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_macro} takes a string in the form
599 accepted by option @option{-D} and unconditionally defines the macro.
601 @code{builtin_macro_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_macro_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.
619 With @code{TARGET_OS_CPP_BUILTINS} this macro obsoletes the
620 @code{CPP_PREDEFINES} target macro.
622 @findex TARGET_OS_CPP_BUILTINS
623 @item TARGET_OS_CPP_BUILTINS()
624 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
625 and is used for the target operating system instead.
627 With @code{TARGET_CPU_CPP_BUILTINS} this macro obsoletes the
628 @code{CPP_PREDEFINES} target macro.
630 @findex CPP_PREDEFINES
632 Define this to be a string constant containing @option{-D} options to
633 define the predefined macros that identify this machine and system.
634 These macros will be predefined unless the @option{-ansi} option (or a
635 @option{-std} option for strict ISO C conformance) is specified.
637 In addition, a parallel set of macros are predefined, whose names are
638 made by appending @samp{__} at the beginning and at the end. These
639 @samp{__} macros are permitted by the ISO standard, so they are
640 predefined regardless of whether @option{-ansi} or a @option{-std} option
643 For example, on the Sun, one can use the following value:
646 "-Dmc68000 -Dsun -Dunix"
649 The result is to define the macros @code{__mc68000__}, @code{__sun__}
650 and @code{__unix__} unconditionally, and the macros @code{mc68000},
651 @code{sun} and @code{unix} provided @option{-ansi} is not specified.
653 @findex extern int target_flags
654 @item extern int target_flags;
655 This declaration should be present.
657 @cindex optional hardware or system features
658 @cindex features, optional, in system conventions
660 This series of macros is to allow compiler command arguments to
661 enable or disable the use of optional features of the target machine.
662 For example, one machine description serves both the 68000 and
663 the 68020; a command argument tells the compiler whether it should
664 use 68020-only instructions or not. This command argument works
665 by means of a macro @code{TARGET_68020} that tests a bit in
668 Define a macro @code{TARGET_@var{featurename}} for each such option.
669 Its definition should test a bit in @code{target_flags}. It is
670 recommended that a helper macro @code{TARGET_MASK_@var{featurename}}
671 is defined for each bit-value to test, and used in
672 @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For
676 #define TARGET_MASK_68020 1
677 #define TARGET_68020 (target_flags & TARGET_MASK_68020)
680 One place where these macros are used is in the condition-expressions
681 of instruction patterns. Note how @code{TARGET_68020} appears
682 frequently in the 68000 machine description file, @file{m68k.md}.
683 Another place they are used is in the definitions of the other
684 macros in the @file{@var{machine}.h} file.
686 @findex TARGET_SWITCHES
687 @item TARGET_SWITCHES
688 This macro defines names of command options to set and clear
689 bits in @code{target_flags}. Its definition is an initializer
690 with a subgrouping for each command option.
692 Each subgrouping contains a string constant, that defines the option
693 name, a number, which contains the bits to set in
694 @code{target_flags}, and a second string which is the description
695 displayed by @option{--help}. If the number is negative then the bits specified
696 by the number are cleared instead of being set. If the description
697 string is present but empty, then no help information will be displayed
698 for that option, but it will not count as an undocumented option. The
699 actual option name is made by appending @samp{-m} to the specified name.
700 Non-empty description strings should be marked with @code{N_(@dots{})} for
701 @command{xgettext}. Please do not mark empty strings because the empty
702 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
703 of the message catalog with meta information, not the empty string.
705 In addition to the description for @option{--help},
706 more detailed documentation for each option should be added to
709 One of the subgroupings should have a null string. The number in
710 this grouping is the default value for @code{target_flags}. Any
711 target options act starting with that value.
713 Here is an example which defines @option{-m68000} and @option{-m68020}
714 with opposite meanings, and picks the latter as the default:
717 #define TARGET_SWITCHES \
718 @{ @{ "68020", TARGET_MASK_68020, "" @}, \
719 @{ "68000", -TARGET_MASK_68020, \
720 N_("Compile for the 68000") @}, \
721 @{ "", TARGET_MASK_68020, "" @}@}
724 @findex TARGET_OPTIONS
726 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
727 options that have values. Its definition is an initializer with a
728 subgrouping for each command option.
730 Each subgrouping contains a string constant, that defines the fixed part
731 of the option name, the address of a variable, and a description string.
732 Non-empty description strings should be marked with @code{N_(@dots{})} for
733 @command{xgettext}. Please do not mark empty strings because the empty
734 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
735 of the message catalog with meta information, not the empty string.
737 The variable, type @code{char *}, is set to the variable part of the
738 given option if the fixed part matches. The actual option name is made
739 by appending @samp{-m} to the specified name. Again, each option should
740 also be documented in @file{invoke.texi}.
742 Here is an example which defines @option{-mshort-data-@var{number}}. If the
743 given option is @option{-mshort-data-512}, the variable @code{m88k_short_data}
744 will be set to the string @code{"512"}.
747 extern char *m88k_short_data;
748 #define TARGET_OPTIONS \
749 @{ @{ "short-data-", &m88k_short_data, \
750 N_("Specify the size of the short data section") @} @}
753 @findex TARGET_VERSION
755 This macro is a C statement to print on @code{stderr} a string
756 describing the particular machine description choice. Every machine
757 description should define @code{TARGET_VERSION}. For example:
761 #define TARGET_VERSION \
762 fprintf (stderr, " (68k, Motorola syntax)");
764 #define TARGET_VERSION \
765 fprintf (stderr, " (68k, MIT syntax)");
769 @findex OVERRIDE_OPTIONS
770 @item OVERRIDE_OPTIONS
771 Sometimes certain combinations of command options do not make sense on
772 a particular target machine. You can define a macro
773 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
774 defined, is executed once just after all the command options have been
777 Don't use this macro to turn on various extra optimizations for
778 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
780 @findex OPTIMIZATION_OPTIONS
781 @item OPTIMIZATION_OPTIONS (@var{level}, @var{size})
782 Some machines may desire to change what optimizations are performed for
783 various optimization levels. This macro, if defined, is executed once
784 just after the optimization level is determined and before the remainder
785 of the command options have been parsed. Values set in this macro are
786 used as the default values for the other command line options.
788 @var{level} is the optimization level specified; 2 if @option{-O2} is
789 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
791 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
793 You should not use this macro to change options that are not
794 machine-specific. These should uniformly selected by the same
795 optimization level on all supported machines. Use this macro to enable
796 machine-specific optimizations.
798 @strong{Do not examine @code{write_symbols} in
799 this macro!} The debugging options are not supposed to alter the
802 @findex CAN_DEBUG_WITHOUT_FP
803 @item CAN_DEBUG_WITHOUT_FP
804 Define this macro if debugging can be performed even without a frame
805 pointer. If this macro is defined, GCC will turn on the
806 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
809 @node Per-Function Data
810 @section Defining data structures for per-function information.
811 @cindex per-function data
812 @cindex data structures
814 If the target needs to store information on a per-function basis, GCC
815 provides a macro and a couple of variables to allow this. Note, just
816 using statics to store the information is a bad idea, since GCC supports
817 nested functions, so you can be halfway through encoding one function
818 when another one comes along.
820 GCC defines a data structure called @code{struct function} which
821 contains all of the data specific to an individual function. This
822 structure contains a field called @code{machine} whose type is
823 @code{struct machine_function *}, which can be used by targets to point
824 to their own specific data.
826 If a target needs per-function specific data it should define the type
827 @code{struct machine_function} and also the macro
828 @code{INIT_EXPANDERS}. This macro should be used to initialize some or
829 all of the function pointers @code{init_machine_status},
830 @code{free_machine_status} and @code{mark_machine_status}. These
831 pointers are explained below.
833 One typical use of per-function, target specific data is to create an
834 RTX to hold the register containing the function's return address. This
835 RTX can then be used to implement the @code{__builtin_return_address}
836 function, for level 0.
838 Note---earlier implementations of GCC used a single data area to hold
839 all of the per-function information. Thus when processing of a nested
840 function began the old per-function data had to be pushed onto a
841 stack, and when the processing was finished, it had to be popped off the
842 stack. GCC used to provide function pointers called
843 @code{save_machine_status} and @code{restore_machine_status} to handle
844 the saving and restoring of the target specific information. Since the
845 single data area approach is no longer used, these pointers are no
848 The macro and function pointers are described below.
851 @findex INIT_EXPANDERS
853 Macro called to initialize any target specific information. This macro
854 is called once per function, before generation of any RTL has begun.
855 The intention of this macro is to allow the initialization of the
856 function pointers below.
858 @findex init_machine_status
859 @item init_machine_status
860 This is a @code{void (*)(struct function *)} function pointer. If this
861 pointer is non-@code{NULL} it will be called once per function, before function
862 compilation starts, in order to allow the target to perform any target
863 specific initialization of the @code{struct function} structure. It is
864 intended that this would be used to initialize the @code{machine} of
867 @findex free_machine_status
868 @item free_machine_status
869 This is a @code{void (*)(struct function *)} function pointer. If this
870 pointer is non-@code{NULL} it will be called once per function, after the
871 function has been compiled, in order to allow any memory allocated
872 during the @code{init_machine_status} function call to be freed.
874 @findex mark_machine_status
875 @item mark_machine_status
876 This is a @code{void (*)(struct function *)} function pointer. If this
877 pointer is non-@code{NULL} it will be called once per function in order to mark
878 any data items in the @code{struct machine_function} structure which
879 need garbage collection.
884 @section Storage Layout
885 @cindex storage layout
887 Note that the definitions of the macros in this table which are sizes or
888 alignments measured in bits do not need to be constant. They can be C
889 expressions that refer to static variables, such as the @code{target_flags}.
890 @xref{Run-time Target}.
893 @findex BITS_BIG_ENDIAN
894 @item BITS_BIG_ENDIAN
895 Define this macro to have the value 1 if the most significant bit in a
896 byte has the lowest number; otherwise define it to have the value zero.
897 This means that bit-field instructions count from the most significant
898 bit. If the machine has no bit-field instructions, then this must still
899 be defined, but it doesn't matter which value it is defined to. This
900 macro need not be a constant.
902 This macro does not affect the way structure fields are packed into
903 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
905 @findex BYTES_BIG_ENDIAN
906 @item BYTES_BIG_ENDIAN
907 Define this macro to have the value 1 if the most significant byte in a
908 word has the lowest number. This macro need not be a constant.
910 @findex WORDS_BIG_ENDIAN
911 @item WORDS_BIG_ENDIAN
912 Define this macro to have the value 1 if, in a multiword object, the
913 most significant word has the lowest number. This applies to both
914 memory locations and registers; GCC fundamentally assumes that the
915 order of words in memory is the same as the order in registers. This
916 macro need not be a constant.
918 @findex LIBGCC2_WORDS_BIG_ENDIAN
919 @item LIBGCC2_WORDS_BIG_ENDIAN
920 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
921 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
922 used only when compiling @file{libgcc2.c}. Typically the value will be set
923 based on preprocessor defines.
925 @findex FLOAT_WORDS_BIG_ENDIAN
926 @item FLOAT_WORDS_BIG_ENDIAN
927 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
928 @code{TFmode} floating point numbers are stored in memory with the word
929 containing the sign bit at the lowest address; otherwise define it to
930 have the value 0. This macro need not be a constant.
932 You need not define this macro if the ordering is the same as for
935 @findex BITS_PER_UNIT
937 Define this macro to be the number of bits in an addressable storage
938 unit (byte). If you do not define this macro the default is 8.
940 @findex BITS_PER_WORD
942 Number of bits in a word. If you do not define this macro, the default
943 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
945 @findex MAX_BITS_PER_WORD
946 @item MAX_BITS_PER_WORD
947 Maximum number of bits in a word. If this is undefined, the default is
948 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
949 largest value that @code{BITS_PER_WORD} can have at run-time.
951 @findex UNITS_PER_WORD
953 Number of storage units in a word; normally 4.
955 @findex MIN_UNITS_PER_WORD
956 @item MIN_UNITS_PER_WORD
957 Minimum number of units in a word. If this is undefined, the default is
958 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
959 smallest value that @code{UNITS_PER_WORD} can have at run-time.
963 Width of a pointer, in bits. You must specify a value no wider than the
964 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
965 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
966 a value the default is @code{BITS_PER_WORD}.
968 @findex POINTERS_EXTEND_UNSIGNED
969 @item POINTERS_EXTEND_UNSIGNED
970 A C expression whose value is greater than zero if pointers that need to be
971 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
972 be zero-extended and zero if they are to be sign-extended. If the value
973 is less then zero then there must be an "ptr_extend" instruction that
974 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
976 You need not define this macro if the @code{POINTER_SIZE} is equal
977 to the width of @code{Pmode}.
