1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,2002
2 @c 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" @}
168 A C string constant that tells the GCC driver program options to
169 pass to CPP@. It can also specify how to translate options you
170 give to GCC into options for GCC to pass to the CPP@.
172 Do not define this macro if it does not need to do anything.
174 @findex CPLUSPLUS_CPP_SPEC
175 @item CPLUSPLUS_CPP_SPEC
176 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
177 than C@. If you do not define this macro, then the value of
178 @code{CPP_SPEC} (if any) will be used instead.
182 A C string constant that tells the GCC driver program options to
183 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
185 It can also specify how to translate options you give to GCC into options
186 for GCC to pass to front ends.
188 Do not define this macro if it does not need to do anything.
192 A C string constant that tells the GCC driver program options to
193 pass to @code{cc1plus}. It can also specify how to translate options you
194 give to GCC into options for GCC to pass to the @code{cc1plus}.
196 Do not define this macro if it does not need to do anything.
197 Note that everything defined in CC1_SPEC is already passed to
198 @code{cc1plus} so there is no need to duplicate the contents of
199 CC1_SPEC in CC1PLUS_SPEC@.
203 A C string constant that tells the GCC driver program options to
204 pass to the assembler. It can also specify how to translate options
205 you give to GCC into options for GCC to pass to the assembler.
206 See the file @file{sun3.h} for an example of this.
208 Do not define this macro if it does not need to do anything.
210 @findex ASM_FINAL_SPEC
212 A C string constant that tells the GCC driver program how to
213 run any programs which cleanup after the normal assembler.
214 Normally, this is not needed. See the file @file{mips.h} for
217 Do not define this macro if it does not need to do anything.
221 A C string constant that tells the GCC driver program options to
222 pass to the linker. It can also specify how to translate options you
223 give to GCC into options for GCC to pass to the linker.
225 Do not define this macro if it does not need to do anything.
229 Another C string constant used much like @code{LINK_SPEC}. The difference
230 between the two is that @code{LIB_SPEC} is used at the end of the
231 command given to the linker.
233 If this macro is not defined, a default is provided that
234 loads the standard C library from the usual place. See @file{gcc.c}.
238 Another C string constant that tells the GCC driver program
239 how and when to place a reference to @file{libgcc.a} into the
240 linker command line. This constant is placed both before and after
241 the value of @code{LIB_SPEC}.
243 If this macro is not defined, the GCC driver provides a default that
244 passes the string @option{-lgcc} to the linker.
246 @findex STARTFILE_SPEC
248 Another C string constant used much like @code{LINK_SPEC}. The
249 difference between the two is that @code{STARTFILE_SPEC} is used at
250 the very beginning of the command given to the linker.
252 If this macro is not defined, a default is provided that loads the
253 standard C startup file from the usual place. See @file{gcc.c}.
257 Another C string constant used much like @code{LINK_SPEC}. The
258 difference between the two is that @code{ENDFILE_SPEC} is used at
259 the very end of the command given to the linker.
261 Do not define this macro if it does not need to do anything.
263 @findex THREAD_MODEL_SPEC
264 @item THREAD_MODEL_SPEC
265 GCC @code{-v} will print the thread model GCC was configured to use.
266 However, this doesn't work on platforms that are multilibbed on thread
267 models, such as AIX 4.3. On such platforms, define
268 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
269 blanks that names one of the recognized thread models. @code{%*}, the
270 default value of this macro, will expand to the value of
271 @code{thread_file} set in @file{config.gcc}.
275 Define this macro to provide additional specifications to put in the
276 @file{specs} file that can be used in various specifications like
279 The definition should be an initializer for an array of structures,
280 containing a string constant, that defines the specification name, and a
281 string constant that provides the specification.
283 Do not define this macro if it does not need to do anything.
285 @code{EXTRA_SPECS} is useful when an architecture contains several
286 related targets, which have various @code{@dots{}_SPECS} which are similar
287 to each other, and the maintainer would like one central place to keep
290 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
291 define either @code{_CALL_SYSV} when the System V calling sequence is
292 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
295 The @file{config/rs6000/rs6000.h} target file defines:
298 #define EXTRA_SPECS \
299 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
301 #define CPP_SYS_DEFAULT ""
304 The @file{config/rs6000/sysv.h} target file defines:
308 "%@{posix: -D_POSIX_SOURCE @} \
309 %@{mcall-sysv: -D_CALL_SYSV @} %@{mcall-aix: -D_CALL_AIX @} \
310 %@{!mcall-sysv: %@{!mcall-aix: %(cpp_sysv_default) @}@} \
311 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
313 #undef CPP_SYSV_DEFAULT
314 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
317 while the @file{config/rs6000/eabiaix.h} target file defines
318 @code{CPP_SYSV_DEFAULT} as:
321 #undef CPP_SYSV_DEFAULT
322 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
325 @findex LINK_LIBGCC_SPECIAL
326 @item LINK_LIBGCC_SPECIAL
327 Define this macro if the driver program should find the library
328 @file{libgcc.a} itself and should not pass @option{-L} options to the
329 linker. If you do not define this macro, the driver program will pass
330 the argument @option{-lgcc} to tell the linker to do the search and will
331 pass @option{-L} options to it.
333 @findex LINK_LIBGCC_SPECIAL_1
334 @item LINK_LIBGCC_SPECIAL_1
335 Define this macro if the driver program should find the library
336 @file{libgcc.a}. If you do not define this macro, the driver program will pass
337 the argument @option{-lgcc} to tell the linker to do the search.
338 This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does
339 not affect @option{-L} options.
341 @findex LINK_GCC_C_SEQUENCE_SPEC
342 @item LINK_GCC_C_SEQUENCE_SPEC
343 The sequence in which libgcc and libc are specified to the linker.
344 By default this is @code{%G %L %G}.
346 @findex LINK_COMMAND_SPEC
347 @item LINK_COMMAND_SPEC
348 A C string constant giving the complete command line need to execute the
349 linker. When you do this, you will need to update your port each time a
350 change is made to the link command line within @file{gcc.c}. Therefore,
351 define this macro only if you need to completely redefine the command
352 line for invoking the linker and there is no other way to accomplish
353 the effect you need. Overriding this macro may be avoidable by overriding
354 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
356 @findex LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
357 @item LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
358 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
359 directories from linking commands. Do not give it a nonzero value if
360 removing duplicate search directories changes the linker's semantics.
362 @findex MULTILIB_DEFAULTS
363 @item MULTILIB_DEFAULTS
364 Define this macro as a C expression for the initializer of an array of
365 string to tell the driver program which options are defaults for this
366 target and thus do not need to be handled specially when using
367 @code{MULTILIB_OPTIONS}.
369 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
370 the target makefile fragment or if none of the options listed in
371 @code{MULTILIB_OPTIONS} are set by default.
372 @xref{Target Fragment}.
374 @findex RELATIVE_PREFIX_NOT_LINKDIR
375 @item RELATIVE_PREFIX_NOT_LINKDIR
376 Define this macro to tell @code{gcc} that it should only translate
377 a @option{-B} prefix into a @option{-L} linker option if the prefix
378 indicates an absolute file name.
380 @findex STANDARD_EXEC_PREFIX
381 @item STANDARD_EXEC_PREFIX
382 Define this macro as a C string constant if you wish to override the
383 standard choice of @file{/usr/local/lib/gcc-lib/} as the default prefix to
384 try when searching for the executable files of the compiler.
386 @findex MD_EXEC_PREFIX
388 If defined, this macro is an additional prefix to try after
389 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
390 when the @option{-b} option is used, or the compiler is built as a cross
391 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
392 to the list of directories used to find the assembler in @file{configure.in}.
394 @findex STANDARD_STARTFILE_PREFIX
395 @item STANDARD_STARTFILE_PREFIX
396 Define this macro as a C string constant if you wish to override the
397 standard choice of @file{/usr/local/lib/} as the default prefix to
398 try when searching for startup files such as @file{crt0.o}.
400 @findex MD_STARTFILE_PREFIX
401 @item MD_STARTFILE_PREFIX
402 If defined, this macro supplies an additional prefix to try after the
403 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
404 @option{-b} option is used, or when the compiler is built as a cross
407 @findex MD_STARTFILE_PREFIX_1
408 @item MD_STARTFILE_PREFIX_1
409 If defined, this macro supplies yet another prefix to try after the
410 standard prefixes. It is not searched when the @option{-b} option is
411 used, or when the compiler is built as a cross compiler.
413 @findex INIT_ENVIRONMENT
414 @item INIT_ENVIRONMENT
415 Define this macro as a C string constant if you wish to set environment
416 variables for programs called by the driver, such as the assembler and
417 loader. The driver passes the value of this macro to @code{putenv} to
418 initialize the necessary environment variables.
420 @findex LOCAL_INCLUDE_DIR
421 @item LOCAL_INCLUDE_DIR
422 Define this macro as a C string constant if you wish to override the
423 standard choice of @file{/usr/local/include} as the default prefix to
424 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
425 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
427 Cross compilers do not search either @file{/usr/local/include} or its
430 @findex MODIFY_TARGET_NAME
431 @item MODIFY_TARGET_NAME
432 Define this macro if you with to define command-line switches that modify the
435 For each switch, you can include a string to be appended to the first
436 part of the configuration name or a string to be deleted from the
437 configuration name, if present. The definition should be an initializer
438 for an array of structures. Each array element should have three
439 elements: the switch name (a string constant, including the initial
440 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
441 indicate whether the string should be inserted or deleted, and the string
442 to be inserted or deleted (a string constant).
444 For example, on a machine where @samp{64} at the end of the
445 configuration name denotes a 64-bit target and you want the @option{-32}
446 and @option{-64} switches to select between 32- and 64-bit targets, you would
450 #define MODIFY_TARGET_NAME \
451 @{ @{ "-32", DELETE, "64"@}, \
452 @{"-64", ADD, "64"@}@}
456 @findex SYSTEM_INCLUDE_DIR
457 @item SYSTEM_INCLUDE_DIR
458 Define this macro as a C string constant if you wish to specify a
459 system-specific directory to search for header files before the standard
460 directory. @code{SYSTEM_INCLUDE_DIR} comes before
461 @code{STANDARD_INCLUDE_DIR} in the search order.
463 Cross compilers do not use this macro and do not search the directory
466 @findex STANDARD_INCLUDE_DIR
467 @item STANDARD_INCLUDE_DIR
468 Define this macro as a C string constant if you wish to override the
469 standard choice of @file{/usr/include} as the default prefix to
470 try when searching for header files.
472 Cross compilers do not use this macro and do not search either
473 @file{/usr/include} or its replacement.
475 @findex STANDARD_INCLUDE_COMPONENT
476 @item STANDARD_INCLUDE_COMPONENT
477 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
478 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
479 If you do not define this macro, no component is used.
481 @findex INCLUDE_DEFAULTS
482 @item INCLUDE_DEFAULTS
483 Define this macro if you wish to override the entire default search path
484 for include files. For a native compiler, the default search path
485 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
486 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
487 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
488 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
489 and specify private search areas for GCC@. The directory
490 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
492 The definition should be an initializer for an array of structures.
493 Each array element should have four elements: the directory name (a
494 string constant), the component name (also a string constant), a flag
495 for C++-only directories,
496 and a flag showing that the includes in the directory don't need to be
497 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
498 the array with a null element.
500 The component name denotes what GNU package the include file is part of,
501 if any, in all upper-case letters. For example, it might be @samp{GCC}
502 or @samp{BINUTILS}. If the package is part of a vendor-supplied
503 operating system, code the component name as @samp{0}.
505 For example, here is the definition used for VAX/VMS:
508 #define INCLUDE_DEFAULTS \
510 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
511 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
512 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
519 Here is the order of prefixes tried for exec files:
523 Any prefixes specified by the user with @option{-B}.
526 The environment variable @code{GCC_EXEC_PREFIX}, if any.
529 The directories specified by the environment variable @code{COMPILER_PATH}.
532 The macro @code{STANDARD_EXEC_PREFIX}.
535 @file{/usr/lib/gcc/}.
538 The macro @code{MD_EXEC_PREFIX}, if any.
541 Here is the order of prefixes tried for startfiles:
545 Any prefixes specified by the user with @option{-B}.
548 The environment variable @code{GCC_EXEC_PREFIX}, if any.
551 The directories specified by the environment variable @code{LIBRARY_PATH}
552 (or port-specific name; native only, cross compilers do not use this).
555 The macro @code{STANDARD_EXEC_PREFIX}.
558 @file{/usr/lib/gcc/}.
561 The macro @code{MD_EXEC_PREFIX}, if any.
564 The macro @code{MD_STARTFILE_PREFIX}, if any.
567 The macro @code{STANDARD_STARTFILE_PREFIX}.
576 @node Run-time Target
577 @section Run-time Target Specification
578 @cindex run-time target specification
579 @cindex predefined macros
580 @cindex target specifications
582 @c prevent bad page break with this line
583 Here are run-time target specifications.
586 @findex TARGET_CPU_CPP_BUILTINS
587 @item TARGET_CPU_CPP_BUILTINS()
588 This function-like macro expands to a block of code that defines
589 built-in preprocessor macros and assertions for the target cpu, using
590 the functions @code{builtin_define}, @code{builtin_define_std} and
591 @code{builtin_assert} defined in @file{c-common.c}. When the front end
592 calls this macro it provides a trailing semicolon, and since it has
593 finished command line option processing your code can use those
596 @code{builtin_assert} takes a string in the form you pass to the
597 command-line option @option{-A}, such as @code{cpu=mips}, and creates
598 the assertion. @code{builtin_define} takes a string in the form
599 accepted by option @option{-D} and unconditionally defines the macro.
601 @code{builtin_define_std} takes a string representing the name of an
602 object-like macro. If it doesn't lie in the user's namespace,
603 @code{builtin_define_std} defines it unconditionally. Otherwise, it
604 defines a version with two leading underscores, and another version
605 with two leading and trailing underscores, and defines the original
606 only if an ISO standard was not requested on the command line. For
607 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
608 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
609 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
610 defines only @code{_ABI64}.
612 You can also test for the C dialect being compiled. The variable
613 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
614 or @code{clk_objective_c}. Note that if we are preprocessing
615 assembler, this variable will be @code{clk_c} but the function-like
616 macro @code{preprocessing_asm_p()} will return true, so you might want
617 to check for that first. If you need to check for strict ANSI, the
618 variable @code{flag_iso} can be used.
620 With @code{TARGET_OS_CPP_BUILTINS} this macro obsoletes the
621 @code{CPP_PREDEFINES} target macro.
623 @findex TARGET_OS_CPP_BUILTINS
624 @item TARGET_OS_CPP_BUILTINS()
625 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
626 and is used for the target operating system instead.
628 With @code{TARGET_CPU_CPP_BUILTINS} this macro obsoletes the
629 @code{CPP_PREDEFINES} target macro.
631 @findex CPP_PREDEFINES
633 Define this to be a string constant containing @option{-D} options to
634 define the predefined macros that identify this machine and system.
635 These macros will be predefined unless the @option{-ansi} option (or a
636 @option{-std} option for strict ISO C conformance) is specified.
638 In addition, a parallel set of macros are predefined, whose names are
639 made by appending @samp{__} at the beginning and at the end. These
640 @samp{__} macros are permitted by the ISO standard, so they are
641 predefined regardless of whether @option{-ansi} or a @option{-std} option
644 For example, on the Sun, one can use the following value:
647 "-Dmc68000 -Dsun -Dunix"
650 The result is to define the macros @code{__mc68000__}, @code{__sun__}
651 and @code{__unix__} unconditionally, and the macros @code{mc68000},
652 @code{sun} and @code{unix} provided @option{-ansi} is not specified.
654 @findex extern int target_flags
655 @item extern int target_flags;
656 This declaration should be present.
658 @cindex optional hardware or system features
659 @cindex features, optional, in system conventions
661 This series of macros is to allow compiler command arguments to
662 enable or disable the use of optional features of the target machine.
663 For example, one machine description serves both the 68000 and
664 the 68020; a command argument tells the compiler whether it should
665 use 68020-only instructions or not. This command argument works
666 by means of a macro @code{TARGET_68020} that tests a bit in
669 Define a macro @code{TARGET_@var{featurename}} for each such option.
670 Its definition should test a bit in @code{target_flags}. It is
671 recommended that a helper macro @code{TARGET_MASK_@var{featurename}}
672 is defined for each bit-value to test, and used in
673 @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For
677 #define TARGET_MASK_68020 1
678 #define TARGET_68020 (target_flags & TARGET_MASK_68020)
681 One place where these macros are used is in the condition-expressions
682 of instruction patterns. Note how @code{TARGET_68020} appears
683 frequently in the 68000 machine description file, @file{m68k.md}.
684 Another place they are used is in the definitions of the other
685 macros in the @file{@var{machine}.h} file.
687 @findex TARGET_SWITCHES
688 @item TARGET_SWITCHES
689 This macro defines names of command options to set and clear
690 bits in @code{target_flags}. Its definition is an initializer
691 with a subgrouping for each command option.
693 Each subgrouping contains a string constant, that defines the option
694 name, a number, which contains the bits to set in
695 @code{target_flags}, and a second string which is the description
696 displayed by @option{--help}. If the number is negative then the bits specified
697 by the number are cleared instead of being set. If the description
698 string is present but empty, then no help information will be displayed
699 for that option, but it will not count as an undocumented option. The
700 actual option name is made by appending @samp{-m} to the specified name.
701 Non-empty description strings should be marked with @code{N_(@dots{})} for
702 @command{xgettext}. Please do not mark empty strings because the empty
703 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
704 of the message catalog with meta information, not the empty string.
706 In addition to the description for @option{--help},
707 more detailed documentation for each option should be added to
710 One of the subgroupings should have a null string. The number in
711 this grouping is the default value for @code{target_flags}. Any
712 target options act starting with that value.
714 Here is an example which defines @option{-m68000} and @option{-m68020}
715 with opposite meanings, and picks the latter as the default:
718 #define TARGET_SWITCHES \
719 @{ @{ "68020", TARGET_MASK_68020, "" @}, \
720 @{ "68000", -TARGET_MASK_68020, \
721 N_("Compile for the 68000") @}, \
722 @{ "", TARGET_MASK_68020, "" @}@}
725 @findex TARGET_OPTIONS
727 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
728 options that have values. Its definition is an initializer with a
729 subgrouping for each command option.
731 Each subgrouping contains a string constant, that defines the fixed part
732 of the option name, the address of a variable, and a description string.
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 The variable, type @code{char *}, is set to the variable part of the
739 given option if the fixed part matches. The actual option name is made
740 by appending @samp{-m} to the specified name. Again, each option should
741 also be documented in @file{invoke.texi}.
743 Here is an example which defines @option{-mshort-data-@var{number}}. If the
744 given option is @option{-mshort-data-512}, the variable @code{m88k_short_data}
745 will be set to the string @code{"512"}.
748 extern char *m88k_short_data;
749 #define TARGET_OPTIONS \
750 @{ @{ "short-data-", &m88k_short_data, \
751 N_("Specify the size of the short data section") @} @}
754 @findex TARGET_VERSION
756 This macro is a C statement to print on @code{stderr} a string
757 describing the particular machine description choice. Every machine
758 description should define @code{TARGET_VERSION}. For example:
762 #define TARGET_VERSION \
763 fprintf (stderr, " (68k, Motorola syntax)");
765 #define TARGET_VERSION \
766 fprintf (stderr, " (68k, MIT syntax)");
770 @findex OVERRIDE_OPTIONS
771 @item OVERRIDE_OPTIONS
772 Sometimes certain combinations of command options do not make sense on
773 a particular target machine. You can define a macro
774 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
775 defined, is executed once just after all the command options have been
778 Don't use this macro to turn on various extra optimizations for
779 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
781 @findex OPTIMIZATION_OPTIONS
782 @item OPTIMIZATION_OPTIONS (@var{level}, @var{size})
783 Some machines may desire to change what optimizations are performed for
784 various optimization levels. This macro, if defined, is executed once
785 just after the optimization level is determined and before the remainder
786 of the command options have been parsed. Values set in this macro are
787 used as the default values for the other command line options.
789 @var{level} is the optimization level specified; 2 if @option{-O2} is
790 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
792 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
794 You should not use this macro to change options that are not
795 machine-specific. These should uniformly selected by the same
796 optimization level on all supported machines. Use this macro to enable
797 machine-specific optimizations.
799 @strong{Do not examine @code{write_symbols} in
800 this macro!} The debugging options are not supposed to alter the
803 @findex CAN_DEBUG_WITHOUT_FP
804 @item CAN_DEBUG_WITHOUT_FP
805 Define this macro if debugging can be performed even without a frame
806 pointer. If this macro is defined, GCC will turn on the
807 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
810 @node Per-Function Data
811 @section Defining data structures for per-function information.
812 @cindex per-function data
813 @cindex data structures
815 If the target needs to store information on a per-function basis, GCC
816 provides a macro and a couple of variables to allow this. Note, just
817 using statics to store the information is a bad idea, since GCC supports
818 nested functions, so you can be halfway through encoding one function
819 when another one comes along.
821 GCC defines a data structure called @code{struct function} which
822 contains all of the data specific to an individual function. This
823 structure contains a field called @code{machine} whose type is
824 @code{struct machine_function *}, which can be used by targets to point
825 to their own specific data.
827 If a target needs per-function specific data it should define the type
828 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
829 This macro should be used to initialize the function pointer
830 @code{init_machine_status}. This pointer is explained below.
832 One typical use of per-function, target specific data is to create an
833 RTX to hold the register containing the function's return address. This
834 RTX can then be used to implement the @code{__builtin_return_address}
835 function, for level 0.
837 Note---earlier implementations of GCC used a single data area to hold
838 all of the per-function information. Thus when processing of a nested
839 function began the old per-function data had to be pushed onto a
840 stack, and when the processing was finished, it had to be popped off the
841 stack. GCC used to provide function pointers called
842 @code{save_machine_status} and @code{restore_machine_status} to handle
843 the saving and restoring of the target specific information. Since the
844 single data area approach is no longer used, these pointers are no
847 The macro and function pointers are described below.
850 @findex INIT_EXPANDERS
852 Macro called to initialize any target specific information. This macro
853 is called once per function, before generation of any RTL has begun.
854 The intention of this macro is to allow the initialization of the
855 function pointers below.
857 @findex init_machine_status
858 @item init_machine_status
859 This is a @code{void (*)(struct function *)} function pointer. If this
860 pointer is non-@code{NULL} it will be called once per function, before function
861 compilation starts, in order to allow the target to perform any target
862 specific initialization of the @code{struct function} structure. It is
863 intended that this would be used to initialize the @code{machine} of
866 @code{struct machine_function} structures are expected to be freed by GC.
867 Generally, any memory that they reference must be allocated by using
868 @code{ggc_alloc}, including the structure itself.
873 @section Storage Layout
874 @cindex storage layout
876 Note that the definitions of the macros in this table which are sizes or
877 alignments measured in bits do not need to be constant. They can be C
878 expressions that refer to static variables, such as the @code{target_flags}.
879 @xref{Run-time Target}.
882 @findex BITS_BIG_ENDIAN
883 @item BITS_BIG_ENDIAN
884 Define this macro to have the value 1 if the most significant bit in a
885 byte has the lowest number; otherwise define it to have the value zero.
886 This means that bit-field instructions count from the most significant
887 bit. If the machine has no bit-field instructions, then this must still
888 be defined, but it doesn't matter which value it is defined to. This
889 macro need not be a constant.
891 This macro does not affect the way structure fields are packed into
892 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
894 @findex BYTES_BIG_ENDIAN
895 @item BYTES_BIG_ENDIAN
896 Define this macro to have the value 1 if the most significant byte in a
897 word has the lowest number. This macro need not be a constant.
899 @findex WORDS_BIG_ENDIAN
900 @item WORDS_BIG_ENDIAN
901 Define this macro to have the value 1 if, in a multiword object, the
902 most significant word has the lowest number. This applies to both
903 memory locations and registers; GCC fundamentally assumes that the
904 order of words in memory is the same as the order in registers. This
905 macro need not be a constant.
907 @findex LIBGCC2_WORDS_BIG_ENDIAN
908 @item LIBGCC2_WORDS_BIG_ENDIAN
909 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
910 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
911 used only when compiling @file{libgcc2.c}. Typically the value will be set
912 based on preprocessor defines.
914 @findex FLOAT_WORDS_BIG_ENDIAN
915 @item FLOAT_WORDS_BIG_ENDIAN
916 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
917 @code{TFmode} floating point numbers are stored in memory with the word
918 containing the sign bit at the lowest address; otherwise define it to
919 have the value 0. This macro need not be a constant.
921 You need not define this macro if the ordering is the same as for
924 @findex BITS_PER_UNIT
926 Define this macro to be the number of bits in an addressable storage
927 unit (byte). If you do not define this macro the default is 8.
929 @findex BITS_PER_WORD
931 Number of bits in a word. If you do not define this macro, the default
932 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
934 @findex MAX_BITS_PER_WORD
935 @item MAX_BITS_PER_WORD
936 Maximum number of bits in a word. If this is undefined, the default is
937 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
938 largest value that @code{BITS_PER_WORD} can have at run-time.
940 @findex UNITS_PER_WORD
942 Number of storage units in a word; normally 4.
944 @findex MIN_UNITS_PER_WORD
945 @item MIN_UNITS_PER_WORD
946 Minimum number of units in a word. If this is undefined, the default is
947 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
948 smallest value that @code{UNITS_PER_WORD} can have at run-time.
952 Width of a pointer, in bits. You must specify a value no wider than the
953 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
954 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
955 a value the default is @code{BITS_PER_WORD}.
957 @findex POINTERS_EXTEND_UNSIGNED
958 @item POINTERS_EXTEND_UNSIGNED
959 A C expression whose value is greater than zero if pointers that need to be
960 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
961 be zero-extended and zero if they are to be sign-extended. If the value
962 is less then zero then there must be an "ptr_extend" instruction that
963 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
965 You need not define this macro if the @code{POINTER_SIZE} is equal
966 to the width of @code{Pmode}.
969 @item PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
970 A macro to update @var{m} and @var{unsignedp} when an object whose type
971 is @var{type} and which has the specified mode and signedness is to be
972 stored in a register. This macro is only called when @var{type} is a
975 On most RISC machines, which only have operations that operate on a full
976 register, define this macro to set @var{m} to @code{word_mode} if
977 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
978 cases, only integer modes should be widened because wider-precision
979 floating-point operations are usually more expensive than their narrower
982 For most machines, the macro definition does not change @var{unsignedp}.
983 However, some machines, have instructions that preferentially handle
984 either signed or unsigned quantities of certain modes. For example, on
985 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
986 sign-extend the result to 64 bits. On such machines, set
987 @var{unsignedp} according to which kind of extension is more efficient.
989 Do not define this macro if it would never modify @var{m}.
991 @findex PROMOTE_FUNCTION_ARGS
992 @item PROMOTE_FUNCTION_ARGS
993 Define this macro if the promotion described by @code{PROMOTE_MODE}
994 should also be done for outgoing function arguments.
996 @findex PROMOTE_FUNCTION_RETURN
997 @item PROMOTE_FUNCTION_RETURN
998 Define this macro if the promotion described by @code{PROMOTE_MODE}
999 should also be done for the return value of functions.
1001 If this macro is defined, @code{FUNCTION_VALUE} must perform the same
1002 promotions done by @code{PROMOTE_MODE}.
1004 @findex PROMOTE_FOR_CALL_ONLY
1005 @item PROMOTE_FOR_CALL_ONLY
1006 Define this macro if the promotion described by @code{PROMOTE_MODE}
1007 should @emph{only} be performed for outgoing function arguments or
1008 function return values, as specified by @code{PROMOTE_FUNCTION_ARGS}
1009 and @code{PROMOTE_FUNCTION_RETURN}, respectively.
1011 @findex PARM_BOUNDARY
1013 Normal alignment required for function parameters on the stack, in
1014 bits. All stack parameters receive at least this much alignment
1015 regardless of data type. On most machines, this is the same as the
1018 @findex STACK_BOUNDARY
1019 @item STACK_BOUNDARY
1020 Define this macro to the minimum alignment enforced by hardware for the
1021 stack pointer on this machine. The definition is a C expression for the
1022 desired alignment (measured in bits). This value is used as a default
1023 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1024 this should be the same as @code{PARM_BOUNDARY}.
1026 @findex PREFERRED_STACK_BOUNDARY
1027 @item PREFERRED_STACK_BOUNDARY
1028 Define this macro if you wish to preserve a certain alignment for the
1029 stack pointer, greater than what the hardware enforces. The definition
1030 is a C expression for the desired alignment (measured in bits). This
1031 macro must evaluate to a value equal to or larger than
1032 @code{STACK_BOUNDARY}.
1034 @findex FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1035 @item FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1036 A C expression that evaluates true if @code{PREFERRED_STACK_BOUNDARY} is
1037 not guaranteed by the runtime and we should emit code to align the stack
1038 at the beginning of @code{main}.
1040 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
1041 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
1042 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
1043 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
1044 be momentarily unaligned while pushing arguments.
1046 @findex FUNCTION_BOUNDARY
1047 @item FUNCTION_BOUNDARY
1048 Alignment required for a function entry point, in bits.
1050 @findex BIGGEST_ALIGNMENT
1051 @item BIGGEST_ALIGNMENT
1052 Biggest alignment that any data type can require on this machine, in bits.
1054 @findex MINIMUM_ATOMIC_ALIGNMENT
1055 @item MINIMUM_ATOMIC_ALIGNMENT
1056 If defined, the smallest alignment, in bits, that can be given to an
1057 object that can be referenced in one operation, without disturbing any
1058 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1059 on machines that don't have byte or half-word store operations.
1061 @findex BIGGEST_FIELD_ALIGNMENT
1062 @item BIGGEST_FIELD_ALIGNMENT
1063 Biggest alignment that any structure or union field can require on this
1064 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1065 structure and union fields only, unless the field alignment has been set
1066 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1068 @findex ADJUST_FIELD_ALIGN
1069 @item ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1070 An expression for the alignment of a structure field @var{field} if the
1071 alignment computed in the usual way (including applying of
1072 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1073 alignment) is @var{computed}. It overrides alignment only if the
1074 field alignment has not been set by the
1075 @code{__attribute__ ((aligned (@var{n})))} construct.
1077 @findex MAX_OFILE_ALIGNMENT
1078 @item MAX_OFILE_ALIGNMENT
1079 Biggest alignment supported by the object file format of this machine.
1080 Use this macro to limit the alignment which can be specified using the
1081 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1082 the default value is @code{BIGGEST_ALIGNMENT}.
1084 @findex DATA_ALIGNMENT
1085 @item DATA_ALIGNMENT (@var{type}, @var{basic-align})
1086 If defined, a C expression to compute the alignment for a variable in
1087 the static store. @var{type} is the data type, and @var{basic-align} is
1088 the alignment that the object would ordinarily have. The value of this
1089 macro is used instead of that alignment to align the object.
1091 If this macro is not defined, then @var{basic-align} is used.
1094 One use of this macro is to increase alignment of medium-size data to
1095 make it all fit in fewer cache lines. Another is to cause character
1096 arrays to be word-aligned so that @code{strcpy} calls that copy
1097 constants to character arrays can be done inline.
1099 @findex CONSTANT_ALIGNMENT
1100 @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1101 If defined, a C expression to compute the alignment given to a constant
1102 that is being placed in memory. @var{constant} is the constant and
1103 @var{basic-align} is the alignment that the object would ordinarily
1104 have. The value of this macro is used instead of that alignment to
1107 If this macro is not defined, then @var{basic-align} is used.
1109 The typical use of this macro is to increase alignment for string
1110 constants to be word aligned so that @code{strcpy} calls that copy
1111 constants can be done inline.
1113 @findex LOCAL_ALIGNMENT
1114 @item LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1115 If defined, a C expression to compute the alignment for a variable in
1116 the local store. @var{type} is the data type, and @var{basic-align} is
1117 the alignment that the object would ordinarily have. The value of this
1118 macro is used instead of that alignment to align the object.
1120 If this macro is not defined, then @var{basic-align} is used.
1122 One use of this macro is to increase alignment of medium-size data to
1123 make it all fit in fewer cache lines.
1125 @findex EMPTY_FIELD_BOUNDARY
1126 @item EMPTY_FIELD_BOUNDARY
1127 Alignment in bits to be given to a structure bit-field that follows an
1128 empty field such as @code{int : 0;}.
1130 Note that @code{PCC_BITFIELD_TYPE_MATTERS} also affects the alignment
1131 that results from an empty field.
1133 @findex STRUCTURE_SIZE_BOUNDARY
1134 @item STRUCTURE_SIZE_BOUNDARY
1135 Number of bits which any structure or union's size must be a multiple of.
1136 Each structure or union's size is rounded up to a multiple of this.
1138 If you do not define this macro, the default is the same as
1139 @code{BITS_PER_UNIT}.
1141 @findex STRICT_ALIGNMENT
1142 @item STRICT_ALIGNMENT
1143 Define this macro to be the value 1 if instructions will fail to work
1144 if given data not on the nominal alignment. If instructions will merely
1145 go slower in that case, define this macro as 0.
1147 @findex PCC_BITFIELD_TYPE_MATTERS
1148 @item PCC_BITFIELD_TYPE_MATTERS
1149 Define this if you wish to imitate the way many other C compilers handle
1150 alignment of bit-fields and the structures that contain them.
1152 The behavior is that the type written for a bit-field (@code{int},
1153 @code{short}, or other integer type) imposes an alignment for the
1154 entire structure, as if the structure really did contain an ordinary
1155 field of that type. In addition, the bit-field is placed within the
1156 structure so that it would fit within such a field, not crossing a
1159 Thus, on most machines, a bit-field whose type is written as @code{int}
1160 would not cross a four-byte boundary, and would force four-byte
1161 alignment for the whole structure. (The alignment used may not be four
1162 bytes; it is controlled by the other alignment parameters.)
1164 If the macro is defined, its definition should be a C expression;
1165 a nonzero value for the expression enables this behavior.
1167 Note that if this macro is not defined, or its value is zero, some
1168 bit-fields may cross more than one alignment boundary. The compiler can
1169 support such references if there are @samp{insv}, @samp{extv}, and
1170 @samp{extzv} insns that can directly reference memory.
1172 The other known way of making bit-fields work is to define
1173 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1174 Then every structure can be accessed with fullwords.
1176 Unless the machine has bit-field instructions or you define
1177 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1178 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1180 If your aim is to make GCC use the same conventions for laying out
1181 bit-fields as are used by another compiler, here is how to investigate
1182 what the other compiler does. Compile and run this program:
1201 printf ("Size of foo1 is %d\n",
1202 sizeof (struct foo1));
1203 printf ("Size of foo2 is %d\n",
1204 sizeof (struct foo2));
1209 If this prints 2 and 5, then the compiler's behavior is what you would
1210 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1212 @findex BITFIELD_NBYTES_LIMITED
1213 @item BITFIELD_NBYTES_LIMITED
1214 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1215 to aligning a bit-field within the structure.
1217 @findex MEMBER_TYPE_FORCES_BLK
1218 @item MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1219 Return 1 if a structure or array containing @var{field} should be accessed using
1222 If @var{field} is the only field in the structure, @var{mode} is its
1223 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1224 case where structures of one field would require the structure's mode to
1225 retain the field's mode.
1227 Normally, this is not needed. See the file @file{c4x.h} for an example
1228 of how to use this macro to prevent a structure having a floating point
1229 field from being accessed in an integer mode.
1231 @findex ROUND_TYPE_SIZE
1232 @item ROUND_TYPE_SIZE (@var{type}, @var{computed}, @var{specified})
1233 Define this macro as an expression for the overall size of a type
1234 (given by @var{type} as a tree node) when the size computed in the
1235 usual way is @var{computed} and the alignment is @var{specified}.
1237 The default is to round @var{computed} up to a multiple of @var{specified}.
1239 @findex ROUND_TYPE_SIZE_UNIT
1240 @item ROUND_TYPE_SIZE_UNIT (@var{type}, @var{computed}, @var{specified})
1241 Similar to @code{ROUND_TYPE_SIZE}, but sizes and alignments are
1242 specified in units (bytes). If you define @code{ROUND_TYPE_SIZE},
1243 you must also define this macro and they must be defined consistently
1246 @findex ROUND_TYPE_ALIGN
1247 @item ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1248 Define this macro as an expression for the alignment of a type (given
1249 by @var{type} as a tree node) if the alignment computed in the usual
1250 way is @var{computed} and the alignment explicitly specified was
1253 The default is to use @var{specified} if it is larger; otherwise, use
1254 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1256 @findex MAX_FIXED_MODE_SIZE
1257 @item MAX_FIXED_MODE_SIZE
1258 An integer expression for the size in bits of the largest integer
1259 machine mode that should actually be used. All integer machine modes of
1260 this size or smaller can be used for structures and unions with the
1261 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1262 (DImode)} is assumed.
1264 @findex VECTOR_MODE_SUPPORTED_P
1265 @item VECTOR_MODE_SUPPORTED_P(@var{mode})
1266 Define this macro to be nonzero if the port is prepared to handle insns
1267 involving vector mode @var{mode}. At the very least, it must have move
1268 patterns for this mode.
1270 @findex STACK_SAVEAREA_MODE
1271 @item STACK_SAVEAREA_MODE (@var{save_level})
1272 If defined, an expression of type @code{enum machine_mode} that
1273 specifies the mode of the save area operand of a
1274 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1275 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1276 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1277 having its mode specified.
1279 You need not define this macro if it always returns @code{Pmode}. You
1280 would most commonly define this macro if the
1281 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1284 @findex STACK_SIZE_MODE
1285 @item STACK_SIZE_MODE
1286 If defined, an expression of type @code{enum machine_mode} that
1287 specifies the mode of the size increment operand of an
1288 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1290 You need not define this macro if it always returns @code{word_mode}.
1291 You would most commonly define this macro if the @code{allocate_stack}
1292 pattern needs to support both a 32- and a 64-bit mode.
1294 @findex CHECK_FLOAT_VALUE
1295 @item CHECK_FLOAT_VALUE (@var{mode}, @var{value}, @var{overflow})
1296 A C statement to validate the value @var{value} (of type
1297 @code{double}) for mode @var{mode}. This means that you check whether
1298 @var{value} fits within the possible range of values for mode
1299 @var{mode} on this target machine. The mode @var{mode} is always
1300 a mode of class @code{MODE_FLOAT}. @var{overflow} is nonzero if
1301 the value is already known to be out of range.
1303 If @var{value} is not valid or if @var{overflow} is nonzero, you should
1304 set @var{overflow} to 1 and then assign some valid value to @var{value}.
1305 Allowing an invalid value to go through the compiler can produce
1306 incorrect assembler code which may even cause Unix assemblers to crash.
1308 This macro need not be defined if there is no work for it to do.
1310 @findex TARGET_FLOAT_FORMAT
1311 @item TARGET_FLOAT_FORMAT
1312 A code distinguishing the floating point format of the target machine.
1313 There are five defined values:
1316 @findex IEEE_FLOAT_FORMAT
1317 @item IEEE_FLOAT_FORMAT
1318 This code indicates IEEE floating point. It is the default; there is no
1319 need to define this macro when the format is IEEE@.
1321 @findex VAX_FLOAT_FORMAT
1322 @item VAX_FLOAT_FORMAT
1323 This code indicates the ``F float'' (for @code{float}) and ``D float''
1324 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
1326 @findex IBM_FLOAT_FORMAT
1327 @item IBM_FLOAT_FORMAT
1328 This code indicates the format used on the IBM System/370.
1330 @findex C4X_FLOAT_FORMAT
1331 @item C4X_FLOAT_FORMAT
1332 This code indicates the format used on the TMS320C3x/C4x.
1334 @findex UNKNOWN_FLOAT_FORMAT
1335 @item UNKNOWN_FLOAT_FORMAT
1336 This code indicates any other format.
1340 formats are actually in use on supported machines, new codes should be
1343 The ordering of the component words of floating point values stored in
1344 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1346 @findex VAX_HALFWORD_ORDER
1347 @item VAX_HALFWORD_ORDER
1348 This macro is only used if @code{TARGET_FLOAT_FORMAT} is
1349 @code{VAX_FLOAT_FORMAT}. If defaulted or defined as 1, the halfwords of
1350 the generated floating point data are in the order used by the VAX. If
1351 defined as 0, they are reversed, which is used by the PDP-11 target.
