and assertions for the target cpu, using
620 the functions @code{builtin_define}, @code{builtin_define_std} and
621 @code{builtin_assert} defined in @file{c-common.c}. When the front end
622 calls this macro it provides a trailing semicolon, and since it has
623 finished command line option processing your code can use those
626 @code{builtin_assert} takes a string in the form you pass to the
627 command-line option @option{-A}, such as @code{cpu=mips}, and creates
628 the assertion. @code{builtin_define} takes a string in the form
629 accepted by option @option{-D} and unconditionally defines the macro.
631 @code{builtin_define_std} takes a string representing the name of an
632 object-like macro. If it doesn't lie in the user's namespace,
633 @code{builtin_define_std} defines it unconditionally. Otherwise, it
634 defines a version with two leading underscores, and another version
635 with two leading and trailing underscores, and defines the original
636 only if an ISO standard was not requested on the command line. For
637 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
638 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
639 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
640 defines only @code{_ABI64}.
642 You can also test for the C dialect being compiled. The variable
643 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
644 or @code{clk_objective_c}. Note that if we are preprocessing
645 assembler, this variable will be @code{clk_c} but the function-like
646 macro @code{preprocessing_asm_p()} will return true, so you might want
647 to check for that first. If you need to check for strict ANSI, the
648 variable @code{flag_iso} can be used. The function-like macro
649 @code{preprocessing_trad_p()} can be used to check for traditional
652 With @code{TARGET_OS_CPP_BUILTINS} this macro obsoletes the
653 @code{CPP_PREDEFINES} target macro.
655 @findex TARGET_OS_CPP_BUILTINS
656 @item TARGET_OS_CPP_BUILTINS()
657 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
658 and is used for the target operating system instead.
660 With @code{TARGET_CPU_CPP_BUILTINS} this macro obsoletes the
661 @code{CPP_PREDEFINES} target macro.
663 @findex CPP_PREDEFINES
665 Define this to be a string constant containing @option{-D} options to
666 define the predefined macros that identify this machine and system.
667 These macros will be predefined unless the @option{-ansi} option (or a
668 @option{-std} option for strict ISO C conformance) is specified.
670 In addition, a parallel set of macros are predefined, whose names are
671 made by appending @samp{__} at the beginning and at the end. These
672 @samp{__} macros are permitted by the ISO standard, so they are
673 predefined regardless of whether @option{-ansi} or a @option{-std} option
676 For example, on the Sun, one can use the following value:
679 "-Dmc68000 -Dsun -Dunix"
682 The result is to define the macros @code{__mc68000__}, @code{__sun__}
683 and @code{__unix__} unconditionally, and the macros @code{mc68000},
684 @code{sun} and @code{unix} provided @option{-ansi} is not specified.
686 @findex extern int target_flags
687 @item extern int target_flags;
688 This declaration should be present.
690 @cindex optional hardware or system features
691 @cindex features, optional, in system conventions
693 This series of macros is to allow compiler command arguments to
694 enable or disable the use of optional features of the target machine.
695 For example, one machine description serves both the 68000 and
696 the 68020; a command argument tells the compiler whether it should
697 use 68020-only instructions or not. This command argument works
698 by means of a macro @code{TARGET_68020} that tests a bit in
701 Define a macro @code{TARGET_@var{featurename}} for each such option.
702 Its definition should test a bit in @code{target_flags}. It is
703 recommended that a helper macro @code{TARGET_MASK_@var{featurename}}
704 is defined for each bit-value to test, and used in
705 @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For
709 #define TARGET_MASK_68020 1
710 #define TARGET_68020 (target_flags & TARGET_MASK_68020)
713 One place where these macros are used is in the condition-expressions
714 of instruction patterns. Note how @code{TARGET_68020} appears
715 frequently in the 68000 machine description file, @file{m68k.md}.
716 Another place they are used is in the definitions of the other
717 macros in the @file{@var{machine}.h} file.
719 @findex TARGET_SWITCHES
720 @item TARGET_SWITCHES
721 This macro defines names of command options to set and clear
722 bits in @code{target_flags}. Its definition is an initializer
723 with a subgrouping for each command option.
725 Each subgrouping contains a string constant, that defines the option
726 name, a number, which contains the bits to set in
727 @code{target_flags}, and a second string which is the description
728 displayed by @option{--help}. If the number is negative then the bits specified
729 by the number are cleared instead of being set. If the description
730 string is present but empty, then no help information will be displayed
731 for that option, but it will not count as an undocumented option. The
732 actual option name is made by appending @samp{-m} to the specified name.
733 Non-empty description strings should be marked with @code{N_(@dots{})} for
734 @command{xgettext}. Please do not mark empty strings because the empty
735 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
736 of the message catalog with meta information, not the empty string.
738 In addition to the description for @option{--help},
739 more detailed documentation for each option should be added to
742 One of the subgroupings should have a null string. The number in
743 this grouping is the default value for @code{target_flags}. Any
744 target options act starting with that value.
746 Here is an example which defines @option{-m68000} and @option{-m68020}
747 with opposite meanings, and picks the latter as the default:
750 #define TARGET_SWITCHES \
751 @{ @{ "68020", TARGET_MASK_68020, "" @}, \
752 @{ "68000", -TARGET_MASK_68020, \
753 N_("Compile for the 68000") @}, \
754 @{ "", TARGET_MASK_68020, "" @}@}
757 @findex TARGET_OPTIONS
759 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
760 options that have values. Its definition is an initializer with a
761 subgrouping for each command option.
763 Each subgrouping contains a string constant, that defines the fixed part
764 of the option name, the address of a variable, and a description string.
765 Non-empty description strings should be marked with @code{N_(@dots{})} for
766 @command{xgettext}. Please do not mark empty strings because the empty
767 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
768 of the message catalog with meta information, not the empty string.
770 The variable, type @code{char *}, is set to the variable part of the
771 given option if the fixed part matches. The actual option name is made
772 by appending @samp{-m} to the specified name. Again, each option should
773 also be documented in @file{invoke.texi}.
775 Here is an example which defines @option{-mshort-data-@var{number}}. If the
776 given option is @option{-mshort-data-512}, the variable @code{m88k_short_data}
777 will be set to the string @code{"512"}.
780 extern char *m88k_short_data;
781 #define TARGET_OPTIONS \
782 @{ @{ "short-data-", &m88k_short_data, \
783 N_("Specify the size of the short data section") @} @}
786 @findex TARGET_VERSION
788 This macro is a C statement to print on @code{stderr} a string
789 describing the particular machine description choice. Every machine
790 description should define @code{TARGET_VERSION}. For example:
794 #define TARGET_VERSION \
795 fprintf (stderr, " (68k, Motorola syntax)");
797 #define TARGET_VERSION \
798 fprintf (stderr, " (68k, MIT syntax)");
802 @findex OVERRIDE_OPTIONS
803 @item OVERRIDE_OPTIONS
804 Sometimes certain combinations of command options do not make sense on
805 a particular target machine. You can define a macro
806 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
807 defined, is executed once just after all the command options have been
810 Don't use this macro to turn on various extra optimizations for
811 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
813 @findex OPTIMIZATION_OPTIONS
814 @item OPTIMIZATION_OPTIONS (@var{level}, @var{size})
815 Some machines may desire to change what optimizations are performed for
816 various optimization levels. This macro, if defined, is executed once
817 just after the optimization level is determined and before the remainder
818 of the command options have been parsed. Values set in this macro are
819 used as the default values for the other command line options.
821 @var{level} is the optimization level specified; 2 if @option{-O2} is
822 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
824 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
826 You should not use this macro to change options that are not
827 machine-specific. These should uniformly selected by the same
828 optimization level on all supported machines. Use this macro to enable
829 machine-specific optimizations.
831 @strong{Do not examine @code{write_symbols} in
832 this macro!} The debugging options are not supposed to alter the
835 @findex CAN_DEBUG_WITHOUT_FP
836 @item CAN_DEBUG_WITHOUT_FP
837 Define this macro if debugging can be performed even without a frame
838 pointer. If this macro is defined, GCC will turn on the
839 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
842 @node Per-Function Data
843 @section Defining data structures for per-function information.
844 @cindex per-function data
845 @cindex data structures
847 If the target needs to store information on a per-function basis, GCC
848 provides a macro and a couple of variables to allow this. Note, just
849 using statics to store the information is a bad idea, since GCC supports
850 nested functions, so you can be halfway through encoding one function
851 when another one comes along.
853 GCC defines a data structure called @code{struct function} which
854 contains all of the data specific to an individual function. This
855 structure contains a field called @code{machine} whose type is
856 @code{struct machine_function *}, which can be used by targets to point
857 to their own specific data.
859 If a target needs per-function specific data it should define the type
860 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
861 This macro should be used to initialize the function pointer
862 @code{init_machine_status}. This pointer is explained below.
864 One typical use of per-function, target specific data is to create an
865 RTX to hold the register containing the function's return address. This
866 RTX can then be used to implement the @code{__builtin_return_address}
867 function, for level 0.
869 Note---earlier implementations of GCC used a single data area to hold
870 all of the per-function information. Thus when processing of a nested
871 function began the old per-function data had to be pushed onto a
872 stack, and when the processing was finished, it had to be popped off the
873 stack. GCC used to provide function pointers called
874 @code{save_machine_status} and @code{restore_machine_status} to handle
875 the saving and restoring of the target specific information. Since the
876 single data area approach is no longer used, these pointers are no
879 The macro and function pointers are described below.
882 @findex INIT_EXPANDERS
884 Macro called to initialize any target specific information. This macro
885 is called once per function, before generation of any RTL has begun.
886 The intention of this macro is to allow the initialization of the
887 function pointers below.
889 @findex init_machine_status
890 @item init_machine_status
891 This is a @code{void (*)(struct function *)} function pointer. If this
892 pointer is non-@code{NULL} it will be called once per function, before function
893 compilation starts, in order to allow the target to perform any target
894 specific initialization of the @code{struct function} structure. It is
895 intended that this would be used to initialize the @code{machine} of
898 @code{struct machine_function} structures are expected to be freed by GC.
899 Generally, any memory that they reference must be allocated by using
900 @code{ggc_alloc}, including the structure itself.
905 @section Storage Layout
906 @cindex storage layout
908 Note that the definitions of the macros in this table which are sizes or
909 alignments measured in bits do not need to be constant. They can be C
910 expressions that refer to static variables, such as the @code{target_flags}.
911 @xref{Run-time Target}.
914 @findex BITS_BIG_ENDIAN
915 @item BITS_BIG_ENDIAN
916 Define this macro to have the value 1 if the most significant bit in a
917 byte has the lowest number; otherwise define it to have the value zero.
918 This means that bit-field instructions count from the most significant
919 bit. If the machine has no bit-field instructions, then this must still
920 be defined, but it doesn't matter which value it is defined to. This
921 macro need not be a constant.
923 This macro does not affect the way structure fields are packed into
924 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
926 @findex BYTES_BIG_ENDIAN
927 @item BYTES_BIG_ENDIAN
928 Define this macro to have the value 1 if the most significant byte in a
929 word has the lowest number. This macro need not be a constant.
931 @findex WORDS_BIG_ENDIAN
932 @item WORDS_BIG_ENDIAN
933 Define this macro to have the value 1 if, in a multiword object, the
934 most significant word has the lowest number. This applies to both
935 memory locations and registers; GCC fundamentally assumes that the
936 order of words in memory is the same as the order in registers. This
937 macro need not be a constant.
939 @findex LIBGCC2_WORDS_BIG_ENDIAN
940 @item LIBGCC2_WORDS_BIG_ENDIAN
941 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
942 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
943 used only when compiling @file{libgcc2.c}. Typically the value will be set
944 based on preprocessor defines.
946 @findex FLOAT_WORDS_BIG_ENDIAN
947 @item FLOAT_WORDS_BIG_ENDIAN
948 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
949 @code{TFmode} floating point numbers are stored in memory with the word
950 containing the sign bit at the lowest address; otherwise define it to
951 have the value 0. This macro need not be a constant.
953 You need not define this macro if the ordering is the same as for
956 @findex BITS_PER_UNIT
958 Define this macro to be the number of bits in an addressable storage
959 unit (byte). If you do not define this macro the default is 8.
961 @findex BITS_PER_WORD
963 Number of bits in a word. If you do not define this macro, the default
964 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
966 @findex MAX_BITS_PER_WORD
967 @item MAX_BITS_PER_WORD
968 Maximum number of bits in a word. If this is undefined, the default is
969 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
970 largest value that @code{BITS_PER_WORD} can have at run-time.
972 @findex UNITS_PER_WORD
974 Number of storage units in a word; normally 4.
976 @findex MIN_UNITS_PER_WORD
977 @item MIN_UNITS_PER_WORD
978 Minimum number of units in a word. If this is undefined, the default is
979 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
980 smallest value that @code{UNITS_PER_WORD} can have at run-time.
984 Width of a pointer, in bits. You must specify a value no wider than the
985 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
986 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
987 a value the default is @code{BITS_PER_WORD}.
989 @findex POINTERS_EXTEND_UNSIGNED
990 @item POINTERS_EXTEND_UNSIGNED
991 A C expression whose value is greater than zero if pointers that need to be
992 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
993 be zero-extended and zero if they are to be sign-extended. If the value
994 is less then zero then there must be an "ptr_extend" instruction that
995 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
997 You need not define this macro if the @code{POINTER_SIZE} is equal
998 to the width of @code{Pmode}.
1000 @findex PROMOTE_MODE
1001 @item PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
1002 A macro to update @var{m} and @var{unsignedp} when an object whose type
1003 is @var{type} and which has the specified mode and signedness is to be
1004 stored in a register. This macro is only called when @var{type} is a
1007 On most RISC machines, which only have operations that operate on a full
1008 register, define this macro to set @var{m} to @code{word_mode} if
1009 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
1010 cases, only integer modes should be widened because wider-precision
1011 floating-point operations are usually more expensive than their narrower
1014 For most machines, the macro definition does not change @var{unsignedp}.
1015 However, some machines, have instructions that preferentially handle
1016 either signed or unsigned quantities of certain modes. For example, on
1017 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
1018 sign-extend the result to 64 bits. On such machines, set
1019 @var{unsignedp} according to which kind of extension is more efficient.
1021 Do not define this macro if it would never modify @var{m}.
1023 @findex PROMOTE_FUNCTION_ARGS
1024 @item PROMOTE_FUNCTION_ARGS
1025 Define this macro if the promotion described by @code{PROMOTE_MODE}
1026 should also be done for outgoing function arguments.
1028 @findex PROMOTE_FUNCTION_RETURN
1029 @item PROMOTE_FUNCTION_RETURN
1030 Define this macro if the promotion described by @code{PROMOTE_MODE}
1031 should also be done for the return value of functions.
1033 If this macro is defined, @code{FUNCTION_VALUE} must perform the same
1034 promotions done by @code{PROMOTE_MODE}.
1036 @findex PROMOTE_FOR_CALL_ONLY
1037 @item PROMOTE_FOR_CALL_ONLY
1038 Define this macro if the promotion described by @code{PROMOTE_MODE}
1039 should @emph{only} be performed for outgoing function arguments or
1040 function return values, as specified by @code{PROMOTE_FUNCTION_ARGS}
1041 and @code{PROMOTE_FUNCTION_RETURN}, respectively.
1043 @findex PARM_BOUNDARY
1045 Normal alignment required for function parameters on the stack, in
1046 bits. All stack parameters receive at least this much alignment
1047 regardless of data type. On most machines, this is the same as the
1050 @findex STACK_BOUNDARY
1051 @item STACK_BOUNDARY
1052 Define this macro to the minimum alignment enforced by hardware for the
1053 stack pointer on this machine. The definition is a C expression for the
1054 desired alignment (measured in bits). This value is used as a default
1055 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1056 this should be the same as @code{PARM_BOUNDARY}.
1058 @findex PREFERRED_STACK_BOUNDARY
1059 @item PREFERRED_STACK_BOUNDARY
1060 Define this macro if you wish to preserve a certain alignment for the
1061 stack pointer, greater than what the hardware enforces. The definition
1062 is a C expression for the desired alignment (measured in bits). This
1063 macro must evaluate to a value equal to or larger than
1064 @code{STACK_BOUNDARY}.
1066 @findex FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1067 @item FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1068 A C expression that evaluates true if @code{PREFERRED_STACK_BOUNDARY} is
1069 not guaranteed by the runtime and we should emit code to align the stack
1070 at the beginning of @code{main}.
1072 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
1073 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
1074 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
1075 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
1076 be momentarily unaligned while pushing arguments.
1078 @findex FUNCTION_BOUNDARY
1079 @item FUNCTION_BOUNDARY
1080 Alignment required for a function entry point, in bits.
1082 @findex BIGGEST_ALIGNMENT
1083 @item BIGGEST_ALIGNMENT
1084 Biggest alignment that any data type can require on this machine, in bits.
1086 @findex MINIMUM_ATOMIC_ALIGNMENT
1087 @item MINIMUM_ATOMIC_ALIGNMENT
1088 If defined, the smallest alignment, in bits, that can be given to an
1089 object that can be referenced in one operation, without disturbing any
1090 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1091 on machines that don't have byte or half-word store operations.
1093 @findex BIGGEST_FIELD_ALIGNMENT
1094 @item BIGGEST_FIELD_ALIGNMENT
1095 Biggest alignment that any structure or union field can require on this
1096 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1097 structure and union fields only, unless the field alignment has been set
1098 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1100 @findex ADJUST_FIELD_ALIGN
1101 @item ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1102 An expression for the alignment of a structure field @var{field} if the
1103 alignment computed in the usual way (including applying of
1104 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1105 alignment) is @var{computed}. It overrides alignment only if the
1106 field alignment has not been set by the
1107 @code{__attribute__ ((aligned (@var{n})))} construct.
1109 @findex MAX_OFILE_ALIGNMENT
1110 @item MAX_OFILE_ALIGNMENT
1111 Biggest alignment supported by the object file format of this machine.
1112 Use this macro to limit the alignment which can be specified using the
1113 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1114 the default value is @code{BIGGEST_ALIGNMENT}.
1116 @findex DATA_ALIGNMENT
1117 @item DATA_ALIGNMENT (@var{type}, @var{basic-align})
1118 If defined, a C expression to compute the alignment for a variable in
1119 the static store. @var{type} is the data type, and @var{basic-align} is
1120 the alignment that the object would ordinarily have. The value of this
1121 macro is used instead of that alignment to align the object.
1123 If this macro is not defined, then @var{basic-align} is used.
1126 One use of this macro is to increase alignment of medium-size data to
1127 make it all fit in fewer cache lines. Another is to cause character
1128 arrays to be word-aligned so that @code{strcpy} calls that copy
1129 constants to character arrays can be done inline.
1131 @findex CONSTANT_ALIGNMENT
1132 @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1133 If defined, a C expression to compute the alignment given to a constant
1134 that is being placed in memory. @var{constant} is the constant and
1135 @var{basic-align} is the alignment that the object would ordinarily
1136 have. The value of this macro is used instead of that alignment to
1139 If this macro is not defined, then @var{basic-align} is used.
1141 The typical use of this macro is to increase alignment for string
1142 constants to be word aligned so that @code{strcpy} calls that copy
1143 constants can be done inline.
1145 @findex LOCAL_ALIGNMENT
1146 @item LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1147 If defined, a C expression to compute the alignment for a variable in
1148 the local store. @var{type} is the data type, and @var{basic-align} is
1149 the alignment that the object would ordinarily have. The value of this
1150 macro is used instead of that alignment to align the object.
1152 If this macro is not defined, then @var{basic-align} is used.
1154 One use of this macro is to increase alignment of medium-size data to
1155 make it all fit in fewer cache lines.
1157 @findex EMPTY_FIELD_BOUNDARY
1158 @item EMPTY_FIELD_BOUNDARY
1159 Alignment in bits to be given to a structure bit-field that follows an
1160 empty field such as @code{int : 0;}.
1162 Note that @code{PCC_BITFIELD_TYPE_MATTERS} also affects the alignment
1163 that results from an empty field.
1165 @findex STRUCTURE_SIZE_BOUNDARY
1166 @item STRUCTURE_SIZE_BOUNDARY
1167 Number of bits which any structure or union's size must be a multiple of.
1168 Each structure or union's size is rounded up to a multiple of this.
1170 If you do not define this macro, the default is the same as
1171 @code{BITS_PER_UNIT}.
1173 @findex STRICT_ALIGNMENT
1174 @item STRICT_ALIGNMENT
1175 Define this macro to be the value 1 if instructions will fail to work
1176 if given data not on the nominal alignment. If instructions will merely
1177 go slower in that case, define this macro as 0.
1179 @findex PCC_BITFIELD_TYPE_MATTERS
1180 @item PCC_BITFIELD_TYPE_MATTERS
1181 Define this if you wish to imitate the way many other C compilers handle
1182 alignment of bit-fields and the structures that contain them.
1184 The behavior is that the type written for a named bit-field (@code{int},
1185 @code{short}, or other integer type) imposes an alignment for the entire
1186 structure, as if the structure really did contain an ordinary field of
1187 that type. In addition, the bit-field is placed within the structure so
1188 that it would fit within such a field, not crossing a boundary for it.
1190 Thus, on most machines, a named bit-field whose type is written as
1191 @code{int} would not cross a four-byte boundary, and would force
1192 four-byte alignment for the whole structure. (The alignment used may
1193 not be four bytes; it is controlled by the other alignment parameters.)
1195 An unnamed bit-field will not affect the alignment of the containing
1198 If the macro is defined, its definition should be a C expression;
1199 a nonzero value for the expression enables this behavior.
1201 Note that if this macro is not defined, or its value is zero, some
1202 bit-fields may cross more than one alignment boundary. The compiler can
1203 support such references if there are @samp{insv}, @samp{extv}, and
1204 @samp{extzv} insns that can directly reference memory.
1206 The other known way of making bit-fields work is to define
1207 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1208 Then every structure can be accessed with fullwords.
1210 Unless the machine has bit-field instructions or you define
1211 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1212 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1214 If your aim is to make GCC use the same conventions for laying out
1215 bit-fields as are used by another compiler, here is how to investigate
1216 what the other compiler does. Compile and run this program:
1235 printf ("Size of foo1 is %d\n",
1236 sizeof (struct foo1));
1237 printf ("Size of foo2 is %d\n",
1238 sizeof (struct foo2));
1243 If this prints 2 and 5, then the compiler's behavior is what you would
1244 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1246 @findex BITFIELD_NBYTES_LIMITED
1247 @item BITFIELD_NBYTES_LIMITED
1248 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1249 to aligning a bit-field within the structure.
1251 @findex MEMBER_TYPE_FORCES_BLK
1252 @item MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1253 Return 1 if a structure or array containing @var{field} should be accessed using
1256 If @var{field} is the only field in the structure, @var{mode} is its
1257 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1258 case where structures of one field would require the structure's mode to
1259 retain the field's mode.
1261 Normally, this is not needed. See the file @file{c4x.h} for an example
1262 of how to use this macro to prevent a structure having a floating point
1263 field from being accessed in an integer mode.
1265 @findex ROUND_TYPE_SIZE
1266 @item ROUND_TYPE_SIZE (@var{type}, @var{computed}, @var{specified})
1267 Define this macro as an expression for the overall size of a type
1268 (given by @var{type} as a tree node) when the size computed in the
1269 usual way is @var{computed} and the alignment is @var{specified}.
1271 The default is to round @var{computed} up to a multiple of @var{specified}.
1273 @findex ROUND_TYPE_SIZE_UNIT
1274 @item ROUND_TYPE_SIZE_UNIT (@var{type}, @var{computed}, @var{specified})
1275 Similar to @code{ROUND_TYPE_SIZE}, but sizes and alignments are
1276 specified in units (bytes). If you define @code{ROUND_TYPE_SIZE},
1277 you must also define this macro and they must be defined consistently
1280 @findex ROUND_TYPE_ALIGN
1281 @item ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1282 Define this macro as an expression for the alignment of a type (given
1283 by @var{type} as a tree node) if the alignment computed in the usual
1284 way is @var{computed} and the alignment explicitly specified was
1287 The default is to use @var{specified} if it is larger; otherwise, use
1288 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1290 @findex MAX_FIXED_MODE_SIZE
1291 @item MAX_FIXED_MODE_SIZE
1292 An integer expression for the size in bits of the largest integer
1293 machine mode that should actually be used. All integer machine modes of
1294 this size or smaller can be used for structures and unions with the
1295 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1296 (DImode)} is assumed.
1298 @findex VECTOR_MODE_SUPPORTED_P
1299 @item VECTOR_MODE_SUPPORTED_P(@var{mode})
1300 Define this macro to be nonzero if the port is prepared to handle insns
1301 involving vector mode @var{mode}. At the very least, it must have move
1302 patterns for this mode.
1304 @findex STACK_SAVEAREA_MODE
1305 @item STACK_SAVEAREA_MODE (@var{save_level})
1306 If defined, an expression of type @code{enum machine_mode} that
1307 specifies the mode of the save area operand of a
1308 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1309 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1310 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1311 having its mode specified.
1313 You need not define this macro if it always returns @code{Pmode}. You
1314 would most commonly define this macro if the
1315 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1318 @findex STACK_SIZE_MODE
1319 @item STACK_SIZE_MODE
1320 If defined, an expression of type @code{enum machine_mode} that
1321 specifies the mode of the size increment operand of an
1322 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1324 You need not define this macro if it always returns @code{word_mode}.
1325 You would most commonly define this macro if the @code{allocate_stack}
1326 pattern needs to support both a 32- and a 64-bit mode.
1328 @findex TARGET_FLOAT_FORMAT
1329 @item TARGET_FLOAT_FORMAT
1330 A code distinguishing the floating point format of the target machine.
1331 There are five defined values:
1334 @findex IEEE_FLOAT_FORMAT
1335 @item IEEE_FLOAT_FORMAT
1336 This code indicates IEEE floating point. It is the default; there is no
1337 need to define this macro when the format is IEEE@.
1339 @findex VAX_FLOAT_FORMAT
1340 @item VAX_FLOAT_FORMAT
1341 This code indicates the ``F float'' (for @code{float}) and ``D float''
1342 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
1344 @findex IBM_FLOAT_FORMAT
1345 @item IBM_FLOAT_FORMAT
1346 This code indicates the format used on the IBM System/370.
1348 @findex C4X_FLOAT_FORMAT
1349 @item C4X_FLOAT_FORMAT
1350 This code indicates the format used on the TMS320C3x/C4x.
1352 @findex UNKNOWN_FLOAT_FORMAT
1353 @item UNKNOWN_FLOAT_FORMAT
1354 This code indicates any other format.
1358 formats are actually in use on supported machines, new codes should be
1361 The ordering of the component words of floating point values stored in
1362 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1364 @findex MODE_HAS_NANS
1365 @item MODE_HAS_NANS (@var{mode})
1366 When defined, this macro should be true if @var{mode} has a NaN
1367 representation. The compiler assumes that NaNs are not equal to
1368 anything (including themselves) and that addition, subtraction,
1369 multiplication and division all return NaNs when one operand is
1372 By default, this macro is true if @var{mode} is a floating-point
1373 mode and the target floating-point format is IEEE@.
1375 @findex MODE_HAS_INFINITIES
1376 @item MODE_HAS_INFINITIES (@var{mode})
1377 This macro should be true if @var{mode} can represent infinity. At
1378 present, the compiler uses this macro to decide whether @samp{x - x}
1379 is always defined. By default, the macro is true when @var{mode}
1380 is a floating-point mode and the target format is IEEE@.
1382 @findex MODE_HAS_SIGNED_ZEROS
1383 @item MODE_HAS_SIGNED_ZEROS (@var{mode})
1384 True if @var{mode} distinguishes between positive and negative zero.
1385 The rules are expected to follow the IEEE standard:
1389 @samp{x + x} has the same sign as @samp{x}.
1392 If the sum of two values with opposite sign is zero, the result is
1393 positive for all rounding modes expect towards @minus{}infinity, for
1394 which it is negative.
1397 The sign of a product or quotient is negative when exactly one
1398 of the operands is negative.
1401 The default definition is true if @var{mode} is a floating-point
1402 mode and the target format is IEEE@.
1404 @findex MODE_HAS_SIGN_DEPENDENT_ROUNDING
1405 @item MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1406 If defined, this macro should be true for @var{mode} if it has at
1407 least one rounding mode in which @samp{x} and @samp{-x} can be
1408 rounded to numbers of different magnitude. Two such modes are
1409 towards @minus{}infinity and towards +infinity.
1411 The default definition of this macro is true if @var{mode} is
1412 a floating-point mode and the target format is IEEE@.
1414 @findex ROUND_TOWARDS_ZERO
1415 @item ROUND_TOWARDS_ZERO
1416 If defined, this macro should be true if the prevailing rounding
1417 mode is towards zero. A true value has the following effects:
1421 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1424 @file{libgcc.a}'s floating-point emulator will round towards zero
1425 rather than towards nearest.
1428 The compiler's floating-point emulator will round towards zero after
1429 doing arithmetic, and when converting from the internal float format to
1433 The macro does not affect the parsing of string literals. When the
1434 primary rounding mode is towards zero, library functions like
1435 @code{strtod} might still round towards nearest, and the compiler's
1436 parser should behave like the target's @code{strtod} where possible.
1438 Not defining this macro is equivalent to returning zero.
1440 @findex LARGEST_EXPONENT_IS_NORMAL
1441 @item LARGEST_EXPONENT_IS_NORMAL (@var{size})
1442 This macro should return true if floats with @var{size}
1443 bits do not have a NaN or infinity representation, but use the largest
1444 exponent for normal numbers instead.
1446 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1447 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1448 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1449 floating-point arithmetic.
1451 The default definition of this macro returns false for all sizes.
1454 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1455 This target hook returns @code{true} if bit-fields in the given
1456 @var{record_type} are to be laid out following the rules of Microsoft
1457 Visual C/C++, namely: (i) a bit-field won't share the same storage
1458 unit with the previous bit-field if their underlying types have
1459 different sizes, and the bit-field will be aligned to the highest
1460 alignment of the underlying types of itself and of the previous
1461 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1462 the whole enclosing structure, even if it is unnamed; except that
1463 (iii) a zero-sized bit-field will be disregarded unless it follows
1464 another bit-field of nonzero size. If this hook returns @code{true},
1465 other macros that control bit-field layout are ignored.
1467 When a bit-field is inserted into a packed record, the whole size
1468 of the underlying type is used by one or more same-size adjacent
1469 bit-fields (that is, if its long:3, 32 bits is used in the record,
1470 and any additional adjacent long bit-fields are packed into the same
1471 chunk of 32 bits. However, if the size changes, a new field of that
1472 size is allocated). In an unpacked record, this is the same as using
1473 alignment, but not equivalent when packing.
1475 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1476 the latter will take precedence. If @samp{__attribute__((packed))} is
1477 used on a single field when MS bit-fields are in use, it will take
1478 precedence for that field, but the alignment of the rest of the structure
1479 may affect its placement.
1483 @section Layout of Source Language Data Types
1485 These macros define the sizes and other characteristics of the standard
1486 basic data types used in programs being compiled. Unlike the macros in
1487 the previous section, these apply to specific features of C and related
1488 languages, rather than to fundamental aspects of storage layout.
1491 @findex INT_TYPE_SIZE
1493 A C expression for the size in bits of the type @code{int} on the
1494 target machine. If you don't define this, the default is one word.
1496 @findex SHORT_TYPE_SIZE
1497 @item SHORT_TYPE_SIZE
1498 A C expression for the size in bits of the type @code{short} on the
1499 target machine. If you don't define this, the default is half a word.
1500 (If this would be less than one storage unit, it is rounded up to one
1503 @findex LONG_TYPE_SIZE
1504 @item LONG_TYPE_SIZE
1505 A C expression for the size in bits of the type @code{long} on the
1506 target machine. If you don't define this, the default is one word.
1508 @findex ADA_LONG_TYPE_SIZE
1509 @item ADA_LONG_TYPE_SIZE
1510 On some machines, the size used for the Ada equivalent of the type
1511 @code{long} by a native Ada compiler differs from that used by C. In
1512 that situation, define this macro to be a C expression to be used for
1513 the size of that type. If you don't define this, the default is the
1514 value of @code{LONG_TYPE_SIZE}.
1516 @findex MAX_LONG_TYPE_SIZE
1517 @item MAX_LONG_TYPE_SIZE
1518 Maximum number for the size in bits of the type @code{long} on the
1519 target machine. If this is undefined, the default is
1520 @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the
1521 largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is
1524 @findex LONG_LONG_TYPE_SIZE
1525 @item LONG_LONG_TYPE_SIZE
1526 A C expression for the size in bits of the type @code{long long} on the
1527 target machine. If you don't define this, the default is two
1528 words. If you want to support GNU Ada on your machine, the value of this
1529 macro must be at least 64.
1531 @findex CHAR_TYPE_SIZE
1532 @item CHAR_TYPE_SIZE
1533 A C expression for the size in bits of the type @code{char} on the
1534 target machine. If you don't define this, the default is
1535 @code{BITS_PER_UNIT}.
1537 @findex BOOL_TYPE_SIZE
1538 @item BOOL_TYPE_SIZE
1539 A C expression for the size in bits of the C++ type @code{bool} and
1540 C99 type @code{_Bool} on the target machine. If you don't define
1541 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1543 @findex FLOAT_TYPE_SIZE
1544 @item FLOAT_TYPE_SIZE
1545 A C expression for the size in bits of the type @code{float} on the
1546 target machine. If you don't define this, the default is one word.
1548 @findex DOUBLE_TYPE_SIZE
1549 @item DOUBLE_TYPE_SIZE
1550 A C expression for the size in bits of the type @code{double} on the
1551 target machine. If you don't define this, the default is two
1554 @findex LONG_DOUBLE_TYPE_SIZE
1555 @item LONG_DOUBLE_TYPE_SIZE
1556 A C expression for the size in bits of the type @code{long double} on
1557 the target machine. If you don't define this, the default is two
1560 @findex MAX_LONG_DOUBLE_TYPE_SIZE
1561 Maximum number for the size in bits of the type @code{long double} on the
1562 target machine. If this is undefined, the default is
1563 @code{LONG_DOUBLE_TYPE_SIZE}. Otherwise, it is the constant value that is
1564 the largest value that @code{LONG_DOUBLE_TYPE_SIZE} can have at run-time.
1565 This is used in @code{cpp}.
1567 @findex TARGET_FLT_EVAL_METHOD
1568 @item TARGET_FLT_EVAL_METHOD
1569 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1570 assuming, if applicable, that the floating-point control word is in its
1571 default state. If you do not define this macro the value of
1572 @code{FLT_EVAL_METHOD} will be zero.
1574 @findex WIDEST_HARDWARE_FP_SIZE
1575 @item WIDEST_HARDWARE_FP_SIZE
1576 A C expression for the size in bits of the widest floating-point format
1577 supported by the hardware. If you define this macro, you must specify a
1578 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1579 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1582 @findex DEFAULT_SIGNED_CHAR
1583 @item DEFAULT_SIGNED_CHAR
1584 An expression whose value is 1 or 0, according to whether the type
1585 @code{char} should be signed or unsigned by default. The user can
1586 always override this default with the options @option{-fsigned-char}
1587 and @option{-funsigned-char}.
1589 @findex DEFAULT_SHORT_ENUMS
1590 @item DEFAULT_SHORT_ENUMS
1591 A C expression to determine whether to give an @code{enum} type
1592 only as many bytes as it takes to represent the range of possible values
1593 of that type. A nonzero value means to do that; a zero value means all
1594 @code{enum} types should be allocated like @code{int}.
1596 If you don't define the macro, the default is 0.
1600 A C expression for a string describing the name of the data type to use
1601 for size values. The typedef name @code{size_t} is defined using the
1602 contents of the string.
1604 The string can contain more than one keyword. If so, separate them with
1605 spaces, and write first any length keyword, then @code{unsigned} if
1606 appropriate, and finally @code{int}. The string must exactly match one
1607 of the data type names defined in the function
1608 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1609 omit @code{int} or change the order---that would cause the compiler to
1612 If you don't define this macro, the default is @code{"long unsigned
1615 @findex PTRDIFF_TYPE
1617 A C expression for a string describing the name of the data type to use
1618 for the result of subtracting two pointers. The typedef name
1619 @code{ptrdiff_t} is defined using the contents of the string. See
1620 @code{SIZE_TYPE} above for more information.
1622 If you don't define this macro, the default is @code{"long int"}.
1626 A C expression for a string describing the name of the data type to use
1627 for wide characters. The typedef name @code{wchar_t} is defined using
1628 the contents of the string. See @code{SIZE_TYPE} above for more
1631 If you don't define this macro, the default is @code{"int"}.
1633 @findex WCHAR_TYPE_SIZE
1634 @item WCHAR_TYPE_SIZE
1635 A C expression for the size in bits of the data type for wide
1636 characters. This is used in @code{cpp}, which cannot make use of
1639 @findex MAX_WCHAR_TYPE_SIZE
1640 @item MAX_WCHAR_TYPE_SIZE
1641 Maximum number for the size in bits of the data type for wide
1642 characters. If this is undefined, the default is
1643 @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1644 largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is
1647 @findex GCOV_TYPE_SIZE
1648 @item GCOV_TYPE_SIZE
1649 A C expression for the size in bits of the type used for gcov counters on the
1650 target machine. If you don't define this, the default is one
1651 @code{LONG_TYPE_SIZE} in case it is greater or equal to 64-bit and
1652 @code{LONG_LONG_TYPE_SIZE} otherwise. You may want to re-define the type to
1653 ensure atomicity for counters in multithreaded programs.
1657 A C expression for a string describing the name of the data type to
1658 use for wide characters passed to @code{printf} and returned from
1659 @code{getwc}. The typedef name @code{wint_t} is defined using the
1660 contents of the string. See @code{SIZE_TYPE} above for more
1663 If you don't define this macro, the default is @code{"unsigned int"}.
1667 A C expression for a string describing the name of the data type that
1668 can represent any value of any standard or extended signed integer type.
1669 The typedef name @code{intmax_t} is defined using the contents of the
1670 string. See @code{SIZE_TYPE} above for more information.
1672 If you don't define this macro, the default is the first of
1673 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1674 much precision as @code{long long int}.
1676 @findex UINTMAX_TYPE
1678 A C expression for a string describing the name of the data type that
1679 can represent any value of any standard or extended unsigned integer
1680 type. The typedef name @code{uintmax_t} is defined using the contents
1681 of the string. See @code{SIZE_TYPE} above for more information.
1683 If you don't define this macro, the default is the first of
1684 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1685 unsigned int"} that has as much precision as @code{long long unsigned
1688 @findex TARGET_PTRMEMFUNC_VBIT_LOCATION
1689 @item TARGET_PTRMEMFUNC_VBIT_LOCATION
1690 The C++ compiler represents a pointer-to-member-function with a struct
1697 ptrdiff_t vtable_index;
1704 The C++ compiler must use one bit to indicate whether the function that
1705 will be called through a pointer-to-member-function is virtual.
1706 Normally, we assume that the low-order bit of a function pointer must
1707 always be zero. Then, by ensuring that the vtable_index is odd, we can
1708 distinguish which variant of the union is in use. But, on some
1709 platforms function pointers can be odd, and so this doesn't work. In
1710 that case, we use the low-order bit of the @code{delta} field, and shift
1711 the remainder of the @code{delta} field to the left.