980 @item PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
981 A macro to update @var{m} and @var{unsignedp} when an object whose type
982 is @var{type} and which has the specified mode and signedness is to be
983 stored in a register. This macro is only called when @var{type} is a
986 On most RISC machines, which only have operations that operate on a full
987 register, define this macro to set @var{m} to @code{word_mode} if
988 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
989 cases, only integer modes should be widened because wider-precision
990 floating-point operations are usually more expensive than their narrower
993 For most machines, the macro definition does not change @var{unsignedp}.
994 However, some machines, have instructions that preferentially handle
995 either signed or unsigned quantities of certain modes. For example, on
996 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
997 sign-extend the result to 64 bits. On such machines, set
998 @var{unsignedp} according to which kind of extension is more efficient.
1000 Do not define this macro if it would never modify @var{m}.
1002 @findex PROMOTE_FUNCTION_ARGS
1003 @item PROMOTE_FUNCTION_ARGS
1004 Define this macro if the promotion described by @code{PROMOTE_MODE}
1005 should also be done for outgoing function arguments.
1007 @findex PROMOTE_FUNCTION_RETURN
1008 @item PROMOTE_FUNCTION_RETURN
1009 Define this macro if the promotion described by @code{PROMOTE_MODE}
1010 should also be done for the return value of functions.
1012 If this macro is defined, @code{FUNCTION_VALUE} must perform the same
1013 promotions done by @code{PROMOTE_MODE}.
1015 @findex PROMOTE_FOR_CALL_ONLY
1016 @item PROMOTE_FOR_CALL_ONLY
1017 Define this macro if the promotion described by @code{PROMOTE_MODE}
1018 should @emph{only} be performed for outgoing function arguments or
1019 function return values, as specified by @code{PROMOTE_FUNCTION_ARGS}
1020 and @code{PROMOTE_FUNCTION_RETURN}, respectively.
1022 @findex PARM_BOUNDARY
1024 Normal alignment required for function parameters on the stack, in
1025 bits. All stack parameters receive at least this much alignment
1026 regardless of data type. On most machines, this is the same as the
1029 @findex STACK_BOUNDARY
1030 @item STACK_BOUNDARY
1031 Define this macro to the minimum alignment enforced by hardware for the
1032 stack pointer on this machine. The definition is a C expression for the
1033 desired alignment (measured in bits). This value is used as a default
1034 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1035 this should be the same as @code{PARM_BOUNDARY}.
1037 @findex PREFERRED_STACK_BOUNDARY
1038 @item PREFERRED_STACK_BOUNDARY
1039 Define this macro if you wish to preserve a certain alignment for the
1040 stack pointer, greater than what the hardware enforces. The definition
1041 is a C expression for the desired alignment (measured in bits). This
1042 macro must evaluate to a value equal to or larger than
1043 @code{STACK_BOUNDARY}.
1045 @findex FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1046 @item FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1047 A C expression that evaluates true if @code{PREFERRED_STACK_BOUNDARY} is
1048 not guaranteed by the runtime and we should emit code to align the stack
1049 at the beginning of @code{main}.
1051 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
1052 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
1053 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
1054 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
1055 be momentarily unaligned while pushing arguments.
1057 @findex FUNCTION_BOUNDARY
1058 @item FUNCTION_BOUNDARY
1059 Alignment required for a function entry point, in bits.
1061 @findex BIGGEST_ALIGNMENT
1062 @item BIGGEST_ALIGNMENT
1063 Biggest alignment that any data type can require on this machine, in bits.
1065 @findex MINIMUM_ATOMIC_ALIGNMENT
1066 @item MINIMUM_ATOMIC_ALIGNMENT
1067 If defined, the smallest alignment, in bits, that can be given to an
1068 object that can be referenced in one operation, without disturbing any
1069 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1070 on machines that don't have byte or half-word store operations.
1072 @findex BIGGEST_FIELD_ALIGNMENT
1073 @item BIGGEST_FIELD_ALIGNMENT
1074 Biggest alignment that any structure or union field can require on this
1075 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1076 structure and union fields only, unless the field alignment has been set
1077 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1079 @findex ADJUST_FIELD_ALIGN
1080 @item ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1081 An expression for the alignment of a structure field @var{field} if the
1082 alignment computed in the usual way is @var{computed}. GCC uses
1083 this value instead of the value in @code{BIGGEST_ALIGNMENT} or
1084 @code{BIGGEST_FIELD_ALIGNMENT}, if defined.
1086 @findex MAX_OFILE_ALIGNMENT
1087 @item MAX_OFILE_ALIGNMENT
1088 Biggest alignment supported by the object file format of this machine.
1089 Use this macro to limit the alignment which can be specified using the
1090 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1091 the default value is @code{BIGGEST_ALIGNMENT}.
1093 @findex DATA_ALIGNMENT
1094 @item DATA_ALIGNMENT (@var{type}, @var{basic-align})
1095 If defined, a C expression to compute the alignment for a variable in
1096 the static store. @var{type} is the data type, and @var{basic-align} is
1097 the alignment that the object would ordinarily have. The value of this
1098 macro is used instead of that alignment to align the object.
1100 If this macro is not defined, then @var{basic-align} is used.
1103 One use of this macro is to increase alignment of medium-size data to
1104 make it all fit in fewer cache lines. Another is to cause character
1105 arrays to be word-aligned so that @code{strcpy} calls that copy
1106 constants to character arrays can be done inline.
1108 @findex CONSTANT_ALIGNMENT
1109 @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1110 If defined, a C expression to compute the alignment given to a constant
1111 that is being placed in memory. @var{constant} is the constant and
1112 @var{basic-align} is the alignment that the object would ordinarily
1113 have. The value of this macro is used instead of that alignment to
1116 If this macro is not defined, then @var{basic-align} is used.
1118 The typical use of this macro is to increase alignment for string
1119 constants to be word aligned so that @code{strcpy} calls that copy
1120 constants can be done inline.
1122 @findex LOCAL_ALIGNMENT
1123 @item LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1124 If defined, a C expression to compute the alignment for a variable in
1125 the local store. @var{type} is the data type, and @var{basic-align} is
1126 the alignment that the object would ordinarily have. The value of this
1127 macro is used instead of that alignment to align the object.
1129 If this macro is not defined, then @var{basic-align} is used.
1131 One use of this macro is to increase alignment of medium-size data to
1132 make it all fit in fewer cache lines.
1134 @findex EMPTY_FIELD_BOUNDARY
1135 @item EMPTY_FIELD_BOUNDARY
1136 Alignment in bits to be given to a structure bit-field that follows an
1137 empty field such as @code{int : 0;}.
1139 Note that @code{PCC_BITFIELD_TYPE_MATTERS} also affects the alignment
1140 that results from an empty field.
1142 @findex STRUCTURE_SIZE_BOUNDARY
1143 @item STRUCTURE_SIZE_BOUNDARY
1144 Number of bits which any structure or union's size must be a multiple of.
1145 Each structure or union's size is rounded up to a multiple of this.
1147 If you do not define this macro, the default is the same as
1148 @code{BITS_PER_UNIT}.
1150 @findex STRICT_ALIGNMENT
1151 @item STRICT_ALIGNMENT
1152 Define this macro to be the value 1 if instructions will fail to work
1153 if given data not on the nominal alignment. If instructions will merely
1154 go slower in that case, define this macro as 0.
1156 @findex PCC_BITFIELD_TYPE_MATTERS
1157 @item PCC_BITFIELD_TYPE_MATTERS
1158 Define this if you wish to imitate the way many other C compilers handle
1159 alignment of bit-fields and the structures that contain them.
1161 The behavior is that the type written for a bit-field (@code{int},
1162 @code{short}, or other integer type) imposes an alignment for the
1163 entire structure, as if the structure really did contain an ordinary
1164 field of that type. In addition, the bit-field is placed within the
1165 structure so that it would fit within such a field, not crossing a
1168 Thus, on most machines, a bit-field whose type is written as @code{int}
1169 would not cross a four-byte boundary, and would force four-byte
1170 alignment for the whole structure. (The alignment used may not be four
1171 bytes; it is controlled by the other alignment parameters.)
1173 If the macro is defined, its definition should be a C expression;
1174 a nonzero value for the expression enables this behavior.
1176 Note that if this macro is not defined, or its value is zero, some
1177 bit-fields may cross more than one alignment boundary. The compiler can
1178 support such references if there are @samp{insv}, @samp{extv}, and
1179 @samp{extzv} insns that can directly reference memory.
1181 The other known way of making bit-fields work is to define
1182 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1183 Then every structure can be accessed with fullwords.
1185 Unless the machine has bit-field instructions or you define
1186 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1187 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1189 If your aim is to make GCC use the same conventions for laying out
1190 bit-fields as are used by another compiler, here is how to investigate
1191 what the other compiler does. Compile and run this program:
1210 printf ("Size of foo1 is %d\n",
1211 sizeof (struct foo1));
1212 printf ("Size of foo2 is %d\n",
1213 sizeof (struct foo2));
1218 If this prints 2 and 5, then the compiler's behavior is what you would
1219 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1221 @findex BITFIELD_NBYTES_LIMITED
1222 @item BITFIELD_NBYTES_LIMITED
1223 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1224 to aligning a bit-field within the structure.
1226 @findex MEMBER_TYPE_FORCES_BLK
1227 @item MEMBER_TYPE_FORCES_BLK (@var{field})
1228 Return 1 if a structure or array containing @var{field} should be accessed using
1231 Normally, this is not needed. See the file @file{c4x.h} for an example
1232 of how to use this macro to prevent a structure having a floating point
1233 field from being accessed in an integer mode.
1235 @findex ROUND_TYPE_SIZE
1236 @item ROUND_TYPE_SIZE (@var{type}, @var{computed}, @var{specified})
1237 Define this macro as an expression for the overall size of a type
1238 (given by @var{type} as a tree node) when the size computed in the
1239 usual way is @var{computed} and the alignment is @var{specified}.
1241 The default is to round @var{computed} up to a multiple of @var{specified}.
1243 @findex ROUND_TYPE_SIZE_UNIT
1244 @item ROUND_TYPE_SIZE_UNIT (@var{type}, @var{computed}, @var{specified})
1245 Similar to @code{ROUND_TYPE_SIZE}, but sizes and alignments are
1246 specified in units (bytes). If you define @code{ROUND_TYPE_SIZE},
1247 you must also define this macro and they must be defined consistently
1250 @findex ROUND_TYPE_ALIGN
1251 @item ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1252 Define this macro as an expression for the alignment of a type (given
1253 by @var{type} as a tree node) if the alignment computed in the usual
1254 way is @var{computed} and the alignment explicitly specified was
1257 The default is to use @var{specified} if it is larger; otherwise, use
1258 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1260 @findex MAX_FIXED_MODE_SIZE
1261 @item MAX_FIXED_MODE_SIZE
1262 An integer expression for the size in bits of the largest integer
1263 machine mode that should actually be used. All integer machine modes of
1264 this size or smaller can be used for structures and unions with the
1265 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1266 (DImode)} is assumed.
1268 @findex VECTOR_MODE_SUPPORTED_P
1269 @item VECTOR_MODE_SUPPORTED_P(@var{mode})
1270 Define this macro to be nonzero if the port is prepared to handle insns
1271 involving vector mode @var{mode}. At the very least, it must have move
1272 patterns for this mode.
1274 @findex STACK_SAVEAREA_MODE
1275 @item STACK_SAVEAREA_MODE (@var{save_level})
1276 If defined, an expression of type @code{enum machine_mode} that
1277 specifies the mode of the save area operand of a
1278 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1279 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1280 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1281 having its mode specified.
1283 You need not define this macro if it always returns @code{Pmode}. You
1284 would most commonly define this macro if the
1285 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1288 @findex STACK_SIZE_MODE
1289 @item STACK_SIZE_MODE
1290 If defined, an expression of type @code{enum machine_mode} that
1291 specifies the mode of the size increment operand of an
1292 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1294 You need not define this macro if it always returns @code{word_mode}.
1295 You would most commonly define this macro if the @code{allocate_stack}
1296 pattern needs to support both a 32- and a 64-bit mode.
1298 @findex CHECK_FLOAT_VALUE
1299 @item CHECK_FLOAT_VALUE (@var{mode}, @var{value}, @var{overflow})
1300 A C statement to validate the value @var{value} (of type
1301 @code{double}) for mode @var{mode}. This means that you check whether
1302 @var{value} fits within the possible range of values for mode
1303 @var{mode} on this target machine. The mode @var{mode} is always
1304 a mode of class @code{MODE_FLOAT}. @var{overflow} is nonzero if
1305 the value is already known to be out of range.
1307 If @var{value} is not valid or if @var{overflow} is nonzero, you should
1308 set @var{overflow} to 1 and then assign some valid value to @var{value}.
1309 Allowing an invalid value to go through the compiler can produce
1310 incorrect assembler code which may even cause Unix assemblers to crash.
1312 This macro need not be defined if there is no work for it to do.
1314 @findex TARGET_FLOAT_FORMAT
1315 @item TARGET_FLOAT_FORMAT
1316 A code distinguishing the floating point format of the target machine.
1317 There are five defined values:
1320 @findex IEEE_FLOAT_FORMAT
1321 @item IEEE_FLOAT_FORMAT
1322 This code indicates IEEE floating point. It is the default; there is no
1323 need to define this macro when the format is IEEE@.