1353 @findex MODE_HAS_NANS
1354 @item MODE_HAS_NANS (@var{mode})
1355 When defined, this macro should be true if @var{mode} has a NaN
1356 representation. The compiler assumes that NaNs are not equal to
1357 anything (including themselves) and that addition, subtraction,
1358 multiplication and division all return NaNs when one operand is
1361 By default, this macro is true if @var{mode} is a floating-point
1362 mode and the target floating-point format is IEEE@.
1364 @findex MODE_HAS_INFINITIES
1365 @item MODE_HAS_INFINITIES (@var{mode})
1366 This macro should be true if @var{mode} can represent infinity. At
1367 present, the compiler uses this macro to decide whether @samp{x - x}
1368 is always defined. By default, the macro is true when @var{mode}
1369 is a floating-point mode and the target format is IEEE@.
1371 @findex MODE_HAS_SIGNED_ZEROS
1372 @item MODE_HAS_SIGNED_ZEROS (@var{mode})
1373 True if @var{mode} distinguishes between positive and negative zero.
1374 The rules are expected to follow the IEEE standard:
1378 @samp{x + x} has the same sign as @samp{x}.
1381 If the sum of two values with opposite sign is zero, the result is
1382 positive for all rounding modes expect towards @minus{}infinity, for
1383 which it is negative.
1386 The sign of a product or quotient is negative when exactly one
1387 of the operands is negative.
1390 The default definition is true if @var{mode} is a floating-point
1391 mode and the target format is IEEE@.
1393 @findex MODE_HAS_SIGN_DEPENDENT_ROUNDING
1394 @item MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1395 If defined, this macro should be true for @var{mode} if it has at
1396 least one rounding mode in which @samp{x} and @samp{-x} can be
1397 rounded to numbers of different magnitude. Two such modes are
1398 towards @minus{}infinity and towards +infinity.
1400 The default definition of this macro is true if @var{mode} is
1401 a floating-point mode and the target format is IEEE@.
1403 @findex ROUND_TOWARDS_ZERO
1404 @item ROUND_TOWARDS_ZERO
1405 If defined, this macro should be true if the prevailing rounding
1406 mode is towards zero. A true value has the following effects:
1410 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1413 @file{libgcc.a}'s floating-point emulator will round towards zero
1414 rather than towards nearest.
1417 The compiler's floating-point emulator will round towards zero after
1418 doing arithmetic, and when converting from the internal float format to
1422 The macro does not affect the parsing of string literals. When the
1423 primary rounding mode is towards zero, library functions like
1424 @code{strtod} might still round towards nearest, and the compiler's
1425 parser should behave like the target's @code{strtod} where possible.
1427 Not defining this macro is equivalent to returning zero.
1429 @findex LARGEST_EXPONENT_IS_NORMAL
1430 @item LARGEST_EXPONENT_IS_NORMAL (@var{size})
1431 This macro should return true if floats with @var{size}
1432 bits do not have a NaN or infinity representation, but use the largest
1433 exponent for normal numbers instead.
1435 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1436 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1437 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1438 floating-point arithmetic.
1440 The default definition of this macro returns false for all sizes.
1443 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1444 This target hook returns @code{true} if bit-fields in the given
1445 @var{record_type} are to be laid out following the rules of Microsoft
1446 Visual C/C++, namely: (i) a bit-field won't share the same storage
1447 unit with the previous bit-field if their underlying types have
1448 different sizes, and the bit-field will be aligned to the highest
1449 alignment of the underlying types of itself and of the previous
1450 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1451 the whole enclosing structure, even if it is unnamed; except that
1452 (iii) a zero-sized bit-field will be disregarded unless it follows
1453 another bit-field of non-zero size. If this hook returns @code{true},
1454 other macros that control bit-field layout are ignored.
1458 @section Layout of Source Language Data Types
1460 These macros define the sizes and other characteristics of the standard
1461 basic data types used in programs being compiled. Unlike the macros in
1462 the previous section, these apply to specific features of C and related
1463 languages, rather than to fundamental aspects of storage layout.
1466 @findex INT_TYPE_SIZE
1468 A C expression for the size in bits of the type @code{int} on the
1469 target machine. If you don't define this, the default is one word.
1471 @findex SHORT_TYPE_SIZE
1472 @item SHORT_TYPE_SIZE
1473 A C expression for the size in bits of the type @code{short} on the
1474 target machine. If you don't define this, the default is half a word.
1475 (If this would be less than one storage unit, it is rounded up to one
1478 @findex LONG_TYPE_SIZE
1479 @item LONG_TYPE_SIZE
1480 A C expression for the size in bits of the type @code{long} on the
1481 target machine. If you don't define this, the default is one word.
1483 @findex ADA_LONG_TYPE_SIZE
1484 @item ADA_LONG_TYPE_SIZE
1485 On some machines, the size used for the Ada equivalent of the type
1486 @code{long} by a native Ada compiler differs from that used by C. In
1487 that situation, define this macro to be a C expression to be used for
1488 the size of that type. If you don't define this, the default is the
1489 value of @code{LONG_TYPE_SIZE}.
1491 @findex MAX_LONG_TYPE_SIZE
1492 @item MAX_LONG_TYPE_SIZE
1493 Maximum number for the size in bits of the type @code{long} on the
1494 target machine. If this is undefined, the default is
1495 @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the
1496 largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is
1499 @findex LONG_LONG_TYPE_SIZE
1500 @item LONG_LONG_TYPE_SIZE
1501 A C expression for the size in bits of the type @code{long long} on the
1502 target machine. If you don't define this, the default is two
1503 words. If you want to support GNU Ada on your machine, the value of this
1504 macro must be at least 64.
1506 @findex CHAR_TYPE_SIZE
1507 @item CHAR_TYPE_SIZE
1508 A C expression for the size in bits of the type @code{char} on the
1509 target machine. If you don't define this, the default is
1510 @code{BITS_PER_UNIT}.
1512 @findex BOOL_TYPE_SIZE
1513 @item BOOL_TYPE_SIZE
1514 A C expression for the size in bits of the C++ type @code{bool} and
1515 C99 type @code{_Bool} on the target machine. If you don't define
1516 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1518 @findex FLOAT_TYPE_SIZE
1519 @item FLOAT_TYPE_SIZE
1520 A C expression for the size in bits of the type @code{float} on the
1521 target machine. If you don't define this, the default is one word.
1523 @findex DOUBLE_TYPE_SIZE
1524 @item DOUBLE_TYPE_SIZE
1525 A C expression for the size in bits of the type @code{double} on the
1526 target machine. If you don't define this, the default is two
1529 @findex LONG_DOUBLE_TYPE_SIZE
1530 @item LONG_DOUBLE_TYPE_SIZE
1531 A C expression for the size in bits of the type @code{long double} on
1532 the target machine. If you don't define this, the default is two
1535 @findex MAX_LONG_DOUBLE_TYPE_SIZE
1536 Maximum number for the size in bits of the type @code{long double} on the
1537 target machine. If this is undefined, the default is
1538 @code{LONG_DOUBLE_TYPE_SIZE}. Otherwise, it is the constant value that is
1539 the largest value that @code{LONG_DOUBLE_TYPE_SIZE} can have at run-time.
1540 This is used in @code{cpp}.
1542 @findex INTEL_EXTENDED_IEEE_FORMAT
1543 Define this macro to be 1 if the target machine uses 80-bit floating-point
1544 values with 128-bit size and alignment. This is used in @file{real.c}.
1546 @findex WIDEST_HARDWARE_FP_SIZE
1547 @item WIDEST_HARDWARE_FP_SIZE
1548 A C expression for the size in bits of the widest floating-point format
1549 supported by the hardware. If you define this macro, you must specify a
1550 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1551 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1554 @findex DEFAULT_SIGNED_CHAR
1555 @item DEFAULT_SIGNED_CHAR
1556 An expression whose value is 1 or 0, according to whether the type
1557 @code{char} should be signed or unsigned by default. The user can
1558 always override this default with the options @option{-fsigned-char}
1559 and @option{-funsigned-char}.
1561 @findex DEFAULT_SHORT_ENUMS
1562 @item DEFAULT_SHORT_ENUMS
1563 A C expression to determine whether to give an @code{enum} type
1564 only as many bytes as it takes to represent the range of possible values
1565 of that type. A nonzero value means to do that; a zero value means all
1566 @code{enum} types should be allocated like @code{int}.
1568 If you don't define the macro, the default is 0.
1572 A C expression for a string describing the name of the data type to use
1573 for size values. The typedef name @code{size_t} is defined using the
1574 contents of the string.
1576 The string can contain more than one keyword. If so, separate them with
1577 spaces, and write first any length keyword, then @code{unsigned} if
1578 appropriate, and finally @code{int}. The string must exactly match one
1579 of the data type names defined in the function
1580 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1581 omit @code{int} or change the order---that would cause the compiler to
1584 If you don't define this macro, the default is @code{"long unsigned
1587 @findex PTRDIFF_TYPE
1589 A C expression for a string describing the name of the data type to use
1590 for the result of subtracting two pointers. The typedef name
1591 @code{ptrdiff_t} is defined using the contents of the string. See
1592 @code{SIZE_TYPE} above for more information.
1594 If you don't define this macro, the default is @code{"long int"}.
1598 A C expression for a string describing the name of the data type to use
1599 for wide characters. The typedef name @code{wchar_t} is defined using
1600 the contents of the string. See @code{SIZE_TYPE} above for more
1603 If you don't define this macro, the default is @code{"int"}.
1605 @findex WCHAR_TYPE_SIZE
1606 @item WCHAR_TYPE_SIZE
1607 A C expression for the size in bits of the data type for wide
1608 characters. This is used in @code{cpp}, which cannot make use of
1611 @findex MAX_WCHAR_TYPE_SIZE
1612 @item MAX_WCHAR_TYPE_SIZE
1613 Maximum number for the size in bits of the data type for wide
1614 characters. If this is undefined, the default is
1615 @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1616 largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is
1619 @findex GCOV_TYPE_SIZE
1620 @item GCOV_TYPE_SIZE
1621 A C expression for the size in bits of the type used for gcov counters on the
1622 target machine. If you don't define this, the default is one
1623 @code{LONG_TYPE_SIZE} in case it is greater or equal to 64-bit and
1624 @code{LONG_LONG_TYPE_SIZE} otherwise. You may want to re-define the type to
1625 ensure atomicity for counters in multithreaded programs.
1629 A C expression for a string describing the name of the data type to
1630 use for wide characters passed to @code{printf} and returned from
1631 @code{getwc}. The typedef name @code{wint_t} is defined using the
1632 contents of the string. See @code{SIZE_TYPE} above for more
1635 If you don't define this macro, the default is @code{"unsigned int"}.
1639 A C expression for a string describing the name of the data type that
1640 can represent any value of any standard or extended signed integer type.
1641 The typedef name @code{intmax_t} is defined using the contents of the
1642 string. See @code{SIZE_TYPE} above for more information.
1644 If you don't define this macro, the default is the first of
1645 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1646 much precision as @code{long long int}.
1648 @findex UINTMAX_TYPE
1650 A C expression for a string describing the name of the data type that
1651 can represent any value of any standard or extended unsigned integer
1652 type. The typedef name @code{uintmax_t} is defined using the contents
1653 of the string. See @code{SIZE_TYPE} above for more information.
1655 If you don't define this macro, the default is the first of
1656 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1657 unsigned int"} that has as much precision as @code{long long unsigned
1660 @findex TARGET_PTRMEMFUNC_VBIT_LOCATION
1661 @item TARGET_PTRMEMFUNC_VBIT_LOCATION
1662 The C++ compiler represents a pointer-to-member-function with a struct
1669 ptrdiff_t vtable_index;
1676 The C++ compiler must use one bit to indicate whether the function that
1677 will be called through a pointer-to-member-function is virtual.
1678 Normally, we assume that the low-order bit of a function pointer must
1679 always be zero. Then, by ensuring that the vtable_index is odd, we can
1680 distinguish which variant of the union is in use. But, on some
1681 platforms function pointers can be odd, and so this doesn't work. In
1682 that case, we use the low-order bit of the @code{delta} field, and shift
1683 the remainder of the @code{delta} field to the left.
1685 GCC will automatically make the right selection about where to store
1686 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1687 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1688 set such that functions always start at even addresses, but the lowest
1689 bit of pointers to functions indicate whether the function at that
1690 address is in ARM or Thumb mode. If this is the case of your
1691 architecture, you should define this macro to
1692 @code{ptrmemfunc_vbit_in_delta}.
1694 In general, you should not have to define this macro. On architectures
1695 in which function addresses are always even, according to
1696 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1697 @code{ptrmemfunc_vbit_in_pfn}.
1699 @findex TARGET_VTABLE_USES_DESCRIPTORS
1700 @item TARGET_VTABLE_USES_DESCRIPTORS
1701 Normally, the C++ compiler uses function pointers in vtables. This
1702 macro allows the target to change to use ``function descriptors''
1703 instead. Function descriptors are found on targets for whom a
1704 function pointer is actually a small data structure. Normally the
1705 data structure consists of the actual code address plus a data
1706 pointer to which the function's data is relative.
1708 If vtables are used, the value of this macro should be the number
1709 of words that the function descriptor occupies.
1711 @findex TARGET_VTABLE_ENTRY_ALIGN
1712 @item TARGET_VTABLE_ENTRY_ALIGN
1713 By default, the vtable entries are void pointers, the so the alignment
1714 is the same as pointer alignment. The value of this macro specifies
1715 the alignment of the vtable entry in bits. It should be defined only
1716 when special alignment is necessary. */
1718 @findex TARGET_VTABLE_DATA_ENTRY_DISTANCE
1719 @item TARGET_VTABLE_DATA_ENTRY_DISTANCE
1720 There are a few non-descriptor entries in the vtable at offsets below
1721 zero. If these entries must be padded (say, to preserve the alignment
1722 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1723 of words in each data entry.
1726 @node Escape Sequences
1727 @section Target Character Escape Sequences
1728 @cindex escape sequences
1730 By default, GCC assumes that the C character escape sequences take on
1731 their ASCII values for the target. If this is not correct, you must
1732 explicitly define all of the macros below.
1737 A C constant expression for the integer value for escape sequence
1742 A C constant expression for the integer value of the target escape
1743 character. As an extension, GCC evaluates the escape sequences
1744 @samp{\e} and @samp{\E} to this.
1748 @findex TARGET_NEWLINE
1751 @itemx TARGET_NEWLINE
1752 C constant expressions for the integer values for escape sequences
1753 @samp{\b}, @samp{\t} and @samp{\n}.
1761 C constant expressions for the integer values for escape sequences
1762 @samp{\v}, @samp{\f} and @samp{\r}.
1766 @section Register Usage
1767 @cindex register usage
1769 This section explains how to describe what registers the target machine
1770 has, and how (in general) they can be used.
1772 The description of which registers a specific instruction can use is
1773 done with register classes; see @ref{Register Classes}. For information
1774 on using registers to access a stack frame, see @ref{Frame Registers}.
1775 For passing values in registers, see @ref{Register Arguments}.
1776 For returning values in registers, see @ref{Scalar Return}.
1779 * Register Basics:: Number and kinds of registers.
1780 * Allocation Order:: Order in which registers are allocated.
1781 * Values in Registers:: What kinds of values each reg can hold.
1782 * Leaf Functions:: Renumbering registers for leaf functions.
1783 * Stack Registers:: Handling a register stack such as 80387.
1786 @node Register Basics
1787 @subsection Basic Characteristics of Registers
1789 @c prevent bad page break with this line
1790 Registers have various characteristics.
1793 @findex FIRST_PSEUDO_REGISTER
1794 @item FIRST_PSEUDO_REGISTER
1795 Number of hardware registers known to the compiler. They receive
1796 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1797 pseudo register's number really is assigned the number
1798 @code{FIRST_PSEUDO_REGISTER}.
1800 @item FIXED_REGISTERS
1801 @findex FIXED_REGISTERS
1802 @cindex fixed register
1803 An initializer that says which registers are used for fixed purposes
1804 all throughout the compiled code and are therefore not available for
1805 general allocation. These would include the stack pointer, the frame
1806 pointer (except on machines where that can be used as a general
1807 register when no frame pointer is needed), the program counter on
1808 machines where that is considered one of the addressable registers,
1809 and any other numbered register with a standard use.
1811 This information is expressed as a sequence of numbers, separated by
1812 commas and surrounded by braces. The @var{n}th number is 1 if
1813 register @var{n} is fixed, 0 otherwise.
1815 The table initialized from this macro, and the table initialized by
1816 the following one, may be overridden at run time either automatically,
1817 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1818 the user with the command options @option{-ffixed-@var{reg}},
1819 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1821 @findex CALL_USED_REGISTERS
1822 @item CALL_USED_REGISTERS
1823 @cindex call-used register
1824 @cindex call-clobbered register
1825 @cindex call-saved register
1826 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1827 clobbered (in general) by function calls as well as for fixed
1828 registers. This macro therefore identifies the registers that are not
1829 available for general allocation of values that must live across
1832 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1833 automatically saves it on function entry and restores it on function
1834 exit, if the register is used within the function.
1836 @findex CALL_REALLY_USED_REGISTERS
1837 @item CALL_REALLY_USED_REGISTERS
1838 @cindex call-used register
1839 @cindex call-clobbered register
1840 @cindex call-saved register
1841 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1842 that the entire set of @code{FIXED_REGISTERS} be included.
1843 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1844 This macro is optional. If not specified, it defaults to the value
1845 of @code{CALL_USED_REGISTERS}.
1847 @findex HARD_REGNO_CALL_PART_CLOBBERED
1848 @item HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1849 @cindex call-used register
1850 @cindex call-clobbered register
1851 @cindex call-saved register
1852 A C expression that is nonzero if it is not permissible to store a
1853 value of mode @var{mode} in hard register number @var{regno} across a
1854 call without some part of it being clobbered. For most machines this
1855 macro need not be defined. It is only required for machines that do not
1856 preserve the entire contents of a register across a call.
1858 @findex CONDITIONAL_REGISTER_USAGE
1860 @findex call_used_regs
1861 @item CONDITIONAL_REGISTER_USAGE
1862 Zero or more C statements that may conditionally modify five variables
1863 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1864 @code{reg_names}, and @code{reg_class_contents}, to take into account
1865 any dependence of these register sets on target flags. The first three
1866 of these are of type @code{char []} (interpreted as Boolean vectors).
1867 @code{global_regs} is a @code{const char *[]}, and
1868 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1869 called, @code{fixed_regs}, @code{call_used_regs},
1870 @code{reg_class_contents}, and @code{reg_names} have been initialized
1871 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1872 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1873 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1874 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1875 command options have been applied.
1877 You need not define this macro if it has no work to do.
1879 @cindex disabling certain registers
1880 @cindex controlling register usage
1881 If the usage of an entire class of registers depends on the target
1882 flags, you may indicate this to GCC by using this macro to modify
1883 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1884 registers in the classes which should not be used by GCC@. Also define
1885 the macro @code{REG_CLASS_FROM_LETTER} to return @code{NO_REGS} if it
1886 is called with a letter for a class that shouldn't be used.
1888 (However, if this class is not included in @code{GENERAL_REGS} and all
1889 of the insn patterns whose constraints permit this class are
1890 controlled by target switches, then GCC will automatically avoid using
1891 these registers when the target switches are opposed to them.)
1893 @findex NON_SAVING_SETJMP
1894 @item NON_SAVING_SETJMP
1895 If this macro is defined and has a nonzero value, it means that
1896 @code{setjmp} and related functions fail to save the registers, or that
1897 @code{longjmp} fails to restore them. To compensate, the compiler
1898 avoids putting variables in registers in functions that use
1901 @findex INCOMING_REGNO
1902 @item INCOMING_REGNO (@var{out})
1903 Define this macro if the target machine has register windows. This C
1904 expression returns the register number as seen by the called function
1905 corresponding to the register number @var{out} as seen by the calling
1906 function. Return @var{out} if register number @var{out} is not an
1909 @findex OUTGOING_REGNO
1910 @item OUTGOING_REGNO (@var{in})
1911 Define this macro if the target machine has register windows. This C
1912 expression returns the register number as seen by the calling function
1913 corresponding to the register number @var{in} as seen by the called
1914 function. Return @var{in} if register number @var{in} is not an inbound
1918 @item LOCAL_REGNO (@var{regno})
1919 Define this macro if the target machine has register windows. This C
1920 expression returns true if the register is call-saved but is in the
1921 register window. Unlike most call-saved registers, such registers
1922 need not be explicitly restored on function exit or during non-local
1928 If the program counter has a register number, define this as that
1929 register number. Otherwise, do not define it.
1933 @node Allocation Order
1934 @subsection Order of Allocation of Registers
1935 @cindex order of register allocation
1936 @cindex register allocation order
1938 @c prevent bad page break with this line
1939 Registers are allocated in order.
1942 @findex REG_ALLOC_ORDER
1943 @item REG_ALLOC_ORDER
1944 If defined, an initializer for a vector of integers, containing the
1945 numbers of hard registers in the order in which GCC should prefer
1946 to use them (from most preferred to least).
1948 If this macro is not defined, registers are used lowest numbered first
1949 (all else being equal).
1951 One use of this macro is on machines where the highest numbered
1952 registers must always be saved and the save-multiple-registers
1953 instruction supports only sequences of consecutive registers. On such
1954 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1955 the highest numbered allocable register first.
1957 @findex ORDER_REGS_FOR_LOCAL_ALLOC
1958 @item ORDER_REGS_FOR_LOCAL_ALLOC
1959 A C statement (sans semicolon) to choose the order in which to allocate
1960 hard registers for pseudo-registers local to a basic block.
1962 Store the desired register order in the array @code{reg_alloc_order}.
1963 Element 0 should be the register to allocate first; element 1, the next
1964 register; and so on.
1966 The macro body should not assume anything about the contents of
1967 @code{reg_alloc_order} before execution of the macro.
1969 On most machines, it is not necessary to define this macro.
1972 @node Values in Registers
1973 @subsection How Values Fit in Registers
1975 This section discusses the macros that describe which kinds of values
1976 (specifically, which machine modes) each register can hold, and how many
1977 consecutive registers are needed for a given mode.
1980 @findex HARD_REGNO_NREGS
1981 @item HARD_REGNO_NREGS (@var{regno}, @var{mode})
1982 A C expression for the number of consecutive hard registers, starting
1983 at register number @var{regno}, required to hold a value of mode
1986 On a machine where all registers are exactly one word, a suitable
1987 definition of this macro is
1990 #define HARD_REGNO_NREGS(REGNO, MODE) \
1991 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
1995 @findex HARD_REGNO_MODE_OK
1996 @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
1997 A C expression that is nonzero if it is permissible to store a value
1998 of mode @var{mode} in hard register number @var{regno} (or in several
1999 registers starting with that one). For a machine where all registers
2000 are equivalent, a suitable definition is
2003 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2006 You need not include code to check for the numbers of fixed registers,
2007 because the allocation mechanism considers them to be always occupied.
2009 @cindex register pairs
2010 On some machines, double-precision values must be kept in even/odd
2011 register pairs. You can implement that by defining this macro to reject
2012 odd register numbers for such modes.
2014 The minimum requirement for a mode to be OK in a register is that the
2015 @samp{mov@var{mode}} instruction pattern support moves between the
2016 register and other hard register in the same class and that moving a
2017 value into the register and back out not alter it.
2019 Since the same instruction used to move @code{word_mode} will work for
2020 all narrower integer modes, it is not necessary on any machine for
2021 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2022 you define patterns @samp{movhi}, etc., to take advantage of this. This
2023 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2024 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2027 Many machines have special registers for floating point arithmetic.
2028 Often people assume that floating point machine modes are allowed only
2029 in floating point registers. This is not true. Any registers that
2030 can hold integers can safely @emph{hold} a floating point machine
2031 mode, whether or not floating arithmetic can be done on it in those
2032 registers. Integer move instructions can be used to move the values.
2034 On some machines, though, the converse is true: fixed-point machine
2035 modes may not go in floating registers. This is true if the floating
2036 registers normalize any value stored in them, because storing a
2037 non-floating value there would garble it. In this case,
2038 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2039 floating registers. But if the floating registers do not automatically
2040 normalize, if you can store any bit pattern in one and retrieve it
2041 unchanged without a trap, then any machine mode may go in a floating
2042 register, so you can define this macro to say so.
2044 The primary significance of special floating registers is rather that
2045 they are the registers acceptable in floating point arithmetic
2046 instructions. However, this is of no concern to
2047 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2048 constraints for those instructions.
2050 On some machines, the floating registers are especially slow to access,
2051 so that it is better to store a value in a stack frame than in such a
2052 register if floating point arithmetic is not being done. As long as the
2053 floating registers are not in class @code{GENERAL_REGS}, they will not
2054 be used unless some pattern's constraint asks for one.
2056 @findex MODES_TIEABLE_P
2057 @item MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2058 A C expression that is nonzero if a value of mode
2059 @var{mode1} is accessible in mode @var{mode2} without copying.
2061 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2062 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2063 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2064 should be nonzero. If they differ for any @var{r}, you should define
2065 this macro to return zero unless some other mechanism ensures the
2066 accessibility of the value in a narrower mode.
2068 You should define this macro to return nonzero in as many cases as
2069 possible since doing so will allow GCC to perform better register
2072 @findex AVOID_CCMODE_COPIES
2073 @item AVOID_CCMODE_COPIES
2074 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2075 registers. You should only define this macro if support for copying to/from
2076 @code{CCmode} is incomplete.
2079 @node Leaf Functions
2080 @subsection Handling Leaf Functions
2082 @cindex leaf functions
2083 @cindex functions, leaf
2084 On some machines, a leaf function (i.e., one which makes no calls) can run
2085 more efficiently if it does not make its own register window. Often this
2086 means it is required to receive its arguments in the registers where they
2087 are passed by the caller, instead of the registers where they would
2090 The special treatment for leaf functions generally applies only when
2091 other conditions are met; for example, often they may use only those
2092 registers for its own variables and temporaries. We use the term ``leaf
2093 function'' to mean a function that is suitable for this special
2094 handling, so that functions with no calls are not necessarily ``leaf
2097 GCC assigns register numbers before it knows whether the function is
2098 suitable for leaf function treatment. So it needs to renumber the
2099 registers in order to output a leaf function. The following macros
2103 @findex LEAF_REGISTERS
2104 @item LEAF_REGISTERS
2105 Name of a char vector, indexed by hard register number, which
2106 contains 1 for a register that is allowable in a candidate for leaf
2109 If leaf function treatment involves renumbering the registers, then the
2110 registers marked here should be the ones before renumbering---those that
2111 GCC would ordinarily allocate. The registers which will actually be
2112 used in the assembler code, after renumbering, should not be marked with 1
2115 Define this macro only if the target machine offers a way to optimize
2116 the treatment of leaf functions.
2118 @findex LEAF_REG_REMAP
2119 @item LEAF_REG_REMAP (@var{regno})
2120 A C expression whose value is the register number to which @var{regno}
2121 should be renumbered, when a function is treated as a leaf function.
2123 If @var{regno} is a register number which should not appear in a leaf
2124 function before renumbering, then the expression should yield @minus{}1, which
2125 will cause the compiler to abort.
2127 Define this macro only if the target machine offers a way to optimize the
2128 treatment of leaf functions, and registers need to be renumbered to do
2132 @findex current_function_is_leaf
2133 @findex current_function_uses_only_leaf_regs
2134 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2135 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2136 specially. They can test the C variable @code{current_function_is_leaf}
2137 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2138 set prior to local register allocation and is valid for the remaining
2139 compiler passes. They can also test the C variable
2140 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2141 functions which only use leaf registers.
2142 @code{current_function_uses_only_leaf_regs} is valid after reload and is
2143 only useful if @code{LEAF_REGISTERS} is defined.
2144 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2145 @c of the next paragraph?! --mew 2feb93
2147 @node Stack Registers
2148 @subsection Registers That Form a Stack
2150 There are special features to handle computers where some of the
2151 ``registers'' form a stack, as in the 80387 coprocessor for the 80386.
2152 Stack registers are normally written by pushing onto the stack, and are
2153 numbered relative to the top of the stack.
2155 Currently, GCC can only handle one group of stack-like registers, and
2156 they must be consecutively numbered.
2161 Define this if the machine has any stack-like registers.
2163 @findex FIRST_STACK_REG
2164 @item FIRST_STACK_REG
2165 The number of the first stack-like register. This one is the top
2168 @findex LAST_STACK_REG
2169 @item LAST_STACK_REG
2170 The number of the last stack-like register. This one is the bottom of
2174 @node Register Classes
2175 @section Register Classes
2176 @cindex register class definitions
2177 @cindex class definitions, register
2179 On many machines, the numbered registers are not all equivalent.
2180 For example, certain registers may not be allowed for indexed addressing;
2181 certain registers may not be allowed in some instructions. These machine
2182 restrictions are described to the compiler using @dfn{register classes}.
2184 You define a number of register classes, giving each one a name and saying
2185 which of the registers belong to it. Then you can specify register classes
2186 that are allowed as operands to particular instruction patterns.
2190 In general, each register will belong to several classes. In fact, one
2191 class must be named @code{ALL_REGS} and contain all the registers. Another
2192 class must be named @code{NO_REGS} and contain no registers. Often the
2193 union of two classes will be another class; however, this is not required.
2195 @findex GENERAL_REGS
2196 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2197 terribly special about the name, but the operand constraint letters
2198 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2199 the same as @code{ALL_REGS}, just define it as a macro which expands
2202 Order the classes so that if class @var{x} is contained in class @var{y}
2203 then @var{x} has a lower class number than @var{y}.
2205 The way classes other than @code{GENERAL_REGS} are specified in operand
2206 constraints is through machine-dependent operand constraint letters.
2207 You can define such letters to correspond to various classes, then use
2208 them in operand constraints.
2210 You should define a class for the union of two classes whenever some
2211 instruction allows both classes. For example, if an instruction allows
2212 either a floating point (coprocessor) register or a general register for a
2213 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2214 which includes both of them. Otherwise you will get suboptimal code.
2216 You must also specify certain redundant information about the register
2217 classes: for each class, which classes contain it and which ones are
2218 contained in it; for each pair of classes, the largest class contained
2221 When a value occupying several consecutive registers is expected in a
2222 certain class, all the registers used must belong to that class.
2223 Therefore, register classes cannot be used to enforce a requirement for
2224 a register pair to start with an even-numbered register. The way to
2225 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2227 Register classes used for input-operands of bitwise-and or shift
2228 instructions have a special requirement: each such class must have, for
2229 each fixed-point machine mode, a subclass whose registers can transfer that
2230 mode to or from memory. For example, on some machines, the operations for
2231 single-byte values (@code{QImode}) are limited to certain registers. When
2232 this is so, each register class that is used in a bitwise-and or shift
2233 instruction must have a subclass consisting of registers from which
2234 single-byte values can be loaded or stored. This is so that
2235 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2238 @findex enum reg_class
2239 @item enum reg_class
2240 An enumeral type that must be defined with all the register class names
2241 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
2242 must be the last register class, followed by one more enumeral value,
2243 @code{LIM_REG_CLASSES}, which is not a register class but rather
2244 tells how many classes there are.
2246 Each register class has a number, which is the value of casting
2247 the class name to type @code{int}. The number serves as an index
2248 in many of the tables described below.
2250 @findex N_REG_CLASSES
2252 The number of distinct register classes, defined as follows:
2255 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2258 @findex REG_CLASS_NAMES
2259 @item REG_CLASS_NAMES
2260 An initializer containing the names of the register classes as C string
2261 constants. These names are used in writing some of the debugging dumps.
2263 @findex REG_CLASS_CONTENTS
2264 @item REG_CLASS_CONTENTS
2265 An initializer containing the contents of the register classes, as integers
2266 which are bit masks. The @var{n}th integer specifies the contents of class
2267 @var{n}. The way the integer @var{mask} is interpreted is that
2268 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2270 When the machine has more than 32 registers, an integer does not suffice.
2271 Then the integers are replaced by sub-initializers, braced groupings containing
2272 several integers. Each sub-initializer must be suitable as an initializer
2273 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2274 In this situation, the first integer in each sub-initializer corresponds to
2275 registers 0 through 31, the second integer to registers 32 through 63, and
2278 @findex REGNO_REG_CLASS
2279 @item REGNO_REG_CLASS (@var{regno})
2280 A C expression whose value is a register class containing hard register
2281 @var{regno}. In general there is more than one such class; choose a class
2282 which is @dfn{minimal}, meaning that no smaller class also contains the
2285 @findex BASE_REG_CLASS
2286 @item BASE_REG_CLASS
2287 A macro whose definition is the name of the class to which a valid
2288 base register must belong. A base register is one used in an address
2289 which is the register value plus a displacement.
2291 @findex MODE_BASE_REG_CLASS
2292 @item MODE_BASE_REG_CLASS (@var{mode})
2293 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2294 the selection of a base register in a mode depenedent manner. If
2295 @var{mode} is VOIDmode then it should return the same value as
2296 @code{BASE_REG_CLASS}.
2298 @findex INDEX_REG_CLASS
2299 @item INDEX_REG_CLASS
2300 A macro whose definition is the name of the class to which a valid
2301 index register must belong. An index register is one used in an
2302 address where its value is either multiplied by a scale factor or
2303 added to another register (as well as added to a displacement).
2305 @findex REG_CLASS_FROM_LETTER
2306 @item REG_CLASS_FROM_LETTER (@var{char})
2307 A C expression which defines the machine-dependent operand constraint
2308 letters for register classes. If @var{char} is such a letter, the
2309 value should be the register class corresponding to it. Otherwise,
2310 the value should be @code{NO_REGS}. The register letter @samp{r},
2311 corresponding to class @code{GENERAL_REGS}, will not be passed
2312 to this macro; you do not need to handle it.
2314 @findex REGNO_OK_FOR_BASE_P
2315 @item REGNO_OK_FOR_BASE_P (@var{num})
2316 A C expression which is nonzero if register number @var{num} is
2317 suitable for use as a base register in operand addresses. It may be
2318 either a suitable hard register or a pseudo register that has been
2319 allocated such a hard register.
2321 @findex REGNO_MODE_OK_FOR_BASE_P
2322 @item REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2323 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2324 that expression may examine the mode of the memory reference in
2325 @var{mode}. You should define this macro if the mode of the memory
2326 reference affects whether a register may be used as a base register. If
2327 you define this macro, the compiler will use it instead of
2328 @code{REGNO_OK_FOR_BASE_P}.
2330 @findex REGNO_OK_FOR_INDEX_P
2331 @item REGNO_OK_FOR_INDEX_P (@var{num})
2332 A C expression which is nonzero if register number @var{num} is
2333 suitable for use as an index register in operand addresses. It may be
2334 either a suitable hard register or a pseudo register that has been
2335 allocated such a hard register.
2337 The difference between an index register and a base register is that
2338 the index register may be scaled. If an address involves the sum of
2339 two registers, neither one of them scaled, then either one may be
2340 labeled the ``base'' and the other the ``index''; but whichever
2341 labeling is used must fit the machine's constraints of which registers
2342 may serve in each capacity. The compiler will try both labelings,
2343 looking for one that is valid, and will reload one or both registers
2344 only if neither labeling works.
2346 @findex PREFERRED_RELOAD_CLASS
2347 @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2348 A C expression that places additional restrictions on the register class
2349 to use when it is necessary to copy value @var{x} into a register in class
2350 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2351 another, smaller class. On many machines, the following definition is
2355 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2358 Sometimes returning a more restrictive class makes better code. For
2359 example, on the 68000, when @var{x} is an integer constant that is in range
2360 for a @samp{moveq} instruction, the value of this macro is always
2361 @code{DATA_REGS} as long as @var{class} includes the data registers.
2362 Requiring a data register guarantees that a @samp{moveq} will be used.
2364 If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
2365 you can force @var{x} into a memory constant. This is useful on
2366 certain machines where immediate floating values cannot be loaded into
2367 certain kinds of registers.
2369 @findex PREFERRED_OUTPUT_RELOAD_CLASS
2370 @item PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2371 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2372 input reloads. If you don't define this macro, the default is to use
2373 @var{class}, unchanged.
2375 @findex LIMIT_RELOAD_CLASS
2376 @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2377 A C expression that places additional restrictions on the register class
2378 to use when it is necessary to be able to hold a value of mode
2379 @var{mode} in a reload register for which class @var{class} would
2382 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2383 there are certain modes that simply can't go in certain reload classes.
2385 The value is a register class; perhaps @var{class}, or perhaps another,
2388 Don't define this macro unless the target machine has limitations which
2389 require the macro to do something nontrivial.
2391 @findex SECONDARY_RELOAD_CLASS
2392 @findex SECONDARY_INPUT_RELOAD_CLASS
2393 @findex SECONDARY_OUTPUT_RELOAD_CLASS
2394 @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2395 @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2396 @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2397 Many machines have some registers that cannot be copied directly to or
2398 from memory or even from other types of registers. An example is the
2399 @samp{MQ} register, which on most machines, can only be copied to or
2400 from general registers, but not memory. Some machines allow copying all
2401 registers to and from memory, but require a scratch register for stores
2402 to some memory locations (e.g., those with symbolic address on the RT,
2403 and those with certain symbolic address on the Sparc when compiling
2404 PIC)@. In some cases, both an intermediate and a scratch register are
2407 You should define these macros to indicate to the reload phase that it may
2408 need to allocate at least one register for a reload in addition to the
2409 register to contain the data. Specifically, if copying @var{x} to a
2410 register @var{class} in @var{mode} requires an intermediate register,
2411 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2412 largest register class all of whose registers can be used as
2413 intermediate registers or scratch registers.
2415 If copying a register @var{class} in @var{mode} to @var{x} requires an
2416 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2417 should be defined to return the largest register class required. If the
2418 requirements for input and output reloads are the same, the macro
2419 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2422 The values returned by these macros are often @code{GENERAL_REGS}.
2423 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2424 can be directly copied to or from a register of @var{class} in
2425 @var{mode} without requiring a scratch register. Do not define this
2426 macro if it would always return @code{NO_REGS}.
2428 If a scratch register is required (either with or without an
2429 intermediate register), you should define patterns for
2430 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2431 (@pxref{Standard Names}. These patterns, which will normally be
2432 implemented with a @code{define_expand}, should be similar to the
2433 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2436 Define constraints for the reload register and scratch register that
2437 contain a single register class. If the original reload register (whose
2438 class is @var{class}) can meet the constraint given in the pattern, the
2439 value returned by these macros is used for the class of the scratch
2440 register. Otherwise, two additional reload registers are required.
2441 Their classes are obtained from the constraints in the insn pattern.
2443 @var{x} might be a pseudo-register or a @code{subreg} of a
2444 pseudo-register, which could either be in a hard register or in memory.
2445 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2446 in memory and the hard register number if it is in a register.
2448 These macros should not be used in the case where a particular class of
2449 registers can only be copied to memory and not to another class of
2450 registers. In that case, secondary reload registers are not needed and
2451 would not be helpful. Instead, a stack location must be used to perform
2452 the copy and the @code{mov@var{m}} pattern should use memory as an
2453 intermediate storage. This case often occurs between floating-point and
2456 @findex SECONDARY_MEMORY_NEEDED
2457 @item SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2458 Certain machines have the property that some registers cannot be copied
2459 to some other registers without using memory. Define this macro on
2460 those machines to be a C expression that is nonzero if objects of mode
2461 @var{m} in registers of @var{class1} can only be copied to registers of
2462 class @var{class2} by storing a register of @var{class1} into memory
2463 and loading that memory location into a register of @var{class2}.
2465 Do not define this macro if its value would always be zero.
2467 @findex SECONDARY_MEMORY_NEEDED_RTX
2468 @item SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2469 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2470 allocates a stack slot for a memory location needed for register copies.
2471 If this macro is defined, the compiler instead uses the memory location
2472 defined by this macro.
2474 Do not define this macro if you do not define
2475 @code{SECONDARY_MEMORY_NEEDED}.
2477 @findex SECONDARY_MEMORY_NEEDED_MODE
2478 @item SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2479 When the compiler needs a secondary memory location to copy between two
2480 registers of mode @var{mode}, it normally allocates sufficient memory to
2481 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2482 load operations in a mode that many bits wide and whose class is the
2483 same as that of @var{mode}.