1713 GCC will automatically make the right selection about where to store
1714 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1715 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1716 set such that functions always start at even addresses, but the lowest
1717 bit of pointers to functions indicate whether the function at that
1718 address is in ARM or Thumb mode. If this is the case of your
1719 architecture, you should define this macro to
1720 @code{ptrmemfunc_vbit_in_delta}.
1722 In general, you should not have to define this macro. On architectures
1723 in which function addresses are always even, according to
1724 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1725 @code{ptrmemfunc_vbit_in_pfn}.
1727 @findex TARGET_VTABLE_USES_DESCRIPTORS
1728 @item TARGET_VTABLE_USES_DESCRIPTORS
1729 Normally, the C++ compiler uses function pointers in vtables. This
1730 macro allows the target to change to use ``function descriptors''
1731 instead. Function descriptors are found on targets for whom a
1732 function pointer is actually a small data structure. Normally the
1733 data structure consists of the actual code address plus a data
1734 pointer to which the function's data is relative.
1736 If vtables are used, the value of this macro should be the number
1737 of words that the function descriptor occupies.
1739 @findex TARGET_VTABLE_ENTRY_ALIGN
1740 @item TARGET_VTABLE_ENTRY_ALIGN
1741 By default, the vtable entries are void pointers, the so the alignment
1742 is the same as pointer alignment. The value of this macro specifies
1743 the alignment of the vtable entry in bits. It should be defined only
1744 when special alignment is necessary. */
1746 @findex TARGET_VTABLE_DATA_ENTRY_DISTANCE
1747 @item TARGET_VTABLE_DATA_ENTRY_DISTANCE
1748 There are a few non-descriptor entries in the vtable at offsets below
1749 zero. If these entries must be padded (say, to preserve the alignment
1750 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1751 of words in each data entry.
1754 @node Escape Sequences
1755 @section Target Character Escape Sequences
1756 @cindex escape sequences
1758 By default, GCC assumes that the C character escape sequences take on
1759 their ASCII values for the target. If this is not correct, you must
1760 explicitly define all of the macros below.
1765 A C constant expression for the integer value for escape sequence
1770 A C constant expression for the integer value of the target escape
1771 character. As an extension, GCC evaluates the escape sequences
1772 @samp{\e} and @samp{\E} to this.
1776 @findex TARGET_NEWLINE
1779 @itemx TARGET_NEWLINE
1780 C constant expressions for the integer values for escape sequences
1781 @samp{\b}, @samp{\t} and @samp{\n}.
1789 C constant expressions for the integer values for escape sequences
1790 @samp{\v}, @samp{\f} and @samp{\r}.
1794 @section Register Usage
1795 @cindex register usage
1797 This section explains how to describe what registers the target machine
1798 has, and how (in general) they can be used.
1800 The description of which registers a specific instruction can use is
1801 done with register classes; see @ref{Register Classes}. For information
1802 on using registers to access a stack frame, see @ref{Frame Registers}.
1803 For passing values in registers, see @ref{Register Arguments}.
1804 For returning values in registers, see @ref{Scalar Return}.
1807 * Register Basics:: Number and kinds of registers.
1808 * Allocation Order:: Order in which registers are allocated.
1809 * Values in Registers:: What kinds of values each reg can hold.
1810 * Leaf Functions:: Renumbering registers for leaf functions.
1811 * Stack Registers:: Handling a register stack such as 80387.
1814 @node Register Basics
1815 @subsection Basic Characteristics of Registers
1817 @c prevent bad page break with this line
1818 Registers have various characteristics.
1821 @findex FIRST_PSEUDO_REGISTER
1822 @item FIRST_PSEUDO_REGISTER
1823 Number of hardware registers known to the compiler. They receive
1824 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1825 pseudo register's number really is assigned the number
1826 @code{FIRST_PSEUDO_REGISTER}.
1828 @item FIXED_REGISTERS
1829 @findex FIXED_REGISTERS
1830 @cindex fixed register
1831 An initializer that says which registers are used for fixed purposes
1832 all throughout the compiled code and are therefore not available for
1833 general allocation. These would include the stack pointer, the frame
1834 pointer (except on machines where that can be used as a general
1835 register when no frame pointer is needed), the program counter on
1836 machines where that is considered one of the addressable registers,
1837 and any other numbered register with a standard use.
1839 This information is expressed as a sequence of numbers, separated by
1840 commas and surrounded by braces. The @var{n}th number is 1 if
1841 register @var{n} is fixed, 0 otherwise.
1843 The table initialized from this macro, and the table initialized by
1844 the following one, may be overridden at run time either automatically,
1845 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1846 the user with the command options @option{-ffixed-@var{reg}},
1847 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1849 @findex CALL_USED_REGISTERS
1850 @item CALL_USED_REGISTERS
1851 @cindex call-used register
1852 @cindex call-clobbered register
1853 @cindex call-saved register
1854 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1855 clobbered (in general) by function calls as well as for fixed
1856 registers. This macro therefore identifies the registers that are not
1857 available for general allocation of values that must live across
1860 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1861 automatically saves it on function entry and restores it on function
1862 exit, if the register is used within the function.
1864 @findex CALL_REALLY_USED_REGISTERS
1865 @item CALL_REALLY_USED_REGISTERS
1866 @cindex call-used register
1867 @cindex call-clobbered register
1868 @cindex call-saved register
1869 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1870 that the entire set of @code{FIXED_REGISTERS} be included.
1871 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1872 This macro is optional. If not specified, it defaults to the value
1873 of @code{CALL_USED_REGISTERS}.
1875 @findex HARD_REGNO_CALL_PART_CLOBBERED
1876 @item HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1877 @cindex call-used register
1878 @cindex call-clobbered register
1879 @cindex call-saved register
1880 A C expression that is nonzero if it is not permissible to store a
1881 value of mode @var{mode} in hard register number @var{regno} across a
1882 call without some part of it being clobbered. For most machines this
1883 macro need not be defined. It is only required for machines that do not
1884 preserve the entire contents of a register across a call.
1886 @findex CONDITIONAL_REGISTER_USAGE
1888 @findex call_used_regs
1889 @item CONDITIONAL_REGISTER_USAGE
1890 Zero or more C statements that may conditionally modify five variables
1891 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1892 @code{reg_names}, and @code{reg_class_contents}, to take into account
1893 any dependence of these register sets on target flags. The first three
1894 of these are of type @code{char []} (interpreted as Boolean vectors).
1895 @code{global_regs} is a @code{const char *[]}, and
1896 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1897 called, @code{fixed_regs}, @code{call_used_regs},
1898 @code{reg_class_contents}, and @code{reg_names} have been initialized
1899 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1900 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1901 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1902 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1903 command options have been applied.
1905 You need not define this macro if it has no work to do.
1907 @cindex disabling certain registers
1908 @cindex controlling register usage
1909 If the usage of an entire class of registers depends on the target
1910 flags, you may indicate this to GCC by using this macro to modify
1911 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1912 registers in the classes which should not be used by GCC@. Also define
1913 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1914 to return @code{NO_REGS} if it
1915 is called with a letter for a class that shouldn't be used.
1917 (However, if this class is not included in @code{GENERAL_REGS} and all
1918 of the insn patterns whose constraints permit this class are
1919 controlled by target switches, then GCC will automatically avoid using
1920 these registers when the target switches are opposed to them.)
1922 @findex NON_SAVING_SETJMP
1923 @item NON_SAVING_SETJMP
1924 If this macro is defined and has a nonzero value, it means that
1925 @code{setjmp} and related functions fail to save the registers, or that
1926 @code{longjmp} fails to restore them. To compensate, the compiler
1927 avoids putting variables in registers in functions that use
1930 @findex INCOMING_REGNO
1931 @item INCOMING_REGNO (@var{out})
1932 Define this macro if the target machine has register windows. This C
1933 expression returns the register number as seen by the called function
1934 corresponding to the register number @var{out} as seen by the calling
1935 function. Return @var{out} if register number @var{out} is not an
1938 @findex OUTGOING_REGNO
1939 @item OUTGOING_REGNO (@var{in})
1940 Define this macro if the target machine has register windows. This C
1941 expression returns the register number as seen by the calling function
1942 corresponding to the register number @var{in} as seen by the called
1943 function. Return @var{in} if register number @var{in} is not an inbound
1947 @item LOCAL_REGNO (@var{regno})
1948 Define this macro if the target machine has register windows. This C
1949 expression returns true if the register is call-saved but is in the
1950 register window. Unlike most call-saved registers, such registers
1951 need not be explicitly restored on function exit or during non-local
1957 If the program counter has a register number, define this as that
1958 register number. Otherwise, do not define it.
1962 @node Allocation Order
1963 @subsection Order of Allocation of Registers
1964 @cindex order of register allocation
1965 @cindex register allocation order
1967 @c prevent bad page break with this line
1968 Registers are allocated in order.
1971 @findex REG_ALLOC_ORDER
1972 @item REG_ALLOC_ORDER
1973 If defined, an initializer for a vector of integers, containing the
1974 numbers of hard registers in the order in which GCC should prefer
1975 to use them (from most preferred to least).
1977 If this macro is not defined, registers are used lowest numbered first
1978 (all else being equal).
1980 One use of this macro is on machines where the highest numbered
1981 registers must always be saved and the save-multiple-registers
1982 instruction supports only sequences of consecutive registers. On such
1983 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1984 the highest numbered allocable register first.
1986 @findex ORDER_REGS_FOR_LOCAL_ALLOC
1987 @item ORDER_REGS_FOR_LOCAL_ALLOC
1988 A C statement (sans semicolon) to choose the order in which to allocate
1989 hard registers for pseudo-registers local to a basic block.
1991 Store the desired register order in the array @code{reg_alloc_order}.
1992 Element 0 should be the register to allocate first; element 1, the next
1993 register; and so on.
1995 The macro body should not assume anything about the contents of
1996 @code{reg_alloc_order} before execution of the macro.
1998 On most machines, it is not necessary to define this macro.
2001 @node Values in Registers
2002 @subsection How Values Fit in Registers
2004 This section discusses the macros that describe which kinds of values
2005 (specifically, which machine modes) each register can hold, and how many
2006 consecutive registers are needed for a given mode.
2009 @findex HARD_REGNO_NREGS
2010 @item HARD_REGNO_NREGS (@var{regno}, @var{mode})
2011 A C expression for the number of consecutive hard registers, starting
2012 at register number @var{regno}, required to hold a value of mode
2015 On a machine where all registers are exactly one word, a suitable
2016 definition of this macro is
2019 #define HARD_REGNO_NREGS(REGNO, MODE) \
2020 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2024 @findex HARD_REGNO_MODE_OK
2025 @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2026 A C expression that is nonzero if it is permissible to store a value
2027 of mode @var{mode} in hard register number @var{regno} (or in several
2028 registers starting with that one). For a machine where all registers
2029 are equivalent, a suitable definition is
2032 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2035 You need not include code to check for the numbers of fixed registers,
2036 because the allocation mechanism considers them to be always occupied.
2038 @cindex register pairs
2039 On some machines, double-precision values must be kept in even/odd
2040 register pairs. You can implement that by defining this macro to reject
2041 odd register numbers for such modes.
2043 The minimum requirement for a mode to be OK in a register is that the
2044 @samp{mov@var{mode}} instruction pattern support moves between the
2045 register and other hard register in the same class and that moving a
2046 value into the register and back out not alter it.
2048 Since the same instruction used to move @code{word_mode} will work for
2049 all narrower integer modes, it is not necessary on any machine for
2050 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2051 you define patterns @samp{movhi}, etc., to take advantage of this. This
2052 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2053 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2056 Many machines have special registers for floating point arithmetic.
2057 Often people assume that floating point machine modes are allowed only
2058 in floating point registers. This is not true. Any registers that
2059 can hold integers can safely @emph{hold} a floating point machine
2060 mode, whether or not floating arithmetic can be done on it in those
2061 registers. Integer move instructions can be used to move the values.
2063 On some machines, though, the converse is true: fixed-point machine
2064 modes may not go in floating registers. This is true if the floating
2065 registers normalize any value stored in them, because storing a
2066 non-floating value there would garble it. In this case,
2067 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2068 floating registers. But if the floating registers do not automatically
2069 normalize, if you can store any bit pattern in one and retrieve it
2070 unchanged without a trap, then any machine mode may go in a floating
2071 register, so you can define this macro to say so.
2073 The primary significance of special floating registers is rather that
2074 they are the registers acceptable in floating point arithmetic
2075 instructions. However, this is of no concern to
2076 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2077 constraints for those instructions.
2079 On some machines, the floating registers are especially slow to access,
2080 so that it is better to store a value in a stack frame than in such a
2081 register if floating point arithmetic is not being done. As long as the
2082 floating registers are not in class @code{GENERAL_REGS}, they will not
2083 be used unless some pattern's constraint asks for one.
2085 @findex MODES_TIEABLE_P
2086 @item MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2087 A C expression that is nonzero if a value of mode
2088 @var{mode1} is accessible in mode @var{mode2} without copying.
2090 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2091 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2092 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2093 should be nonzero. If they differ for any @var{r}, you should define
2094 this macro to return zero unless some other mechanism ensures the
2095 accessibility of the value in a narrower mode.
2097 You should define this macro to return nonzero in as many cases as
2098 possible since doing so will allow GCC to perform better register
2101 @findex AVOID_CCMODE_COPIES
2102 @item AVOID_CCMODE_COPIES
2103 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2104 registers. You should only define this macro if support for copying to/from
2105 @code{CCmode} is incomplete.
2108 @node Leaf Functions
2109 @subsection Handling Leaf Functions
2111 @cindex leaf functions
2112 @cindex functions, leaf
2113 On some machines, a leaf function (i.e., one which makes no calls) can run
2114 more efficiently if it does not make its own register window. Often this
2115 means it is required to receive its arguments in the registers where they
2116 are passed by the caller, instead of the registers where they would
2119 The special treatment for leaf functions generally applies only when
2120 other conditions are met; for example, often they may use only those
2121 registers for its own variables and temporaries. We use the term ``leaf
2122 function'' to mean a function that is suitable for this special
2123 handling, so that functions with no calls are not necessarily ``leaf
2126 GCC assigns register numbers before it knows whether the function is
2127 suitable for leaf function treatment. So it needs to renumber the
2128 registers in order to output a leaf function. The following macros
2132 @findex LEAF_REGISTERS
2133 @item LEAF_REGISTERS
2134 Name of a char vector, indexed by hard register number, which
2135 contains 1 for a register that is allowable in a candidate for leaf
2138 If leaf function treatment involves renumbering the registers, then the
2139 registers marked here should be the ones before renumbering---those that
2140 GCC would ordinarily allocate. The registers which will actually be
2141 used in the assembler code, after renumbering, should not be marked with 1
2144 Define this macro only if the target machine offers a way to optimize
2145 the treatment of leaf functions.
2147 @findex LEAF_REG_REMAP
2148 @item LEAF_REG_REMAP (@var{regno})
2149 A C expression whose value is the register number to which @var{regno}
2150 should be renumbered, when a function is treated as a leaf function.
2152 If @var{regno} is a register number which should not appear in a leaf
2153 function before renumbering, then the expression should yield @minus{}1, which
2154 will cause the compiler to abort.
2156 Define this macro only if the target machine offers a way to optimize the
2157 treatment of leaf functions, and registers need to be renumbered to do
2161 @findex current_function_is_leaf
2162 @findex current_function_uses_only_leaf_regs
2163 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2164 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2165 specially. They can test the C variable @code{current_function_is_leaf}
2166 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2167 set prior to local register allocation and is valid for the remaining
2168 compiler passes. They can also test the C variable
2169 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2170 functions which only use leaf registers.
2171 @code{current_function_uses_only_leaf_regs} is valid after reload and is
2172 only useful if @code{LEAF_REGISTERS} is defined.
2173 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2174 @c of the next paragraph?! --mew 2feb93
2176 @node Stack Registers
2177 @subsection Registers That Form a Stack
2179 There are special features to handle computers where some of the
2180 ``registers'' form a stack, as in the 80387 coprocessor for the 80386.
2181 Stack registers are normally written by pushing onto the stack, and are
2182 numbered relative to the top of the stack.
2184 Currently, GCC can only handle one group of stack-like registers, and
2185 they must be consecutively numbered.
2190 Define this if the machine has any stack-like registers.
2192 @findex FIRST_STACK_REG
2193 @item FIRST_STACK_REG
2194 The number of the first stack-like register. This one is the top
2197 @findex LAST_STACK_REG
2198 @item LAST_STACK_REG
2199 The number of the last stack-like register. This one is the bottom of
2203 @node Register Classes
2204 @section Register Classes
2205 @cindex register class definitions
2206 @cindex class definitions, register
2208 On many machines, the numbered registers are not all equivalent.
2209 For example, certain registers may not be allowed for indexed addressing;
2210 certain registers may not be allowed in some instructions. These machine
2211 restrictions are described to the compiler using @dfn{register classes}.
2213 You define a number of register classes, giving each one a name and saying
2214 which of the registers belong to it. Then you can specify register classes
2215 that are allowed as operands to particular instruction patterns.
2219 In general, each register will belong to several classes. In fact, one
2220 class must be named @code{ALL_REGS} and contain all the registers. Another
2221 class must be named @code{NO_REGS} and contain no registers. Often the
2222 union of two classes will be another class; however, this is not required.
2224 @findex GENERAL_REGS
2225 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2226 terribly special about the name, but the operand constraint letters
2227 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2228 the same as @code{ALL_REGS}, just define it as a macro which expands
2231 Order the classes so that if class @var{x} is contained in class @var{y}
2232 then @var{x} has a lower class number than @var{y}.
2234 The way classes other than @code{GENERAL_REGS} are specified in operand
2235 constraints is through machine-dependent operand constraint letters.
2236 You can define such letters to correspond to various classes, then use
2237 them in operand constraints.
2239 You should define a class for the union of two classes whenever some
2240 instruction allows both classes. For example, if an instruction allows
2241 either a floating point (coprocessor) register or a general register for a
2242 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2243 which includes both of them. Otherwise you will get suboptimal code.
2245 You must also specify certain redundant information about the register
2246 classes: for each class, which classes contain it and which ones are
2247 contained in it; for each pair of classes, the largest class contained
2250 When a value occupying several consecutive registers is expected in a
2251 certain class, all the registers used must belong to that class.
2252 Therefore, register classes cannot be used to enforce a requirement for
2253 a register pair to start with an even-numbered register. The way to
2254 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2256 Register classes used for input-operands of bitwise-and or shift
2257 instructions have a special requirement: each such class must have, for
2258 each fixed-point machine mode, a subclass whose registers can transfer that
2259 mode to or from memory. For example, on some machines, the operations for
2260 single-byte values (@code{QImode}) are limited to certain registers. When
2261 this is so, each register class that is used in a bitwise-and or shift
2262 instruction must have a subclass consisting of registers from which
2263 single-byte values can be loaded or stored. This is so that
2264 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2267 @findex enum reg_class
2268 @item enum reg_class
2269 An enumeral type that must be defined with all the register class names
2270 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
2271 must be the last register class, followed by one more enumeral value,
2272 @code{LIM_REG_CLASSES}, which is not a register class but rather
2273 tells how many classes there are.
2275 Each register class has a number, which is the value of casting
2276 the class name to type @code{int}. The number serves as an index
2277 in many of the tables described below.
2279 @findex N_REG_CLASSES
2281 The number of distinct register classes, defined as follows:
2284 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2287 @findex REG_CLASS_NAMES
2288 @item REG_CLASS_NAMES
2289 An initializer containing the names of the register classes as C string
2290 constants. These names are used in writing some of the debugging dumps.
2292 @findex REG_CLASS_CONTENTS
2293 @item REG_CLASS_CONTENTS
2294 An initializer containing the contents of the register classes, as integers
2295 which are bit masks. The @var{n}th integer specifies the contents of class
2296 @var{n}. The way the integer @var{mask} is interpreted is that
2297 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2299 When the machine has more than 32 registers, an integer does not suffice.
2300 Then the integers are replaced by sub-initializers, braced groupings containing
2301 several integers. Each sub-initializer must be suitable as an initializer
2302 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2303 In this situation, the first integer in each sub-initializer corresponds to
2304 registers 0 through 31, the second integer to registers 32 through 63, and
2307 @findex REGNO_REG_CLASS
2308 @item REGNO_REG_CLASS (@var{regno})
2309 A C expression whose value is a register class containing hard register
2310 @var{regno}. In general there is more than one such class; choose a class
2311 which is @dfn{minimal}, meaning that no smaller class also contains the
2314 @findex BASE_REG_CLASS
2315 @item BASE_REG_CLASS
2316 A macro whose definition is the name of the class to which a valid
2317 base register must belong. A base register is one used in an address
2318 which is the register value plus a displacement.
2320 @findex MODE_BASE_REG_CLASS
2321 @item MODE_BASE_REG_CLASS (@var{mode})
2322 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2323 the selection of a base register in a mode dependent manner. If
2324 @var{mode} is VOIDmode then it should return the same value as
2325 @code{BASE_REG_CLASS}.
2327 @findex INDEX_REG_CLASS
2328 @item INDEX_REG_CLASS
2329 A macro whose definition is the name of the class to which a valid
2330 index register must belong. An index register is one used in an
2331 address where its value is either multiplied by a scale factor or
2332 added to another register (as well as added to a displacement).
2334 @findex CONSTRAINT_LEN
2335 @item CONSTRAINT_LEN (@var{char}, @var{str})
2336 For the constraint at the start of @var{str}, which starts with the letter
2337 @var{c}, return the length. This allows you to have register class /
2338 constant / extra constraints that are longer than a single letter;
2339 you don't need to define this macro if you can do with single-letter
2340 constraints only. The definition of this macro should use
2341 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2342 to handle specially.
2343 There are some sanity checks in genoutput.c that check the constraint lengths
2344 for the md file, so you can also use this macro to help you while you are
2345 transitioning from a byzantine single-letter-constraint scheme: when you
2346 return a negative length for a constraint you want to re-use, genoutput
2347 will complain about every instance where it is used in the md file.
2349 @findex REG_CLASS_FROM_LETTER
2350 @item REG_CLASS_FROM_LETTER (@var{char})
2351 A C expression which defines the machine-dependent operand constraint
2352 letters for register classes. If @var{char} is such a letter, the
2353 value should be the register class corresponding to it. Otherwise,
2354 the value should be @code{NO_REGS}. The register letter @samp{r},
2355 corresponding to class @code{GENERAL_REGS}, will not be passed
2356 to this macro; you do not need to handle it.
2358 @findex REG_CLASS_FROM_CONSTRAINT
2359 @item REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2360 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2361 passed in @var{str}, so that you can use suffixes to distinguish between
2364 @findex REGNO_OK_FOR_BASE_P
2365 @item REGNO_OK_FOR_BASE_P (@var{num})
2366 A C expression which is nonzero if register number @var{num} is
2367 suitable for use as a base register in operand addresses. It may be
2368 either a suitable hard register or a pseudo register that has been
2369 allocated such a hard register.
2371 @findex REGNO_MODE_OK_FOR_BASE_P
2372 @item REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2373 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2374 that expression may examine the mode of the memory reference in
2375 @var{mode}. You should define this macro if the mode of the memory
2376 reference affects whether a register may be used as a base register. If
2377 you define this macro, the compiler will use it instead of
2378 @code{REGNO_OK_FOR_BASE_P}.
2380 @findex REGNO_OK_FOR_INDEX_P
2381 @item REGNO_OK_FOR_INDEX_P (@var{num})
2382 A C expression which is nonzero if register number @var{num} is
2383 suitable for use as an index register in operand addresses. It may be
2384 either a suitable hard register or a pseudo register that has been
2385 allocated such a hard register.
2387 The difference between an index register and a base register is that
2388 the index register may be scaled. If an address involves the sum of
2389 two registers, neither one of them scaled, then either one may be
2390 labeled the ``base'' and the other the ``index''; but whichever
2391 labeling is used must fit the machine's constraints of which registers
2392 may serve in each capacity. The compiler will try both labelings,
2393 looking for one that is valid, and will reload one or both registers
2394 only if neither labeling works.
2396 @findex PREFERRED_RELOAD_CLASS
2397 @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2398 A C expression that places additional restrictions on the register class
2399 to use when it is necessary to copy value @var{x} into a register in class
2400 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2401 another, smaller class. On many machines, the following definition is
2405 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2408 Sometimes returning a more restrictive class makes better code. For
2409 example, on the 68000, when @var{x} is an integer constant that is in range
2410 for a @samp{moveq} instruction, the value of this macro is always
2411 @code{DATA_REGS} as long as @var{class} includes the data registers.
2412 Requiring a data register guarantees that a @samp{moveq} will be used.
2414 If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
2415 you can force @var{x} into a memory constant. This is useful on
2416 certain machines where immediate floating values cannot be loaded into
2417 certain kinds of registers.
2419 @findex PREFERRED_OUTPUT_RELOAD_CLASS
2420 @item PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2421 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2422 input reloads. If you don't define this macro, the default is to use
2423 @var{class}, unchanged.
2425 @findex LIMIT_RELOAD_CLASS
2426 @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2427 A C expression that places additional restrictions on the register class
2428 to use when it is necessary to be able to hold a value of mode
2429 @var{mode} in a reload register for which class @var{class} would
2432 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2433 there are certain modes that simply can't go in certain reload classes.
2435 The value is a register class; perhaps @var{class}, or perhaps another,
2438 Don't define this macro unless the target machine has limitations which
2439 require the macro to do something nontrivial.
2441 @findex SECONDARY_RELOAD_CLASS
2442 @findex SECONDARY_INPUT_RELOAD_CLASS
2443 @findex SECONDARY_OUTPUT_RELOAD_CLASS
2444 @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2445 @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2446 @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2447 Many machines have some registers that cannot be copied directly to or
2448 from memory or even from other types of registers. An example is the
2449 @samp{MQ} register, which on most machines, can only be copied to or
2450 from general registers, but not memory. Some machines allow copying all
2451 registers to and from memory, but require a scratch register for stores
2452 to some memory locations (e.g., those with symbolic address on the RT,
2453 and those with certain symbolic address on the SPARC when compiling
2454 PIC)@. In some cases, both an intermediate and a scratch register are
2457 You should define these macros to indicate to the reload phase that it may
2458 need to allocate at least one register for a reload in addition to the
2459 register to contain the data. Specifically, if copying @var{x} to a
2460 register @var{class} in @var{mode} requires an intermediate register,
2461 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2462 largest register class all of whose registers can be used as
2463 intermediate registers or scratch registers.
2465 If copying a register @var{class} in @var{mode} to @var{x} requires an
2466 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2467 should be defined to return the largest register class required. If the
2468 requirements for input and output reloads are the same, the macro
2469 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2472 The values returned by these macros are often @code{GENERAL_REGS}.
2473 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2474 can be directly copied to or from a register of @var{class} in
2475 @var{mode} without requiring a scratch register. Do not define this
2476 macro if it would always return @code{NO_REGS}.
2478 If a scratch register is required (either with or without an
2479 intermediate register), you should define patterns for
2480 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2481 (@pxref{Standard Names}. These patterns, which will normally be
2482 implemented with a @code{define_expand}, should be similar to the
2483 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2486 Define constraints for the reload register and scratch register that
2487 contain a single register class. If the original reload register (whose
2488 class is @var{class}) can meet the constraint given in the pattern, the
2489 value returned by these macros is used for the class of the scratch
2490 register. Otherwise, two additional reload registers are required.
2491 Their classes are obtained from the constraints in the insn pattern.
2493 @var{x} might be a pseudo-register or a @code{subreg} of a
2494 pseudo-register, which could either be in a hard register or in memory.
2495 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2496 in memory and the hard register number if it is in a register.
2498 These macros should not be used in the case where a particular class of
2499 registers can only be copied to memory and not to another class of
2500 registers. In that case, secondary reload registers are not needed and
2501 would not be helpful. Instead, a stack location must be used to perform
2502 the copy and the @code{mov@var{m}} pattern should use memory as an
2503 intermediate storage. This case often occurs between floating-point and
2506 @findex SECONDARY_MEMORY_NEEDED
2507 @item SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2508 Certain machines have the property that some registers cannot be copied
2509 to some other registers without using memory. Define this macro on
2510 those machines to be a C expression that is nonzero if objects of mode
2511 @var{m} in registers of @var{class1} can only be copied to registers of
2512 class @var{class2} by storing a register of @var{class1} into memory
2513 and loading that memory location into a register of @var{class2}.
2515 Do not define this macro if its value would always be zero.
2517 @findex SECONDARY_MEMORY_NEEDED_RTX
2518 @item SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2519 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2520 allocates a stack slot for a memory location needed for register copies.
2521 If this macro is defined, the compiler instead uses the memory location
2522 defined by this macro.
2524 Do not define this macro if you do not define
2525 @code{SECONDARY_MEMORY_NEEDED}.
2527 @findex SECONDARY_MEMORY_NEEDED_MODE
2528 @item SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2529 When the compiler needs a secondary memory location to copy between two
2530 registers of mode @var{mode}, it normally allocates sufficient memory to
2531 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2532 load operations in a mode that many bits wide and whose class is the
2533 same as that of @var{mode}.
2535 This is right thing to do on most machines because it ensures that all
2536 bits of the register are copied and prevents accesses to the registers
2537 in a narrower mode, which some machines prohibit for floating-point
2540 However, this default behavior is not correct on some machines, such as
2541 the DEC Alpha, that store short integers in floating-point registers
2542 differently than in integer registers. On those machines, the default
2543 widening will not work correctly and you must define this macro to
2544 suppress that widening in some cases. See the file @file{alpha.h} for
2547 Do not define this macro if you do not define
2548 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2549 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2551 @findex SMALL_REGISTER_CLASSES
2552 @item SMALL_REGISTER_CLASSES
2553 On some machines, it is risky to let hard registers live across arbitrary
2554 insns. Typically, these machines have instructions that require values
2555 to be in specific registers (like an accumulator), and reload will fail
2556 if the required hard register is used for another purpose across such an
2559 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2560 value on these machines. When this macro has a nonzero value, the
2561 compiler will try to minimize the lifetime of hard registers.
2563 It is always safe to define this macro with a nonzero value, but if you
2564 unnecessarily define it, you will reduce the amount of optimizations
2565 that can be performed in some cases. If you do not define this macro
2566 with a nonzero value when it is required, the compiler will run out of
2567 spill registers and print a fatal error message. For most machines, you
2568 should not define this macro at all.
2570 @findex CLASS_LIKELY_SPILLED_P
2571 @item CLASS_LIKELY_SPILLED_P (@var{class})
2572 A C expression whose value is nonzero if pseudos that have been assigned
2573 to registers of class @var{class} would likely be spilled because
2574 registers of @var{class} are needed for spill registers.
2576 The default value of this macro returns 1 if @var{class} has exactly one
2577 register and zero otherwise. On most machines, this default should be
2578 used. Only define this macro to some other expression if pseudos
2579 allocated by @file{local-alloc.c} end up in memory because their hard
2580 registers were needed for spill registers. If this macro returns nonzero
2581 for those classes, those pseudos will only be allocated by
2582 @file{global.c}, which knows how to reallocate the pseudo to another
2583 register. If there would not be another register available for
2584 reallocation, you should not change the definition of this macro since
2585 the only effect of such a definition would be to slow down register
2588 @findex CLASS_MAX_NREGS
2589 @item CLASS_MAX_NREGS (@var{class}, @var{mode})
2590 A C expression for the maximum number of consecutive registers
2591 of class @var{class} needed to hold a value of mode @var{mode}.
2593 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2594 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2595 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2596 @var{mode})} for all @var{regno} values in the class @var{class}.
2598 This macro helps control the handling of multiple-word values
2601 @item CANNOT_CHANGE_MODE_CLASS(@var{from}, @var{to})
2602 If defined, a C expression that returns a register class for which
2603 a change from mode @var{from} to mode @var{to} is invalid, otherwise the
2604 macro returns @code{NO_REGS}.
2606 For the example, loading 32-bit integer or floating-point objects into
2607 floating-point registers on the Alpha extends them to 64 bits.
2608 Therefore loading a 64-bit object and then storing it as a 32-bit object
2609 does not store the low-order 32 bits, as would be the case for a normal
2610 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2614 #define CANNOT_CHANGE_MODE_CLASS \
2615 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) ? FLOAT_REGS : NO_REGS)
2619 Three other special macros describe which operands fit which constraint
2623 @findex CONST_OK_FOR_LETTER_P
2624 @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2625 A C expression that defines the machine-dependent operand constraint
2626 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2627 particular ranges of integer values. If @var{c} is one of those
2628 letters, the expression should check that @var{value}, an integer, is in
2629 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2630 not one of those letters, the value should be 0 regardless of
2633 @findex CONST_OK_FOR_CONSTRAINT_P
2634 @item CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2635 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2636 string passed in @var{str}, so that you can use suffixes to distinguish
2637 between different variants.
2639 @findex CONST_DOUBLE_OK_FOR_LETTER_P
2640 @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2641 A C expression that defines the machine-dependent operand constraint
2642 letters that specify particular ranges of @code{const_double} values
2643 (@samp{G} or @samp{H}).
2645 If @var{c} is one of those letters, the expression should check that
2646 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2647 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2648 letters, the value should be 0 regardless of @var{value}.
2650 @code{const_double} is used for all floating-point constants and for
2651 @code{DImode} fixed-point constants. A given letter can accept either
2652 or both kinds of values. It can use @code{GET_MODE} to distinguish
2653 between these kinds.
2655 @findex CONST_DOUBLE_OK_FOR_CONSTRAINT_P
2656 @item CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2657 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2658 string passed in @var{str}, so that you can use suffixes to distinguish
2659 between different variants.
2661 @findex EXTRA_CONSTRAINT
2662 @item EXTRA_CONSTRAINT (@var{value}, @var{c})
2663 A C expression that defines the optional machine-dependent constraint
2664 letters that can be used to segregate specific types of operands, usually
2665 memory references, for the target machine. Any letter that is not
2666 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2667 @code{REG_CLASS_FROM_CONSTRAINT}
2668 may be used. Normally this macro will not be defined.
2670 If it is required for a particular target machine, it should return 1
2671 if @var{value} corresponds to the operand type represented by the
2672 constraint letter @var{c}. If @var{c} is not defined as an extra
2673 constraint, the value returned should be 0 regardless of @var{value}.
2675 For example, on the ROMP, load instructions cannot have their output
2676 in r0 if the memory reference contains a symbolic address. Constraint
2677 letter @samp{Q} is defined as representing a memory address that does
2678 @emph{not} contain a symbolic address. An alternative is specified with
2679 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2680 alternative specifies @samp{m} on the input and a register class that
2681 does not include r0 on the output.
2683 @findex EXTRA_CONSTRAINT_STR
2684 @item EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2685 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2686 in @var{str}, so that you can use suffixes to distinguish between different
2689 @findex EXTRA_MEMORY_CONSTRAINT
2690 @item EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2691 A C expression that defines the optional machine-dependent constraint
2692 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2693 be treated like memory constraints by the reload pass.
2695 It should return 1 if the operand type represented by the constraint
2696 at the start of @var{str}, the first letter of which is the letter @var{c},
2697 comprises a subset of all memory references including
2698 all those whose address is simply a base register. This allows the reload
2699 pass to reload an operand, if it does not directly correspond to the operand
2700 type of @var{c}, by copying its address into a base register.
2702 For example, on the S/390, some instructions do not accept arbitrary
2703 memory references, but only those that do not make use of an index
2704 register. The constraint letter @samp{Q} is defined via
2705 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2706 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2707 a @samp{Q} constraint can handle any memory operand, because the
2708 reload pass knows it can be reloaded by copying the memory address
2709 into a base register if required. This is analogous to the way
2710 a @samp{o} constraint can handle any memory operand.
2712 @findex EXTRA_ADDRESS_CONSTRAINT
2713 @item EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2714 A C expression that defines the optional machine-dependent constraint
2715 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2716 @code{EXTRA_CONSTRAINT_STR}, that should
2717 be treated like address constraints by the reload pass.
2719 It should return 1 if the operand type represented by the constraint
2720 at the start of @var{str}, which starts with the letter @var{c}, comprises
2721 a subset of all memory addresses including
2722 all those that consist of just a base register. This allows the reload
2723 pass to reload an operand, if it does not directly correspond to the operand
2724 type of @var{str}, by copying it into a base register.
2726 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2727 be used with the @code{address_operand} predicate. It is treated
2728 analogously to the @samp{p} constraint.
2731 @node Stack and Calling
2732 @section Stack Layout and Calling Conventions
2733 @cindex calling conventions
2735 @c prevent bad page break with this line
2736 This describes the stack layout and calling conventions.
2740 * Exception Handling::
2745 * Register Arguments::
2747 * Aggregate Return::
2755 @subsection Basic Stack Layout
2756 @cindex stack frame layout
2757 @cindex frame layout
2759 @c prevent bad page break with this line
2760 Here is the basic stack layout.
2763 @findex STACK_GROWS_DOWNWARD
2764 @item STACK_GROWS_DOWNWARD
2765 Define this macro if pushing a word onto the stack moves the stack
2766 pointer to a smaller address.
2768 When we say, ``define this macro if @dots{},'' it means that the
2769 compiler checks this macro only with @code{#ifdef} so the precise
2770 definition used does not matter.
2772 @findex STACK_PUSH_CODE
2773 @item STACK_PUSH_CODE
2775 This macro defines the operation used when something is pushed
2776 on the stack. In RTL, a push operation will be
2777 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2779 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2780 and @code{POST_INC}. Which of these is correct depends on
2781 the stack direction and on whether the stack pointer points
2782 to the last item on the stack or whether it points to the
2783 space for the next item on the stack.
2785 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2786 defined, which is almost always right, and @code{PRE_INC} otherwise,
2787 which is often wrong.
2789 @findex FRAME_GROWS_DOWNWARD
2790 @item FRAME_GROWS_DOWNWARD
2791 Define this macro if the addresses of local variable slots are at negative
2792 offsets from the frame pointer.
2794 @findex ARGS_GROW_DOWNWARD
2795 @item ARGS_GROW_DOWNWARD
2796 Define this macro if successive arguments to a function occupy decreasing
2797 addresses on the stack.
2799 @findex STARTING_FRAME_OFFSET
2800 @item STARTING_FRAME_OFFSET
2801 Offset from the frame pointer to the first local variable slot to be allocated.
2803 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2804 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2805 Otherwise, it is found by adding the length of the first slot to the
2806 value @code{STARTING_FRAME_OFFSET}.
2807 @c i'm not sure if the above is still correct.. had to change it to get
2808 @c rid of an overfull. --mew 2feb93
2810 @findex STACK_POINTER_OFFSET
2811 @item STACK_POINTER_OFFSET
2812 Offset from the stack pointer register to the first location at which
2813 outgoing arguments are placed. If not specified, the default value of
2814 zero is used. This is the proper value for most machines.
2816 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2817 the first location at which outgoing arguments are placed.
2819 @findex FIRST_PARM_OFFSET
2820 @item FIRST_PARM_OFFSET (@var{fundecl})
2821 Offset from the argument pointer register to the first argument's
2822 address. On some machines it may depend on the data type of the
2825 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2826 the first argument's address.