1325 @findex VAX_FLOAT_FORMAT
1326 @item VAX_FLOAT_FORMAT
1327 This code indicates the ``D float'' format used on the VAX@.
1329 @findex IBM_FLOAT_FORMAT
1330 @item IBM_FLOAT_FORMAT
1331 This code indicates the format used on the IBM System/370.
1333 @findex C4X_FLOAT_FORMAT
1334 @item C4X_FLOAT_FORMAT
1335 This code indicates the format used on the TMS320C3x/C4x.
1337 @findex UNKNOWN_FLOAT_FORMAT
1338 @item UNKNOWN_FLOAT_FORMAT
1339 This code indicates any other format.
1342 The value of this macro is compared with @code{HOST_FLOAT_FORMAT}, which
1343 is defined by the @command{configure} script, to determine whether the
1344 target machine has the same format as the host machine. If any other
1345 formats are actually in use on supported machines, new codes should be
1348 The ordering of the component words of floating point values stored in
1349 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1351 @findex MODE_HAS_NANS
1352 @item MODE_HAS_NANS (@var{mode})
1353 When defined, this macro should be true if @var{mode} has a NaN
1354 representation. The compiler assumes that NaNs are not equal to
1355 anything (including themselves) and that addition, subtraction,
1356 multiplication and division all return NaNs when one operand is
1359 By default, this macro is true if @var{mode} is a floating-point
1360 mode and the target floating-point format is IEEE@.
1362 @findex MODE_HAS_INFINITIES
1363 @item MODE_HAS_INFINITIES (@var{mode})
1364 This macro should be true if @var{mode} can represent infinity. At
1365 present, the compiler uses this macro to decide whether @samp{x - x}
1366 is always defined. By default, the macro is true when @var{mode}
1367 is a floating-point mode and the target format is IEEE@.
1369 @findex MODE_HAS_SIGNED_ZEROS
1370 @item MODE_HAS_SIGNED_ZEROS (@var{mode})
1371 True if @var{mode} distinguishes between positive and negative zero.
1372 The rules are expected to follow the IEEE standard:
1376 @samp{x + x} has the same sign as @samp{x}.
1379 If the sum of two values with opposite sign is zero, the result is
1380 positive for all rounding modes expect towards @minus{}infinity, for
1381 which it is negative.
1384 The sign of a product or quotient is negative when exactly one
1385 of the operands is negative.
1388 The default definition is true if @var{mode} is a floating-point
1389 mode and the target format is IEEE@.
1391 @findex MODE_HAS_SIGN_DEPENDENT_ROUNDING
1392 @item MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1393 If defined, this macro should be true for @var{mode} if it has at
1394 least one rounding mode in which @samp{x} and @samp{-x} can be
1395 rounded to numbers of different magnitude. Two such modes are
1396 towards @minus{}infinity and towards +infinity.
1398 The default definition of this macro is true if @var{mode} is
1399 a floating-point mode and the target format is IEEE@.
1401 @findex ROUND_TOWARDS_ZERO
1402 @item ROUND_TOWARDS_ZERO
1403 If defined, this macro should be true if the prevailing rounding
1404 mode is towards zero. A true value has the following effects:
1408 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1411 @file{libgcc.a}'s floating-point emulator will round towards zero
1412 rather than towards nearest.
1415 The compiler's floating-point emulator will round towards zero after
1416 doing arithmetic, and when converting from the internal float format to
1420 The macro does not affect the parsing of string literals. When the
1421 primary rounding mode is towards zero, library functions like
1422 @code{strtod} might still round towards nearest, and the compiler's
1423 parser should behave like the target's @code{strtod} where possible.
1425 Not defining this macro is equivalent to returning zero.
1427 @findex LARGEST_EXPONENT_IS_NORMAL
1428 @item LARGEST_EXPONENT_IS_NORMAL (@var{size})
1429 This macro should only be defined when the target float format is
1430 described as IEEE@. It should return true if floats with @var{size}
1431 bits do not have a NaN or infinity representation, but use the largest
1432 exponent for normal numbers instead.
1434 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1435 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1436 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1437 floating-point arithmetic.
1439 The default definition of this macro returns false for all sizes.
1442 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1443 This target hook returns @code{true} if bit-fields in the given
1444 @var{record_type} are to be laid out following the rules of Microsoft
1445 Visual C/C++, namely: (i) a bit-field won't share the same storage
1446 unit with the previous bit-field if their underlying types have
1447 different sizes, and the bit-field will be aligned to the highest
1448 alignment of the underlying types of itself and of the previous
1449 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1450 the whole enclosing structure, even if it is unnamed; except that
1451 (iii) a zero-sized bit-field will be disregarded unless it follows
1452 another bit-field of non-zero size. If this hook returns @code{true},
1453 other macros that control bit-field layout are ignored.
1457 @section Layout of Source Language Data Types
1459 These macros define the sizes and other characteristics of the standard
1460 basic data types used in programs being compiled. Unlike the macros in
1461 the previous section, these apply to specific features of C and related
1462 languages, rather than to fundamental aspects of storage layout.
1465 @findex INT_TYPE_SIZE
1467 A C expression for the size in bits of the type @code{int} on the
1468 target machine. If you don't define this, the default is one word.
1470 @findex SHORT_TYPE_SIZE
1471 @item SHORT_TYPE_SIZE
1472 A C expression for the size in bits of the type @code{short} on the
1473 target machine. If you don't define this, the default is half a word.
1474 (If this would be less than one storage unit, it is rounded up to one
1477 @findex LONG_TYPE_SIZE
1478 @item LONG_TYPE_SIZE
1479 A C expression for the size in bits of the type @code{long} on the
1480 target machine. If you don't define this, the default is one word.
1482 @findex ADA_LONG_TYPE_SIZE
1483 @item ADA_LONG_TYPE_SIZE
1484 On some machines, the size used for the Ada equivalent of the type
1485 @code{long} by a native Ada compiler differs from that used by C. In
1486 that situation, define this macro to be a C expression to be used for
1487 the size of that type. If you don't define this, the default is the
1488 value of @code{LONG_TYPE_SIZE}.
1490 @findex MAX_LONG_TYPE_SIZE
1491 @item MAX_LONG_TYPE_SIZE
1492 Maximum number for the size in bits of the type @code{long} on the
1493 target machine. If this is undefined, the default is
1494 @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the
1495 largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is
1498 @findex LONG_LONG_TYPE_SIZE
1499 @item LONG_LONG_TYPE_SIZE
1500 A C expression for the size in bits of the type @code{long long} on the
1501 target machine. If you don't define this, the default is two
1502 words. If you want to support GNU Ada on your machine, the value of this
1503 macro must be at least 64.
1505 @findex CHAR_TYPE_SIZE
1506 @item CHAR_TYPE_SIZE
1507 A C expression for the size in bits of the type @code{char} on the
1508 target machine. If you don't define this, the default is
1509 @code{BITS_PER_UNIT}.
1511 @findex BOOL_TYPE_SIZE
1512 @item BOOL_TYPE_SIZE
1513 A C expression for the size in bits of the C++ type @code{bool} and
1514 C99 type @code{_Bool} on the target machine. If you don't define
1515 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1517 @findex FLOAT_TYPE_SIZE
1518 @item FLOAT_TYPE_SIZE
1519 A C expression for the size in bits of the type @code{float} on the
1520 target machine. If you don't define this, the default is one word.
1522 @findex DOUBLE_TYPE_SIZE
1523 @item DOUBLE_TYPE_SIZE
1524 A C expression for the size in bits of the type @code{double} on the
1525 target machine. If you don't define this, the default is two
1528 @findex LONG_DOUBLE_TYPE_SIZE
1529 @item LONG_DOUBLE_TYPE_SIZE
1530 A C expression for the size in bits of the type @code{long double} on
1531 the target machine. If you don't define this, the default is two
1534 @findex MAX_LONG_DOUBLE_TYPE_SIZE
1535 Maximum number for the size in bits of the type @code{long double} on the
1536 target machine. If this is undefined, the default is
1537 @code{LONG_DOUBLE_TYPE_SIZE}. Otherwise, it is the constant value that is
1538 the largest value that @code{LONG_DOUBLE_TYPE_SIZE} can have at run-time.
1539 This is used in @code{cpp}.
1541 @findex INTEL_EXTENDED_IEEE_FORMAT
1542 Define this macro to be 1 if the target machine uses 80-bit floating-point
1543 values with 128-bit size and alignment. This is used in @file{real.c}.
1545 @findex WIDEST_HARDWARE_FP_SIZE
1546 @item WIDEST_HARDWARE_FP_SIZE
1547 A C expression for the size in bits of the widest floating-point format
1548 supported by the hardware. If you define this macro, you must specify a
1549 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1550 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1553 @findex DEFAULT_SIGNED_CHAR
1554 @item DEFAULT_SIGNED_CHAR
1555 An expression whose value is 1 or 0, according to whether the type
1556 @code{char} should be signed or unsigned by default. The user can
1557 always override this default with the options @option{-fsigned-char}
1558 and @option{-funsigned-char}.
1560 @findex DEFAULT_SHORT_ENUMS
1561 @item DEFAULT_SHORT_ENUMS
1562 A C expression to determine whether to give an @code{enum} type
1563 only as many bytes as it takes to represent the range of possible values
1564 of that type. A nonzero value means to do that; a zero value means all
1565 @code{enum} types should be allocated like @code{int}.
1567 If you don't define the macro, the default is 0.
1571 A C expression for a string describing the name of the data type to use
1572 for size values. The typedef name @code{size_t} is defined using the
1573 contents of the string.
1575 The string can contain more than one keyword. If so, separate them with
1576 spaces, and write first any length keyword, then @code{unsigned} if
1577 appropriate, and finally @code{int}. The string must exactly match one
1578 of the data type names defined in the function
1579 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1580 omit @code{int} or change the order---that would cause the compiler to
1583 If you don't define this macro, the default is @code{"long unsigned
1586 @findex PTRDIFF_TYPE
1588 A C expression for a string describing the name of the data type to use
1589 for the result of subtracting two pointers. The typedef name
1590 @code{ptrdiff_t} is defined using the contents of the string. See
1591 @code{SIZE_TYPE} above for more information.
1593 If you don't define this macro, the default is @code{"long int"}.
1597 A C expression for a string describing the name of the data type to use
1598 for wide characters. The typedef name @code{wchar_t} is defined using
1599 the contents of the string. See @code{SIZE_TYPE} above for more
1602 If you don't define this macro, the default is @code{"int"}.
1604 @findex WCHAR_TYPE_SIZE
1605 @item WCHAR_TYPE_SIZE
1606 A C expression for the size in bits of the data type for wide
1607 characters. This is used in @code{cpp}, which cannot make use of
1610 @findex MAX_WCHAR_TYPE_SIZE
1611 @item MAX_WCHAR_TYPE_SIZE
1612 Maximum number for the size in bits of the data type for wide
1613 characters. If this is undefined, the default is
1614 @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1615 largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is
1618 @findex GCOV_TYPE_SIZE
1619 @item GCOV_TYPE_SIZE
1620 A C expression for the size in bits of the type used for gcov counters on the
1621 target machine. If you don't define this, the default is one
1622 @code{LONG_TYPE_SIZE} in case it is greater or equal to 64-bit and
1623 @code{LONG_LONG_TYPE_SIZE} otherwise. You may want to re-define the type to
1624 ensure atomicity for counters in multithreaded programs.
1628 A C expression for a string describing the name of the data type to
1629 use for wide characters passed to @code{printf} and returned from
1630 @code{getwc}. The typedef name @code{wint_t} is defined using the
1631 contents of the string. See @code{SIZE_TYPE} above for more
1634 If you don't define this macro, the default is @code{"unsigned int"}.
1638 A C expression for a string describing the name of the data type that
1639 can represent any value of any standard or extended signed integer type.
1640 The typedef name @code{intmax_t} is defined using the contents of the
1641 string. See @code{SIZE_TYPE} above for more information.
1643 If you don't define this macro, the default is the first of
1644 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1645 much precision as @code{long long int}.
1647 @findex UINTMAX_TYPE
1649 A C expression for a string describing the name of the data type that
1650 can represent any value of any standard or extended unsigned integer
1651 type. The typedef name @code{uintmax_t} is defined using the contents
1652 of the string. See @code{SIZE_TYPE} above for more information.
1654 If you don't define this macro, the default is the first of
1655 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1656 unsigned int"} that has as much precision as @code{long long unsigned
1659 @findex TARGET_PTRMEMFUNC_VBIT_LOCATION
1660 @item TARGET_PTRMEMFUNC_VBIT_LOCATION
1661 The C++ compiler represents a pointer-to-member-function with a struct
1668 ptrdiff_t vtable_index;
1675 The C++ compiler must use one bit to indicate whether the function that
1676 will be called through a pointer-to-member-function is virtual.
1677 Normally, we assume that the low-order bit of a function pointer must
1678 always be zero. Then, by ensuring that the vtable_index is odd, we can
1679 distinguish which variant of the union is in use. But, on some
1680 platforms function pointers can be odd, and so this doesn't work. In
1681 that case, we use the low-order bit of the @code{delta} field, and shift
1682 the remainder of the @code{delta} field to the left.