2485 This is right thing to do on most machines because it ensures that all
2486 bits of the register are copied and prevents accesses to the registers
2487 in a narrower mode, which some machines prohibit for floating-point
2490 However, this default behavior is not correct on some machines, such as
2491 the DEC Alpha, that store short integers in floating-point registers
2492 differently than in integer registers. On those machines, the default
2493 widening will not work correctly and you must define this macro to
2494 suppress that widening in some cases. See the file @file{alpha.h} for
2497 Do not define this macro if you do not define
2498 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2499 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2501 @findex SMALL_REGISTER_CLASSES
2502 @item SMALL_REGISTER_CLASSES
2503 On some machines, it is risky to let hard registers live across arbitrary
2504 insns. Typically, these machines have instructions that require values
2505 to be in specific registers (like an accumulator), and reload will fail
2506 if the required hard register is used for another purpose across such an
2509 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2510 value on these machines. When this macro has a nonzero value, the
2511 compiler will try to minimize the lifetime of hard registers.
2513 It is always safe to define this macro with a nonzero value, but if you
2514 unnecessarily define it, you will reduce the amount of optimizations
2515 that can be performed in some cases. If you do not define this macro
2516 with a nonzero value when it is required, the compiler will run out of
2517 spill registers and print a fatal error message. For most machines, you
2518 should not define this macro at all.
2520 @findex CLASS_LIKELY_SPILLED_P
2521 @item CLASS_LIKELY_SPILLED_P (@var{class})
2522 A C expression whose value is nonzero if pseudos that have been assigned
2523 to registers of class @var{class} would likely be spilled because
2524 registers of @var{class} are needed for spill registers.
2526 The default value of this macro returns 1 if @var{class} has exactly one
2527 register and zero otherwise. On most machines, this default should be
2528 used. Only define this macro to some other expression if pseudos
2529 allocated by @file{local-alloc.c} end up in memory because their hard
2530 registers were needed for spill registers. If this macro returns nonzero
2531 for those classes, those pseudos will only be allocated by
2532 @file{global.c}, which knows how to reallocate the pseudo to another
2533 register. If there would not be another register available for
2534 reallocation, you should not change the definition of this macro since
2535 the only effect of such a definition would be to slow down register
2538 @findex CLASS_MAX_NREGS
2539 @item CLASS_MAX_NREGS (@var{class}, @var{mode})
2540 A C expression for the maximum number of consecutive registers
2541 of class @var{class} needed to hold a value of mode @var{mode}.
2543 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2544 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2545 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2546 @var{mode})} for all @var{regno} values in the class @var{class}.
2548 This macro helps control the handling of multiple-word values
2551 @item CLASS_CANNOT_CHANGE_MODE
2552 If defined, a C expression for a class that contains registers for
2553 which the compiler may not change modes arbitrarily.
2555 @item CLASS_CANNOT_CHANGE_MODE_P(@var{from}, @var{to})
2556 A C expression that is true if, for a register in
2557 @code{CLASS_CANNOT_CHANGE_MODE}, the requested mode punning is invalid.
2559 For the example, loading 32-bit integer or floating-point objects into
2560 floating-point registers on the Alpha extends them to 64 bits.
2561 Therefore loading a 64-bit object and then storing it as a 32-bit object
2562 does not store the low-order 32 bits, as would be the case for a normal
2563 register. Therefore, @file{alpha.h} defines @code{CLASS_CANNOT_CHANGE_MODE}
2564 as @code{FLOAT_REGS} and @code{CLASS_CANNOT_CHANGE_MODE_P} restricts
2565 mode changes to same-size modes.
2567 Compare this to IA-64, which extends floating-point values to 82-bits,
2568 and stores 64-bit integers in a different format than 64-bit doubles.
2569 Therefore @code{CLASS_CANNOT_CHANGE_MODE_P} is always true.
2572 Three other special macros describe which operands fit which constraint
2576 @findex CONST_OK_FOR_LETTER_P
2577 @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2578 A C expression that defines the machine-dependent operand constraint
2579 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2580 particular ranges of integer values. If @var{c} is one of those
2581 letters, the expression should check that @var{value}, an integer, is in
2582 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2583 not one of those letters, the value should be 0 regardless of
2586 @findex CONST_DOUBLE_OK_FOR_LETTER_P
2587 @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2588 A C expression that defines the machine-dependent operand constraint
2589 letters that specify particular ranges of @code{const_double} values
2590 (@samp{G} or @samp{H}).
2592 If @var{c} is one of those letters, the expression should check that
2593 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2594 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2595 letters, the value should be 0 regardless of @var{value}.
2597 @code{const_double} is used for all floating-point constants and for
2598 @code{DImode} fixed-point constants. A given letter can accept either
2599 or both kinds of values. It can use @code{GET_MODE} to distinguish
2600 between these kinds.
2602 @findex EXTRA_CONSTRAINT
2603 @item EXTRA_CONSTRAINT (@var{value}, @var{c})
2604 A C expression that defines the optional machine-dependent constraint
2605 letters that can be used to segregate specific types of operands, usually
2606 memory references, for the target machine. Any letter that is not
2607 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER}
2608 may be used. Normally this macro will not be defined.
2610 If it is required for a particular target machine, it should return 1
2611 if @var{value} corresponds to the operand type represented by the
2612 constraint letter @var{c}. If @var{c} is not defined as an extra
2613 constraint, the value returned should be 0 regardless of @var{value}.
2615 For example, on the ROMP, load instructions cannot have their output
2616 in r0 if the memory reference contains a symbolic address. Constraint
2617 letter @samp{Q} is defined as representing a memory address that does
2618 @emph{not} contain a symbolic address. An alternative is specified with
2619 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2620 alternative specifies @samp{m} on the input and a register class that
2621 does not include r0 on the output.
2623 @findex EXTRA_MEMORY_CONSTRAINT
2624 @item EXTRA_MEMORY_CONSTRAINT (@var{c})
2625 A C expression that defines the optional machine-dependent constraint
2626 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2627 be treated like memory constraints by the reload pass.
2629 It should return 1 if the operand type represented by the constraint
2630 letter @var{c} comprises a subset of all memory references including
2631 all those whose address is simply a base register. This allows the reload
2632 pass to reload an operand, if it does not directly correspond to the operand
2633 type of @var{c}, by copying its address into a base register.
2635 For example, on the S/390, some instructions do not accept arbitrary
2636 memory references, but only those that do not make use of an index
2637 register. The constraint letter @samp{Q} is defined via
2638 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2639 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2640 a @samp{Q} constraint can handle any memory operand, because the
2641 reload pass knows it can be reloaded by copying the memory address
2642 into a base register if required. This is analogous to the way
2643 a @samp{o} constraint can handle any memory operand.
2645 @findex EXTRA_ADDRESS_CONSTRAINT
2646 @item EXTRA_ADDRESS_CONSTRAINT (@var{c})
2647 A C expression that defines the optional machine-dependent constraint
2648 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2649 be treated like address constraints by the reload pass.
2651 It should return 1 if the operand type represented by the constraint
2652 letter @var{c} comprises a subset of all memory addresses including
2653 all those that consist of just a base register. This allows the reload
2654 pass to reload an operand, if it does not directly correspond to the operand
2655 type of @var{c}, by copying it into a base register.
2657 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2658 be used with the @code{address_operand} predicate. It is treated
2659 analogously to the @samp{p} constraint.
2662 @node Stack and Calling
2663 @section Stack Layout and Calling Conventions
2664 @cindex calling conventions
2666 @c prevent bad page break with this line
2667 This describes the stack layout and calling conventions.
2671 * Exception Handling::
2676 * Register Arguments::
2678 * Aggregate Return::
2686 @subsection Basic Stack Layout
2687 @cindex stack frame layout
2688 @cindex frame layout
2690 @c prevent bad page break with this line
2691 Here is the basic stack layout.
2694 @findex STACK_GROWS_DOWNWARD
2695 @item STACK_GROWS_DOWNWARD
2696 Define this macro if pushing a word onto the stack moves the stack
2697 pointer to a smaller address.
2699 When we say, ``define this macro if @dots{},'' it means that the
2700 compiler checks this macro only with @code{#ifdef} so the precise
2701 definition used does not matter.
2703 @findex STACK_PUSH_CODE
2704 @item STACK_PUSH_CODE
2706 This macro defines the operation used when something is pushed
2707 on the stack. In RTL, a push operation will be
2708 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2710 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2711 and @code{POST_INC}. Which of these is correct depends on
2712 the stack direction and on whether the stack pointer points
2713 to the last item on the stack or whether it points to the
2714 space for the next item on the stack.
2716 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2717 defined, which is almost always right, and @code{PRE_INC} otherwise,
2718 which is often wrong.
2720 @findex FRAME_GROWS_DOWNWARD
2721 @item FRAME_GROWS_DOWNWARD
2722 Define this macro if the addresses of local variable slots are at negative
2723 offsets from the frame pointer.
2725 @findex ARGS_GROW_DOWNWARD
2726 @item ARGS_GROW_DOWNWARD
2727 Define this macro if successive arguments to a function occupy decreasing
2728 addresses on the stack.
2730 @findex STARTING_FRAME_OFFSET
2731 @item STARTING_FRAME_OFFSET
2732 Offset from the frame pointer to the first local variable slot to be allocated.
2734 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2735 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2736 Otherwise, it is found by adding the length of the first slot to the
2737 value @code{STARTING_FRAME_OFFSET}.
2738 @c i'm not sure if the above is still correct.. had to change it to get
2739 @c rid of an overfull. --mew 2feb93
2741 @findex STACK_POINTER_OFFSET
2742 @item STACK_POINTER_OFFSET
2743 Offset from the stack pointer register to the first location at which
2744 outgoing arguments are placed. If not specified, the default value of
2745 zero is used. This is the proper value for most machines.
2747 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2748 the first location at which outgoing arguments are placed.
2750 @findex FIRST_PARM_OFFSET
2751 @item FIRST_PARM_OFFSET (@var{fundecl})
2752 Offset from the argument pointer register to the first argument's
2753 address. On some machines it may depend on the data type of the
2756 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2757 the first argument's address.
2759 @findex STACK_DYNAMIC_OFFSET
2760 @item STACK_DYNAMIC_OFFSET (@var{fundecl})
2761 Offset from the stack pointer register to an item dynamically allocated
2762 on the stack, e.g., by @code{alloca}.
2764 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2765 length of the outgoing arguments. The default is correct for most
2766 machines. See @file{function.c} for details.
2768 @findex DYNAMIC_CHAIN_ADDRESS
2769 @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2770 A C expression whose value is RTL representing the address in a stack
2771 frame where the pointer to the caller's frame is stored. Assume that
2772 @var{frameaddr} is an RTL expression for the address of the stack frame
2775 If you don't define this macro, the default is to return the value
2776 of @var{frameaddr}---that is, the stack frame address is also the
2777 address of the stack word that points to the previous frame.
2779 @findex SETUP_FRAME_ADDRESSES
2780 @item SETUP_FRAME_ADDRESSES
2781 If defined, a C expression that produces the machine-specific code to
2782 setup the stack so that arbitrary frames can be accessed. For example,
2783 on the Sparc, we must flush all of the register windows to the stack
2784 before we can access arbitrary stack frames. You will seldom need to
2787 @findex BUILTIN_SETJMP_FRAME_VALUE
2788 @item BUILTIN_SETJMP_FRAME_VALUE
2789 If defined, a C expression that contains an rtx that is used to store
2790 the address of the current frame into the built in @code{setjmp} buffer.
2791 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2792 machines. One reason you may need to define this macro is if
2793 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2795 @findex RETURN_ADDR_RTX
2796 @item RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2797 A C expression whose value is RTL representing the value of the return
2798 address for the frame @var{count} steps up from the current frame, after
2799 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2800 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2801 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2803 The value of the expression must always be the correct address when
2804 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2805 determine the return address of other frames.
2807 @findex RETURN_ADDR_IN_PREVIOUS_FRAME
2808 @item RETURN_ADDR_IN_PREVIOUS_FRAME
2809 Define this if the return address of a particular stack frame is accessed
2810 from the frame pointer of the previous stack frame.
2812 @findex INCOMING_RETURN_ADDR_RTX
2813 @item INCOMING_RETURN_ADDR_RTX
2814 A C expression whose value is RTL representing the location of the
2815 incoming return address at the beginning of any function, before the
2816 prologue. This RTL is either a @code{REG}, indicating that the return
2817 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2820 You only need to define this macro if you want to support call frame
2821 debugging information like that provided by DWARF 2.
2823 If this RTL is a @code{REG}, you should also define
2824 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2826 @findex INCOMING_FRAME_SP_OFFSET
2827 @item INCOMING_FRAME_SP_OFFSET
2828 A C expression whose value is an integer giving the offset, in bytes,
2829 from the value of the stack pointer register to the top of the stack
2830 frame at the beginning of any function, before the prologue. The top of
2831 the frame is defined to be the value of the stack pointer in the
2832 previous frame, just before the call instruction.
2834 You only need to define this macro if you want to support call frame
2835 debugging information like that provided by DWARF 2.
2837 @findex ARG_POINTER_CFA_OFFSET
2838 @item ARG_POINTER_CFA_OFFSET (@var{fundecl})
2839 A C expression whose value is an integer giving the offset, in bytes,
2840 from the argument pointer to the canonical frame address (cfa). The
2841 final value should coincide with that calculated by
2842 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2843 during virtual register instantiation.
2845 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
2846 which is correct for most machines; in general, the arguments are found
2847 immediately before the stack frame. Note that this is not the case on
2848 some targets that save registers into the caller's frame, such as SPARC
2849 and rs6000, and so such targets need to define this macro.
2851 You only need to define this macro if the default is incorrect, and you
2852 want to support call frame debugging information like that provided by
2857 Define this macro if the stack size for the target is very small. This
2858 has the effect of disabling gcc's built-in @samp{alloca}, though
2859 @samp{__builtin_alloca} is not affected.
2862 @node Exception Handling
2863 @subsection Exception Handling Support
2864 @cindex exception handling
2867 @findex EH_RETURN_DATA_REGNO
2868 @item EH_RETURN_DATA_REGNO (@var{N})
2869 A C expression whose value is the @var{N}th register number used for
2870 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2871 @var{N} registers are usable.
2873 The exception handling library routines communicate with the exception
2874 handlers via a set of agreed upon registers. Ideally these registers
2875 should be call-clobbered; it is possible to use call-saved registers,
2876 but may negatively impact code size. The target must support at least
2877 2 data registers, but should define 4 if there are enough free registers.
2879 You must define this macro if you want to support call frame exception
2880 handling like that provided by DWARF 2.
2882 @findex EH_RETURN_STACKADJ_RTX
2883 @item EH_RETURN_STACKADJ_RTX
2884 A C expression whose value is RTL representing a location in which
2885 to store a stack adjustment to be applied before function return.
2886 This is used to unwind the stack to an exception handler's call frame.
2887 It will be assigned zero on code paths that return normally.
2889 Typically this is a call-clobbered hard register that is otherwise
2890 untouched by the epilogue, but could also be a stack slot.
2892 You must define this macro if you want to support call frame exception
2893 handling like that provided by DWARF 2.
2895 @findex EH_RETURN_HANDLER_RTX
2896 @item EH_RETURN_HANDLER_RTX
2897 A C expression whose value is RTL representing a location in which
2898 to store the address of an exception handler to which we should
2899 return. It will not be assigned on code paths that return normally.
2901 Typically this is the location in the call frame at which the normal
2902 return address is stored. For targets that return by popping an
2903 address off the stack, this might be a memory address just below
2904 the @emph{target} call frame rather than inside the current call
2905 frame. @code{EH_RETURN_STACKADJ_RTX} will have already been assigned,
2906 so it may be used to calculate the location of the target call frame.
2908 Some targets have more complex requirements than storing to an
2909 address calculable during initial code generation. In that case
2910 the @code{eh_return} instruction pattern should be used instead.
2912 If you want to support call frame exception handling, you must
2913 define either this macro or the @code{eh_return} instruction pattern.
2915 @findex ASM_PREFERRED_EH_DATA_FORMAT
2916 @item ASM_PREFERRED_EH_DATA_FORMAT(@var{code}, @var{global})
2917 This macro chooses the encoding of pointers embedded in the exception
2918 handling sections. If at all possible, this should be defined such
2919 that the exception handling section will not require dynamic relocations,
2920 and so may be read-only.
2922 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
2923 @var{global} is true if the symbol may be affected by dynamic relocations.
2924 The macro should return a combination of the @code{DW_EH_PE_*} defines
2925 as found in @file{dwarf2.h}.
2927 If this macro is not defined, pointers will not be encoded but
2928 represented directly.
2930 @findex ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX
2931 @item ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX(@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
2932 This macro allows the target to emit whatever special magic is required
2933 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
2934 Generic code takes care of pc-relative and indirect encodings; this must
2935 be defined if the target uses text-relative or data-relative encodings.
2937 This is a C statement that branches to @var{done} if the format was
2938 handled. @var{encoding} is the format chosen, @var{size} is the number
2939 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
2942 @findex MD_FALLBACK_FRAME_STATE_FOR
2943 @item MD_FALLBACK_FRAME_STATE_FOR(@var{context}, @var{fs}, @var{success})
2944 This macro allows the target to add cpu and operating system specific
2945 code to the call-frame unwinder for use when there is no unwind data
2946 available. The most common reason to implement this macro is to unwind
2947 through signal frames.
2949 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
2950 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
2951 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
2952 for the address of the code being executed and @code{context->cfa} for
2953 the stack pointer value. If the frame can be decoded, the register save
2954 addresses should be updated in @var{fs} and the macro should branch to
2955 @var{success}. If the frame cannot be decoded, the macro should do
2959 @node Stack Checking
2960 @subsection Specifying How Stack Checking is Done
2962 GCC will check that stack references are within the boundaries of
2963 the stack, if the @option{-fstack-check} is specified, in one of three ways:
2967 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
2968 will assume that you have arranged for stack checking to be done at
2969 appropriate places in the configuration files, e.g., in
2970 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
2974 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
2975 called @code{check_stack} in your @file{md} file, GCC will call that
2976 pattern with one argument which is the address to compare the stack
2977 value against. You must arrange for this pattern to report an error if
2978 the stack pointer is out of range.
2981 If neither of the above are true, GCC will generate code to periodically
2982 ``probe'' the stack pointer using the values of the macros defined below.
2985 Normally, you will use the default values of these macros, so GCC
2986 will use the third approach.
2989 @findex STACK_CHECK_BUILTIN
2990 @item STACK_CHECK_BUILTIN
2991 A nonzero value if stack checking is done by the configuration files in a
2992 machine-dependent manner. You should define this macro if stack checking
2993 is require by the ABI of your machine or if you would like to have to stack
2994 checking in some more efficient way than GCC's portable approach.
2995 The default value of this macro is zero.
2997 @findex STACK_CHECK_PROBE_INTERVAL
2998 @item STACK_CHECK_PROBE_INTERVAL
2999 An integer representing the interval at which GCC must generate stack
3000 probe instructions. You will normally define this macro to be no larger
3001 than the size of the ``guard pages'' at the end of a stack area. The
3002 default value of 4096 is suitable for most systems.
3004 @findex STACK_CHECK_PROBE_LOAD
3005 @item STACK_CHECK_PROBE_LOAD
3006 A integer which is nonzero if GCC should perform the stack probe
3007 as a load instruction and zero if GCC should use a store instruction.
3008 The default is zero, which is the most efficient choice on most systems.
3010 @findex STACK_CHECK_PROTECT
3011 @item STACK_CHECK_PROTECT
3012 The number of bytes of stack needed to recover from a stack overflow,
3013 for languages where such a recovery is supported. The default value of
3014 75 words should be adequate for most machines.
3016 @findex STACK_CHECK_MAX_FRAME_SIZE
3017 @item STACK_CHECK_MAX_FRAME_SIZE
3018 The maximum size of a stack frame, in bytes. GCC will generate probe
3019 instructions in non-leaf functions to ensure at least this many bytes of
3020 stack are available. If a stack frame is larger than this size, stack
3021 checking will not be reliable and GCC will issue a warning. The
3022 default is chosen so that GCC only generates one instruction on most
3023 systems. You should normally not change the default value of this macro.
3025 @findex STACK_CHECK_FIXED_FRAME_SIZE
3026 @item STACK_CHECK_FIXED_FRAME_SIZE
3027 GCC uses this value to generate the above warning message. It
3028 represents the amount of fixed frame used by a function, not including
3029 space for any callee-saved registers, temporaries and user variables.
3030 You need only specify an upper bound for this amount and will normally
3031 use the default of four words.
3033 @findex STACK_CHECK_MAX_VAR_SIZE
3034 @item STACK_CHECK_MAX_VAR_SIZE
3035 The maximum size, in bytes, of an object that GCC will place in the
3036 fixed area of the stack frame when the user specifies
3037 @option{-fstack-check}.
3038 GCC computed the default from the values of the above macros and you will
3039 normally not need to override that default.
3043 @node Frame Registers
3044 @subsection Registers That Address the Stack Frame
3046 @c prevent bad page break with this line
3047 This discusses registers that address the stack frame.
3050 @findex STACK_POINTER_REGNUM
3051 @item STACK_POINTER_REGNUM
3052 The register number of the stack pointer register, which must also be a
3053 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3054 the hardware determines which register this is.
3056 @findex FRAME_POINTER_REGNUM
3057 @item FRAME_POINTER_REGNUM
3058 The register number of the frame pointer register, which is used to
3059 access automatic variables in the stack frame. On some machines, the
3060 hardware determines which register this is. On other machines, you can
3061 choose any register you wish for this purpose.
3063 @findex HARD_FRAME_POINTER_REGNUM
3064 @item HARD_FRAME_POINTER_REGNUM
3065 On some machines the offset between the frame pointer and starting
3066 offset of the automatic variables is not known until after register
3067 allocation has been done (for example, because the saved registers are
3068 between these two locations). On those machines, define
3069 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3070 be used internally until the offset is known, and define
3071 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3072 used for the frame pointer.
3074 You should define this macro only in the very rare circumstances when it
3075 is not possible to calculate the offset between the frame pointer and
3076 the automatic variables until after register allocation has been
3077 completed. When this macro is defined, you must also indicate in your
3078 definition of @code{ELIMINABLE_REGS} how to eliminate
3079 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3080 or @code{STACK_POINTER_REGNUM}.
3082 Do not define this macro if it would be the same as
3083 @code{FRAME_POINTER_REGNUM}.
3085 @findex ARG_POINTER_REGNUM
3086 @item ARG_POINTER_REGNUM
3087 The register number of the arg pointer register, which is used to access
3088 the function's argument list. On some machines, this is the same as the
3089 frame pointer register. On some machines, the hardware determines which
3090 register this is. On other machines, you can choose any register you
3091 wish for this purpose. If this is not the same register as the frame
3092 pointer register, then you must mark it as a fixed register according to
3093 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3094 (@pxref{Elimination}).
3096 @findex RETURN_ADDRESS_POINTER_REGNUM
3097 @item RETURN_ADDRESS_POINTER_REGNUM
3098 The register number of the return address pointer register, which is used to
3099 access the current function's return address from the stack. On some
3100 machines, the return address is not at a fixed offset from the frame
3101 pointer or stack pointer or argument pointer. This register can be defined
3102 to point to the return address on the stack, and then be converted by
3103 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3105 Do not define this macro unless there is no other way to get the return
3106 address from the stack.
3108 @findex STATIC_CHAIN_REGNUM
3109 @findex STATIC_CHAIN_INCOMING_REGNUM
3110 @item STATIC_CHAIN_REGNUM
3111 @itemx STATIC_CHAIN_INCOMING_REGNUM
3112 Register numbers used for passing a function's static chain pointer. If
3113 register windows are used, the register number as seen by the called
3114 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3115 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3116 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3119 The static chain register need not be a fixed register.
3121 If the static chain is passed in memory, these macros should not be
3122 defined; instead, the next two macros should be defined.
3124 @findex STATIC_CHAIN
3125 @findex STATIC_CHAIN_INCOMING
3127 @itemx STATIC_CHAIN_INCOMING
3128 If the static chain is passed in memory, these macros provide rtx giving
3129 @code{mem} expressions that denote where they are stored.
3130 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3131 as seen by the calling and called functions, respectively. Often the former
3132 will be at an offset from the stack pointer and the latter at an offset from
3135 @findex stack_pointer_rtx
3136 @findex frame_pointer_rtx
3137 @findex arg_pointer_rtx
3138 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3139 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3140 macros and should be used to refer to those items.
3142 If the static chain is passed in a register, the two previous macros should
3145 @findex DWARF_FRAME_REGISTERS
3146 @item DWARF_FRAME_REGISTERS
3147 This macro specifies the maximum number of hard registers that can be
3148 saved in a call frame. This is used to size data structures used in
3149 DWARF2 exception handling.
3151 Prior to GCC 3.0, this macro was needed in order to establish a stable
3152 exception handling ABI in the face of adding new hard registers for ISA
3153 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3154 in the number of hard registers. Nevertheless, this macro can still be
3155 used to reduce the runtime memory requirements of the exception handling
3156 routines, which can be substantial if the ISA contains a lot of
3157 registers that are not call-saved.
3159 If this macro is not defined, it defaults to
3160 @code{FIRST_PSEUDO_REGISTER}.
3162 @findex PRE_GCC3_DWARF_FRAME_REGISTERS
3163 @item PRE_GCC3_DWARF_FRAME_REGISTERS
3165 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3166 for backward compatibility in pre GCC 3.0 compiled code.
3168 If this macro is not defined, it defaults to
3169 @code{DWARF_FRAME_REGISTERS}.
3174 @subsection Eliminating Frame Pointer and Arg Pointer
3176 @c prevent bad page break with this line
3177 This is about eliminating the frame pointer and arg pointer.
3180 @findex FRAME_POINTER_REQUIRED
3181 @item FRAME_POINTER_REQUIRED
3182 A C expression which is nonzero if a function must have and use a frame
3183 pointer. This expression is evaluated in the reload pass. If its value is
3184 nonzero the function will have a frame pointer.
3186 The expression can in principle examine the current function and decide
3187 according to the facts, but on most machines the constant 0 or the
3188 constant 1 suffices. Use 0 when the machine allows code to be generated
3189 with no frame pointer, and doing so saves some time or space. Use 1
3190 when there is no possible advantage to avoiding a frame pointer.
3192 In certain cases, the compiler does not know how to produce valid code
3193 without a frame pointer. The compiler recognizes those cases and
3194 automatically gives the function a frame pointer regardless of what
3195 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3198 In a function that does not require a frame pointer, the frame pointer
3199 register can be allocated for ordinary usage, unless you mark it as a
3200 fixed register. See @code{FIXED_REGISTERS} for more information.
3202 @findex INITIAL_FRAME_POINTER_OFFSET
3203 @findex get_frame_size
3204 @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3205 A C statement to store in the variable @var{depth-var} the difference
3206 between the frame pointer and the stack pointer values immediately after
3207 the function prologue. The value would be computed from information
3208 such as the result of @code{get_frame_size ()} and the tables of
3209 registers @code{regs_ever_live} and @code{call_used_regs}.
3211 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3212 need not be defined. Otherwise, it must be defined even if
3213 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3214 case, you may set @var{depth-var} to anything.
3216 @findex ELIMINABLE_REGS
3217 @item ELIMINABLE_REGS
3218 If defined, this macro specifies a table of register pairs used to
3219 eliminate unneeded registers that point into the stack frame. If it is not
3220 defined, the only elimination attempted by the compiler is to replace
3221 references to the frame pointer with references to the stack pointer.
3223 The definition of this macro is a list of structure initializations, each
3224 of which specifies an original and replacement register.
3226 On some machines, the position of the argument pointer is not known until
3227 the compilation is completed. In such a case, a separate hard register
3228 must be used for the argument pointer. This register can be eliminated by
3229 replacing it with either the frame pointer or the argument pointer,
3230 depending on whether or not the frame pointer has been eliminated.
3232 In this case, you might specify:
3234 #define ELIMINABLE_REGS \
3235 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3236 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3237 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3240 Note that the elimination of the argument pointer with the stack pointer is
3241 specified first since that is the preferred elimination.
3243 @findex CAN_ELIMINATE
3244 @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3245 A C expression that returns nonzero if the compiler is allowed to try
3246 to replace register number @var{from-reg} with register number
3247 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3248 is defined, and will usually be the constant 1, since most of the cases
3249 preventing register elimination are things that the compiler already
3252 @findex INITIAL_ELIMINATION_OFFSET
3253 @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3254 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3255 specifies the initial difference between the specified pair of
3256 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3260 @node Stack Arguments
3261 @subsection Passing Function Arguments on the Stack
3262 @cindex arguments on stack
3263 @cindex stack arguments
3265 The macros in this section control how arguments are passed
3266 on the stack. See the following section for other macros that
3267 control passing certain arguments in registers.
3270 @findex PROMOTE_PROTOTYPES
3271 @item PROMOTE_PROTOTYPES
3272 A C expression whose value is nonzero if an argument declared in
3273 a prototype as an integral type smaller than @code{int} should
3274 actually be passed as an @code{int}. In addition to avoiding
3275 errors in certain cases of mismatch, it also makes for better
3276 code on certain machines. If the macro is not defined in target
3277 header files, it defaults to 0.
3281 A C expression. If nonzero, push insns will be used to pass
3283 If the target machine does not have a push instruction, set it to zero.
3284 That directs GCC to use an alternate strategy: to
3285 allocate the entire argument block and then store the arguments into
3286 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3287 On some machines, the definition
3289 @findex PUSH_ROUNDING
3290 @item PUSH_ROUNDING (@var{npushed})
3291 A C expression that is the number of bytes actually pushed onto the
3292 stack when an instruction attempts to push @var{npushed} bytes.
3294 On some machines, the definition
3297 #define PUSH_ROUNDING(BYTES) (BYTES)
3301 will suffice. But on other machines, instructions that appear
3302 to push one byte actually push two bytes in an attempt to maintain
3303 alignment. Then the definition should be
3306 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3309 @findex ACCUMULATE_OUTGOING_ARGS
3310 @findex current_function_outgoing_args_size
3311 @item ACCUMULATE_OUTGOING_ARGS
3312 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3313 will be computed and placed into the variable
3314 @code{current_function_outgoing_args_size}. No space will be pushed
3315 onto the stack for each call; instead, the function prologue should
3316 increase the stack frame size by this amount.
3318 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3321 @findex REG_PARM_STACK_SPACE
3322 @item REG_PARM_STACK_SPACE (@var{fndecl})
3323 Define this macro if functions should assume that stack space has been
3324 allocated for arguments even when their values are passed in
3327 The value of this macro is the size, in bytes, of the area reserved for
3328 arguments passed in registers for the function represented by @var{fndecl},
3329 which can be zero if GCC is calling a library function.
3331 This space can be allocated by the caller, or be a part of the
3332 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3334 @c above is overfull. not sure what to do. --mew 5feb93 did
3335 @c something, not sure if it looks good. --mew 10feb93
3337 @findex MAYBE_REG_PARM_STACK_SPACE
3338 @findex FINAL_REG_PARM_STACK_SPACE
3339 @item MAYBE_REG_PARM_STACK_SPACE
3340 @itemx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
3341 Define these macros in addition to the one above if functions might
3342 allocate stack space for arguments even when their values are passed
3343 in registers. These should be used when the stack space allocated
3344 for arguments in registers is not a simple constant independent of the
3345 function declaration.
3347 The value of the first macro is the size, in bytes, of the area that
3348 we should initially assume would be reserved for arguments passed in registers.
3350 The value of the second macro is the actual size, in bytes, of the area
3351 that will be reserved for arguments passed in registers. This takes two
3352 arguments: an integer representing the number of bytes of fixed sized
3353 arguments on the stack, and a tree representing the number of bytes of
3354 variable sized arguments on the stack.
3356 When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
3357 called for libcall functions, the current function, or for a function
3358 being called when it is known that such stack space must be allocated.
3359 In each case this value can be easily computed.
3361 When deciding whether a called function needs such stack space, and how
3362 much space to reserve, GCC uses these two macros instead of
3363 @code{REG_PARM_STACK_SPACE}.
3365 @findex OUTGOING_REG_PARM_STACK_SPACE
3366 @item OUTGOING_REG_PARM_STACK_SPACE
3367 Define this if it is the responsibility of the caller to allocate the area
3368 reserved for arguments passed in registers.
3370 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3371 whether the space for these arguments counts in the value of
3372 @code{current_function_outgoing_args_size}.
3374 @findex STACK_PARMS_IN_REG_PARM_AREA
3375 @item STACK_PARMS_IN_REG_PARM_AREA
3376 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3377 stack parameters don't skip the area specified by it.
3378 @c i changed this, makes more sens and it should have taken care of the
3379 @c overfull.. not as specific, tho. --mew 5feb93
3381 Normally, when a parameter is not passed in registers, it is placed on the
3382 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3383 suppresses this behavior and causes the parameter to be passed on the
3384 stack in its natural location.
3386 @findex RETURN_POPS_ARGS
3387 @item RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3388 A C expression that should indicate the number of bytes of its own
3389 arguments that a function pops on returning, or 0 if the
3390 function pops no arguments and the caller must therefore pop them all
3391 after the function returns.
3393 @var{fundecl} is a C variable whose value is a tree node that describes
3394 the function in question. Normally it is a node of type
3395 @code{FUNCTION_DECL} that describes the declaration of the function.
3396 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3398 @var{funtype} is a C variable whose value is a tree node that
3399 describes the function in question. Normally it is a node of type
3400 @code{FUNCTION_TYPE} that describes the data type of the function.
3401 From this it is possible to obtain the data types of the value and
3402 arguments (if known).
3404 When a call to a library function is being considered, @var{fundecl}
3405 will contain an identifier node for the library function. Thus, if
3406 you need to distinguish among various library functions, you can do so
3407 by their names. Note that ``library function'' in this context means
3408 a function used to perform arithmetic, whose name is known specially
3409 in the compiler and was not mentioned in the C code being compiled.
3411 @var{stack-size} is the number of bytes of arguments passed on the
3412 stack. If a variable number of bytes is passed, it is zero, and
3413 argument popping will always be the responsibility of the calling function.
3415 On the VAX, all functions always pop their arguments, so the definition
3416 of this macro is @var{stack-size}. On the 68000, using the standard
3417 calling convention, no functions pop their arguments, so the value of
3418 the macro is always 0 in this case. But an alternative calling
3419 convention is available in which functions that take a fixed number of
3420 arguments pop them but other functions (such as @code{printf}) pop
3421 nothing (the caller pops all). When this convention is in use,
3422 @var{funtype} is examined to determine whether a function takes a fixed
3423 number of arguments.
3425 @findex CALL_POPS_ARGS
3426 @item CALL_POPS_ARGS (@var{cum})
3427 A C expression that should indicate the number of bytes a call sequence
3428 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3429 when compiling a function call.
3431 @var{cum} is the variable in which all arguments to the called function
3432 have been accumulated.
3434 On certain architectures, such as the SH5, a call trampoline is used
3435 that pops certain registers off the stack, depending on the arguments
3436 that have been passed to the function. Since this is a property of the
3437 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3442 @node Register Arguments
3443 @subsection Passing Arguments in Registers
3444 @cindex arguments in registers
3445 @cindex registers arguments
3447 This section describes the macros which let you control how various
3448 types of arguments are passed in registers or how they are arranged in
3452 @findex FUNCTION_ARG
3453 @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3454 A C expression that controls whether a function argument is passed
3455 in a register, and which register.
3457 The arguments are @var{cum}, which summarizes all the previous
3458 arguments; @var{mode}, the machine mode of the argument; @var{type},
3459 the data type of the argument as a tree node or 0 if that is not known
3460 (which happens for C support library functions); and @var{named},
3461 which is 1 for an ordinary argument and 0 for nameless arguments that
3462 correspond to @samp{@dots{}} in the called function's prototype.
3463 @var{type} can be an incomplete type if a syntax error has previously
3466 The value of the expression is usually either a @code{reg} RTX for the
3467 hard register in which to pass the argument, or zero to pass the
3468 argument on the stack.
3470 For machines like the VAX and 68000, where normally all arguments are
3471 pushed, zero suffices as a definition.
3473 The value of the expression can also be a @code{parallel} RTX@. This is
3474 used when an argument is passed in multiple locations. The mode of the
3475 of the @code{parallel} should be the mode of the entire argument. The
3476 @code{parallel} holds any number of @code{expr_list} pairs; each one
3477 describes where part of the argument is passed. In each
3478 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3479 register in which to pass this part of the argument, and the mode of the
3480 register RTX indicates how large this part of the argument is. The
3481 second operand of the @code{expr_list} is a @code{const_int} which gives
3482 the offset in bytes into the entire argument of where this part starts.
3483 As a special exception the first @code{expr_list} in the @code{parallel}
3484 RTX may have a first operand of zero. This indicates that the entire
3485 argument is also stored on the stack.
3487 The last time this macro is called, it is called with @code{MODE ==
3488 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3489 pattern as operands 2 and 3 respectively.
3491 @cindex @file{stdarg.h} and register arguments
3492 The usual way to make the ISO library @file{stdarg.h} work on a machine
3493 where some arguments are usually passed in registers, is to cause
3494 nameless arguments to be passed on the stack instead. This is done
3495 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3497 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3498 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3499 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
3500 in the definition of this macro to determine if this argument is of a
3501 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3502 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3503 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3504 defined, the argument will be computed in the stack and then loaded into
3507 @findex MUST_PASS_IN_STACK
3508 @item MUST_PASS_IN_STACK (@var{mode}, @var{type})
3509 Define as a C expression that evaluates to nonzero if we do not know how
3510 to pass TYPE solely in registers. The file @file{expr.h} defines a
3511 definition that is usually appropriate, refer to @file{expr.h} for additional
3514 @findex FUNCTION_INCOMING_ARG
3515 @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3516 Define this macro if the target machine has ``register windows'', so
3517 that the register in which a function sees an arguments is not
3518 necessarily the same as the one in which the caller passed the
3521 For such machines, @code{FUNCTION_ARG} computes the register in which
3522 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3523 be defined in a similar fashion to tell the function being called
3524 where the arguments will arrive.
3526 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3527 serves both purposes.
3529 @findex FUNCTION_ARG_PARTIAL_NREGS
3530 @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
3531 A C expression for the number of words, at the beginning of an
3532 argument, that must be put in registers. The value must be zero for
3533 arguments that are passed entirely in registers or that are entirely
3534 pushed on the stack.
3536 On some machines, certain arguments must be passed partially in
3537 registers and partially in memory. On these machines, typically the
3538 first @var{n} words of arguments are passed in registers, and the rest
3539 on the stack. If a multi-word argument (a @code{double} or a
3540 structure) crosses that boundary, its first few words must be passed
3541 in registers and the rest must be pushed. This macro tells the
3542 compiler when this occurs, and how many of the words should go in
3545 @code{FUNCTION_ARG} for these arguments should return the first
3546 register to be used by the caller for this argument; likewise
3547 @code{FUNCTION_INCOMING_ARG}, for the called function.
3549 @findex FUNCTION_ARG_PASS_BY_REFERENCE
3550 @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3551 A C expression that indicates when an argument must be passed by reference.
3552 If nonzero for an argument, a copy of that argument is made in memory and a
3553 pointer to the argument is passed instead of the argument itself.
3554 The pointer is passed in whatever way is appropriate for passing a pointer
3557 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
3558 definition of this macro might be
3560 #define FUNCTION_ARG_PASS_BY_REFERENCE\
3561 (CUM, MODE, TYPE, NAMED) \
3562 MUST_PASS_IN_STACK (MODE, TYPE)
3564 @c this is *still* too long. --mew 5feb93
3566 @findex FUNCTION_ARG_CALLEE_COPIES
3567 @item FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
3568 If defined, a C expression that indicates when it is the called function's
3569 responsibility to make a copy of arguments passed by invisible reference.
3570 Normally, the caller makes a copy and passes the address of the copy to the
3571 routine being called. When @code{FUNCTION_ARG_CALLEE_COPIES} is defined and is
3572 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
3573 ``live'' value. The called function must not modify this value. If it can be
3574 determined that the value won't be modified, it need not make a copy;
3575 otherwise a copy must be made.
3577 @findex FUNCTION_ARG_REG_LITTLE_ENDIAN
3578 @item FUNCTION_ARG_REG_LITTLE_ENDIAN
3579 If defined TRUE on a big-endian system then structure arguments passed
3580 (and returned) in registers are passed in a little-endian manner instead of
3581 the big-endian manner. On the HP-UX IA64 and PA64 platforms structures are
3582 aligned differently then integral values and setting this value to true will
3583 allow for the special handling of structure arguments and return values.