2828 @findex STACK_DYNAMIC_OFFSET
2829 @item STACK_DYNAMIC_OFFSET (@var{fundecl})
2830 Offset from the stack pointer register to an item dynamically allocated
2831 on the stack, e.g., by @code{alloca}.
2833 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2834 length of the outgoing arguments. The default is correct for most
2835 machines. See @file{function.c} for details.
2837 @findex DYNAMIC_CHAIN_ADDRESS
2838 @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2839 A C expression whose value is RTL representing the address in a stack
2840 frame where the pointer to the caller's frame is stored. Assume that
2841 @var{frameaddr} is an RTL expression for the address of the stack frame
2844 If you don't define this macro, the default is to return the value
2845 of @var{frameaddr}---that is, the stack frame address is also the
2846 address of the stack word that points to the previous frame.
2848 @findex SETUP_FRAME_ADDRESSES
2849 @item SETUP_FRAME_ADDRESSES
2850 If defined, a C expression that produces the machine-specific code to
2851 setup the stack so that arbitrary frames can be accessed. For example,
2852 on the SPARC, we must flush all of the register windows to the stack
2853 before we can access arbitrary stack frames. You will seldom need to
2856 @findex BUILTIN_SETJMP_FRAME_VALUE
2857 @item BUILTIN_SETJMP_FRAME_VALUE
2858 If defined, a C expression that contains an rtx that is used to store
2859 the address of the current frame into the built in @code{setjmp} buffer.
2860 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2861 machines. One reason you may need to define this macro is if
2862 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2864 @findex RETURN_ADDR_RTX
2865 @item RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2866 A C expression whose value is RTL representing the value of the return
2867 address for the frame @var{count} steps up from the current frame, after
2868 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2869 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2870 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2872 The value of the expression must always be the correct address when
2873 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2874 determine the return address of other frames.
2876 @findex RETURN_ADDR_IN_PREVIOUS_FRAME
2877 @item RETURN_ADDR_IN_PREVIOUS_FRAME
2878 Define this if the return address of a particular stack frame is accessed
2879 from the frame pointer of the previous stack frame.
2881 @findex INCOMING_RETURN_ADDR_RTX
2882 @item INCOMING_RETURN_ADDR_RTX
2883 A C expression whose value is RTL representing the location of the
2884 incoming return address at the beginning of any function, before the
2885 prologue. This RTL is either a @code{REG}, indicating that the return
2886 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2889 You only need to define this macro if you want to support call frame
2890 debugging information like that provided by DWARF 2.
2892 If this RTL is a @code{REG}, you should also define
2893 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2895 @findex INCOMING_FRAME_SP_OFFSET
2896 @item INCOMING_FRAME_SP_OFFSET
2897 A C expression whose value is an integer giving the offset, in bytes,
2898 from the value of the stack pointer register to the top of the stack
2899 frame at the beginning of any function, before the prologue. The top of
2900 the frame is defined to be the value of the stack pointer in the
2901 previous frame, just before the call instruction.
2903 You only need to define this macro if you want to support call frame
2904 debugging information like that provided by DWARF 2.
2906 @findex ARG_POINTER_CFA_OFFSET
2907 @item ARG_POINTER_CFA_OFFSET (@var{fundecl})
2908 A C expression whose value is an integer giving the offset, in bytes,
2909 from the argument pointer to the canonical frame address (cfa). The
2910 final value should coincide with that calculated by
2911 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2912 during virtual register instantiation.
2914 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
2915 which is correct for most machines; in general, the arguments are found
2916 immediately before the stack frame. Note that this is not the case on
2917 some targets that save registers into the caller's frame, such as SPARC
2918 and rs6000, and so such targets need to define this macro.
2920 You only need to define this macro if the default is incorrect, and you
2921 want to support call frame debugging information like that provided by
2926 Define this macro if the stack size for the target is very small. This
2927 has the effect of disabling gcc's built-in @samp{alloca}, though
2928 @samp{__builtin_alloca} is not affected.
2931 @node Exception Handling
2932 @subsection Exception Handling Support
2933 @cindex exception handling
2936 @findex EH_RETURN_DATA_REGNO
2937 @item EH_RETURN_DATA_REGNO (@var{N})
2938 A C expression whose value is the @var{N}th register number used for
2939 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2940 @var{N} registers are usable.
2942 The exception handling library routines communicate with the exception
2943 handlers via a set of agreed upon registers. Ideally these registers
2944 should be call-clobbered; it is possible to use call-saved registers,
2945 but may negatively impact code size. The target must support at least
2946 2 data registers, but should define 4 if there are enough free registers.
2948 You must define this macro if you want to support call frame exception
2949 handling like that provided by DWARF 2.
2951 @findex EH_RETURN_STACKADJ_RTX
2952 @item EH_RETURN_STACKADJ_RTX
2953 A C expression whose value is RTL representing a location in which
2954 to store a stack adjustment to be applied before function return.
2955 This is used to unwind the stack to an exception handler's call frame.
2956 It will be assigned zero on code paths that return normally.
2958 Typically this is a call-clobbered hard register that is otherwise
2959 untouched by the epilogue, but could also be a stack slot.
2961 You must define this macro if you want to support call frame exception
2962 handling like that provided by DWARF 2.
2964 @findex EH_RETURN_HANDLER_RTX
2965 @item EH_RETURN_HANDLER_RTX
2966 A C expression whose value is RTL representing a location in which
2967 to store the address of an exception handler to which we should
2968 return. It will not be assigned on code paths that return normally.
2970 Typically this is the location in the call frame at which the normal
2971 return address is stored. For targets that return by popping an
2972 address off the stack, this might be a memory address just below
2973 the @emph{target} call frame rather than inside the current call
2974 frame. @code{EH_RETURN_STACKADJ_RTX} will have already been assigned,
2975 so it may be used to calculate the location of the target call frame.
2977 Some targets have more complex requirements than storing to an
2978 address calculable during initial code generation. In that case
2979 the @code{eh_return} instruction pattern should be used instead.
2981 If you want to support call frame exception handling, you must
2982 define either this macro or the @code{eh_return} instruction pattern.
2984 @findex ASM_PREFERRED_EH_DATA_FORMAT
2985 @item ASM_PREFERRED_EH_DATA_FORMAT(@var{code}, @var{global})
2986 This macro chooses the encoding of pointers embedded in the exception
2987 handling sections. If at all possible, this should be defined such
2988 that the exception handling section will not require dynamic relocations,
2989 and so may be read-only.
2991 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
2992 @var{global} is true if the symbol may be affected by dynamic relocations.
2993 The macro should return a combination of the @code{DW_EH_PE_*} defines
2994 as found in @file{dwarf2.h}.
2996 If this macro is not defined, pointers will not be encoded but
2997 represented directly.
2999 @findex ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX
3000 @item ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX(@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3001 This macro allows the target to emit whatever special magic is required
3002 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3003 Generic code takes care of pc-relative and indirect encodings; this must
3004 be defined if the target uses text-relative or data-relative encodings.
3006 This is a C statement that branches to @var{done} if the format was
3007 handled. @var{encoding} is the format chosen, @var{size} is the number
3008 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3011 @findex MD_FALLBACK_FRAME_STATE_FOR
3012 @item MD_FALLBACK_FRAME_STATE_FOR(@var{context}, @var{fs}, @var{success})
3013 This macro allows the target to add cpu and operating system specific
3014 code to the call-frame unwinder for use when there is no unwind data
3015 available. The most common reason to implement this macro is to unwind
3016 through signal frames.
3018 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
3019 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3020 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3021 for the address of the code being executed and @code{context->cfa} for
3022 the stack pointer value. If the frame can be decoded, the register save
3023 addresses should be updated in @var{fs} and the macro should branch to
3024 @var{success}. If the frame cannot be decoded, the macro should do
3028 @node Stack Checking
3029 @subsection Specifying How Stack Checking is Done
3031 GCC will check that stack references are within the boundaries of
3032 the stack, if the @option{-fstack-check} is specified, in one of three ways:
3036 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3037 will assume that you have arranged for stack checking to be done at
3038 appropriate places in the configuration files, e.g., in
3039 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
3043 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
3044 called @code{check_stack} in your @file{md} file, GCC will call that
3045 pattern with one argument which is the address to compare the stack
3046 value against. You must arrange for this pattern to report an error if
3047 the stack pointer is out of range.
3050 If neither of the above are true, GCC will generate code to periodically
3051 ``probe'' the stack pointer using the values of the macros defined below.
3054 Normally, you will use the default values of these macros, so GCC
3055 will use the third approach.
3058 @findex STACK_CHECK_BUILTIN
3059 @item STACK_CHECK_BUILTIN
3060 A nonzero value if stack checking is done by the configuration files in a
3061 machine-dependent manner. You should define this macro if stack checking
3062 is require by the ABI of your machine or if you would like to have to stack
3063 checking in some more efficient way than GCC's portable approach.
3064 The default value of this macro is zero.
3066 @findex STACK_CHECK_PROBE_INTERVAL
3067 @item STACK_CHECK_PROBE_INTERVAL
3068 An integer representing the interval at which GCC must generate stack
3069 probe instructions. You will normally define this macro to be no larger
3070 than the size of the ``guard pages'' at the end of a stack area. The
3071 default value of 4096 is suitable for most systems.
3073 @findex STACK_CHECK_PROBE_LOAD
3074 @item STACK_CHECK_PROBE_LOAD
3075 A integer which is nonzero if GCC should perform the stack probe
3076 as a load instruction and zero if GCC should use a store instruction.
3077 The default is zero, which is the most efficient choice on most systems.
3079 @findex STACK_CHECK_PROTECT
3080 @item STACK_CHECK_PROTECT
3081 The number of bytes of stack needed to recover from a stack overflow,
3082 for languages where such a recovery is supported. The default value of
3083 75 words should be adequate for most machines.
3085 @findex STACK_CHECK_MAX_FRAME_SIZE
3086 @item STACK_CHECK_MAX_FRAME_SIZE
3087 The maximum size of a stack frame, in bytes. GCC will generate probe
3088 instructions in non-leaf functions to ensure at least this many bytes of
3089 stack are available. If a stack frame is larger than this size, stack
3090 checking will not be reliable and GCC will issue a warning. The
3091 default is chosen so that GCC only generates one instruction on most
3092 systems. You should normally not change the default value of this macro.
3094 @findex STACK_CHECK_FIXED_FRAME_SIZE
3095 @item STACK_CHECK_FIXED_FRAME_SIZE
3096 GCC uses this value to generate the above warning message. It
3097 represents the amount of fixed frame used by a function, not including
3098 space for any callee-saved registers, temporaries and user variables.
3099 You need only specify an upper bound for this amount and will normally
3100 use the default of four words.
3102 @findex STACK_CHECK_MAX_VAR_SIZE
3103 @item STACK_CHECK_MAX_VAR_SIZE
3104 The maximum size, in bytes, of an object that GCC will place in the
3105 fixed area of the stack frame when the user specifies
3106 @option{-fstack-check}.
3107 GCC computed the default from the values of the above macros and you will
3108 normally not need to override that default.
3112 @node Frame Registers
3113 @subsection Registers That Address the Stack Frame
3115 @c prevent bad page break with this line
3116 This discusses registers that address the stack frame.
3119 @findex STACK_POINTER_REGNUM
3120 @item STACK_POINTER_REGNUM
3121 The register number of the stack pointer register, which must also be a
3122 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3123 the hardware determines which register this is.
3125 @findex FRAME_POINTER_REGNUM
3126 @item FRAME_POINTER_REGNUM
3127 The register number of the frame pointer register, which is used to
3128 access automatic variables in the stack frame. On some machines, the
3129 hardware determines which register this is. On other machines, you can
3130 choose any register you wish for this purpose.
3132 @findex HARD_FRAME_POINTER_REGNUM
3133 @item HARD_FRAME_POINTER_REGNUM
3134 On some machines the offset between the frame pointer and starting
3135 offset of the automatic variables is not known until after register
3136 allocation has been done (for example, because the saved registers are
3137 between these two locations). On those machines, define
3138 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3139 be used internally until the offset is known, and define
3140 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3141 used for the frame pointer.
3143 You should define this macro only in the very rare circumstances when it
3144 is not possible to calculate the offset between the frame pointer and
3145 the automatic variables until after register allocation has been
3146 completed. When this macro is defined, you must also indicate in your
3147 definition of @code{ELIMINABLE_REGS} how to eliminate
3148 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3149 or @code{STACK_POINTER_REGNUM}.
3151 Do not define this macro if it would be the same as
3152 @code{FRAME_POINTER_REGNUM}.
3154 @findex ARG_POINTER_REGNUM
3155 @item ARG_POINTER_REGNUM
3156 The register number of the arg pointer register, which is used to access
3157 the function's argument list. On some machines, this is the same as the
3158 frame pointer register. On some machines, the hardware determines which
3159 register this is. On other machines, you can choose any register you
3160 wish for this purpose. If this is not the same register as the frame
3161 pointer register, then you must mark it as a fixed register according to
3162 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3163 (@pxref{Elimination}).
3165 @findex RETURN_ADDRESS_POINTER_REGNUM
3166 @item RETURN_ADDRESS_POINTER_REGNUM
3167 The register number of the return address pointer register, which is used to
3168 access the current function's return address from the stack. On some
3169 machines, the return address is not at a fixed offset from the frame
3170 pointer or stack pointer or argument pointer. This register can be defined
3171 to point to the return address on the stack, and then be converted by
3172 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3174 Do not define this macro unless there is no other way to get the return
3175 address from the stack.
3177 @findex STATIC_CHAIN_REGNUM
3178 @findex STATIC_CHAIN_INCOMING_REGNUM
3179 @item STATIC_CHAIN_REGNUM
3180 @itemx STATIC_CHAIN_INCOMING_REGNUM
3181 Register numbers used for passing a function's static chain pointer. If
3182 register windows are used, the register number as seen by the called
3183 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3184 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3185 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3188 The static chain register need not be a fixed register.
3190 If the static chain is passed in memory, these macros should not be
3191 defined; instead, the next two macros should be defined.
3193 @findex STATIC_CHAIN
3194 @findex STATIC_CHAIN_INCOMING
3196 @itemx STATIC_CHAIN_INCOMING
3197 If the static chain is passed in memory, these macros provide rtx giving
3198 @code{mem} expressions that denote where they are stored.
3199 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3200 as seen by the calling and called functions, respectively. Often the former
3201 will be at an offset from the stack pointer and the latter at an offset from
3204 @findex stack_pointer_rtx
3205 @findex frame_pointer_rtx
3206 @findex arg_pointer_rtx
3207 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3208 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3209 macros and should be used to refer to those items.
3211 If the static chain is passed in a register, the two previous macros should
3214 @findex DWARF_FRAME_REGISTERS
3215 @item DWARF_FRAME_REGISTERS
3216 This macro specifies the maximum number of hard registers that can be
3217 saved in a call frame. This is used to size data structures used in
3218 DWARF2 exception handling.
3220 Prior to GCC 3.0, this macro was needed in order to establish a stable
3221 exception handling ABI in the face of adding new hard registers for ISA
3222 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3223 in the number of hard registers. Nevertheless, this macro can still be
3224 used to reduce the runtime memory requirements of the exception handling
3225 routines, which can be substantial if the ISA contains a lot of
3226 registers that are not call-saved.
3228 If this macro is not defined, it defaults to
3229 @code{FIRST_PSEUDO_REGISTER}.
3231 @findex PRE_GCC3_DWARF_FRAME_REGISTERS
3232 @item PRE_GCC3_DWARF_FRAME_REGISTERS
3234 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3235 for backward compatibility in pre GCC 3.0 compiled code.
3237 If this macro is not defined, it defaults to
3238 @code{DWARF_FRAME_REGISTERS}.
3243 @subsection Eliminating Frame Pointer and Arg Pointer
3245 @c prevent bad page break with this line
3246 This is about eliminating the frame pointer and arg pointer.
3249 @findex FRAME_POINTER_REQUIRED
3250 @item FRAME_POINTER_REQUIRED
3251 A C expression which is nonzero if a function must have and use a frame
3252 pointer. This expression is evaluated in the reload pass. If its value is
3253 nonzero the function will have a frame pointer.
3255 The expression can in principle examine the current function and decide
3256 according to the facts, but on most machines the constant 0 or the
3257 constant 1 suffices. Use 0 when the machine allows code to be generated
3258 with no frame pointer, and doing so saves some time or space. Use 1
3259 when there is no possible advantage to avoiding a frame pointer.
3261 In certain cases, the compiler does not know how to produce valid code
3262 without a frame pointer. The compiler recognizes those cases and
3263 automatically gives the function a frame pointer regardless of what
3264 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3267 In a function that does not require a frame pointer, the frame pointer
3268 register can be allocated for ordinary usage, unless you mark it as a
3269 fixed register. See @code{FIXED_REGISTERS} for more information.
3271 @findex INITIAL_FRAME_POINTER_OFFSET
3272 @findex get_frame_size
3273 @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3274 A C statement to store in the variable @var{depth-var} the difference
3275 between the frame pointer and the stack pointer values immediately after
3276 the function prologue. The value would be computed from information
3277 such as the result of @code{get_frame_size ()} and the tables of
3278 registers @code{regs_ever_live} and @code{call_used_regs}.
3280 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3281 need not be defined. Otherwise, it must be defined even if
3282 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3283 case, you may set @var{depth-var} to anything.
3285 @findex ELIMINABLE_REGS
3286 @item ELIMINABLE_REGS
3287 If defined, this macro specifies a table of register pairs used to
3288 eliminate unneeded registers that point into the stack frame. If it is not
3289 defined, the only elimination attempted by the compiler is to replace
3290 references to the frame pointer with references to the stack pointer.
3292 The definition of this macro is a list of structure initializations, each
3293 of which specifies an original and replacement register.
3295 On some machines, the position of the argument pointer is not known until
3296 the compilation is completed. In such a case, a separate hard register
3297 must be used for the argument pointer. This register can be eliminated by
3298 replacing it with either the frame pointer or the argument pointer,
3299 depending on whether or not the frame pointer has been eliminated.
3301 In this case, you might specify:
3303 #define ELIMINABLE_REGS \
3304 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3305 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3306 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3309 Note that the elimination of the argument pointer with the stack pointer is
3310 specified first since that is the preferred elimination.
3312 @findex CAN_ELIMINATE
3313 @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3314 A C expression that returns nonzero if the compiler is allowed to try
3315 to replace register number @var{from-reg} with register number
3316 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3317 is defined, and will usually be the constant 1, since most of the cases
3318 preventing register elimination are things that the compiler already
3321 @findex INITIAL_ELIMINATION_OFFSET
3322 @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3323 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3324 specifies the initial difference between the specified pair of
3325 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3329 @node Stack Arguments
3330 @subsection Passing Function Arguments on the Stack
3331 @cindex arguments on stack
3332 @cindex stack arguments
3334 The macros in this section control how arguments are passed
3335 on the stack. See the following section for other macros that
3336 control passing certain arguments in registers.
3339 @findex PROMOTE_PROTOTYPES
3340 @item PROMOTE_PROTOTYPES
3341 A C expression whose value is nonzero if an argument declared in
3342 a prototype as an integral type smaller than @code{int} should
3343 actually be passed as an @code{int}. In addition to avoiding
3344 errors in certain cases of mismatch, it also makes for better
3345 code on certain machines. If the macro is not defined in target
3346 header files, it defaults to 0.
3350 A C expression. If nonzero, push insns will be used to pass
3352 If the target machine does not have a push instruction, set it to zero.
3353 That directs GCC to use an alternate strategy: to
3354 allocate the entire argument block and then store the arguments into
3355 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3357 @findex PUSH_ROUNDING
3358 @item PUSH_ROUNDING (@var{npushed})
3359 A C expression that is the number of bytes actually pushed onto the
3360 stack when an instruction attempts to push @var{npushed} bytes.
3362 On some machines, the definition
3365 #define PUSH_ROUNDING(BYTES) (BYTES)
3369 will suffice. But on other machines, instructions that appear
3370 to push one byte actually push two bytes in an attempt to maintain
3371 alignment. Then the definition should be
3374 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3377 @findex ACCUMULATE_OUTGOING_ARGS
3378 @findex current_function_outgoing_args_size
3379 @item ACCUMULATE_OUTGOING_ARGS
3380 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3381 will be computed and placed into the variable
3382 @code{current_function_outgoing_args_size}. No space will be pushed
3383 onto the stack for each call; instead, the function prologue should
3384 increase the stack frame size by this amount.
3386 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3389 @findex REG_PARM_STACK_SPACE
3390 @item REG_PARM_STACK_SPACE (@var{fndecl})
3391 Define this macro if functions should assume that stack space has been
3392 allocated for arguments even when their values are passed in
3395 The value of this macro is the size, in bytes, of the area reserved for
3396 arguments passed in registers for the function represented by @var{fndecl},
3397 which can be zero if GCC is calling a library function.
3399 This space can be allocated by the caller, or be a part of the
3400 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3402 @c above is overfull. not sure what to do. --mew 5feb93 did
3403 @c something, not sure if it looks good. --mew 10feb93
3405 @findex MAYBE_REG_PARM_STACK_SPACE
3406 @findex FINAL_REG_PARM_STACK_SPACE
3407 @item MAYBE_REG_PARM_STACK_SPACE
3408 @itemx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
3409 Define these macros in addition to the one above if functions might
3410 allocate stack space for arguments even when their values are passed
3411 in registers. These should be used when the stack space allocated
3412 for arguments in registers is not a simple constant independent of the
3413 function declaration.
3415 The value of the first macro is the size, in bytes, of the area that
3416 we should initially assume would be reserved for arguments passed in registers.
3418 The value of the second macro is the actual size, in bytes, of the area
3419 that will be reserved for arguments passed in registers. This takes two
3420 arguments: an integer representing the number of bytes of fixed sized
3421 arguments on the stack, and a tree representing the number of bytes of
3422 variable sized arguments on the stack.
3424 When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
3425 called for libcall functions, the current function, or for a function
3426 being called when it is known that such stack space must be allocated.
3427 In each case this value can be easily computed.
3429 When deciding whether a called function needs such stack space, and how
3430 much space to reserve, GCC uses these two macros instead of
3431 @code{REG_PARM_STACK_SPACE}.
3433 @findex OUTGOING_REG_PARM_STACK_SPACE
3434 @item OUTGOING_REG_PARM_STACK_SPACE
3435 Define this if it is the responsibility of the caller to allocate the area
3436 reserved for arguments passed in registers.
3438 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3439 whether the space for these arguments counts in the value of
3440 @code{current_function_outgoing_args_size}.
3442 @findex STACK_PARMS_IN_REG_PARM_AREA
3443 @item STACK_PARMS_IN_REG_PARM_AREA
3444 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3445 stack parameters don't skip the area specified by it.
3446 @c i changed this, makes more sens and it should have taken care of the
3447 @c overfull.. not as specific, tho. --mew 5feb93
3449 Normally, when a parameter is not passed in registers, it is placed on the
3450 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3451 suppresses this behavior and causes the parameter to be passed on the
3452 stack in its natural location.
3454 @findex RETURN_POPS_ARGS
3455 @item RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3456 A C expression that should indicate the number of bytes of its own
3457 arguments that a function pops on returning, or 0 if the
3458 function pops no arguments and the caller must therefore pop them all
3459 after the function returns.
3461 @var{fundecl} is a C variable whose value is a tree node that describes
3462 the function in question. Normally it is a node of type
3463 @code{FUNCTION_DECL} that describes the declaration of the function.
3464 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3466 @var{funtype} is a C variable whose value is a tree node that
3467 describes the function in question. Normally it is a node of type
3468 @code{FUNCTION_TYPE} that describes the data type of the function.
3469 From this it is possible to obtain the data types of the value and
3470 arguments (if known).
3472 When a call to a library function is being considered, @var{fundecl}
3473 will contain an identifier node for the library function. Thus, if
3474 you need to distinguish among various library functions, you can do so
3475 by their names. Note that ``library function'' in this context means
3476 a function used to perform arithmetic, whose name is known specially
3477 in the compiler and was not mentioned in the C code being compiled.
3479 @var{stack-size} is the number of bytes of arguments passed on the
3480 stack. If a variable number of bytes is passed, it is zero, and
3481 argument popping will always be the responsibility of the calling function.
3483 On the VAX, all functions always pop their arguments, so the definition
3484 of this macro is @var{stack-size}. On the 68000, using the standard
3485 calling convention, no functions pop their arguments, so the value of
3486 the macro is always 0 in this case. But an alternative calling
3487 convention is available in which functions that take a fixed number of
3488 arguments pop them but other functions (such as @code{printf}) pop
3489 nothing (the caller pops all). When this convention is in use,
3490 @var{funtype} is examined to determine whether a function takes a fixed
3491 number of arguments.
3493 @findex CALL_POPS_ARGS
3494 @item CALL_POPS_ARGS (@var{cum})
3495 A C expression that should indicate the number of bytes a call sequence
3496 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3497 when compiling a function call.
3499 @var{cum} is the variable in which all arguments to the called function
3500 have been accumulated.
3502 On certain architectures, such as the SH5, a call trampoline is used
3503 that pops certain registers off the stack, depending on the arguments
3504 that have been passed to the function. Since this is a property of the
3505 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3510 @node Register Arguments
3511 @subsection Passing Arguments in Registers
3512 @cindex arguments in registers
3513 @cindex registers arguments
3515 This section describes the macros which let you control how various
3516 types of arguments are passed in registers or how they are arranged in
3520 @findex FUNCTION_ARG
3521 @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3522 A C expression that controls whether a function argument is passed
3523 in a register, and which register.
3525 The arguments are @var{cum}, which summarizes all the previous
3526 arguments; @var{mode}, the machine mode of the argument; @var{type},
3527 the data type of the argument as a tree node or 0 if that is not known
3528 (which happens for C support library functions); and @var{named},
3529 which is 1 for an ordinary argument and 0 for nameless arguments that
3530 correspond to @samp{@dots{}} in the called function's prototype.
3531 @var{type} can be an incomplete type if a syntax error has previously
3534 The value of the expression is usually either a @code{reg} RTX for the
3535 hard register in which to pass the argument, or zero to pass the
3536 argument on the stack.
3538 For machines like the VAX and 68000, where normally all arguments are
3539 pushed, zero suffices as a definition.
3541 The value of the expression can also be a @code{parallel} RTX@. This is
3542 used when an argument is passed in multiple locations. The mode of the
3543 of the @code{parallel} should be the mode of the entire argument. The
3544 @code{parallel} holds any number of @code{expr_list} pairs; each one
3545 describes where part of the argument is passed. In each
3546 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3547 register in which to pass this part of the argument, and the mode of the
3548 register RTX indicates how large this part of the argument is. The
3549 second operand of the @code{expr_list} is a @code{const_int} which gives
3550 the offset in bytes into the entire argument of where this part starts.
3551 As a special exception the first @code{expr_list} in the @code{parallel}
3552 RTX may have a first operand of zero. This indicates that the entire
3553 argument is also stored on the stack.
3555 The last time this macro is called, it is called with @code{MODE ==
3556 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3557 pattern as operands 2 and 3 respectively.
3559 @cindex @file{stdarg.h} and register arguments
3560 The usual way to make the ISO library @file{stdarg.h} work on a machine
3561 where some arguments are usually passed in registers, is to cause
3562 nameless arguments to be passed on the stack instead. This is done
3563 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3565 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3566 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3567 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
3568 in the definition of this macro to determine if this argument is of a
3569 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3570 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3571 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3572 defined, the argument will be computed in the stack and then loaded into
3575 @findex MUST_PASS_IN_STACK
3576 @item MUST_PASS_IN_STACK (@var{mode}, @var{type})
3577 Define as a C expression that evaluates to nonzero if we do not know how
3578 to pass TYPE solely in registers. The file @file{expr.h} defines a
3579 definition that is usually appropriate, refer to @file{expr.h} for additional
3582 @findex FUNCTION_INCOMING_ARG
3583 @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3584 Define this macro if the target machine has ``register windows'', so
3585 that the register in which a function sees an arguments is not
3586 necessarily the same as the one in which the caller passed the
3589 For such machines, @code{FUNCTION_ARG} computes the register in which
3590 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3591 be defined in a similar fashion to tell the function being called
3592 where the arguments will arrive.
3594 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3595 serves both purposes.
3597 @findex FUNCTION_ARG_PARTIAL_NREGS
3598 @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
3599 A C expression for the number of words, at the beginning of an
3600 argument, that must be put in registers. The value must be zero for
3601 arguments that are passed entirely in registers or that are entirely
3602 pushed on the stack.
3604 On some machines, certain arguments must be passed partially in
3605 registers and partially in memory. On these machines, typically the
3606 first @var{n} words of arguments are passed in registers, and the rest
3607 on the stack. If a multi-word argument (a @code{double} or a
3608 structure) crosses that boundary, its first few words must be passed
3609 in registers and the rest must be pushed. This macro tells the
3610 compiler when this occurs, and how many of the words should go in
3613 @code{FUNCTION_ARG} for these arguments should return the first
3614 register to be used by the caller for this argument; likewise
3615 @code{FUNCTION_INCOMING_ARG}, for the called function.
3617 @findex FUNCTION_ARG_PASS_BY_REFERENCE
3618 @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3619 A C expression that indicates when an argument must be passed by reference.
3620 If nonzero for an argument, a copy of that argument is made in memory and a
3621 pointer to the argument is passed instead of the argument itself.
3622 The pointer is passed in whatever way is appropriate for passing a pointer
3625 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
3626 definition of this macro might be
3628 #define FUNCTION_ARG_PASS_BY_REFERENCE\
3629 (CUM, MODE, TYPE, NAMED) \
3630 MUST_PASS_IN_STACK (MODE, TYPE)
3632 @c this is *still* too long. --mew 5feb93
3634 @findex FUNCTION_ARG_CALLEE_COPIES
3635 @item FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
3636 If defined, a C expression that indicates when it is the called function's
3637 responsibility to make a copy of arguments passed by invisible reference.
3638 Normally, the caller makes a copy and passes the address of the copy to the
3639 routine being called. When @code{FUNCTION_ARG_CALLEE_COPIES} is defined and is
3640 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
3641 ``live'' value. The called function must not modify this value. If it can be
3642 determined that the value won't be modified, it need not make a copy;
3643 otherwise a copy must be made.
3645 @findex CUMULATIVE_ARGS
3646 @item CUMULATIVE_ARGS
3647 A C type for declaring a variable that is used as the first argument of
3648 @code{FUNCTION_ARG} and other related values. For some target machines,
3649 the type @code{int} suffices and can hold the number of bytes of
3652 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3653 arguments that have been passed on the stack. The compiler has other
3654 variables to keep track of that. For target machines on which all
3655 arguments are passed on the stack, there is no need to store anything in
3656 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3657 should not be empty, so use @code{int}.
3659 @findex INIT_CUMULATIVE_ARGS
3660 @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{indirect})
3661 A C statement (sans semicolon) for initializing the variable @var{cum}
3662 for the state at the beginning of the argument list. The variable has
3663 type @code{CUMULATIVE_ARGS}. The value of @var{fntype} is the tree node
3664 for the data type of the function which will receive the args, or 0
3665 if the args are to a compiler support library function. The value of
3666 @var{indirect} is nonzero when processing an indirect call, for example
3667 a call through a function pointer. The value of @var{indirect} is zero
3668 for a call to an explicitly named function, a library function call, or when
3669 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3672 When processing a call to a compiler support library function,
3673 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3674 contains the name of the function, as a string. @var{libname} is 0 when
3675 an ordinary C function call is being processed. Thus, each time this
3676 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3677 never both of them at once.
3679 @findex INIT_CUMULATIVE_LIBCALL_ARGS
3680 @item INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3681 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3682 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3683 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3684 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3685 0)} is used instead.
3687 @findex INIT_CUMULATIVE_INCOMING_ARGS
3688 @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3689 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3690 finding the arguments for the function being compiled. If this macro is
3691 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3693 The value passed for @var{libname} is always 0, since library routines
3694 with special calling conventions are never compiled with GCC@. The
3695 argument @var{libname} exists for symmetry with
3696 @code{INIT_CUMULATIVE_ARGS}.
3697 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3698 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3700 @findex FUNCTION_ARG_ADVANCE
3701 @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3702 A C statement (sans semicolon) to update the summarizer variable
3703 @var{cum} to advance past an argument in the argument list. The
3704 values @var{mode}, @var{type} and @var{named} describe that argument.
3705 Once this is done, the variable @var{cum} is suitable for analyzing
3706 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
3708 This macro need not do anything if the argument in question was passed
3709 on the stack. The compiler knows how to track the amount of stack space
3710 used for arguments without any special help.
3712 @findex FUNCTION_ARG_PADDING
3713 @item FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3714 If defined, a C expression which determines whether, and in which direction,
3715 to pad out an argument with extra space. The value should be of type
3716 @code{enum direction}: either @code{upward} to pad above the argument,
3717 @code{downward} to pad below, or @code{none} to inhibit padding.
3719 The @emph{amount} of padding is always just enough to reach the next
3720 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3723 This macro has a default definition which is right for most systems.
3724 For little-endian machines, the default is to pad upward. For
3725 big-endian machines, the default is to pad downward for an argument of
3726 constant size shorter than an @code{int}, and upward otherwise.
3728 @findex PAD_VARARGS_DOWN
3729 @item PAD_VARARGS_DOWN
3730 If defined, a C expression which determines whether the default
3731 implementation of va_arg will attempt to pad down before reading the
3732 next argument, if that argument is smaller than its aligned space as
3733 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3734 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3736 @findex FUNCTION_ARG_BOUNDARY
3737 @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3738 If defined, a C expression that gives the alignment boundary, in bits,
3739 of an argument with the specified mode and type. If it is not defined,
3740 @code{PARM_BOUNDARY} is used for all arguments.
3742 @findex FUNCTION_ARG_REGNO_P
3743 @item FUNCTION_ARG_REGNO_P (@var{regno})
3744 A C expression that is nonzero if @var{regno} is the number of a hard
3745 register in which function arguments are sometimes passed. This does
3746 @emph{not} include implicit arguments such as the static chain and
3747 the structure-value address. On many machines, no registers can be
3748 used for this purpose since all function arguments are pushed on the
3751 @findex LOAD_ARGS_REVERSED
3752 @item LOAD_ARGS_REVERSED
3753 If defined, the order in which arguments are loaded into their
3754 respective argument registers is reversed so that the last
3755 argument is loaded first. This macro only affects arguments
3756 passed in registers.
3761 @subsection How Scalar Function Values Are Returned
3762 @cindex return values in registers
3763 @cindex values, returned by functions
3764 @cindex scalars, returned as values
3766 This section discusses the macros that control returning scalars as
3767 values---values that can fit in registers.
3770 @findex FUNCTION_VALUE
3771 @item FUNCTION_VALUE (@var{valtype}, @var{func})
3772 A C expression to create an RTX representing the place where a
3773 function returns a value of data type @var{valtype}. @var{valtype} is
3774 a tree node representing a data type. Write @code{TYPE_MODE
3775 (@var{valtype})} to get the machine mode used to represent that type.
3776 On many machines, only the mode is relevant. (Actually, on most
3777 machines, scalar values are returned in the same place regardless of
3780 The value of the expression is usually a @code{reg} RTX for the hard
3781 register where the return value is stored. The value can also be a
3782 @code{parallel} RTX, if the return value is in multiple places. See
3783 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3785 If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
3786 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3789 If the precise function being called is known, @var{func} is a tree
3790 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3791 pointer. This makes it possible to use a different value-returning
3792 convention for specific functions when all their calls are
3795 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3796 types, because these are returned in another way. See
3797 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3799 @findex FUNCTION_OUTGOING_VALUE
3800 @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3801 Define this macro if the target machine has ``register windows''
3802 so that the register in which a function returns its value is not
3803 the same as the one in which the caller sees the value.
3805 For such machines, @code{FUNCTION_VALUE} computes the register in which
3806 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3807 defined in a similar fashion to tell the function where to put the
3810 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3811 @code{FUNCTION_VALUE} serves both purposes.
3813 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3814 aggregate data types, because these are returned in another way. See
3815 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3817 @findex LIBCALL_VALUE
3818 @item LIBCALL_VALUE (@var{mode})
3819 A C expression to create an RTX representing the place where a library
3820 function returns a value of mode @var{mode}. If the precise function
3821 being called is known, @var{func} is a tree node
3822 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3823 pointer. This makes it possible to use a different value-returning
3824 convention for specific functions when all their calls are
3827 Note that ``library function'' in this context means a compiler
3828 support routine, used to perform arithmetic, whose name is known
3829 specially by the compiler and was not mentioned in the C code being
3832 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3833 data types, because none of the library functions returns such types.
3835 @findex FUNCTION_VALUE_REGNO_P
3836 @item FUNCTION_VALUE_REGNO_P (@var{regno})
3837 A C expression that is nonzero if @var{regno} is the number of a hard
3838 register in which the values of called function may come back.
3840 A register whose use for returning values is limited to serving as the
3841 second of a pair (for a value of type @code{double}, say) need not be
3842 recognized by this macro. So for most machines, this definition
3846 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3849 If the machine has register windows, so that the caller and the called
3850 function use different registers for the return value, this macro
3851 should recognize only the caller's register numbers.
3853 @findex APPLY_RESULT_SIZE
3854 @item APPLY_RESULT_SIZE
3855 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3856 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3857 saving and restoring an arbitrary return value.
3860 @node Aggregate Return
3861 @subsection How Large Values Are Returned
3862 @cindex aggregates as return values
3863 @cindex large return values
3864 @cindex returning aggregate values
3865 @cindex structure value address
3867 When a function value's mode is @code{BLKmode} (and in some other
3868 cases), the value is not returned according to @code{FUNCTION_VALUE}
3869 (@pxref{Scalar Return}). Instead, the caller passes the address of a
3870 block of memory in which the value should be stored. This address
3871 is called the @dfn{structure value address}.
3873 This section describes how to control returning structure values in
3877 @findex RETURN_IN_MEMORY
3878 @item RETURN_IN_MEMORY (@var{type})
3879 A C expression which can inhibit the returning of certain function
3880 values in registers, based on the type of value. A nonzero value says
3881 to return the function value in memory, just as large structures are
3882 always returned. Here @var{type} will be a C expression of type
3883 @code{tree}, representing the data type of the value.
3885 Note that values of mode @code{BLKmode} must be explicitly handled
3886 by this macro. Also, the option @option{-fpcc-struct-return}
3887 takes effect regardless of this macro. On most systems, it is
3888 possible to leave the macro undefined; this causes a default
3889 definition to be used, whose value is the constant 1 for @code{BLKmode}
3890 values, and 0 otherwise.
3892 Do not use this macro to indicate that structures and unions should always
3893 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
3896 @findex DEFAULT_PCC_STRUCT_RETURN
3897 @item DEFAULT_PCC_STRUCT_RETURN
3898 Define this macro to be 1 if all structure and union return values must be
3899 in memory. Since this results in slower code, this should be defined
3900 only if needed for compatibility with other compilers or with an ABI@.