1684 GCC will automatically make the right selection about where to store
1685 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1686 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1687 set such that functions always start at even addresses, but the lowest
1688 bit of pointers to functions indicate whether the function at that
1689 address is in ARM or Thumb mode. If this is the case of your
1690 architecture, you should define this macro to
1691 @code{ptrmemfunc_vbit_in_delta}.
1693 In general, you should not have to define this macro. On architectures
1694 in which function addresses are always even, according to
1695 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1696 @code{ptrmemfunc_vbit_in_pfn}.
1698 @findex TARGET_VTABLE_USES_DESCRIPTORS
1699 @item TARGET_VTABLE_USES_DESCRIPTORS
1700 Normally, the C++ compiler uses function pointers in vtables. This
1701 macro allows the target to change to use ``function descriptors''
1702 instead. Function descriptors are found on targets for whom a
1703 function pointer is actually a small data structure. Normally the
1704 data structure consists of the actual code address plus a data
1705 pointer to which the function's data is relative.
1707 If vtables are used, the value of this macro should be the number
1708 of words that the function descriptor occupies.
1711 @node Escape Sequences
1712 @section Target Character Escape Sequences
1713 @cindex escape sequences
1715 By default, GCC assumes that the C character escape sequences take on
1716 their ASCII values for the target. If this is not correct, you must
1717 explicitly define all of the macros below.
1722 A C constant expression for the integer value for escape sequence
1727 A C constant expression for the integer value of the target escape
1728 character. As an extension, GCC evaluates the escape sequences
1729 @samp{\e} and @samp{\E} to this.
1733 @findex TARGET_NEWLINE
1736 @itemx TARGET_NEWLINE
1737 C constant expressions for the integer values for escape sequences
1738 @samp{\b}, @samp{\t} and @samp{\n}.
1746 C constant expressions for the integer values for escape sequences
1747 @samp{\v}, @samp{\f} and @samp{\r}.
1751 @section Register Usage
1752 @cindex register usage
1754 This section explains how to describe what registers the target machine
1755 has, and how (in general) they can be used.
1757 The description of which registers a specific instruction can use is
1758 done with register classes; see @ref{Register Classes}. For information
1759 on using registers to access a stack frame, see @ref{Frame Registers}.
1760 For passing values in registers, see @ref{Register Arguments}.
1761 For returning values in registers, see @ref{Scalar Return}.
1764 * Register Basics:: Number and kinds of registers.
1765 * Allocation Order:: Order in which registers are allocated.
1766 * Values in Registers:: What kinds of values each reg can hold.
1767 * Leaf Functions:: Renumbering registers for leaf functions.
1768 * Stack Registers:: Handling a register stack such as 80387.
1771 @node Register Basics
1772 @subsection Basic Characteristics of Registers
1774 @c prevent bad page break with this line
1775 Registers have various characteristics.
1778 @findex FIRST_PSEUDO_REGISTER
1779 @item FIRST_PSEUDO_REGISTER
1780 Number of hardware registers known to the compiler. They receive
1781 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1782 pseudo register's number really is assigned the number
1783 @code{FIRST_PSEUDO_REGISTER}.
1785 @item FIXED_REGISTERS
1786 @findex FIXED_REGISTERS
1787 @cindex fixed register
1788 An initializer that says which registers are used for fixed purposes
1789 all throughout the compiled code and are therefore not available for
1790 general allocation. These would include the stack pointer, the frame
1791 pointer (except on machines where that can be used as a general
1792 register when no frame pointer is needed), the program counter on
1793 machines where that is considered one of the addressable registers,
1794 and any other numbered register with a standard use.
1796 This information is expressed as a sequence of numbers, separated by
1797 commas and surrounded by braces. The @var{n}th number is 1 if
1798 register @var{n} is fixed, 0 otherwise.
1800 The table initialized from this macro, and the table initialized by
1801 the following one, may be overridden at run time either automatically,
1802 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1803 the user with the command options @option{-ffixed-@var{reg}},
1804 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1806 @findex CALL_USED_REGISTERS
1807 @item CALL_USED_REGISTERS
1808 @cindex call-used register
1809 @cindex call-clobbered register
1810 @cindex call-saved register
1811 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1812 clobbered (in general) by function calls as well as for fixed
1813 registers. This macro therefore identifies the registers that are not
1814 available for general allocation of values that must live across
1817 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1818 automatically saves it on function entry and restores it on function
1819 exit, if the register is used within the function.
1821 @findex CALL_REALLY_USED_REGISTERS
1822 @item CALL_REALLY_USED_REGISTERS
1823 @cindex call-used register
1824 @cindex call-clobbered register
1825 @cindex call-saved register
1826 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1827 that the entire set of @code{FIXED_REGISTERS} be included.
1828 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1829 This macro is optional. If not specified, it defaults to the value
1830 of @code{CALL_USED_REGISTERS}.
1832 @findex HARD_REGNO_CALL_PART_CLOBBERED
1833 @item HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1834 @cindex call-used register
1835 @cindex call-clobbered register
1836 @cindex call-saved register
1837 A C expression that is nonzero if it is not permissible to store a
1838 value of mode @var{mode} in hard register number @var{regno} across a
1839 call without some part of it being clobbered. For most machines this
1840 macro need not be defined. It is only required for machines that do not
1841 preserve the entire contents of a register across a call.
1843 @findex CONDITIONAL_REGISTER_USAGE
1845 @findex call_used_regs
1846 @item CONDITIONAL_REGISTER_USAGE
1847 Zero or more C statements that may conditionally modify five variables
1848 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1849 @code{reg_names}, and @code{reg_class_contents}, to take into account
1850 any dependence of these register sets on target flags. The first three
1851 of these are of type @code{char []} (interpreted as Boolean vectors).
1852 @code{global_regs} is a @code{const char *[]}, and
1853 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1854 called, @code{fixed_regs}, @code{call_used_regs},
1855 @code{reg_class_contents}, and @code{reg_names} have been initialized
1856 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1857 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1858 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1859 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1860 command options have been applied.
1862 You need not define this macro if it has no work to do.
1864 @cindex disabling certain registers
1865 @cindex controlling register usage
1866 If the usage of an entire class of registers depends on the target
1867 flags, you may indicate this to GCC by using this macro to modify
1868 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1869 registers in the classes which should not be used by GCC@. Also define
1870 the macro @code{REG_CLASS_FROM_LETTER} to return @code{NO_REGS} if it
1871 is called with a letter for a class that shouldn't be used.
1873 (However, if this class is not included in @code{GENERAL_REGS} and all
1874 of the insn patterns whose constraints permit this class are
1875 controlled by target switches, then GCC will automatically avoid using
1876 these registers when the target switches are opposed to them.)
1878 @findex NON_SAVING_SETJMP
1879 @item NON_SAVING_SETJMP
1880 If this macro is defined and has a nonzero value, it means that
1881 @code{setjmp} and related functions fail to save the registers, or that
1882 @code{longjmp} fails to restore them. To compensate, the compiler
1883 avoids putting variables in registers in functions that use
1886 @findex INCOMING_REGNO
1887 @item INCOMING_REGNO (@var{out})
1888 Define this macro if the target machine has register windows. This C
1889 expression returns the register number as seen by the called function
1890 corresponding to the register number @var{out} as seen by the calling
1891 function. Return @var{out} if register number @var{out} is not an
1894 @findex OUTGOING_REGNO
1895 @item OUTGOING_REGNO (@var{in})
1896 Define this macro if the target machine has register windows. This C
1897 expression returns the register number as seen by the calling function
1898 corresponding to the register number @var{in} as seen by the called
1899 function. Return @var{in} if register number @var{in} is not an inbound
1903 @item LOCAL_REGNO (@var{regno})
1904 Define this macro if the target machine has register windows. This C
1905 expression returns true if the register is call-saved but is in the
1906 register window. Unlike most call-saved registers, such registers
1907 need not be explicitly restored on function exit or during non-local
1913 If the program counter has a register number, define this as that
1914 register number. Otherwise, do not define it.
1918 @node Allocation Order
1919 @subsection Order of Allocation of Registers
1920 @cindex order of register allocation
1921 @cindex register allocation order
1923 @c prevent bad page break with this line
1924 Registers are allocated in order.
1927 @findex REG_ALLOC_ORDER
1928 @item REG_ALLOC_ORDER
1929 If defined, an initializer for a vector of integers, containing the
1930 numbers of hard registers in the order in which GCC should prefer
1931 to use them (from most preferred to least).
1933 If this macro is not defined, registers are used lowest numbered first
1934 (all else being equal).
1936 One use of this macro is on machines where the highest numbered
1937 registers must always be saved and the save-multiple-registers
1938 instruction supports only sequences of consecutive registers. On such
1939 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1940 the highest numbered allocable register first.
1942 @findex ORDER_REGS_FOR_LOCAL_ALLOC
1943 @item ORDER_REGS_FOR_LOCAL_ALLOC
1944 A C statement (sans semicolon) to choose the order in which to allocate
1945 hard registers for pseudo-registers local to a basic block.
1947 Store the desired register order in the array @code{reg_alloc_order}.
1948 Element 0 should be the register to allocate first; element 1, the next
1949 register; and so on.
1951 The macro body should not assume anything about the contents of
1952 @code{reg_alloc_order} before execution of the macro.
1954 On most machines, it is not necessary to define this macro.
1957 @node Values in Registers
1958 @subsection How Values Fit in Registers
1960 This section discusses the macros that describe which kinds of values
1961 (specifically, which machine modes) each register can hold, and how many
1962 consecutive registers are needed for a given mode.
1965 @findex HARD_REGNO_NREGS
1966 @item HARD_REGNO_NREGS (@var{regno}, @var{mode})
1967 A C expression for the number of consecutive hard registers, starting
1968 at register number @var{regno}, required to hold a value of mode
1971 On a machine where all registers are exactly one word, a suitable
1972 definition of this macro is
1975 #define HARD_REGNO_NREGS(REGNO, MODE) \
1976 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
1980 @findex HARD_REGNO_MODE_OK
1981 @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
1982 A C expression that is nonzero if it is permissible to store a value
1983 of mode @var{mode} in hard register number @var{regno} (or in several
1984 registers starting with that one). For a machine where all registers
1985 are equivalent, a suitable definition is
1988 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
1991 You need not include code to check for the numbers of fixed registers,
1992 because the allocation mechanism considers them to be always occupied.
1994 @cindex register pairs
1995 On some machines, double-precision values must be kept in even/odd
1996 register pairs. You can implement that by defining this macro to reject
1997 odd register numbers for such modes.
1999 The minimum requirement for a mode to be OK in a register is that the
2000 @samp{mov@var{mode}} instruction pattern support moves between the
2001 register and other hard register in the same class and that moving a
2002 value into the register and back out not alter it.
2004 Since the same instruction used to move @code{word_mode} will work for
2005 all narrower integer modes, it is not necessary on any machine for
2006 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2007 you define patterns @samp{movhi}, etc., to take advantage of this. This
2008 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2009 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2012 Many machines have special registers for floating point arithmetic.
2013 Often people assume that floating point machine modes are allowed only
2014 in floating point registers. This is not true. Any registers that
2015 can hold integers can safely @emph{hold} a floating point machine
2016 mode, whether or not floating arithmetic can be done on it in those
2017 registers. Integer move instructions can be used to move the values.
2019 On some machines, though, the converse is true: fixed-point machine
2020 modes may not go in floating registers. This is true if the floating
2021 registers normalize any value stored in them, because storing a
2022 non-floating value there would garble it. In this case,
2023 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2024 floating registers. But if the floating registers do not automatically
2025 normalize, if you can store any bit pattern in one and retrieve it
2026 unchanged without a trap, then any machine mode may go in a floating
2027 register, so you can define this macro to say so.
2029 The primary significance of special floating registers is rather that
2030 they are the registers acceptable in floating point arithmetic
2031 instructions. However, this is of no concern to
2032 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2033 constraints for those instructions.
2035 On some machines, the floating registers are especially slow to access,
2036 so that it is better to store a value in a stack frame than in such a
2037 register if floating point arithmetic is not being done. As long as the
2038 floating registers are not in class @code{GENERAL_REGS}, they will not
2039 be used unless some pattern's constraint asks for one.
2041 @findex MODES_TIEABLE_P
2042 @item MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2043 A C expression that is nonzero if a value of mode
2044 @var{mode1} is accessible in mode @var{mode2} without copying.
2046 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2047 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2048 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2049 should be nonzero. If they differ for any @var{r}, you should define
2050 this macro to return zero unless some other mechanism ensures the
2051 accessibility of the value in a narrower mode.
2053 You should define this macro to return nonzero in as many cases as
2054 possible since doing so will allow GCC to perform better register
2057 @findex AVOID_CCMODE_COPIES
2058 @item AVOID_CCMODE_COPIES
2059 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2060 registers. You should only define this macro if support for copying to/from
2061 @code{CCmode} is incomplete.