3585 @findex CUMULATIVE_ARGS
3586 @item CUMULATIVE_ARGS
3587 A C type for declaring a variable that is used as the first argument of
3588 @code{FUNCTION_ARG} and other related values. For some target machines,
3589 the type @code{int} suffices and can hold the number of bytes of
3592 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3593 arguments that have been passed on the stack. The compiler has other
3594 variables to keep track of that. For target machines on which all
3595 arguments are passed on the stack, there is no need to store anything in
3596 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3597 should not be empty, so use @code{int}.
3599 @findex INIT_CUMULATIVE_ARGS
3600 @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{indirect})
3601 A C statement (sans semicolon) for initializing the variable @var{cum}
3602 for the state at the beginning of the argument list. The variable has
3603 type @code{CUMULATIVE_ARGS}. The value of @var{fntype} is the tree node
3604 for the data type of the function which will receive the args, or 0
3605 if the args are to a compiler support library function. The value of
3606 @var{indirect} is nonzero when processing an indirect call, for example
3607 a call through a function pointer. The value of @var{indirect} is zero
3608 for a call to an explicitly named function, a library function call, or when
3609 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3612 When processing a call to a compiler support library function,
3613 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3614 contains the name of the function, as a string. @var{libname} is 0 when
3615 an ordinary C function call is being processed. Thus, each time this
3616 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3617 never both of them at once.
3619 @findex INIT_CUMULATIVE_LIBCALL_ARGS
3620 @item INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3621 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3622 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3623 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3624 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3625 0)} is used instead.
3627 @findex INIT_CUMULATIVE_INCOMING_ARGS
3628 @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3629 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3630 finding the arguments for the function being compiled. If this macro is
3631 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3633 The value passed for @var{libname} is always 0, since library routines
3634 with special calling conventions are never compiled with GCC@. The
3635 argument @var{libname} exists for symmetry with
3636 @code{INIT_CUMULATIVE_ARGS}.
3637 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3638 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3640 @findex FUNCTION_ARG_ADVANCE
3641 @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3642 A C statement (sans semicolon) to update the summarizer variable
3643 @var{cum} to advance past an argument in the argument list. The
3644 values @var{mode}, @var{type} and @var{named} describe that argument.
3645 Once this is done, the variable @var{cum} is suitable for analyzing
3646 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
3648 This macro need not do anything if the argument in question was passed
3649 on the stack. The compiler knows how to track the amount of stack space
3650 used for arguments without any special help.
3652 @findex FUNCTION_ARG_PADDING
3653 @item FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3654 If defined, a C expression which determines whether, and in which direction,
3655 to pad out an argument with extra space. The value should be of type
3656 @code{enum direction}: either @code{upward} to pad above the argument,
3657 @code{downward} to pad below, or @code{none} to inhibit padding.
3659 The @emph{amount} of padding is always just enough to reach the next
3660 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3663 This macro has a default definition which is right for most systems.
3664 For little-endian machines, the default is to pad upward. For
3665 big-endian machines, the default is to pad downward for an argument of
3666 constant size shorter than an @code{int}, and upward otherwise.
3668 @findex PAD_VARARGS_DOWN
3669 @item PAD_VARARGS_DOWN
3670 If defined, a C expression which determines whether the default
3671 implementation of va_arg will attempt to pad down before reading the
3672 next argument, if that argument is smaller than its aligned space as
3673 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3674 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3676 @findex FUNCTION_ARG_BOUNDARY
3677 @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3678 If defined, a C expression that gives the alignment boundary, in bits,
3679 of an argument with the specified mode and type. If it is not defined,
3680 @code{PARM_BOUNDARY} is used for all arguments.
3682 @findex FUNCTION_ARG_REGNO_P
3683 @item FUNCTION_ARG_REGNO_P (@var{regno})
3684 A C expression that is nonzero if @var{regno} is the number of a hard
3685 register in which function arguments are sometimes passed. This does
3686 @emph{not} include implicit arguments such as the static chain and
3687 the structure-value address. On many machines, no registers can be
3688 used for this purpose since all function arguments are pushed on the
3691 @findex LOAD_ARGS_REVERSED
3692 @item LOAD_ARGS_REVERSED
3693 If defined, the order in which arguments are loaded into their
3694 respective argument registers is reversed so that the last
3695 argument is loaded first. This macro only affects arguments
3696 passed in registers.
3701 @subsection How Scalar Function Values Are Returned
3702 @cindex return values in registers
3703 @cindex values, returned by functions
3704 @cindex scalars, returned as values
3706 This section discusses the macros that control returning scalars as
3707 values---values that can fit in registers.
3710 @findex FUNCTION_VALUE
3711 @item FUNCTION_VALUE (@var{valtype}, @var{func})
3712 A C expression to create an RTX representing the place where a
3713 function returns a value of data type @var{valtype}. @var{valtype} is
3714 a tree node representing a data type. Write @code{TYPE_MODE
3715 (@var{valtype})} to get the machine mode used to represent that type.
3716 On many machines, only the mode is relevant. (Actually, on most
3717 machines, scalar values are returned in the same place regardless of
3720 The value of the expression is usually a @code{reg} RTX for the hard
3721 register where the return value is stored. The value can also be a
3722 @code{parallel} RTX, if the return value is in multiple places. See
3723 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3725 If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
3726 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3729 If the precise function being called is known, @var{func} is a tree
3730 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3731 pointer. This makes it possible to use a different value-returning
3732 convention for specific functions when all their calls are
3735 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3736 types, because these are returned in another way. See
3737 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3739 @findex FUNCTION_OUTGOING_VALUE
3740 @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3741 Define this macro if the target machine has ``register windows''
3742 so that the register in which a function returns its value is not
3743 the same as the one in which the caller sees the value.
3745 For such machines, @code{FUNCTION_VALUE} computes the register in which
3746 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3747 defined in a similar fashion to tell the function where to put the
3750 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3751 @code{FUNCTION_VALUE} serves both purposes.
3753 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3754 aggregate data types, because these are returned in another way. See
3755 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3757 @findex LIBCALL_VALUE
3758 @item LIBCALL_VALUE (@var{mode})
3759 A C expression to create an RTX representing the place where a library
3760 function returns a value of mode @var{mode}. If the precise function
3761 being called is known, @var{func} is a tree node
3762 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3763 pointer. This makes it possible to use a different value-returning
3764 convention for specific functions when all their calls are
3767 Note that ``library function'' in this context means a compiler
3768 support routine, used to perform arithmetic, whose name is known
3769 specially by the compiler and was not mentioned in the C code being
3772 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3773 data types, because none of the library functions returns such types.
3775 @findex FUNCTION_VALUE_REGNO_P
3776 @item FUNCTION_VALUE_REGNO_P (@var{regno})
3777 A C expression that is nonzero if @var{regno} is the number of a hard
3778 register in which the values of called function may come back.
3780 A register whose use for returning values is limited to serving as the
3781 second of a pair (for a value of type @code{double}, say) need not be
3782 recognized by this macro. So for most machines, this definition
3786 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3789 If the machine has register windows, so that the caller and the called
3790 function use different registers for the return value, this macro
3791 should recognize only the caller's register numbers.
3793 @findex APPLY_RESULT_SIZE
3794 @item APPLY_RESULT_SIZE
3795 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3796 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3797 saving and restoring an arbitrary return value.
3800 @node Aggregate Return
3801 @subsection How Large Values Are Returned
3802 @cindex aggregates as return values
3803 @cindex large return values
3804 @cindex returning aggregate values
3805 @cindex structure value address
3807 When a function value's mode is @code{BLKmode} (and in some other
3808 cases), the value is not returned according to @code{FUNCTION_VALUE}
3809 (@pxref{Scalar Return}). Instead, the caller passes the address of a
3810 block of memory in which the value should be stored. This address
3811 is called the @dfn{structure value address}.
3813 This section describes how to control returning structure values in
3817 @findex RETURN_IN_MEMORY
3818 @item RETURN_IN_MEMORY (@var{type})
3819 A C expression which can inhibit the returning of certain function
3820 values in registers, based on the type of value. A nonzero value says
3821 to return the function value in memory, just as large structures are
3822 always returned. Here @var{type} will be a C expression of type
3823 @code{tree}, representing the data type of the value.
3825 Note that values of mode @code{BLKmode} must be explicitly handled
3826 by this macro. Also, the option @option{-fpcc-struct-return}
3827 takes effect regardless of this macro. On most systems, it is
3828 possible to leave the macro undefined; this causes a default
3829 definition to be used, whose value is the constant 1 for @code{BLKmode}
3830 values, and 0 otherwise.
3832 Do not use this macro to indicate that structures and unions should always
3833 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
3836 @findex DEFAULT_PCC_STRUCT_RETURN
3837 @item DEFAULT_PCC_STRUCT_RETURN
3838 Define this macro to be 1 if all structure and union return values must be
3839 in memory. Since this results in slower code, this should be defined
3840 only if needed for compatibility with other compilers or with an ABI@.
3841 If you define this macro to be 0, then the conventions used for structure
3842 and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
3844 If not defined, this defaults to the value 1.
3846 @findex STRUCT_VALUE_REGNUM
3847 @item STRUCT_VALUE_REGNUM
3848 If the structure value address is passed in a register, then
3849 @code{STRUCT_VALUE_REGNUM} should be the number of that register.
3851 @findex STRUCT_VALUE
3853 If the structure value address is not passed in a register, define
3854 @code{STRUCT_VALUE} as an expression returning an RTX for the place
3855 where the address is passed. If it returns 0, the address is passed as
3856 an ``invisible'' first argument.
3858 @findex STRUCT_VALUE_INCOMING_REGNUM
3859 @item STRUCT_VALUE_INCOMING_REGNUM
3860 On some architectures the place where the structure value address
3861 is found by the called function is not the same place that the
3862 caller put it. This can be due to register windows, or it could
3863 be because the function prologue moves it to a different place.
3865 If the incoming location of the structure value address is in a
3866 register, define this macro as the register number.
3868 @findex STRUCT_VALUE_INCOMING
3869 @item STRUCT_VALUE_INCOMING
3870 If the incoming location is not a register, then you should define
3871 @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the
3872 called function should find the value. If it should find the value on
3873 the stack, define this to create a @code{mem} which refers to the frame
3874 pointer. A definition of 0 means that the address is passed as an
3875 ``invisible'' first argument.
3877 @findex PCC_STATIC_STRUCT_RETURN
3878 @item PCC_STATIC_STRUCT_RETURN
3879 Define this macro if the usual system convention on the target machine
3880 for returning structures and unions is for the called function to return
3881 the address of a static variable containing the value.
3883 Do not define this if the usual system convention is for the caller to
3884 pass an address to the subroutine.
3886 This macro has effect in @option{-fpcc-struct-return} mode, but it does
3887 nothing when you use @option{-freg-struct-return} mode.
3891 @subsection Caller-Saves Register Allocation
3893 If you enable it, GCC can save registers around function calls. This
3894 makes it possible to use call-clobbered registers to hold variables that
3895 must live across calls.
3898 @findex DEFAULT_CALLER_SAVES
3899 @item DEFAULT_CALLER_SAVES
3900 Define this macro if function calls on the target machine do not preserve
3901 any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
3902 for all registers. When defined, this macro enables @option{-fcaller-saves}
3903 by default for all optimization levels. It has no effect for optimization
3904 levels 2 and higher, where @option{-fcaller-saves} is the default.
3906 @findex CALLER_SAVE_PROFITABLE
3907 @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
3908 A C expression to determine whether it is worthwhile to consider placing
3909 a pseudo-register in a call-clobbered hard register and saving and
3910 restoring it around each function call. The expression should be 1 when
3911 this is worth doing, and 0 otherwise.
3913 If you don't define this macro, a default is used which is good on most
3914 machines: @code{4 * @var{calls} < @var{refs}}.
3916 @findex HARD_REGNO_CALLER_SAVE_MODE
3917 @item HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3918 A C expression specifying which mode is required for saving @var{nregs}
3919 of a pseudo-register in call-clobbered hard register @var{regno}. If
3920 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3921 returned. For most machines this macro need not be defined since GCC
3922 will select the smallest suitable mode.
3925 @node Function Entry
3926 @subsection Function Entry and Exit
3927 @cindex function entry and exit
3931 This section describes the macros that output function entry
3932 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3934 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
3935 If defined, a function that outputs the assembler code for entry to a
3936 function. The prologue is responsible for setting up the stack frame,
3937 initializing the frame pointer register, saving registers that must be
3938 saved, and allocating @var{size} additional bytes of storage for the
3939 local variables. @var{size} is an integer. @var{file} is a stdio
3940 stream to which the assembler code should be output.
3942 The label for the beginning of the function need not be output by this
3943 macro. That has already been done when the macro is run.
3945 @findex regs_ever_live
3946 To determine which registers to save, the macro can refer to the array
3947 @code{regs_ever_live}: element @var{r} is nonzero if hard register
3948 @var{r} is used anywhere within the function. This implies the function
3949 prologue should save register @var{r}, provided it is not one of the
3950 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
3951 @code{regs_ever_live}.)
3953 On machines that have ``register windows'', the function entry code does
3954 not save on the stack the registers that are in the windows, even if
3955 they are supposed to be preserved by function calls; instead it takes
3956 appropriate steps to ``push'' the register stack, if any non-call-used
3957 registers are used in the function.
3959 @findex frame_pointer_needed
3960 On machines where functions may or may not have frame-pointers, the
3961 function entry code must vary accordingly; it must set up the frame
3962 pointer if one is wanted, and not otherwise. To determine whether a
3963 frame pointer is in wanted, the macro can refer to the variable
3964 @code{frame_pointer_needed}. The variable's value will be 1 at run
3965 time in a function that needs a frame pointer. @xref{Elimination}.
3967 The function entry code is responsible for allocating any stack space
3968 required for the function. This stack space consists of the regions
3969 listed below. In most cases, these regions are allocated in the
3970 order listed, with the last listed region closest to the top of the
3971 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
3972 the highest address if it is not defined). You can use a different order
3973 for a machine if doing so is more convenient or required for
3974 compatibility reasons. Except in cases where required by standard
3975 or by a debugger, there is no reason why the stack layout used by GCC
3976 need agree with that used by other compilers for a machine.
3979 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
3980 If defined, a function that outputs assembler code at the end of a
3981 prologue. This should be used when the function prologue is being
3982 emitted as RTL, and you have some extra assembler that needs to be
3983 emitted. @xref{prologue instruction pattern}.
3986 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
3987 If defined, a function that outputs assembler code at the start of an
3988 epilogue. This should be used when the function epilogue is being
3989 emitted as RTL, and you have some extra assembler that needs to be
3990 emitted. @xref{epilogue instruction pattern}.
3993 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
3994 If defined, a function that outputs the assembler code for exit from a
3995 function. The epilogue is responsible for restoring the saved
3996 registers and stack pointer to their values when the function was
3997 called, and returning control to the caller. This macro takes the
3998 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
3999 registers to restore are determined from @code{regs_ever_live} and
4000 @code{CALL_USED_REGISTERS} in the same way.
4002 On some machines, there is a single instruction that does all the work
4003 of returning from the function. On these machines, give that
4004 instruction the name @samp{return} and do not define the macro
4005 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4007 Do not define a pattern named @samp{return} if you want the
4008 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4009 switches to control whether return instructions or epilogues are used,
4010 define a @samp{return} pattern with a validity condition that tests the
4011 target switches appropriately. If the @samp{return} pattern's validity
4012 condition is false, epilogues will be used.
4014 On machines where functions may or may not have frame-pointers, the
4015 function exit code must vary accordingly. Sometimes the code for these
4016 two cases is completely different. To determine whether a frame pointer
4017 is wanted, the macro can refer to the variable
4018 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4019 a function that needs a frame pointer.
4021 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4022 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4023 The C variable @code{current_function_is_leaf} is nonzero for such a
4024 function. @xref{Leaf Functions}.
4026 On some machines, some functions pop their arguments on exit while
4027 others leave that for the caller to do. For example, the 68020 when
4028 given @option{-mrtd} pops arguments in functions that take a fixed
4029 number of arguments.
4031 @findex current_function_pops_args
4032 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4033 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4034 needs to know what was decided. The variable that is called
4035 @code{current_function_pops_args} is the number of bytes of its
4036 arguments that a function should pop. @xref{Scalar Return}.
4037 @c what is the "its arguments" in the above sentence referring to, pray
4038 @c tell? --mew 5feb93
4045 @findex current_function_pretend_args_size
4046 A region of @code{current_function_pretend_args_size} bytes of
4047 uninitialized space just underneath the first argument arriving on the
4048 stack. (This may not be at the very start of the allocated stack region
4049 if the calling sequence has pushed anything else since pushing the stack
4050 arguments. But usually, on such machines, nothing else has been pushed
4051 yet, because the function prologue itself does all the pushing.) This
4052 region is used on machines where an argument may be passed partly in
4053 registers and partly in memory, and, in some cases to support the
4054 features in @code{<stdarg.h>}.
4057 An area of memory used to save certain registers used by the function.
4058 The size of this area, which may also include space for such things as
4059 the return address and pointers to previous stack frames, is
4060 machine-specific and usually depends on which registers have been used
4061 in the function. Machines with register windows often do not require
4065 A region of at least @var{size} bytes, possibly rounded up to an allocation
4066 boundary, to contain the local variables of the function. On some machines,
4067 this region and the save area may occur in the opposite order, with the
4068 save area closer to the top of the stack.
4071 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4072 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4073 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4074 argument lists of the function. @xref{Stack Arguments}.
4077 Normally, it is necessary for the macros
4078 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4079 @code{TARGET_ASM_FUNCTION_EPILOGUE} to treat leaf functions specially.
4080 The C variable @code{current_function_is_leaf} is nonzero for such a
4083 @findex EXIT_IGNORE_STACK
4084 @item EXIT_IGNORE_STACK
4085 Define this macro as a C expression that is nonzero if the return
4086 instruction or the function epilogue ignores the value of the stack
4087 pointer; in other words, if it is safe to delete an instruction to
4088 adjust the stack pointer before a return from the function.
4090 Note that this macro's value is relevant only for functions for which
4091 frame pointers are maintained. It is never safe to delete a final
4092 stack adjustment in a function that has no frame pointer, and the
4093 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4095 @findex EPILOGUE_USES
4096 @item EPILOGUE_USES (@var{regno})
4097 Define this macro as a C expression that is nonzero for registers that are
4098 used by the epilogue or the @samp{return} pattern. The stack and frame
4099 pointer registers are already be assumed to be used as needed.
4102 @item EH_USES (@var{regno})
4103 Define this macro as a C expression that is nonzero for registers that are
4104 used by the exception handling mechanism, and so should be considered live
4105 on entry to an exception edge.
4107 @findex DELAY_SLOTS_FOR_EPILOGUE
4108 @item DELAY_SLOTS_FOR_EPILOGUE
4109 Define this macro if the function epilogue contains delay slots to which
4110 instructions from the rest of the function can be ``moved''. The
4111 definition should be a C expression whose value is an integer
4112 representing the number of delay slots there.
4114 @findex ELIGIBLE_FOR_EPILOGUE_DELAY
4115 @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4116 A C expression that returns 1 if @var{insn} can be placed in delay
4117 slot number @var{n} of the epilogue.
4119 The argument @var{n} is an integer which identifies the delay slot now
4120 being considered (since different slots may have different rules of
4121 eligibility). It is never negative and is always less than the number
4122 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4123 If you reject a particular insn for a given delay slot, in principle, it
4124 may be reconsidered for a subsequent delay slot. Also, other insns may
4125 (at least in principle) be considered for the so far unfilled delay
4128 @findex current_function_epilogue_delay_list
4129 @findex final_scan_insn
4130 The insns accepted to fill the epilogue delay slots are put in an RTL
4131 list made with @code{insn_list} objects, stored in the variable
4132 @code{current_function_epilogue_delay_list}. The insn for the first
4133 delay slot comes first in the list. Your definition of the macro
4134 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4135 outputting the insns in this list, usually by calling
4136 @code{final_scan_insn}.
4138 You need not define this macro if you did not define
4139 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4141 @findex ASM_OUTPUT_MI_THUNK
4142 @item ASM_OUTPUT_MI_THUNK (@var{file}, @var{thunk_fndecl}, @var{delta}, @var{function})
4143 A C compound statement that outputs the assembler code for a thunk
4144 function, used to implement C++ virtual function calls with multiple
4145 inheritance. The thunk acts as a wrapper around a virtual function,
4146 adjusting the implicit object parameter before handing control off to
4149 First, emit code to add the integer @var{delta} to the location that
4150 contains the incoming first argument. Assume that this argument
4151 contains a pointer, and is the one used to pass the @code{this} pointer
4152 in C++. This is the incoming argument @emph{before} the function prologue,
4153 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4154 all other incoming arguments.
4156 After the addition, emit code to jump to @var{function}, which is a
4157 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4158 not touch the return address. Hence returning from @var{FUNCTION} will
4159 return to whoever called the current @samp{thunk}.
4161 The effect must be as if @var{function} had been called directly with
4162 the adjusted first argument. This macro is responsible for emitting all
4163 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4164 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4166 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4167 have already been extracted from it.) It might possibly be useful on
4168 some targets, but probably not.
4170 If you do not define this macro, the target-independent code in the C++
4171 front end will generate a less efficient heavyweight thunk that calls
4172 @var{function} instead of jumping to it. The generic approach does
4173 not support varargs.
4177 @subsection Generating Code for Profiling
4178 @cindex profiling, code generation
4180 These macros will help you generate code for profiling.
4183 @findex FUNCTION_PROFILER
4184 @item FUNCTION_PROFILER (@var{file}, @var{labelno})
4185 A C statement or compound statement to output to @var{file} some
4186 assembler code to call the profiling subroutine @code{mcount}.
4189 The details of how @code{mcount} expects to be called are determined by
4190 your operating system environment, not by GCC@. To figure them out,
4191 compile a small program for profiling using the system's installed C
4192 compiler and look at the assembler code that results.
4194 Older implementations of @code{mcount} expect the address of a counter
4195 variable to be loaded into some register. The name of this variable is
4196 @samp{LP} followed by the number @var{labelno}, so you would generate
4197 the name using @samp{LP%d} in a @code{fprintf}.
4199 @findex PROFILE_HOOK
4201 A C statement or compound statement to output to @var{file} some assembly
4202 code to call the profiling subroutine @code{mcount} even the target does
4203 not support profiling.
4205 @findex NO_PROFILE_COUNTERS
4206 @item NO_PROFILE_COUNTERS
4207 Define this macro if the @code{mcount} subroutine on your system does
4208 not need a counter variable allocated for each function. This is true
4209 for almost all modern implementations. If you define this macro, you
4210 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
4212 @findex PROFILE_BEFORE_PROLOGUE
4213 @item PROFILE_BEFORE_PROLOGUE
4214 Define this macro if the code for function profiling should come before
4215 the function prologue. Normally, the profiling code comes after.
4219 @subsection Permitting tail calls
4223 @findex FUNCTION_OK_FOR_SIBCALL
4224 @item FUNCTION_OK_FOR_SIBCALL (@var{decl})
4225 A C expression that evaluates to true if it is ok to perform a sibling
4226 call to @var{decl} from the current function.
4228 It is not uncommon for limitations of calling conventions to prevent
4229 tail calls to functions outside the current unit of translation, or
4230 during PIC compilation. Use this macro to enforce these restrictions,
4231 as the @code{sibcall} md pattern can not fail, or fall over to a
4236 @section Implementing the Varargs Macros
4237 @cindex varargs implementation
4239 GCC comes with an implementation of @code{<varargs.h>} and
4240 @code{<stdarg.h>} that work without change on machines that pass arguments
4241 on the stack. Other machines require their own implementations of
4242 varargs, and the two machine independent header files must have
4243 conditionals to include it.
4245 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4246 the calling convention for @code{va_start}. The traditional
4247 implementation takes just one argument, which is the variable in which
4248 to store the argument pointer. The ISO implementation of
4249 @code{va_start} takes an additional second argument. The user is
4250 supposed to write the last named argument of the function here.
4252 However, @code{va_start} should not use this argument. The way to find
4253 the end of the named arguments is with the built-in functions described
4257 @findex __builtin_saveregs
4258 @item __builtin_saveregs ()
4259 Use this built-in function to save the argument registers in memory so
4260 that the varargs mechanism can access them. Both ISO and traditional
4261 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4262 you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
4264 On some machines, @code{__builtin_saveregs} is open-coded under the
4265 control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
4266 it calls a routine written in assembler language, found in
4269 Code generated for the call to @code{__builtin_saveregs} appears at the
4270 beginning of the function, as opposed to where the call to
4271 @code{__builtin_saveregs} is written, regardless of what the code is.
4272 This is because the registers must be saved before the function starts
4273 to use them for its own purposes.
4274 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4277 @findex __builtin_args_info
4278 @item __builtin_args_info (@var{category})
4279 Use this built-in function to find the first anonymous arguments in
4282 In general, a machine may have several categories of registers used for
4283 arguments, each for a particular category of data types. (For example,
4284 on some machines, floating-point registers are used for floating-point
4285 arguments while other arguments are passed in the general registers.)
4286 To make non-varargs functions use the proper calling convention, you
4287 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4288 registers in each category have been used so far
4290 @code{__builtin_args_info} accesses the same data structure of type
4291 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4292 with it, with @var{category} specifying which word to access. Thus, the
4293 value indicates the first unused register in a given category.
4295 Normally, you would use @code{__builtin_args_info} in the implementation
4296 of @code{va_start}, accessing each category just once and storing the
4297 value in the @code{va_list} object. This is because @code{va_list} will
4298 have to update the values, and there is no way to alter the
4299 values accessed by @code{__builtin_args_info}.
4301 @findex __builtin_next_arg
4302 @item __builtin_next_arg (@var{lastarg})
4303 This is the equivalent of @code{__builtin_args_info}, for stack
4304 arguments. It returns the address of the first anonymous stack
4305 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4306 returns the address of the location above the first anonymous stack
4307 argument. Use it in @code{va_start} to initialize the pointer for
4308 fetching arguments from the stack. Also use it in @code{va_start} to
4309 verify that the second parameter @var{lastarg} is the last named argument
4310 of the current function.
4312 @findex __builtin_classify_type
4313 @item __builtin_classify_type (@var{object})
4314 Since each machine has its own conventions for which data types are
4315 passed in which kind of register, your implementation of @code{va_arg}
4316 has to embody these conventions. The easiest way to categorize the
4317 specified data type is to use @code{__builtin_classify_type} together
4318 with @code{sizeof} and @code{__alignof__}.
4320 @code{__builtin_classify_type} ignores the value of @var{object},
4321 considering only its data type. It returns an integer describing what
4322 kind of type that is---integer, floating, pointer, structure, and so on.
4324 The file @file{typeclass.h} defines an enumeration that you can use to
4325 interpret the values of @code{__builtin_classify_type}.
4328 These machine description macros help implement varargs:
4331 @findex EXPAND_BUILTIN_SAVEREGS
4332 @item EXPAND_BUILTIN_SAVEREGS ()
4333 If defined, is a C expression that produces the machine-specific code
4334 for a call to @code{__builtin_saveregs}. This code will be moved to the
4335 very beginning of the function, before any parameter access are made.
4336 The return value of this function should be an RTX that contains the
4337 value to use as the return of @code{__builtin_saveregs}.
4339 @findex SETUP_INCOMING_VARARGS
4340 @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type}, @var{pretend_args_size}, @var{second_time})
4341 This macro offers an alternative to using @code{__builtin_saveregs} and
4342 defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the
4343 anonymous register arguments into the stack so that all the arguments
4344 appear to have been passed consecutively on the stack. Once this is
4345 done, you can use the standard implementation of varargs that works for
4346 machines that pass all their arguments on the stack.
4348 The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data
4349 structure, containing the values that are obtained after processing the
4350 named arguments. The arguments @var{mode} and @var{type} describe the
4351 last named argument---its machine mode and its data type as a tree node.
4353 The macro implementation should do two things: first, push onto the
4354 stack all the argument registers @emph{not} used for the named
4355 arguments, and second, store the size of the data thus pushed into the
4356 @code{int}-valued variable whose name is supplied as the argument
4357 @var{pretend_args_size}. The value that you store here will serve as
4358 additional offset for setting up the stack frame.
4360 Because you must generate code to push the anonymous arguments at
4361 compile time without knowing their data types,
4362 @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just
4363 a single category of argument register and use it uniformly for all data
4366 If the argument @var{second_time} is nonzero, it means that the
4367 arguments of the function are being analyzed for the second time. This
4368 happens for an inline function, which is not actually compiled until the
4369 end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should
4370 not generate any instructions in this case.
4372 @findex STRICT_ARGUMENT_NAMING
4373 @item STRICT_ARGUMENT_NAMING
4374 Define this macro to be a nonzero value if the location where a function
4375 argument is passed depends on whether or not it is a named argument.
4377 This macro controls how the @var{named} argument to @code{FUNCTION_ARG}
4378 is set for varargs and stdarg functions. If this macro returns a
4379 nonzero value, the @var{named} argument is always true for named
4380 arguments, and false for unnamed arguments. If it returns a value of
4381 zero, but @code{SETUP_INCOMING_VARARGS} is defined, then all arguments
4382 are treated as named. Otherwise, all named arguments except the last
4383 are treated as named.
4385 You need not define this macro if it always returns zero.
4387 @findex PRETEND_OUTGOING_VARARGS_NAMED
4388 @item PRETEND_OUTGOING_VARARGS_NAMED
4389 If you need to conditionally change ABIs so that one works with
4390 @code{SETUP_INCOMING_VARARGS}, but the other works like neither
4391 @code{SETUP_INCOMING_VARARGS} nor @code{STRICT_ARGUMENT_NAMING} was
4392 defined, then define this macro to return nonzero if
4393 @code{SETUP_INCOMING_VARARGS} is used, zero otherwise.
4394 Otherwise, you should not define this macro.
4398 @section Trampolines for Nested Functions
4399 @cindex trampolines for nested functions
4400 @cindex nested functions, trampolines for
4402 A @dfn{trampoline} is a small piece of code that is created at run time
4403 when the address of a nested function is taken. It normally resides on
4404 the stack, in the stack frame of the containing function. These macros
4405 tell GCC how to generate code to allocate and initialize a
4408 The instructions in the trampoline must do two things: load a constant
4409 address into the static chain register, and jump to the real address of
4410 the nested function. On CISC machines such as the m68k, this requires
4411 two instructions, a move immediate and a jump. Then the two addresses
4412 exist in the trampoline as word-long immediate operands. On RISC
4413 machines, it is often necessary to load each address into a register in
4414 two parts. Then pieces of each address form separate immediate
4417 The code generated to initialize the trampoline must store the variable
4418 parts---the static chain value and the function address---into the
4419 immediate operands of the instructions. On a CISC machine, this is
4420 simply a matter of copying each address to a memory reference at the
4421 proper offset from the start of the trampoline. On a RISC machine, it
4422 may be necessary to take out pieces of the address and store them
4426 @findex TRAMPOLINE_TEMPLATE
4427 @item TRAMPOLINE_TEMPLATE (@var{file})
4428 A C statement to output, on the stream @var{file}, assembler code for a
4429 block of data that contains the constant parts of a trampoline. This
4430 code should not include a label---the label is taken care of
4433 If you do not define this macro, it means no template is needed
4434 for the target. Do not define this macro on systems where the block move
4435 code to copy the trampoline into place would be larger than the code
4436 to generate it on the spot.
4438 @findex TRAMPOLINE_SECTION
4439 @item TRAMPOLINE_SECTION
4440 The name of a subroutine to switch to the section in which the
4441 trampoline template is to be placed (@pxref{Sections}). The default is
4442 a value of @samp{readonly_data_section}, which places the trampoline in
4443 the section containing read-only data.
4445 @findex TRAMPOLINE_SIZE
4446 @item TRAMPOLINE_SIZE
4447 A C expression for the size in bytes of the trampoline, as an integer.
4449 @findex TRAMPOLINE_ALIGNMENT
4450 @item TRAMPOLINE_ALIGNMENT
4451 Alignment required for trampolines, in bits.
4453 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4454 is used for aligning trampolines.
4456 @findex INITIALIZE_TRAMPOLINE
4457 @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4458 A C statement to initialize the variable parts of a trampoline.
4459 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4460 an RTX for the address of the nested function; @var{static_chain} is an
4461 RTX for the static chain value that should be passed to the function
4464 @findex TRAMPOLINE_ADJUST_ADDRESS
4465 @item TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4466 A C statement that should perform any machine-specific adjustment in
4467 the address of the trampoline. Its argument contains the address that
4468 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4469 used for a function call should be different from the address in which
4470 the template was stored, the different address should be assigned to
4471 @var{addr}. If this macro is not defined, @var{addr} will be used for
4474 @findex ALLOCATE_TRAMPOLINE
4475 @item ALLOCATE_TRAMPOLINE (@var{fp})
4476 A C expression to allocate run-time space for a trampoline. The
4477 expression value should be an RTX representing a memory reference to the
4478 space for the trampoline.
4480 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4481 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4482 If this macro is not defined, by default the trampoline is allocated as
4483 a stack slot. This default is right for most machines. The exceptions
4484 are machines where it is impossible to execute instructions in the stack
4485 area. On such machines, you may have to implement a separate stack,
4486 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4487 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4489 @var{fp} points to a data structure, a @code{struct function}, which
4490 describes the compilation status of the immediate containing function of
4491 the function which the trampoline is for. Normally (when
4492 @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the
4493 trampoline is in the stack frame of this containing function. Other
4494 allocation strategies probably must do something analogous with this
4498 Implementing trampolines is difficult on many machines because they have
4499 separate instruction and data caches. Writing into a stack location
4500 fails to clear the memory in the instruction cache, so when the program
4501 jumps to that location, it executes the old contents.
4503 Here are two possible solutions. One is to clear the relevant parts of
4504 the instruction cache whenever a trampoline is set up. The other is to
4505 make all trampolines identical, by having them jump to a standard
4506 subroutine. The former technique makes trampoline execution faster; the
4507 latter makes initialization faster.
4509 To clear the instruction cache when a trampoline is initialized, define
4510 the following macros which describe the shape of the cache.
4513 @findex INSN_CACHE_SIZE
4514 @item INSN_CACHE_SIZE
4515 The total size in bytes of the cache.
4517 @findex INSN_CACHE_LINE_WIDTH
4518 @item INSN_CACHE_LINE_WIDTH
4519 The length in bytes of each cache line. The cache is divided into cache
4520 lines which are disjoint slots, each holding a contiguous chunk of data
4521 fetched from memory. Each time data is brought into the cache, an
4522 entire line is read at once. The data loaded into a cache line is
4523 always aligned on a boundary equal to the line size.
4525 @findex INSN_CACHE_DEPTH
4526 @item INSN_CACHE_DEPTH
4527 The number of alternative cache lines that can hold any particular memory
4531 Alternatively, if the machine has system calls or instructions to clear
4532 the instruction cache directly, you can define the following macro.
4535 @findex CLEAR_INSN_CACHE
4536 @item CLEAR_INSN_CACHE (@var{beg}, @var{end})
4537 If defined, expands to a C expression clearing the @emph{instruction
4538 cache} in the specified interval. If it is not defined, and the macro
4539 @code{INSN_CACHE_SIZE} is defined, some generic code is generated to clear the
4540 cache. The definition of this macro would typically be a series of
4541 @code{asm} statements. Both @var{beg} and @var{end} are both pointer
4545 To use a standard subroutine, define the following macro. In addition,
4546 you must make sure that the instructions in a trampoline fill an entire
4547 cache line with identical instructions, or else ensure that the
4548 beginning of the trampoline code is always aligned at the same point in
4549 its cache line. Look in @file{m68k.h} as a guide.
4552 @findex TRANSFER_FROM_TRAMPOLINE
4553 @item TRANSFER_FROM_TRAMPOLINE
4554 Define this macro if trampolines need a special subroutine to do their
4555 work. The macro should expand to a series of @code{asm} statements
4556 which will be compiled with GCC@. They go in a library function named
4557 @code{__transfer_from_trampoline}.
4559 If you need to avoid executing the ordinary prologue code of a compiled
4560 C function when you jump to the subroutine, you can do so by placing a
4561 special label of your own in the assembler code. Use one @code{asm}
4562 statement to generate an assembler label, and another to make the label
4563 global. Then trampolines can use that label to jump directly to your
4564 special assembler code.
4568 @section Implicit Calls to Library Routines
4569 @cindex library subroutine names
4570 @cindex @file{libgcc.a}
4572 @c prevent bad page break with this line
4573 Here is an explanation of implicit calls to library routines.
4576 @findex MULSI3_LIBCALL
4577 @item MULSI3_LIBCALL
4578 A C string constant giving the name of the function to call for
4579 multiplication of one signed full-word by another. If you do not
4580 define this macro, the default name is used, which is @code{__mulsi3},
4581 a function defined in @file{libgcc.a}.
4583 @findex DIVSI3_LIBCALL
4584 @item DIVSI3_LIBCALL
4585 A C string constant giving the name of the function to call for
4586 division of one signed full-word by another. If you do not define
4587 this macro, the default name is used, which is @code{__divsi3}, a
4588 function defined in @file{libgcc.a}.
4590 @findex UDIVSI3_LIBCALL
4591 @item UDIVSI3_LIBCALL
4592 A C string constant giving the name of the function to call for
4593 division of one unsigned full-word by another. If you do not define
4594 this macro, the default name is used, which is @code{__udivsi3}, a
4595 function defined in @file{libgcc.a}.
4597 @findex MODSI3_LIBCALL
4598 @item MODSI3_LIBCALL
4599 A C string constant giving the name of the function to call for the
4600 remainder in division of one signed full-word by another. If you do
4601 not define this macro, the default name is used, which is
4602 @code{__modsi3}, a function defined in @file{libgcc.a}.
4604 @findex UMODSI3_LIBCALL
4605 @item UMODSI3_LIBCALL
4606 A C string constant giving the name of the function to call for the
4607 remainder in division of one unsigned full-word by another. If you do
4608 not define this macro, the default name is used, which is
4609 @code{__umodsi3}, a function defined in @file{libgcc.a}.
4611 @findex MULDI3_LIBCALL
4612 @item MULDI3_LIBCALL
4613 A C string constant giving the name of the function to call for
4614 multiplication of one signed double-word by another. If you do not
4615 define this macro, the default name is used, which is @code{__muldi3},
4616 a function defined in @file{libgcc.a}.
4618 @findex DIVDI3_LIBCALL
4619 @item DIVDI3_LIBCALL
4620 A C string constant giving the name of the function to call for
4621 division of one signed double-word by another. If you do not define
4622 this macro, the default name is used, which is @code{__divdi3}, a
4623 function defined in @file{libgcc.a}.
4625 @findex UDIVDI3_LIBCALL
4626 @item UDIVDI3_LIBCALL
4627 A C string constant giving the name of the function to call for
4628 division of one unsigned full-word by another. If you do not define
4629 this macro, the default name is used, which is @code{__udivdi3}, a
4630 function defined in @file{libgcc.a}.
4632 @findex MODDI3_LIBCALL
4633 @item MODDI3_LIBCALL
4634 A C string constant giving the name of the function to call for the
4635 remainder in division of one signed double-word by another. If you do
4636 not define this macro, the default name is used, which is
4637 @code{__moddi3}, a function defined in @file{libgcc.a}.
4639 @findex UMODDI3_LIBCALL
4640 @item UMODDI3_LIBCALL
4641 A C string constant giving the name of the function to call for the
4642 remainder in division of one unsigned full-word by another. If you do
4643 not define this macro, the default name is used, which is
4644 @code{__umoddi3}, a function defined in @file{libgcc.a}.
4646 @findex DECLARE_LIBRARY_RENAMES
4647 @item DECLARE_LIBRARY_RENAMES
4648 This macro, if defined, should expand to a piece of C code that will get
4649 expanded when compiling functions for libgcc.a. It can be used to
4650 provide alternate names for gcc's internal library functions if there
4651 are ABI-mandated names that the compiler should provide.
4653 @findex INIT_TARGET_OPTABS
4654 @item INIT_TARGET_OPTABS
4655 Define this macro as a C statement that declares additional library
4656 routines renames existing ones. @code{init_optabs} calls this macro after
4657 initializing all the normal library routines.
4659 @findex FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4660 @item FLOAT_LIB_COMPARE_RETURNS_BOOL
4661 Define this macro as a C statement that returns nonzero if a call to
4662 the floating point comparison library function will return a boolean
4663 value that indicates the result of the comparison. It should return
4664 zero if one of gcc's own libgcc functions is called.