3901 If you define this macro to be 0, then the conventions used for structure
3902 and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
3904 If not defined, this defaults to the value 1.
3906 @findex STRUCT_VALUE_REGNUM
3907 @item STRUCT_VALUE_REGNUM
3908 If the structure value address is passed in a register, then
3909 @code{STRUCT_VALUE_REGNUM} should be the number of that register.
3911 @findex STRUCT_VALUE
3913 If the structure value address is not passed in a register, define
3914 @code{STRUCT_VALUE} as an expression returning an RTX for the place
3915 where the address is passed. If it returns 0, the address is passed as
3916 an ``invisible'' first argument.
3918 @findex STRUCT_VALUE_INCOMING_REGNUM
3919 @item STRUCT_VALUE_INCOMING_REGNUM
3920 On some architectures the place where the structure value address
3921 is found by the called function is not the same place that the
3922 caller put it. This can be due to register windows, or it could
3923 be because the function prologue moves it to a different place.
3925 If the incoming location of the structure value address is in a
3926 register, define this macro as the register number.
3928 @findex STRUCT_VALUE_INCOMING
3929 @item STRUCT_VALUE_INCOMING
3930 If the incoming location is not a register, then you should define
3931 @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the
3932 called function should find the value. If it should find the value on
3933 the stack, define this to create a @code{mem} which refers to the frame
3934 pointer. A definition of 0 means that the address is passed as an
3935 ``invisible'' first argument.
3937 @findex PCC_STATIC_STRUCT_RETURN
3938 @item PCC_STATIC_STRUCT_RETURN
3939 Define this macro if the usual system convention on the target machine
3940 for returning structures and unions is for the called function to return
3941 the address of a static variable containing the value.
3943 Do not define this if the usual system convention is for the caller to
3944 pass an address to the subroutine.
3946 This macro has effect in @option{-fpcc-struct-return} mode, but it does
3947 nothing when you use @option{-freg-struct-return} mode.
3951 @subsection Caller-Saves Register Allocation
3953 If you enable it, GCC can save registers around function calls. This
3954 makes it possible to use call-clobbered registers to hold variables that
3955 must live across calls.
3958 @findex DEFAULT_CALLER_SAVES
3959 @item DEFAULT_CALLER_SAVES
3960 Define this macro if function calls on the target machine do not preserve
3961 any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
3962 for all registers. When defined, this macro enables @option{-fcaller-saves}
3963 by default for all optimization levels. It has no effect for optimization
3964 levels 2 and higher, where @option{-fcaller-saves} is the default.
3966 @findex CALLER_SAVE_PROFITABLE
3967 @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
3968 A C expression to determine whether it is worthwhile to consider placing
3969 a pseudo-register in a call-clobbered hard register and saving and
3970 restoring it around each function call. The expression should be 1 when
3971 this is worth doing, and 0 otherwise.
3973 If you don't define this macro, a default is used which is good on most
3974 machines: @code{4 * @var{calls} < @var{refs}}.
3976 @findex HARD_REGNO_CALLER_SAVE_MODE
3977 @item HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3978 A C expression specifying which mode is required for saving @var{nregs}
3979 of a pseudo-register in call-clobbered hard register @var{regno}. If
3980 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3981 returned. For most machines this macro need not be defined since GCC
3982 will select the smallest suitable mode.
3985 @node Function Entry
3986 @subsection Function Entry and Exit
3987 @cindex function entry and exit
3991 This section describes the macros that output function entry
3992 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3994 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
3995 If defined, a function that outputs the assembler code for entry to a
3996 function. The prologue is responsible for setting up the stack frame,
3997 initializing the frame pointer register, saving registers that must be
3998 saved, and allocating @var{size} additional bytes of storage for the
3999 local variables. @var{size} is an integer. @var{file} is a stdio
4000 stream to which the assembler code should be output.
4002 The label for the beginning of the function need not be output by this
4003 macro. That has already been done when the macro is run.
4005 @findex regs_ever_live
4006 To determine which registers to save, the macro can refer to the array
4007 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4008 @var{r} is used anywhere within the function. This implies the function
4009 prologue should save register @var{r}, provided it is not one of the
4010 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4011 @code{regs_ever_live}.)
4013 On machines that have ``register windows'', the function entry code does
4014 not save on the stack the registers that are in the windows, even if
4015 they are supposed to be preserved by function calls; instead it takes
4016 appropriate steps to ``push'' the register stack, if any non-call-used
4017 registers are used in the function.
4019 @findex frame_pointer_needed
4020 On machines where functions may or may not have frame-pointers, the
4021 function entry code must vary accordingly; it must set up the frame
4022 pointer if one is wanted, and not otherwise. To determine whether a
4023 frame pointer is in wanted, the macro can refer to the variable
4024 @code{frame_pointer_needed}. The variable's value will be 1 at run
4025 time in a function that needs a frame pointer. @xref{Elimination}.
4027 The function entry code is responsible for allocating any stack space
4028 required for the function. This stack space consists of the regions
4029 listed below. In most cases, these regions are allocated in the
4030 order listed, with the last listed region closest to the top of the
4031 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4032 the highest address if it is not defined). You can use a different order
4033 for a machine if doing so is more convenient or required for
4034 compatibility reasons. Except in cases where required by standard
4035 or by a debugger, there is no reason why the stack layout used by GCC
4036 need agree with that used by other compilers for a machine.
4039 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4040 If defined, a function that outputs assembler code at the end of a
4041 prologue. This should be used when the function prologue is being
4042 emitted as RTL, and you have some extra assembler that needs to be
4043 emitted. @xref{prologue instruction pattern}.
4046 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4047 If defined, a function that outputs assembler code at the start of an
4048 epilogue. This should be used when the function epilogue is being
4049 emitted as RTL, and you have some extra assembler that needs to be
4050 emitted. @xref{epilogue instruction pattern}.
4053 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4054 If defined, a function that outputs the assembler code for exit from a
4055 function. The epilogue is responsible for restoring the saved
4056 registers and stack pointer to their values when the function was
4057 called, and returning control to the caller. This macro takes the
4058 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4059 registers to restore are determined from @code{regs_ever_live} and
4060 @code{CALL_USED_REGISTERS} in the same way.
4062 On some machines, there is a single instruction that does all the work
4063 of returning from the function. On these machines, give that
4064 instruction the name @samp{return} and do not define the macro
4065 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4067 Do not define a pattern named @samp{return} if you want the
4068 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4069 switches to control whether return instructions or epilogues are used,
4070 define a @samp{return} pattern with a validity condition that tests the
4071 target switches appropriately. If the @samp{return} pattern's validity
4072 condition is false, epilogues will be used.
4074 On machines where functions may or may not have frame-pointers, the
4075 function exit code must vary accordingly. Sometimes the code for these
4076 two cases is completely different. To determine whether a frame pointer
4077 is wanted, the macro can refer to the variable
4078 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4079 a function that needs a frame pointer.
4081 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4082 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4083 The C variable @code{current_function_is_leaf} is nonzero for such a
4084 function. @xref{Leaf Functions}.
4086 On some machines, some functions pop their arguments on exit while
4087 others leave that for the caller to do. For example, the 68020 when
4088 given @option{-mrtd} pops arguments in functions that take a fixed
4089 number of arguments.
4091 @findex current_function_pops_args
4092 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4093 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4094 needs to know what was decided. The variable that is called
4095 @code{current_function_pops_args} is the number of bytes of its
4096 arguments that a function should pop. @xref{Scalar Return}.
4097 @c what is the "its arguments" in the above sentence referring to, pray
4098 @c tell? --mew 5feb93
4105 @findex current_function_pretend_args_size
4106 A region of @code{current_function_pretend_args_size} bytes of
4107 uninitialized space just underneath the first argument arriving on the
4108 stack. (This may not be at the very start of the allocated stack region
4109 if the calling sequence has pushed anything else since pushing the stack
4110 arguments. But usually, on such machines, nothing else has been pushed
4111 yet, because the function prologue itself does all the pushing.) This
4112 region is used on machines where an argument may be passed partly in
4113 registers and partly in memory, and, in some cases to support the
4114 features in @code{<stdarg.h>}.
4117 An area of memory used to save certain registers used by the function.
4118 The size of this area, which may also include space for such things as
4119 the return address and pointers to previous stack frames, is
4120 machine-specific and usually depends on which registers have been used
4121 in the function. Machines with register windows often do not require
4125 A region of at least @var{size} bytes, possibly rounded up to an allocation
4126 boundary, to contain the local variables of the function. On some machines,
4127 this region and the save area may occur in the opposite order, with the
4128 save area closer to the top of the stack.
4131 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4132 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4133 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4134 argument lists of the function. @xref{Stack Arguments}.
4137 Normally, it is necessary for the macros
4138 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4139 @code{TARGET_ASM_FUNCTION_EPILOGUE} to treat leaf functions specially.
4140 The C variable @code{current_function_is_leaf} is nonzero for such a
4143 @findex EXIT_IGNORE_STACK
4144 @item EXIT_IGNORE_STACK
4145 Define this macro as a C expression that is nonzero if the return
4146 instruction or the function epilogue ignores the value of the stack
4147 pointer; in other words, if it is safe to delete an instruction to
4148 adjust the stack pointer before a return from the function.
4150 Note that this macro's value is relevant only for functions for which
4151 frame pointers are maintained. It is never safe to delete a final
4152 stack adjustment in a function that has no frame pointer, and the
4153 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4155 @findex EPILOGUE_USES
4156 @item EPILOGUE_USES (@var{regno})
4157 Define this macro as a C expression that is nonzero for registers that are
4158 used by the epilogue or the @samp{return} pattern. The stack and frame
4159 pointer registers are already be assumed to be used as needed.
4162 @item EH_USES (@var{regno})
4163 Define this macro as a C expression that is nonzero for registers that are
4164 used by the exception handling mechanism, and so should be considered live
4165 on entry to an exception edge.
4167 @findex DELAY_SLOTS_FOR_EPILOGUE
4168 @item DELAY_SLOTS_FOR_EPILOGUE
4169 Define this macro if the function epilogue contains delay slots to which
4170 instructions from the rest of the function can be ``moved''. The
4171 definition should be a C expression whose value is an integer
4172 representing the number of delay slots there.
4174 @findex ELIGIBLE_FOR_EPILOGUE_DELAY
4175 @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4176 A C expression that returns 1 if @var{insn} can be placed in delay
4177 slot number @var{n} of the epilogue.
4179 The argument @var{n} is an integer which identifies the delay slot now
4180 being considered (since different slots may have different rules of
4181 eligibility). It is never negative and is always less than the number
4182 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4183 If you reject a particular insn for a given delay slot, in principle, it
4184 may be reconsidered for a subsequent delay slot. Also, other insns may
4185 (at least in principle) be considered for the so far unfilled delay
4188 @findex current_function_epilogue_delay_list
4189 @findex final_scan_insn
4190 The insns accepted to fill the epilogue delay slots are put in an RTL
4191 list made with @code{insn_list} objects, stored in the variable
4192 @code{current_function_epilogue_delay_list}. The insn for the first
4193 delay slot comes first in the list. Your definition of the macro
4194 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4195 outputting the insns in this list, usually by calling
4196 @code{final_scan_insn}.
4198 You need not define this macro if you did not define
4199 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4203 @findex TARGET_ASM_OUTPUT_MI_THUNK
4204 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, tree @var{function})
4205 A function that outputs the assembler code for a thunk
4206 function, used to implement C++ virtual function calls with multiple
4207 inheritance. The thunk acts as a wrapper around a virtual function,
4208 adjusting the implicit object parameter before handing control off to
4211 First, emit code to add the integer @var{delta} to the location that
4212 contains the incoming first argument. Assume that this argument
4213 contains a pointer, and is the one used to pass the @code{this} pointer
4214 in C++. This is the incoming argument @emph{before} the function prologue,
4215 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4216 all other incoming arguments.
4218 After the addition, emit code to jump to @var{function}, which is a
4219 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4220 not touch the return address. Hence returning from @var{FUNCTION} will
4221 return to whoever called the current @samp{thunk}.
4223 The effect must be as if @var{function} had been called directly with
4224 the adjusted first argument. This macro is responsible for emitting all
4225 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4226 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4228 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4229 have already been extracted from it.) It might possibly be useful on
4230 some targets, but probably not.
4232 If you do not define this macro, the target-independent code in the C++
4233 front end will generate a less efficient heavyweight thunk that calls
4234 @var{function} instead of jumping to it. The generic approach does
4235 not support varargs.
4238 @findex TARGET_ASM_OUTPUT_MI_VCALL_THUNK
4239 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_VCALL_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, int @var{vcall_offset}, tree @var{function})
4240 A function like @code{TARGET_ASM_OUTPUT_MI_THUNK}, except that if
4241 @var{vcall_offset} is nonzero, an additional adjustment should be made
4242 after adding @code{delta}. In particular, if @var{p} is the
4243 adjusted pointer, the following adjustment should be made:
4246 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4250 If this function is defined, it will always be used in place of
4251 @code{TARGET_ASM_OUTPUT_MI_THUNK}.
4256 @subsection Generating Code for Profiling
4257 @cindex profiling, code generation
4259 These macros will help you generate code for profiling.
4262 @findex FUNCTION_PROFILER
4263 @item FUNCTION_PROFILER (@var{file}, @var{labelno})
4264 A C statement or compound statement to output to @var{file} some
4265 assembler code to call the profiling subroutine @code{mcount}.
4268 The details of how @code{mcount} expects to be called are determined by
4269 your operating system environment, not by GCC@. To figure them out,
4270 compile a small program for profiling using the system's installed C
4271 compiler and look at the assembler code that results.
4273 Older implementations of @code{mcount} expect the address of a counter
4274 variable to be loaded into some register. The name of this variable is
4275 @samp{LP} followed by the number @var{labelno}, so you would generate
4276 the name using @samp{LP%d} in a @code{fprintf}.
4278 @findex PROFILE_HOOK
4280 A C statement or compound statement to output to @var{file} some assembly
4281 code to call the profiling subroutine @code{mcount} even the target does
4282 not support profiling.
4284 @findex NO_PROFILE_COUNTERS
4285 @item NO_PROFILE_COUNTERS
4286 Define this macro if the @code{mcount} subroutine on your system does
4287 not need a counter variable allocated for each function. This is true
4288 for almost all modern implementations. If you define this macro, you
4289 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
4291 @findex PROFILE_BEFORE_PROLOGUE
4292 @item PROFILE_BEFORE_PROLOGUE
4293 Define this macro if the code for function profiling should come before
4294 the function prologue. Normally, the profiling code comes after.
4298 @subsection Permitting tail calls
4301 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4302 True if it is ok to do sibling call optimization for the specified
4303 call expression @var{exp}. @var{decl} will be the called function,
4304 or @code{NULL} if this is an indirect call.
4306 It is not uncommon for limitations of calling conventions to prevent
4307 tail calls to functions outside the current unit of translation, or
4308 during PIC compilation. The hook is used to enforce these restrictions,
4309 as the @code{sibcall} md pattern can not fail, or fall over to a
4310 ``normal'' call. The criteria for successful sibling call optimization
4311 may vary greatly between different architectures.
4315 @section Implementing the Varargs Macros
4316 @cindex varargs implementation
4318 GCC comes with an implementation of @code{<varargs.h>} and
4319 @code{<stdarg.h>} that work without change on machines that pass arguments
4320 on the stack. Other machines require their own implementations of
4321 varargs, and the two machine independent header files must have
4322 conditionals to include it.
4324 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4325 the calling convention for @code{va_start}. The traditional
4326 implementation takes just one argument, which is the variable in which
4327 to store the argument pointer. The ISO implementation of
4328 @code{va_start} takes an additional second argument. The user is
4329 supposed to write the last named argument of the function here.
4331 However, @code{va_start} should not use this argument. The way to find
4332 the end of the named arguments is with the built-in functions described
4336 @findex __builtin_saveregs
4337 @item __builtin_saveregs ()
4338 Use this built-in function to save the argument registers in memory so
4339 that the varargs mechanism can access them. Both ISO and traditional
4340 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4341 you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
4343 On some machines, @code{__builtin_saveregs} is open-coded under the
4344 control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
4345 it calls a routine written in assembler language, found in
4348 Code generated for the call to @code{__builtin_saveregs} appears at the
4349 beginning of the function, as opposed to where the call to
4350 @code{__builtin_saveregs} is written, regardless of what the code is.
4351 This is because the registers must be saved before the function starts
4352 to use them for its own purposes.
4353 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4356 @findex __builtin_args_info
4357 @item __builtin_args_info (@var{category})
4358 Use this built-in function to find the first anonymous arguments in
4361 In general, a machine may have several categories of registers used for
4362 arguments, each for a particular category of data types. (For example,
4363 on some machines, floating-point registers are used for floating-point
4364 arguments while other arguments are passed in the general registers.)
4365 To make non-varargs functions use the proper calling convention, you
4366 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4367 registers in each category have been used so far
4369 @code{__builtin_args_info} accesses the same data structure of type
4370 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4371 with it, with @var{category} specifying which word to access. Thus, the
4372 value indicates the first unused register in a given category.
4374 Normally, you would use @code{__builtin_args_info} in the implementation
4375 of @code{va_start}, accessing each category just once and storing the
4376 value in the @code{va_list} object. This is because @code{va_list} will
4377 have to update the values, and there is no way to alter the
4378 values accessed by @code{__builtin_args_info}.
4380 @findex __builtin_next_arg
4381 @item __builtin_next_arg (@var{lastarg})
4382 This is the equivalent of @code{__builtin_args_info}, for stack
4383 arguments. It returns the address of the first anonymous stack
4384 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4385 returns the address of the location above the first anonymous stack
4386 argument. Use it in @code{va_start} to initialize the pointer for
4387 fetching arguments from the stack. Also use it in @code{va_start} to
4388 verify that the second parameter @var{lastarg} is the last named argument
4389 of the current function.
4391 @findex __builtin_classify_type
4392 @item __builtin_classify_type (@var{object})
4393 Since each machine has its own conventions for which data types are
4394 passed in which kind of register, your implementation of @code{va_arg}
4395 has to embody these conventions. The easiest way to categorize the
4396 specified data type is to use @code{__builtin_classify_type} together
4397 with @code{sizeof} and @code{__alignof__}.
4399 @code{__builtin_classify_type} ignores the value of @var{object},
4400 considering only its data type. It returns an integer describing what
4401 kind of type that is---integer, floating, pointer, structure, and so on.
4403 The file @file{typeclass.h} defines an enumeration that you can use to
4404 interpret the values of @code{__builtin_classify_type}.
4407 These machine description macros help implement varargs:
4410 @findex EXPAND_BUILTIN_SAVEREGS
4411 @item EXPAND_BUILTIN_SAVEREGS ()
4412 If defined, is a C expression that produces the machine-specific code
4413 for a call to @code{__builtin_saveregs}. This code will be moved to the
4414 very beginning of the function, before any parameter access are made.
4415 The return value of this function should be an RTX that contains the
4416 value to use as the return of @code{__builtin_saveregs}.
4418 @findex SETUP_INCOMING_VARARGS
4419 @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type}, @var{pretend_args_size}, @var{second_time})
4420 This macro offers an alternative to using @code{__builtin_saveregs} and
4421 defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the
4422 anonymous register arguments into the stack so that all the arguments
4423 appear to have been passed consecutively on the stack. Once this is
4424 done, you can use the standard implementation of varargs that works for
4425 machines that pass all their arguments on the stack.
4427 The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data
4428 structure, containing the values that are obtained after processing the
4429 named arguments. The arguments @var{mode} and @var{type} describe the
4430 last named argument---its machine mode and its data type as a tree node.
4432 The macro implementation should do two things: first, push onto the
4433 stack all the argument registers @emph{not} used for the named
4434 arguments, and second, store the size of the data thus pushed into the
4435 @code{int}-valued variable whose name is supplied as the argument
4436 @var{pretend_args_size}. The value that you store here will serve as
4437 additional offset for setting up the stack frame.
4439 Because you must generate code to push the anonymous arguments at
4440 compile time without knowing their data types,
4441 @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just
4442 a single category of argument register and use it uniformly for all data
4445 If the argument @var{second_time} is nonzero, it means that the
4446 arguments of the function are being analyzed for the second time. This
4447 happens for an inline function, which is not actually compiled until the
4448 end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should
4449 not generate any instructions in this case.
4451 @findex STRICT_ARGUMENT_NAMING
4452 @item STRICT_ARGUMENT_NAMING
4453 Define this macro to be a nonzero value if the location where a function
4454 argument is passed depends on whether or not it is a named argument.
4456 This macro controls how the @var{named} argument to @code{FUNCTION_ARG}
4457 is set for varargs and stdarg functions. If this macro returns a
4458 nonzero value, the @var{named} argument is always true for named
4459 arguments, and false for unnamed arguments. If it returns a value of
4460 zero, but @code{SETUP_INCOMING_VARARGS} is defined, then all arguments
4461 are treated as named. Otherwise, all named arguments except the last
4462 are treated as named.
4464 You need not define this macro if it always returns zero.
4466 @findex PRETEND_OUTGOING_VARARGS_NAMED
4467 @item PRETEND_OUTGOING_VARARGS_NAMED
4468 If you need to conditionally change ABIs so that one works with
4469 @code{SETUP_INCOMING_VARARGS}, but the other works like neither
4470 @code{SETUP_INCOMING_VARARGS} nor @code{STRICT_ARGUMENT_NAMING} was
4471 defined, then define this macro to return nonzero if
4472 @code{SETUP_INCOMING_VARARGS} is used, zero otherwise.
4473 Otherwise, you should not define this macro.
4477 @section Trampolines for Nested Functions
4478 @cindex trampolines for nested functions
4479 @cindex nested functions, trampolines for
4481 A @dfn{trampoline} is a small piece of code that is created at run time
4482 when the address of a nested function is taken. It normally resides on
4483 the stack, in the stack frame of the containing function. These macros
4484 tell GCC how to generate code to allocate and initialize a
4487 The instructions in the trampoline must do two things: load a constant
4488 address into the static chain register, and jump to the real address of
4489 the nested function. On CISC machines such as the m68k, this requires
4490 two instructions, a move immediate and a jump. Then the two addresses
4491 exist in the trampoline as word-long immediate operands. On RISC
4492 machines, it is often necessary to load each address into a register in
4493 two parts. Then pieces of each address form separate immediate
4496 The code generated to initialize the trampoline must store the variable
4497 parts---the static chain value and the function address---into the
4498 immediate operands of the instructions. On a CISC machine, this is
4499 simply a matter of copying each address to a memory reference at the
4500 proper offset from the start of the trampoline. On a RISC machine, it
4501 may be necessary to take out pieces of the address and store them
4505 @findex TRAMPOLINE_TEMPLATE
4506 @item TRAMPOLINE_TEMPLATE (@var{file})
4507 A C statement to output, on the stream @var{file}, assembler code for a
4508 block of data that contains the constant parts of a trampoline. This
4509 code should not include a label---the label is taken care of
4512 If you do not define this macro, it means no template is needed
4513 for the target. Do not define this macro on systems where the block move
4514 code to copy the trampoline into place would be larger than the code
4515 to generate it on the spot.
4517 @findex TRAMPOLINE_SECTION
4518 @item TRAMPOLINE_SECTION
4519 The name of a subroutine to switch to the section in which the
4520 trampoline template is to be placed (@pxref{Sections}). The default is
4521 a value of @samp{readonly_data_section}, which places the trampoline in
4522 the section containing read-only data.
4524 @findex TRAMPOLINE_SIZE
4525 @item TRAMPOLINE_SIZE
4526 A C expression for the size in bytes of the trampoline, as an integer.
4528 @findex TRAMPOLINE_ALIGNMENT
4529 @item TRAMPOLINE_ALIGNMENT
4530 Alignment required for trampolines, in bits.
4532 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4533 is used for aligning trampolines.
4535 @findex INITIALIZE_TRAMPOLINE
4536 @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4537 A C statement to initialize the variable parts of a trampoline.
4538 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4539 an RTX for the address of the nested function; @var{static_chain} is an
4540 RTX for the static chain value that should be passed to the function
4543 @findex TRAMPOLINE_ADJUST_ADDRESS
4544 @item TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4545 A C statement that should perform any machine-specific adjustment in
4546 the address of the trampoline. Its argument contains the address that
4547 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4548 used for a function call should be different from the address in which
4549 the template was stored, the different address should be assigned to
4550 @var{addr}. If this macro is not defined, @var{addr} will be used for
4553 @findex ALLOCATE_TRAMPOLINE
4554 @item ALLOCATE_TRAMPOLINE (@var{fp})
4555 A C expression to allocate run-time space for a trampoline. The
4556 expression value should be an RTX representing a memory reference to the
4557 space for the trampoline.
4559 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4560 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4561 If this macro is not defined, by default the trampoline is allocated as
4562 a stack slot. This default is right for most machines. The exceptions
4563 are machines where it is impossible to execute instructions in the stack
4564 area. On such machines, you may have to implement a separate stack,
4565 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4566 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4568 @var{fp} points to a data structure, a @code{struct function}, which
4569 describes the compilation status of the immediate containing function of
4570 the function which the trampoline is for. Normally (when
4571 @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the
4572 trampoline is in the stack frame of this containing function. Other
4573 allocation strategies probably must do something analogous with this
4577 Implementing trampolines is difficult on many machines because they have
4578 separate instruction and data caches. Writing into a stack location
4579 fails to clear the memory in the instruction cache, so when the program
4580 jumps to that location, it executes the old contents.
4582 Here are two possible solutions. One is to clear the relevant parts of
4583 the instruction cache whenever a trampoline is set up. The other is to
4584 make all trampolines identical, by having them jump to a standard
4585 subroutine. The former technique makes trampoline execution faster; the
4586 latter makes initialization faster.
4588 To clear the instruction cache when a trampoline is initialized, define
4589 the following macros which describe the shape of the cache.
4592 @findex INSN_CACHE_SIZE
4593 @item INSN_CACHE_SIZE
4594 The total size in bytes of the cache.
4596 @findex INSN_CACHE_LINE_WIDTH
4597 @item INSN_CACHE_LINE_WIDTH
4598 The length in bytes of each cache line. The cache is divided into cache
4599 lines which are disjoint slots, each holding a contiguous chunk of data
4600 fetched from memory. Each time data is brought into the cache, an
4601 entire line is read at once. The data loaded into a cache line is
4602 always aligned on a boundary equal to the line size.
4604 @findex INSN_CACHE_DEPTH
4605 @item INSN_CACHE_DEPTH
4606 The number of alternative cache lines that can hold any particular memory
4610 Alternatively, if the machine has system calls or instructions to clear
4611 the instruction cache directly, you can define the following macro.
4614 @findex CLEAR_INSN_CACHE
4615 @item CLEAR_INSN_CACHE (@var{beg}, @var{end})
4616 If defined, expands to a C expression clearing the @emph{instruction
4617 cache} in the specified interval. If it is not defined, and the macro
4618 @code{INSN_CACHE_SIZE} is defined, some generic code is generated to clear the
4619 cache. The definition of this macro would typically be a series of
4620 @code{asm} statements. Both @var{beg} and @var{end} are both pointer
4624 To use a standard subroutine, define the following macro. In addition,
4625 you must make sure that the instructions in a trampoline fill an entire
4626 cache line with identical instructions, or else ensure that the
4627 beginning of the trampoline code is always aligned at the same point in
4628 its cache line. Look in @file{m68k.h} as a guide.
4631 @findex TRANSFER_FROM_TRAMPOLINE
4632 @item TRANSFER_FROM_TRAMPOLINE
4633 Define this macro if trampolines need a special subroutine to do their
4634 work. The macro should expand to a series of @code{asm} statements
4635 which will be compiled with GCC@. They go in a library function named
4636 @code{__transfer_from_trampoline}.
4638 If you need to avoid executing the ordinary prologue code of a compiled
4639 C function when you jump to the subroutine, you can do so by placing a
4640 special label of your own in the assembler code. Use one @code{asm}
4641 statement to generate an assembler label, and another to make the label
4642 global. Then trampolines can use that label to jump directly to your
4643 special assembler code.
4647 @section Implicit Calls to Library Routines
4648 @cindex library subroutine names
4649 @cindex @file{libgcc.a}
4651 @c prevent bad page break with this line
4652 Here is an explanation of implicit calls to library routines.
4655 @findex MULSI3_LIBCALL
4656 @item MULSI3_LIBCALL
4657 A C string constant giving the name of the function to call for
4658 multiplication of one signed full-word by another. If you do not
4659 define this macro, the default name is used, which is @code{__mulsi3},
4660 a function defined in @file{libgcc.a}.
4662 @findex DIVSI3_LIBCALL
4663 @item DIVSI3_LIBCALL
4664 A C string constant giving the name of the function to call for
4665 division of one signed full-word by another. If you do not define
4666 this macro, the default name is used, which is @code{__divsi3}, a
4667 function defined in @file{libgcc.a}.
4669 @findex UDIVSI3_LIBCALL
4670 @item UDIVSI3_LIBCALL
4671 A C string constant giving the name of the function to call for
4672 division of one unsigned full-word by another. If you do not define
4673 this macro, the default name is used, which is @code{__udivsi3}, a
4674 function defined in @file{libgcc.a}.
4676 @findex MODSI3_LIBCALL
4677 @item MODSI3_LIBCALL
4678 A C string constant giving the name of the function to call for the
4679 remainder in division of one signed full-word by another. If you do
4680 not define this macro, the default name is used, which is
4681 @code{__modsi3}, a function defined in @file{libgcc.a}.
4683 @findex UMODSI3_LIBCALL
4684 @item UMODSI3_LIBCALL
4685 A C string constant giving the name of the function to call for the
4686 remainder in division of one unsigned full-word by another. If you do
4687 not define this macro, the default name is used, which is
4688 @code{__umodsi3}, a function defined in @file{libgcc.a}.
4690 @findex MULDI3_LIBCALL
4691 @item MULDI3_LIBCALL
4692 A C string constant giving the name of the function to call for
4693 multiplication of one signed double-word by another. If you do not
4694 define this macro, the default name is used, which is @code{__muldi3},
4695 a function defined in @file{libgcc.a}.
4697 @findex DIVDI3_LIBCALL
4698 @item DIVDI3_LIBCALL
4699 A C string constant giving the name of the function to call for
4700 division of one signed double-word by another. If you do not define
4701 this macro, the default name is used, which is @code{__divdi3}, a
4702 function defined in @file{libgcc.a}.
4704 @findex UDIVDI3_LIBCALL
4705 @item UDIVDI3_LIBCALL
4706 A C string constant giving the name of the function to call for
4707 division of one unsigned full-word by another. If you do not define
4708 this macro, the default name is used, which is @code{__udivdi3}, a
4709 function defined in @file{libgcc.a}.
4711 @findex MODDI3_LIBCALL
4712 @item MODDI3_LIBCALL
4713 A C string constant giving the name of the function to call for the
4714 remainder in division of one signed double-word by another. If you do
4715 not define this macro, the default name is used, which is
4716 @code{__moddi3}, a function defined in @file{libgcc.a}.
4718 @findex UMODDI3_LIBCALL
4719 @item UMODDI3_LIBCALL
4720 A C string constant giving the name of the function to call for the
4721 remainder in division of one unsigned full-word by another. If you do
4722 not define this macro, the default name is used, which is
4723 @code{__umoddi3}, a function defined in @file{libgcc.a}.
4725 @findex DECLARE_LIBRARY_RENAMES
4726 @item DECLARE_LIBRARY_RENAMES
4727 This macro, if defined, should expand to a piece of C code that will get
4728 expanded when compiling functions for libgcc.a. It can be used to
4729 provide alternate names for gcc's internal library functions if there
4730 are ABI-mandated names that the compiler should provide.
4732 @findex INIT_TARGET_OPTABS
4733 @item INIT_TARGET_OPTABS
4734 Define this macro as a C statement that declares additional library
4735 routines renames existing ones. @code{init_optabs} calls this macro after
4736 initializing all the normal library routines.
4738 @findex FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4739 @item FLOAT_LIB_COMPARE_RETURNS_BOOL
4740 Define this macro as a C statement that returns nonzero if a call to
4741 the floating point comparison library function will return a boolean
4742 value that indicates the result of the comparison. It should return
4743 zero if one of gcc's own libgcc functions is called.
4745 Most ports don't need to define this macro.
4748 @cindex @code{EDOM}, implicit usage
4750 The value of @code{EDOM} on the target machine, as a C integer constant
4751 expression. If you don't define this macro, GCC does not attempt to
4752 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4753 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4756 If you do not define @code{TARGET_EDOM}, then compiled code reports
4757 domain errors by calling the library function and letting it report the
4758 error. If mathematical functions on your system use @code{matherr} when
4759 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4760 that @code{matherr} is used normally.
4762 @findex GEN_ERRNO_RTX
4763 @cindex @code{errno}, implicit usage
4765 Define this macro as a C expression to create an rtl expression that
4766 refers to the global ``variable'' @code{errno}. (On certain systems,
4767 @code{errno} may not actually be a variable.) If you don't define this
4768 macro, a reasonable default is used.
4770 @findex TARGET_MEM_FUNCTIONS
4771 @cindex @code{bcopy}, implicit usage
4772 @cindex @code{memcpy}, implicit usage
4773 @cindex @code{memmove}, implicit usage
4774 @cindex @code{bzero}, implicit usage
4775 @cindex @code{memset}, implicit usage
4776 @item TARGET_MEM_FUNCTIONS
4777 Define this macro if GCC should generate calls to the ISO C
4778 (and System V) library functions @code{memcpy}, @code{memmove} and
4779 @code{memset} rather than the BSD functions @code{bcopy} and @code{bzero}.
4781 @findex LIBGCC_NEEDS_DOUBLE
4782 @item LIBGCC_NEEDS_DOUBLE
4783 Define this macro if @code{float} arguments cannot be passed to library
4784 routines (so they must be converted to @code{double}). This macro
4785 affects both how library calls are generated and how the library
4786 routines in @file{libgcc.a} accept their arguments. It is useful on
4787 machines where floating and fixed point arguments are passed
4788 differently, such as the i860.
4790 @findex NEXT_OBJC_RUNTIME
4791 @item NEXT_OBJC_RUNTIME
4792 Define this macro to generate code for Objective-C message sending using
4793 the calling convention of the NeXT system. This calling convention
4794 involves passing the object, the selector and the method arguments all
4795 at once to the method-lookup library function.
4797 The default calling convention passes just the object and the selector
4798 to the lookup function, which returns a pointer to the method.
4801 @node Addressing Modes
4802 @section Addressing Modes
4803 @cindex addressing modes
4805 @c prevent bad page break with this line
4806 This is about addressing modes.
4809 @findex HAVE_PRE_INCREMENT
4810 @findex HAVE_PRE_DECREMENT
4811 @findex HAVE_POST_INCREMENT
4812 @findex HAVE_POST_DECREMENT
4813 @item HAVE_PRE_INCREMENT
4814 @itemx HAVE_PRE_DECREMENT
4815 @itemx HAVE_POST_INCREMENT
4816 @itemx HAVE_POST_DECREMENT
4817 A C expression that is nonzero if the machine supports pre-increment,
4818 pre-decrement, post-increment, or post-decrement addressing respectively.
4820 @findex HAVE_POST_MODIFY_DISP
4821 @findex HAVE_PRE_MODIFY_DISP
4822 @item HAVE_PRE_MODIFY_DISP
4823 @itemx HAVE_POST_MODIFY_DISP
4824 A C expression that is nonzero if the machine supports pre- or
4825 post-address side-effect generation involving constants other than
4826 the size of the memory operand.
4828 @findex HAVE_POST_MODIFY_REG
4829 @findex HAVE_PRE_MODIFY_REG
4830 @item HAVE_PRE_MODIFY_REG
4831 @itemx HAVE_POST_MODIFY_REG
4832 A C expression that is nonzero if the machine supports pre- or
4833 post-address side-effect generation involving a register displacement.
4835 @findex CONSTANT_ADDRESS_P
4836 @item CONSTANT_ADDRESS_P (@var{x})
4837 A C expression that is 1 if the RTX @var{x} is a constant which
4838 is a valid address. On most machines, this can be defined as
4839 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4840 in which constant addresses are supported.
4843 @code{CONSTANT_P} accepts integer-values expressions whose values are
4844 not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and
4845 @code{high} expressions and @code{const} arithmetic expressions, in
4846 addition to @code{const_int} and @code{const_double} expressions.
4848 @findex MAX_REGS_PER_ADDRESS
4849 @item MAX_REGS_PER_ADDRESS
4850 A number, the maximum number of registers that can appear in a valid
4851 memory address. Note that it is up to you to specify a value equal to
4852 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4855 @findex GO_IF_LEGITIMATE_ADDRESS
4856 @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4857 A C compound statement with a conditional @code{goto @var{label};}
4858 executed if @var{x} (an RTX) is a legitimate memory address on the
4859 target machine for a memory operand of mode @var{mode}.
4861 It usually pays to define several simpler macros to serve as
4862 subroutines for this one. Otherwise it may be too complicated to
4865 This macro must exist in two variants: a strict variant and a
4866 non-strict one. The strict variant is used in the reload pass. It
4867 must be defined so that any pseudo-register that has not been
4868 allocated a hard register is considered a memory reference. In
4869 contexts where some kind of register is required, a pseudo-register
4870 with no hard register must be rejected.
4872 The non-strict variant is used in other passes. It must be defined to
4873 accept all pseudo-registers in every context where some kind of
4874 register is required.
4876 @findex REG_OK_STRICT
4877 Compiler source files that want to use the strict variant of this
4878 macro define the macro @code{REG_OK_STRICT}. You should use an
4879 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4880 in that case and the non-strict variant otherwise.
4882 Subroutines to check for acceptable registers for various purposes (one
4883 for base registers, one for index registers, and so on) are typically
4884 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4885 Then only these subroutine macros need have two variants; the higher
4886 levels of macros may be the same whether strict or not.
4888 Normally, constant addresses which are the sum of a @code{symbol_ref}
4889 and an integer are stored inside a @code{const} RTX to mark them as
4890 constant. Therefore, there is no need to recognize such sums
4891 specifically as legitimate addresses. Normally you would simply
4892 recognize any @code{const} as legitimate.
4894 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4895 sums that are not marked with @code{const}. It assumes that a naked
4896 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4897 naked constant sums as illegitimate addresses, so that none of them will
4898 be given to @code{PRINT_OPERAND_ADDRESS}.
4900 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
4901 On some machines, whether a symbolic address is legitimate depends on
4902 the section that the address refers to. On these machines, define the
4903 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
4904 into the @code{symbol_ref}, and then check for it here. When you see a
4905 @code{const}, you will have to look inside it to find the
4906 @code{symbol_ref} in order to determine the section. @xref{Assembler
4909 @findex saveable_obstack
4910 The best way to modify the name string is by adding text to the
4911 beginning, with suitable punctuation to prevent any ambiguity. Allocate
4912 the new name in @code{saveable_obstack}. You will have to modify
4913 @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and
4914 output the name accordingly, and define @code{TARGET_STRIP_NAME_ENCODING}
4915 to access the original name string.
4917 You can check the information stored here into the @code{symbol_ref} in
4918 the definitions of the macros @code{GO_IF_LEGITIMATE_ADDRESS} and
4919 @code{PRINT_OPERAND_ADDRESS}.