2064 @node Leaf Functions
2065 @subsection Handling Leaf Functions
2067 @cindex leaf functions
2068 @cindex functions, leaf
2069 On some machines, a leaf function (i.e., one which makes no calls) can run
2070 more efficiently if it does not make its own register window. Often this
2071 means it is required to receive its arguments in the registers where they
2072 are passed by the caller, instead of the registers where they would
2075 The special treatment for leaf functions generally applies only when
2076 other conditions are met; for example, often they may use only those
2077 registers for its own variables and temporaries. We use the term ``leaf
2078 function'' to mean a function that is suitable for this special
2079 handling, so that functions with no calls are not necessarily ``leaf
2082 GCC assigns register numbers before it knows whether the function is
2083 suitable for leaf function treatment. So it needs to renumber the
2084 registers in order to output a leaf function. The following macros
2088 @findex LEAF_REGISTERS
2089 @item LEAF_REGISTERS
2090 Name of a char vector, indexed by hard register number, which
2091 contains 1 for a register that is allowable in a candidate for leaf
2094 If leaf function treatment involves renumbering the registers, then the
2095 registers marked here should be the ones before renumbering---those that
2096 GCC would ordinarily allocate. The registers which will actually be
2097 used in the assembler code, after renumbering, should not be marked with 1
2100 Define this macro only if the target machine offers a way to optimize
2101 the treatment of leaf functions.
2103 @findex LEAF_REG_REMAP
2104 @item LEAF_REG_REMAP (@var{regno})
2105 A C expression whose value is the register number to which @var{regno}
2106 should be renumbered, when a function is treated as a leaf function.
2108 If @var{regno} is a register number which should not appear in a leaf
2109 function before renumbering, then the expression should yield @minus{}1, which
2110 will cause the compiler to abort.
2112 Define this macro only if the target machine offers a way to optimize the
2113 treatment of leaf functions, and registers need to be renumbered to do
2117 @findex current_function_is_leaf
2118 @findex current_function_uses_only_leaf_regs
2119 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2120 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2121 specially. They can test the C variable @code{current_function_is_leaf}
2122 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2123 set prior to local register allocation and is valid for the remaining
2124 compiler passes. They can also test the C variable
2125 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2126 functions which only use leaf registers.
2127 @code{current_function_uses_only_leaf_regs} is valid after reload and is
2128 only useful if @code{LEAF_REGISTERS} is defined.
2129 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2130 @c of the next paragraph?! --mew 2feb93
2132 @node Stack Registers
2133 @subsection Registers That Form a Stack
2135 There are special features to handle computers where some of the
2136 ``registers'' form a stack, as in the 80387 coprocessor for the 80386.
2137 Stack registers are normally written by pushing onto the stack, and are
2138 numbered relative to the top of the stack.
2140 Currently, GCC can only handle one group of stack-like registers, and
2141 they must be consecutively numbered.
2146 Define this if the machine has any stack-like registers.
2148 @findex FIRST_STACK_REG
2149 @item FIRST_STACK_REG
2150 The number of the first stack-like register. This one is the top
2153 @findex LAST_STACK_REG
2154 @item LAST_STACK_REG
2155 The number of the last stack-like register. This one is the bottom of
2159 @node Register Classes
2160 @section Register Classes
2161 @cindex register class definitions
2162 @cindex class definitions, register
2164 On many machines, the numbered registers are not all equivalent.
2165 For example, certain registers may not be allowed for indexed addressing;
2166 certain registers may not be allowed in some instructions. These machine
2167 restrictions are described to the compiler using @dfn{register classes}.
2169 You define a number of register classes, giving each one a name and saying
2170 which of the registers belong to it. Then you can specify register classes
2171 that are allowed as operands to particular instruction patterns.
2175 In general, each register will belong to several classes. In fact, one
2176 class must be named @code{ALL_REGS} and contain all the registers. Another
2177 class must be named @code{NO_REGS} and contain no registers. Often the
2178 union of two classes will be another class; however, this is not required.
2180 @findex GENERAL_REGS
2181 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2182 terribly special about the name, but the operand constraint letters
2183 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2184 the same as @code{ALL_REGS}, just define it as a macro which expands
2187 Order the classes so that if class @var{x} is contained in class @var{y}
2188 then @var{x} has a lower class number than @var{y}.
2190 The way classes other than @code{GENERAL_REGS} are specified in operand
2191 constraints is through machine-dependent operand constraint letters.
2192 You can define such letters to correspond to various classes, then use
2193 them in operand constraints.
2195 You should define a class for the union of two classes whenever some
2196 instruction allows both classes. For example, if an instruction allows
2197 either a floating point (coprocessor) register or a general register for a
2198 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2199 which includes both of them. Otherwise you will get suboptimal code.
2201 You must also specify certain redundant information about the register
2202 classes: for each class, which classes contain it and which ones are
2203 contained in it; for each pair of classes, the largest class contained
2206 When a value occupying several consecutive registers is expected in a
2207 certain class, all the registers used must belong to that class.
2208 Therefore, register classes cannot be used to enforce a requirement for
2209 a register pair to start with an even-numbered register. The way to
2210 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2212 Register classes used for input-operands of bitwise-and or shift
2213 instructions have a special requirement: each such class must have, for
2214 each fixed-point machine mode, a subclass whose registers can transfer that
2215 mode to or from memory. For example, on some machines, the operations for
2216 single-byte values (@code{QImode}) are limited to certain registers. When
2217 this is so, each register class that is used in a bitwise-and or shift
2218 instruction must have a subclass consisting of registers from which
2219 single-byte values can be loaded or stored. This is so that
2220 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2223 @findex enum reg_class
2224 @item enum reg_class
2225 An enumeral type that must be defined with all the register class names
2226 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
2227 must be the last register class, followed by one more enumeral value,
2228 @code{LIM_REG_CLASSES}, which is not a register class but rather
2229 tells how many classes there are.
2231 Each register class has a number, which is the value of casting
2232 the class name to type @code{int}. The number serves as an index
2233 in many of the tables described below.
2235 @findex N_REG_CLASSES
2237 The number of distinct register classes, defined as follows:
2240 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2243 @findex REG_CLASS_NAMES
2244 @item REG_CLASS_NAMES
2245 An initializer containing the names of the register classes as C string
2246 constants. These names are used in writing some of the debugging dumps.
2248 @findex REG_CLASS_CONTENTS
2249 @item REG_CLASS_CONTENTS
2250 An initializer containing the contents of the register classes, as integers
2251 which are bit masks. The @var{n}th integer specifies the contents of class
2252 @var{n}. The way the integer @var{mask} is interpreted is that
2253 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2255 When the machine has more than 32 registers, an integer does not suffice.
2256 Then the integers are replaced by sub-initializers, braced groupings containing
2257 several integers. Each sub-initializer must be suitable as an initializer
2258 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2259 In this situation, the first integer in each sub-initializer corresponds to
2260 registers 0 through 31, the second integer to registers 32 through 63, and
2263 @findex REGNO_REG_CLASS
2264 @item REGNO_REG_CLASS (@var{regno})
2265 A C expression whose value is a register class containing hard register
2266 @var{regno}. In general there is more than one such class; choose a class
2267 which is @dfn{minimal}, meaning that no smaller class also contains the
2270 @findex BASE_REG_CLASS
2271 @item BASE_REG_CLASS
2272 A macro whose definition is the name of the class to which a valid
2273 base register must belong. A base register is one used in an address
2274 which is the register value plus a displacement.
2276 @findex MODE_BASE_REG_CLASS
2277 @item MODE_BASE_REG_CLASS (@var{mode})
2278 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2279 the selection of a base register in a mode depenedent manner. If
2280 @var{mode} is VOIDmode then it should return the same value as
2281 @code{BASE_REG_CLASS}.
2283 @findex INDEX_REG_CLASS
2284 @item INDEX_REG_CLASS
2285 A macro whose definition is the name of the class to which a valid
2286 index register must belong. An index register is one used in an
2287 address where its value is either multiplied by a scale factor or
2288 added to another register (as well as added to a displacement).
2290 @findex REG_CLASS_FROM_LETTER
2291 @item REG_CLASS_FROM_LETTER (@var{char})
2292 A C expression which defines the machine-dependent operand constraint
2293 letters for register classes. If @var{char} is such a letter, the
2294 value should be the register class corresponding to it. Otherwise,
2295 the value should be @code{NO_REGS}. The register letter @samp{r},
2296 corresponding to class @code{GENERAL_REGS}, will not be passed
2297 to this macro; you do not need to handle it.
2299 @findex REGNO_OK_FOR_BASE_P
2300 @item REGNO_OK_FOR_BASE_P (@var{num})
2301 A C expression which is nonzero if register number @var{num} is
2302 suitable for use as a base register in operand addresses. It may be
2303 either a suitable hard register or a pseudo register that has been
2304 allocated such a hard register.
2306 @findex REGNO_MODE_OK_FOR_BASE_P
2307 @item REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2308 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2309 that expression may examine the mode of the memory reference in
2310 @var{mode}. You should define this macro if the mode of the memory
2311 reference affects whether a register may be used as a base register. If
2312 you define this macro, the compiler will use it instead of
2313 @code{REGNO_OK_FOR_BASE_P}.
2315 @findex REGNO_OK_FOR_INDEX_P
2316 @item REGNO_OK_FOR_INDEX_P (@var{num})
2317 A C expression which is nonzero if register number @var{num} is
2318 suitable for use as an index register in operand addresses. It may be
2319 either a suitable hard register or a pseudo register that has been
2320 allocated such a hard register.
2322 The difference between an index register and a base register is that
2323 the index register may be scaled. If an address involves the sum of
2324 two registers, neither one of them scaled, then either one may be
2325 labeled the ``base'' and the other the ``index''; but whichever
2326 labeling is used must fit the machine's constraints of which registers
2327 may serve in each capacity. The compiler will try both labelings,
2328 looking for one that is valid, and will reload one or both registers
2329 only if neither labeling works.
2331 @findex PREFERRED_RELOAD_CLASS
2332 @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2333 A C expression that places additional restrictions on the register class
2334 to use when it is necessary to copy value @var{x} into a register in class
2335 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2336 another, smaller class. On many machines, the following definition is
2340 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2343 Sometimes returning a more restrictive class makes better code. For
2344 example, on the 68000, when @var{x} is an integer constant that is in range
2345 for a @samp{moveq} instruction, the value of this macro is always
2346 @code{DATA_REGS} as long as @var{class} includes the data registers.
2347 Requiring a data register guarantees that a @samp{moveq} will be used.
2349 If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
2350 you can force @var{x} into a memory constant. This is useful on
2351 certain machines where immediate floating values cannot be loaded into
2352 certain kinds of registers.
2354 @findex PREFERRED_OUTPUT_RELOAD_CLASS
2355 @item PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2356 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2357 input reloads. If you don't define this macro, the default is to use
2358 @var{class}, unchanged.
2360 @findex LIMIT_RELOAD_CLASS
2361 @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2362 A C expression that places additional restrictions on the register class
2363 to use when it is necessary to be able to hold a value of mode
2364 @var{mode} in a reload register for which class @var{class} would
2367 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2368 there are certain modes that simply can't go in certain reload classes.
2370 The value is a register class; perhaps @var{class}, or perhaps another,
2373 Don't define this macro unless the target machine has limitations which
2374 require the macro to do something nontrivial.
2376 @findex SECONDARY_RELOAD_CLASS
2377 @findex SECONDARY_INPUT_RELOAD_CLASS
2378 @findex SECONDARY_OUTPUT_RELOAD_CLASS
2379 @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2380 @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2381 @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2382 Many machines have some registers that cannot be copied directly to or
2383 from memory or even from other types of registers. An example is the
2384 @samp{MQ} register, which on most machines, can only be copied to or
2385 from general registers, but not memory. Some machines allow copying all
2386 registers to and from memory, but require a scratch register for stores
2387 to some memory locations (e.g., those with symbolic address on the RT,
2388 and those with certain symbolic address on the Sparc when compiling
2389 PIC)@. In some cases, both an intermediate and a scratch register are
2392 You should define these macros to indicate to the reload phase that it may
2393 need to allocate at least one register for a reload in addition to the
2394 register to contain the data. Specifically, if copying @var{x} to a
2395 register @var{class} in @var{mode} requires an intermediate register,
2396 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2397 largest register class all of whose registers can be used as
2398 intermediate registers or scratch registers.
2400 If copying a register @var{class} in @var{mode} to @var{x} requires an
2401 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2402 should be defined to return the largest register class required. If the
2403 requirements for input and output reloads are the same, the macro
2404 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2407 The values returned by these macros are often @code{GENERAL_REGS}.
2408 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2409 can be directly copied to or from a register of @var{class} in
2410 @var{mode} without requiring a scratch register. Do not define this
2411 macro if it would always return @code{NO_REGS}.
2413 If a scratch register is required (either with or without an
2414 intermediate register), you should define patterns for
2415 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2416 (@pxref{Standard Names}. These patterns, which will normally be
2417 implemented with a @code{define_expand}, should be similar to the
2418 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2421 Define constraints for the reload register and scratch register that
2422 contain a single register class. If the original reload register (whose
2423 class is @var{class}) can meet the constraint given in the pattern, the
2424 value returned by these macros is used for the class of the scratch
2425 register. Otherwise, two additional reload registers are required.