4666 Most ports don't need to define this macro.
4669 @cindex @code{EDOM}, implicit usage
4671 The value of @code{EDOM} on the target machine, as a C integer constant
4672 expression. If you don't define this macro, GCC does not attempt to
4673 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4674 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4677 If you do not define @code{TARGET_EDOM}, then compiled code reports
4678 domain errors by calling the library function and letting it report the
4679 error. If mathematical functions on your system use @code{matherr} when
4680 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4681 that @code{matherr} is used normally.
4683 @findex GEN_ERRNO_RTX
4684 @cindex @code{errno}, implicit usage
4686 Define this macro as a C expression to create an rtl expression that
4687 refers to the global ``variable'' @code{errno}. (On certain systems,
4688 @code{errno} may not actually be a variable.) If you don't define this
4689 macro, a reasonable default is used.
4691 @findex TARGET_MEM_FUNCTIONS
4692 @cindex @code{bcopy}, implicit usage
4693 @cindex @code{memcpy}, implicit usage
4694 @cindex @code{memmove}, implicit usage
4695 @cindex @code{bzero}, implicit usage
4696 @cindex @code{memset}, implicit usage
4697 @item TARGET_MEM_FUNCTIONS
4698 Define this macro if GCC should generate calls to the ISO C
4699 (and System V) library functions @code{memcpy}, @code{memmove} and
4700 @code{memset} rather than the BSD functions @code{bcopy} and @code{bzero}.
4702 @findex LIBGCC_NEEDS_DOUBLE
4703 @item LIBGCC_NEEDS_DOUBLE
4704 Define this macro if @code{float} arguments cannot be passed to library
4705 routines (so they must be converted to @code{double}). This macro
4706 affects both how library calls are generated and how the library
4707 routines in @file{libgcc.a} accept their arguments. It is useful on
4708 machines where floating and fixed point arguments are passed
4709 differently, such as the i860.
4711 @findex NEXT_OBJC_RUNTIME
4712 @item NEXT_OBJC_RUNTIME
4713 Define this macro to generate code for Objective-C message sending using
4714 the calling convention of the NeXT system. This calling convention
4715 involves passing the object, the selector and the method arguments all
4716 at once to the method-lookup library function.
4718 The default calling convention passes just the object and the selector
4719 to the lookup function, which returns a pointer to the method.
4722 @node Addressing Modes
4723 @section Addressing Modes
4724 @cindex addressing modes
4726 @c prevent bad page break with this line
4727 This is about addressing modes.
4730 @findex HAVE_PRE_INCREMENT
4731 @findex HAVE_PRE_DECREMENT
4732 @findex HAVE_POST_INCREMENT
4733 @findex HAVE_POST_DECREMENT
4734 @item HAVE_PRE_INCREMENT
4735 @itemx HAVE_PRE_DECREMENT
4736 @itemx HAVE_POST_INCREMENT
4737 @itemx HAVE_POST_DECREMENT
4738 A C expression that is nonzero if the machine supports pre-increment,
4739 pre-decrement, post-increment, or post-decrement addressing respectively.
4741 @findex HAVE_POST_MODIFY_DISP
4742 @findex HAVE_PRE_MODIFY_DISP
4743 @item HAVE_PRE_MODIFY_DISP
4744 @itemx HAVE_POST_MODIFY_DISP
4745 A C expression that is nonzero if the machine supports pre- or
4746 post-address side-effect generation involving constants other than
4747 the size of the memory operand.
4749 @findex HAVE_POST_MODIFY_REG
4750 @findex HAVE_PRE_MODIFY_REG
4751 @item HAVE_PRE_MODIFY_REG
4752 @itemx HAVE_POST_MODIFY_REG
4753 A C expression that is nonzero if the machine supports pre- or
4754 post-address side-effect generation involving a register displacement.
4756 @findex CONSTANT_ADDRESS_P
4757 @item CONSTANT_ADDRESS_P (@var{x})
4758 A C expression that is 1 if the RTX @var{x} is a constant which
4759 is a valid address. On most machines, this can be defined as
4760 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4761 in which constant addresses are supported.
4764 @code{CONSTANT_P} accepts integer-values expressions whose values are
4765 not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and
4766 @code{high} expressions and @code{const} arithmetic expressions, in
4767 addition to @code{const_int} and @code{const_double} expressions.
4769 @findex MAX_REGS_PER_ADDRESS
4770 @item MAX_REGS_PER_ADDRESS
4771 A number, the maximum number of registers that can appear in a valid
4772 memory address. Note that it is up to you to specify a value equal to
4773 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4776 @findex GO_IF_LEGITIMATE_ADDRESS
4777 @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4778 A C compound statement with a conditional @code{goto @var{label};}
4779 executed if @var{x} (an RTX) is a legitimate memory address on the
4780 target machine for a memory operand of mode @var{mode}.
4782 It usually pays to define several simpler macros to serve as
4783 subroutines for this one. Otherwise it may be too complicated to
4786 This macro must exist in two variants: a strict variant and a
4787 non-strict one. The strict variant is used in the reload pass. It
4788 must be defined so that any pseudo-register that has not been
4789 allocated a hard register is considered a memory reference. In
4790 contexts where some kind of register is required, a pseudo-register
4791 with no hard register must be rejected.
4793 The non-strict variant is used in other passes. It must be defined to
4794 accept all pseudo-registers in every context where some kind of
4795 register is required.
4797 @findex REG_OK_STRICT
4798 Compiler source files that want to use the strict variant of this
4799 macro define the macro @code{REG_OK_STRICT}. You should use an
4800 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4801 in that case and the non-strict variant otherwise.
4803 Subroutines to check for acceptable registers for various purposes (one
4804 for base registers, one for index registers, and so on) are typically
4805 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4806 Then only these subroutine macros need have two variants; the higher
4807 levels of macros may be the same whether strict or not.
4809 Normally, constant addresses which are the sum of a @code{symbol_ref}
4810 and an integer are stored inside a @code{const} RTX to mark them as
4811 constant. Therefore, there is no need to recognize such sums
4812 specifically as legitimate addresses. Normally you would simply
4813 recognize any @code{const} as legitimate.
4815 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4816 sums that are not marked with @code{const}. It assumes that a naked
4817 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4818 naked constant sums as illegitimate addresses, so that none of them will
4819 be given to @code{PRINT_OPERAND_ADDRESS}.
4821 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
4822 On some machines, whether a symbolic address is legitimate depends on
4823 the section that the address refers to. On these machines, define the
4824 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
4825 into the @code{symbol_ref}, and then check for it here. When you see a
4826 @code{const}, you will have to look inside it to find the
4827 @code{symbol_ref} in order to determine the section. @xref{Assembler
4830 @findex saveable_obstack
4831 The best way to modify the name string is by adding text to the
4832 beginning, with suitable punctuation to prevent any ambiguity. Allocate
4833 the new name in @code{saveable_obstack}. You will have to modify
4834 @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and
4835 output the name accordingly, and define @code{TARGET_STRIP_NAME_ENCODING}
4836 to access the original name string.
4838 You can check the information stored here into the @code{symbol_ref} in
4839 the definitions of the macros @code{GO_IF_LEGITIMATE_ADDRESS} and
4840 @code{PRINT_OPERAND_ADDRESS}.
4842 @findex REG_OK_FOR_BASE_P
4843 @item REG_OK_FOR_BASE_P (@var{x})
4844 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4845 RTX) is valid for use as a base register. For hard registers, it
4846 should always accept those which the hardware permits and reject the
4847 others. Whether the macro accepts or rejects pseudo registers must be
4848 controlled by @code{REG_OK_STRICT} as described above. This usually
4849 requires two variant definitions, of which @code{REG_OK_STRICT}
4850 controls the one actually used.
4852 @findex REG_MODE_OK_FOR_BASE_P
4853 @item REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4854 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4855 that expression may examine the mode of the memory reference in
4856 @var{mode}. You should define this macro if the mode of the memory
4857 reference affects whether a register may be used as a base register. If
4858 you define this macro, the compiler will use it instead of
4859 @code{REG_OK_FOR_BASE_P}.
4861 @findex REG_OK_FOR_INDEX_P
4862 @item REG_OK_FOR_INDEX_P (@var{x})
4863 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4864 RTX) is valid for use as an index register.
4866 The difference between an index register and a base register is that
4867 the index register may be scaled. If an address involves the sum of
4868 two registers, neither one of them scaled, then either one may be
4869 labeled the ``base'' and the other the ``index''; but whichever
4870 labeling is used must fit the machine's constraints of which registers
4871 may serve in each capacity. The compiler will try both labelings,
4872 looking for one that is valid, and will reload one or both registers
4873 only if neither labeling works.
4875 @findex FIND_BASE_TERM
4876 @item FIND_BASE_TERM (@var{x})
4877 A C expression to determine the base term of address @var{x}.
4878 This macro is used in only one place: `find_base_term' in alias.c.
4880 It is always safe for this macro to not be defined. It exists so
4881 that alias analysis can understand machine-dependent addresses.
4883 The typical use of this macro is to handle addresses containing
4884 a label_ref or symbol_ref within an UNSPEC@.
4886 @findex LEGITIMIZE_ADDRESS
4887 @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4888 A C compound statement that attempts to replace @var{x} with a valid
4889 memory address for an operand of mode @var{mode}. @var{win} will be a
4890 C statement label elsewhere in the code; the macro definition may use
4893 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4897 to avoid further processing if the address has become legitimate.
4899 @findex break_out_memory_refs
4900 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4901 and @var{oldx} will be the operand that was given to that function to produce
4904 The code generated by this macro should not alter the substructure of
4905 @var{x}. If it transforms @var{x} into a more legitimate form, it
4906 should assign @var{x} (which will always be a C variable) a new value.
4908 It is not necessary for this macro to come up with a legitimate
4909 address. The compiler has standard ways of doing so in all cases. In
4910 fact, it is safe for this macro to do nothing. But often a
4911 machine-dependent strategy can generate better code.
4913 @findex LEGITIMIZE_RELOAD_ADDRESS
4914 @item LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4915 A C compound statement that attempts to replace @var{x}, which is an address
4916 that needs reloading, with a valid memory address for an operand of mode
4917 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4918 It is not necessary to define this macro, but it might be useful for
4919 performance reasons.
4921 For example, on the i386, it is sometimes possible to use a single
4922 reload register instead of two by reloading a sum of two pseudo
4923 registers into a register. On the other hand, for number of RISC
4924 processors offsets are limited so that often an intermediate address
4925 needs to be generated in order to address a stack slot. By defining
4926 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
4927 generated for adjacent some stack slots can be made identical, and thus
4930 @emph{Note}: This macro should be used with caution. It is necessary
4931 to know something of how reload works in order to effectively use this,
4932 and it is quite easy to produce macros that build in too much knowledge
4933 of reload internals.
4935 @emph{Note}: This macro must be able to reload an address created by a
4936 previous invocation of this macro. If it fails to handle such addresses
4937 then the compiler may generate incorrect code or abort.
4940 The macro definition should use @code{push_reload} to indicate parts that
4941 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4942 suitable to be passed unaltered to @code{push_reload}.
4944 The code generated by this macro must not alter the substructure of
4945 @var{x}. If it transforms @var{x} into a more legitimate form, it
4946 should assign @var{x} (which will always be a C variable) a new value.
4947 This also applies to parts that you change indirectly by calling
4950 @findex strict_memory_address_p
4951 The macro definition may use @code{strict_memory_address_p} to test if
4952 the address has become legitimate.
4955 If you want to change only a part of @var{x}, one standard way of doing
4956 this is to use @code{copy_rtx}. Note, however, that is unshares only a
4957 single level of rtl. Thus, if the part to be changed is not at the
4958 top level, you'll need to replace first the top level.
4959 It is not necessary for this macro to come up with a legitimate
4960 address; but often a machine-dependent strategy can generate better code.
4962 @findex GO_IF_MODE_DEPENDENT_ADDRESS
4963 @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
4964 A C statement or compound statement with a conditional @code{goto
4965 @var{label};} executed if memory address @var{x} (an RTX) can have
4966 different meanings depending on the machine mode of the memory
4967 reference it is used for or if the address is valid for some modes
4970 Autoincrement and autodecrement addresses typically have mode-dependent
4971 effects because the amount of the increment or decrement is the size
4972 of the operand being addressed. Some machines have other mode-dependent
4973 addresses. Many RISC machines have no mode-dependent addresses.
4975 You may assume that @var{addr} is a valid address for the machine.
4977 @findex LEGITIMATE_CONSTANT_P
4978 @item LEGITIMATE_CONSTANT_P (@var{x})
4979 A C expression that is nonzero if @var{x} is a legitimate constant for
4980 an immediate operand on the target machine. You can assume that
4981 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
4982 @samp{1} is a suitable definition for this macro on machines where
4983 anything @code{CONSTANT_P} is valid.
4986 @node Condition Code
4987 @section Condition Code Status
4988 @cindex condition code status
4990 @c prevent bad page break with this line
4991 This describes the condition code status.
4994 The file @file{conditions.h} defines a variable @code{cc_status} to
4995 describe how the condition code was computed (in case the interpretation of
4996 the condition code depends on the instruction that it was set by). This
4997 variable contains the RTL expressions on which the condition code is
4998 currently based, and several standard flags.
5000 Sometimes additional machine-specific flags must be defined in the machine
5001 description header file. It can also add additional machine-specific
5002 information by defining @code{CC_STATUS_MDEP}.
5005 @findex CC_STATUS_MDEP
5006 @item CC_STATUS_MDEP
5007 C code for a data type which is used for declaring the @code{mdep}
5008 component of @code{cc_status}. It defaults to @code{int}.
5010 This macro is not used on machines that do not use @code{cc0}.
5012 @findex CC_STATUS_MDEP_INIT
5013 @item CC_STATUS_MDEP_INIT
5014 A C expression to initialize the @code{mdep} field to ``empty''.
5015 The default definition does nothing, since most machines don't use
5016 the field anyway. If you want to use the field, you should probably
5017 define this macro to initialize it.
5019 This macro is not used on machines that do not use @code{cc0}.
5021 @findex NOTICE_UPDATE_CC
5022 @item NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5023 A C compound statement to set the components of @code{cc_status}
5024 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5025 this macro's responsibility to recognize insns that set the condition
5026 code as a byproduct of other activity as well as those that explicitly
5029 This macro is not used on machines that do not use @code{cc0}.
5031 If there are insns that do not set the condition code but do alter
5032 other machine registers, this macro must check to see whether they
5033 invalidate the expressions that the condition code is recorded as
5034 reflecting. For example, on the 68000, insns that store in address
5035 registers do not set the condition code, which means that usually
5036 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5037 insns. But suppose that the previous insn set the condition code
5038 based on location @samp{a4@@(102)} and the current insn stores a new
5039 value in @samp{a4}. Although the condition code is not changed by
5040 this, it will no longer be true that it reflects the contents of
5041 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5042 @code{cc_status} in this case to say that nothing is known about the
5043 condition code value.
5045 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5046 with the results of peephole optimization: insns whose patterns are
5047 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5048 constants which are just the operands. The RTL structure of these
5049 insns is not sufficient to indicate what the insns actually do. What
5050 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5051 @code{CC_STATUS_INIT}.
5053 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5054 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5055 @samp{cc}. This avoids having detailed information about patterns in
5056 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5058 @findex EXTRA_CC_MODES
5059 @item EXTRA_CC_MODES
5060 Condition codes are represented in registers by machine modes of class
5061 @code{MODE_CC}. By default, there is just one mode, @code{CCmode}, with
5062 this class. If you need more such modes, create a file named
5063 @file{@var{machine}-modes.def} in your @file{config/@var{machine}}
5064 directory (@pxref{Back End, , Anatomy of a Target Back End}), containing
5065 a list of these modes. Each entry in the list should be a call to the
5066 macro @code{CC}. This macro takes one argument, which is the name of
5067 the mode: it should begin with @samp{CC}. Do not put quotation marks
5068 around the name, or include the trailing @samp{mode}; these are
5069 automatically added. There should not be anything else in the file
5072 A sample @file{@var{machine}-modes.def} file might look like this:
5075 CC (CC_NOOV) /* @r{Comparison only valid if there was no overflow.} */
5076 CC (CCFP) /* @r{Floating point comparison that cannot trap.} */
5077 CC (CCFPE) /* @r{Floating point comparison that may trap.} */
5080 When you create this file, the macro @code{EXTRA_CC_MODES} is
5081 automatically defined by @command{configure}, with value @samp{1}.
5083 @findex SELECT_CC_MODE
5084 @item SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5085 Returns a mode from class @code{MODE_CC} to be used when comparison
5086 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5087 example, on the Sparc, @code{SELECT_CC_MODE} is defined as (see
5088 @pxref{Jump Patterns} for a description of the reason for this
5092 #define SELECT_CC_MODE(OP,X,Y) \
5093 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5094 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5095 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5096 || GET_CODE (X) == NEG) \
5097 ? CC_NOOVmode : CCmode))
5100 You need not define this macro if @code{EXTRA_CC_MODES} is not defined.
5102 @findex CANONICALIZE_COMPARISON
5103 @item CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5104 On some machines not all possible comparisons are defined, but you can
5105 convert an invalid comparison into a valid one. For example, the Alpha
5106 does not have a @code{GT} comparison, but you can use an @code{LT}
5107 comparison instead and swap the order of the operands.
5109 On such machines, define this macro to be a C statement to do any
5110 required conversions. @var{code} is the initial comparison code
5111 and @var{op0} and @var{op1} are the left and right operands of the
5112 comparison, respectively. You should modify @var{code}, @var{op0}, and
5113 @var{op1} as required.
5115 GCC will not assume that the comparison resulting from this macro is
5116 valid but will see if the resulting insn matches a pattern in the
5119 You need not define this macro if it would never change the comparison
5122 @findex REVERSIBLE_CC_MODE
5123 @item REVERSIBLE_CC_MODE (@var{mode})
5124 A C expression whose value is one if it is always safe to reverse a
5125 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5126 can ever return @var{mode} for a floating-point inequality comparison,
5127 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5129 You need not define this macro if it would always returns zero or if the
5130 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5131 For example, here is the definition used on the Sparc, where floating-point
5132 inequality comparisons are always given @code{CCFPEmode}:
5135 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5138 @findex REVERSE_CONDITION (@var{code}, @var{mode})
5139 A C expression whose value is reversed condition code of the @var{code} for
5140 comparison done in CC_MODE @var{mode}. The macro is used only in case
5141 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5142 machine has some non-standard way how to reverse certain conditionals. For
5143 instance in case all floating point conditions are non-trapping, compiler may
5144 freely convert unordered compares to ordered one. Then definition may look
5148 #define REVERSE_CONDITION(CODE, MODE) \
5149 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5150 : reverse_condition_maybe_unordered (CODE))
5153 @findex REVERSE_CONDEXEC_PREDICATES_P
5154 @item REVERSE_CONDEXEC_PREDICATES_P (@var{code1}, @var{code2})
5155 A C expression that returns true if the conditional execution predicate
5156 @var{code1} is the inverse of @var{code2} and vice versa. Define this to
5157 return 0 if the target has conditional execution predicates that cannot be
5158 reversed safely. If no expansion is specified, this macro is defined as
5162 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5163 ((x) == reverse_condition (y))
5169 @section Describing Relative Costs of Operations
5170 @cindex costs of instructions
5171 @cindex relative costs
5172 @cindex speed of instructions
5174 These macros let you describe the relative speed of various operations
5175 on the target machine.
5179 @item CONST_COSTS (@var{x}, @var{code}, @var{outer_code})
5180 A part of a C @code{switch} statement that describes the relative costs
5181 of constant RTL expressions. It must contain @code{case} labels for
5182 expression codes @code{const_int}, @code{const}, @code{symbol_ref},
5183 @code{label_ref} and @code{const_double}. Each case must ultimately
5184 reach a @code{return} statement to return the relative cost of the use
5185 of that kind of constant value in an expression. The cost may depend on
5186 the precise value of the constant, which is available for examination in
5187 @var{x}, and the rtx code of the expression in which it is contained,
5188 found in @var{outer_code}.
5190 @var{code} is the expression code---redundant, since it can be
5191 obtained with @code{GET_CODE (@var{x})}.
5194 @findex COSTS_N_INSNS
5195 @item RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
5196 Like @code{CONST_COSTS} but applies to nonconstant RTL expressions.
5197 This can be used, for example, to indicate how costly a multiply
5198 instruction is. In writing this macro, you can use the construct
5199 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
5200 instructions. @var{outer_code} is the code of the expression in which
5201 @var{x} is contained.
5203 This macro is optional; do not define it if the default cost assumptions
5204 are adequate for the target machine.
5206 @findex DEFAULT_RTX_COSTS
5207 @item DEFAULT_RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
5208 This macro, if defined, is called for any case not handled by the
5209 @code{RTX_COSTS} or @code{CONST_COSTS} macros. This eliminates the need
5210 to put case labels into the macro, but the code, or any functions it
5211 calls, must assume that the RTL in @var{x} could be of any type that has
5212 not already been handled. The arguments are the same as for
5213 @code{RTX_COSTS}, and the macro should execute a return statement giving
5214 the cost of any RTL expressions that it can handle. The default cost
5215 calculation is used for any RTL for which this macro does not return a
5218 This macro is optional; do not define it if the default cost assumptions
5219 are adequate for the target machine.
5221 @findex ADDRESS_COST
5222 @item ADDRESS_COST (@var{address})
5223 An expression giving the cost of an addressing mode that contains
5224 @var{address}. If not defined, the cost is computed from
5225 the @var{address} expression and the @code{CONST_COSTS} values.
5227 For most CISC machines, the default cost is a good approximation of the
5228 true cost of the addressing mode. However, on RISC machines, all
5229 instructions normally have the same length and execution time. Hence
5230 all addresses will have equal costs.
5232 In cases where more than one form of an address is known, the form with
5233 the lowest cost will be used. If multiple forms have the same, lowest,
5234 cost, the one that is the most complex will be used.
5236 For example, suppose an address that is equal to the sum of a register
5237 and a constant is used twice in the same basic block. When this macro
5238 is not defined, the address will be computed in a register and memory
5239 references will be indirect through that register. On machines where
5240 the cost of the addressing mode containing the sum is no higher than
5241 that of a simple indirect reference, this will produce an additional
5242 instruction and possibly require an additional register. Proper
5243 specification of this macro eliminates this overhead for such machines.
5245 Similar use of this macro is made in strength reduction of loops.
5247 @var{address} need not be valid as an address. In such a case, the cost
5248 is not relevant and can be any value; invalid addresses need not be
5249 assigned a different cost.
5251 On machines where an address involving more than one register is as
5252 cheap as an address computation involving only one register, defining
5253 @code{ADDRESS_COST} to reflect this can cause two registers to be live
5254 over a region of code where only one would have been if
5255 @code{ADDRESS_COST} were not defined in that manner. This effect should
5256 be considered in the definition of this macro. Equivalent costs should
5257 probably only be given to addresses with different numbers of registers
5258 on machines with lots of registers.
5260 This macro will normally either not be defined or be defined as a
5263 @findex REGISTER_MOVE_COST
5264 @item REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5265 A C expression for the cost of moving data of mode @var{mode} from a
5266 register in class @var{from} to one in class @var{to}. The classes are
5267 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5268 value of 2 is the default; other values are interpreted relative to
5271 It is not required that the cost always equal 2 when @var{from} is the
5272 same as @var{to}; on some machines it is expensive to move between
5273 registers if they are not general registers.
5275 If reload sees an insn consisting of a single @code{set} between two
5276 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5277 classes returns a value of 2, reload does not check to ensure that the
5278 constraints of the insn are met. Setting a cost of other than 2 will
5279 allow reload to verify that the constraints are met. You should do this
5280 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5282 @findex MEMORY_MOVE_COST
5283 @item MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5284 A C expression for the cost of moving data of mode @var{mode} between a
5285 register of class @var{class} and memory; @var{in} is zero if the value
5286 is to be written to memory, nonzero if it is to be read in. This cost
5287 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5288 registers and memory is more expensive than between two registers, you
5289 should define this macro to express the relative cost.
5291 If you do not define this macro, GCC uses a default cost of 4 plus
5292 the cost of copying via a secondary reload register, if one is
5293 needed. If your machine requires a secondary reload register to copy
5294 between memory and a register of @var{class} but the reload mechanism is
5295 more complex than copying via an intermediate, define this macro to
5296 reflect the actual cost of the move.
5298 GCC defines the function @code{memory_move_secondary_cost} if
5299 secondary reloads are needed. It computes the costs due to copying via
5300 a secondary register. If your machine copies from memory using a
5301 secondary register in the conventional way but the default base value of
5302 4 is not correct for your machine, define this macro to add some other
5303 value to the result of that function. The arguments to that function
5304 are the same as to this macro.
5308 A C expression for the cost of a branch instruction. A value of 1 is
5309 the default; other values are interpreted relative to that.
5312 Here are additional macros which do not specify precise relative costs,
5313 but only that certain actions are more expensive than GCC would
5317 @findex SLOW_BYTE_ACCESS
5318 @item SLOW_BYTE_ACCESS
5319 Define this macro as a C expression which is nonzero if accessing less
5320 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5321 faster than accessing a word of memory, i.e., if such access
5322 require more than one instruction or if there is no difference in cost
5323 between byte and (aligned) word loads.
5325 When this macro is not defined, the compiler will access a field by
5326 finding the smallest containing object; when it is defined, a fullword
5327 load will be used if alignment permits. Unless bytes accesses are
5328 faster than word accesses, using word accesses is preferable since it
5329 may eliminate subsequent memory access if subsequent accesses occur to
5330 other fields in the same word of the structure, but to different bytes.
5332 @findex SLOW_UNALIGNED_ACCESS
5333 @item SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5334 Define this macro to be the value 1 if memory accesses described by the
5335 @var{mode} and @var{alignment} parameters have a cost many times greater
5336 than aligned accesses, for example if they are emulated in a trap
5339 When this macro is nonzero, the compiler will act as if
5340 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5341 moves. This can cause significantly more instructions to be produced.
5342 Therefore, do not set this macro nonzero if unaligned accesses only add a
5343 cycle or two to the time for a memory access.
5345 If the value of this macro is always zero, it need not be defined. If
5346 this macro is defined, it should produce a nonzero value when
5347 @code{STRICT_ALIGNMENT} is nonzero.
5349 @findex DONT_REDUCE_ADDR
5350 @item DONT_REDUCE_ADDR
5351 Define this macro to inhibit strength reduction of memory addresses.
5352 (On some machines, such strength reduction seems to do harm rather
5357 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5358 which a sequence of insns should be generated instead of a
5359 string move insn or a library call. Increasing the value will always
5360 make code faster, but eventually incurs high cost in increased code size.
5362 Note that on machines where the corresponding move insn is a
5363 @code{define_expand} that emits a sequence of insns, this macro counts
5364 the number of such sequences.
5366 If you don't define this, a reasonable default is used.
5368 @findex MOVE_BY_PIECES_P
5369 @item MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5370 A C expression used to determine whether @code{move_by_pieces} will be used to
5371 copy a chunk of memory, or whether some other block move mechanism
5372 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5373 than @code{MOVE_RATIO}.
5375 @findex MOVE_MAX_PIECES
5376 @item MOVE_MAX_PIECES
5377 A C expression used by @code{move_by_pieces} to determine the largest unit
5378 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5382 The threshold of number of scalar move insns, @emph{below} which a sequence
5383 of insns should be generated to clear memory instead of a string clear insn
5384 or a library call. Increasing the value will always make code faster, but
5385 eventually incurs high cost in increased code size.
5387 If you don't define this, a reasonable default is used.
5389 @findex CLEAR_BY_PIECES_P
5390 @item CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5391 A C expression used to determine whether @code{clear_by_pieces} will be used
5392 to clear a chunk of memory, or whether some other block clear mechanism
5393 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5394 than @code{CLEAR_RATIO}.
5396 @findex USE_LOAD_POST_INCREMENT
5397 @item USE_LOAD_POST_INCREMENT (@var{mode})
5398 A C expression used to determine whether a load postincrement is a good
5399 thing to use for a given mode. Defaults to the value of
5400 @code{HAVE_POST_INCREMENT}.
5402 @findex USE_LOAD_POST_DECREMENT
5403 @item USE_LOAD_POST_DECREMENT (@var{mode})
5404 A C expression used to determine whether a load postdecrement is a good
5405 thing to use for a given mode. Defaults to the value of
5406 @code{HAVE_POST_DECREMENT}.
5408 @findex USE_LOAD_PRE_INCREMENT
5409 @item USE_LOAD_PRE_INCREMENT (@var{mode})
5410 A C expression used to determine whether a load preincrement is a good
5411 thing to use for a given mode. Defaults to the value of
5412 @code{HAVE_PRE_INCREMENT}.
5414 @findex USE_LOAD_PRE_DECREMENT
5415 @item USE_LOAD_PRE_DECREMENT (@var{mode})
5416 A C expression used to determine whether a load predecrement is a good
5417 thing to use for a given mode. Defaults to the value of
5418 @code{HAVE_PRE_DECREMENT}.
5420 @findex USE_STORE_POST_INCREMENT
5421 @item USE_STORE_POST_INCREMENT (@var{mode})
5422 A C expression used to determine whether a store postincrement is a good
5423 thing to use for a given mode. Defaults to the value of
5424 @code{HAVE_POST_INCREMENT}.
5426 @findex USE_STORE_POST_DECREMENT
5427 @item USE_STORE_POST_DECREMENT (@var{mode})
5428 A C expression used to determine whether a store postdecrement is a good
5429 thing to use for a given mode. Defaults to the value of
5430 @code{HAVE_POST_DECREMENT}.
5432 @findex USE_STORE_PRE_INCREMENT
5433 @item USE_STORE_PRE_INCREMENT (@var{mode})
5434 This macro is used to determine whether a store preincrement is a good
5435 thing to use for a given mode. Defaults to the value of
5436 @code{HAVE_PRE_INCREMENT}.
5438 @findex USE_STORE_PRE_DECREMENT
5439 @item USE_STORE_PRE_DECREMENT (@var{mode})
5440 This macro is used to determine whether a store predecrement is a good
5441 thing to use for a given mode. Defaults to the value of
5442 @code{HAVE_PRE_DECREMENT}.
5444 @findex NO_FUNCTION_CSE
5445 @item NO_FUNCTION_CSE
5446 Define this macro if it is as good or better to call a constant
5447 function address than to call an address kept in a register.
5449 @findex NO_RECURSIVE_FUNCTION_CSE
5450 @item NO_RECURSIVE_FUNCTION_CSE
5451 Define this macro if it is as good or better for a function to call
5452 itself with an explicit address than to call an address kept in a
5457 @section Adjusting the Instruction Scheduler
5459 The instruction scheduler may need a fair amount of machine-specific
5460 adjustment in order to produce good code. GCC provides several target
5461 hooks for this purpose. It is usually enough to define just a few of
5462 them: try the first ones in this list first.
5464 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
5465 This hook returns the maximum number of instructions that can ever
5466 issue at the same time on the target machine. The default is one.
5467 Although the insn scheduler can define itself the possibility of issue
5468 an insn on the same cycle, the value can serve as an additional
5469 constraint to issue insns on the same simulated processor cycle (see
5470 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
5471 This value must be constant over the entire compilation. If you need
5472 it to vary depending on what the instructions are, you must use
5473 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
5475 For the automaton based pipeline interface, you could define this hook
5476 to return the value of the macro @code{MAX_DFA_ISSUE_RATE}.
5479 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
5480 This hook is executed by the scheduler after it has scheduled an insn
5481 from the ready list. It should return the number of insns which can
5482 still be issued in the current cycle. Normally this is
5483 @samp{@w{@var{more} - 1}}. You should define this hook if some insns
5484 take more machine resources than others, so that fewer insns can follow
5485 them in the same cycle. @var{file} is either a null pointer, or a stdio
5486 stream to write any debug output to. @var{verbose} is the verbose level
5487 provided by @option{-fsched-verbose-@var{n}}. @var{insn} is the
5488 instruction that was scheduled.
5491 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
5492 This function corrects the value of @var{cost} based on the
5493 relationship between @var{insn} and @var{dep_insn} through the
5494 dependence @var{link}. It should return the new value. The default
5495 is to make no adjustment to @var{cost}. This can be used for example
5496 to specify to the scheduler using the traditional pipeline description
5497 that an output- or anti-dependence does not incur the same cost as a
5498 data-dependence. If the scheduler using the automaton based pipeline
5499 description, the cost of anti-dependence is zero and the cost of
5500 output-dependence is maximum of one and the difference of latency
5501 times of the first and the second insns. If these values are not
5502 acceptable, you could use the hook to modify them too. See also
5503 @pxref{Automaton pipeline description}.
5506 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
5507 This hook adjusts the integer scheduling priority @var{priority} of
5508 @var{insn}. It should return the new priority. Reduce the priority to
5509 execute @var{insn} earlier, increase the priority to execute @var{insn}
5510 later. Do not define this hook if you do not need to adjust the
5511 scheduling priorities of insns.
5514 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
5515 This hook is executed by the scheduler after it has scheduled the ready
5516 list, to allow the machine description to reorder it (for example to
5517 combine two small instructions together on @samp{VLIW} machines).
5518 @var{file} is either a null pointer, or a stdio stream to write any
5519 debug output to. @var{verbose} is the verbose level provided by
5520 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
5521 list of instructions that are ready to be scheduled. @var{n_readyp} is
5522 a pointer to the number of elements in the ready list. The scheduler
5523 reads the ready list in reverse order, starting with
5524 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
5525 is the timer tick of the scheduler. You may modify the ready list and
5526 the number of ready insns. The return value is the number of insns that
5527 can issue this cycle; normally this is just @code{issue_rate}. See also
5528 @samp{TARGET_SCHED_REORDER2}.
5531 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
5532 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
5533 function is called whenever the scheduler starts a new cycle. This one
5534 is called once per iteration over a cycle, immediately after
5535 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
5536 return the number of insns to be scheduled in the same cycle. Defining
5537 this hook can be useful if there are frequent situations where
5538 scheduling one insn causes other insns to become ready in the same
5539 cycle. These other insns can then be taken into account properly.
5542 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
5543 This hook is executed by the scheduler at the beginning of each block of
5544 instructions that are to be scheduled. @var{file} is either a null
5545 pointer, or a stdio stream to write any debug output to. @var{verbose}
5546 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5547 @var{max_ready} is the maximum number of insns in the current scheduling
5548 region that can be live at the same time. This can be used to allocate
5549 scratch space if it is needed, e.g. by @samp{TARGET_SCHED_REORDER}.
5552 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
5553 This hook is executed by the scheduler at the end of each block of
5554 instructions that are to be scheduled. It can be used to perform
5555 cleanup of any actions done by the other scheduling hooks. @var{file}
5556 is either a null pointer, or a stdio stream to write any debug output
5557 to. @var{verbose} is the verbose level provided by
5558 @option{-fsched-verbose-@var{n}}.
5561 @deftypefn {Target Hook} int TARGET_SCHED_USE_DFA_PIPELINE_INTERFACE (void)
5562 This hook is called many times during insn scheduling. If the hook
5563 returns nonzero, the automaton based pipeline description is used for
5564 insn scheduling. Otherwise the traditional pipeline description is
5565 used. The default is usage of the traditional pipeline description.
5567 You should also remember that to simplify the insn scheduler sources
5568 an empty traditional pipeline description interface is generated even
5569 if there is no a traditional pipeline description in the @file{.md}
5570 file. The same is true for the automaton based pipeline description.
5571 That means that you should be accurate in defining the hook.
5574 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
5575 The hook returns an RTL insn. The automaton state used in the
5576 pipeline hazard recognizer is changed as if the insn were scheduled
5577 when the new simulated processor cycle starts. Usage of the hook may
5578 simplify the automaton pipeline description for some @acronym{VLIW}
5579 processors. If the hook is defined, it is used only for the automaton
5580 based pipeline description. The default is not to change the state
5581 when the new simulated processor cycle starts.
5584 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
5585 The hook can be used to initialize data used by the previous hook.
5588 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
5589 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
5590 to changed the state as if the insn were scheduled when the new
5591 simulated processor cycle finishes.
5594 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
5595 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
5596 used to initialize data used by the previous hook.
5599 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
5600 This hook controls better choosing an insn from the ready insn queue
5601 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
5602 chooses the first insn from the queue. If the hook returns a positive
5603 value, an additional scheduler code tries all permutations of
5604 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
5605 subsequent ready insns to choose an insn whose issue will result in
5606 maximal number of issued insns on the same cycle. For the
5607 @acronym{VLIW} processor, the code could actually solve the problem of
5608 packing simple insns into the @acronym{VLIW} insn. Of course, if the
5609 rules of @acronym{VLIW} packing are described in the automaton.
5611 This code also could be used for superscalar @acronym{RISC}
5612 processors. Let us consider a superscalar @acronym{RISC} processor
5613 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
5614 @var{B}, some insns can be executed only in pipelines @var{B} or
5615 @var{C}, and one insn can be executed in pipeline @var{B}. The
5616 processor may issue the 1st insn into @var{A} and the 2nd one into
5617 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
5618 until the next cycle. If the scheduler issues the 3rd insn the first,
5619 the processor could issue all 3 insns per cycle.
5621 Actually this code demonstrates advantages of the automaton based
5622 pipeline hazard recognizer. We try quickly and easy many insn
5623 schedules to choose the best one.
5625 The default is no multipass scheduling.
5628 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_BUBBLES (void)
5629 The @acronym{DFA}-based scheduler could take the insertion of nop
5630 operations for better insn scheduling into account. It can be done
5631 only if the multi-pass insn scheduling works (see hook
5632 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD}).
5634 Let us consider a @acronym{VLIW} processor insn with 3 slots. Each
5635 insn can be placed only in one of the three slots. We have 3 ready
5636 insns @var{A}, @var{B}, and @var{C}. @var{A} and @var{C} can be
5637 placed only in the 1st slot, @var{B} can be placed only in the 3rd
5638 slot. We described the automaton which does not permit empty slot
5639 gaps between insns (usually such description is simpler). Without
5640 this code the scheduler would place each insn in 3 separate
5641 @acronym{VLIW} insns. If the scheduler places a nop insn into the 2nd
5642 slot, it could place the 3 insns into 2 @acronym{VLIW} insns. What is
5643 the nop insn is returned by hook @samp{TARGET_SCHED_DFA_BUBBLE}. Hook
5644 @samp{TARGET_SCHED_INIT_DFA_BUBBLES} can be used to initialize or
5645 create the nop insns.
5647 You should remember that the scheduler does not insert the nop insns.
5648 It is not wise because of the following optimizations. The scheduler
5649 only considers such possibility to improve the result schedule. The
5650 nop insns should be inserted lately, e.g. on the final phase.
5653 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_BUBBLE (int @var{index})
5654 This hook @samp{FIRST_CYCLE_MULTIPASS_SCHEDULING} is used to insert
5655 nop operations for better insn scheduling when @acronym{DFA}-based
5656 scheduler makes multipass insn scheduling (see also description of
5657 hook @samp{TARGET_SCHED_INIT_DFA_BUBBLES}). This hook
5658 returns a nop insn with given @var{index}. The indexes start with
5659 zero. The hook should return @code{NULL} if there are no more nop
5660 insns with indexes greater than given index.
5663 Macros in the following table are generated by the program
5664 @file{genattr} and can be useful for writing the hooks.
5667 @findex TRADITIONAL_PIPELINE_INTERFACE
5668 @item TRADITIONAL_PIPELINE_INTERFACE
5669 The macro definition is generated if there is a traditional pipeline
5670 description in @file{.md} file. You should also remember that to
5671 simplify the insn scheduler sources an empty traditional pipeline
5672 description interface is generated even if there is no a traditional
5673 pipeline description in the @file{.md} file. The macro can be used to
5674 distinguish the two types of the traditional interface.
5676 @findex DFA_PIPELINE_INTERFACE
5677 @item DFA_PIPELINE_INTERFACE
5678 The macro definition is generated if there is an automaton pipeline
5679 description in @file{.md} file. You should also remember that to
5680 simplify the insn scheduler sources an empty automaton pipeline
5681 description interface is generated even if there is no an automaton
5682 pipeline description in the @file{.md} file. The macro can be used to
5683 distinguish the two types of the automaton interface.
5685 @findex MAX_DFA_ISSUE_RATE
5686 @item MAX_DFA_ISSUE_RATE
5687 The macro definition is generated in the automaton based pipeline
5688 description interface. Its value is calculated from the automaton
5689 based pipeline description and is equal to maximal number of all insns
5690 described in constructions @samp{define_insn_reservation} which can be
5691 issued on the same processor cycle.