4921 @findex REG_OK_FOR_BASE_P
4922 @item REG_OK_FOR_BASE_P (@var{x})
4923 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4924 RTX) is valid for use as a base register. For hard registers, it
4925 should always accept those which the hardware permits and reject the
4926 others. Whether the macro accepts or rejects pseudo registers must be
4927 controlled by @code{REG_OK_STRICT} as described above. This usually
4928 requires two variant definitions, of which @code{REG_OK_STRICT}
4929 controls the one actually used.
4931 @findex REG_MODE_OK_FOR_BASE_P
4932 @item REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4933 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4934 that expression may examine the mode of the memory reference in
4935 @var{mode}. You should define this macro if the mode of the memory
4936 reference affects whether a register may be used as a base register. If
4937 you define this macro, the compiler will use it instead of
4938 @code{REG_OK_FOR_BASE_P}.
4940 @findex REG_OK_FOR_INDEX_P
4941 @item REG_OK_FOR_INDEX_P (@var{x})
4942 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4943 RTX) is valid for use as an index register.
4945 The difference between an index register and a base register is that
4946 the index register may be scaled. If an address involves the sum of
4947 two registers, neither one of them scaled, then either one may be
4948 labeled the ``base'' and the other the ``index''; but whichever
4949 labeling is used must fit the machine's constraints of which registers
4950 may serve in each capacity. The compiler will try both labelings,
4951 looking for one that is valid, and will reload one or both registers
4952 only if neither labeling works.
4954 @findex FIND_BASE_TERM
4955 @item FIND_BASE_TERM (@var{x})
4956 A C expression to determine the base term of address @var{x}.
4957 This macro is used in only one place: `find_base_term' in alias.c.
4959 It is always safe for this macro to not be defined. It exists so
4960 that alias analysis can understand machine-dependent addresses.
4962 The typical use of this macro is to handle addresses containing
4963 a label_ref or symbol_ref within an UNSPEC@.
4965 @findex LEGITIMIZE_ADDRESS
4966 @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4967 A C compound statement that attempts to replace @var{x} with a valid
4968 memory address for an operand of mode @var{mode}. @var{win} will be a
4969 C statement label elsewhere in the code; the macro definition may use
4972 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4976 to avoid further processing if the address has become legitimate.
4978 @findex break_out_memory_refs
4979 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4980 and @var{oldx} will be the operand that was given to that function to produce
4983 The code generated by this macro should not alter the substructure of
4984 @var{x}. If it transforms @var{x} into a more legitimate form, it
4985 should assign @var{x} (which will always be a C variable) a new value.
4987 It is not necessary for this macro to come up with a legitimate
4988 address. The compiler has standard ways of doing so in all cases. In
4989 fact, it is safe for this macro to do nothing. But often a
4990 machine-dependent strategy can generate better code.
4992 @findex LEGITIMIZE_RELOAD_ADDRESS
4993 @item LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4994 A C compound statement that attempts to replace @var{x}, which is an address
4995 that needs reloading, with a valid memory address for an operand of mode
4996 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4997 It is not necessary to define this macro, but it might be useful for
4998 performance reasons.
5000 For example, on the i386, it is sometimes possible to use a single
5001 reload register instead of two by reloading a sum of two pseudo
5002 registers into a register. On the other hand, for number of RISC
5003 processors offsets are limited so that often an intermediate address
5004 needs to be generated in order to address a stack slot. By defining
5005 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5006 generated for adjacent some stack slots can be made identical, and thus
5009 @emph{Note}: This macro should be used with caution. It is necessary
5010 to know something of how reload works in order to effectively use this,
5011 and it is quite easy to produce macros that build in too much knowledge
5012 of reload internals.
5014 @emph{Note}: This macro must be able to reload an address created by a
5015 previous invocation of this macro. If it fails to handle such addresses
5016 then the compiler may generate incorrect code or abort.
5019 The macro definition should use @code{push_reload} to indicate parts that
5020 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5021 suitable to be passed unaltered to @code{push_reload}.
5023 The code generated by this macro must not alter the substructure of
5024 @var{x}. If it transforms @var{x} into a more legitimate form, it
5025 should assign @var{x} (which will always be a C variable) a new value.
5026 This also applies to parts that you change indirectly by calling
5029 @findex strict_memory_address_p
5030 The macro definition may use @code{strict_memory_address_p} to test if
5031 the address has become legitimate.
5034 If you want to change only a part of @var{x}, one standard way of doing
5035 this is to use @code{copy_rtx}. Note, however, that is unshares only a
5036 single level of rtl. Thus, if the part to be changed is not at the
5037 top level, you'll need to replace first the top level.
5038 It is not necessary for this macro to come up with a legitimate
5039 address; but often a machine-dependent strategy can generate better code.
5041 @findex GO_IF_MODE_DEPENDENT_ADDRESS
5042 @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5043 A C statement or compound statement with a conditional @code{goto
5044 @var{label};} executed if memory address @var{x} (an RTX) can have
5045 different meanings depending on the machine mode of the memory
5046 reference it is used for or if the address is valid for some modes
5049 Autoincrement and autodecrement addresses typically have mode-dependent
5050 effects because the amount of the increment or decrement is the size
5051 of the operand being addressed. Some machines have other mode-dependent
5052 addresses. Many RISC machines have no mode-dependent addresses.
5054 You may assume that @var{addr} is a valid address for the machine.
5056 @findex LEGITIMATE_CONSTANT_P
5057 @item LEGITIMATE_CONSTANT_P (@var{x})
5058 A C expression that is nonzero if @var{x} is a legitimate constant for
5059 an immediate operand on the target machine. You can assume that
5060 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5061 @samp{1} is a suitable definition for this macro on machines where
5062 anything @code{CONSTANT_P} is valid.
5065 @node Condition Code
5066 @section Condition Code Status
5067 @cindex condition code status
5069 @c prevent bad page break with this line
5070 This describes the condition code status.
5073 The file @file{conditions.h} defines a variable @code{cc_status} to
5074 describe how the condition code was computed (in case the interpretation of
5075 the condition code depends on the instruction that it was set by). This
5076 variable contains the RTL expressions on which the condition code is
5077 currently based, and several standard flags.
5079 Sometimes additional machine-specific flags must be defined in the machine
5080 description header file. It can also add additional machine-specific
5081 information by defining @code{CC_STATUS_MDEP}.
5084 @findex CC_STATUS_MDEP
5085 @item CC_STATUS_MDEP
5086 C code for a data type which is used for declaring the @code{mdep}
5087 component of @code{cc_status}. It defaults to @code{int}.
5089 This macro is not used on machines that do not use @code{cc0}.
5091 @findex CC_STATUS_MDEP_INIT
5092 @item CC_STATUS_MDEP_INIT
5093 A C expression to initialize the @code{mdep} field to ``empty''.
5094 The default definition does nothing, since most machines don't use
5095 the field anyway. If you want to use the field, you should probably
5096 define this macro to initialize it.
5098 This macro is not used on machines that do not use @code{cc0}.
5100 @findex NOTICE_UPDATE_CC
5101 @item NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5102 A C compound statement to set the components of @code{cc_status}
5103 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5104 this macro's responsibility to recognize insns that set the condition
5105 code as a byproduct of other activity as well as those that explicitly
5108 This macro is not used on machines that do not use @code{cc0}.
5110 If there are insns that do not set the condition code but do alter
5111 other machine registers, this macro must check to see whether they
5112 invalidate the expressions that the condition code is recorded as
5113 reflecting. For example, on the 68000, insns that store in address
5114 registers do not set the condition code, which means that usually
5115 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5116 insns. But suppose that the previous insn set the condition code
5117 based on location @samp{a4@@(102)} and the current insn stores a new
5118 value in @samp{a4}. Although the condition code is not changed by
5119 this, it will no longer be true that it reflects the contents of
5120 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5121 @code{cc_status} in this case to say that nothing is known about the
5122 condition code value.
5124 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5125 with the results of peephole optimization: insns whose patterns are
5126 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5127 constants which are just the operands. The RTL structure of these
5128 insns is not sufficient to indicate what the insns actually do. What
5129 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5130 @code{CC_STATUS_INIT}.
5132 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5133 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5134 @samp{cc}. This avoids having detailed information about patterns in
5135 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5137 @findex EXTRA_CC_MODES
5138 @item EXTRA_CC_MODES
5139 Condition codes are represented in registers by machine modes of class
5140 @code{MODE_CC}. By default, there is just one mode, @code{CCmode}, with
5141 this class. If you need more such modes, create a file named
5142 @file{@var{machine}-modes.def} in your @file{config/@var{machine}}
5143 directory (@pxref{Back End, , Anatomy of a Target Back End}), containing
5144 a list of these modes. Each entry in the list should be a call to the
5145 macro @code{CC}. This macro takes one argument, which is the name of
5146 the mode: it should begin with @samp{CC}. Do not put quotation marks
5147 around the name, or include the trailing @samp{mode}; these are
5148 automatically added. There should not be anything else in the file
5151 A sample @file{@var{machine}-modes.def} file might look like this:
5154 CC (CC_NOOV) /* @r{Comparison only valid if there was no overflow.} */
5155 CC (CCFP) /* @r{Floating point comparison that cannot trap.} */
5156 CC (CCFPE) /* @r{Floating point comparison that may trap.} */
5159 When you create this file, the macro @code{EXTRA_CC_MODES} is
5160 automatically defined by @command{configure}, with value @samp{1}.
5162 @findex SELECT_CC_MODE
5163 @item SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5164 Returns a mode from class @code{MODE_CC} to be used when comparison
5165 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5166 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5167 @pxref{Jump Patterns} for a description of the reason for this
5171 #define SELECT_CC_MODE(OP,X,Y) \
5172 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5173 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5174 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5175 || GET_CODE (X) == NEG) \
5176 ? CC_NOOVmode : CCmode))
5179 You need not define this macro if @code{EXTRA_CC_MODES} is not defined.
5181 @findex CANONICALIZE_COMPARISON
5182 @item CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5183 On some machines not all possible comparisons are defined, but you can
5184 convert an invalid comparison into a valid one. For example, the Alpha
5185 does not have a @code{GT} comparison, but you can use an @code{LT}
5186 comparison instead and swap the order of the operands.
5188 On such machines, define this macro to be a C statement to do any
5189 required conversions. @var{code} is the initial comparison code
5190 and @var{op0} and @var{op1} are the left and right operands of the
5191 comparison, respectively. You should modify @var{code}, @var{op0}, and
5192 @var{op1} as required.
5194 GCC will not assume that the comparison resulting from this macro is
5195 valid but will see if the resulting insn matches a pattern in the
5198 You need not define this macro if it would never change the comparison
5201 @findex REVERSIBLE_CC_MODE
5202 @item REVERSIBLE_CC_MODE (@var{mode})
5203 A C expression whose value is one if it is always safe to reverse a
5204 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5205 can ever return @var{mode} for a floating-point inequality comparison,
5206 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5208 You need not define this macro if it would always returns zero or if the
5209 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5210 For example, here is the definition used on the SPARC, where floating-point
5211 inequality comparisons are always given @code{CCFPEmode}:
5214 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5217 @findex REVERSE_CONDITION (@var{code}, @var{mode})
5218 A C expression whose value is reversed condition code of the @var{code} for
5219 comparison done in CC_MODE @var{mode}. The macro is used only in case
5220 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5221 machine has some non-standard way how to reverse certain conditionals. For
5222 instance in case all floating point conditions are non-trapping, compiler may
5223 freely convert unordered compares to ordered one. Then definition may look
5227 #define REVERSE_CONDITION(CODE, MODE) \
5228 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5229 : reverse_condition_maybe_unordered (CODE))
5232 @findex REVERSE_CONDEXEC_PREDICATES_P
5233 @item REVERSE_CONDEXEC_PREDICATES_P (@var{code1}, @var{code2})
5234 A C expression that returns true if the conditional execution predicate
5235 @var{code1} is the inverse of @var{code2} and vice versa. Define this to
5236 return 0 if the target has conditional execution predicates that cannot be
5237 reversed safely. If no expansion is specified, this macro is defined as
5241 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5242 ((x) == reverse_condition (y))
5248 @section Describing Relative Costs of Operations
5249 @cindex costs of instructions
5250 @cindex relative costs
5251 @cindex speed of instructions
5253 These macros let you describe the relative speed of various operations
5254 on the target machine.
5258 @item CONST_COSTS (@var{x}, @var{code}, @var{outer_code})
5259 A part of a C @code{switch} statement that describes the relative costs
5260 of constant RTL expressions. It must contain @code{case} labels for
5261 expression codes @code{const_int}, @code{const}, @code{symbol_ref},
5262 @code{label_ref} and @code{const_double}. Each case must ultimately
5263 reach a @code{return} statement to return the relative cost of the use
5264 of that kind of constant value in an expression. The cost may depend on
5265 the precise value of the constant, which is available for examination in
5266 @var{x}, and the rtx code of the expression in which it is contained,
5267 found in @var{outer_code}.
5269 @var{code} is the expression code---redundant, since it can be
5270 obtained with @code{GET_CODE (@var{x})}.
5273 @findex COSTS_N_INSNS
5274 @item RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
5275 Like @code{CONST_COSTS} but applies to nonconstant RTL expressions.
5276 This can be used, for example, to indicate how costly a multiply
5277 instruction is. In writing this macro, you can use the construct
5278 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
5279 instructions. @var{outer_code} is the code of the expression in which
5280 @var{x} is contained.
5282 This macro is optional; do not define it if the default cost assumptions
5283 are adequate for the target machine.
5285 @findex DEFAULT_RTX_COSTS
5286 @item DEFAULT_RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
5287 This macro, if defined, is called for any case not handled by the
5288 @code{RTX_COSTS} or @code{CONST_COSTS} macros. This eliminates the need
5289 to put case labels into the macro, but the code, or any functions it
5290 calls, must assume that the RTL in @var{x} could be of any type that has
5291 not already been handled. The arguments are the same as for
5292 @code{RTX_COSTS}, and the macro should execute a return statement giving
5293 the cost of any RTL expressions that it can handle. The default cost
5294 calculation is used for any RTL for which this macro does not return a
5297 This macro is optional; do not define it if the default cost assumptions
5298 are adequate for the target machine.
5300 @findex ADDRESS_COST
5301 @item ADDRESS_COST (@var{address})
5302 An expression giving the cost of an addressing mode that contains
5303 @var{address}. If not defined, the cost is computed from
5304 the @var{address} expression and the @code{CONST_COSTS} values.
5306 For most CISC machines, the default cost is a good approximation of the
5307 true cost of the addressing mode. However, on RISC machines, all
5308 instructions normally have the same length and execution time. Hence
5309 all addresses will have equal costs.
5311 In cases where more than one form of an address is known, the form with
5312 the lowest cost will be used. If multiple forms have the same, lowest,
5313 cost, the one that is the most complex will be used.
5315 For example, suppose an address that is equal to the sum of a register
5316 and a constant is used twice in the same basic block. When this macro
5317 is not defined, the address will be computed in a register and memory
5318 references will be indirect through that register. On machines where
5319 the cost of the addressing mode containing the sum is no higher than
5320 that of a simple indirect reference, this will produce an additional
5321 instruction and possibly require an additional register. Proper
5322 specification of this macro eliminates this overhead for such machines.
5324 Similar use of this macro is made in strength reduction of loops.
5326 @var{address} need not be valid as an address. In such a case, the cost
5327 is not relevant and can be any value; invalid addresses need not be
5328 assigned a different cost.
5330 On machines where an address involving more than one register is as
5331 cheap as an address computation involving only one register, defining
5332 @code{ADDRESS_COST} to reflect this can cause two registers to be live
5333 over a region of code where only one would have been if
5334 @code{ADDRESS_COST} were not defined in that manner. This effect should
5335 be considered in the definition of this macro. Equivalent costs should
5336 probably only be given to addresses with different numbers of registers
5337 on machines with lots of registers.
5339 This macro will normally either not be defined or be defined as a
5342 @findex REGISTER_MOVE_COST
5343 @item REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5344 A C expression for the cost of moving data of mode @var{mode} from a
5345 register in class @var{from} to one in class @var{to}. The classes are
5346 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5347 value of 2 is the default; other values are interpreted relative to
5350 It is not required that the cost always equal 2 when @var{from} is the
5351 same as @var{to}; on some machines it is expensive to move between
5352 registers if they are not general registers.
5354 If reload sees an insn consisting of a single @code{set} between two
5355 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5356 classes returns a value of 2, reload does not check to ensure that the
5357 constraints of the insn are met. Setting a cost of other than 2 will
5358 allow reload to verify that the constraints are met. You should do this
5359 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5361 @findex MEMORY_MOVE_COST
5362 @item MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5363 A C expression for the cost of moving data of mode @var{mode} between a
5364 register of class @var{class} and memory; @var{in} is zero if the value
5365 is to be written to memory, nonzero if it is to be read in. This cost
5366 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5367 registers and memory is more expensive than between two registers, you
5368 should define this macro to express the relative cost.
5370 If you do not define this macro, GCC uses a default cost of 4 plus
5371 the cost of copying via a secondary reload register, if one is
5372 needed. If your machine requires a secondary reload register to copy
5373 between memory and a register of @var{class} but the reload mechanism is
5374 more complex than copying via an intermediate, define this macro to
5375 reflect the actual cost of the move.
5377 GCC defines the function @code{memory_move_secondary_cost} if
5378 secondary reloads are needed. It computes the costs due to copying via
5379 a secondary register. If your machine copies from memory using a
5380 secondary register in the conventional way but the default base value of
5381 4 is not correct for your machine, define this macro to add some other
5382 value to the result of that function. The arguments to that function
5383 are the same as to this macro.
5387 A C expression for the cost of a branch instruction. A value of 1 is
5388 the default; other values are interpreted relative to that.
5391 Here are additional macros which do not specify precise relative costs,
5392 but only that certain actions are more expensive than GCC would
5396 @findex SLOW_BYTE_ACCESS
5397 @item SLOW_BYTE_ACCESS
5398 Define this macro as a C expression which is nonzero if accessing less
5399 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5400 faster than accessing a word of memory, i.e., if such access
5401 require more than one instruction or if there is no difference in cost
5402 between byte and (aligned) word loads.
5404 When this macro is not defined, the compiler will access a field by
5405 finding the smallest containing object; when it is defined, a fullword
5406 load will be used if alignment permits. Unless bytes accesses are
5407 faster than word accesses, using word accesses is preferable since it
5408 may eliminate subsequent memory access if subsequent accesses occur to
5409 other fields in the same word of the structure, but to different bytes.
5411 @findex SLOW_UNALIGNED_ACCESS
5412 @item SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5413 Define this macro to be the value 1 if memory accesses described by the
5414 @var{mode} and @var{alignment} parameters have a cost many times greater
5415 than aligned accesses, for example if they are emulated in a trap
5418 When this macro is nonzero, the compiler will act as if
5419 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5420 moves. This can cause significantly more instructions to be produced.
5421 Therefore, do not set this macro nonzero if unaligned accesses only add a
5422 cycle or two to the time for a memory access.
5424 If the value of this macro is always zero, it need not be defined. If
5425 this macro is defined, it should produce a nonzero value when
5426 @code{STRICT_ALIGNMENT} is nonzero.
5428 @findex DONT_REDUCE_ADDR
5429 @item DONT_REDUCE_ADDR
5430 Define this macro to inhibit strength reduction of memory addresses.
5431 (On some machines, such strength reduction seems to do harm rather
5436 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5437 which a sequence of insns should be generated instead of a
5438 string move insn or a library call. Increasing the value will always
5439 make code faster, but eventually incurs high cost in increased code size.
5441 Note that on machines where the corresponding move insn is a
5442 @code{define_expand} that emits a sequence of insns, this macro counts
5443 the number of such sequences.
5445 If you don't define this, a reasonable default is used.
5447 @findex MOVE_BY_PIECES_P
5448 @item MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5449 A C expression used to determine whether @code{move_by_pieces} will be used to
5450 copy a chunk of memory, or whether some other block move mechanism
5451 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5452 than @code{MOVE_RATIO}.
5454 @findex MOVE_MAX_PIECES
5455 @item MOVE_MAX_PIECES
5456 A C expression used by @code{move_by_pieces} to determine the largest unit
5457 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5461 The threshold of number of scalar move insns, @emph{below} which a sequence
5462 of insns should be generated to clear memory instead of a string clear insn
5463 or a library call. Increasing the value will always make code faster, but
5464 eventually incurs high cost in increased code size.
5466 If you don't define this, a reasonable default is used.
5468 @findex CLEAR_BY_PIECES_P
5469 @item CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5470 A C expression used to determine whether @code{clear_by_pieces} will be used
5471 to clear a chunk of memory, or whether some other block clear mechanism
5472 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5473 than @code{CLEAR_RATIO}.
5475 @findex STORE_BY_PIECES_P
5476 @item STORE_BY_PIECES_P (@var{size}, @var{alignment})
5477 A C expression used to determine whether @code{store_by_pieces} will be
5478 used to set a chunk of memory to a constant value, or whether some other
5479 mechanism will be used. Used by @code{__builtin_memset} when storing
5480 values other than constant zero and by @code{__builtin_strcpy} when
5481 when called with a constant source string.
5482 Defaults to @code{MOVE_BY_PIECES_P}.
5484 @findex USE_LOAD_POST_INCREMENT
5485 @item USE_LOAD_POST_INCREMENT (@var{mode})
5486 A C expression used to determine whether a load postincrement is a good
5487 thing to use for a given mode. Defaults to the value of
5488 @code{HAVE_POST_INCREMENT}.
5490 @findex USE_LOAD_POST_DECREMENT
5491 @item USE_LOAD_POST_DECREMENT (@var{mode})
5492 A C expression used to determine whether a load postdecrement is a good
5493 thing to use for a given mode. Defaults to the value of
5494 @code{HAVE_POST_DECREMENT}.
5496 @findex USE_LOAD_PRE_INCREMENT
5497 @item USE_LOAD_PRE_INCREMENT (@var{mode})
5498 A C expression used to determine whether a load preincrement is a good
5499 thing to use for a given mode. Defaults to the value of
5500 @code{HAVE_PRE_INCREMENT}.
5502 @findex USE_LOAD_PRE_DECREMENT
5503 @item USE_LOAD_PRE_DECREMENT (@var{mode})
5504 A C expression used to determine whether a load predecrement is a good
5505 thing to use for a given mode. Defaults to the value of
5506 @code{HAVE_PRE_DECREMENT}.
5508 @findex USE_STORE_POST_INCREMENT
5509 @item USE_STORE_POST_INCREMENT (@var{mode})
5510 A C expression used to determine whether a store postincrement is a good
5511 thing to use for a given mode. Defaults to the value of
5512 @code{HAVE_POST_INCREMENT}.
5514 @findex USE_STORE_POST_DECREMENT
5515 @item USE_STORE_POST_DECREMENT (@var{mode})
5516 A C expression used to determine whether a store postdecrement is a good
5517 thing to use for a given mode. Defaults to the value of
5518 @code{HAVE_POST_DECREMENT}.
5520 @findex USE_STORE_PRE_INCREMENT
5521 @item USE_STORE_PRE_INCREMENT (@var{mode})
5522 This macro is used to determine whether a store preincrement is a good
5523 thing to use for a given mode. Defaults to the value of
5524 @code{HAVE_PRE_INCREMENT}.
5526 @findex USE_STORE_PRE_DECREMENT
5527 @item USE_STORE_PRE_DECREMENT (@var{mode})
5528 This macro is used to determine whether a store predecrement is a good
5529 thing to use for a given mode. Defaults to the value of
5530 @code{HAVE_PRE_DECREMENT}.
5532 @findex NO_FUNCTION_CSE
5533 @item NO_FUNCTION_CSE
5534 Define this macro if it is as good or better to call a constant
5535 function address than to call an address kept in a register.
5537 @findex NO_RECURSIVE_FUNCTION_CSE
5538 @item NO_RECURSIVE_FUNCTION_CSE
5539 Define this macro if it is as good or better for a function to call
5540 itself with an explicit address than to call an address kept in a
5545 @section Adjusting the Instruction Scheduler
5547 The instruction scheduler may need a fair amount of machine-specific
5548 adjustment in order to produce good code. GCC provides several target
5549 hooks for this purpose. It is usually enough to define just a few of
5550 them: try the first ones in this list first.
5552 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
5553 This hook returns the maximum number of instructions that can ever
5554 issue at the same time on the target machine. The default is one.
5555 Although the insn scheduler can define itself the possibility of issue
5556 an insn on the same cycle, the value can serve as an additional
5557 constraint to issue insns on the same simulated processor cycle (see
5558 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
5559 This value must be constant over the entire compilation. If you need
5560 it to vary depending on what the instructions are, you must use
5561 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
5563 For the automaton based pipeline interface, you could define this hook
5564 to return the value of the macro @code{MAX_DFA_ISSUE_RATE}.
5567 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
5568 This hook is executed by the scheduler after it has scheduled an insn
5569 from the ready list. It should return the number of insns which can
5570 still be issued in the current cycle. The default is
5571 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
5572 @code{USE}, which normally are not counted against the issue rate.
5573 You should define this hook if some insns take more machine resources
5574 than others, so that fewer insns can follow them in the same cycle.
5575 @var{file} is either a null pointer, or a stdio stream to write any
5576 debug output to. @var{verbose} is the verbose level provided by
5577 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
5581 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
5582 This function corrects the value of @var{cost} based on the
5583 relationship between @var{insn} and @var{dep_insn} through the
5584 dependence @var{link}. It should return the new value. The default
5585 is to make no adjustment to @var{cost}. This can be used for example
5586 to specify to the scheduler using the traditional pipeline description
5587 that an output- or anti-dependence does not incur the same cost as a
5588 data-dependence. If the scheduler using the automaton based pipeline
5589 description, the cost of anti-dependence is zero and the cost of
5590 output-dependence is maximum of one and the difference of latency
5591 times of the first and the second insns. If these values are not
5592 acceptable, you could use the hook to modify them too. See also
5593 @pxref{Automaton pipeline description}.
5596 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
5597 This hook adjusts the integer scheduling priority @var{priority} of
5598 @var{insn}. It should return the new priority. Reduce the priority to
5599 execute @var{insn} earlier, increase the priority to execute @var{insn}
5600 later. Do not define this hook if you do not need to adjust the
5601 scheduling priorities of insns.
5604 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
5605 This hook is executed by the scheduler after it has scheduled the ready
5606 list, to allow the machine description to reorder it (for example to
5607 combine two small instructions together on @samp{VLIW} machines).
5608 @var{file} is either a null pointer, or a stdio stream to write any
5609 debug output to. @var{verbose} is the verbose level provided by
5610 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
5611 list of instructions that are ready to be scheduled. @var{n_readyp} is
5612 a pointer to the number of elements in the ready list. The scheduler
5613 reads the ready list in reverse order, starting with
5614 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
5615 is the timer tick of the scheduler. You may modify the ready list and
5616 the number of ready insns. The return value is the number of insns that
5617 can issue this cycle; normally this is just @code{issue_rate}. See also
5618 @samp{TARGET_SCHED_REORDER2}.
5621 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
5622 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
5623 function is called whenever the scheduler starts a new cycle. This one
5624 is called once per iteration over a cycle, immediately after
5625 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
5626 return the number of insns to be scheduled in the same cycle. Defining
5627 this hook can be useful if there are frequent situations where
5628 scheduling one insn causes other insns to become ready in the same
5629 cycle. These other insns can then be taken into account properly.
5632 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
5633 This hook is called after evaluation forward dependencies of insns in
5634 chain given by two parameter values (@var{head} and @var{tail}
5635 correspondingly) but before insns scheduling of the insn chain. For
5636 example, it can be used for better insn classification if it requires
5637 analysis of dependencies. This hook can use backward and forward
5638 dependencies of the insn scheduler because they are already
5642 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
5643 This hook is executed by the scheduler at the beginning of each block of
5644 instructions that are to be scheduled. @var{file} is either a null
5645 pointer, or a stdio stream to write any debug output to. @var{verbose}
5646 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5647 @var{max_ready} is the maximum number of insns in the current scheduling
5648 region that can be live at the same time. This can be used to allocate
5649 scratch space if it is needed, e.g. by @samp{TARGET_SCHED_REORDER}.
5652 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
5653 This hook is executed by the scheduler at the end of each block of
5654 instructions that are to be scheduled. It can be used to perform
5655 cleanup of any actions done by the other scheduling hooks. @var{file}
5656 is either a null pointer, or a stdio stream to write any debug output
5657 to. @var{verbose} is the verbose level provided by
5658 @option{-fsched-verbose-@var{n}}.
5661 @deftypefn {Target Hook} int TARGET_SCHED_USE_DFA_PIPELINE_INTERFACE (void)
5662 This hook is called many times during insn scheduling. If the hook
5663 returns nonzero, the automaton based pipeline description is used for
5664 insn scheduling. Otherwise the traditional pipeline description is
5665 used. The default is usage of the traditional pipeline description.
5667 You should also remember that to simplify the insn scheduler sources
5668 an empty traditional pipeline description interface is generated even
5669 if there is no a traditional pipeline description in the @file{.md}
5670 file. The same is true for the automaton based pipeline description.
5671 That means that you should be accurate in defining the hook.
5674 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
5675 The hook returns an RTL insn. The automaton state used in the
5676 pipeline hazard recognizer is changed as if the insn were scheduled
5677 when the new simulated processor cycle starts. Usage of the hook may
5678 simplify the automaton pipeline description for some @acronym{VLIW}
5679 processors. If the hook is defined, it is used only for the automaton
5680 based pipeline description. The default is not to change the state
5681 when the new simulated processor cycle starts.
5684 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
5685 The hook can be used to initialize data used by the previous hook.
5688 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
5689 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
5690 to changed the state as if the insn were scheduled when the new
5691 simulated processor cycle finishes.
5694 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
5695 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
5696 used to initialize data used by the previous hook.
5699 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
5700 This hook controls better choosing an insn from the ready insn queue
5701 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
5702 chooses the first insn from the queue. If the hook returns a positive
5703 value, an additional scheduler code tries all permutations of
5704 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
5705 subsequent ready insns to choose an insn whose issue will result in
5706 maximal number of issued insns on the same cycle. For the
5707 @acronym{VLIW} processor, the code could actually solve the problem of
5708 packing simple insns into the @acronym{VLIW} insn. Of course, if the
5709 rules of @acronym{VLIW} packing are described in the automaton.
5711 This code also could be used for superscalar @acronym{RISC}
5712 processors. Let us consider a superscalar @acronym{RISC} processor
5713 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
5714 @var{B}, some insns can be executed only in pipelines @var{B} or
5715 @var{C}, and one insn can be executed in pipeline @var{B}. The
5716 processor may issue the 1st insn into @var{A} and the 2nd one into
5717 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
5718 until the next cycle. If the scheduler issues the 3rd insn the first,
5719 the processor could issue all 3 insns per cycle.
5721 Actually this code demonstrates advantages of the automaton based
5722 pipeline hazard recognizer. We try quickly and easy many insn
5723 schedules to choose the best one.
5725 The default is no multipass scheduling.
5728 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
5730 This hook controls what insns from the ready insn queue will be
5731 considered for the multipass insn scheduling. If the hook returns
5732 zero for insn passed as the parameter, the insn will be not chosen to
5735 The default is that any ready insns can be choosen to be issued.
5738 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
5740 This hook is called by the insn scheduler before issuing insn passed
5741 as the third parameter on given cycle. If the hook returns nonzero,
5742 the insn is not issued on given processors cycle. Instead of that,
5743 the processor cycle is advanced. If the value passed through the last
5744 parameter is zero, the insn ready queue is not sorted on the new cycle
5745 start as usually. The first parameter passes file for debugging
5746 output. The second one passes the scheduler verbose level of the
5747 debugging output. The forth and the fifth parameter values are
5748 correspondingly processor cycle on which the previous insn has been
5749 issued and the current processor cycle.
5752 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_BUBBLES (void)
5753 The @acronym{DFA}-based scheduler could take the insertion of nop
5754 operations for better insn scheduling into account. It can be done
5755 only if the multi-pass insn scheduling works (see hook
5756 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD}).
5758 Let us consider a @acronym{VLIW} processor insn with 3 slots. Each
5759 insn can be placed only in one of the three slots. We have 3 ready
5760 insns @var{A}, @var{B}, and @var{C}. @var{A} and @var{C} can be
5761 placed only in the 1st slot, @var{B} can be placed only in the 3rd
5762 slot. We described the automaton which does not permit empty slot
5763 gaps between insns (usually such description is simpler). Without
5764 this code the scheduler would place each insn in 3 separate
5765 @acronym{VLIW} insns. If the scheduler places a nop insn into the 2nd
5766 slot, it could place the 3 insns into 2 @acronym{VLIW} insns. What is
5767 the nop insn is returned by hook @samp{TARGET_SCHED_DFA_BUBBLE}. Hook
5768 @samp{TARGET_SCHED_INIT_DFA_BUBBLES} can be used to initialize or
5769 create the nop insns.
5771 You should remember that the scheduler does not insert the nop insns.
5772 It is not wise because of the following optimizations. The scheduler
5773 only considers such possibility to improve the result schedule. The
5774 nop insns should be inserted lately, e.g. on the final phase.
5777 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_BUBBLE (int @var{index})
5778 This hook @samp{FIRST_CYCLE_MULTIPASS_SCHEDULING} is used to insert
5779 nop operations for better insn scheduling when @acronym{DFA}-based
5780 scheduler makes multipass insn scheduling (see also description of
5781 hook @samp{TARGET_SCHED_INIT_DFA_BUBBLES}). This hook
5782 returns a nop insn with given @var{index}. The indexes start with
5783 zero. The hook should return @code{NULL} if there are no more nop
5784 insns with indexes greater than given index.
5787 Macros in the following table are generated by the program
5788 @file{genattr} and can be useful for writing the hooks.
5791 @findex TRADITIONAL_PIPELINE_INTERFACE
5792 @item TRADITIONAL_PIPELINE_INTERFACE
5793 The macro definition is generated if there is a traditional pipeline
5794 description in @file{.md} file. You should also remember that to
5795 simplify the insn scheduler sources an empty traditional pipeline
5796 description interface is generated even if there is no a traditional
5797 pipeline description in the @file{.md} file. The macro can be used to
5798 distinguish the two types of the traditional interface.
5800 @findex DFA_PIPELINE_INTERFACE
5801 @item DFA_PIPELINE_INTERFACE
5802 The macro definition is generated if there is an automaton pipeline
5803 description in @file{.md} file. You should also remember that to
5804 simplify the insn scheduler sources an empty automaton pipeline
5805 description interface is generated even if there is no an automaton
5806 pipeline description in the @file{.md} file. The macro can be used to
5807 distinguish the two types of the automaton interface.
5809 @findex MAX_DFA_ISSUE_RATE
5810 @item MAX_DFA_ISSUE_RATE
5811 The macro definition is generated in the automaton based pipeline
5812 description interface. Its value is calculated from the automaton
5813 based pipeline description and is equal to maximal number of all insns
5814 described in constructions @samp{define_insn_reservation} which can be
5815 issued on the same processor cycle.
5820 @section Dividing the Output into Sections (Texts, Data, @dots{})
5821 @c the above section title is WAY too long. maybe cut the part between
5822 @c the (...)? --mew 10feb93
5824 An object file is divided into sections containing different types of
5825 data. In the most common case, there are three sections: the @dfn{text
5826 section}, which holds instructions and read-only data; the @dfn{data
5827 section}, which holds initialized writable data; and the @dfn{bss
5828 section}, which holds uninitialized data. Some systems have other kinds
5831 The compiler must tell the assembler when to switch sections. These
5832 macros control what commands to output to tell the assembler this. You
5833 can also define additional sections.
5836 @findex TEXT_SECTION_ASM_OP
5837 @item TEXT_SECTION_ASM_OP
5838 A C expression whose value is a string, including spacing, containing the
5839 assembler operation that should precede instructions and read-only data.
5840 Normally @code{"\t.text"} is right.
5842 @findex TEXT_SECTION
5844 A C statement that switches to the default section containing instructions.
5845 Normally this is not needed, as simply defining @code{TEXT_SECTION_ASM_OP}
5846 is enough. The MIPS port uses this to sort all functions after all data
5849 @findex HOT_TEXT_SECTION_NAME
5850 @item HOT_TEXT_SECTION_NAME
5851 If defined, a C string constant for the name of the section containing most
5852 frequently executed functions of the program. If not defined, GCC will provide
5853 a default definition if the target supports named sections.
5855 @findex UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5856 @item UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5857 If defined, a C string constant for the name of the section containing unlikely
5858 executed functions in the program.
5860 @findex DATA_SECTION_ASM_OP
5861 @item DATA_SECTION_ASM_OP
5862 A C expression whose value is a string, including spacing, containing the
5863 assembler operation to identify the following data as writable initialized
5864 data. Normally @code{"\t.data"} is right.
5866 @findex READONLY_DATA_SECTION_ASM_OP
5867 @item READONLY_DATA_SECTION_ASM_OP
5868 A C expression whose value is a string, including spacing, containing the
5869 assembler operation to identify the following data as read-only initialized
5872 @findex READONLY_DATA_SECTION
5873 @item READONLY_DATA_SECTION
5874 A macro naming a function to call to switch to the proper section for
5875 read-only data. The default is to use @code{READONLY_DATA_SECTION_ASM_OP}
5876 if defined, else fall back to @code{text_section}.
5878 The most common definition will be @code{data_section}, if the target
5879 does not have a special read-only data section, and does not put data
5880 in the text section.
5882 @findex SHARED_SECTION_ASM_OP
5883 @item SHARED_SECTION_ASM_OP
5884 If defined, a C expression whose value is a string, including spacing,
5885 containing the assembler operation to identify the following data as
5886 shared data. If not defined, @code{DATA_SECTION_ASM_OP} will be used.
5888 @findex BSS_SECTION_ASM_OP
5889 @item BSS_SECTION_ASM_OP
5890 If defined, a C expression whose value is a string, including spacing,
5891 containing the assembler operation to identify the following data as
5892 uninitialized global data. If not defined, and neither
5893 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5894 uninitialized global data will be output in the data section if
5895 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5898 @findex SHARED_BSS_SECTION_ASM_OP
5899 @item SHARED_BSS_SECTION_ASM_OP
5900 If defined, a C expression whose value is a string, including spacing,
5901 containing the assembler operation to identify the following data as
5902 uninitialized global shared data. If not defined, and
5903 @code{BSS_SECTION_ASM_OP} is, the latter will be used.
5905 @findex INIT_SECTION_ASM_OP
5906 @item INIT_SECTION_ASM_OP
5907 If defined, a C expression whose value is a string, including spacing,
5908 containing the assembler operation to identify the following data as
5909 initialization code. If not defined, GCC will assume such a section does
5912 @findex FINI_SECTION_ASM_OP
5913 @item FINI_SECTION_ASM_OP
5914 If defined, a C expression whose value is a string, including spacing,
5915 containing the assembler operation to identify the following data as
5916 finalization code. If not defined, GCC will assume such a section does
5919 @findex CRT_CALL_STATIC_FUNCTION
5920 @item CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
5921 If defined, an ASM statement that switches to a different section
5922 via @var{section_op}, calls @var{function}, and switches back to
5923 the text section. This is used in @file{crtstuff.c} if
5924 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
5925 to initialization and finalization functions from the init and fini
5926 sections. By default, this macro uses a simple function call. Some
5927 ports need hand-crafted assembly code to avoid dependencies on
5928 registers initialized in the function prologue or to ensure that
5929 constant pools don't end up too far way in the text section.