2426 Their classes are obtained from the constraints in the insn pattern.
2428 @var{x} might be a pseudo-register or a @code{subreg} of a
2429 pseudo-register, which could either be in a hard register or in memory.
2430 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2431 in memory and the hard register number if it is in a register.
2433 These macros should not be used in the case where a particular class of
2434 registers can only be copied to memory and not to another class of
2435 registers. In that case, secondary reload registers are not needed and
2436 would not be helpful. Instead, a stack location must be used to perform
2437 the copy and the @code{mov@var{m}} pattern should use memory as an
2438 intermediate storage. This case often occurs between floating-point and
2441 @findex SECONDARY_MEMORY_NEEDED
2442 @item SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2443 Certain machines have the property that some registers cannot be copied
2444 to some other registers without using memory. Define this macro on
2445 those machines to be a C expression that is nonzero if objects of mode
2446 @var{m} in registers of @var{class1} can only be copied to registers of
2447 class @var{class2} by storing a register of @var{class1} into memory
2448 and loading that memory location into a register of @var{class2}.
2450 Do not define this macro if its value would always be zero.
2452 @findex SECONDARY_MEMORY_NEEDED_RTX
2453 @item SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2454 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2455 allocates a stack slot for a memory location needed for register copies.
2456 If this macro is defined, the compiler instead uses the memory location
2457 defined by this macro.
2459 Do not define this macro if you do not define
2460 @code{SECONDARY_MEMORY_NEEDED}.
2462 @findex SECONDARY_MEMORY_NEEDED_MODE
2463 @item SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2464 When the compiler needs a secondary memory location to copy between two
2465 registers of mode @var{mode}, it normally allocates sufficient memory to
2466 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2467 load operations in a mode that many bits wide and whose class is the
2468 same as that of @var{mode}.
2470 This is right thing to do on most machines because it ensures that all
2471 bits of the register are copied and prevents accesses to the registers
2472 in a narrower mode, which some machines prohibit for floating-point
2475 However, this default behavior is not correct on some machines, such as
2476 the DEC Alpha, that store short integers in floating-point registers
2477 differently than in integer registers. On those machines, the default
2478 widening will not work correctly and you must define this macro to
2479 suppress that widening in some cases. See the file @file{alpha.h} for
2482 Do not define this macro if you do not define
2483 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2484 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2486 @findex SMALL_REGISTER_CLASSES
2487 @item SMALL_REGISTER_CLASSES
2488 On some machines, it is risky to let hard registers live across arbitrary
2489 insns. Typically, these machines have instructions that require values
2490 to be in specific registers (like an accumulator), and reload will fail
2491 if the required hard register is used for another purpose across such an
2494 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2495 value on these machines. When this macro has a nonzero value, the
2496 compiler will try to minimize the lifetime of hard registers.
2498 It is always safe to define this macro with a nonzero value, but if you
2499 unnecessarily define it, you will reduce the amount of optimizations
2500 that can be performed in some cases. If you do not define this macro
2501 with a nonzero value when it is required, the compiler will run out of
2502 spill registers and print a fatal error message. For most machines, you
2503 should not define this macro at all.
2505 @findex CLASS_LIKELY_SPILLED_P
2506 @item CLASS_LIKELY_SPILLED_P (@var{class})
2507 A C expression whose value is nonzero if pseudos that have been assigned
2508 to registers of class @var{class} would likely be spilled because
2509 registers of @var{class} are needed for spill registers.
2511 The default value of this macro returns 1 if @var{class} has exactly one
2512 register and zero otherwise. On most machines, this default should be
2513 used. Only define this macro to some other expression if pseudos
2514 allocated by @file{local-alloc.c} end up in memory because their hard
2515 registers were needed for spill registers. If this macro returns nonzero
2516 for those classes, those pseudos will only be allocated by
2517 @file{global.c}, which knows how to reallocate the pseudo to another
2518 register. If there would not be another register available for
2519 reallocation, you should not change the definition of this macro since
2520 the only effect of such a definition would be to slow down register
2523 @findex CLASS_MAX_NREGS
2524 @item CLASS_MAX_NREGS (@var{class}, @var{mode})
2525 A C expression for the maximum number of consecutive registers
2526 of class @var{class} needed to hold a value of mode @var{mode}.
2528 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2529 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2530 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2531 @var{mode})} for all @var{regno} values in the class @var{class}.
2533 This macro helps control the handling of multiple-word values
2536 @item CLASS_CANNOT_CHANGE_MODE
2537 If defined, a C expression for a class that contains registers for
2538 which the compiler may not change modes arbitrarily.
2540 @item CLASS_CANNOT_CHANGE_MODE_P(@var{from}, @var{to})
2541 A C expression that is true if, for a register in
2542 @code{CLASS_CANNOT_CHANGE_MODE}, the requested mode punning is invalid.
2544 For the example, loading 32-bit integer or floating-point objects into
2545 floating-point registers on the Alpha extends them to 64-bits.
2546 Therefore loading a 64-bit object and then storing it as a 32-bit object
2547 does not store the low-order 32-bits, as would be the case for a normal
2548 register. Therefore, @file{alpha.h} defines @code{CLASS_CANNOT_CHANGE_MODE}
2549 as @code{FLOAT_REGS} and @code{CLASS_CANNOT_CHANGE_MODE_P} restricts
2550 mode changes to same-size modes.
2552 Compare this to IA-64, which extends floating-point values to 82-bits,
2553 and stores 64-bit integers in a different format than 64-bit doubles.
2554 Therefore @code{CLASS_CANNOT_CHANGE_MODE_P} is always true.
2557 Three other special macros describe which operands fit which constraint
2561 @findex CONST_OK_FOR_LETTER_P
2562 @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2563 A C expression that defines the machine-dependent operand constraint
2564 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2565 particular ranges of integer values. If @var{c} is one of those
2566 letters, the expression should check that @var{value}, an integer, is in
2567 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2568 not one of those letters, the value should be 0 regardless of
2571 @findex CONST_DOUBLE_OK_FOR_LETTER_P
2572 @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2573 A C expression that defines the machine-dependent operand constraint
2574 letters that specify particular ranges of @code{const_double} values
2575 (@samp{G} or @samp{H}).
2577 If @var{c} is one of those letters, the expression should check that
2578 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2579 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2580 letters, the value should be 0 regardless of @var{value}.
2582 @code{const_double} is used for all floating-point constants and for
2583 @code{DImode} fixed-point constants. A given letter can accept either
2584 or both kinds of values. It can use @code{GET_MODE} to distinguish
2585 between these kinds.
2587 @findex EXTRA_CONSTRAINT
2588 @item EXTRA_CONSTRAINT (@var{value}, @var{c})
2589 A C expression that defines the optional machine-dependent constraint
2590 letters that can be used to segregate specific types of operands, usually
2591 memory references, for the target machine. Any letter that is not
2592 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER}
2593 may be used. Normally this macro will not be defined.
2595 If it is required for a particular target machine, it should return 1
2596 if @var{value} corresponds to the operand type represented by the
2597 constraint letter @var{c}. If @var{c} is not defined as an extra
2598 constraint, the value returned should be 0 regardless of @var{value}.
2600 For example, on the ROMP, load instructions cannot have their output
2601 in r0 if the memory reference contains a symbolic address. Constraint
2602 letter @samp{Q} is defined as representing a memory address that does
2603 @emph{not} contain a symbolic address. An alternative is specified with
2604 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2605 alternative specifies @samp{m} on the input and a register class that
2606 does not include r0 on the output.
2609 @node Stack and Calling
2610 @section Stack Layout and Calling Conventions
2611 @cindex calling conventions
2613 @c prevent bad page break with this line
2614 This describes the stack layout and calling conventions.
2618 * Exception Handling::
2623 * Register Arguments::
2625 * Aggregate Return::
2633 @subsection Basic Stack Layout
2634 @cindex stack frame layout
2635 @cindex frame layout
2637 @c prevent bad page break with this line
2638 Here is the basic stack layout.
2641 @findex STACK_GROWS_DOWNWARD
2642 @item STACK_GROWS_DOWNWARD
2643 Define this macro if pushing a word onto the stack moves the stack
2644 pointer to a smaller address.
2646 When we say, ``define this macro if @dots{},'' it means that the
2647 compiler checks this macro only with @code{#ifdef} so the precise
2648 definition used does not matter.
2650 @findex STACK_PUSH_CODE
2651 @item STACK_PUSH_CODE
2653 This macro defines the operation used when something is pushed
2654 on the stack. In RTL, a push operation will be
2655 @code{(set (mem (STACK_PUSH_CODE (reg sp))) ...)}
2657 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2658 and @code{POST_INC}. Which of these is correct depends on
2659 the stack direction and on whether the stack pointer points
2660 to the last item on the stack or whether it points to the
2661 space for the next item on the stack.
2663 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2664 defined, which is almost always right, and @code{PRE_INC} otherwise,
2665 which is often wrong.
2667 @findex FRAME_GROWS_DOWNWARD
2668 @item FRAME_GROWS_DOWNWARD
2669 Define this macro if the addresses of local variable slots are at negative
2670 offsets from the frame pointer.
2672 @findex ARGS_GROW_DOWNWARD
2673 @item ARGS_GROW_DOWNWARD
2674 Define this macro if successive arguments to a function occupy decreasing
2675 addresses on the stack.
2677 @findex STARTING_FRAME_OFFSET
2678 @item STARTING_FRAME_OFFSET
2679 Offset from the frame pointer to the first local variable slot to be allocated.
2681 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2682 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2683 Otherwise, it is found by adding the length of the first slot to the
2684 value @code{STARTING_FRAME_OFFSET}.
2685 @c i'm not sure if the above is still correct.. had to change it to get
2686 @c rid of an overfull. --mew 2feb93
2688 @findex STACK_POINTER_OFFSET
2689 @item STACK_POINTER_OFFSET
2690 Offset from the stack pointer register to the first location at which
2691 outgoing arguments are placed. If not specified, the default value of
2692 zero is used. This is the proper value for most machines.
2694 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2695 the first location at which outgoing arguments are placed.
2697 @findex FIRST_PARM_OFFSET
2698 @item FIRST_PARM_OFFSET (@var{fundecl})
2699 Offset from the argument pointer register to the first argument's
2700 address. On some machines it may depend on the data type of the
2703 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2704 the first argument's address.
2706 @findex STACK_DYNAMIC_OFFSET
2707 @item STACK_DYNAMIC_OFFSET (@var{fundecl})
2708 Offset from the stack pointer register to an item dynamically allocated
2709 on the stack, e.g., by @code{alloca}.
2711 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2712 length of the outgoing arguments. The default is correct for most
2713 machines. See @file{function.c} for details.
2715 @findex DYNAMIC_CHAIN_ADDRESS
2716 @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2717 A C expression whose value is RTL representing the address in a stack
2718 frame where the pointer to the caller's frame is stored. Assume that
2719 @var{frameaddr} is an RTL expression for the address of the stack frame
2722 If you don't define this macro, the default is to return the value
2723 of @var{frameaddr}---that is, the stack frame address is also the
2724 address of the stack word that points to the previous frame.
2726 @findex SETUP_FRAME_ADDRESSES
2727 @item SETUP_FRAME_ADDRESSES
2728 If defined, a C expression that produces the machine-specific code to
2729 setup the stack so that arbitrary frames can be accessed. For example,
2730 on the Sparc, we must flush all of the register windows to the stack
2731 before we can access arbitrary stack frames. You will seldom need to
2734 @findex BUILTIN_SETJMP_FRAME_VALUE
2735 @item BUILTIN_SETJMP_FRAME_VALUE
2736 If defined, a C expression that contains an rtx that is used to store
2737 the address of the current frame into the built in @code{setjmp} buffer.
2738 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2739 machines. One reason you may need to define this macro is if
2740 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2742 @findex RETURN_ADDR_RTX
2743 @item RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2744 A C expression whose value is RTL representing the value of the return
2745 address for the frame @var{count} steps up from the current frame, after
2746 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2747 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2748 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2750 The value of the expression must always be the correct address when
2751 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2752 determine the return address of other frames.
2754 @findex RETURN_ADDR_IN_PREVIOUS_FRAME
2755 @item RETURN_ADDR_IN_PREVIOUS_FRAME
2756 Define this if the return address of a particular stack frame is accessed
2757 from the frame pointer of the previous stack frame.
2759 @findex INCOMING_RETURN_ADDR_RTX
2760 @item INCOMING_RETURN_ADDR_RTX
2761 A C expression whose value is RTL representing the location of the
2762 incoming return address at the beginning of any function, before the
2763 prologue. This RTL is either a @code{REG}, indicating that the return
2764 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2767 You only need to define this macro if you want to support call frame
2768 debugging information like that provided by DWARF 2.