5696 @section Dividing the Output into Sections (Texts, Data, @dots{})
5697 @c the above section title is WAY too long. maybe cut the part between
5698 @c the (...)? --mew 10feb93
5700 An object file is divided into sections containing different types of
5701 data. In the most common case, there are three sections: the @dfn{text
5702 section}, which holds instructions and read-only data; the @dfn{data
5703 section}, which holds initialized writable data; and the @dfn{bss
5704 section}, which holds uninitialized data. Some systems have other kinds
5707 The compiler must tell the assembler when to switch sections. These
5708 macros control what commands to output to tell the assembler this. You
5709 can also define additional sections.
5712 @findex TEXT_SECTION_ASM_OP
5713 @item TEXT_SECTION_ASM_OP
5714 A C expression whose value is a string, including spacing, containing the
5715 assembler operation that should precede instructions and read-only data.
5716 Normally @code{"\t.text"} is right.
5718 @findex TEXT_SECTION
5720 A C statement that switches to the default section containing instructions.
5721 Normally this is not needed, as simply defining @code{TEXT_SECTION_ASM_OP}
5722 is enough. The MIPS port uses this to sort all functions after all data
5725 @findex HOT_TEXT_SECTION_NAME
5726 @item HOT_TEXT_SECTION_NAME
5727 If defined, a C string constant for the name of the section containing most
5728 frequently executed functions of the program. If not defined, GCC will provide
5729 a default definition if the target supports named sections.
5731 @findex UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5732 @item UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5733 If defined, a C string constant for the name of the section containing unlikely
5734 executed functions in the program.
5736 @findex DATA_SECTION_ASM_OP
5737 @item DATA_SECTION_ASM_OP
5738 A C expression whose value is a string, including spacing, containing the
5739 assembler operation to identify the following data as writable initialized
5740 data. Normally @code{"\t.data"} is right.
5742 @findex READONLY_DATA_SECTION_ASM_OP
5743 @item READONLY_DATA_SECTION_ASM_OP
5744 A C expression whose value is a string, including spacing, containing the
5745 assembler operation to identify the following data as read-only initialized
5748 @findex READONLY_DATA_SECTION
5749 @item READONLY_DATA_SECTION
5750 A macro naming a function to call to switch to the proper section for
5751 read-only data. The default is to use @code{READONLY_DATA_SECTION_ASM_OP}
5752 if defined, else fall back to @code{text_section}.
5754 The most common definition will be @code{data_section}, if the target
5755 does not have a special read-only data section, and does not put data
5756 in the text section.
5758 @findex SHARED_SECTION_ASM_OP
5759 @item SHARED_SECTION_ASM_OP
5760 If defined, a C expression whose value is a string, including spacing,
5761 containing the assembler operation to identify the following data as
5762 shared data. If not defined, @code{DATA_SECTION_ASM_OP} will be used.
5764 @findex BSS_SECTION_ASM_OP
5765 @item BSS_SECTION_ASM_OP
5766 If defined, a C expression whose value is a string, including spacing,
5767 containing the assembler operation to identify the following data as
5768 uninitialized global data. If not defined, and neither
5769 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5770 uninitialized global data will be output in the data section if
5771 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5774 @findex SHARED_BSS_SECTION_ASM_OP
5775 @item SHARED_BSS_SECTION_ASM_OP
5776 If defined, a C expression whose value is a string, including spacing,
5777 containing the assembler operation to identify the following data as
5778 uninitialized global shared data. If not defined, and
5779 @code{BSS_SECTION_ASM_OP} is, the latter will be used.
5781 @findex INIT_SECTION_ASM_OP
5782 @item INIT_SECTION_ASM_OP
5783 If defined, a C expression whose value is a string, including spacing,
5784 containing the assembler operation to identify the following data as
5785 initialization code. If not defined, GCC will assume such a section does
5788 @findex FINI_SECTION_ASM_OP
5789 @item FINI_SECTION_ASM_OP
5790 If defined, a C expression whose value is a string, including spacing,
5791 containing the assembler operation to identify the following data as
5792 finalization code. If not defined, GCC will assume such a section does
5795 @findex CRT_CALL_STATIC_FUNCTION
5796 @item CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
5797 If defined, an ASM statement that switches to a different section
5798 via @var{section_op}, calls @var{function}, and switches back to
5799 the text section. This is used in @file{crtstuff.c} if
5800 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
5801 to initialization and finalization functions from the init and fini
5802 sections. By default, this macro uses a simple function call. Some
5803 ports need hand-crafted assembly code to avoid dependencies on
5804 registers initialized in the function prologue or to ensure that
5805 constant pools don't end up too far way in the text section.
5807 @findex FORCE_CODE_SECTION_ALIGN
5808 @item FORCE_CODE_SECTION_ALIGN
5809 If defined, an ASM statement that aligns a code section to some
5810 arbitrary boundary. This is used to force all fragments of the
5811 @code{.init} and @code{.fini} sections to have to same alignment
5812 and thus prevent the linker from having to add any padding.
5814 @findex EXTRA_SECTIONS
5817 @item EXTRA_SECTIONS
5818 A list of names for sections other than the standard two, which are
5819 @code{in_text} and @code{in_data}. You need not define this macro
5820 on a system with no other sections (that GCC needs to use).
5822 @findex EXTRA_SECTION_FUNCTIONS
5823 @findex text_section
5824 @findex data_section
5825 @item EXTRA_SECTION_FUNCTIONS
5826 One or more functions to be defined in @file{varasm.c}. These
5827 functions should do jobs analogous to those of @code{text_section} and
5828 @code{data_section}, for your additional sections. Do not define this
5829 macro if you do not define @code{EXTRA_SECTIONS}.
5831 @findex JUMP_TABLES_IN_TEXT_SECTION
5832 @item JUMP_TABLES_IN_TEXT_SECTION
5833 Define this macro to be an expression with a nonzero value if jump
5834 tables (for @code{tablejump} insns) should be output in the text
5835 section, along with the assembler instructions. Otherwise, the
5836 readonly data section is used.
5838 This macro is irrelevant if there is no separate readonly data section.
5841 @deftypefn {Target Hook} void TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
5842 Switches to the appropriate section for output of @var{exp}. You can
5843 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
5844 some sort. @var{reloc} indicates whether the initial value of @var{exp}
5845 requires link-time relocations. Bit 0 is set when variable contains
5846 local relocations only, while bit 1 is set for global relocations.
5847 Select the section by calling @code{data_section} or one of the
5848 alternatives for other sections. @var{align} is the constant alignment
5851 The default version of this function takes care of putting read-only
5852 variables in @code{readonly_data_section}.
5855 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
5856 Build up a unique section name, expressed as a @code{STRING_CST} node,
5857 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
5858 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
5859 the initial value of @var{exp} requires link-time relocations.
5861 The default version of this function appends the symbol name to the
5862 ELF section name that would normally be used for the symbol. For
5863 example, the function @code{foo} would be placed in @code{.text.foo}.
5864 Whatever the actual target object format, this is often good enough.
5867 @deftypefn {Target Hook} void TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
5868 Switches to the appropriate section for output of constant pool entry
5869 @var{x} in @var{mode}. You can assume that @var{x} is some kind of
5870 constant in RTL@. The argument @var{mode} is redundant except in the
5871 case of a @code{const_int} rtx. Select the section by calling
5872 @code{readonly_data_section} or one of the alternatives for other
5873 sections. @var{align} is the constant alignment in bits.
5875 The default version of this function takes care of putting symbolic
5876 constants in @code{flag_pic} mode in @code{data_section} and everything
5877 else in @code{readonly_data_section}.
5880 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, int @var{new_decl_p})
5881 Define this hook if references to a symbol or a constant must be
5882 treated differently depending on something about the variable or
5883 function named by the symbol (such as what section it is in).
5885 The hook is executed under two circumstances. One is immediately after
5886 the rtl for @var{decl} that represents a variable or a function has been
5887 created and stored in @code{DECL_RTL(@var{decl})}. The value of the rtl
5888 will be a @code{mem} whose address is a @code{symbol_ref}. The other is
5889 immediately after the rtl for @var{decl} that represents a constant has
5890 been created and stored in @code{TREE_CST_RTL (@var{decl})}. The macro
5891 is called once for each distinct constant in a source file.
5893 The @var{new_decl_p} argument will be true if this is the first time
5894 that @code{ENCODE_SECTION_INFO} has been invoked on this decl. It will
5895 be false for subsequent invocations, which will happen for duplicate
5896 declarations. Whether or not anything must be done for the duplicate
5897 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
5899 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
5900 The usual thing for this hook to do is to record a flag in the
5901 @code{symbol_ref} (such as @code{SYMBOL_REF_FLAG}) or to store a
5902 modified name string in the @code{symbol_ref} (if one bit is not
5903 enough information).
5906 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
5907 Decode @var{name} and return the real name part, sans
5908 the characters that @code{TARGET_ENCODE_SECTION_INFO}
5912 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
5913 Returns true if @var{exp} should be placed into a ``small data'' section.
5914 The default version of this hook always returns false.
5917 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
5918 Returns true if @var{exp} names an object for which name resolution
5919 rules must resolve to the current ``module'' (dynamic shared library
5920 or executable image).
5922 The default version of this hook implements the name resolution rules
5923 for ELF, which has a looser model of global name binding than other
5924 currently supported object file formats.
5928 @section Position Independent Code
5929 @cindex position independent code
5932 This section describes macros that help implement generation of position
5933 independent code. Simply defining these macros is not enough to
5934 generate valid PIC; you must also add support to the macros
5935 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
5936 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
5937 @samp{movsi} to do something appropriate when the source operand
5938 contains a symbolic address. You may also need to alter the handling of
5939 switch statements so that they use relative addresses.
5940 @c i rearranged the order of the macros above to try to force one of
5941 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
5944 @findex PIC_OFFSET_TABLE_REGNUM
5945 @item PIC_OFFSET_TABLE_REGNUM
5946 The register number of the register used to address a table of static
5947 data addresses in memory. In some cases this register is defined by a
5948 processor's ``application binary interface'' (ABI)@. When this macro
5949 is defined, RTL is generated for this register once, as with the stack
5950 pointer and frame pointer registers. If this macro is not defined, it
5951 is up to the machine-dependent files to allocate such a register (if
5952 necessary). Note that this register must be fixed when in use (e.g.@:
5953 when @code{flag_pic} is true).
5955 @findex PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5956 @item PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5957 Define this macro if the register defined by
5958 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
5959 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
5961 @findex FINALIZE_PIC
5963 By generating position-independent code, when two different programs (A
5964 and B) share a common library (libC.a), the text of the library can be
5965 shared whether or not the library is linked at the same address for both
5966 programs. In some of these environments, position-independent code
5967 requires not only the use of different addressing modes, but also
5968 special code to enable the use of these addressing modes.
5970 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
5971 codes once the function is being compiled into assembly code, but not
5972 before. (It is not done before, because in the case of compiling an
5973 inline function, it would lead to multiple PIC prologues being
5974 included in functions which used inline functions and were compiled to
5977 @findex LEGITIMATE_PIC_OPERAND_P
5978 @item LEGITIMATE_PIC_OPERAND_P (@var{x})
5979 A C expression that is nonzero if @var{x} is a legitimate immediate
5980 operand on the target machine when generating position independent code.
5981 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
5982 check this. You can also assume @var{flag_pic} is true, so you need not
5983 check it either. You need not define this macro if all constants
5984 (including @code{SYMBOL_REF}) can be immediate operands when generating
5985 position independent code.
5988 @node Assembler Format
5989 @section Defining the Output Assembler Language
5991 This section describes macros whose principal purpose is to describe how
5992 to write instructions in assembler language---rather than what the
5996 * File Framework:: Structural information for the assembler file.
5997 * Data Output:: Output of constants (numbers, strings, addresses).
5998 * Uninitialized Data:: Output of uninitialized variables.
5999 * Label Output:: Output and generation of labels.
6000 * Initialization:: General principles of initialization
6001 and termination routines.
6002 * Macros for Initialization::
6003 Specific macros that control the handling of
6004 initialization and termination routines.
6005 * Instruction Output:: Output of actual instructions.
6006 * Dispatch Tables:: Output of jump tables.
6007 * Exception Region Output:: Output of exception region code.
6008 * Alignment Output:: Pseudo ops for alignment and skipping data.
6011 @node File Framework
6012 @subsection The Overall Framework of an Assembler File
6013 @cindex assembler format
6014 @cindex output of assembler code
6016 @c prevent bad page break with this line
6017 This describes the overall framework of an assembler file.
6020 @findex ASM_FILE_START
6021 @item ASM_FILE_START (@var{stream})
6022 A C expression which outputs to the stdio stream @var{stream}
6023 some appropriate text to go at the start of an assembler file.
6025 Normally this macro is defined to output a line containing
6026 @samp{#NO_APP}, which is a comment that has no effect on most
6027 assemblers but tells the GNU assembler that it can save time by not
6028 checking for certain assembler constructs.
6030 On systems that use SDB, it is necessary to output certain commands;
6031 see @file{attasm.h}.
6033 @findex ASM_FILE_END
6034 @item ASM_FILE_END (@var{stream})
6035 A C expression which outputs to the stdio stream @var{stream}
6036 some appropriate text to go at the end of an assembler file.
6038 If this macro is not defined, the default is to output nothing
6039 special at the end of the file. Most systems don't require any
6042 On systems that use SDB, it is necessary to output certain commands;
6043 see @file{attasm.h}.
6045 @findex ASM_COMMENT_START
6046 @item ASM_COMMENT_START
6047 A C string constant describing how to begin a comment in the target
6048 assembler language. The compiler assumes that the comment will end at
6049 the end of the line.
6053 A C string constant for text to be output before each @code{asm}
6054 statement or group of consecutive ones. Normally this is
6055 @code{"#APP"}, which is a comment that has no effect on most
6056 assemblers but tells the GNU assembler that it must check the lines
6057 that follow for all valid assembler constructs.
6061 A C string constant for text to be output after each @code{asm}
6062 statement or group of consecutive ones. Normally this is
6063 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6064 time-saving assumptions that are valid for ordinary compiler output.
6066 @findex ASM_OUTPUT_SOURCE_FILENAME
6067 @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6068 A C statement to output COFF information or DWARF debugging information
6069 which indicates that filename @var{name} is the current source file to
6070 the stdio stream @var{stream}.
6072 This macro need not be defined if the standard form of output
6073 for the file format in use is appropriate.
6075 @findex OUTPUT_QUOTED_STRING
6076 @item OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6077 A C statement to output the string @var{string} to the stdio stream
6078 @var{stream}. If you do not call the function @code{output_quoted_string}
6079 in your config files, GCC will only call it to output filenames to
6080 the assembler source. So you can use it to canonicalize the format
6081 of the filename using this macro.
6083 @findex ASM_OUTPUT_SOURCE_LINE
6084 @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
6085 A C statement to output DBX or SDB debugging information before code
6086 for line number @var{line} of the current source file to the
6087 stdio stream @var{stream}.
6089 This macro need not be defined if the standard form of debugging
6090 information for the debugger in use is appropriate.
6092 @findex ASM_OUTPUT_IDENT
6093 @item ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6094 A C statement to output something to the assembler file to handle a
6095 @samp{#ident} directive containing the text @var{string}. If this
6096 macro is not defined, nothing is output for a @samp{#ident} directive.
6098 @findex OBJC_PROLOGUE
6100 A C statement to output any assembler statements which are required to
6101 precede any Objective-C object definitions or message sending. The
6102 statement is executed only when compiling an Objective-C program.
6105 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6106 Output assembly directives to switch to section @var{name}. The section
6107 should have attributes as specified by @var{flags}, which is a bit mask
6108 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6109 is nonzero, it contains an alignment in bytes to be used for the section,
6110 otherwise some target default should be used. Only targets that must
6111 specify an alignment within the section directive need pay attention to
6112 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6115 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6116 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6119 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6120 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6121 based on a variable or function decl, a section name, and whether or not the
6122 declaration's initializer may contain runtime relocations. @var{decl} may be
6123 null, in which case read-write data should be assumed.
6125 The default version if this function handles choosing code vs data,
6126 read-only vs read-write data, and @code{flag_pic}. You should only
6127 need to override this if your target has special flags that might be
6128 set via @code{__attribute__}.
6133 @subsection Output of Data
6136 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6137 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6138 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6139 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6140 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6141 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6142 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6143 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6144 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6145 These hooks specify assembly directives for creating certain kinds
6146 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6147 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6148 aligned two-byte object, and so on. Any of the hooks may be
6149 @code{NULL}, indicating that no suitable directive is available.
6151 The compiler will print these strings at the start of a new line,
6152 followed immediately by the object's initial value. In most cases,
6153 the string should contain a tab, a pseudo-op, and then another tab.
6156 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6157 The @code{assemble_integer} function uses this hook to output an
6158 integer object. @var{x} is the object's value, @var{size} is its size
6159 in bytes and @var{aligned_p} indicates whether it is aligned. The
6160 function should return @code{true} if it was able to output the
6161 object. If it returns false, @code{assemble_integer} will try to
6162 split the object into smaller parts.
6164 The default implementation of this hook will use the
6165 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6166 when the relevant string is @code{NULL}.
6170 @findex OUTPUT_ADDR_CONST_EXTRA
6171 @item OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6172 A C statement to recognize @var{rtx} patterns that
6173 @code{output_addr_const} can't deal with, and output assembly code to
6174 @var{stream} corresponding to the pattern @var{x}. This may be used to
6175 allow machine-dependent @code{UNSPEC}s to appear within constants.
6177 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6178 @code{goto fail}, so that a standard error message is printed. If it
6179 prints an error message itself, by calling, for example,
6180 @code{output_operand_lossage}, it may just complete normally.
6182 @findex ASM_OUTPUT_ASCII
6183 @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6184 A C statement to output to the stdio stream @var{stream} an assembler
6185 instruction to assemble a string constant containing the @var{len}
6186 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6187 @code{char *} and @var{len} a C expression of type @code{int}.
6189 If the assembler has a @code{.ascii} pseudo-op as found in the
6190 Berkeley Unix assembler, do not define the macro
6191 @code{ASM_OUTPUT_ASCII}.
6193 @findex ASM_OUTPUT_FDESC
6194 @item ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6195 A C statement to output word @var{n} of a function descriptor for
6196 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6197 is defined, and is otherwise unused.
6199 @findex CONSTANT_POOL_BEFORE_FUNCTION
6200 @item CONSTANT_POOL_BEFORE_FUNCTION
6201 You may define this macro as a C expression. You should define the
6202 expression to have a nonzero value if GCC should output the constant
6203 pool for a function before the code for the function, or a zero value if
6204 GCC should output the constant pool after the function. If you do
6205 not define this macro, the usual case, GCC will output the constant
6206 pool before the function.
6208 @findex ASM_OUTPUT_POOL_PROLOGUE
6209 @item ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
6210 A C statement to output assembler commands to define the start of the
6211 constant pool for a function. @var{funname} is a string giving
6212 the name of the function. Should the return type of the function
6213 be required, it can be obtained via @var{fundecl}. @var{size}
6214 is the size, in bytes, of the constant pool that will be written
6215 immediately after this call.
6217 If no constant-pool prefix is required, the usual case, this macro need
6220 @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY
6221 @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
6222 A C statement (with or without semicolon) to output a constant in the
6223 constant pool, if it needs special treatment. (This macro need not do
6224 anything for RTL expressions that can be output normally.)
6226 The argument @var{file} is the standard I/O stream to output the
6227 assembler code on. @var{x} is the RTL expression for the constant to
6228 output, and @var{mode} is the machine mode (in case @var{x} is a
6229 @samp{const_int}). @var{align} is the required alignment for the value
6230 @var{x}; you should output an assembler directive to force this much
6233 The argument @var{labelno} is a number to use in an internal label for
6234 the address of this pool entry. The definition of this macro is
6235 responsible for outputting the label definition at the proper place.
6236 Here is how to do this:
6239 ASM_OUTPUT_INTERNAL_LABEL (@var{file}, "LC", @var{labelno});
6242 When you output a pool entry specially, you should end with a
6243 @code{goto} to the label @var{jumpto}. This will prevent the same pool
6244 entry from being output a second time in the usual manner.
6246 You need not define this macro if it would do nothing.
6248 @findex CONSTANT_AFTER_FUNCTION_P
6249 @item CONSTANT_AFTER_FUNCTION_P (@var{exp})
6250 Define this macro as a C expression which is nonzero if the constant
6251 @var{exp}, of type @code{tree}, should be output after the code for a
6252 function. The compiler will normally output all constants before the
6253 function; you need not define this macro if this is OK@.
6255 @findex ASM_OUTPUT_POOL_EPILOGUE
6256 @item ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
6257 A C statement to output assembler commands to at the end of the constant
6258 pool for a function. @var{funname} is a string giving the name of the
6259 function. Should the return type of the function be required, you can
6260 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
6261 constant pool that GCC wrote immediately before this call.
6263 If no constant-pool epilogue is required, the usual case, you need not
6266 @findex IS_ASM_LOGICAL_LINE_SEPARATOR
6267 @item IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
6268 Define this macro as a C expression which is nonzero if @var{C} is
6269 used as a logical line separator by the assembler.
6271 If you do not define this macro, the default is that only
6272 the character @samp{;} is treated as a logical line separator.
6275 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
6276 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
6277 These target hooks are C string constants, describing the syntax in the
6278 assembler for grouping arithmetic expressions. If not overridden, they
6279 default to normal parentheses, which is correct for most assemblers.
6282 These macros are provided by @file{real.h} for writing the definitions
6283 of @code{ASM_OUTPUT_DOUBLE} and the like:
6286 @item REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
6287 @itemx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
6288 @itemx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
6289 @findex REAL_VALUE_TO_TARGET_SINGLE
6290 @findex REAL_VALUE_TO_TARGET_DOUBLE
6291 @findex REAL_VALUE_TO_TARGET_LONG_DOUBLE
6292 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
6293 floating point representation, and store its bit pattern in the variable
6294 @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE}, this variable should
6295 be a simple @code{long int}. For the others, it should be an array of
6296 @code{long int}. The number of elements in this array is determined by
6297 the size of the desired target floating point data type: 32 bits of it
6298 go in each @code{long int} array element. Each array element holds 32
6299 bits of the result, even if @code{long int} is wider than 32 bits on the
6302 The array element values are designed so that you can print them out
6303 using @code{fprintf} in the order they should appear in the target
6306 @item REAL_VALUE_TO_DECIMAL (@var{x}, @var{format}, @var{string})
6307 @findex REAL_VALUE_TO_DECIMAL
6308 This macro converts @var{x}, of type @code{REAL_VALUE_TYPE}, to a
6309 decimal number and stores it as a string into @var{string}.
6310 You must pass, as @var{string}, the address of a long enough block
6311 of space to hold the result.
6313 The argument @var{format} is a @code{printf}-specification that serves
6314 as a suggestion for how to format the output string.
6317 @node Uninitialized Data
6318 @subsection Output of Uninitialized Variables
6320 Each of the macros in this section is used to do the whole job of
6321 outputting a single uninitialized variable.
6324 @findex ASM_OUTPUT_COMMON
6325 @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6326 A C statement (sans semicolon) to output to the stdio stream
6327 @var{stream} the assembler definition of a common-label named
6328 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6329 is the size rounded up to whatever alignment the caller wants.
6331 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6332 output the name itself; before and after that, output the additional
6333 assembler syntax for defining the name, and a newline.
6335 This macro controls how the assembler definitions of uninitialized
6336 common global variables are output.
6338 @findex ASM_OUTPUT_ALIGNED_COMMON
6339 @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
6340 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
6341 separate, explicit argument. If you define this macro, it is used in
6342 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
6343 handling the required alignment of the variable. The alignment is specified
6344 as the number of bits.
6346 @findex ASM_OUTPUT_ALIGNED_DECL_COMMON
6347 @item ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6348 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
6349 variable to be output, if there is one, or @code{NULL_TREE} if there
6350 is no corresponding variable. If you define this macro, GCC will use it
6351 in place of both @code{ASM_OUTPUT_COMMON} and
6352 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
6353 the variable's decl in order to chose what to output.
6355 @findex ASM_OUTPUT_SHARED_COMMON
6356 @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6357 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
6358 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
6361 @findex ASM_OUTPUT_BSS
6362 @item ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6363 A C statement (sans semicolon) to output to the stdio stream
6364 @var{stream} the assembler definition of uninitialized global @var{decl} named
6365 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6366 is the size rounded up to whatever alignment the caller wants.
6368 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
6369 defining this macro. If unable, use the expression
6370 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
6371 before and after that, output the additional assembler syntax for defining
6372 the name, and a newline.
6374 This macro controls how the assembler definitions of uninitialized global
6375 variables are output. This macro exists to properly support languages like
6376 C++ which do not have @code{common} data. However, this macro currently
6377 is not defined for all targets. If this macro and
6378 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
6379 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
6380 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
6382 @findex ASM_OUTPUT_ALIGNED_BSS
6383 @item ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6384 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
6385 separate, explicit argument. If you define this macro, it is used in
6386 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
6387 handling the required alignment of the variable. The alignment is specified
6388 as the number of bits.
6390 Try to use function @code{asm_output_aligned_bss} defined in file
6391 @file{varasm.c} when defining this macro.
6393 @findex ASM_OUTPUT_SHARED_BSS
6394 @item ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6395 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
6396 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
6399 @findex ASM_OUTPUT_LOCAL
6400 @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6401 A C statement (sans semicolon) to output to the stdio stream
6402 @var{stream} the assembler definition of a local-common-label named
6403 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6404 is the size rounded up to whatever alignment the caller wants.
6406 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6407 output the name itself; before and after that, output the additional
6408 assembler syntax for defining the name, and a newline.
6410 This macro controls how the assembler definitions of uninitialized
6411 static variables are output.
6413 @findex ASM_OUTPUT_ALIGNED_LOCAL
6414 @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
6415 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
6416 separate, explicit argument. If you define this macro, it is used in
6417 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
6418 handling the required alignment of the variable. The alignment is specified
6419 as the number of bits.
6421 @findex ASM_OUTPUT_ALIGNED_DECL_LOCAL
6422 @item ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6423 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
6424 variable to be output, if there is one, or @code{NULL_TREE} if there
6425 is no corresponding variable. If you define this macro, GCC will use it
6426 in place of both @code{ASM_OUTPUT_DECL} and
6427 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
6428 the variable's decl in order to chose what to output.
6430 @findex ASM_OUTPUT_SHARED_LOCAL
6431 @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6432 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
6433 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
6438 @subsection Output and Generation of Labels
6440 @c prevent bad page break with this line
6441 This is about outputting labels.
6444 @findex ASM_OUTPUT_LABEL
6445 @findex assemble_name
6446 @item ASM_OUTPUT_LABEL (@var{stream}, @var{name})
6447 A C statement (sans semicolon) to output to the stdio stream
6448 @var{stream} the assembler definition of a label named @var{name}.
6449 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6450 output the name itself; before and after that, output the additional
6451 assembler syntax for defining the name, and a newline. A default
6452 definition of this macro is provided which is correct for most systems.
6456 A C string containing the appropriate assembler directive to specify the
6457 size of a symbol, without any arguments. On systems that use ELF, the
6458 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
6459 systems, the default is not to define this macro.
6461 Define this macro only if it is correct to use the default definitions
6462 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
6463 for your system. If you need your own custom definitions of those
6464 macros, or if you do not need explicit symbol sizes at all, do not
6467 @findex ASM_OUTPUT_SIZE_DIRECTIVE
6468 @item ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
6469 A C statement (sans semicolon) to output to the stdio stream
6470 @var{stream} a directive telling the assembler that the size of the
6471 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
6472 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6475 @findex ASM_OUTPUT_MEASURED_SIZE
6476 @item ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
6477 A C statement (sans semicolon) to output to the stdio stream
6478 @var{stream} a directive telling the assembler to calculate the size of
6479 the symbol @var{name} by subtracting its address from the current
6482 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6483 provided. The default assumes that the assembler recognizes a special
6484 @samp{.} symbol as referring to the current address, and can calculate
6485 the difference between this and another symbol. If your assembler does
6486 not recognize @samp{.} or cannot do calculations with it, you will need
6487 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
6491 A C string containing the appropriate assembler directive to specify the
6492 type of a symbol, without any arguments. On systems that use ELF, the
6493 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
6494 systems, the default is not to define this macro.
6496 Define this macro only if it is correct to use the default definition of
6497 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6498 custom definition of this macro, or if you do not need explicit symbol
6499 types at all, do not define this macro.
6501 @findex TYPE_OPERAND_FMT
6502 @item TYPE_OPERAND_FMT
6503 A C string which specifies (using @code{printf} syntax) the format of
6504 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
6505 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
6506 the default is not to define this macro.
6508 Define this macro only if it is correct to use the default definition of
6509 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6510 custom definition of this macro, or if you do not need explicit symbol
6511 types at all, do not define this macro.
6513 @findex ASM_OUTPUT_TYPE_DIRECTIVE
6514 @item ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
6515 A C statement (sans semicolon) to output to the stdio stream
6516 @var{stream} a directive telling the assembler that the type of the
6517 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
6518 that string is always either @samp{"function"} or @samp{"object"}, but
6519 you should not count on this.
6521 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
6522 definition of this macro is provided.
6524 @findex ASM_DECLARE_FUNCTION_NAME
6525 @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
6526 A C statement (sans semicolon) to output to the stdio stream
6527 @var{stream} any text necessary for declaring the name @var{name} of a
6528 function which is being defined. This macro is responsible for
6529 outputting the label definition (perhaps using
6530 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
6531 @code{FUNCTION_DECL} tree node representing the function.
6533 If this macro is not defined, then the function name is defined in the
6534 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6536 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6539 @findex ASM_DECLARE_FUNCTION_SIZE
6540 @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
6541 A C statement (sans semicolon) to output to the stdio stream
6542 @var{stream} any text necessary for declaring the size of a function
6543 which is being defined. The argument @var{name} is the name of the
6544 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
6545 representing the function.
6547 If this macro is not defined, then the function size is not defined.
6549 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
6552 @findex ASM_DECLARE_OBJECT_NAME
6553 @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6554 A C statement (sans semicolon) to output to the stdio stream
6555 @var{stream} any text necessary for declaring the name @var{name} of an
6556 initialized variable which is being defined. This macro must output the
6557 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6558 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6560 If this macro is not defined, then the variable name is defined in the
6561 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6563 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
6564 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
6566 @findex ASM_DECLARE_REGISTER_GLOBAL
6567 @item ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6568 A C statement (sans semicolon) to output to the stdio stream
6569 @var{stream} any text necessary for claiming a register @var{regno}
6570 for a global variable @var{decl} with name @var{name}.
6572 If you don't define this macro, that is equivalent to defining it to do
6575 @findex ASM_FINISH_DECLARE_OBJECT
6576 @item ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6577 A C statement (sans semicolon) to finish up declaring a variable name
6578 once the compiler has processed its initializer fully and thus has had a
6579 chance to determine the size of an array when controlled by an
6580 initializer. This is used on systems where it's necessary to declare
6581 something about the size of the object.
6583 If you don't define this macro, that is equivalent to defining it to do
6586 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
6587 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
6590 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
6591 This target hook is a function to output to the stdio stream
6592 @var{stream} some commands that will make the label @var{name} global;
6593 that is, available for reference from other files.
6595 The default implementation relies on a proper definition of
6596 @code{GLOBAL_ASM_OP}.
6600 @findex ASM_WEAKEN_LABEL
6601 @item ASM_WEAKEN_LABEL (@var{stream}, @var{name})
6602 A C statement (sans semicolon) to output to the stdio stream
6603 @var{stream} some commands that will make the label @var{name} weak;
6604 that is, available for reference from other files but only used if
6605 no other definition is available. Use the expression
6606 @code{assemble_name (@var{stream}, @var{name})} to output the name
6607 itself; before and after that, output the additional assembler syntax
6608 for making that name weak, and a newline.
6610 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
6611 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
6614 @findex ASM_WEAKEN_DECL
6615 @item ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
6616 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
6617 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
6618 or variable decl. If @var{value} is not @code{NULL}, this C statement
6619 should output to the stdio stream @var{stream} assembler code which
6620 defines (equates) the weak symbol @var{name} to have the value
6621 @var{value}. If @var{value} is @code{NULL}, it should output commands
6622 to make @var{name} weak.
6624 @findex SUPPORTS_WEAK
6626 A C expression which evaluates to true if the target supports weak symbols.
6628 If you don't define this macro, @file{defaults.h} provides a default
6629 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
6630 is defined, the default definition is @samp{1}; otherwise, it is
6631 @samp{0}. Define this macro if you want to control weak symbol support
6632 with a compiler flag such as @option{-melf}.
6634 @findex MAKE_DECL_ONE_ONLY (@var{decl})
6635 @item MAKE_DECL_ONE_ONLY
6636 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
6637 public symbol such that extra copies in multiple translation units will
6638 be discarded by the linker. Define this macro if your object file
6639 format provides support for this concept, such as the @samp{COMDAT}
6640 section flags in the Microsoft Windows PE/COFF format, and this support
6641 requires changes to @var{decl}, such as putting it in a separate section.
6643 @findex SUPPORTS_ONE_ONLY
6644 @item SUPPORTS_ONE_ONLY
6645 A C expression which evaluates to true if the target supports one-only
6648 If you don't define this macro, @file{varasm.c} provides a default
6649 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6650 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6651 you want to control one-only symbol support with a compiler flag, or if
6652 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6653 be emitted as one-only.
6655 @findex ASM_OUTPUT_EXTERNAL
6656 @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6657 A C statement (sans semicolon) to output to the stdio stream
6658 @var{stream} any text necessary for declaring the name of an external
6659 symbol named @var{name} which is referenced in this compilation but
6660 not defined. The value of @var{decl} is the tree node for the
6663 This macro need not be defined if it does not need to output anything.
6664 The GNU assembler and most Unix assemblers don't require anything.
6666 @findex ASM_OUTPUT_EXTERNAL_LIBCALL
6667 @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref})
6668 A C statement (sans semicolon) to output on @var{stream} an assembler
6669 pseudo-op to declare a library function name external. The name of the
6670 library function is given by @var{symref}, which has type @code{rtx} and
6671 is a @code{symbol_ref}.
6673 This macro need not be defined if it does not need to output anything.
6674 The GNU assembler and most Unix assemblers don't require anything.
6676 @findex ASM_OUTPUT_LABELREF
6677 @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6678 A C statement (sans semicolon) to output to the stdio stream
6679 @var{stream} a reference in assembler syntax to a label named
6680 @var{name}. This should add @samp{_} to the front of the name, if that
6681 is customary on your operating system, as it is in most Berkeley Unix
6682 systems. This macro is used in @code{assemble_name}.
6684 @findex ASM_OUTPUT_SYMBOL_REF
6685 @item ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6686 A C statement (sans semicolon) to output a reference to
6687 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
6688 will be used to output the name of the symbol. This macro may be used
6689 to modify the way a symbol is referenced depending on information
6690 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
6692 @findex ASM_OUTPUT_LABEL_REF
6693 @item ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
6694 A C statement (sans semicolon) to output a reference to @var{buf}, the
6695 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
6696 @code{assemble_name} will be used to output the name of the symbol.
6697 This macro is not used by @code{output_asm_label}, or the @code{%l}
6698 specifier that calls it; the intention is that this macro should be set
6699 when it is necessary to output a label differently when its address is
6702 @findex ASM_OUTPUT_INTERNAL_LABEL
6703 @item ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{prefix}, @var{num})
6704 A C statement to output to the stdio stream @var{stream} a label whose
6705 name is made from the string @var{prefix} and the number @var{num}.
6707 It is absolutely essential that these labels be distinct from the labels
6708 used for user-level functions and variables. Otherwise, certain programs
6709 will have name conflicts with internal labels.
6711 It is desirable to exclude internal labels from the symbol table of the
6712 object file. Most assemblers have a naming convention for labels that
6713 should be excluded; on many systems, the letter @samp{L} at the
6714 beginning of a label has this effect. You should find out what
6715 convention your system uses, and follow it.
6717 The usual definition of this macro is as follows:
6720 fprintf (@var{stream}, "L%s%d:\n", @var{prefix}, @var{num})
6723 @findex ASM_OUTPUT_DEBUG_LABEL
6724 @item ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
6725 A C statement to output to the stdio stream @var{stream} a debug info
6726 label whose name is made from the string @var{prefix} and the number
6727 @var{num}. This is useful for VLIW targets, where debug info labels
6728 may need to be treated differently than branch target labels. On some
6729 systems, branch target labels must be at the beginning of instruction
6730 bundles, but debug info labels can occur in the middle of instruction
6733 If this macro is not defined, then @code{ASM_OUTPUT_INTERNAL_LABEL} will be
6736 @findex ASM_GENERATE_INTERNAL_LABEL
6737 @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
6738 A C statement to store into the string @var{string} a label whose name
6739 is made from the string @var{prefix} and the number @var{num}.
6741 This string, when output subsequently by @code{assemble_name}, should
6742 produce the output that @code{ASM_OUTPUT_INTERNAL_LABEL} would produce
6743 with the same @var{prefix} and @var{num}.
6745 If the string begins with @samp{*}, then @code{assemble_name} will
6746 output the rest of the string unchanged. It is often convenient for
6747 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
6748 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
6749 to output the string, and may change it. (Of course,
6750 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
6751 you should know what it does on your machine.)
6753 @findex ASM_FORMAT_PRIVATE_NAME
6754 @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
6755 A C expression to assign to @var{outvar} (which is a variable of type
6756 @code{char *}) a newly allocated string made from the string
6757 @var{name} and the number @var{number}, with some suitable punctuation
6758 added. Use @code{alloca} to get space for the string.
6760 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
6761 produce an assembler label for an internal static variable whose name is
6762 @var{name}. Therefore, the string must be such as to result in valid
6763 assembler code. The argument @var{number} is different each time this
6764 macro is executed; it prevents conflicts between similarly-named
6765 internal static variables in different scopes.
6767 Ideally this string should not be a valid C identifier, to prevent any
6768 conflict with the user's own symbols. Most assemblers allow periods
6769 or percent signs in assembler symbols; putting at least one of these
6770 between the name and the number will suffice.
6772 @findex ASM_OUTPUT_DEF
6773 @item ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
6774 A C statement to output to the stdio stream @var{stream} assembler code
6775 which defines (equates) the symbol @var{name} to have the value @var{value}.
6778 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6779 correct for most systems.
6781 @findex ASM_OUTPUT_DEF_FROM_DECLS
6782 @item ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
6783 A C statement to output to the stdio stream @var{stream} assembler code
6784 which defines (equates) the symbol whose tree node is @var{decl_of_name}
6785 to have the value of the tree node @var{decl_of_value}. This macro will
6786 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
6787 the tree nodes are available.
6790 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6791 correct for most systems.
6793 @findex ASM_OUTPUT_WEAK_ALIAS
6794 @item ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
6795 A C statement to output to the stdio stream @var{stream} assembler code
6796 which defines (equates) the weak symbol @var{name} to have the value
6797 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
6798 an undefined weak symbol.
6800 Define this macro if the target only supports weak aliases; define
6801 @code{ASM_OUTPUT_DEF} instead if possible.
6803 @findex OBJC_GEN_METHOD_LABEL
6804 @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
6805 Define this macro to override the default assembler names used for
6806 Objective-C methods.
6808 The default name is a unique method number followed by the name of the
6809 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
6810 the category is also included in the assembler name (e.g.@:
6813 These names are safe on most systems, but make debugging difficult since
6814 the method's selector is not present in the name. Therefore, particular
6815 systems define other ways of computing names.
6817 @var{buf} is an expression of type @code{char *} which gives you a
6818 buffer in which to store the name; its length is as long as
6819 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
6820 50 characters extra.
6822 The argument @var{is_inst} specifies whether the method is an instance
6823 method or a class method; @var{class_name} is the name of the class;
6824 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
6825 in a category); and @var{sel_name} is the name of the selector.
6827 On systems where the assembler can handle quoted names, you can use this
6828 macro to provide more human-readable names.
6830 @findex ASM_DECLARE_CLASS_REFERENCE
6831 @item ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
6832 A C statement (sans semicolon) to output to the stdio stream
6833 @var{stream} commands to declare that the label @var{name} is an
6834 Objective-C class reference. This is only needed for targets whose
6835 linkers have special support for NeXT-style runtimes.