5931 @findex FORCE_CODE_SECTION_ALIGN
5932 @item FORCE_CODE_SECTION_ALIGN
5933 If defined, an ASM statement that aligns a code section to some
5934 arbitrary boundary. This is used to force all fragments of the
5935 @code{.init} and @code{.fini} sections to have to same alignment
5936 and thus prevent the linker from having to add any padding.
5938 @findex EXTRA_SECTIONS
5941 @item EXTRA_SECTIONS
5942 A list of names for sections other than the standard two, which are
5943 @code{in_text} and @code{in_data}. You need not define this macro
5944 on a system with no other sections (that GCC needs to use).
5946 @findex EXTRA_SECTION_FUNCTIONS
5947 @findex text_section
5948 @findex data_section
5949 @item EXTRA_SECTION_FUNCTIONS
5950 One or more functions to be defined in @file{varasm.c}. These
5951 functions should do jobs analogous to those of @code{text_section} and
5952 @code{data_section}, for your additional sections. Do not define this
5953 macro if you do not define @code{EXTRA_SECTIONS}.
5955 @findex JUMP_TABLES_IN_TEXT_SECTION
5956 @item JUMP_TABLES_IN_TEXT_SECTION
5957 Define this macro to be an expression with a nonzero value if jump
5958 tables (for @code{tablejump} insns) should be output in the text
5959 section, along with the assembler instructions. Otherwise, the
5960 readonly data section is used.
5962 This macro is irrelevant if there is no separate readonly data section.
5965 @deftypefn {Target Hook} void TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
5966 Switches to the appropriate section for output of @var{exp}. You can
5967 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
5968 some sort. @var{reloc} indicates whether the initial value of @var{exp}
5969 requires link-time relocations. Bit 0 is set when variable contains
5970 local relocations only, while bit 1 is set for global relocations.
5971 Select the section by calling @code{data_section} or one of the
5972 alternatives for other sections. @var{align} is the constant alignment
5975 The default version of this function takes care of putting read-only
5976 variables in @code{readonly_data_section}.
5979 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
5980 Build up a unique section name, expressed as a @code{STRING_CST} node,
5981 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
5982 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
5983 the initial value of @var{exp} requires link-time relocations.
5985 The default version of this function appends the symbol name to the
5986 ELF section name that would normally be used for the symbol. For
5987 example, the function @code{foo} would be placed in @code{.text.foo}.
5988 Whatever the actual target object format, this is often good enough.
5991 @deftypefn {Target Hook} void TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
5992 Switches to the appropriate section for output of constant pool entry
5993 @var{x} in @var{mode}. You can assume that @var{x} is some kind of
5994 constant in RTL@. The argument @var{mode} is redundant except in the
5995 case of a @code{const_int} rtx. Select the section by calling
5996 @code{readonly_data_section} or one of the alternatives for other
5997 sections. @var{align} is the constant alignment in bits.
5999 The default version of this function takes care of putting symbolic
6000 constants in @code{flag_pic} mode in @code{data_section} and everything
6001 else in @code{readonly_data_section}.
6004 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, int @var{new_decl_p})
6005 Define this hook if references to a symbol or a constant must be
6006 treated differently depending on something about the variable or
6007 function named by the symbol (such as what section it is in).
6009 The hook is executed under two circumstances. One is immediately after
6010 the rtl for @var{decl} that represents a variable or a function has been
6011 created and stored in @code{DECL_RTL(@var{decl})}. The value of the rtl
6012 will be a @code{mem} whose address is a @code{symbol_ref}. The other is
6013 immediately after the rtl for @var{decl} that represents a constant has
6014 been created and stored in @code{TREE_CST_RTL (@var{decl})}. The macro
6015 is called once for each distinct constant in a source file.
6017 The @var{new_decl_p} argument will be true if this is the first time
6018 that @code{ENCODE_SECTION_INFO} has been invoked on this decl. It will
6019 be false for subsequent invocations, which will happen for duplicate
6020 declarations. Whether or not anything must be done for the duplicate
6021 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
6023 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
6024 The usual thing for this hook to do is to record a flag in the
6025 @code{symbol_ref} (such as @code{SYMBOL_REF_FLAG}) or to store a
6026 modified name string in the @code{symbol_ref} (if one bit is not
6027 enough information).
6030 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
6031 Decode @var{name} and return the real name part, sans
6032 the characters that @code{TARGET_ENCODE_SECTION_INFO}
6036 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
6037 Returns true if @var{exp} should be placed into a ``small data'' section.
6038 The default version of this hook always returns false.
6041 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
6042 Contains the value true if the target places read-only
6043 ``small data'' into a separate section. The default value is false.
6046 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
6047 Returns true if @var{exp} names an object for which name resolution
6048 rules must resolve to the current ``module'' (dynamic shared library
6049 or executable image).
6051 The default version of this hook implements the name resolution rules
6052 for ELF, which has a looser model of global name binding than other
6053 currently supported object file formats.
6056 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
6057 Contains the value true if the target supports thread-local storage.
6058 The default value is false.
6063 @section Position Independent Code
6064 @cindex position independent code
6067 This section describes macros that help implement generation of position
6068 independent code. Simply defining these macros is not enough to
6069 generate valid PIC; you must also add support to the macros
6070 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
6071 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
6072 @samp{movsi} to do something appropriate when the source operand
6073 contains a symbolic address. You may also need to alter the handling of
6074 switch statements so that they use relative addresses.
6075 @c i rearranged the order of the macros above to try to force one of
6076 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
6079 @findex PIC_OFFSET_TABLE_REGNUM
6080 @item PIC_OFFSET_TABLE_REGNUM
6081 The register number of the register used to address a table of static
6082 data addresses in memory. In some cases this register is defined by a
6083 processor's ``application binary interface'' (ABI)@. When this macro
6084 is defined, RTL is generated for this register once, as with the stack
6085 pointer and frame pointer registers. If this macro is not defined, it
6086 is up to the machine-dependent files to allocate such a register (if
6087 necessary). Note that this register must be fixed when in use (e.g.@:
6088 when @code{flag_pic} is true).
6090 @findex PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6091 @item PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6092 Define this macro if the register defined by
6093 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
6094 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
6096 @findex FINALIZE_PIC
6098 By generating position-independent code, when two different programs (A
6099 and B) share a common library (libC.a), the text of the library can be
6100 shared whether or not the library is linked at the same address for both
6101 programs. In some of these environments, position-independent code
6102 requires not only the use of different addressing modes, but also
6103 special code to enable the use of these addressing modes.
6105 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
6106 codes once the function is being compiled into assembly code, but not
6107 before. (It is not done before, because in the case of compiling an
6108 inline function, it would lead to multiple PIC prologues being
6109 included in functions which used inline functions and were compiled to
6112 @findex LEGITIMATE_PIC_OPERAND_P
6113 @item LEGITIMATE_PIC_OPERAND_P (@var{x})
6114 A C expression that is nonzero if @var{x} is a legitimate immediate
6115 operand on the target machine when generating position independent code.
6116 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6117 check this. You can also assume @var{flag_pic} is true, so you need not
6118 check it either. You need not define this macro if all constants
6119 (including @code{SYMBOL_REF}) can be immediate operands when generating
6120 position independent code.
6123 @node Assembler Format
6124 @section Defining the Output Assembler Language
6126 This section describes macros whose principal purpose is to describe how
6127 to write instructions in assembler language---rather than what the
6131 * File Framework:: Structural information for the assembler file.
6132 * Data Output:: Output of constants (numbers, strings, addresses).
6133 * Uninitialized Data:: Output of uninitialized variables.
6134 * Label Output:: Output and generation of labels.
6135 * Initialization:: General principles of initialization
6136 and termination routines.
6137 * Macros for Initialization::
6138 Specific macros that control the handling of
6139 initialization and termination routines.
6140 * Instruction Output:: Output of actual instructions.
6141 * Dispatch Tables:: Output of jump tables.
6142 * Exception Region Output:: Output of exception region code.
6143 * Alignment Output:: Pseudo ops for alignment and skipping data.
6146 @node File Framework
6147 @subsection The Overall Framework of an Assembler File
6148 @cindex assembler format
6149 @cindex output of assembler code
6151 @c prevent bad page break with this line
6152 This describes the overall framework of an assembler file.
6155 @findex ASM_FILE_START
6156 @item ASM_FILE_START (@var{stream})
6157 A C expression which outputs to the stdio stream @var{stream}
6158 some appropriate text to go at the start of an assembler file.
6160 Normally this macro is defined to output a line containing
6161 @samp{#NO_APP}, which is a comment that has no effect on most
6162 assemblers but tells the GNU assembler that it can save time by not
6163 checking for certain assembler constructs.
6165 On systems that use SDB, it is necessary to output certain commands;
6166 see @file{attasm.h}.
6168 @findex ASM_FILE_END
6169 @item ASM_FILE_END (@var{stream})
6170 A C expression which outputs to the stdio stream @var{stream}
6171 some appropriate text to go at the end of an assembler file.
6173 If this macro is not defined, the default is to output nothing
6174 special at the end of the file. Most systems don't require any
6177 On systems that use SDB, it is necessary to output certain commands;
6178 see @file{attasm.h}.
6180 @findex ASM_COMMENT_START
6181 @item ASM_COMMENT_START
6182 A C string constant describing how to begin a comment in the target
6183 assembler language. The compiler assumes that the comment will end at
6184 the end of the line.
6188 A C string constant for text to be output before each @code{asm}
6189 statement or group of consecutive ones. Normally this is
6190 @code{"#APP"}, which is a comment that has no effect on most
6191 assemblers but tells the GNU assembler that it must check the lines
6192 that follow for all valid assembler constructs.
6196 A C string constant for text to be output after each @code{asm}
6197 statement or group of consecutive ones. Normally this is
6198 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6199 time-saving assumptions that are valid for ordinary compiler output.
6201 @findex ASM_OUTPUT_SOURCE_FILENAME
6202 @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6203 A C statement to output COFF information or DWARF debugging information
6204 which indicates that filename @var{name} is the current source file to
6205 the stdio stream @var{stream}.
6207 This macro need not be defined if the standard form of output
6208 for the file format in use is appropriate.
6210 @findex OUTPUT_QUOTED_STRING
6211 @item OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6212 A C statement to output the string @var{string} to the stdio stream
6213 @var{stream}. If you do not call the function @code{output_quoted_string}
6214 in your config files, GCC will only call it to output filenames to
6215 the assembler source. So you can use it to canonicalize the format
6216 of the filename using this macro.
6218 @findex ASM_OUTPUT_SOURCE_LINE
6219 @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
6220 A C statement to output DBX or SDB debugging information before code
6221 for line number @var{line} of the current source file to the
6222 stdio stream @var{stream}.
6224 This macro need not be defined if the standard form of debugging
6225 information for the debugger in use is appropriate.
6227 @findex ASM_OUTPUT_IDENT
6228 @item ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6229 A C statement to output something to the assembler file to handle a
6230 @samp{#ident} directive containing the text @var{string}. If this
6231 macro is not defined, nothing is output for a @samp{#ident} directive.
6233 @findex OBJC_PROLOGUE
6235 A C statement to output any assembler statements which are required to
6236 precede any Objective-C object definitions or message sending. The
6237 statement is executed only when compiling an Objective-C program.
6240 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6241 Output assembly directives to switch to section @var{name}. The section
6242 should have attributes as specified by @var{flags}, which is a bit mask
6243 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6244 is nonzero, it contains an alignment in bytes to be used for the section,
6245 otherwise some target default should be used. Only targets that must
6246 specify an alignment within the section directive need pay attention to
6247 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6250 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6251 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6254 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6255 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6256 based on a variable or function decl, a section name, and whether or not the
6257 declaration's initializer may contain runtime relocations. @var{decl} may be
6258 null, in which case read-write data should be assumed.
6260 The default version if this function handles choosing code vs data,
6261 read-only vs read-write data, and @code{flag_pic}. You should only
6262 need to override this if your target has special flags that might be
6263 set via @code{__attribute__}.
6268 @subsection Output of Data
6271 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6272 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6273 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6274 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6275 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6276 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6277 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6278 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6279 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6280 These hooks specify assembly directives for creating certain kinds
6281 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6282 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6283 aligned two-byte object, and so on. Any of the hooks may be
6284 @code{NULL}, indicating that no suitable directive is available.
6286 The compiler will print these strings at the start of a new line,
6287 followed immediately by the object's initial value. In most cases,
6288 the string should contain a tab, a pseudo-op, and then another tab.
6291 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6292 The @code{assemble_integer} function uses this hook to output an
6293 integer object. @var{x} is the object's value, @var{size} is its size
6294 in bytes and @var{aligned_p} indicates whether it is aligned. The
6295 function should return @code{true} if it was able to output the
6296 object. If it returns false, @code{assemble_integer} will try to
6297 split the object into smaller parts.
6299 The default implementation of this hook will use the
6300 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6301 when the relevant string is @code{NULL}.
6305 @findex OUTPUT_ADDR_CONST_EXTRA
6306 @item OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6307 A C statement to recognize @var{rtx} patterns that
6308 @code{output_addr_const} can't deal with, and output assembly code to
6309 @var{stream} corresponding to the pattern @var{x}. This may be used to
6310 allow machine-dependent @code{UNSPEC}s to appear within constants.
6312 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6313 @code{goto fail}, so that a standard error message is printed. If it
6314 prints an error message itself, by calling, for example,
6315 @code{output_operand_lossage}, it may just complete normally.
6317 @findex ASM_OUTPUT_ASCII
6318 @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6319 A C statement to output to the stdio stream @var{stream} an assembler
6320 instruction to assemble a string constant containing the @var{len}
6321 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6322 @code{char *} and @var{len} a C expression of type @code{int}.
6324 If the assembler has a @code{.ascii} pseudo-op as found in the
6325 Berkeley Unix assembler, do not define the macro
6326 @code{ASM_OUTPUT_ASCII}.
6328 @findex ASM_OUTPUT_FDESC
6329 @item ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6330 A C statement to output word @var{n} of a function descriptor for
6331 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6332 is defined, and is otherwise unused.
6334 @findex CONSTANT_POOL_BEFORE_FUNCTION
6335 @item CONSTANT_POOL_BEFORE_FUNCTION
6336 You may define this macro as a C expression. You should define the
6337 expression to have a nonzero value if GCC should output the constant
6338 pool for a function before the code for the function, or a zero value if
6339 GCC should output the constant pool after the function. If you do
6340 not define this macro, the usual case, GCC will output the constant
6341 pool before the function.
6343 @findex ASM_OUTPUT_POOL_PROLOGUE
6344 @item ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
6345 A C statement to output assembler commands to define the start of the
6346 constant pool for a function. @var{funname} is a string giving
6347 the name of the function. Should the return type of the function
6348 be required, it can be obtained via @var{fundecl}. @var{size}
6349 is the size, in bytes, of the constant pool that will be written
6350 immediately after this call.
6352 If no constant-pool prefix is required, the usual case, this macro need
6355 @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY
6356 @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
6357 A C statement (with or without semicolon) to output a constant in the
6358 constant pool, if it needs special treatment. (This macro need not do
6359 anything for RTL expressions that can be output normally.)
6361 The argument @var{file} is the standard I/O stream to output the
6362 assembler code on. @var{x} is the RTL expression for the constant to
6363 output, and @var{mode} is the machine mode (in case @var{x} is a
6364 @samp{const_int}). @var{align} is the required alignment for the value
6365 @var{x}; you should output an assembler directive to force this much
6368 The argument @var{labelno} is a number to use in an internal label for
6369 the address of this pool entry. The definition of this macro is
6370 responsible for outputting the label definition at the proper place.
6371 Here is how to do this:
6374 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
6377 When you output a pool entry specially, you should end with a
6378 @code{goto} to the label @var{jumpto}. This will prevent the same pool
6379 entry from being output a second time in the usual manner.
6381 You need not define this macro if it would do nothing.
6383 @findex CONSTANT_AFTER_FUNCTION_P
6384 @item CONSTANT_AFTER_FUNCTION_P (@var{exp})
6385 Define this macro as a C expression which is nonzero if the constant
6386 @var{exp}, of type @code{tree}, should be output after the code for a
6387 function. The compiler will normally output all constants before the
6388 function; you need not define this macro if this is OK@.
6390 @findex ASM_OUTPUT_POOL_EPILOGUE
6391 @item ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
6392 A C statement to output assembler commands to at the end of the constant
6393 pool for a function. @var{funname} is a string giving the name of the
6394 function. Should the return type of the function be required, you can
6395 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
6396 constant pool that GCC wrote immediately before this call.
6398 If no constant-pool epilogue is required, the usual case, you need not
6401 @findex IS_ASM_LOGICAL_LINE_SEPARATOR
6402 @item IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
6403 Define this macro as a C expression which is nonzero if @var{C} is
6404 used as a logical line separator by the assembler.
6406 If you do not define this macro, the default is that only
6407 the character @samp{;} is treated as a logical line separator.
6410 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
6411 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
6412 These target hooks are C string constants, describing the syntax in the
6413 assembler for grouping arithmetic expressions. If not overridden, they
6414 default to normal parentheses, which is correct for most assemblers.
6417 These macros are provided by @file{real.h} for writing the definitions
6418 of @code{ASM_OUTPUT_DOUBLE} and the like:
6421 @item REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
6422 @itemx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
6423 @itemx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
6424 @findex REAL_VALUE_TO_TARGET_SINGLE
6425 @findex REAL_VALUE_TO_TARGET_DOUBLE
6426 @findex REAL_VALUE_TO_TARGET_LONG_DOUBLE
6427 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
6428 floating point representation, and store its bit pattern in the variable
6429 @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE}, this variable should
6430 be a simple @code{long int}. For the others, it should be an array of
6431 @code{long int}. The number of elements in this array is determined by
6432 the size of the desired target floating point data type: 32 bits of it
6433 go in each @code{long int} array element. Each array element holds 32
6434 bits of the result, even if @code{long int} is wider than 32 bits on the
6437 The array element values are designed so that you can print them out
6438 using @code{fprintf} in the order they should appear in the target
6442 @node Uninitialized Data
6443 @subsection Output of Uninitialized Variables
6445 Each of the macros in this section is used to do the whole job of
6446 outputting a single uninitialized variable.
6449 @findex ASM_OUTPUT_COMMON
6450 @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6451 A C statement (sans semicolon) to output to the stdio stream
6452 @var{stream} the assembler definition of a common-label named
6453 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6454 is the size rounded up to whatever alignment the caller wants.
6456 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6457 output the name itself; before and after that, output the additional
6458 assembler syntax for defining the name, and a newline.
6460 This macro controls how the assembler definitions of uninitialized
6461 common global variables are output.
6463 @findex ASM_OUTPUT_ALIGNED_COMMON
6464 @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
6465 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
6466 separate, explicit argument. If you define this macro, it is used in
6467 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
6468 handling the required alignment of the variable. The alignment is specified
6469 as the number of bits.
6471 @findex ASM_OUTPUT_ALIGNED_DECL_COMMON
6472 @item ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6473 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
6474 variable to be output, if there is one, or @code{NULL_TREE} if there
6475 is no corresponding variable. If you define this macro, GCC will use it
6476 in place of both @code{ASM_OUTPUT_COMMON} and
6477 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
6478 the variable's decl in order to chose what to output.
6480 @findex ASM_OUTPUT_SHARED_COMMON
6481 @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6482 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
6483 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
6486 @findex ASM_OUTPUT_BSS
6487 @item ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6488 A C statement (sans semicolon) to output to the stdio stream
6489 @var{stream} the assembler definition of uninitialized global @var{decl} named
6490 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6491 is the size rounded up to whatever alignment the caller wants.
6493 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
6494 defining this macro. If unable, use the expression
6495 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
6496 before and after that, output the additional assembler syntax for defining
6497 the name, and a newline.
6499 This macro controls how the assembler definitions of uninitialized global
6500 variables are output. This macro exists to properly support languages like
6501 C++ which do not have @code{common} data. However, this macro currently
6502 is not defined for all targets. If this macro and
6503 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
6504 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
6505 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
6507 @findex ASM_OUTPUT_ALIGNED_BSS
6508 @item ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6509 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
6510 separate, explicit argument. If you define this macro, it is used in
6511 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
6512 handling the required alignment of the variable. The alignment is specified
6513 as the number of bits.
6515 Try to use function @code{asm_output_aligned_bss} defined in file
6516 @file{varasm.c} when defining this macro.
6518 @findex ASM_OUTPUT_SHARED_BSS
6519 @item ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6520 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
6521 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
6524 @findex ASM_OUTPUT_LOCAL
6525 @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6526 A C statement (sans semicolon) to output to the stdio stream
6527 @var{stream} the assembler definition of a local-common-label named
6528 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6529 is the size rounded up to whatever alignment the caller wants.
6531 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6532 output the name itself; before and after that, output the additional
6533 assembler syntax for defining the name, and a newline.
6535 This macro controls how the assembler definitions of uninitialized
6536 static variables are output.
6538 @findex ASM_OUTPUT_ALIGNED_LOCAL
6539 @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
6540 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
6541 separate, explicit argument. If you define this macro, it is used in
6542 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
6543 handling the required alignment of the variable. The alignment is specified
6544 as the number of bits.
6546 @findex ASM_OUTPUT_ALIGNED_DECL_LOCAL
6547 @item ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6548 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
6549 variable to be output, if there is one, or @code{NULL_TREE} if there
6550 is no corresponding variable. If you define this macro, GCC will use it
6551 in place of both @code{ASM_OUTPUT_DECL} and
6552 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
6553 the variable's decl in order to chose what to output.
6555 @findex ASM_OUTPUT_SHARED_LOCAL
6556 @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6557 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
6558 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
6563 @subsection Output and Generation of Labels
6565 @c prevent bad page break with this line
6566 This is about outputting labels.
6569 @findex ASM_OUTPUT_LABEL
6570 @findex assemble_name
6571 @item ASM_OUTPUT_LABEL (@var{stream}, @var{name})
6572 A C statement (sans semicolon) to output to the stdio stream
6573 @var{stream} the assembler definition of a label named @var{name}.
6574 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6575 output the name itself; before and after that, output the additional
6576 assembler syntax for defining the name, and a newline. A default
6577 definition of this macro is provided which is correct for most systems.
6581 A C string containing the appropriate assembler directive to specify the
6582 size of a symbol, without any arguments. On systems that use ELF, the
6583 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
6584 systems, the default is not to define this macro.
6586 Define this macro only if it is correct to use the default definitions
6587 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
6588 for your system. If you need your own custom definitions of those
6589 macros, or if you do not need explicit symbol sizes at all, do not
6592 @findex ASM_OUTPUT_SIZE_DIRECTIVE
6593 @item ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
6594 A C statement (sans semicolon) to output to the stdio stream
6595 @var{stream} a directive telling the assembler that the size of the
6596 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
6597 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6600 @findex ASM_OUTPUT_MEASURED_SIZE
6601 @item ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
6602 A C statement (sans semicolon) to output to the stdio stream
6603 @var{stream} a directive telling the assembler to calculate the size of
6604 the symbol @var{name} by subtracting its address from the current
6607 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6608 provided. The default assumes that the assembler recognizes a special
6609 @samp{.} symbol as referring to the current address, and can calculate
6610 the difference between this and another symbol. If your assembler does
6611 not recognize @samp{.} or cannot do calculations with it, you will need
6612 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
6616 A C string containing the appropriate assembler directive to specify the
6617 type of a symbol, without any arguments. On systems that use ELF, the
6618 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
6619 systems, the default is not to define this macro.
6621 Define this macro only if it is correct to use the default definition of
6622 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6623 custom definition of this macro, or if you do not need explicit symbol
6624 types at all, do not define this macro.
6626 @findex TYPE_OPERAND_FMT
6627 @item TYPE_OPERAND_FMT
6628 A C string which specifies (using @code{printf} syntax) the format of
6629 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
6630 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
6631 the default is not to define this macro.
6633 Define this macro only if it is correct to use the default definition of
6634 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6635 custom definition of this macro, or if you do not need explicit symbol
6636 types at all, do not define this macro.
6638 @findex ASM_OUTPUT_TYPE_DIRECTIVE
6639 @item ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
6640 A C statement (sans semicolon) to output to the stdio stream
6641 @var{stream} a directive telling the assembler that the type of the
6642 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
6643 that string is always either @samp{"function"} or @samp{"object"}, but
6644 you should not count on this.
6646 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
6647 definition of this macro is provided.
6649 @findex ASM_DECLARE_FUNCTION_NAME
6650 @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
6651 A C statement (sans semicolon) to output to the stdio stream
6652 @var{stream} any text necessary for declaring the name @var{name} of a
6653 function which is being defined. This macro is responsible for
6654 outputting the label definition (perhaps using
6655 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
6656 @code{FUNCTION_DECL} tree node representing the function.
6658 If this macro is not defined, then the function name is defined in the
6659 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6661 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6664 @findex ASM_DECLARE_FUNCTION_SIZE
6665 @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
6666 A C statement (sans semicolon) to output to the stdio stream
6667 @var{stream} any text necessary for declaring the size of a function
6668 which is being defined. The argument @var{name} is the name of the
6669 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
6670 representing the function.
6672 If this macro is not defined, then the function size is not defined.
6674 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
6677 @findex ASM_DECLARE_OBJECT_NAME
6678 @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6679 A C statement (sans semicolon) to output to the stdio stream
6680 @var{stream} any text necessary for declaring the name @var{name} of an
6681 initialized variable which is being defined. This macro must output the
6682 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6683 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6685 If this macro is not defined, then the variable name is defined in the
6686 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6688 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
6689 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
6691 @findex ASM_DECLARE_REGISTER_GLOBAL
6692 @item ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6693 A C statement (sans semicolon) to output to the stdio stream
6694 @var{stream} any text necessary for claiming a register @var{regno}
6695 for a global variable @var{decl} with name @var{name}.
6697 If you don't define this macro, that is equivalent to defining it to do
6700 @findex ASM_FINISH_DECLARE_OBJECT
6701 @item ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6702 A C statement (sans semicolon) to finish up declaring a variable name
6703 once the compiler has processed its initializer fully and thus has had a
6704 chance to determine the size of an array when controlled by an
6705 initializer. This is used on systems where it's necessary to declare
6706 something about the size of the object.
6708 If you don't define this macro, that is equivalent to defining it to do
6711 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
6712 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
6715 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
6716 This target hook is a function to output to the stdio stream
6717 @var{stream} some commands that will make the label @var{name} global;
6718 that is, available for reference from other files.
6720 The default implementation relies on a proper definition of
6721 @code{GLOBAL_ASM_OP}.
6725 @findex ASM_WEAKEN_LABEL
6726 @item ASM_WEAKEN_LABEL (@var{stream}, @var{name})
6727 A C statement (sans semicolon) to output to the stdio stream
6728 @var{stream} some commands that will make the label @var{name} weak;
6729 that is, available for reference from other files but only used if
6730 no other definition is available. Use the expression
6731 @code{assemble_name (@var{stream}, @var{name})} to output the name
6732 itself; before and after that, output the additional assembler syntax
6733 for making that name weak, and a newline.
6735 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
6736 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
6739 @findex ASM_WEAKEN_DECL
6740 @item ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
6741 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
6742 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
6743 or variable decl. If @var{value} is not @code{NULL}, this C statement
6744 should output to the stdio stream @var{stream} assembler code which
6745 defines (equates) the weak symbol @var{name} to have the value
6746 @var{value}. If @var{value} is @code{NULL}, it should output commands
6747 to make @var{name} weak.
6749 @findex SUPPORTS_WEAK
6751 A C expression which evaluates to true if the target supports weak symbols.
6753 If you don't define this macro, @file{defaults.h} provides a default
6754 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
6755 is defined, the default definition is @samp{1}; otherwise, it is
6756 @samp{0}. Define this macro if you want to control weak symbol support
6757 with a compiler flag such as @option{-melf}.
6759 @findex MAKE_DECL_ONE_ONLY (@var{decl})
6760 @item MAKE_DECL_ONE_ONLY
6761 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
6762 public symbol such that extra copies in multiple translation units will
6763 be discarded by the linker. Define this macro if your object file
6764 format provides support for this concept, such as the @samp{COMDAT}
6765 section flags in the Microsoft Windows PE/COFF format, and this support
6766 requires changes to @var{decl}, such as putting it in a separate section.
6768 @findex SUPPORTS_ONE_ONLY
6769 @item SUPPORTS_ONE_ONLY
6770 A C expression which evaluates to true if the target supports one-only
6773 If you don't define this macro, @file{varasm.c} provides a default
6774 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6775 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6776 you want to control one-only symbol support with a compiler flag, or if
6777 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6778 be emitted as one-only.
6780 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
6781 This target hook is a function to output to @var{asm_out_file} some
6782 commands that will make the symbol(s) associated with @var{decl} have
6783 hidden, protected or internal visibility as specified by @var{visibility}.
6786 @findex ASM_OUTPUT_EXTERNAL
6787 @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6788 A C statement (sans semicolon) to output to the stdio stream
6789 @var{stream} any text necessary for declaring the name of an external
6790 symbol named @var{name} which is referenced in this compilation but
6791 not defined. The value of @var{decl} is the tree node for the
6794 This macro need not be defined if it does not need to output anything.
6795 The GNU assembler and most Unix assemblers don't require anything.
6797 @findex ASM_OUTPUT_EXTERNAL_LIBCALL
6798 @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref})
6799 A C statement (sans semicolon) to output on @var{stream} an assembler
6800 pseudo-op to declare a library function name external. The name of the
6801 library function is given by @var{symref}, which has type @code{rtx} and
6802 is a @code{symbol_ref}.
6804 This macro need not be defined if it does not need to output anything.
6805 The GNU assembler and most Unix assemblers don't require anything.
6807 @findex ASM_OUTPUT_LABELREF
6808 @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6809 A C statement (sans semicolon) to output to the stdio stream
6810 @var{stream} a reference in assembler syntax to a label named
6811 @var{name}. This should add @samp{_} to the front of the name, if that
6812 is customary on your operating system, as it is in most Berkeley Unix
6813 systems. This macro is used in @code{assemble_name}.
6815 @findex ASM_OUTPUT_SYMBOL_REF
6816 @item ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6817 A C statement (sans semicolon) to output a reference to
6818 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
6819 will be used to output the name of the symbol. This macro may be used
6820 to modify the way a symbol is referenced depending on information
6821 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
6823 @findex ASM_OUTPUT_LABEL_REF
6824 @item ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
6825 A C statement (sans semicolon) to output a reference to @var{buf}, the
6826 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
6827 @code{assemble_name} will be used to output the name of the symbol.
6828 This macro is not used by @code{output_asm_label}, or the @code{%l}
6829 specifier that calls it; the intention is that this macro should be set
6830 when it is necessary to output a label differently when its address is
6834 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
6835 A function to output to the stdio stream @var{stream} a label whose
6836 name is made from the string @var{prefix} and the number @var{labelno}.
6838 It is absolutely essential that these labels be distinct from the labels
6839 used for user-level functions and variables. Otherwise, certain programs
6840 will have name conflicts with internal labels.
6842 It is desirable to exclude internal labels from the symbol table of the
6843 object file. Most assemblers have a naming convention for labels that
6844 should be excluded; on many systems, the letter @samp{L} at the
6845 beginning of a label has this effect. You should find out what
6846 convention your system uses, and follow it.
6848 The default version of this function utilizes ASM_GENERATE_INTERNAL_LABEL.
6853 @findex ASM_OUTPUT_DEBUG_LABEL
6854 @item ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
6855 A C statement to output to the stdio stream @var{stream} a debug info
6856 label whose name is made from the string @var{prefix} and the number
6857 @var{num}. This is useful for VLIW targets, where debug info labels
6858 may need to be treated differently than branch target labels. On some
6859 systems, branch target labels must be at the beginning of instruction
6860 bundles, but debug info labels can occur in the middle of instruction
6863 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
6866 @findex ASM_GENERATE_INTERNAL_LABEL
6867 @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
6868 A C statement to store into the string @var{string} a label whose name
6869 is made from the string @var{prefix} and the number @var{num}.
6871 This string, when output subsequently by @code{assemble_name}, should
6872 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
6873 with the same @var{prefix} and @var{num}.
6875 If the string begins with @samp{*}, then @code{assemble_name} will
6876 output the rest of the string unchanged. It is often convenient for
6877 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
6878 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
6879 to output the string, and may change it. (Of course,
6880 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
6881 you should know what it does on your machine.)
6883 @findex ASM_FORMAT_PRIVATE_NAME
6884 @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
6885 A C expression to assign to @var{outvar} (which is a variable of type
6886 @code{char *}) a newly allocated string made from the string
6887 @var{name} and the number @var{number}, with some suitable punctuation
6888 added. Use @code{alloca} to get space for the string.
6890 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
6891 produce an assembler label for an internal static variable whose name is
6892 @var{name}. Therefore, the string must be such as to result in valid
6893 assembler code. The argument @var{number} is different each time this
6894 macro is executed; it prevents conflicts between similarly-named
6895 internal static variables in different scopes.
6897 Ideally this string should not be a valid C identifier, to prevent any
6898 conflict with the user's own symbols. Most assemblers allow periods
6899 or percent signs in assembler symbols; putting at least one of these
6900 between the name and the number will suffice.
6902 If this macro is not defined, a default definition will be provided
6903 which is correct for most systems.
6905 @findex ASM_OUTPUT_DEF
6906 @item ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
6907 A C statement to output to the stdio stream @var{stream} assembler code
6908 which defines (equates) the symbol @var{name} to have the value @var{value}.
6911 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6912 correct for most systems.
6914 @findex ASM_OUTPUT_DEF_FROM_DECLS
6915 @item ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
6916 A C statement to output to the stdio stream @var{stream} assembler code
6917 which defines (equates) the symbol whose tree node is @var{decl_of_name}
6918 to have the value of the tree node @var{decl_of_value}. This macro will
6919 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
6920 the tree nodes are available.
6923 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6924 correct for most systems.
6926 @findex ASM_OUTPUT_WEAK_ALIAS
6927 @item ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
6928 A C statement to output to the stdio stream @var{stream} assembler code
6929 which defines (equates) the weak symbol @var{name} to have the value
6930 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
6931 an undefined weak symbol.
6933 Define this macro if the target only supports weak aliases; define
6934 @code{ASM_OUTPUT_DEF} instead if possible.
6936 @findex OBJC_GEN_METHOD_LABEL
6937 @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
6938 Define this macro to override the default assembler names used for
6939 Objective-C methods.
6941 The default name is a unique method number followed by the name of the
6942 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
6943 the category is also included in the assembler name (e.g.@:
6946 These names are safe on most systems, but make debugging difficult since
6947 the method's selector is not present in the name. Therefore, particular
6948 systems define other ways of computing names.
6950 @var{buf} is an expression of type @code{char *} which gives you a
6951 buffer in which to store the name; its length is as long as
6952 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
6953 50 characters extra.
6955 The argument @var{is_inst} specifies whether the method is an instance
6956 method or a class method; @var{class_name} is the name of the class;
6957 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
6958 in a category); and @var{sel_name} is the name of the selector.
6960 On systems where the assembler can handle quoted names, you can use this
6961 macro to provide more human-readable names.
6963 @findex ASM_DECLARE_CLASS_REFERENCE
6964 @item ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
6965 A C statement (sans semicolon) to output to the stdio stream
6966 @var{stream} commands to declare that the label @var{name} is an
6967 Objective-C class reference. This is only needed for targets whose
6968 linkers have special support for NeXT-style runtimes.
6970 @findex ASM_DECLARE_UNRESOLVED_REFERENCE
6971 @item ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
6972 A C statement (sans semicolon) to output to the stdio stream
6973 @var{stream} commands to declare that the label @var{name} is an
6974 unresolved Objective-C class reference. This is only needed for targets
6975 whose linkers have special support for NeXT-style runtimes.
6978 @node Initialization
6979 @subsection How Initialization Functions Are Handled
6980 @cindex initialization routines
6981 @cindex termination routines
6982 @cindex constructors, output of
6983 @cindex destructors, output of
6985 The compiled code for certain languages includes @dfn{constructors}
6986 (also called @dfn{initialization routines})---functions to initialize
6987 data in the program when the program is started. These functions need
6988 to be called before the program is ``started''---that is to say, before
6989 @code{main} is called.
6991 Compiling some languages generates @dfn{destructors} (also called
6992 @dfn{termination routines}) that should be called when the program
6995 To make the initialization and termination functions work, the compiler
6996 must output something in the assembler code to cause those functions to
6997 be called at the appropriate time. When you port the compiler to a new
6998 system, you need to specify how to do this.
7000 There are two major ways that GCC currently supports the execution of
7001 initialization and termination functions. Each way has two variants.
7002 Much of the structure is common to all four variations.
7004 @findex __CTOR_LIST__
7005 @findex __DTOR_LIST__
7006 The linker must build two lists of these functions---a list of
7007 initialization functions, called @code{__CTOR_LIST__}, and a list of
7008 termination functions, called @code{__DTOR_LIST__}.
7010 Each list always begins with an ignored function pointer (which may hold
7011 0, @minus{}1, or a count of the function pointers after it, depending on
7012 the environment). This is followed by a series of zero or more function
7013 pointers to constructors (or destructors), followed by a function
7014 pointer containing zero.
7016 Depending on the operating system and its executable file format, either
7017 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
7018 time and exit time. Constructors are called in reverse order of the
7019 list; destructors in forward order.
7021 The best way to handle static constructors works only for object file
7022 formats which provide arbitrarily-named sections. A section is set
7023 aside for a list of constructors, and another for a list of destructors.
7024 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
7025 object file that defines an initialization function also puts a word in
7026 the constructor section to point to that function. The linker
7027 accumulates all these words into one contiguous @samp{.ctors} section.
7028 Termination functions are handled similarly.
7030 This method will be chosen as the default by @file{target-def.h} if
7031 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
7032 support arbitrary sections, but does support special designated
7033 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
7034 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
7036 When arbitrary sections are available, there are two variants, depending
7037 upon how the code in @file{crtstuff.c} is called. On systems that
7038 support a @dfn{.init} section which is executed at program startup,
7039 parts of @file{crtstuff.c} are compiled into that section. The
7040 program is linked by the @code{gcc} driver like this:
7043 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
7046 The prologue of a function (@code{__init}) appears in the @code{.init}
7047 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
7048 for the function @code{__fini} in the @dfn{.fini} section. Normally these
7049 files are provided by the operating system or by the GNU C library, but
7050 are provided by GCC for a few targets.
7052 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
7053 compiled from @file{crtstuff.c}. They contain, among other things, code
7054 fragments within the @code{.init} and @code{.fini} sections that branch
7055 to routines in the @code{.text} section. The linker will pull all parts
7056 of a section together, which results in a complete @code{__init} function
7057 that invokes the routines we need at startup.