2770 If this RTL is a @code{REG}, you should also define
2771 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2773 @findex INCOMING_FRAME_SP_OFFSET
2774 @item INCOMING_FRAME_SP_OFFSET
2775 A C expression whose value is an integer giving the offset, in bytes,
2776 from the value of the stack pointer register to the top of the stack
2777 frame at the beginning of any function, before the prologue. The top of
2778 the frame is defined to be the value of the stack pointer in the
2779 previous frame, just before the call instruction.
2781 You only need to define this macro if you want to support call frame
2782 debugging information like that provided by DWARF 2.
2784 @findex ARG_POINTER_CFA_OFFSET
2785 @item ARG_POINTER_CFA_OFFSET (@var{fundecl})
2786 A C expression whose value is an integer giving the offset, in bytes,
2787 from the argument pointer to the canonical frame address (cfa). The
2788 final value should coincide with that calculated by
2789 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2790 during virtual register instantiation.
2792 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
2793 which is correct for most machines; in general, the arguments are found
2794 immediately before the stack frame. Note that this is not the case on
2795 some targets that save registers into the caller's frame, such as SPARC
2796 and rs6000, and so such targets need to define this macro.
2798 You only need to define this macro if the default is incorrect, and you
2799 want to support call frame debugging information like that provided by
2804 Define this macro if the stack size for the target is very small. This
2805 has the effect of disabling gcc's built-in @samp{alloca}, though
2806 @samp{__builtin_alloca} is not affected.
2809 @node Exception Handling
2810 @subsection Exception Handling Support
2811 @cindex exception handling
2814 @findex EH_RETURN_DATA_REGNO
2815 @item EH_RETURN_DATA_REGNO (@var{N})
2816 A C expression whose value is the @var{N}th register number used for
2817 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2818 @var{N} registers are usable.
2820 The exception handling library routines communicate with the exception
2821 handlers via a set of agreed upon registers. Ideally these registers
2822 should be call-clobbered; it is possible to use call-saved registers,
2823 but may negatively impact code size. The target must support at least
2824 2 data registers, but should define 4 if there are enough free registers.
2826 You must define this macro if you want to support call frame exception
2827 handling like that provided by DWARF 2.
2829 @findex EH_RETURN_STACKADJ_RTX
2830 @item EH_RETURN_STACKADJ_RTX
2831 A C expression whose value is RTL representing a location in which
2832 to store a stack adjustment to be applied before function return.
2833 This is used to unwind the stack to an exception handler's call frame.
2834 It will be assigned zero on code paths that return normally.
2836 Typically this is a call-clobbered hard register that is otherwise
2837 untouched by the epilogue, but could also be a stack slot.
2839 You must define this macro if you want to support call frame exception
2840 handling like that provided by DWARF 2.
2842 @findex EH_RETURN_HANDLER_RTX
2843 @item EH_RETURN_HANDLER_RTX
2844 A C expression whose value is RTL representing a location in which
2845 to store the address of an exception handler to which we should
2846 return. It will not be assigned on code paths that return normally.
2848 Typically this is the location in the call frame at which the normal
2849 return address is stored. For targets that return by popping an
2850 address off the stack, this might be a memory address just below
2851 the @emph{target} call frame rather than inside the current call
2852 frame. @code{EH_RETURN_STACKADJ_RTX} will have already been assigned,
2853 so it may be used to calculate the location of the target call frame.
2855 Some targets have more complex requirements than storing to an
2856 address calculable during initial code generation. In that case
2857 the @code{eh_return} instruction pattern should be used instead.
2859 If you want to support call frame exception handling, you must
2860 define either this macro or the @code{eh_return} instruction pattern.
2862 @findex ASM_PREFERRED_EH_DATA_FORMAT
2863 @item ASM_PREFERRED_EH_DATA_FORMAT(@var{code}, @var{global})
2864 This macro chooses the encoding of pointers embedded in the exception
2865 handling sections. If at all possible, this should be defined such
2866 that the exception handling section will not require dynamic relocations,
2867 and so may be read-only.
2869 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
2870 @var{global} is true if the symbol may be affected by dynamic relocations.
2871 The macro should return a combination of the @code{DW_EH_PE_*} defines
2872 as found in @file{dwarf2.h}.
2874 If this macro is not defined, pointers will not be encoded but
2875 represented directly.
2877 @findex ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX
2878 @item ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX(@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
2879 This macro allows the target to emit whatever special magic is required
2880 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
2881 Generic code takes care of pc-relative and indirect encodings; this must
2882 be defined if the target uses text-relative or data-relative encodings.
2884 This is a C statement that branches to @var{done} if the format was
2885 handled. @var{encoding} is the format chosen, @var{size} is the number
2886 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
2889 @findex MD_FALLBACK_FRAME_STATE_FOR
2890 @item MD_FALLBACK_FRAME_STATE_FOR(@var{context}, @var{fs}, @var{success})
2891 This macro allows the target to add cpu and operating system specific
2892 code to the call-frame unwinder for use when there is no unwind data
2893 available. The most common reason to implement this macro is to unwind
2894 through signal frames.
2896 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
2897 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
2898 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
2899 for the address of the code being executed and @code{context->cfa} for
2900 the stack pointer value. If the frame can be decoded, the register save
2901 addresses should be updated in @var{fs} and the macro should branch to
2902 @var{success}. If the frame cannot be decoded, the macro should do
2906 @node Stack Checking
2907 @subsection Specifying How Stack Checking is Done
2909 GCC will check that stack references are within the boundaries of
2910 the stack, if the @option{-fstack-check} is specified, in one of three ways:
2914 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
2915 will assume that you have arranged for stack checking to be done at
2916 appropriate places in the configuration files, e.g., in
2917 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
2921 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
2922 called @code{check_stack} in your @file{md} file, GCC will call that
2923 pattern with one argument which is the address to compare the stack
2924 value against. You must arrange for this pattern to report an error if
2925 the stack pointer is out of range.
2928 If neither of the above are true, GCC will generate code to periodically
2929 ``probe'' the stack pointer using the values of the macros defined below.
2932 Normally, you will use the default values of these macros, so GCC
2933 will use the third approach.
2936 @findex STACK_CHECK_BUILTIN
2937 @item STACK_CHECK_BUILTIN
2938 A nonzero value if stack checking is done by the configuration files in a
2939 machine-dependent manner. You should define this macro if stack checking
2940 is require by the ABI of your machine or if you would like to have to stack
2941 checking in some more efficient way than GCC's portable approach.
2942 The default value of this macro is zero.
2944 @findex STACK_CHECK_PROBE_INTERVAL
2945 @item STACK_CHECK_PROBE_INTERVAL
2946 An integer representing the interval at which GCC must generate stack
2947 probe instructions. You will normally define this macro to be no larger
2948 than the size of the ``guard pages'' at the end of a stack area. The
2949 default value of 4096 is suitable for most systems.
2951 @findex STACK_CHECK_PROBE_LOAD
2952 @item STACK_CHECK_PROBE_LOAD
2953 A integer which is nonzero if GCC should perform the stack probe
2954 as a load instruction and zero if GCC should use a store instruction.
2955 The default is zero, which is the most efficient choice on most systems.
2957 @findex STACK_CHECK_PROTECT
2958 @item STACK_CHECK_PROTECT
2959 The number of bytes of stack needed to recover from a stack overflow,
2960 for languages where such a recovery is supported. The default value of
2961 75 words should be adequate for most machines.
2963 @findex STACK_CHECK_MAX_FRAME_SIZE
2964 @item STACK_CHECK_MAX_FRAME_SIZE
2965 The maximum size of a stack frame, in bytes. GCC will generate probe
2966 instructions in non-leaf functions to ensure at least this many bytes of
2967 stack are available. If a stack frame is larger than this size, stack
2968 checking will not be reliable and GCC will issue a warning. The
2969 default is chosen so that GCC only generates one instruction on most
2970 systems. You should normally not change the default value of this macro.
2972 @findex STACK_CHECK_FIXED_FRAME_SIZE
2973 @item STACK_CHECK_FIXED_FRAME_SIZE
2974 GCC uses this value to generate the above warning message. It
2975 represents the amount of fixed frame used by a function, not including
2976 space for any callee-saved registers, temporaries and user variables.
2977 You need only specify an upper bound for this amount and will normally
2978 use the default of four words.
2980 @findex STACK_CHECK_MAX_VAR_SIZE
2981 @item STACK_CHECK_MAX_VAR_SIZE
2982 The maximum size, in bytes, of an object that GCC will place in the
2983 fixed area of the stack frame when the user specifies
2984 @option{-fstack-check}.
2985 GCC computed the default from the values of the above macros and you will
2986 normally not need to override that default.
2990 @node Frame Registers
2991 @subsection Registers That Address the Stack Frame
2993 @c prevent bad page break with this line
2994 This discusses registers that address the stack frame.
2997 @findex STACK_POINTER_REGNUM
2998 @item STACK_POINTER_REGNUM
2999 The register number of the stack pointer register, which must also be a
3000 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3001 the hardware determines which register this is.
3003 @findex FRAME_POINTER_REGNUM
3004 @item FRAME_POINTER_REGNUM
3005 The register number of the frame pointer register, which is used to
3006 access automatic variables in the stack frame. On some machines, the
3007 hardware determines which register this is. On other machines, you can
3008 choose any register you wish for this purpose.
3010 @findex HARD_FRAME_POINTER_REGNUM
3011 @item HARD_FRAME_POINTER_REGNUM
3012 On some machines the offset between the frame pointer and starting
3013 offset of the automatic variables is not known until after register
3014 allocation has been done (for example, because the saved registers are
3015 between these two locations). On those machines, define
3016 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3017 be used internally until the offset is known, and define
3018 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3019 used for the frame pointer.
3021 You should define this macro only in the very rare circumstances when it
3022 is not possible to calculate the offset between the frame pointer and
3023 the automatic variables until after register allocation has been
3024 completed. When this macro is defined, you must also indicate in your
3025 definition of @code{ELIMINABLE_REGS} how to eliminate
3026 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3027 or @code{STACK_POINTER_REGNUM}.
3029 Do not define this macro if it would be the same as
3030 @code{FRAME_POINTER_REGNUM}.
3032 @findex ARG_POINTER_REGNUM
3033 @item ARG_POINTER_REGNUM
3034 The register number of the arg pointer register, which is used to access
3035 the function's argument list. On some machines, this is the same as the
3036 frame pointer register. On some machines, the hardware determines which
3037 register this is. On other machines, you can choose any register you
3038 wish for this purpose. If this is not the same register as the frame
3039 pointer register, then you must mark it as a fixed register according to
3040 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3041 (@pxref{Elimination}).
3043 @findex RETURN_ADDRESS_POINTER_REGNUM
3044 @item RETURN_ADDRESS_POINTER_REGNUM
3045 The register number of the return address pointer register, which is used to
3046 access the current function's return address from the stack. On some
3047 machines, the return address is not at a fixed offset from the frame
3048 pointer or stack pointer or argument pointer. This register can be defined
3049 to point to the return address on the stack, and then be converted by
3050 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3052 Do not define this macro unless there is no other way to get the return
3053 address from the stack.
3055 @findex STATIC_CHAIN_REGNUM
3056 @findex STATIC_CHAIN_INCOMING_REGNUM
3057 @item STATIC_CHAIN_REGNUM
3058 @itemx STATIC_CHAIN_INCOMING_REGNUM
3059 Register numbers used for passing a function's static chain pointer. If
3060 register windows are used, the register number as seen by the called
3061 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3062 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3063 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3066 The static chain register need not be a fixed register.
3068 If the static chain is passed in memory, these macros should not be
3069 defined; instead, the next two macros should be defined.
3071 @findex STATIC_CHAIN
3072 @findex STATIC_CHAIN_INCOMING
3074 @itemx STATIC_CHAIN_INCOMING
3075 If the static chain is passed in memory, these macros provide rtx giving
3076 @code{mem} expressions that denote where they are stored.
3077 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3078 as seen by the calling and called functions, respectively. Often the former
3079 will be at an offset from the stack pointer and the latter at an offset from
3082 @findex stack_pointer_rtx
3083 @findex frame_pointer_rtx
3084 @findex arg_pointer_rtx
3085 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3086 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3087 macros and should be used to refer to those items.
3089 If the static chain is passed in a register, the two previous macros should
3092 @findex DWARF_FRAME_REGISTERS
3093 @item DWARF_FRAME_REGISTERS
3094 This macro specifies the maximum number of hard registers that can be
3095 saved in a call frame. This is used to size data structures used in
3096 DWARF2 exception handling.
3098 Prior to GCC 3.0, this macro was needed in order to establish a stable
3099 exception handling ABI in the face of adding new hard registers for ISA
3100 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3101 in the number of hard registers. Nevertheless, this macro can still be
3102 used to reduce the runtime memory requirements of the exception handling
3103 routines, which can be substantial if the ISA contains a lot of
3104 registers that are not call-saved.
3106 If this macro is not defined, it defaults to
3107 @code{FIRST_PSEUDO_REGISTER}.
3109 @findex PRE_GCC3_DWARF_FRAME_REGISTERS
3110 @item PRE_GCC3_DWARF_FRAME_REGISTERS
3112 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3113 for backward compatibility in pre GCC 3.0 compiled code.