6837 @findex ASM_DECLARE_UNRESOLVED_REFERENCE
6838 @item ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
6839 A C statement (sans semicolon) to output to the stdio stream
6840 @var{stream} commands to declare that the label @var{name} is an
6841 unresolved Objective-C class reference. This is only needed for targets
6842 whose linkers have special support for NeXT-style runtimes.
6845 @node Initialization
6846 @subsection How Initialization Functions Are Handled
6847 @cindex initialization routines
6848 @cindex termination routines
6849 @cindex constructors, output of
6850 @cindex destructors, output of
6852 The compiled code for certain languages includes @dfn{constructors}
6853 (also called @dfn{initialization routines})---functions to initialize
6854 data in the program when the program is started. These functions need
6855 to be called before the program is ``started''---that is to say, before
6856 @code{main} is called.
6858 Compiling some languages generates @dfn{destructors} (also called
6859 @dfn{termination routines}) that should be called when the program
6862 To make the initialization and termination functions work, the compiler
6863 must output something in the assembler code to cause those functions to
6864 be called at the appropriate time. When you port the compiler to a new
6865 system, you need to specify how to do this.
6867 There are two major ways that GCC currently supports the execution of
6868 initialization and termination functions. Each way has two variants.
6869 Much of the structure is common to all four variations.
6871 @findex __CTOR_LIST__
6872 @findex __DTOR_LIST__
6873 The linker must build two lists of these functions---a list of
6874 initialization functions, called @code{__CTOR_LIST__}, and a list of
6875 termination functions, called @code{__DTOR_LIST__}.
6877 Each list always begins with an ignored function pointer (which may hold
6878 0, @minus{}1, or a count of the function pointers after it, depending on
6879 the environment). This is followed by a series of zero or more function
6880 pointers to constructors (or destructors), followed by a function
6881 pointer containing zero.
6883 Depending on the operating system and its executable file format, either
6884 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
6885 time and exit time. Constructors are called in reverse order of the
6886 list; destructors in forward order.
6888 The best way to handle static constructors works only for object file
6889 formats which provide arbitrarily-named sections. A section is set
6890 aside for a list of constructors, and another for a list of destructors.
6891 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
6892 object file that defines an initialization function also puts a word in
6893 the constructor section to point to that function. The linker
6894 accumulates all these words into one contiguous @samp{.ctors} section.
6895 Termination functions are handled similarly.
6897 This method will be chosen as the default by @file{target-def.h} if
6898 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
6899 support arbitrary sections, but does support special designated
6900 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
6901 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
6903 When arbitrary sections are available, there are two variants, depending
6904 upon how the code in @file{crtstuff.c} is called. On systems that
6905 support a @dfn{.init} section which is executed at program startup,
6906 parts of @file{crtstuff.c} are compiled into that section. The
6907 program is linked by the @code{gcc} driver like this:
6910 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
6913 The prologue of a function (@code{__init}) appears in the @code{.init}
6914 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
6915 for the function @code{__fini} in the @dfn{.fini} section. Normally these
6916 files are provided by the operating system or by the GNU C library, but
6917 are provided by GCC for a few targets.
6919 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
6920 compiled from @file{crtstuff.c}. They contain, among other things, code
6921 fragments within the @code{.init} and @code{.fini} sections that branch
6922 to routines in the @code{.text} section. The linker will pull all parts
6923 of a section together, which results in a complete @code{__init} function
6924 that invokes the routines we need at startup.
6926 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
6929 If no init section is available, when GCC compiles any function called
6930 @code{main} (or more accurately, any function designated as a program
6931 entry point by the language front end calling @code{expand_main_function}),
6932 it inserts a procedure call to @code{__main} as the first executable code
6933 after the function prologue. The @code{__main} function is defined
6934 in @file{libgcc2.c} and runs the global constructors.
6936 In file formats that don't support arbitrary sections, there are again
6937 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
6938 and an `a.out' format must be used. In this case,
6939 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
6940 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
6941 and with the address of the void function containing the initialization
6942 code as its value. The GNU linker recognizes this as a request to add
6943 the value to a @dfn{set}; the values are accumulated, and are eventually
6944 placed in the executable as a vector in the format described above, with
6945 a leading (ignored) count and a trailing zero element.
6946 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
6947 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
6948 the compilation of @code{main} to call @code{__main} as above, starting
6949 the initialization process.
6951 The last variant uses neither arbitrary sections nor the GNU linker.
6952 This is preferable when you want to do dynamic linking and when using
6953 file formats which the GNU linker does not support, such as `ECOFF'@. In
6954 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
6955 termination functions are recognized simply by their names. This requires
6956 an extra program in the linkage step, called @command{collect2}. This program
6957 pretends to be the linker, for use with GCC; it does its job by running
6958 the ordinary linker, but also arranges to include the vectors of
6959 initialization and termination functions. These functions are called
6960 via @code{__main} as described above. In order to use this method,
6961 @code{use_collect2} must be defined in the target in @file{config.gcc}.
6964 The following section describes the specific macros that control and
6965 customize the handling of initialization and termination functions.
6968 @node Macros for Initialization
6969 @subsection Macros Controlling Initialization Routines
6971 Here are the macros that control how the compiler handles initialization
6972 and termination functions:
6975 @findex INIT_SECTION_ASM_OP
6976 @item INIT_SECTION_ASM_OP
6977 If defined, a C string constant, including spacing, for the assembler
6978 operation to identify the following data as initialization code. If not
6979 defined, GCC will assume such a section does not exist. When you are
6980 using special sections for initialization and termination functions, this
6981 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
6982 run the initialization functions.
6984 @item HAS_INIT_SECTION
6985 @findex HAS_INIT_SECTION
6986 If defined, @code{main} will not call @code{__main} as described above.
6987 This macro should be defined for systems that control start-up code
6988 on a symbol-by-symbol basis, such as OSF/1, and should not
6989 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
6991 @item LD_INIT_SWITCH
6992 @findex LD_INIT_SWITCH
6993 If defined, a C string constant for a switch that tells the linker that
6994 the following symbol is an initialization routine.
6996 @item LD_FINI_SWITCH
6997 @findex LD_FINI_SWITCH
6998 If defined, a C string constant for a switch that tells the linker that
6999 the following symbol is a finalization routine.
7001 @item COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7002 If defined, a C statement that will write a function that can be
7003 automatically called when a shared library is loaded. The function
7004 should call @var{func}, which takes no arguments. If not defined, and
7005 the object format requires an explicit initialization function, then a
7006 function called @code{_GLOBAL__DI} will be generated.
7008 This function and the following one are used by collect2 when linking a
7009 shared library that needs constructors or destructors, or has DWARF2
7010 exception tables embedded in the code.
7012 @item COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7013 If defined, a C statement that will write a function that can be
7014 automatically called when a shared library is unloaded. The function
7015 should call @var{func}, which takes no arguments. If not defined, and
7016 the object format requires an explicit finalization function, then a
7017 function called @code{_GLOBAL__DD} will be generated.
7020 @findex INVOKE__main
7021 If defined, @code{main} will call @code{__main} despite the presence of
7022 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7023 where the init section is not actually run automatically, but is still
7024 useful for collecting the lists of constructors and destructors.
7026 @item SUPPORTS_INIT_PRIORITY
7027 @findex SUPPORTS_INIT_PRIORITY
7028 If nonzero, the C++ @code{init_priority} attribute is supported and the
7029 compiler should emit instructions to control the order of initialization
7030 of objects. If zero, the compiler will issue an error message upon
7031 encountering an @code{init_priority} attribute.
7034 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7035 This value is true if the target supports some ``native'' method of
7036 collecting constructors and destructors to be run at startup and exit.
7037 It is false if we must use @command{collect2}.
7040 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7041 If defined, a function that outputs assembler code to arrange to call
7042 the function referenced by @var{symbol} at initialization time.
7044 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7045 no arguments and with no return value. If the target supports initialization
7046 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7047 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7049 If this macro is not defined by the target, a suitable default will
7050 be chosen if (1) the target supports arbitrary section names, (2) the
7051 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7055 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7056 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7057 functions rather than initialization functions.
7060 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7061 generated for the generated object file will have static linkage.
7063 If your system uses @command{collect2} as the means of processing
7064 constructors, then that program normally uses @command{nm} to scan
7065 an object file for constructor functions to be called.
7067 On certain kinds of systems, you can define these macros to make
7068 @command{collect2} work faster (and, in some cases, make it work at all):
7071 @findex OBJECT_FORMAT_COFF
7072 @item OBJECT_FORMAT_COFF
7073 Define this macro if the system uses COFF (Common Object File Format)
7074 object files, so that @command{collect2} can assume this format and scan
7075 object files directly for dynamic constructor/destructor functions.
7077 @findex OBJECT_FORMAT_ROSE
7078 @item OBJECT_FORMAT_ROSE
7079 Define this macro if the system uses ROSE format object files, so that
7080 @command{collect2} can assume this format and scan object files directly
7081 for dynamic constructor/destructor functions.
7083 These macros are effective only in a native compiler; @command{collect2} as
7084 part of a cross compiler always uses @command{nm} for the target machine.
7086 @findex REAL_NM_FILE_NAME
7087 @item REAL_NM_FILE_NAME
7088 Define this macro as a C string constant containing the file name to use
7089 to execute @command{nm}. The default is to search the path normally for
7092 If your system supports shared libraries and has a program to list the
7093 dynamic dependencies of a given library or executable, you can define
7094 these macros to enable support for running initialization and
7095 termination functions in shared libraries:
7099 Define this macro to a C string constant containing the name of the program
7100 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7102 @findex PARSE_LDD_OUTPUT
7103 @item PARSE_LDD_OUTPUT (@var{ptr})
7104 Define this macro to be C code that extracts filenames from the output
7105 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7106 of type @code{char *} that points to the beginning of a line of output
7107 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7108 code must advance @var{ptr} to the beginning of the filename on that
7109 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7112 @node Instruction Output
7113 @subsection Output of Assembler Instructions
7115 @c prevent bad page break with this line
7116 This describes assembler instruction output.
7119 @findex REGISTER_NAMES
7120 @item REGISTER_NAMES
7121 A C initializer containing the assembler's names for the machine
7122 registers, each one as a C string constant. This is what translates
7123 register numbers in the compiler into assembler language.
7125 @findex ADDITIONAL_REGISTER_NAMES
7126 @item ADDITIONAL_REGISTER_NAMES
7127 If defined, a C initializer for an array of structures containing a name
7128 and a register number. This macro defines additional names for hard
7129 registers, thus allowing the @code{asm} option in declarations to refer
7130 to registers using alternate names.
7132 @findex ASM_OUTPUT_OPCODE
7133 @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7134 Define this macro if you are using an unusual assembler that
7135 requires different names for the machine instructions.
7137 The definition is a C statement or statements which output an
7138 assembler instruction opcode to the stdio stream @var{stream}. The
7139 macro-operand @var{ptr} is a variable of type @code{char *} which
7140 points to the opcode name in its ``internal'' form---the form that is
7141 written in the machine description. The definition should output the
7142 opcode name to @var{stream}, performing any translation you desire, and
7143 increment the variable @var{ptr} to point at the end of the opcode
7144 so that it will not be output twice.
7146 In fact, your macro definition may process less than the entire opcode
7147 name, or more than the opcode name; but if you want to process text
7148 that includes @samp{%}-sequences to substitute operands, you must take
7149 care of the substitution yourself. Just be sure to increment
7150 @var{ptr} over whatever text should not be output normally.
7152 @findex recog_data.operand
7153 If you need to look at the operand values, they can be found as the
7154 elements of @code{recog_data.operand}.
7156 If the macro definition does nothing, the instruction is output
7159 @findex FINAL_PRESCAN_INSN
7160 @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7161 If defined, a C statement to be executed just prior to the output of
7162 assembler code for @var{insn}, to modify the extracted operands so
7163 they will be output differently.
7165 Here the argument @var{opvec} is the vector containing the operands
7166 extracted from @var{insn}, and @var{noperands} is the number of
7167 elements of the vector which contain meaningful data for this insn.
7168 The contents of this vector are what will be used to convert the insn
7169 template into assembler code, so you can change the assembler output
7170 by changing the contents of the vector.
7172 This macro is useful when various assembler syntaxes share a single
7173 file of instruction patterns; by defining this macro differently, you
7174 can cause a large class of instructions to be output differently (such
7175 as with rearranged operands). Naturally, variations in assembler
7176 syntax affecting individual insn patterns ought to be handled by
7177 writing conditional output routines in those patterns.
7179 If this macro is not defined, it is equivalent to a null statement.
7181 @findex FINAL_PRESCAN_LABEL
7182 @item FINAL_PRESCAN_LABEL
7183 If defined, @code{FINAL_PRESCAN_INSN} will be called on each
7184 @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and
7185 @var{noperands} will be zero.
7187 @findex PRINT_OPERAND
7188 @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
7189 A C compound statement to output to stdio stream @var{stream} the
7190 assembler syntax for an instruction operand @var{x}. @var{x} is an
7193 @var{code} is a value that can be used to specify one of several ways
7194 of printing the operand. It is used when identical operands must be
7195 printed differently depending on the context. @var{code} comes from
7196 the @samp{%} specification that was used to request printing of the
7197 operand. If the specification was just @samp{%@var{digit}} then
7198 @var{code} is 0; if the specification was @samp{%@var{ltr}
7199 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
7202 If @var{x} is a register, this macro should print the register's name.
7203 The names can be found in an array @code{reg_names} whose type is
7204 @code{char *[]}. @code{reg_names} is initialized from
7205 @code{REGISTER_NAMES}.
7207 When the machine description has a specification @samp{%@var{punct}}
7208 (a @samp{%} followed by a punctuation character), this macro is called
7209 with a null pointer for @var{x} and the punctuation character for
7212 @findex PRINT_OPERAND_PUNCT_VALID_P
7213 @item PRINT_OPERAND_PUNCT_VALID_P (@var{code})
7214 A C expression which evaluates to true if @var{code} is a valid
7215 punctuation character for use in the @code{PRINT_OPERAND} macro. If
7216 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
7217 punctuation characters (except for the standard one, @samp{%}) are used
7220 @findex PRINT_OPERAND_ADDRESS
7221 @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
7222 A C compound statement to output to stdio stream @var{stream} the
7223 assembler syntax for an instruction operand that is a memory reference
7224 whose address is @var{x}. @var{x} is an RTL expression.
7226 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
7227 On some machines, the syntax for a symbolic address depends on the
7228 section that the address refers to. On these machines, define the hook
7229 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
7230 @code{symbol_ref}, and then check for it here. @xref{Assembler Format}.
7232 @findex DBR_OUTPUT_SEQEND
7233 @findex dbr_sequence_length
7234 @item DBR_OUTPUT_SEQEND(@var{file})
7235 A C statement, to be executed after all slot-filler instructions have
7236 been output. If necessary, call @code{dbr_sequence_length} to
7237 determine the number of slots filled in a sequence (zero if not
7238 currently outputting a sequence), to decide how many no-ops to output,
7241 Don't define this macro if it has nothing to do, but it is helpful in
7242 reading assembly output if the extent of the delay sequence is made
7243 explicit (e.g.@: with white space).
7245 @findex final_sequence
7246 Note that output routines for instructions with delay slots must be
7247 prepared to deal with not being output as part of a sequence
7248 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
7249 found.) The variable @code{final_sequence} is null when not
7250 processing a sequence, otherwise it contains the @code{sequence} rtx
7253 @findex REGISTER_PREFIX
7254 @findex LOCAL_LABEL_PREFIX
7255 @findex USER_LABEL_PREFIX
7256 @findex IMMEDIATE_PREFIX
7258 @item REGISTER_PREFIX
7259 @itemx LOCAL_LABEL_PREFIX
7260 @itemx USER_LABEL_PREFIX
7261 @itemx IMMEDIATE_PREFIX
7262 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
7263 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
7264 @file{final.c}). These are useful when a single @file{md} file must
7265 support multiple assembler formats. In that case, the various @file{tm.h}
7266 files can define these macros differently.
7268 @item ASM_FPRINTF_EXTENSIONS(@var{file}, @var{argptr}, @var{format})
7269 @findex ASM_FPRINTF_EXTENSIONS
7270 If defined this macro should expand to a series of @code{case}
7271 statements which will be parsed inside the @code{switch} statement of
7272 the @code{asm_fprintf} function. This allows targets to define extra
7273 printf formats which may useful when generating their assembler
7274 statements. Note that upper case letters are reserved for future
7275 generic extensions to asm_fprintf, and so are not available to target
7276 specific code. The output file is given by the parameter @var{file}.
7277 The varargs input pointer is @var{argptr} and the rest of the format
7278 string, starting the character after the one that is being switched
7279 upon, is pointed to by @var{format}.
7281 @findex ASSEMBLER_DIALECT
7282 @item ASSEMBLER_DIALECT
7283 If your target supports multiple dialects of assembler language (such as
7284 different opcodes), define this macro as a C expression that gives the
7285 numeric index of the assembler language dialect to use, with zero as the
7288 If this macro is defined, you may use constructs of the form
7290 @samp{@{option0|option1|option2@dots{}@}}
7293 in the output templates of patterns (@pxref{Output Template}) or in the
7294 first argument of @code{asm_fprintf}. This construct outputs
7295 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
7296 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
7297 within these strings retain their usual meaning. If there are fewer
7298 alternatives within the braces than the value of
7299 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
7301 If you do not define this macro, the characters @samp{@{}, @samp{|} and
7302 @samp{@}} do not have any special meaning when used in templates or
7303 operands to @code{asm_fprintf}.
7305 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
7306 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
7307 the variations in assembler language syntax with that mechanism. Define
7308 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
7309 if the syntax variant are larger and involve such things as different
7310 opcodes or operand order.
7312 @findex ASM_OUTPUT_REG_PUSH
7313 @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
7314 A C expression to output to @var{stream} some assembler code
7315 which will push hard register number @var{regno} onto the stack.
7316 The code need not be optimal, since this macro is used only when
7319 @findex ASM_OUTPUT_REG_POP
7320 @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
7321 A C expression to output to @var{stream} some assembler code
7322 which will pop hard register number @var{regno} off of the stack.
7323 The code need not be optimal, since this macro is used only when
7327 @node Dispatch Tables
7328 @subsection Output of Dispatch Tables
7330 @c prevent bad page break with this line
7331 This concerns dispatch tables.
7334 @cindex dispatch table
7335 @findex ASM_OUTPUT_ADDR_DIFF_ELT
7336 @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
7337 A C statement to output to the stdio stream @var{stream} an assembler
7338 pseudo-instruction to generate a difference between two labels.
7339 @var{value} and @var{rel} are the numbers of two internal labels. The
7340 definitions of these labels are output using
7341 @code{ASM_OUTPUT_INTERNAL_LABEL}, and they must be printed in the same
7342 way here. For example,
7345 fprintf (@var{stream}, "\t.word L%d-L%d\n",
7346 @var{value}, @var{rel})
7349 You must provide this macro on machines where the addresses in a
7350 dispatch table are relative to the table's own address. If defined, GCC
7351 will also use this macro on all machines when producing PIC@.
7352 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
7353 mode and flags can be read.
7355 @findex ASM_OUTPUT_ADDR_VEC_ELT
7356 @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
7357 This macro should be provided on machines where the addresses
7358 in a dispatch table are absolute.
7360 The definition should be a C statement to output to the stdio stream
7361 @var{stream} an assembler pseudo-instruction to generate a reference to
7362 a label. @var{value} is the number of an internal label whose
7363 definition is output using @code{ASM_OUTPUT_INTERNAL_LABEL}.
7367 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
7370 @findex ASM_OUTPUT_CASE_LABEL
7371 @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
7372 Define this if the label before a jump-table needs to be output
7373 specially. The first three arguments are the same as for
7374 @code{ASM_OUTPUT_INTERNAL_LABEL}; the fourth argument is the
7375 jump-table which follows (a @code{jump_insn} containing an
7376 @code{addr_vec} or @code{addr_diff_vec}).
7378 This feature is used on system V to output a @code{swbeg} statement
7381 If this macro is not defined, these labels are output with
7382 @code{ASM_OUTPUT_INTERNAL_LABEL}.
7384 @findex ASM_OUTPUT_CASE_END
7385 @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
7386 Define this if something special must be output at the end of a
7387 jump-table. The definition should be a C statement to be executed
7388 after the assembler code for the table is written. It should write
7389 the appropriate code to stdio stream @var{stream}. The argument
7390 @var{table} is the jump-table insn, and @var{num} is the label-number
7391 of the preceding label.
7393 If this macro is not defined, nothing special is output at the end of
7397 @node Exception Region Output
7398 @subsection Assembler Commands for Exception Regions
7400 @c prevent bad page break with this line
7402 This describes commands marking the start and the end of an exception
7406 @findex EH_FRAME_SECTION_NAME
7407 @item EH_FRAME_SECTION_NAME
7408 If defined, a C string constant for the name of the section containing
7409 exception handling frame unwind information. If not defined, GCC will
7410 provide a default definition if the target supports named sections.
7411 @file{crtstuff.c} uses this macro to switch to the appropriate section.
7413 You should define this symbol if your target supports DWARF 2 frame
7414 unwind information and the default definition does not work.
7416 @findex EH_FRAME_IN_DATA_SECTION
7417 @item EH_FRAME_IN_DATA_SECTION
7418 If defined, DWARF 2 frame unwind information will be placed in the
7419 data section even though the target supports named sections. This
7420 might be necessary, for instance, if the system linker does garbage
7421 collection and sections cannot be marked as not to be collected.
7423 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
7426 @findex MASK_RETURN_ADDR
7427 @item MASK_RETURN_ADDR
7428 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
7429 that it does not contain any extraneous set bits in it.
7431 @findex DWARF2_UNWIND_INFO
7432 @item DWARF2_UNWIND_INFO
7433 Define this macro to 0 if your target supports DWARF 2 frame unwind
7434 information, but it does not yet work with exception handling.
7435 Otherwise, if your target supports this information (if it defines
7436 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
7437 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
7440 If this macro is defined to 1, the DWARF 2 unwinder will be the default
7441 exception handling mechanism; otherwise, @code{setjmp}/@code{longjmp} will be used by
7444 If this macro is defined to anything, the DWARF 2 unwinder will be used
7445 instead of inline unwinders and @code{__unwind_function} in the non-@code{setjmp} case.
7447 @findex DWARF_CIE_DATA_ALIGNMENT
7448 @item DWARF_CIE_DATA_ALIGNMENT
7449 This macro need only be defined if the target might save registers in the
7450 function prologue at an offset to the stack pointer that is not aligned to
7451 @code{UNITS_PER_WORD}. The definition should be the negative minimum
7452 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
7453 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
7454 the target supports DWARF 2 frame unwind information.
7458 @deftypefn {Target Hook} void TARGET_ASM_EXCEPTION_SECTION ()
7459 If defined, a function that switches to the section in which the main
7460 exception table is to be placed (@pxref{Sections}). The default is a
7461 function that switches to a section named @code{.gcc_except_table} on
7462 machines that support named sections via
7463 @code{TARGET_ASM_NAMED_SECTION}, otherwise if @option{-fpic} or
7464 @option{-fPIC} is in effect, the @code{data_section}, otherwise the
7465 @code{readonly_data_section}.
7468 @deftypefn {Target Hook} void TARGET_ASM_EH_FRAME_SECTION ()
7469 If defined, a function that switches to the section in which the DWARF 2
7470 frame unwind information to be placed (@pxref{Sections}). The default
7471 is a function that outputs a standard GAS section directive, if
7472 @code{EH_FRAME_SECTION_NAME} is defined, or else a data section
7473 directive followed by a synthetic label.
7476 @node Alignment Output
7477 @subsection Assembler Commands for Alignment
7479 @c prevent bad page break with this line
7480 This describes commands for alignment.
7484 @item JUMP_ALIGN (@var{label})
7485 The alignment (log base 2) to put in front of @var{label}, which is
7486 a common destination of jumps and has no fallthru incoming edge.
7488 This macro need not be defined if you don't want any special alignment
7489 to be done at such a time. Most machine descriptions do not currently
7492 Unless it's necessary to inspect the @var{label} parameter, it is better
7493 to set the variable @var{align_jumps} in the target's
7494 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7495 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
7497 @findex LABEL_ALIGN_AFTER_BARRIER
7498 @item LABEL_ALIGN_AFTER_BARRIER (@var{label})
7499 The alignment (log base 2) to put in front of @var{label}, which follows
7502 This macro need not be defined if you don't want any special alignment
7503 to be done at such a time. Most machine descriptions do not currently
7506 @findex LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7507 @item LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7508 The maximum number of bytes to skip when applying
7509 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
7510 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7513 @item LOOP_ALIGN (@var{label})
7514 The alignment (log base 2) to put in front of @var{label}, which follows
7515 a @code{NOTE_INSN_LOOP_BEG} note.
7517 This macro need not be defined if you don't want any special alignment
7518 to be done at such a time. Most machine descriptions do not currently
7521 Unless it's necessary to inspect the @var{label} parameter, it is better
7522 to set the variable @code{align_loops} in the target's
7523 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7524 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
7526 @findex LOOP_ALIGN_MAX_SKIP
7527 @item LOOP_ALIGN_MAX_SKIP
7528 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
7529 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7532 @item LABEL_ALIGN (@var{label})
7533 The alignment (log base 2) to put in front of @var{label}.
7534 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
7535 the maximum of the specified values is used.
7537 Unless it's necessary to inspect the @var{label} parameter, it is better
7538 to set the variable @code{align_labels} in the target's
7539 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7540 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
7542 @findex LABEL_ALIGN_MAX_SKIP
7543 @item LABEL_ALIGN_MAX_SKIP
7544 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
7545 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7547 @findex ASM_OUTPUT_SKIP
7548 @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
7549 A C statement to output to the stdio stream @var{stream} an assembler
7550 instruction to advance the location counter by @var{nbytes} bytes.
7551 Those bytes should be zero when loaded. @var{nbytes} will be a C
7552 expression of type @code{int}.
7554 @findex ASM_NO_SKIP_IN_TEXT
7555 @item ASM_NO_SKIP_IN_TEXT
7556 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
7557 text section because it fails to put zeros in the bytes that are skipped.
7558 This is true on many Unix systems, where the pseudo--op to skip bytes
7559 produces no-op instructions rather than zeros when used in the text
7562 @findex ASM_OUTPUT_ALIGN
7563 @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
7564 A C statement to output to the stdio stream @var{stream} an assembler
7565 command to advance the location counter to a multiple of 2 to the
7566 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
7568 @findex ASM_OUTPUT_MAX_SKIP_ALIGN
7569 @item ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
7570 A C statement to output to the stdio stream @var{stream} an assembler
7571 command to advance the location counter to a multiple of 2 to the
7572 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
7573 satisfy the alignment request. @var{power} and @var{max_skip} will be
7574 a C expression of type @code{int}.
7578 @node Debugging Info
7579 @section Controlling Debugging Information Format
7581 @c prevent bad page break with this line
7582 This describes how to specify debugging information.
7585 * All Debuggers:: Macros that affect all debugging formats uniformly.
7586 * DBX Options:: Macros enabling specific options in DBX format.
7587 * DBX Hooks:: Hook macros for varying DBX format.
7588 * File Names and DBX:: Macros controlling output of file names in DBX format.
7589 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
7590 * VMS Debug:: Macros for VMS debug format.
7594 @subsection Macros Affecting All Debugging Formats
7596 @c prevent bad page break with this line
7597 These macros affect all debugging formats.
7600 @findex DBX_REGISTER_NUMBER
7601 @item DBX_REGISTER_NUMBER (@var{regno})
7602 A C expression that returns the DBX register number for the compiler
7603 register number @var{regno}. In the default macro provided, the value
7604 of this expression will be @var{regno} itself. But sometimes there are
7605 some registers that the compiler knows about and DBX does not, or vice
7606 versa. In such cases, some register may need to have one number in the
7607 compiler and another for DBX@.
7609 If two registers have consecutive numbers inside GCC, and they can be
7610 used as a pair to hold a multiword value, then they @emph{must} have
7611 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
7612 Otherwise, debuggers will be unable to access such a pair, because they
7613 expect register pairs to be consecutive in their own numbering scheme.
7615 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
7616 does not preserve register pairs, then what you must do instead is
7617 redefine the actual register numbering scheme.
7619 @findex DEBUGGER_AUTO_OFFSET
7620 @item DEBUGGER_AUTO_OFFSET (@var{x})
7621 A C expression that returns the integer offset value for an automatic
7622 variable having address @var{x} (an RTL expression). The default
7623 computation assumes that @var{x} is based on the frame-pointer and
7624 gives the offset from the frame-pointer. This is required for targets
7625 that produce debugging output for DBX or COFF-style debugging output
7626 for SDB and allow the frame-pointer to be eliminated when the
7627 @option{-g} options is used.
7629 @findex DEBUGGER_ARG_OFFSET
7630 @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
7631 A C expression that returns the integer offset value for an argument
7632 having address @var{x} (an RTL expression). The nominal offset is
7635 @findex PREFERRED_DEBUGGING_TYPE
7636 @item PREFERRED_DEBUGGING_TYPE
7637 A C expression that returns the type of debugging output GCC should
7638 produce when the user specifies just @option{-g}. Define
7639 this if you have arranged for GCC to support more than one format of
7640 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7641 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
7642 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
7644 When the user specifies @option{-ggdb}, GCC normally also uses the
7645 value of this macro to select the debugging output format, but with two
7646 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined and
7647 @code{LINKER_DOES_NOT_WORK_WITH_DWARF2} is not defined, GCC uses the
7648 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7649 defined, GCC uses @code{DBX_DEBUG}.
7651 The value of this macro only affects the default debugging output; the
7652 user can always get a specific type of output by using @option{-gstabs},
7653 @option{-gcoff}, @option{-gdwarf-1}, @option{-gdwarf-2}, @option{-gxcoff},
7658 @subsection Specific Options for DBX Output
7660 @c prevent bad page break with this line
7661 These are specific options for DBX output.
7664 @findex DBX_DEBUGGING_INFO
7665 @item DBX_DEBUGGING_INFO
7666 Define this macro if GCC should produce debugging output for DBX
7667 in response to the @option{-g} option.
7669 @findex XCOFF_DEBUGGING_INFO
7670 @item XCOFF_DEBUGGING_INFO
7671 Define this macro if GCC should produce XCOFF format debugging output
7672 in response to the @option{-g} option. This is a variant of DBX format.
7674 @findex DEFAULT_GDB_EXTENSIONS
7675 @item DEFAULT_GDB_EXTENSIONS
7676 Define this macro to control whether GCC should by default generate
7677 GDB's extended version of DBX debugging information (assuming DBX-format
7678 debugging information is enabled at all). If you don't define the
7679 macro, the default is 1: always generate the extended information
7680 if there is any occasion to.
7682 @findex DEBUG_SYMS_TEXT
7683 @item DEBUG_SYMS_TEXT
7684 Define this macro if all @code{.stabs} commands should be output while
7685 in the text section.
7687 @findex ASM_STABS_OP
7689 A C string constant, including spacing, naming the assembler pseudo op to
7690 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
7691 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
7692 applies only to DBX debugging information format.
7694 @findex ASM_STABD_OP
7696 A C string constant, including spacing, naming the assembler pseudo op to
7697 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
7698 value is the current location. If you don't define this macro,
7699 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
7702 @findex ASM_STABN_OP
7704 A C string constant, including spacing, naming the assembler pseudo op to
7705 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
7706 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
7707 macro applies only to DBX debugging information format.
7709 @findex DBX_NO_XREFS
7711 Define this macro if DBX on your system does not support the construct
7712 @samp{xs@var{tagname}}. On some systems, this construct is used to
7713 describe a forward reference to a structure named @var{tagname}.
7714 On other systems, this construct is not supported at all.
7716 @findex DBX_CONTIN_LENGTH
7717 @item DBX_CONTIN_LENGTH
7718 A symbol name in DBX-format debugging information is normally
7719 continued (split into two separate @code{.stabs} directives) when it
7720 exceeds a certain length (by default, 80 characters). On some
7721 operating systems, DBX requires this splitting; on others, splitting
7722 must not be done. You can inhibit splitting by defining this macro
7723 with the value zero. You can override the default splitting-length by
7724 defining this macro as an expression for the length you desire.
7726 @findex DBX_CONTIN_CHAR
7727 @item DBX_CONTIN_CHAR
7728 Normally continuation is indicated by adding a @samp{\} character to
7729 the end of a @code{.stabs} string when a continuation follows. To use
7730 a different character instead, define this macro as a character
7731 constant for the character you want to use. Do not define this macro
7732 if backslash is correct for your system.
7734 @findex DBX_STATIC_STAB_DATA_SECTION
7735 @item DBX_STATIC_STAB_DATA_SECTION
7736 Define this macro if it is necessary to go to the data section before
7737 outputting the @samp{.stabs} pseudo-op for a non-global static
7740 @findex DBX_TYPE_DECL_STABS_CODE
7741 @item DBX_TYPE_DECL_STABS_CODE
7742 The value to use in the ``code'' field of the @code{.stabs} directive
7743 for a typedef. The default is @code{N_LSYM}.
7745 @findex DBX_STATIC_CONST_VAR_CODE
7746 @item DBX_STATIC_CONST_VAR_CODE
7747 The value to use in the ``code'' field of the @code{.stabs} directive
7748 for a static variable located in the text section. DBX format does not
7749 provide any ``right'' way to do this. The default is @code{N_FUN}.
7751 @findex DBX_REGPARM_STABS_CODE
7752 @item DBX_REGPARM_STABS_CODE
7753 The value to use in the ``code'' field of the @code{.stabs} directive
7754 for a parameter passed in registers. DBX format does not provide any
7755 ``right'' way to do this. The default is @code{N_RSYM}.
7757 @findex DBX_REGPARM_STABS_LETTER
7758 @item DBX_REGPARM_STABS_LETTER
7759 The letter to use in DBX symbol data to identify a symbol as a parameter
7760 passed in registers. DBX format does not customarily provide any way to
7761 do this. The default is @code{'P'}.
7763 @findex DBX_MEMPARM_STABS_LETTER
7764 @item DBX_MEMPARM_STABS_LETTER
7765 The letter to use in DBX symbol data to identify a symbol as a stack
7766 parameter. The default is @code{'p'}.
7768 @findex DBX_FUNCTION_FIRST
7769 @item DBX_FUNCTION_FIRST
7770 Define this macro if the DBX information for a function and its
7771 arguments should precede the assembler code for the function. Normally,
7772 in DBX format, the debugging information entirely follows the assembler
7775 @findex DBX_LBRAC_FIRST
7776 @item DBX_LBRAC_FIRST
7777 Define this macro if the @code{N_LBRAC} symbol for a block should
7778 precede the debugging information for variables and functions defined in
7779 that block. Normally, in DBX format, the @code{N_LBRAC} symbol comes
7782 @findex DBX_BLOCKS_FUNCTION_RELATIVE
7783 @item DBX_BLOCKS_FUNCTION_RELATIVE
7784 Define this macro if the value of a symbol describing the scope of a
7785 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
7786 of the enclosing function. Normally, GCC uses an absolute address.
7788 @findex DBX_USE_BINCL
7790 Define this macro if GCC should generate @code{N_BINCL} and
7791 @code{N_EINCL} stabs for included header files, as on Sun systems. This
7792 macro also directs GCC to output a type number as a pair of a file
7793 number and a type number within the file. Normally, GCC does not
7794 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
7795 number for a type number.
7799 @subsection Open-Ended Hooks for DBX Format
7801 @c prevent bad page break with this line
7802 These are hooks for DBX format.
7805 @findex DBX_OUTPUT_LBRAC
7806 @item DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
7807 Define this macro to say how to output to @var{stream} the debugging
7808 information for the start of a scope level for variable names. The
7809 argument @var{name} is the name of an assembler symbol (for use with
7810 @code{assemble_name}) whose value is the address where the scope begins.
7812 @findex DBX_OUTPUT_RBRAC
7813 @item DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
7814 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
7816 @findex DBX_OUTPUT_ENUM
7817 @item DBX_OUTPUT_ENUM (@var{stream}, @var{type})
7818 Define this macro if the target machine requires special handling to
7819 output an enumeration type. The definition should be a C statement
7820 (sans semicolon) to output the appropriate information to @var{stream}
7821 for the type @var{type}.
7823 @findex DBX_OUTPUT_FUNCTION_END
7824 @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
7825 Define this macro if the target machine requires special output at the
7826 end of the debugging information for a function. The definition should
7827 be a C statement (sans semicolon) to output the appropriate information
7828 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
7831 @findex DBX_OUTPUT_STANDARD_TYPES
7832 @item DBX_OUTPUT_STANDARD_TYPES (@var{syms})
7833 Define this macro if you need to control the order of output of the
7834 standard data types at the beginning of compilation. The argument
7835 @var{syms} is a @code{tree} which is a chain of all the predefined
7836 global symbols, including names of data types.
7838 Normally, DBX output starts with definitions of the types for integers
7839 and characters, followed by all the other predefined types of the
7840 particular language in no particular order.
7842 On some machines, it is necessary to output different particular types
7843 first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
7844 those symbols in the necessary order. Any predefined types that you
7845 don't explicitly output will be output afterward in no particular order.
7847 Be careful not to define this macro so that it works only for C@. There
7848 are no global variables to access most of the built-in types, because
7849 another language may have another set of types. The way to output a
7850 particular type is to look through @var{syms} to see if you can find it.
7856 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7857 if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
7859 dbxout_symbol (decl);
7865 This does nothing if the expected type does not exist.
7867 See the function @code{init_decl_processing} in @file{c-decl.c} to find
7868 the names to use for all the built-in C types.
7870 Here is another way of finding a particular type:
7872 @c this is still overfull. --mew 10feb93
7876 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7877 if (TREE_CODE (decl) == TYPE_DECL
7878 && (TREE_CODE (TREE_TYPE (decl))
7880 && TYPE_PRECISION (TREE_TYPE (decl)) == 16
7881 && TYPE_UNSIGNED (TREE_TYPE (decl)))
7883 /* @r{This must be @code{unsigned short}.} */
7884 dbxout_symbol (decl);
7890 @findex NO_DBX_FUNCTION_END
7891 @item NO_DBX_FUNCTION_END
7892 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
7893 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
7894 On those machines, define this macro to turn this feature off without
7895 disturbing the rest of the gdb extensions.
7899 @node File Names and DBX
7900 @subsection File Names in DBX Format
7902 @c prevent bad page break with this line
7903 This describes file names in DBX format.
7906 @findex DBX_WORKING_DIRECTORY
7907 @item DBX_WORKING_DIRECTORY
7908 Define this if DBX wants to have the current directory recorded in each
7911 Note that the working directory is always recorded if GDB extensions are
7914 @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME
7915 @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
7916 A C statement to output DBX debugging information to the stdio stream
7917 @var{stream} which indicates that file @var{name} is the main source
7918 file---the file specified as the input file for compilation.
7919 This macro is called only once, at the beginning of compilation.
7921 This macro need not be defined if the standard form of output
7922 for DBX debugging information is appropriate.
7924 @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY
7925 @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
7926 A C statement to output DBX debugging information to the stdio stream
7927 @var{stream} which indicates that the current directory during
7928 compilation is named @var{name}.
7930 This macro need not be defined if the standard form of output
7931 for DBX debugging information is appropriate.
7933 @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END
7934 @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
7935 A C statement to output DBX debugging information at the end of
7936 compilation of the main source file @var{name}.
7938 If you don't define this macro, nothing special is output at the end
7939 of compilation, which is correct for most machines.
7941 @findex DBX_OUTPUT_SOURCE_FILENAME
7942 @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7943 A C statement to output DBX debugging information to the stdio stream
7944 @var{stream} which indicates that file @var{name} is the current source
7945 file. This output is generated each time input shifts to a different
7946 source file as a result of @samp{#include}, the end of an included file,
7947 or a @samp{#line} command.
7949 This macro need not be defined if the standard form of output
7950 for DBX debugging information is appropriate.
7955 @subsection Macros for SDB and DWARF Output
7957 @c prevent bad page break with this line
7958 Here are macros for SDB and DWARF output.
7961 @findex SDB_DEBUGGING_INFO
7962 @item SDB_DEBUGGING_INFO
7963 Define this macro if GCC should produce COFF-style debugging output
7964 for SDB in response to the @option{-g} option.
7966 @findex DWARF_DEBUGGING_INFO
7967 @item DWARF_DEBUGGING_INFO
7968 Define this macro if GCC should produce dwarf format debugging output
7969 in response to the @option{-g} option.