7059 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
7062 If no init section is available, when GCC compiles any function called
7063 @code{main} (or more accurately, any function designated as a program
7064 entry point by the language front end calling @code{expand_main_function}),
7065 it inserts a procedure call to @code{__main} as the first executable code
7066 after the function prologue. The @code{__main} function is defined
7067 in @file{libgcc2.c} and runs the global constructors.
7069 In file formats that don't support arbitrary sections, there are again
7070 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
7071 and an `a.out' format must be used. In this case,
7072 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
7073 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
7074 and with the address of the void function containing the initialization
7075 code as its value. The GNU linker recognizes this as a request to add
7076 the value to a @dfn{set}; the values are accumulated, and are eventually
7077 placed in the executable as a vector in the format described above, with
7078 a leading (ignored) count and a trailing zero element.
7079 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
7080 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
7081 the compilation of @code{main} to call @code{__main} as above, starting
7082 the initialization process.
7084 The last variant uses neither arbitrary sections nor the GNU linker.
7085 This is preferable when you want to do dynamic linking and when using
7086 file formats which the GNU linker does not support, such as `ECOFF'@. In
7087 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
7088 termination functions are recognized simply by their names. This requires
7089 an extra program in the linkage step, called @command{collect2}. This program
7090 pretends to be the linker, for use with GCC; it does its job by running
7091 the ordinary linker, but also arranges to include the vectors of
7092 initialization and termination functions. These functions are called
7093 via @code{__main} as described above. In order to use this method,
7094 @code{use_collect2} must be defined in the target in @file{config.gcc}.
7097 The following section describes the specific macros that control and
7098 customize the handling of initialization and termination functions.
7101 @node Macros for Initialization
7102 @subsection Macros Controlling Initialization Routines
7104 Here are the macros that control how the compiler handles initialization
7105 and termination functions:
7108 @findex INIT_SECTION_ASM_OP
7109 @item INIT_SECTION_ASM_OP
7110 If defined, a C string constant, including spacing, for the assembler
7111 operation to identify the following data as initialization code. If not
7112 defined, GCC will assume such a section does not exist. When you are
7113 using special sections for initialization and termination functions, this
7114 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7115 run the initialization functions.
7117 @item HAS_INIT_SECTION
7118 @findex HAS_INIT_SECTION
7119 If defined, @code{main} will not call @code{__main} as described above.
7120 This macro should be defined for systems that control start-up code
7121 on a symbol-by-symbol basis, such as OSF/1, and should not
7122 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7124 @item LD_INIT_SWITCH
7125 @findex LD_INIT_SWITCH
7126 If defined, a C string constant for a switch that tells the linker that
7127 the following symbol is an initialization routine.
7129 @item LD_FINI_SWITCH
7130 @findex LD_FINI_SWITCH
7131 If defined, a C string constant for a switch that tells the linker that
7132 the following symbol is a finalization routine.
7134 @item COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7135 If defined, a C statement that will write a function that can be
7136 automatically called when a shared library is loaded. The function
7137 should call @var{func}, which takes no arguments. If not defined, and
7138 the object format requires an explicit initialization function, then a
7139 function called @code{_GLOBAL__DI} will be generated.
7141 This function and the following one are used by collect2 when linking a
7142 shared library that needs constructors or destructors, or has DWARF2
7143 exception tables embedded in the code.
7145 @item COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7146 If defined, a C statement that will write a function that can be
7147 automatically called when a shared library is unloaded. The function
7148 should call @var{func}, which takes no arguments. If not defined, and
7149 the object format requires an explicit finalization function, then a
7150 function called @code{_GLOBAL__DD} will be generated.
7153 @findex INVOKE__main
7154 If defined, @code{main} will call @code{__main} despite the presence of
7155 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7156 where the init section is not actually run automatically, but is still
7157 useful for collecting the lists of constructors and destructors.
7159 @item SUPPORTS_INIT_PRIORITY
7160 @findex SUPPORTS_INIT_PRIORITY
7161 If nonzero, the C++ @code{init_priority} attribute is supported and the
7162 compiler should emit instructions to control the order of initialization
7163 of objects. If zero, the compiler will issue an error message upon
7164 encountering an @code{init_priority} attribute.
7167 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7168 This value is true if the target supports some ``native'' method of
7169 collecting constructors and destructors to be run at startup and exit.
7170 It is false if we must use @command{collect2}.
7173 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7174 If defined, a function that outputs assembler code to arrange to call
7175 the function referenced by @var{symbol} at initialization time.
7177 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7178 no arguments and with no return value. If the target supports initialization
7179 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7180 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7182 If this macro is not defined by the target, a suitable default will
7183 be chosen if (1) the target supports arbitrary section names, (2) the
7184 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7188 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7189 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7190 functions rather than initialization functions.
7193 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7194 generated for the generated object file will have static linkage.
7196 If your system uses @command{collect2} as the means of processing
7197 constructors, then that program normally uses @command{nm} to scan
7198 an object file for constructor functions to be called.
7200 On certain kinds of systems, you can define these macros to make
7201 @command{collect2} work faster (and, in some cases, make it work at all):
7204 @findex OBJECT_FORMAT_COFF
7205 @item OBJECT_FORMAT_COFF
7206 Define this macro if the system uses COFF (Common Object File Format)
7207 object files, so that @command{collect2} can assume this format and scan
7208 object files directly for dynamic constructor/destructor functions.
7210 @findex OBJECT_FORMAT_ROSE
7211 @item OBJECT_FORMAT_ROSE
7212 Define this macro if the system uses ROSE format object files, so that
7213 @command{collect2} can assume this format and scan object files directly
7214 for dynamic constructor/destructor functions.
7216 These macros are effective only in a native compiler; @command{collect2} as
7217 part of a cross compiler always uses @command{nm} for the target machine.
7219 @findex REAL_NM_FILE_NAME
7220 @item REAL_NM_FILE_NAME
7221 Define this macro as a C string constant containing the file name to use
7222 to execute @command{nm}. The default is to search the path normally for
7225 If your system supports shared libraries and has a program to list the
7226 dynamic dependencies of a given library or executable, you can define
7227 these macros to enable support for running initialization and
7228 termination functions in shared libraries:
7232 Define this macro to a C string constant containing the name of the program
7233 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7235 @findex PARSE_LDD_OUTPUT
7236 @item PARSE_LDD_OUTPUT (@var{ptr})
7237 Define this macro to be C code that extracts filenames from the output
7238 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7239 of type @code{char *} that points to the beginning of a line of output
7240 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7241 code must advance @var{ptr} to the beginning of the filename on that
7242 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7245 @node Instruction Output
7246 @subsection Output of Assembler Instructions
7248 @c prevent bad page break with this line
7249 This describes assembler instruction output.
7252 @findex REGISTER_NAMES
7253 @item REGISTER_NAMES
7254 A C initializer containing the assembler's names for the machine
7255 registers, each one as a C string constant. This is what translates
7256 register numbers in the compiler into assembler language.
7258 @findex ADDITIONAL_REGISTER_NAMES
7259 @item ADDITIONAL_REGISTER_NAMES
7260 If defined, a C initializer for an array of structures containing a name
7261 and a register number. This macro defines additional names for hard
7262 registers, thus allowing the @code{asm} option in declarations to refer
7263 to registers using alternate names.
7265 @findex ASM_OUTPUT_OPCODE
7266 @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7267 Define this macro if you are using an unusual assembler that
7268 requires different names for the machine instructions.
7270 The definition is a C statement or statements which output an
7271 assembler instruction opcode to the stdio stream @var{stream}. The
7272 macro-operand @var{ptr} is a variable of type @code{char *} which
7273 points to the opcode name in its ``internal'' form---the form that is
7274 written in the machine description. The definition should output the
7275 opcode name to @var{stream}, performing any translation you desire, and
7276 increment the variable @var{ptr} to point at the end of the opcode
7277 so that it will not be output twice.
7279 In fact, your macro definition may process less than the entire opcode
7280 name, or more than the opcode name; but if you want to process text
7281 that includes @samp{%}-sequences to substitute operands, you must take
7282 care of the substitution yourself. Just be sure to increment
7283 @var{ptr} over whatever text should not be output normally.
7285 @findex recog_data.operand
7286 If you need to look at the operand values, they can be found as the
7287 elements of @code{recog_data.operand}.
7289 If the macro definition does nothing, the instruction is output
7292 @findex FINAL_PRESCAN_INSN
7293 @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7294 If defined, a C statement to be executed just prior to the output of
7295 assembler code for @var{insn}, to modify the extracted operands so
7296 they will be output differently.
7298 Here the argument @var{opvec} is the vector containing the operands
7299 extracted from @var{insn}, and @var{noperands} is the number of
7300 elements of the vector which contain meaningful data for this insn.
7301 The contents of this vector are what will be used to convert the insn
7302 template into assembler code, so you can change the assembler output
7303 by changing the contents of the vector.
7305 This macro is useful when various assembler syntaxes share a single
7306 file of instruction patterns; by defining this macro differently, you
7307 can cause a large class of instructions to be output differently (such
7308 as with rearranged operands). Naturally, variations in assembler
7309 syntax affecting individual insn patterns ought to be handled by
7310 writing conditional output routines in those patterns.
7312 If this macro is not defined, it is equivalent to a null statement.
7314 @findex FINAL_PRESCAN_LABEL
7315 @item FINAL_PRESCAN_LABEL
7316 If defined, @code{FINAL_PRESCAN_INSN} will be called on each
7317 @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and
7318 @var{noperands} will be zero.
7320 @findex PRINT_OPERAND
7321 @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
7322 A C compound statement to output to stdio stream @var{stream} the
7323 assembler syntax for an instruction operand @var{x}. @var{x} is an
7326 @var{code} is a value that can be used to specify one of several ways
7327 of printing the operand. It is used when identical operands must be
7328 printed differently depending on the context. @var{code} comes from
7329 the @samp{%} specification that was used to request printing of the
7330 operand. If the specification was just @samp{%@var{digit}} then
7331 @var{code} is 0; if the specification was @samp{%@var{ltr}
7332 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
7335 If @var{x} is a register, this macro should print the register's name.
7336 The names can be found in an array @code{reg_names} whose type is
7337 @code{char *[]}. @code{reg_names} is initialized from
7338 @code{REGISTER_NAMES}.
7340 When the machine description has a specification @samp{%@var{punct}}
7341 (a @samp{%} followed by a punctuation character), this macro is called
7342 with a null pointer for @var{x} and the punctuation character for
7345 @findex PRINT_OPERAND_PUNCT_VALID_P
7346 @item PRINT_OPERAND_PUNCT_VALID_P (@var{code})
7347 A C expression which evaluates to true if @var{code} is a valid
7348 punctuation character for use in the @code{PRINT_OPERAND} macro. If
7349 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
7350 punctuation characters (except for the standard one, @samp{%}) are used
7353 @findex PRINT_OPERAND_ADDRESS
7354 @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
7355 A C compound statement to output to stdio stream @var{stream} the
7356 assembler syntax for an instruction operand that is a memory reference
7357 whose address is @var{x}. @var{x} is an RTL expression.
7359 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
7360 On some machines, the syntax for a symbolic address depends on the
7361 section that the address refers to. On these machines, define the hook
7362 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
7363 @code{symbol_ref}, and then check for it here. @xref{Assembler Format}.
7365 @findex DBR_OUTPUT_SEQEND
7366 @findex dbr_sequence_length
7367 @item DBR_OUTPUT_SEQEND(@var{file})
7368 A C statement, to be executed after all slot-filler instructions have
7369 been output. If necessary, call @code{dbr_sequence_length} to
7370 determine the number of slots filled in a sequence (zero if not
7371 currently outputting a sequence), to decide how many no-ops to output,
7374 Don't define this macro if it has nothing to do, but it is helpful in
7375 reading assembly output if the extent of the delay sequence is made
7376 explicit (e.g.@: with white space).
7378 @findex final_sequence
7379 Note that output routines for instructions with delay slots must be
7380 prepared to deal with not being output as part of a sequence
7381 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
7382 found.) The variable @code{final_sequence} is null when not
7383 processing a sequence, otherwise it contains the @code{sequence} rtx
7386 @findex REGISTER_PREFIX
7387 @findex LOCAL_LABEL_PREFIX
7388 @findex USER_LABEL_PREFIX
7389 @findex IMMEDIATE_PREFIX
7391 @item REGISTER_PREFIX
7392 @itemx LOCAL_LABEL_PREFIX
7393 @itemx USER_LABEL_PREFIX
7394 @itemx IMMEDIATE_PREFIX
7395 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
7396 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
7397 @file{final.c}). These are useful when a single @file{md} file must
7398 support multiple assembler formats. In that case, the various @file{tm.h}
7399 files can define these macros differently.
7401 @item ASM_FPRINTF_EXTENSIONS(@var{file}, @var{argptr}, @var{format})
7402 @findex ASM_FPRINTF_EXTENSIONS
7403 If defined this macro should expand to a series of @code{case}
7404 statements which will be parsed inside the @code{switch} statement of
7405 the @code{asm_fprintf} function. This allows targets to define extra
7406 printf formats which may useful when generating their assembler
7407 statements. Note that upper case letters are reserved for future
7408 generic extensions to asm_fprintf, and so are not available to target
7409 specific code. The output file is given by the parameter @var{file}.
7410 The varargs input pointer is @var{argptr} and the rest of the format
7411 string, starting the character after the one that is being switched
7412 upon, is pointed to by @var{format}.
7414 @findex ASSEMBLER_DIALECT
7415 @item ASSEMBLER_DIALECT
7416 If your target supports multiple dialects of assembler language (such as
7417 different opcodes), define this macro as a C expression that gives the
7418 numeric index of the assembler language dialect to use, with zero as the
7421 If this macro is defined, you may use constructs of the form
7423 @samp{@{option0|option1|option2@dots{}@}}
7426 in the output templates of patterns (@pxref{Output Template}) or in the
7427 first argument of @code{asm_fprintf}. This construct outputs
7428 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
7429 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
7430 within these strings retain their usual meaning. If there are fewer
7431 alternatives within the braces than the value of
7432 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
7434 If you do not define this macro, the characters @samp{@{}, @samp{|} and
7435 @samp{@}} do not have any special meaning when used in templates or
7436 operands to @code{asm_fprintf}.
7438 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
7439 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
7440 the variations in assembler language syntax with that mechanism. Define
7441 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
7442 if the syntax variant are larger and involve such things as different
7443 opcodes or operand order.
7445 @findex ASM_OUTPUT_REG_PUSH
7446 @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
7447 A C expression to output to @var{stream} some assembler code
7448 which will push hard register number @var{regno} onto the stack.
7449 The code need not be optimal, since this macro is used only when
7452 @findex ASM_OUTPUT_REG_POP
7453 @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
7454 A C expression to output to @var{stream} some assembler code
7455 which will pop hard register number @var{regno} off of the stack.
7456 The code need not be optimal, since this macro is used only when
7460 @node Dispatch Tables
7461 @subsection Output of Dispatch Tables
7463 @c prevent bad page break with this line
7464 This concerns dispatch tables.
7467 @cindex dispatch table
7468 @findex ASM_OUTPUT_ADDR_DIFF_ELT
7469 @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
7470 A C statement to output to the stdio stream @var{stream} an assembler
7471 pseudo-instruction to generate a difference between two labels.
7472 @var{value} and @var{rel} are the numbers of two internal labels. The
7473 definitions of these labels are output using
7474 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
7475 way here. For example,
7478 fprintf (@var{stream}, "\t.word L%d-L%d\n",
7479 @var{value}, @var{rel})
7482 You must provide this macro on machines where the addresses in a
7483 dispatch table are relative to the table's own address. If defined, GCC
7484 will also use this macro on all machines when producing PIC@.
7485 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
7486 mode and flags can be read.
7488 @findex ASM_OUTPUT_ADDR_VEC_ELT
7489 @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
7490 This macro should be provided on machines where the addresses
7491 in a dispatch table are absolute.
7493 The definition should be a C statement to output to the stdio stream
7494 @var{stream} an assembler pseudo-instruction to generate a reference to
7495 a label. @var{value} is the number of an internal label whose
7496 definition is output using @code{(*targetm.asm_out.internal_label)}.
7500 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
7503 @findex ASM_OUTPUT_CASE_LABEL
7504 @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
7505 Define this if the label before a jump-table needs to be output
7506 specially. The first three arguments are the same as for
7507 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
7508 jump-table which follows (a @code{jump_insn} containing an
7509 @code{addr_vec} or @code{addr_diff_vec}).
7511 This feature is used on system V to output a @code{swbeg} statement
7514 If this macro is not defined, these labels are output with
7515 @code{(*targetm.asm_out.internal_label)}.
7517 @findex ASM_OUTPUT_CASE_END
7518 @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
7519 Define this if something special must be output at the end of a
7520 jump-table. The definition should be a C statement to be executed
7521 after the assembler code for the table is written. It should write
7522 the appropriate code to stdio stream @var{stream}. The argument
7523 @var{table} is the jump-table insn, and @var{num} is the label-number
7524 of the preceding label.
7526 If this macro is not defined, nothing special is output at the end of
7530 @node Exception Region Output
7531 @subsection Assembler Commands for Exception Regions
7533 @c prevent bad page break with this line
7535 This describes commands marking the start and the end of an exception
7539 @findex EH_FRAME_SECTION_NAME
7540 @item EH_FRAME_SECTION_NAME
7541 If defined, a C string constant for the name of the section containing
7542 exception handling frame unwind information. If not defined, GCC will
7543 provide a default definition if the target supports named sections.
7544 @file{crtstuff.c} uses this macro to switch to the appropriate section.
7546 You should define this symbol if your target supports DWARF 2 frame
7547 unwind information and the default definition does not work.
7549 @findex EH_FRAME_IN_DATA_SECTION
7550 @item EH_FRAME_IN_DATA_SECTION
7551 If defined, DWARF 2 frame unwind information will be placed in the
7552 data section even though the target supports named sections. This
7553 might be necessary, for instance, if the system linker does garbage
7554 collection and sections cannot be marked as not to be collected.
7556 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
7559 @findex MASK_RETURN_ADDR
7560 @item MASK_RETURN_ADDR
7561 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
7562 that it does not contain any extraneous set bits in it.
7564 @findex DWARF2_UNWIND_INFO
7565 @item DWARF2_UNWIND_INFO
7566 Define this macro to 0 if your target supports DWARF 2 frame unwind
7567 information, but it does not yet work with exception handling.
7568 Otherwise, if your target supports this information (if it defines
7569 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
7570 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
7573 If this macro is defined to 1, the DWARF 2 unwinder will be the default
7574 exception handling mechanism; otherwise, @code{setjmp}/@code{longjmp} will be used by
7577 If this macro is defined to anything, the DWARF 2 unwinder will be used
7578 instead of inline unwinders and @code{__unwind_function} in the non-@code{setjmp} case.
7580 @findex DWARF_CIE_DATA_ALIGNMENT
7581 @item DWARF_CIE_DATA_ALIGNMENT
7582 This macro need only be defined if the target might save registers in the
7583 function prologue at an offset to the stack pointer that is not aligned to
7584 @code{UNITS_PER_WORD}. The definition should be the negative minimum
7585 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
7586 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
7587 the target supports DWARF 2 frame unwind information.
7591 @deftypefn {Target Hook} void TARGET_ASM_EXCEPTION_SECTION ()
7592 If defined, a function that switches to the section in which the main
7593 exception table is to be placed (@pxref{Sections}). The default is a
7594 function that switches to a section named @code{.gcc_except_table} on
7595 machines that support named sections via
7596 @code{TARGET_ASM_NAMED_SECTION}, otherwise if @option{-fpic} or
7597 @option{-fPIC} is in effect, the @code{data_section}, otherwise the
7598 @code{readonly_data_section}.
7601 @deftypefn {Target Hook} void TARGET_ASM_EH_FRAME_SECTION ()
7602 If defined, a function that switches to the section in which the DWARF 2
7603 frame unwind information to be placed (@pxref{Sections}). The default
7604 is a function that outputs a standard GAS section directive, if
7605 @code{EH_FRAME_SECTION_NAME} is defined, or else a data section
7606 directive followed by a synthetic label.
7609 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
7610 Contains the value true if the target should add a zero word onto the
7611 end of a Dwarf-2 frame info section when used for exception handling.
7612 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
7616 @node Alignment Output
7617 @subsection Assembler Commands for Alignment
7619 @c prevent bad page break with this line
7620 This describes commands for alignment.
7624 @item JUMP_ALIGN (@var{label})
7625 The alignment (log base 2) to put in front of @var{label}, which is
7626 a common destination of jumps and has no fallthru incoming edge.
7628 This macro need not be defined if you don't want any special alignment
7629 to be done at such a time. Most machine descriptions do not currently
7632 Unless it's necessary to inspect the @var{label} parameter, it is better
7633 to set the variable @var{align_jumps} in the target's
7634 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7635 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
7637 @findex LABEL_ALIGN_AFTER_BARRIER
7638 @item LABEL_ALIGN_AFTER_BARRIER (@var{label})
7639 The alignment (log base 2) to put in front of @var{label}, which follows
7642 This macro need not be defined if you don't want any special alignment
7643 to be done at such a time. Most machine descriptions do not currently
7646 @findex LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7647 @item LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7648 The maximum number of bytes to skip when applying
7649 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
7650 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7653 @item LOOP_ALIGN (@var{label})
7654 The alignment (log base 2) to put in front of @var{label}, which follows
7655 a @code{NOTE_INSN_LOOP_BEG} note.
7657 This macro need not be defined if you don't want any special alignment
7658 to be done at such a time. Most machine descriptions do not currently
7661 Unless it's necessary to inspect the @var{label} parameter, it is better
7662 to set the variable @code{align_loops} in the target's
7663 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7664 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
7666 @findex LOOP_ALIGN_MAX_SKIP
7667 @item LOOP_ALIGN_MAX_SKIP
7668 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
7669 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7672 @item LABEL_ALIGN (@var{label})
7673 The alignment (log base 2) to put in front of @var{label}.
7674 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
7675 the maximum of the specified values is used.
7677 Unless it's necessary to inspect the @var{label} parameter, it is better
7678 to set the variable @code{align_labels} in the target's
7679 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7680 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
7682 @findex LABEL_ALIGN_MAX_SKIP
7683 @item LABEL_ALIGN_MAX_SKIP
7684 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
7685 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7687 @findex ASM_OUTPUT_SKIP
7688 @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
7689 A C statement to output to the stdio stream @var{stream} an assembler
7690 instruction to advance the location counter by @var{nbytes} bytes.
7691 Those bytes should be zero when loaded. @var{nbytes} will be a C
7692 expression of type @code{int}.
7694 @findex ASM_NO_SKIP_IN_TEXT
7695 @item ASM_NO_SKIP_IN_TEXT
7696 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
7697 text section because it fails to put zeros in the bytes that are skipped.
7698 This is true on many Unix systems, where the pseudo--op to skip bytes
7699 produces no-op instructions rather than zeros when used in the text
7702 @findex ASM_OUTPUT_ALIGN
7703 @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
7704 A C statement to output to the stdio stream @var{stream} an assembler
7705 command to advance the location counter to a multiple of 2 to the
7706 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
7708 @findex ASM_OUTPUT_ALIGN_WITH_NOP
7709 @item ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
7710 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
7711 for padding, if necessary.
7713 @findex ASM_OUTPUT_MAX_SKIP_ALIGN
7714 @item ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
7715 A C statement to output to the stdio stream @var{stream} an assembler
7716 command to advance the location counter to a multiple of 2 to the
7717 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
7718 satisfy the alignment request. @var{power} and @var{max_skip} will be
7719 a C expression of type @code{int}.
7723 @node Debugging Info
7724 @section Controlling Debugging Information Format
7726 @c prevent bad page break with this line
7727 This describes how to specify debugging information.
7730 * All Debuggers:: Macros that affect all debugging formats uniformly.
7731 * DBX Options:: Macros enabling specific options in DBX format.
7732 * DBX Hooks:: Hook macros for varying DBX format.
7733 * File Names and DBX:: Macros controlling output of file names in DBX format.
7734 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
7735 * VMS Debug:: Macros for VMS debug format.
7739 @subsection Macros Affecting All Debugging Formats
7741 @c prevent bad page break with this line
7742 These macros affect all debugging formats.
7745 @findex DBX_REGISTER_NUMBER
7746 @item DBX_REGISTER_NUMBER (@var{regno})
7747 A C expression that returns the DBX register number for the compiler
7748 register number @var{regno}. In the default macro provided, the value
7749 of this expression will be @var{regno} itself. But sometimes there are
7750 some registers that the compiler knows about and DBX does not, or vice
7751 versa. In such cases, some register may need to have one number in the
7752 compiler and another for DBX@.
7754 If two registers have consecutive numbers inside GCC, and they can be
7755 used as a pair to hold a multiword value, then they @emph{must} have
7756 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
7757 Otherwise, debuggers will be unable to access such a pair, because they
7758 expect register pairs to be consecutive in their own numbering scheme.
7760 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
7761 does not preserve register pairs, then what you must do instead is
7762 redefine the actual register numbering scheme.
7764 @findex DEBUGGER_AUTO_OFFSET
7765 @item DEBUGGER_AUTO_OFFSET (@var{x})
7766 A C expression that returns the integer offset value for an automatic
7767 variable having address @var{x} (an RTL expression). The default
7768 computation assumes that @var{x} is based on the frame-pointer and
7769 gives the offset from the frame-pointer. This is required for targets
7770 that produce debugging output for DBX or COFF-style debugging output
7771 for SDB and allow the frame-pointer to be eliminated when the
7772 @option{-g} options is used.
7774 @findex DEBUGGER_ARG_OFFSET
7775 @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
7776 A C expression that returns the integer offset value for an argument
7777 having address @var{x} (an RTL expression). The nominal offset is
7780 @findex PREFERRED_DEBUGGING_TYPE
7781 @item PREFERRED_DEBUGGING_TYPE
7782 A C expression that returns the type of debugging output GCC should
7783 produce when the user specifies just @option{-g}. Define
7784 this if you have arranged for GCC to support more than one format of
7785 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7786 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
7787 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
7789 When the user specifies @option{-ggdb}, GCC normally also uses the
7790 value of this macro to select the debugging output format, but with two
7791 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined and
7792 @code{LINKER_DOES_NOT_WORK_WITH_DWARF2} is not defined, GCC uses the
7793 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7794 defined, GCC uses @code{DBX_DEBUG}.
7796 The value of this macro only affects the default debugging output; the
7797 user can always get a specific type of output by using @option{-gstabs},
7798 @option{-gcoff}, @option{-gdwarf-1}, @option{-gdwarf-2}, @option{-gxcoff},
7803 @subsection Specific Options for DBX Output
7805 @c prevent bad page break with this line
7806 These are specific options for DBX output.
7809 @findex DBX_DEBUGGING_INFO
7810 @item DBX_DEBUGGING_INFO
7811 Define this macro if GCC should produce debugging output for DBX
7812 in response to the @option{-g} option.
7814 @findex XCOFF_DEBUGGING_INFO
7815 @item XCOFF_DEBUGGING_INFO
7816 Define this macro if GCC should produce XCOFF format debugging output
7817 in response to the @option{-g} option. This is a variant of DBX format.
7819 @findex DEFAULT_GDB_EXTENSIONS
7820 @item DEFAULT_GDB_EXTENSIONS
7821 Define this macro to control whether GCC should by default generate
7822 GDB's extended version of DBX debugging information (assuming DBX-format
7823 debugging information is enabled at all). If you don't define the
7824 macro, the default is 1: always generate the extended information
7825 if there is any occasion to.
7827 @findex DEBUG_SYMS_TEXT
7828 @item DEBUG_SYMS_TEXT
7829 Define this macro if all @code{.stabs} commands should be output while
7830 in the text section.
7832 @findex ASM_STABS_OP
7834 A C string constant, including spacing, naming the assembler pseudo op to
7835 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
7836 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
7837 applies only to DBX debugging information format.
7839 @findex ASM_STABD_OP
7841 A C string constant, including spacing, naming the assembler pseudo op to
7842 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
7843 value is the current location. If you don't define this macro,
7844 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
7847 @findex ASM_STABN_OP
7849 A C string constant, including spacing, naming the assembler pseudo op to
7850 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
7851 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
7852 macro applies only to DBX debugging information format.
7854 @findex DBX_NO_XREFS
7856 Define this macro if DBX on your system does not support the construct
7857 @samp{xs@var{tagname}}. On some systems, this construct is used to
7858 describe a forward reference to a structure named @var{tagname}.
7859 On other systems, this construct is not supported at all.
7861 @findex DBX_CONTIN_LENGTH
7862 @item DBX_CONTIN_LENGTH
7863 A symbol name in DBX-format debugging information is normally
7864 continued (split into two separate @code{.stabs} directives) when it
7865 exceeds a certain length (by default, 80 characters). On some
7866 operating systems, DBX requires this splitting; on others, splitting
7867 must not be done. You can inhibit splitting by defining this macro
7868 with the value zero. You can override the default splitting-length by
7869 defining this macro as an expression for the length you desire.
7871 @findex DBX_CONTIN_CHAR
7872 @item DBX_CONTIN_CHAR
7873 Normally continuation is indicated by adding a @samp{\} character to
7874 the end of a @code{.stabs} string when a continuation follows. To use
7875 a different character instead, define this macro as a character
7876 constant for the character you want to use. Do not define this macro
7877 if backslash is correct for your system.
7879 @findex DBX_STATIC_STAB_DATA_SECTION
7880 @item DBX_STATIC_STAB_DATA_SECTION
7881 Define this macro if it is necessary to go to the data section before
7882 outputting the @samp{.stabs} pseudo-op for a non-global static
7885 @findex DBX_TYPE_DECL_STABS_CODE
7886 @item DBX_TYPE_DECL_STABS_CODE
7887 The value to use in the ``code'' field of the @code{.stabs} directive
7888 for a typedef. The default is @code{N_LSYM}.
7890 @findex DBX_STATIC_CONST_VAR_CODE
7891 @item DBX_STATIC_CONST_VAR_CODE
7892 The value to use in the ``code'' field of the @code{.stabs} directive
7893 for a static variable located in the text section. DBX format does not
7894 provide any ``right'' way to do this. The default is @code{N_FUN}.
7896 @findex DBX_REGPARM_STABS_CODE
7897 @item DBX_REGPARM_STABS_CODE
7898 The value to use in the ``code'' field of the @code{.stabs} directive
7899 for a parameter passed in registers. DBX format does not provide any
7900 ``right'' way to do this. The default is @code{N_RSYM}.
7902 @findex DBX_REGPARM_STABS_LETTER
7903 @item DBX_REGPARM_STABS_LETTER
7904 The letter to use in DBX symbol data to identify a symbol as a parameter
7905 passed in registers. DBX format does not customarily provide any way to
7906 do this. The default is @code{'P'}.
7908 @findex DBX_MEMPARM_STABS_LETTER
7909 @item DBX_MEMPARM_STABS_LETTER
7910 The letter to use in DBX symbol data to identify a symbol as a stack
7911 parameter. The default is @code{'p'}.
7913 @findex DBX_FUNCTION_FIRST
7914 @item DBX_FUNCTION_FIRST
7915 Define this macro if the DBX information for a function and its
7916 arguments should precede the assembler code for the function. Normally,
7917 in DBX format, the debugging information entirely follows the assembler
7920 @findex DBX_LBRAC_FIRST
7921 @item DBX_LBRAC_FIRST
7922 Define this macro if the @code{N_LBRAC} symbol for a block should
7923 precede the debugging information for variables and functions defined in
7924 that block. Normally, in DBX format, the @code{N_LBRAC} symbol comes
7927 @findex DBX_BLOCKS_FUNCTION_RELATIVE
7928 @item DBX_BLOCKS_FUNCTION_RELATIVE
7929 Define this macro if the value of a symbol describing the scope of a
7930 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
7931 of the enclosing function. Normally, GCC uses an absolute address.
7933 @findex DBX_USE_BINCL
7935 Define this macro if GCC should generate @code{N_BINCL} and
7936 @code{N_EINCL} stabs for included header files, as on Sun systems. This
7937 macro also directs GCC to output a type number as a pair of a file
7938 number and a type number within the file. Normally, GCC does not
7939 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
7940 number for a type number.
7944 @subsection Open-Ended Hooks for DBX Format
7946 @c prevent bad page break with this line
7947 These are hooks for DBX format.
7950 @findex DBX_OUTPUT_LBRAC
7951 @item DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
7952 Define this macro to say how to output to @var{stream} the debugging
7953 information for the start of a scope level for variable names. The
7954 argument @var{name} is the name of an assembler symbol (for use with
7955 @code{assemble_name}) whose value is the address where the scope begins.
7957 @findex DBX_OUTPUT_RBRAC
7958 @item DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
7959 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
7961 @findex DBX_OUTPUT_NFUN
7962 @item DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
7963 Define this macro if the target machine requires special handling to
7964 output an @code{N_FUN} entry for the function @var{decl}.
7966 @findex DBX_OUTPUT_ENUM
7967 @item DBX_OUTPUT_ENUM (@var{stream}, @var{type})
7968 Define this macro if the target machine requires special handling to
7969 output an enumeration type. The definition should be a C statement
7970 (sans semicolon) to output the appropriate information to @var{stream}
7971 for the type @var{type}.
7973 @findex DBX_OUTPUT_FUNCTION_END
7974 @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
7975 Define this macro if the target machine requires special output at the
7976 end of the debugging information for a function. The definition should
7977 be a C statement (sans semicolon) to output the appropriate information
7978 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
7981 @findex DBX_OUTPUT_STANDARD_TYPES
7982 @item DBX_OUTPUT_STANDARD_TYPES (@var{syms})
7983 Define this macro if you need to control the order of output of the
7984 standard data types at the beginning of compilation. The argument
7985 @var{syms} is a @code{tree} which is a chain of all the predefined
7986 global symbols, including names of data types.
7988 Normally, DBX output starts with definitions of the types for integers
7989 and characters, followed by all the other predefined types of the
7990 particular language in no particular order.
7992 On some machines, it is necessary to output different particular types
7993 first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
7994 those symbols in the necessary order. Any predefined types that you
7995 don't explicitly output will be output afterward in no particular order.
7997 Be careful not to define this macro so that it works only for C@. There
7998 are no global variables to access most of the built-in types, because
7999 another language may have another set of types. The way to output a
8000 particular type is to look through @var{syms} to see if you can find it.
8006 for (decl = syms; decl; decl = TREE_CHAIN (decl))
8007 if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
8009 dbxout_symbol (decl);
8015 This does nothing if the expected type does not exist.
8017 See the function @code{init_decl_processing} in @file{c-decl.c} to find
8018 the names to use for all the built-in C types.
8020 Here is another way of finding a particular type:
8022 @c this is still overfull. --mew 10feb93
8026 for (decl = syms; decl; decl = TREE_CHAIN (decl))
8027 if (TREE_CODE (decl) == TYPE_DECL
8028 && (TREE_CODE (TREE_TYPE (decl))
8030 && TYPE_PRECISION (TREE_TYPE (decl)) == 16
8031 && TYPE_UNSIGNED (TREE_TYPE (decl)))
8033 /* @r{This must be @code{unsigned short}.} */
8034 dbxout_symbol (decl);
8040 @findex NO_DBX_FUNCTION_END
8041 @item NO_DBX_FUNCTION_END
8042 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
8043 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
8044 On those machines, define this macro to turn this feature off without
8045 disturbing the rest of the gdb extensions.
8049 @node File Names and DBX
8050 @subsection File Names in DBX Format
8052 @c prevent bad page break with this line
8053 This describes file names in DBX format.
8056 @findex DBX_WORKING_DIRECTORY
8057 @item DBX_WORKING_DIRECTORY
8058 Define this if DBX wants to have the current directory recorded in each
8061 Note that the working directory is always recorded if GDB extensions are
8064 @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME
8065 @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
8066 A C statement to output DBX debugging information to the stdio stream
8067 @var{stream} which indicates that file @var{name} is the main source
8068 file---the file specified as the input file for compilation.
8069 This macro is called only once, at the beginning of compilation.
8071 This macro need not be defined if the standard form of output
8072 for DBX debugging information is appropriate.
8074 @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY
8075 @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
8076 A C statement to output DBX debugging information to the stdio stream
8077 @var{stream} which indicates that the current directory during
8078 compilation is named @var{name}.
8080 This macro need not be defined if the standard form of output
8081 for DBX debugging information is appropriate.
8083 @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END
8084 @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
8085 A C statement to output DBX debugging information at the end of
8086 compilation of the main source file @var{name}.
8088 If you don't define this macro, nothing special is output at the end
8089 of compilation, which is correct for most machines.
8091 @findex DBX_OUTPUT_SOURCE_FILENAME
8092 @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
8093 A C statement to output DBX debugging information to the stdio stream
8094 @var{stream} which indicates that file @var{name} is the current source
8095 file. This output is generated each time input shifts to a different
8096 source file as a result of @samp{#include}, the end of an included file,
8097 or a @samp{#line} command.
8099 This macro need not be defined if the standard form of output
8100 for DBX debugging information is appropriate.
8105 @subsection Macros for SDB and DWARF Output
8107 @c prevent bad page break with this line
8108 Here are macros for SDB and DWARF output.
8111 @findex SDB_DEBUGGING_INFO
8112 @item SDB_DEBUGGING_INFO
8113 Define this macro if GCC should produce COFF-style debugging output
8114 for SDB in response to the @option{-g} option.
8116 @findex DWARF_DEBUGGING_INFO
8117 @item DWARF_DEBUGGING_INFO
8118 Define this macro if GCC should produce dwarf format debugging output
8119 in response to the @option{-g} option.
8121 @findex DWARF2_DEBUGGING_INFO
8122 @item DWARF2_DEBUGGING_INFO
8123 Define this macro if GCC should produce dwarf version 2 format
8124 debugging output in response to the @option{-g} option.
8126 To support optional call frame debugging information, you must also
8127 define @code{INCOMING_RETURN_ADDR_RTX} and either set
8128 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
8129 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
8130 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
8132 @findex DWARF2_FRAME_INFO
8133 @item DWARF2_FRAME_INFO
8134 Define this macro to a nonzero value if GCC should always output
8135 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
8136 (@pxref{Exception Region Output} is nonzero, GCC will output this
8137 information not matter how you define @code{DWARF2_FRAME_INFO}.
8139 @findex LINKER_DOES_NOT_WORK_WITH_DWARF2
8140 @item LINKER_DOES_NOT_WORK_WITH_DWARF2
8141 Define this macro if the linker does not work with Dwarf version 2.
8142 Normally, if the user specifies only @option{-ggdb} GCC will use Dwarf
8143 version 2 if available; this macro disables this. See the description
8144 of the @code{PREFERRED_DEBUGGING_TYPE} macro for more details.
8146 @findex DWARF2_GENERATE_TEXT_SECTION_LABEL
8147 @item DWARF2_GENERATE_TEXT_SECTION_LABEL
8148 By default, the Dwarf 2 debugging information generator will generate a
8149 label to mark the beginning of the text section. If it is better simply
8150 to use the name of the text section itself, rather than an explicit label,
8151 to indicate the beginning of the text section, define this macro to zero.
8153 @findex DWARF2_ASM_LINE_DEBUG_INFO
8154 @item DWARF2_ASM_LINE_DEBUG_INFO
8155 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8156 line debug info sections. This will result in much more compact line number
8157 tables, and hence is desirable if it works.