7971 @findex DWARF2_DEBUGGING_INFO
7972 @item DWARF2_DEBUGGING_INFO
7973 Define this macro if GCC should produce dwarf version 2 format
7974 debugging output in response to the @option{-g} option.
7976 To support optional call frame debugging information, you must also
7977 define @code{INCOMING_RETURN_ADDR_RTX} and either set
7978 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
7979 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
7980 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
7982 @findex DWARF2_FRAME_INFO
7983 @item DWARF2_FRAME_INFO
7984 Define this macro to a nonzero value if GCC should always output
7985 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
7986 (@pxref{Exception Region Output} is nonzero, GCC will output this
7987 information not matter how you define @code{DWARF2_FRAME_INFO}.
7989 @findex LINKER_DOES_NOT_WORK_WITH_DWARF2
7990 @item LINKER_DOES_NOT_WORK_WITH_DWARF2
7991 Define this macro if the linker does not work with Dwarf version 2.
7992 Normally, if the user specifies only @option{-ggdb} GCC will use Dwarf
7993 version 2 if available; this macro disables this. See the description
7994 of the @code{PREFERRED_DEBUGGING_TYPE} macro for more details.
7996 @findex DWARF2_GENERATE_TEXT_SECTION_LABEL
7997 @item DWARF2_GENERATE_TEXT_SECTION_LABEL
7998 By default, the Dwarf 2 debugging information generator will generate a
7999 label to mark the beginning of the text section. If it is better simply
8000 to use the name of the text section itself, rather than an explicit label,
8001 to indicate the beginning of the text section, define this macro to zero.
8003 @findex DWARF2_ASM_LINE_DEBUG_INFO
8004 @item DWARF2_ASM_LINE_DEBUG_INFO
8005 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8006 line debug info sections. This will result in much more compact line number
8007 tables, and hence is desirable if it works.
8009 @findex PUT_SDB_@dots{}
8010 @item PUT_SDB_@dots{}
8011 Define these macros to override the assembler syntax for the special
8012 SDB assembler directives. See @file{sdbout.c} for a list of these
8013 macros and their arguments. If the standard syntax is used, you need
8014 not define them yourself.
8018 Some assemblers do not support a semicolon as a delimiter, even between
8019 SDB assembler directives. In that case, define this macro to be the
8020 delimiter to use (usually @samp{\n}). It is not necessary to define
8021 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8024 @findex SDB_GENERATE_FAKE
8025 @item SDB_GENERATE_FAKE
8026 Define this macro to override the usual method of constructing a dummy
8027 name for anonymous structure and union types. See @file{sdbout.c} for
8030 @findex SDB_ALLOW_UNKNOWN_REFERENCES
8031 @item SDB_ALLOW_UNKNOWN_REFERENCES
8032 Define this macro to allow references to unknown structure,
8033 union, or enumeration tags to be emitted. Standard COFF does not
8034 allow handling of unknown references, MIPS ECOFF has support for
8037 @findex SDB_ALLOW_FORWARD_REFERENCES
8038 @item SDB_ALLOW_FORWARD_REFERENCES
8039 Define this macro to allow references to structure, union, or
8040 enumeration tags that have not yet been seen to be handled. Some
8041 assemblers choke if forward tags are used, while some require it.
8046 @subsection Macros for VMS Debug Format
8048 @c prevent bad page break with this line
8049 Here are macros for VMS debug format.
8052 @findex VMS_DEBUGGING_INFO
8053 @item VMS_DEBUGGING_INFO
8054 Define this macro if GCC should produce debugging output for VMS
8055 in response to the @option{-g} option. The default behavior for VMS
8056 is to generate minimal debug info for a traceback in the absence of
8057 @option{-g} unless explicitly overridden with @option{-g0}. This
8058 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8059 @code{OVERRIDE_OPTIONS}.
8062 @node Floating Point
8063 @section Cross Compilation and Floating Point
8064 @cindex cross compilation and floating point
8065 @cindex floating point and cross compilation
8067 While all modern machines use twos-complement representation for integers,
8068 there are a variety of representations for floating point numbers. This
8069 means that in a cross-compiler the representation of floating point numbers
8070 in the compiled program may be different from that used in the machine
8071 doing the compilation.
8073 Because different representation systems may offer different amounts of
8074 range and precision, all floating point constants must be represented in
8075 the target machine's format. Therefore, the cross compiler cannot
8076 safely use the host machine's floating point arithmetic; it must emulate
8077 the target's arithmetic. To ensure consistency, GCC always uses
8078 emulation to work with floating point values, even when the host and
8079 target floating point formats are identical.
8081 The following macros are provided by @file{real.h} for the compiler to
8082 use. All parts of the compiler which generate or optimize
8083 floating-point calculations must use these macros. They may evaluate
8084 their operands more than once, so operands must not have side effects.
8086 @defmac REAL_VALUE_TYPE
8087 The C data type to be used to hold a floating point value in the target
8088 machine's format. Typically this is a @code{struct} containing an
8089 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8093 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8094 Compares for equality the two values, @var{x} and @var{y}. If the target
8095 floating point format supports negative zeroes and/or NaNs,
8096 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8097 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8100 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8101 Tests whether @var{x} is less than @var{y}.
8105 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_LDEXP (REAL_VALUE_TYPE @var{x}, int @var{scale})
8106 Multiplies @var{x} by 2 raised to the power @var{scale}.
8109 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8110 Truncates @var{x} to a signed integer, rounding toward zero.
8113 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8114 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8115 @var{x} is negative, returns zero.
8118 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_RNDZINT (REAL_VALUE_TYPE @var{x})
8119 Rounds the target-machine floating point value @var{x} towards zero to an
8120 integer value, but leaves it represented as a floating point number.
8123 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_UNSIGNED_RNDZINT (REAL_VALUE_TYPE @var{x})
8124 Rounds the target-machine floating point value @var{x} towards zero to an
8125 unsigned integer value, but leaves it represented as a floating point
8126 number. If @var{x} is negative, returns (positive) zero.
8129 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8130 Converts @var{string} into a floating point number in the target machine's
8131 representation for mode @var{mode}. This routine can handle both
8132 decimal and hexadecimal floating point constants, using the syntax
8133 defined by the C language for both.
8136 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8137 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8140 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8141 Determines whether @var{x} represents infinity (positive or negative).
8144 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8145 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8148 @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})
8149 Calculates an arithmetic operation on the two floating point values
8150 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8153 The operation to be performed is specified by @var{code}. Only the
8154 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8155 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8157 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8158 target's floating point format cannot represent infinity, it will call
8159 @code{abort}. Callers should check for this situation first, using
8160 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8163 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8164 Returns the negative of the floating point value @var{x}.
8167 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8168 Returns the absolute value of @var{x}.
8171 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8172 Truncates the floating point value @var{x} to fit in @var{mode}. The
8173 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8174 appropriate bit pattern to be output asa floating constant whose
8175 precision accords with mode @var{mode}.
8178 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8179 Converts a floating point value @var{x} into a double-precision integer
8180 which is then stored into @var{low} and @var{high}. If the value is not
8181 integral, it is truncated.
8184 @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})
8185 @findex REAL_VALUE_FROM_INT
8186 Converts a double-precision integer found in @var{low} and @var{high},
8187 into a floating point value which is then stored into @var{x}. The
8188 value is truncated to fit in mode @var{mode}.
8191 @node Mode Switching
8192 @section Mode Switching Instructions
8193 @cindex mode switching
8194 The following macros control mode switching optimizations:
8197 @findex OPTIMIZE_MODE_SWITCHING
8198 @item OPTIMIZE_MODE_SWITCHING (@var{entity})
8199 Define this macro if the port needs extra instructions inserted for mode
8200 switching in an optimizing compilation.
8202 For an example, the SH4 can perform both single and double precision
8203 floating point operations, but to perform a single precision operation,
8204 the FPSCR PR bit has to be cleared, while for a double precision
8205 operation, this bit has to be set. Changing the PR bit requires a general
8206 purpose register as a scratch register, hence these FPSCR sets have to
8207 be inserted before reload, i.e.@: you can't put this into instruction emitting
8208 or @code{MACHINE_DEPENDENT_REORG}.
8210 You can have multiple entities that are mode-switched, and select at run time
8211 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
8212 return nonzero for any @var{entity} that needs mode-switching.
8213 If you define this macro, you also have to define
8214 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
8215 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
8216 @code{NORMAL_MODE} is optional.
8218 @findex NUM_MODES_FOR_MODE_SWITCHING
8219 @item NUM_MODES_FOR_MODE_SWITCHING
8220 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
8221 initializer for an array of integers. Each initializer element
8222 N refers to an entity that needs mode switching, and specifies the number
8223 of different modes that might need to be set for this entity.
8224 The position of the initializer in the initializer - starting counting at
8225 zero - determines the integer that is used to refer to the mode-switched
8227 In macros that take mode arguments / yield a mode result, modes are
8228 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
8229 switch is needed / supplied.
8232 @item MODE_NEEDED (@var{entity}, @var{insn})
8233 @var{entity} is an integer specifying a mode-switched entity. If
8234 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
8235 return an integer value not larger than the corresponding element in
8236 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
8237 be switched into prior to the execution of @var{insn}.
8240 @item NORMAL_MODE (@var{entity})
8241 If this macro is defined, it is evaluated for every @var{entity} that needs
8242 mode switching. It should evaluate to an integer, which is a mode that
8243 @var{entity} is assumed to be switched to at function entry and exit.
8245 @findex MODE_PRIORITY_TO_MODE
8246 @item MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
8247 This macro specifies the order in which modes for @var{entity} are processed.
8248 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
8249 lowest. The value of the macro should be an integer designating a mode
8250 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
8251 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
8252 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
8254 @findex EMIT_MODE_SET
8255 @item EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
8256 Generate one or more insns to set @var{entity} to @var{mode}.
8257 @var{hard_reg_live} is the set of hard registers live at the point where
8258 the insn(s) are to be inserted.
8261 @node Target Attributes
8262 @section Defining target-specific uses of @code{__attribute__}
8263 @cindex target attributes
8264 @cindex machine attributes
8265 @cindex attributes, target-specific
8267 Target-specific attributes may be defined for functions, data and types.
8268 These are described using the following target hooks; they also need to
8269 be documented in @file{extend.texi}.
8271 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
8272 If defined, this target hook points to an array of @samp{struct
8273 attribute_spec} (defined in @file{tree.h}) specifying the machine
8274 specific attributes for this target and some of the restrictions on the
8275 entities to which these attributes are applied and the arguments they
8279 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8280 If defined, this target hook is a function which returns zero if the attributes on
8281 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8282 and two if they are nearly compatible (which causes a warning to be
8283 generated). If this is not defined, machine-specific attributes are
8284 supposed always to be compatible.
8287 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
8288 If defined, this target hook is a function which assigns default attributes to
8289 newly defined @var{type}.
8292 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8293 Define this target hook if the merging of type attributes needs special
8294 handling. If defined, the result is a list of the combined
8295 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
8296 that @code{comptypes} has already been called and returned 1. This
8297 function may call @code{merge_attributes} to handle machine-independent
8301 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
8302 Define this target hook if the merging of decl attributes needs special
8303 handling. If defined, the result is a list of the combined
8304 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
8305 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
8306 when this is needed are when one attribute overrides another, or when an
8307 attribute is nullified by a subsequent definition. This function may
8308 call @code{merge_attributes} to handle machine-independent merging.
8310 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
8311 If the only target-specific handling you require is @samp{dllimport} for
8312 Windows targets, you should define the macro
8313 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. This links in a function
8314 called @code{merge_dllimport_decl_attributes} which can then be defined
8315 as the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. This is done
8316 in @file{i386/cygwin.h} and @file{i386/i386.c}, for example.
8319 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
8320 Define this target hook if you want to be able to add attributes to a decl
8321 when it is being created. This is normally useful for back ends which
8322 wish to implement a pragma by using the attributes which correspond to
8323 the pragma's effect. The @var{node} argument is the decl which is being
8324 created. The @var{attr_ptr} argument is a pointer to the attribute list
8325 for this decl. The list itself should not be modified, since it may be
8326 shared with other decls, but attributes may be chained on the head of
8327 the list and @code{*@var{attr_ptr}} modified to point to the new
8328 attributes, or a copy of the list may be made if further changes are
8332 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
8334 This target hook returns @code{true} if it is ok to inline @var{fndecl}
8335 into the current function, despite its having target-specific
8336 attributes, @code{false} otherwise. By default, if a function has a
8337 target specific attribute attached to it, it will not be inlined.
8340 @node MIPS Coprocessors
8341 @section Defining coprocessor specifics for MIPS targets.
8342 @cindex MIPS coprocessor-definition macros
8344 The MIPS specification allows MIPS implementations to have as many as 4
8345 coprocessors, each with as many as 32 private registers. gcc supports
8346 accessing these registers and transferring values between the registers
8347 and memory using asm-ized variables. For example:
8350 register unsigned int cp0count asm ("c0r1");
8356 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
8357 names may be added as described below, or the default names may be
8358 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
8360 Coprocessor registers are assumed to be epilogue-used; sets to them will
8361 be preserved even if it does not appear that the register is used again
8362 later in the function.
8364 Another note: according to the MIPS spec, coprocessor 1 (if present) is
8365 the FPU. One accesses COP1 registers through standard mips
8366 floating-point support; they are not included in this mechanism.
8368 There is one macro used in defining the MIPS coprocessor interface which
8369 you may want to override in subtargets; it is described below.
8373 @item ALL_COP_ADDITIONAL_REGISTER_NAMES
8374 @findex ALL_COP_ADDITIONAL_REGISTER_NAMES
8375 A comma-separated list (with leading comma) of pairs describing the
8376 alternate names of coprocessor registers. The format of each entry should be
8378 @{ @var{alternatename}, @var{register_number}@}
8385 @section Miscellaneous Parameters
8386 @cindex parameters, miscellaneous
8388 @c prevent bad page break with this line
8389 Here are several miscellaneous parameters.
8392 @item PREDICATE_CODES
8393 @findex PREDICATE_CODES
8394 Define this if you have defined special-purpose predicates in the file
8395 @file{@var{machine}.c}. This macro is called within an initializer of an
8396 array of structures. The first field in the structure is the name of a
8397 predicate and the second field is an array of rtl codes. For each
8398 predicate, list all rtl codes that can be in expressions matched by the
8399 predicate. The list should have a trailing comma. Here is an example
8400 of two entries in the list for a typical RISC machine:
8403 #define PREDICATE_CODES \
8404 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
8405 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
8408 Defining this macro does not affect the generated code (however,
8409 incorrect definitions that omit an rtl code that may be matched by the
8410 predicate can cause the compiler to malfunction). Instead, it allows
8411 the table built by @file{genrecog} to be more compact and efficient,
8412 thus speeding up the compiler. The most important predicates to include
8413 in the list specified by this macro are those used in the most insn
8416 For each predicate function named in @code{PREDICATE_CODES}, a
8417 declaration will be generated in @file{insn-codes.h}.
8419 @item SPECIAL_MODE_PREDICATES
8420 @findex SPECIAL_MODE_PREDICATES
8421 Define this if you have special predicates that know special things
8422 about modes. Genrecog will warn about certain forms of
8423 @code{match_operand} without a mode; if the operand predicate is
8424 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
8427 Here is an example from the IA-32 port (@code{ext_register_operand}
8428 specially checks for @code{HImode} or @code{SImode} in preparation
8429 for a byte extraction from @code{%ah} etc.).
8432 #define SPECIAL_MODE_PREDICATES \
8433 "ext_register_operand",
8436 @findex CASE_VECTOR_MODE
8437 @item CASE_VECTOR_MODE
8438 An alias for a machine mode name. This is the machine mode that
8439 elements of a jump-table should have.
8441 @findex CASE_VECTOR_SHORTEN_MODE
8442 @item CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
8443 Optional: return the preferred mode for an @code{addr_diff_vec}
8444 when the minimum and maximum offset are known. If you define this,
8445 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
8446 To make this work, you also have to define @code{INSN_ALIGN} and
8447 make the alignment for @code{addr_diff_vec} explicit.
8448 The @var{body} argument is provided so that the offset_unsigned and scale
8449 flags can be updated.
8451 @findex CASE_VECTOR_PC_RELATIVE
8452 @item CASE_VECTOR_PC_RELATIVE
8453 Define this macro to be a C expression to indicate when jump-tables
8454 should contain relative addresses. If jump-tables never contain
8455 relative addresses, then you need not define this macro.
8457 @findex CASE_DROPS_THROUGH
8458 @item CASE_DROPS_THROUGH
8459 Define this if control falls through a @code{case} insn when the index
8460 value is out of range. This means the specified default-label is
8461 actually ignored by the @code{case} insn proper.
8463 @findex CASE_VALUES_THRESHOLD
8464 @item CASE_VALUES_THRESHOLD
8465 Define this to be the smallest number of different values for which it
8466 is best to use a jump-table instead of a tree of conditional branches.
8467 The default is four for machines with a @code{casesi} instruction and
8468 five otherwise. This is best for most machines.
8470 @findex WORD_REGISTER_OPERATIONS
8471 @item WORD_REGISTER_OPERATIONS
8472 Define this macro if operations between registers with integral mode
8473 smaller than a word are always performed on the entire register.
8474 Most RISC machines have this property and most CISC machines do not.
8476 @findex LOAD_EXTEND_OP
8477 @item LOAD_EXTEND_OP (@var{mode})
8478 Define this macro to be a C expression indicating when insns that read
8479 memory in @var{mode}, an integral mode narrower than a word, set the
8480 bits outside of @var{mode} to be either the sign-extension or the
8481 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
8482 of @var{mode} for which the
8483 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
8484 @code{NIL} for other modes.
8486 This macro is not called with @var{mode} non-integral or with a width
8487 greater than or equal to @code{BITS_PER_WORD}, so you may return any
8488 value in this case. Do not define this macro if it would always return
8489 @code{NIL}. On machines where this macro is defined, you will normally
8490 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
8492 @findex SHORT_IMMEDIATES_SIGN_EXTEND
8493 @item SHORT_IMMEDIATES_SIGN_EXTEND
8494 Define this macro if loading short immediate values into registers sign
8497 @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC
8498 @item FIXUNS_TRUNC_LIKE_FIX_TRUNC
8499 Define this macro if the same instructions that convert a floating
8500 point number to a signed fixed point number also convert validly to an
8505 The maximum number of bytes that a single instruction can move quickly
8506 between memory and registers or between two memory locations.
8508 @findex MAX_MOVE_MAX
8510 The maximum number of bytes that a single instruction can move quickly
8511 between memory and registers or between two memory locations. If this
8512 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
8513 constant value that is the largest value that @code{MOVE_MAX} can have
8516 @findex SHIFT_COUNT_TRUNCATED
8517 @item SHIFT_COUNT_TRUNCATED
8518 A C expression that is nonzero if on this machine the number of bits
8519 actually used for the count of a shift operation is equal to the number
8520 of bits needed to represent the size of the object being shifted. When
8521 this macro is nonzero, the compiler will assume that it is safe to omit
8522 a sign-extend, zero-extend, and certain bitwise `and' instructions that
8523 truncates the count of a shift operation. On machines that have
8524 instructions that act on bit-fields at variable positions, which may
8525 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
8526 also enables deletion of truncations of the values that serve as
8527 arguments to bit-field instructions.
8529 If both types of instructions truncate the count (for shifts) and
8530 position (for bit-field operations), or if no variable-position bit-field
8531 instructions exist, you should define this macro.
8533 However, on some machines, such as the 80386 and the 680x0, truncation
8534 only applies to shift operations and not the (real or pretended)
8535 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
8536 such machines. Instead, add patterns to the @file{md} file that include
8537 the implied truncation of the shift instructions.
8539 You need not define this macro if it would always have the value of zero.
8541 @findex TRULY_NOOP_TRUNCATION
8542 @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
8543 A C expression which is nonzero if on this machine it is safe to
8544 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
8545 bits (where @var{outprec} is smaller than @var{inprec}) by merely
8546 operating on it as if it had only @var{outprec} bits.
8548 On many machines, this expression can be 1.
8550 @c rearranged this, removed the phrase "it is reported that". this was
8551 @c to fix an overfull hbox. --mew 10feb93
8552 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
8553 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
8554 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
8555 such cases may improve things.
8557 @findex STORE_FLAG_VALUE
8558 @item STORE_FLAG_VALUE
8559 A C expression describing the value returned by a comparison operator
8560 with an integral mode and stored by a store-flag instruction
8561 (@samp{s@var{cond}}) when the condition is true. This description must
8562 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
8563 comparison operators whose results have a @code{MODE_INT} mode.
8565 A value of 1 or @minus{}1 means that the instruction implementing the
8566 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
8567 and 0 when the comparison is false. Otherwise, the value indicates
8568 which bits of the result are guaranteed to be 1 when the comparison is
8569 true. This value is interpreted in the mode of the comparison
8570 operation, which is given by the mode of the first operand in the
8571 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
8572 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
8575 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
8576 generate code that depends only on the specified bits. It can also
8577 replace comparison operators with equivalent operations if they cause
8578 the required bits to be set, even if the remaining bits are undefined.
8579 For example, on a machine whose comparison operators return an
8580 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
8581 @samp{0x80000000}, saying that just the sign bit is relevant, the
8585 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
8592 (ashift:SI @var{x} (const_int @var{n}))
8596 where @var{n} is the appropriate shift count to move the bit being
8597 tested into the sign bit.
8599 There is no way to describe a machine that always sets the low-order bit
8600 for a true value, but does not guarantee the value of any other bits,
8601 but we do not know of any machine that has such an instruction. If you
8602 are trying to port GCC to such a machine, include an instruction to
8603 perform a logical-and of the result with 1 in the pattern for the
8604 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
8606 Often, a machine will have multiple instructions that obtain a value
8607 from a comparison (or the condition codes). Here are rules to guide the
8608 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
8613 Use the shortest sequence that yields a valid definition for
8614 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
8615 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
8616 comparison operators to do so because there may be opportunities to
8617 combine the normalization with other operations.
8620 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
8621 slightly preferred on machines with expensive jumps and 1 preferred on
8625 As a second choice, choose a value of @samp{0x80000001} if instructions
8626 exist that set both the sign and low-order bits but do not define the
8630 Otherwise, use a value of @samp{0x80000000}.
8633 Many machines can produce both the value chosen for
8634 @code{STORE_FLAG_VALUE} and its negation in the same number of
8635 instructions. On those machines, you should also define a pattern for
8636 those cases, e.g., one matching
8639 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
8642 Some machines can also perform @code{and} or @code{plus} operations on
8643 condition code values with less instructions than the corresponding
8644 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
8645 machines, define the appropriate patterns. Use the names @code{incscc}
8646 and @code{decscc}, respectively, for the patterns which perform
8647 @code{plus} or @code{minus} operations on condition code values. See
8648 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
8649 find such instruction sequences on other machines.
8651 You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
8654 @findex FLOAT_STORE_FLAG_VALUE
8655 @item FLOAT_STORE_FLAG_VALUE (@var{mode})
8656 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
8657 returned when comparison operators with floating-point results are true.
8658 Define this macro on machine that have comparison operations that return
8659 floating-point values. If there are no such operations, do not define
8664 An alias for the machine mode for pointers. On most machines, define
8665 this to be the integer mode corresponding to the width of a hardware
8666 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
8667 On some machines you must define this to be one of the partial integer
8668 modes, such as @code{PSImode}.
8670 The width of @code{Pmode} must be at least as large as the value of
8671 @code{POINTER_SIZE}. If it is not equal, you must define the macro
8672 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
8675 @findex FUNCTION_MODE
8677 An alias for the machine mode used for memory references to functions
8678 being called, in @code{call} RTL expressions. On most machines this
8679 should be @code{QImode}.
8681 @findex INTEGRATE_THRESHOLD
8682 @item INTEGRATE_THRESHOLD (@var{decl})
8683 A C expression for the maximum number of instructions above which the
8684 function @var{decl} should not be inlined. @var{decl} is a
8685 @code{FUNCTION_DECL} node.
8687 The default definition of this macro is 64 plus 8 times the number of
8688 arguments that the function accepts. Some people think a larger
8689 threshold should be used on RISC machines.
8691 @findex STDC_0_IN_SYSTEM_HEADERS
8692 @item STDC_0_IN_SYSTEM_HEADERS
8693 In normal operation, the preprocessor expands @code{__STDC__} to the
8694 constant 1, to signify that GCC conforms to ISO Standard C@. On some
8695 hosts, like Solaris, the system compiler uses a different convention,
8696 where @code{__STDC__} is normally 0, but is 1 if the user specifies
8697 strict conformance to the C Standard.
8699 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
8700 convention when processing system header files, but when processing user
8701 files @code{__STDC__} will always expand to 1.
8703 @findex NO_IMPLICIT_EXTERN_C
8704 @item NO_IMPLICIT_EXTERN_C
8705 Define this macro if the system header files support C++ as well as C@.
8706 This macro inhibits the usual method of using system header files in
8707 C++, which is to pretend that the file's contents are enclosed in
8708 @samp{extern "C" @{@dots{}@}}.
8710 @findex HANDLE_PRAGMA
8711 @item HANDLE_PRAGMA (@var{getc}, @var{ungetc}, @var{name})
8712 This macro is no longer supported. You must use
8713 @code{REGISTER_TARGET_PRAGMAS} instead.
8715 @findex REGISTER_TARGET_PRAGMAS
8718 @item REGISTER_TARGET_PRAGMAS (@var{pfile})
8719 Define this macro if you want to implement any target-specific pragmas.
8720 If defined, it is a C expression which makes a series of calls to
8721 @code{cpp_register_pragma} for each pragma, with @var{pfile} passed as
8722 the first argument to to these functions. The macro may also do any
8723 setup required for the pragmas.
8725 The primary reason to define this macro is to provide compatibility with
8726 other compilers for the same target. In general, we discourage
8727 definition of target-specific pragmas for GCC@.
8729 If the pragma can be implemented by attributes then you should consider
8730 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
8732 Preprocessor macros that appear on pragma lines are not expanded. All
8733 @samp{#pragma} directives that do not match any registered pragma are
8734 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
8736 @deftypefun void cpp_register_pragma (cpp_reader *@var{pfile}, const char *@var{space}, const char *@var{name}, void (*@var{callback}) (cpp_reader *))
8738 Each call to @code{cpp_register_pragma} establishes one pragma. The
8739 @var{callback} routine will be called when the preprocessor encounters a
8743 #pragma [@var{space}] @var{name} @dots{}
8746 @var{space} is the case-sensitive namespace of the pragma, or
8747 @code{NULL} to put the pragma in the global namespace. The callback
8748 routine receives @var{pfile} as its first argument, which can be passed
8749 on to cpplib's functions if necessary. You can lex tokens after the
8750 @var{name} by calling @code{c_lex}. Tokens that are not read by the
8751 callback will be silently ignored. The end of the line is indicated by
8752 a token of type @code{CPP_EOF}.
8754 For an example use of this routine, see @file{c4x.h} and the callback
8755 routines defined in @file{c4x-c.c}.
8757 Note that the use of @code{c_lex} is specific to the C and C++
8758 compilers. It will not work in the Java or Fortran compilers, or any
8759 other language compilers for that matter. Thus if @code{c_lex} is going
8760 to be called from target-specific code, it must only be done so when
8761 building the C and C++ compilers. This can be done by defining the
8762 variables @code{c_target_objs} and @code{cxx_target_objs} in the
8763 target entry in the @file{config.gcc} file. These variables should name
8764 the target-specific, language-specific object file which contains the
8765 code that uses @code{c_lex}. Note it will also be necessary to add a
8766 rule to the makefile fragment pointed to by @code{tmake_file} that shows
8767 how to build this object file.
8770 @findex HANDLE_SYSV_PRAGMA
8773 @item HANDLE_SYSV_PRAGMA
8774 Define this macro (to a value of 1) if you want the System V style
8775 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
8776 [=<value>]} to be supported by gcc.
8778 The pack pragma specifies the maximum alignment (in bytes) of fields
8779 within a structure, in much the same way as the @samp{__aligned__} and
8780 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
8781 the behavior to the default.
8783 The weak pragma only works if @code{SUPPORTS_WEAK} and
8784 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
8785 of specifically named weak labels, optionally with a value.
8787 @findex HANDLE_PRAGMA_PACK_PUSH_POP
8790 @item HANDLE_PRAGMA_PACK_PUSH_POP
8791 Define this macro (to a value of 1) if you want to support the Win32
8792 style pragmas @samp{#pragma pack(push,@var{n})} and @samp{#pragma
8793 pack(pop)}. The @samp{pack(push,@var{n})} pragma specifies the maximum alignment
8794 (in bytes) of fields within a structure, in much the same way as the
8795 @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
8796 pack value of zero resets the behavior to the default. Successive
8797 invocations of this pragma cause the previous values to be stacked, so
8798 that invocations of @samp{#pragma pack(pop)} will return to the previous
8801 @findex DOLLARS_IN_IDENTIFIERS
8802 @item DOLLARS_IN_IDENTIFIERS
8803 Define this macro to control use of the character @samp{$} in identifier
8804 names. 0 means @samp{$} is not allowed by default; 1 means it is allowed.
8805 1 is the default; there is no need to define this macro in that case.
8806 This macro controls the compiler proper; it does not affect the preprocessor.
8808 @findex NO_DOLLAR_IN_LABEL
8809 @item NO_DOLLAR_IN_LABEL
8810 Define this macro if the assembler does not accept the character
8811 @samp{$} in label names. By default constructors and destructors in
8812 G++ have @samp{$} in the identifiers. If this macro is defined,
8813 @samp{.} is used instead.
8815 @findex NO_DOT_IN_LABEL
8816 @item NO_DOT_IN_LABEL
8817 Define this macro if the assembler does not accept the character
8818 @samp{.} in label names. By default constructors and destructors in G++
8819 have names that use @samp{.}. If this macro is defined, these names
8820 are rewritten to avoid @samp{.}.
8822 @findex DEFAULT_MAIN_RETURN
8823 @item DEFAULT_MAIN_RETURN
8824 Define this macro if the target system expects every program's @code{main}
8825 function to return a standard ``success'' value by default (if no other
8826 value is explicitly returned).
8828 The definition should be a C statement (sans semicolon) to generate the
8829 appropriate rtl instructions. It is used only when compiling the end of
8834 Define this if the target system lacks the function @code{atexit}
8835 from the ISO C standard. If this macro is defined, a default definition
8836 will be provided to support C++. If @code{ON_EXIT} is not defined,
8837 a default @code{exit} function will also be provided.
8841 Define this macro if the target has another way to implement atexit
8842 functionality without replacing @code{exit}. For instance, SunOS 4 has
8843 a similar @code{on_exit} library function.
8845 The definition should be a functional macro which can be used just like
8846 the @code{atexit} function.
8850 Define this if your @code{exit} function needs to do something
8851 besides calling an external function @code{_cleanup} before
8852 terminating with @code{_exit}. The @code{EXIT_BODY} macro is
8853 only needed if @code{NEED_ATEXIT} is defined and @code{ON_EXIT} is not
8856 @findex INSN_SETS_ARE_DELAYED
8857 @item INSN_SETS_ARE_DELAYED (@var{insn})
8858 Define this macro as a C expression that is nonzero if it is safe for the
8859 delay slot scheduler to place instructions in the delay slot of @var{insn},
8860 even if they appear to use a resource set or clobbered in @var{insn}.
8861 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
8862 every @code{call_insn} has this behavior. On machines where some @code{insn}
8863 or @code{jump_insn} is really a function call and hence has this behavior,
8864 you should define this macro.
8866 You need not define this macro if it would always return zero.
8868 @findex INSN_REFERENCES_ARE_DELAYED
8869 @item INSN_REFERENCES_ARE_DELAYED (@var{insn})
8870 Define this macro as a C expression that is nonzero if it is safe for the
8871 delay slot scheduler to place instructions in the delay slot of @var{insn},
8872 even if they appear to set or clobber a resource referenced in @var{insn}.
8873 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
8874 some @code{insn} or @code{jump_insn} is really a function call and its operands
8875 are registers whose use is actually in the subroutine it calls, you should
8876 define this macro. Doing so allows the delay slot scheduler to move
8877 instructions which copy arguments into the argument registers into the delay
8880 You need not define this macro if it would always return zero.
8882 @findex MACHINE_DEPENDENT_REORG
8883 @item MACHINE_DEPENDENT_REORG (@var{insn})
8884 In rare cases, correct code generation requires extra machine
8885 dependent processing between the second jump optimization pass and
8886 delayed branch scheduling. On those machines, define this macro as a C
8887 statement to act on the code starting at @var{insn}.
8889 @findex MULTIPLE_SYMBOL_SPACES
8890 @item MULTIPLE_SYMBOL_SPACES
8891 Define this macro if in some cases global symbols from one translation
8892 unit may not be bound to undefined symbols in another translation unit
8893 without user intervention. For instance, under Microsoft Windows
8894 symbols must be explicitly imported from shared libraries (DLLs).
8896 @findex MD_ASM_CLOBBERS
8897 @item MD_ASM_CLOBBERS (@var{clobbers})
8898 A C statement that adds to @var{clobbers} @code{STRING_CST} trees for
8899 any hard regs the port wishes to automatically clobber for all asms.
8901 @findex MAX_INTEGER_COMPUTATION_MODE
8902 @item MAX_INTEGER_COMPUTATION_MODE
8903 Define this to the largest integer machine mode which can be used for
8904 operations other than load, store and copy operations.
8906 You need only define this macro if the target holds values larger than
8907 @code{word_mode} in general purpose registers. Most targets should not define
8910 @findex MATH_LIBRARY
8912 Define this macro as a C string constant for the linker argument to link
8913 in the system math library, or @samp{""} if the target does not have a
8914 separate math library.
8916 You need only define this macro if the default of @samp{"-lm"} is wrong.
8918 @findex LIBRARY_PATH_ENV
8919 @item LIBRARY_PATH_ENV
8920 Define this macro as a C string constant for the environment variable that
8921 specifies where the linker should look for libraries.
8923 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
8926 @findex TARGET_HAS_F_SETLKW
8927 @item TARGET_HAS_F_SETLKW
8928 Define this macro if the target supports file locking with fcntl / F_SETLKW@.
8929 Note that this functionality is part of POSIX@.
8930 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
8931 to use file locking when exiting a program, which avoids race conditions
8932 if the program has forked.
8934 @findex MAX_CONDITIONAL_EXECUTE
8935 @item MAX_CONDITIONAL_EXECUTE
8937 A C expression for the maximum number of instructions to execute via
8938 conditional execution instructions instead of a branch. A value of
8939 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
8940 1 if it does use cc0.
8942 @findex IFCVT_MODIFY_TESTS
8943 @item IFCVT_MODIFY_TESTS(@var{ce_info}, @var{true_expr}, @var{false_expr})
8944 Used if the target needs to perform machine-dependent modifications on the
8945 conditionals used for turning basic blocks into conditionally executed code.
8946 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
8947 contains information about the currently processed blocks. @var{true_expr}
8948 and @var{false_expr} are the tests that are used for converting the
8949 then-block and the else-block, respectively. Set either @var{true_expr} or
8950 @var{false_expr} to a null pointer if the tests cannot be converted.
8952 @findex IFCVT_MODIFY_MULTIPLE_TESTS
8953 @item IFCVT_MODIFY_MULTIPLE_TESTS(@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
8954 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
8955 if-statements into conditions combined by @code{and} and @code{or} operations.
8956 @var{bb} contains the basic block that contains the test that is currently
8957 being processed and about to be turned into a condition.
8959 @findex IFCVT_MODIFY_INSN
8960 @item IFCVT_MODIFY_INSN(@var{ce_info}, @var{pattern}, @var{insn})
8961 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
8962 be converted to conditional execution format. @var{ce_info} points to
8963 a data structure, @code{struct ce_if_block}, which contains information
8964 about the currently processed blocks.
8966 @findex IFCVT_MODIFY_FINAL
8967 @item IFCVT_MODIFY_FINAL(@var{ce_info})
8968 A C expression to perform any final machine dependent modifications in
8969 converting code to conditional execution. The involved basic blocks
8970 can be found in the @code{struct ce_if_block} structure that is pointed
8971 to by @var{ce_info}.
8973 @findex IFCVT_MODIFY_CANCEL
8974 @item IFCVT_MODIFY_CANCEL(@var{ce_info})
8975 A C expression to cancel any machine dependent modifications in
8976 converting code to conditional execution. The involved basic blocks
8977 can be found in the @code{struct ce_if_block} structure that is pointed
8978 to by @var{ce_info}.
8980 @findex IFCVT_INIT_EXTRA_FIELDS
8981 @item IFCVT_INIT_EXTRA_FIELDS(@var{ce_info})
8982 A C expression to initialize any extra fields in a @code{struct ce_if_block}
8983 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
8985 @findex IFCVT_EXTRA_FIELDS
8986 @item IFCVT_EXTRA_FIELDS
8987 If defined, it should expand to a set of field declarations that will be
8988 added to the @code{struct ce_if_block} structure. These should be intialized
8989 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
8993 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
8994 Define this hook if you have any machine-specific built-in functions
8995 that need to be defined. It should be a function that performs the
8998 Machine specific built-in functions can be useful to expand special machine
8999 instructions that would otherwise not normally be generated because
9000 they have no equivalent in the source language (for example, SIMD vector
9001 instructions or prefetch instructions).
9003 To create a built-in function, call the function @code{builtin_function}
9004 which is defined by the language front end. You can use any type nodes set
9005 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
9006 only language front ends that use those two functions will call
9007 @samp{TARGET_INIT_BUILTINS}.
9010 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
9012 Expand a call to a machine specific built-in function that was set up by
9013 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
9014 function call; the result should go to @var{target} if that is
9015 convenient, and have mode @var{mode} if that is convenient.
9016 @var{subtarget} may be used as the target for computing one of
9017 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
9018 ignored. This function should return the result of the call to the
9023 @findex MD_CAN_REDIRECT_BRANCH
9024 @item MD_CAN_REDIRECT_BRANCH(@var{branch1}, @var{branch2})
9026 Take a branch insn in @var{branch1} and another in @var{branch2}.
9027 Return true if redirecting @var{branch1} to the destination of
9028 @var{branch2} is possible.
9030 On some targets, branches may have a limited range. Optimizing the
9031 filling of delay slots can result in branches being redirected, and this
9032 may in turn cause a branch offset to overflow.
9034 @findex ALLOCATE_INITIAL_VALUE
9035 @item ALLOCATE_INITIAL_VALUE(@var{hard_reg})
9037 When the initial value of a hard register has been copied in a pseudo
9038 register, it is often not necessary to actually allocate another register
9039 to this pseudo register, because the original hard register or a stack slot
9040 it has been saved into can be used. @code{ALLOCATE_INITIAL_VALUE}, if
9041 defined, is called at the start of register allocation once for each
9042 hard register that had its initial value copied by using
9043 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
9044 Possible values are @code{NULL_RTX}, if you don't want
9045 to do any special allocation, a @code{REG} rtx---that would typically be
9046 the hard register itself, if it is known not to be clobbered---or a
9048 If you are returning a @code{MEM}, this is only a hint for the allocator;
9049 it might decide to use another register anyways.
9050 You may use @code{current_function_leaf_function} in the definition of the
9051 macro, functions that use @code{REG_N_SETS}, to determine if the hard
9052 register in question will not be clobbered.
9054 @findex TARGET_OBJECT_SUFFIX
9055 @item TARGET_OBJECT_SUFFIX
9056 Define this macro to be a C string representing the suffix for object
9057 files on your target machine. If you do not define this macro, GCC will
9058 use @samp{.o} as the suffix for object files.
9060 @findex TARGET_EXECUTABLE_SUFFIX
9061 @item TARGET_EXECUTABLE_SUFFIX
9062 Define this macro to be a C string representing the suffix to be
9063 automatically added to executable files on your target machine. If you
9064 do not define this macro, GCC will use the null string as the suffix for
9067 @findex COLLECT_EXPORT_LIST
9068 @item COLLECT_EXPORT_LIST
9069 If defined, @code{collect2} will scan the individual object files
9070 specified on its command line and create an export list for the linker.
9071 Define this macro for systems like AIX, where the linker discards
9072 object files that are not referenced from @code{main} and uses export
9077 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
9078 This target hook returns @code{true} past the point in which new jump
9079 instructions could be created. On machines that require a register for
9080 every jump such as the SHmedia ISA of SH5, this point would typically be
9081 reload, so this target hook should be defined to a function such as:
9085 cannot_modify_jumps_past_reload_p ()
9087 return (reload_completed || reload_in_progress);