8159 @findex ASM_OUTPUT_DWARF_DELTA
8160 @item ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8161 A C statement to issue assembly directives that create a difference
8162 between the two given labels, using an integer of the given size.
8164 @findex ASM_OUTPUT_DWARF_OFFSET
8165 @item ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label})
8166 A C statement to issue assembly directives that create a
8167 section-relative reference to the given label, using an integer of the
8170 @findex ASM_OUTPUT_DWARF_PCREL
8171 @item ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8172 A C statement to issue assembly directives that create a self-relative
8173 reference to the given label, using an integer of the given size.
8175 @findex PUT_SDB_@dots{}
8176 @item PUT_SDB_@dots{}
8177 Define these macros to override the assembler syntax for the special
8178 SDB assembler directives. See @file{sdbout.c} for a list of these
8179 macros and their arguments. If the standard syntax is used, you need
8180 not define them yourself.
8184 Some assemblers do not support a semicolon as a delimiter, even between
8185 SDB assembler directives. In that case, define this macro to be the
8186 delimiter to use (usually @samp{\n}). It is not necessary to define
8187 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8190 @findex SDB_GENERATE_FAKE
8191 @item SDB_GENERATE_FAKE
8192 Define this macro to override the usual method of constructing a dummy
8193 name for anonymous structure and union types. See @file{sdbout.c} for
8196 @findex SDB_ALLOW_UNKNOWN_REFERENCES
8197 @item SDB_ALLOW_UNKNOWN_REFERENCES
8198 Define this macro to allow references to unknown structure,
8199 union, or enumeration tags to be emitted. Standard COFF does not
8200 allow handling of unknown references, MIPS ECOFF has support for
8203 @findex SDB_ALLOW_FORWARD_REFERENCES
8204 @item SDB_ALLOW_FORWARD_REFERENCES
8205 Define this macro to allow references to structure, union, or
8206 enumeration tags that have not yet been seen to be handled. Some
8207 assemblers choke if forward tags are used, while some require it.
8212 @subsection Macros for VMS Debug Format
8214 @c prevent bad page break with this line
8215 Here are macros for VMS debug format.
8218 @findex VMS_DEBUGGING_INFO
8219 @item VMS_DEBUGGING_INFO
8220 Define this macro if GCC should produce debugging output for VMS
8221 in response to the @option{-g} option. The default behavior for VMS
8222 is to generate minimal debug info for a traceback in the absence of
8223 @option{-g} unless explicitly overridden with @option{-g0}. This
8224 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8225 @code{OVERRIDE_OPTIONS}.
8228 @node Floating Point
8229 @section Cross Compilation and Floating Point
8230 @cindex cross compilation and floating point
8231 @cindex floating point and cross compilation
8233 While all modern machines use twos-complement representation for integers,
8234 there are a variety of representations for floating point numbers. This
8235 means that in a cross-compiler the representation of floating point numbers
8236 in the compiled program may be different from that used in the machine
8237 doing the compilation.
8239 Because different representation systems may offer different amounts of
8240 range and precision, all floating point constants must be represented in
8241 the target machine's format. Therefore, the cross compiler cannot
8242 safely use the host machine's floating point arithmetic; it must emulate
8243 the target's arithmetic. To ensure consistency, GCC always uses
8244 emulation to work with floating point values, even when the host and
8245 target floating point formats are identical.
8247 The following macros are provided by @file{real.h} for the compiler to
8248 use. All parts of the compiler which generate or optimize
8249 floating-point calculations must use these macros. They may evaluate
8250 their operands more than once, so operands must not have side effects.
8252 @defmac REAL_VALUE_TYPE
8253 The C data type to be used to hold a floating point value in the target
8254 machine's format. Typically this is a @code{struct} containing an
8255 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8259 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8260 Compares for equality the two values, @var{x} and @var{y}. If the target
8261 floating point format supports negative zeroes and/or NaNs,
8262 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8263 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8266 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8267 Tests whether @var{x} is less than @var{y}.
8270 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8271 Truncates @var{x} to a signed integer, rounding toward zero.
8274 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8275 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8276 @var{x} is negative, returns zero.
8279 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8280 Converts @var{string} into a floating point number in the target machine's
8281 representation for mode @var{mode}. This routine can handle both
8282 decimal and hexadecimal floating point constants, using the syntax
8283 defined by the C language for both.
8286 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8287 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8290 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8291 Determines whether @var{x} represents infinity (positive or negative).
8294 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8295 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8298 @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})
8299 Calculates an arithmetic operation on the two floating point values
8300 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8303 The operation to be performed is specified by @var{code}. Only the
8304 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8305 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8307 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8308 target's floating point format cannot represent infinity, it will call
8309 @code{abort}. Callers should check for this situation first, using
8310 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8313 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8314 Returns the negative of the floating point value @var{x}.
8317 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8318 Returns the absolute value of @var{x}.
8321 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8322 Truncates the floating point value @var{x} to fit in @var{mode}. The
8323 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8324 appropriate bit pattern to be output asa floating constant whose
8325 precision accords with mode @var{mode}.
8328 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8329 Converts a floating point value @var{x} into a double-precision integer
8330 which is then stored into @var{low} and @var{high}. If the value is not
8331 integral, it is truncated.
8334 @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})
8335 @findex REAL_VALUE_FROM_INT
8336 Converts a double-precision integer found in @var{low} and @var{high},
8337 into a floating point value which is then stored into @var{x}. The
8338 value is truncated to fit in mode @var{mode}.
8341 @node Mode Switching
8342 @section Mode Switching Instructions
8343 @cindex mode switching
8344 The following macros control mode switching optimizations:
8347 @findex OPTIMIZE_MODE_SWITCHING
8348 @item OPTIMIZE_MODE_SWITCHING (@var{entity})
8349 Define this macro if the port needs extra instructions inserted for mode
8350 switching in an optimizing compilation.
8352 For an example, the SH4 can perform both single and double precision
8353 floating point operations, but to perform a single precision operation,
8354 the FPSCR PR bit has to be cleared, while for a double precision
8355 operation, this bit has to be set. Changing the PR bit requires a general
8356 purpose register as a scratch register, hence these FPSCR sets have to
8357 be inserted before reload, i.e.@: you can't put this into instruction emitting
8358 or @code{MACHINE_DEPENDENT_REORG}.
8360 You can have multiple entities that are mode-switched, and select at run time
8361 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
8362 return nonzero for any @var{entity} that needs mode-switching.
8363 If you define this macro, you also have to define
8364 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
8365 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
8366 @code{NORMAL_MODE} is optional.
8368 @findex NUM_MODES_FOR_MODE_SWITCHING
8369 @item NUM_MODES_FOR_MODE_SWITCHING
8370 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
8371 initializer for an array of integers. Each initializer element
8372 N refers to an entity that needs mode switching, and specifies the number
8373 of different modes that might need to be set for this entity.
8374 The position of the initializer in the initializer - starting counting at
8375 zero - determines the integer that is used to refer to the mode-switched
8377 In macros that take mode arguments / yield a mode result, modes are
8378 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
8379 switch is needed / supplied.
8382 @item MODE_NEEDED (@var{entity}, @var{insn})
8383 @var{entity} is an integer specifying a mode-switched entity. If
8384 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
8385 return an integer value not larger than the corresponding element in
8386 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
8387 be switched into prior to the execution of @var{insn}.
8390 @item NORMAL_MODE (@var{entity})
8391 If this macro is defined, it is evaluated for every @var{entity} that needs
8392 mode switching. It should evaluate to an integer, which is a mode that
8393 @var{entity} is assumed to be switched to at function entry and exit.
8395 @findex MODE_PRIORITY_TO_MODE
8396 @item MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
8397 This macro specifies the order in which modes for @var{entity} are processed.
8398 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
8399 lowest. The value of the macro should be an integer designating a mode
8400 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
8401 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
8402 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
8404 @findex EMIT_MODE_SET
8405 @item EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
8406 Generate one or more insns to set @var{entity} to @var{mode}.
8407 @var{hard_reg_live} is the set of hard registers live at the point where
8408 the insn(s) are to be inserted.
8411 @node Target Attributes
8412 @section Defining target-specific uses of @code{__attribute__}
8413 @cindex target attributes
8414 @cindex machine attributes
8415 @cindex attributes, target-specific
8417 Target-specific attributes may be defined for functions, data and types.
8418 These are described using the following target hooks; they also need to
8419 be documented in @file{extend.texi}.
8421 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
8422 If defined, this target hook points to an array of @samp{struct
8423 attribute_spec} (defined in @file{tree.h}) specifying the machine
8424 specific attributes for this target and some of the restrictions on the
8425 entities to which these attributes are applied and the arguments they
8429 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8430 If defined, this target hook is a function which returns zero if the attributes on
8431 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8432 and two if they are nearly compatible (which causes a warning to be
8433 generated). If this is not defined, machine-specific attributes are
8434 supposed always to be compatible.
8437 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
8438 If defined, this target hook is a function which assigns default attributes to
8439 newly defined @var{type}.
8442 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8443 Define this target hook if the merging of type attributes needs special
8444 handling. If defined, the result is a list of the combined
8445 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
8446 that @code{comptypes} has already been called and returned 1. This
8447 function may call @code{merge_attributes} to handle machine-independent
8451 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
8452 Define this target hook if the merging of decl attributes needs special
8453 handling. If defined, the result is a list of the combined
8454 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
8455 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
8456 when this is needed are when one attribute overrides another, or when an
8457 attribute is nullified by a subsequent definition. This function may
8458 call @code{merge_attributes} to handle machine-independent merging.
8460 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
8461 If the only target-specific handling you require is @samp{dllimport} for
8462 Windows targets, you should define the macro
8463 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. This links in a function
8464 called @code{merge_dllimport_decl_attributes} which can then be defined
8465 as the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. This is done
8466 in @file{i386/cygwin.h} and @file{i386/i386.c}, for example.
8469 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
8470 Define this target hook if you want to be able to add attributes to a decl
8471 when it is being created. This is normally useful for back ends which
8472 wish to implement a pragma by using the attributes which correspond to
8473 the pragma's effect. The @var{node} argument is the decl which is being
8474 created. The @var{attr_ptr} argument is a pointer to the attribute list
8475 for this decl. The list itself should not be modified, since it may be
8476 shared with other decls, but attributes may be chained on the head of
8477 the list and @code{*@var{attr_ptr}} modified to point to the new
8478 attributes, or a copy of the list may be made if further changes are
8482 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
8484 This target hook returns @code{true} if it is ok to inline @var{fndecl}
8485 into the current function, despite its having target-specific
8486 attributes, @code{false} otherwise. By default, if a function has a
8487 target specific attribute attached to it, it will not be inlined.
8490 @node MIPS Coprocessors
8491 @section Defining coprocessor specifics for MIPS targets.
8492 @cindex MIPS coprocessor-definition macros
8494 The MIPS specification allows MIPS implementations to have as many as 4
8495 coprocessors, each with as many as 32 private registers. gcc supports
8496 accessing these registers and transferring values between the registers
8497 and memory using asm-ized variables. For example:
8500 register unsigned int cp0count asm ("c0r1");
8506 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
8507 names may be added as described below, or the default names may be
8508 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
8510 Coprocessor registers are assumed to be epilogue-used; sets to them will
8511 be preserved even if it does not appear that the register is used again
8512 later in the function.
8514 Another note: according to the MIPS spec, coprocessor 1 (if present) is
8515 the FPU. One accesses COP1 registers through standard mips
8516 floating-point support; they are not included in this mechanism.
8518 There is one macro used in defining the MIPS coprocessor interface which
8519 you may want to override in subtargets; it is described below.
8523 @item ALL_COP_ADDITIONAL_REGISTER_NAMES
8524 @findex ALL_COP_ADDITIONAL_REGISTER_NAMES
8525 A comma-separated list (with leading comma) of pairs describing the
8526 alternate names of coprocessor registers. The format of each entry should be
8528 @{ @var{alternatename}, @var{register_number}@}
8535 @section Miscellaneous Parameters
8536 @cindex parameters, miscellaneous
8538 @c prevent bad page break with this line
8539 Here are several miscellaneous parameters.
8542 @item PREDICATE_CODES
8543 @findex PREDICATE_CODES
8544 Define this if you have defined special-purpose predicates in the file
8545 @file{@var{machine}.c}. This macro is called within an initializer of an
8546 array of structures. The first field in the structure is the name of a
8547 predicate and the second field is an array of rtl codes. For each
8548 predicate, list all rtl codes that can be in expressions matched by the
8549 predicate. The list should have a trailing comma. Here is an example
8550 of two entries in the list for a typical RISC machine:
8553 #define PREDICATE_CODES \
8554 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
8555 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
8558 Defining this macro does not affect the generated code (however,
8559 incorrect definitions that omit an rtl code that may be matched by the
8560 predicate can cause the compiler to malfunction). Instead, it allows
8561 the table built by @file{genrecog} to be more compact and efficient,
8562 thus speeding up the compiler. The most important predicates to include
8563 in the list specified by this macro are those used in the most insn
8566 For each predicate function named in @code{PREDICATE_CODES}, a
8567 declaration will be generated in @file{insn-codes.h}.
8569 @item SPECIAL_MODE_PREDICATES
8570 @findex SPECIAL_MODE_PREDICATES
8571 Define this if you have special predicates that know special things
8572 about modes. Genrecog will warn about certain forms of
8573 @code{match_operand} without a mode; if the operand predicate is
8574 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
8577 Here is an example from the IA-32 port (@code{ext_register_operand}
8578 specially checks for @code{HImode} or @code{SImode} in preparation
8579 for a byte extraction from @code{%ah} etc.).
8582 #define SPECIAL_MODE_PREDICATES \
8583 "ext_register_operand",
8586 @findex CASE_VECTOR_MODE
8587 @item CASE_VECTOR_MODE
8588 An alias for a machine mode name. This is the machine mode that
8589 elements of a jump-table should have.
8591 @findex CASE_VECTOR_SHORTEN_MODE
8592 @item CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
8593 Optional: return the preferred mode for an @code{addr_diff_vec}
8594 when the minimum and maximum offset are known. If you define this,
8595 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
8596 To make this work, you also have to define @code{INSN_ALIGN} and
8597 make the alignment for @code{addr_diff_vec} explicit.
8598 The @var{body} argument is provided so that the offset_unsigned and scale
8599 flags can be updated.
8601 @findex CASE_VECTOR_PC_RELATIVE
8602 @item CASE_VECTOR_PC_RELATIVE
8603 Define this macro to be a C expression to indicate when jump-tables
8604 should contain relative addresses. If jump-tables never contain
8605 relative addresses, then you need not define this macro.
8607 @findex CASE_DROPS_THROUGH
8608 @item CASE_DROPS_THROUGH
8609 Define this if control falls through a @code{case} insn when the index
8610 value is out of range. This means the specified default-label is
8611 actually ignored by the @code{case} insn proper.
8613 @findex CASE_VALUES_THRESHOLD
8614 @item CASE_VALUES_THRESHOLD
8615 Define this to be the smallest number of different values for which it
8616 is best to use a jump-table instead of a tree of conditional branches.
8617 The default is four for machines with a @code{casesi} instruction and
8618 five otherwise. This is best for most machines.
8620 @findex WORD_REGISTER_OPERATIONS
8621 @item WORD_REGISTER_OPERATIONS
8622 Define this macro if operations between registers with integral mode
8623 smaller than a word are always performed on the entire register.
8624 Most RISC machines have this property and most CISC machines do not.
8626 @findex LOAD_EXTEND_OP
8627 @item LOAD_EXTEND_OP (@var{mode})
8628 Define this macro to be a C expression indicating when insns that read
8629 memory in @var{mode}, an integral mode narrower than a word, set the
8630 bits outside of @var{mode} to be either the sign-extension or the
8631 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
8632 of @var{mode} for which the
8633 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
8634 @code{NIL} for other modes.
8636 This macro is not called with @var{mode} non-integral or with a width
8637 greater than or equal to @code{BITS_PER_WORD}, so you may return any
8638 value in this case. Do not define this macro if it would always return
8639 @code{NIL}. On machines where this macro is defined, you will normally
8640 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
8642 @findex SHORT_IMMEDIATES_SIGN_EXTEND
8643 @item SHORT_IMMEDIATES_SIGN_EXTEND
8644 Define this macro if loading short immediate values into registers sign
8647 @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC
8648 @item FIXUNS_TRUNC_LIKE_FIX_TRUNC
8649 Define this macro if the same instructions that convert a floating
8650 point number to a signed fixed point number also convert validly to an
8655 The maximum number of bytes that a single instruction can move quickly
8656 between memory and registers or between two memory locations.
8658 @findex MAX_MOVE_MAX
8660 The maximum number of bytes that a single instruction can move quickly
8661 between memory and registers or between two memory locations. If this
8662 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
8663 constant value that is the largest value that @code{MOVE_MAX} can have
8666 @findex SHIFT_COUNT_TRUNCATED
8667 @item SHIFT_COUNT_TRUNCATED
8668 A C expression that is nonzero if on this machine the number of bits
8669 actually used for the count of a shift operation is equal to the number
8670 of bits needed to represent the size of the object being shifted. When
8671 this macro is nonzero, the compiler will assume that it is safe to omit
8672 a sign-extend, zero-extend, and certain bitwise `and' instructions that
8673 truncates the count of a shift operation. On machines that have
8674 instructions that act on bit-fields at variable positions, which may
8675 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
8676 also enables deletion of truncations of the values that serve as
8677 arguments to bit-field instructions.
8679 If both types of instructions truncate the count (for shifts) and
8680 position (for bit-field operations), or if no variable-position bit-field
8681 instructions exist, you should define this macro.
8683 However, on some machines, such as the 80386 and the 680x0, truncation
8684 only applies to shift operations and not the (real or pretended)
8685 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
8686 such machines. Instead, add patterns to the @file{md} file that include
8687 the implied truncation of the shift instructions.
8689 You need not define this macro if it would always have the value of zero.
8691 @findex TRULY_NOOP_TRUNCATION
8692 @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
8693 A C expression which is nonzero if on this machine it is safe to
8694 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
8695 bits (where @var{outprec} is smaller than @var{inprec}) by merely
8696 operating on it as if it had only @var{outprec} bits.
8698 On many machines, this expression can be 1.
8700 @c rearranged this, removed the phrase "it is reported that". this was
8701 @c to fix an overfull hbox. --mew 10feb93
8702 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
8703 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
8704 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
8705 such cases may improve things.
8707 @findex STORE_FLAG_VALUE
8708 @item STORE_FLAG_VALUE
8709 A C expression describing the value returned by a comparison operator
8710 with an integral mode and stored by a store-flag instruction
8711 (@samp{s@var{cond}}) when the condition is true. This description must
8712 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
8713 comparison operators whose results have a @code{MODE_INT} mode.
8715 A value of 1 or @minus{}1 means that the instruction implementing the
8716 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
8717 and 0 when the comparison is false. Otherwise, the value indicates
8718 which bits of the result are guaranteed to be 1 when the comparison is
8719 true. This value is interpreted in the mode of the comparison
8720 operation, which is given by the mode of the first operand in the
8721 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
8722 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
8725 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
8726 generate code that depends only on the specified bits. It can also
8727 replace comparison operators with equivalent operations if they cause
8728 the required bits to be set, even if the remaining bits are undefined.
8729 For example, on a machine whose comparison operators return an
8730 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
8731 @samp{0x80000000}, saying that just the sign bit is relevant, the
8735 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
8742 (ashift:SI @var{x} (const_int @var{n}))
8746 where @var{n} is the appropriate shift count to move the bit being
8747 tested into the sign bit.
8749 There is no way to describe a machine that always sets the low-order bit
8750 for a true value, but does not guarantee the value of any other bits,
8751 but we do not know of any machine that has such an instruction. If you
8752 are trying to port GCC to such a machine, include an instruction to
8753 perform a logical-and of the result with 1 in the pattern for the
8754 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
8756 Often, a machine will have multiple instructions that obtain a value
8757 from a comparison (or the condition codes). Here are rules to guide the
8758 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
8763 Use the shortest sequence that yields a valid definition for
8764 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
8765 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
8766 comparison operators to do so because there may be opportunities to
8767 combine the normalization with other operations.
8770 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
8771 slightly preferred on machines with expensive jumps and 1 preferred on
8775 As a second choice, choose a value of @samp{0x80000001} if instructions
8776 exist that set both the sign and low-order bits but do not define the
8780 Otherwise, use a value of @samp{0x80000000}.
8783 Many machines can produce both the value chosen for
8784 @code{STORE_FLAG_VALUE} and its negation in the same number of
8785 instructions. On those machines, you should also define a pattern for
8786 those cases, e.g., one matching
8789 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
8792 Some machines can also perform @code{and} or @code{plus} operations on
8793 condition code values with less instructions than the corresponding
8794 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
8795 machines, define the appropriate patterns. Use the names @code{incscc}
8796 and @code{decscc}, respectively, for the patterns which perform
8797 @code{plus} or @code{minus} operations on condition code values. See
8798 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
8799 find such instruction sequences on other machines.
8801 You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
8804 @findex FLOAT_STORE_FLAG_VALUE
8805 @item FLOAT_STORE_FLAG_VALUE (@var{mode})
8806 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
8807 returned when comparison operators with floating-point results are true.
8808 Define this macro on machine that have comparison operations that return
8809 floating-point values. If there are no such operations, do not define
8814 An alias for the machine mode for pointers. On most machines, define
8815 this to be the integer mode corresponding to the width of a hardware
8816 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
8817 On some machines you must define this to be one of the partial integer
8818 modes, such as @code{PSImode}.
8820 The width of @code{Pmode} must be at least as large as the value of
8821 @code{POINTER_SIZE}. If it is not equal, you must define the macro
8822 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
8825 @findex FUNCTION_MODE
8827 An alias for the machine mode used for memory references to functions
8828 being called, in @code{call} RTL expressions. On most machines this
8829 should be @code{QImode}.
8831 @findex INTEGRATE_THRESHOLD
8832 @item INTEGRATE_THRESHOLD (@var{decl})
8833 A C expression for the maximum number of instructions above which the
8834 function @var{decl} should not be inlined. @var{decl} is a
8835 @code{FUNCTION_DECL} node.
8837 The default definition of this macro is 64 plus 8 times the number of
8838 arguments that the function accepts. Some people think a larger
8839 threshold should be used on RISC machines.
8841 @findex STDC_0_IN_SYSTEM_HEADERS
8842 @item STDC_0_IN_SYSTEM_HEADERS
8843 In normal operation, the preprocessor expands @code{__STDC__} to the
8844 constant 1, to signify that GCC conforms to ISO Standard C@. On some
8845 hosts, like Solaris, the system compiler uses a different convention,
8846 where @code{__STDC__} is normally 0, but is 1 if the user specifies
8847 strict conformance to the C Standard.
8849 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
8850 convention when processing system header files, but when processing user
8851 files @code{__STDC__} will always expand to 1.
8853 @findex NO_IMPLICIT_EXTERN_C
8854 @item NO_IMPLICIT_EXTERN_C
8855 Define this macro if the system header files support C++ as well as C@.
8856 This macro inhibits the usual method of using system header files in
8857 C++, which is to pretend that the file's contents are enclosed in
8858 @samp{extern "C" @{@dots{}@}}.
8860 @findex HANDLE_PRAGMA
8861 @item HANDLE_PRAGMA (@var{getc}, @var{ungetc}, @var{name})
8862 This macro is no longer supported. You must use
8863 @code{REGISTER_TARGET_PRAGMAS} instead.
8865 @findex REGISTER_TARGET_PRAGMAS
8868 @item REGISTER_TARGET_PRAGMAS ()
8869 Define this macro if you want to implement any target-specific pragmas.
8870 If defined, it is a C expression which makes a series of calls to
8871 @code{c_register_pragma} for each pragma. The macro may also do any
8872 setup required for the pragmas.
8874 The primary reason to define this macro is to provide compatibility with
8875 other compilers for the same target. In general, we discourage
8876 definition of target-specific pragmas for GCC@.
8878 If the pragma can be implemented by attributes then you should consider
8879 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
8881 Preprocessor macros that appear on pragma lines are not expanded. All
8882 @samp{#pragma} directives that do not match any registered pragma are
8883 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
8885 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
8887 Each call to @code{c_register_pragma} establishes one pragma. The
8888 @var{callback} routine will be called when the preprocessor encounters a
8892 #pragma [@var{space}] @var{name} @dots{}
8895 @var{space} is the case-sensitive namespace of the pragma, or
8896 @code{NULL} to put the pragma in the global namespace. The callback
8897 routine receives @var{pfile} as its first argument, which can be passed
8898 on to cpplib's functions if necessary. You can lex tokens after the
8899 @var{name} by calling @code{c_lex}. Tokens that are not read by the
8900 callback will be silently ignored. The end of the line is indicated by
8901 a token of type @code{CPP_EOF}.
8903 For an example use of this routine, see @file{c4x.h} and the callback
8904 routines defined in @file{c4x-c.c}.
8906 Note that the use of @code{c_lex} is specific to the C and C++
8907 compilers. It will not work in the Java or Fortran compilers, or any
8908 other language compilers for that matter. Thus if @code{c_lex} is going
8909 to be called from target-specific code, it must only be done so when
8910 building the C and C++ compilers. This can be done by defining the
8911 variables @code{c_target_objs} and @code{cxx_target_objs} in the
8912 target entry in the @file{config.gcc} file. These variables should name
8913 the target-specific, language-specific object file which contains the
8914 code that uses @code{c_lex}. Note it will also be necessary to add a
8915 rule to the makefile fragment pointed to by @code{tmake_file} that shows
8916 how to build this object file.
8919 @findex HANDLE_SYSV_PRAGMA
8922 @item HANDLE_SYSV_PRAGMA
8923 Define this macro (to a value of 1) if you want the System V style
8924 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
8925 [=<value>]} to be supported by gcc.
8927 The pack pragma specifies the maximum alignment (in bytes) of fields
8928 within a structure, in much the same way as the @samp{__aligned__} and
8929 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
8930 the behavior to the default.
8932 A subtlety for Microsoft Visual C/C++ style bit-field packing
8933 (e.g. -mms-bitfields) for targets that support it:
8934 When a bit-field is inserted into a packed record, the whole size
8935 of the underlying type is used by one or more same-size adjacent
8936 bit-fields (that is, if its long:3, 32 bits is used in the record,
8937 and any additional adjacent long bit-fields are packed into the same
8938 chunk of 32 bits. However, if the size changes, a new field of that
8941 If both MS bit-fields and @samp{__attribute__((packed))} are used,
8942 the latter will take precedence. If @samp{__attribute__((packed))} is
8943 used on a single field when MS bit-fields are in use, it will take
8944 precedence for that field, but the alignment of the rest of the structure
8945 may affect its placement.
8947 The weak pragma only works if @code{SUPPORTS_WEAK} and
8948 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
8949 of specifically named weak labels, optionally with a value.
8951 @findex HANDLE_PRAGMA_PACK_PUSH_POP
8954 @item HANDLE_PRAGMA_PACK_PUSH_POP
8955 Define this macro (to a value of 1) if you want to support the Win32
8956 style pragmas @samp{#pragma pack(push,@var{n})} and @samp{#pragma
8957 pack(pop)}. The @samp{pack(push,@var{n})} pragma specifies the maximum alignment
8958 (in bytes) of fields within a structure, in much the same way as the
8959 @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
8960 pack value of zero resets the behavior to the default. Successive
8961 invocations of this pragma cause the previous values to be stacked, so
8962 that invocations of @samp{#pragma pack(pop)} will return to the previous
8965 @findex DOLLARS_IN_IDENTIFIERS
8966 @item DOLLARS_IN_IDENTIFIERS
8967 Define this macro to control use of the character @samp{$} in identifier
8968 names. 0 means @samp{$} is not allowed by default; 1 means it is allowed.
8969 1 is the default; there is no need to define this macro in that case.
8970 This macro controls the compiler proper; it does not affect the preprocessor.
8972 @findex NO_DOLLAR_IN_LABEL
8973 @item NO_DOLLAR_IN_LABEL
8974 Define this macro if the assembler does not accept the character
8975 @samp{$} in label names. By default constructors and destructors in
8976 G++ have @samp{$} in the identifiers. If this macro is defined,
8977 @samp{.} is used instead.
8979 @findex NO_DOT_IN_LABEL
8980 @item NO_DOT_IN_LABEL
8981 Define this macro if the assembler does not accept the character
8982 @samp{.} in label names. By default constructors and destructors in G++
8983 have names that use @samp{.}. If this macro is defined, these names
8984 are rewritten to avoid @samp{.}.
8986 @findex DEFAULT_MAIN_RETURN
8987 @item DEFAULT_MAIN_RETURN
8988 Define this macro if the target system expects every program's @code{main}
8989 function to return a standard ``success'' value by default (if no other
8990 value is explicitly returned).
8992 The definition should be a C statement (sans semicolon) to generate the
8993 appropriate rtl instructions. It is used only when compiling the end of
8998 Define this if the target system lacks the function @code{atexit}
8999 from the ISO C standard. If this macro is defined, a default definition
9000 will be provided to support C++. If @code{ON_EXIT} is not defined,
9001 a default @code{exit} function will also be provided.
9005 Define this macro if the target has another way to implement atexit
9006 functionality without replacing @code{exit}. For instance, SunOS 4 has
9007 a similar @code{on_exit} library function.
9009 The definition should be a functional macro which can be used just like
9010 the @code{atexit} function.
9014 Define this if your @code{exit} function needs to do something
9015 besides calling an external function @code{_cleanup} before
9016 terminating with @code{_exit}. The @code{EXIT_BODY} macro is
9017 only needed if @code{NEED_ATEXIT} is defined and @code{ON_EXIT} is not
9020 @findex INSN_SETS_ARE_DELAYED
9021 @item INSN_SETS_ARE_DELAYED (@var{insn})
9022 Define this macro as a C expression that is nonzero if it is safe for the
9023 delay slot scheduler to place instructions in the delay slot of @var{insn},
9024 even if they appear to use a resource set or clobbered in @var{insn}.
9025 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
9026 every @code{call_insn} has this behavior. On machines where some @code{insn}
9027 or @code{jump_insn} is really a function call and hence has this behavior,
9028 you should define this macro.
9030 You need not define this macro if it would always return zero.
9032 @findex INSN_REFERENCES_ARE_DELAYED
9033 @item INSN_REFERENCES_ARE_DELAYED (@var{insn})
9034 Define this macro as a C expression that is nonzero if it is safe for the
9035 delay slot scheduler to place instructions in the delay slot of @var{insn},
9036 even if they appear to set or clobber a resource referenced in @var{insn}.
9037 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
9038 some @code{insn} or @code{jump_insn} is really a function call and its operands
9039 are registers whose use is actually in the subroutine it calls, you should
9040 define this macro. Doing so allows the delay slot scheduler to move
9041 instructions which copy arguments into the argument registers into the delay
9044 You need not define this macro if it would always return zero.
9046 @findex MACHINE_DEPENDENT_REORG
9047 @item MACHINE_DEPENDENT_REORG (@var{insn})
9048 In rare cases, correct code generation requires extra machine
9049 dependent processing between the second jump optimization pass and
9050 delayed branch scheduling. On those machines, define this macro as a C
9051 statement to act on the code starting at @var{insn}.
9053 @findex MULTIPLE_SYMBOL_SPACES
9054 @item MULTIPLE_SYMBOL_SPACES
9055 Define this macro if in some cases global symbols from one translation
9056 unit may not be bound to undefined symbols in another translation unit
9057 without user intervention. For instance, under Microsoft Windows
9058 symbols must be explicitly imported from shared libraries (DLLs).
9060 @findex MD_ASM_CLOBBERS
9061 @item MD_ASM_CLOBBERS (@var{clobbers})
9062 A C statement that adds to @var{clobbers} @code{STRING_CST} trees for
9063 any hard regs the port wishes to automatically clobber for all asms.
9065 @findex MAX_INTEGER_COMPUTATION_MODE
9066 @item MAX_INTEGER_COMPUTATION_MODE
9067 Define this to the largest integer machine mode which can be used for
9068 operations other than load, store and copy operations.
9070 You need only define this macro if the target holds values larger than
9071 @code{word_mode} in general purpose registers. Most targets should not define
9074 @findex MATH_LIBRARY
9076 Define this macro as a C string constant for the linker argument to link
9077 in the system math library, or @samp{""} if the target does not have a
9078 separate math library.
9080 You need only define this macro if the default of @samp{"-lm"} is wrong.
9082 @findex LIBRARY_PATH_ENV
9083 @item LIBRARY_PATH_ENV
9084 Define this macro as a C string constant for the environment variable that
9085 specifies where the linker should look for libraries.
9087 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
9090 @findex TARGET_HAS_F_SETLKW
9091 @item TARGET_HAS_F_SETLKW
9092 Define this macro if the target supports file locking with fcntl / F_SETLKW@.
9093 Note that this functionality is part of POSIX@.
9094 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
9095 to use file locking when exiting a program, which avoids race conditions
9096 if the program has forked.
9098 @findex MAX_CONDITIONAL_EXECUTE
9099 @item MAX_CONDITIONAL_EXECUTE
9101 A C expression for the maximum number of instructions to execute via
9102 conditional execution instructions instead of a branch. A value of
9103 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
9104 1 if it does use cc0.
9106 @findex IFCVT_MODIFY_TESTS
9107 @item IFCVT_MODIFY_TESTS(@var{ce_info}, @var{true_expr}, @var{false_expr})
9108 Used if the target needs to perform machine-dependent modifications on the
9109 conditionals used for turning basic blocks into conditionally executed code.
9110 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
9111 contains information about the currently processed blocks. @var{true_expr}
9112 and @var{false_expr} are the tests that are used for converting the
9113 then-block and the else-block, respectively. Set either @var{true_expr} or
9114 @var{false_expr} to a null pointer if the tests cannot be converted.
9116 @findex IFCVT_MODIFY_MULTIPLE_TESTS
9117 @item IFCVT_MODIFY_MULTIPLE_TESTS(@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
9118 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
9119 if-statements into conditions combined by @code{and} and @code{or} operations.
9120 @var{bb} contains the basic block that contains the test that is currently
9121 being processed and about to be turned into a condition.
9123 @findex IFCVT_MODIFY_INSN
9124 @item IFCVT_MODIFY_INSN(@var{ce_info}, @var{pattern}, @var{insn})
9125 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
9126 be converted to conditional execution format. @var{ce_info} points to
9127 a data structure, @code{struct ce_if_block}, which contains information
9128 about the currently processed blocks.
9130 @findex IFCVT_MODIFY_FINAL
9131 @item IFCVT_MODIFY_FINAL(@var{ce_info})
9132 A C expression to perform any final machine dependent modifications in
9133 converting code to conditional execution. The involved basic blocks
9134 can be found in the @code{struct ce_if_block} structure that is pointed
9135 to by @var{ce_info}.
9137 @findex IFCVT_MODIFY_CANCEL
9138 @item IFCVT_MODIFY_CANCEL(@var{ce_info})
9139 A C expression to cancel any machine dependent modifications in
9140 converting code to conditional execution. The involved basic blocks
9141 can be found in the @code{struct ce_if_block} structure that is pointed
9142 to by @var{ce_info}.
9144 @findex IFCVT_INIT_EXTRA_FIELDS
9145 @item IFCVT_INIT_EXTRA_FIELDS(@var{ce_info})
9146 A C expression to initialize any extra fields in a @code{struct ce_if_block}
9147 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
9149 @findex IFCVT_EXTRA_FIELDS
9150 @item IFCVT_EXTRA_FIELDS
9151 If defined, it should expand to a set of field declarations that will be
9152 added to the @code{struct ce_if_block} structure. These should be initialized
9153 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
9157 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
9158 Define this hook if you have any machine-specific built-in functions
9159 that need to be defined. It should be a function that performs the
9162 Machine specific built-in functions can be useful to expand special machine
9163 instructions that would otherwise not normally be generated because
9164 they have no equivalent in the source language (for example, SIMD vector
9165 instructions or prefetch instructions).
9167 To create a built-in function, call the function @code{builtin_function}
9168 which is defined by the language front end. You can use any type nodes set
9169 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
9170 only language front ends that use those two functions will call
9171 @samp{TARGET_INIT_BUILTINS}.
9174 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
9176 Expand a call to a machine specific built-in function that was set up by
9177 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
9178 function call; the result should go to @var{target} if that is
9179 convenient, and have mode @var{mode} if that is convenient.
9180 @var{subtarget} may be used as the target for computing one of
9181 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
9182 ignored. This function should return the result of the call to the
9187 @findex MD_CAN_REDIRECT_BRANCH
9188 @item MD_CAN_REDIRECT_BRANCH(@var{branch1}, @var{branch2})
9190 Take a branch insn in @var{branch1} and another in @var{branch2}.
9191 Return true if redirecting @var{branch1} to the destination of
9192 @var{branch2} is possible.
9194 On some targets, branches may have a limited range. Optimizing the
9195 filling of delay slots can result in branches being redirected, and this
9196 may in turn cause a branch offset to overflow.
9198 @findex ALLOCATE_INITIAL_VALUE
9199 @item ALLOCATE_INITIAL_VALUE(@var{hard_reg})
9201 When the initial value of a hard register has been copied in a pseudo
9202 register, it is often not necessary to actually allocate another register
9203 to this pseudo register, because the original hard register or a stack slot
9204 it has been saved into can be used. @code{ALLOCATE_INITIAL_VALUE}, if
9205 defined, is called at the start of register allocation once for each
9206 hard register that had its initial value copied by using
9207 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
9208 Possible values are @code{NULL_RTX}, if you don't want
9209 to do any special allocation, a @code{REG} rtx---that would typically be
9210 the hard register itself, if it is known not to be clobbered---or a
9212 If you are returning a @code{MEM}, this is only a hint for the allocator;
9213 it might decide to use another register anyways.
9214 You may use @code{current_function_leaf_function} in the definition of the
9215 macro, functions that use @code{REG_N_SETS}, to determine if the hard
9216 register in question will not be clobbered.
9218 @findex TARGET_OBJECT_SUFFIX
9219 @item TARGET_OBJECT_SUFFIX
9220 Define this macro to be a C string representing the suffix for object
9221 files on your target machine. If you do not define this macro, GCC will
9222 use @samp{.o} as the suffix for object files.
9224 @findex TARGET_EXECUTABLE_SUFFIX
9225 @item TARGET_EXECUTABLE_SUFFIX
9226 Define this macro to be a C string representing the suffix to be
9227 automatically added to executable files on your target machine. If you
9228 do not define this macro, GCC will use the null string as the suffix for
9231 @findex COLLECT_EXPORT_LIST
9232 @item COLLECT_EXPORT_LIST
9233 If defined, @code{collect2} will scan the individual object files
9234 specified on its command line and create an export list for the linker.
9235 Define this macro for systems like AIX, where the linker discards
9236 object files that are not referenced from @code{main} and uses export
9241 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
9242 This target hook returns @code{true} past the point in which new jump
9243 instructions could be created. On machines that require a register for
9244 every jump such as the SHmedia ISA of SH5, this point would typically be
9245 reload, so this target hook should be defined to a function such as:
9249 cannot_modify_jumps_past_reload_p ()
9251 return (reload_completed || reload_in_progress);