3115 If this macro is not defined, it defaults to
3116 @code{DWARF_FRAME_REGISTERS}.
3121 @subsection Eliminating Frame Pointer and Arg Pointer
3123 @c prevent bad page break with this line
3124 This is about eliminating the frame pointer and arg pointer.
3127 @findex FRAME_POINTER_REQUIRED
3128 @item FRAME_POINTER_REQUIRED
3129 A C expression which is nonzero if a function must have and use a frame
3130 pointer. This expression is evaluated in the reload pass. If its value is
3131 nonzero the function will have a frame pointer.
3133 The expression can in principle examine the current function and decide
3134 according to the facts, but on most machines the constant 0 or the
3135 constant 1 suffices. Use 0 when the machine allows code to be generated
3136 with no frame pointer, and doing so saves some time or space. Use 1
3137 when there is no possible advantage to avoiding a frame pointer.
3139 In certain cases, the compiler does not know how to produce valid code
3140 without a frame pointer. The compiler recognizes those cases and
3141 automatically gives the function a frame pointer regardless of what
3142 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3145 In a function that does not require a frame pointer, the frame pointer
3146 register can be allocated for ordinary usage, unless you mark it as a
3147 fixed register. See @code{FIXED_REGISTERS} for more information.
3149 @findex INITIAL_FRAME_POINTER_OFFSET
3150 @findex get_frame_size
3151 @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3152 A C statement to store in the variable @var{depth-var} the difference
3153 between the frame pointer and the stack pointer values immediately after
3154 the function prologue. The value would be computed from information
3155 such as the result of @code{get_frame_size ()} and the tables of
3156 registers @code{regs_ever_live} and @code{call_used_regs}.
3158 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3159 need not be defined. Otherwise, it must be defined even if
3160 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3161 case, you may set @var{depth-var} to anything.
3163 @findex ELIMINABLE_REGS
3164 @item ELIMINABLE_REGS
3165 If defined, this macro specifies a table of register pairs used to
3166 eliminate unneeded registers that point into the stack frame. If it is not
3167 defined, the only elimination attempted by the compiler is to replace
3168 references to the frame pointer with references to the stack pointer.
3170 The definition of this macro is a list of structure initializations, each
3171 of which specifies an original and replacement register.
3173 On some machines, the position of the argument pointer is not known until
3174 the compilation is completed. In such a case, a separate hard register
3175 must be used for the argument pointer. This register can be eliminated by
3176 replacing it with either the frame pointer or the argument pointer,
3177 depending on whether or not the frame pointer has been eliminated.
3179 In this case, you might specify:
3181 #define ELIMINABLE_REGS \
3182 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3183 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3184 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3187 Note that the elimination of the argument pointer with the stack pointer is
3188 specified first since that is the preferred elimination.
3190 @findex CAN_ELIMINATE
3191 @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3192 A C expression that returns nonzero if the compiler is allowed to try
3193 to replace register number @var{from-reg} with register number
3194 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3195 is defined, and will usually be the constant 1, since most of the cases
3196 preventing register elimination are things that the compiler already
3199 @findex INITIAL_ELIMINATION_OFFSET
3200 @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3201 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3202 specifies the initial difference between the specified pair of
3203 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3207 @node Stack Arguments
3208 @subsection Passing Function Arguments on the Stack
3209 @cindex arguments on stack
3210 @cindex stack arguments
3212 The macros in this section control how arguments are passed
3213 on the stack. See the following section for other macros that
3214 control passing certain arguments in registers.
3217 @findex PROMOTE_PROTOTYPES
3218 @item PROMOTE_PROTOTYPES
3219 A C expression whose value is nonzero if an argument declared in
3220 a prototype as an integral type smaller than @code{int} should
3221 actually be passed as an @code{int}. In addition to avoiding
3222 errors in certain cases of mismatch, it also makes for better
3223 code on certain machines. If the macro is not defined in target
3224 header files, it defaults to 0.
3228 A C expression. If nonzero, push insns will be used to pass
3230 If the target machine does not have a push instruction, set it to zero.
3231 That directs GCC to use an alternate strategy: to
3232 allocate the entire argument block and then store the arguments into
3233 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3234 On some machines, the definition
3236 @findex PUSH_ROUNDING
3237 @item PUSH_ROUNDING (@var{npushed})
3238 A C expression that is the number of bytes actually pushed onto the
3239 stack when an instruction attempts to push @var{npushed} bytes.
3241 On some machines, the definition
3244 #define PUSH_ROUNDING(BYTES) (BYTES)
3248 will suffice. But on other machines, instructions that appear
3249 to push one byte actually push two bytes in an attempt to maintain
3250 alignment. Then the definition should be
3253 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3256 @findex ACCUMULATE_OUTGOING_ARGS
3257 @findex current_function_outgoing_args_size
3258 @item ACCUMULATE_OUTGOING_ARGS
3259 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3260 will be computed and placed into the variable
3261 @code{current_function_outgoing_args_size}. No space will be pushed
3262 onto the stack for each call; instead, the function prologue should
3263 increase the stack frame size by this amount.
3265 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3268 @findex REG_PARM_STACK_SPACE
3269 @item REG_PARM_STACK_SPACE (@var{fndecl})
3270 Define this macro if functions should assume that stack space has been
3271 allocated for arguments even when their values are passed in
3274 The value of this macro is the size, in bytes, of the area reserved for
3275 arguments passed in registers for the function represented by @var{fndecl},
3276 which can be zero if GCC is calling a library function.
3278 This space can be allocated by the caller, or be a part of the
3279 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3281 @c above is overfull. not sure what to do. --mew 5feb93 did
3282 @c something, not sure if it looks good. --mew 10feb93
3284 @findex MAYBE_REG_PARM_STACK_SPACE
3285 @findex FINAL_REG_PARM_STACK_SPACE
3286 @item MAYBE_REG_PARM_STACK_SPACE
3287 @itemx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
3288 Define these macros in addition to the one above if functions might
3289 allocate stack space for arguments even when their values are passed
3290 in registers. These should be used when the stack space allocated
3291 for arguments in registers is not a simple constant independent of the
3292 function declaration.
3294 The value of the first macro is the size, in bytes, of the area that
3295 we should initially assume would be reserved for arguments passed in registers.
3297 The value of the second macro is the actual size, in bytes, of the area
3298 that will be reserved for arguments passed in registers. This takes two
3299 arguments: an integer representing the number of bytes of fixed sized
3300 arguments on the stack, and a tree representing the number of bytes of
3301 variable sized arguments on the stack.
3303 When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
3304 called for libcall functions, the current function, or for a function
3305 being called when it is known that such stack space must be allocated.
3306 In each case this value can be easily computed.
3308 When deciding whether a called function needs such stack space, and how
3309 much space to reserve, GCC uses these two macros instead of
3310 @code{REG_PARM_STACK_SPACE}.
3312 @findex OUTGOING_REG_PARM_STACK_SPACE
3313 @item OUTGOING_REG_PARM_STACK_SPACE
3314 Define this if it is the responsibility of the caller to allocate the area
3315 reserved for arguments passed in registers.
3317 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3318 whether the space for these arguments counts in the value of
3319 @code{current_function_outgoing_args_size}.
3321 @findex STACK_PARMS_IN_REG_PARM_AREA
3322 @item STACK_PARMS_IN_REG_PARM_AREA
3323 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3324 stack parameters don't skip the area specified by it.
3325 @c i changed this, makes more sens and it should have taken care of the
3326 @c overfull.. not as specific, tho. --mew 5feb93
3328 Normally, when a parameter is not passed in registers, it is placed on the
3329 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3330 suppresses this behavior and causes the parameter to be passed on the
3331 stack in its natural location.
3333 @findex RETURN_POPS_ARGS
3334 @item RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3335 A C expression that should indicate the number of bytes of its own
3336 arguments that a function pops on returning, or 0 if the
3337 function pops no arguments and the caller must therefore pop them all
3338 after the function returns.
3340 @var{fundecl} is a C variable whose value is a tree node that describes
3341 the function in question. Normally it is a node of type
3342 @code{FUNCTION_DECL} that describes the declaration of the function.
3343 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3345 @var{funtype} is a C variable whose value is a tree node that
3346 describes the function in question. Normally it is a node of type
3347 @code{FUNCTION_TYPE} that describes the data type of the function.
3348 From this it is possible to obtain the data types of the value and
3349 arguments (if known).
3351 When a call to a library function is being considered, @var{fundecl}
3352 will contain an identifier node for the library function. Thus, if
3353 you need to distinguish among various library functions, you can do so
3354 by their names. Note that ``library function'' in this context means
3355 a function used to perform arithmetic, whose name is known specially
3356 in the compiler and was not mentioned in the C code being compiled.
3358 @var{stack-size} is the number of bytes of arguments passed on the
3359 stack. If a variable number of bytes is passed, it is zero, and
3360 argument popping will always be the responsibility of the calling function.
3362 On the VAX, all functions always pop their arguments, so the definition
3363 of this macro is @var{stack-size}. On the 68000, using the standard
3364 calling convention, no functions pop their arguments, so the value of
3365 the macro is always 0 in this case. But an alternative calling
3366 convention is available in which functions that take a fixed number of
3367 arguments pop them but other functions (such as @code{printf}) pop
3368 nothing (the caller pops all). When this convention is in use,
3369 @var{funtype} is examined to determine whether a function takes a fixed
3370 number of arguments.
3372 @findex CALL_POPS_ARGS
3373 @item CALL_POPS_ARGS (@var{cum})
3374 A C expression that should indicate the number of bytes a call sequence
3375 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3376 when compiling a function call.
3378 @var{cum} is the variable in which all arguments to the called function
3379 have been accumulated.
3381 On certain architectures, such as the SH5, a call trampoline is used
3382 that pops certain registers off the stack, depending on the arguments
3383 that have been passed to the function. Since this is a property of the
3384 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3389 @node Register Arguments
3390 @subsection Passing Arguments in Registers
3391 @cindex arguments in registers
3392 @cindex registers arguments
3394 This section describes the macros which let you control how various
3395 types of arguments are passed in registers or how they are arranged in
3399 @findex FUNCTION_ARG
3400 @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3401 A C expression that controls whether a function argument is passed
3402 in a register, and which register.
3404 The arguments are @var{cum}, which summarizes all the previous
3405 arguments; @var{mode}, the machine mode of the argument; @var{type},
3406 the data type of the argument as a tree node or 0 if that is not known
3407 (which happens for C support library functions); and @var{named},
3408 which is 1 for an ordinary argument and 0 for nameless arguments that
3409 correspond to @samp{@dots{}} in the called function's prototype.
3410 @var{type} can be an incomplete type if a syntax error has previously
3413 The value of the expression is usually either a @code{reg} RTX for the
3414 hard register in which to pass the argument, or zero to pass the
3415 argument on the stack.
3417 For machines like the VAX and 68000, where normally all arguments are
3418 pushed, zero suffices as a definition.
3420 The value of the expression can also be a @code{parallel} RTX@. This is
3421 used when an argument is passed in multiple locations. The mode of the
3422 of the @code{parallel} should be the mode of the entire argument. The
3423 @code{parallel} holds any number of @code{expr_list} pairs; each one
3424 describes where part of the argument is passed. In each
3425 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3426 register in which to pass this part of the argument, and the mode of the
3427 register RTX indicates how large this part of the argument is. The
3428 second operand of the @code{expr_list} is a @code{const_int} which gives
3429 the offset in bytes into the entire argument of where this part starts.
3430 As a special exception the first @code{expr_list} in the @code{parallel}
3431 RTX may have a first operand of zero. This indicates that the entire
3432 argument is also stored on the stack.
3434 The last time this macro is called, it is called with @code{MODE ==
3435 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3436 pattern as operands 2 and 3 respectively.
3438 @cindex @file{stdarg.h} and register arguments
3439 The usual way to make the ISO library @file{stdarg.h} work on a machine
3440 where some arguments are usually passed in registers, is to cause
3441 nameless arguments to be passed on the stack instead. This is done
3442 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3444 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3445 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3446 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
3447 in the definition of this macro to determine if this argument is of a
3448 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3449 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3450 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3451 defined, the argument will be computed in the stack and then loaded into
3454 @findex MUST_PASS_IN_STACK
3455 @item MUST_PASS_IN_STACK (@var{mode}, @var{type})
3456 Define as a C expression that evaluates to nonzero if we do not know how
3457 to pass TYPE solely in registers. The file @file{expr.h} defines a
3458 definition that is usually appropriate, refer to @file{expr.h} for additional
3461 @findex FUNCTION_INCOMING_ARG
3462 @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3463 Define this macro if the target machine has ``register windows'', so
3464 that the register in which a function sees an arguments is not
3465 necessarily the same as the one in which the caller passed the
3468 For such machines, @code{FUNCTION_ARG} computes the register in which
3469 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3470 be defined in a similar fashion to tell the function being called
3471 where the arguments will arrive.
3473 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3474 serves both purposes.
3476 @findex FUNCTION_ARG_PARTIAL_NREGS
3477 @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
3478 A C expression for the number of words, at the beginning of an
3479 argument, that must be put in registers. The value must be zero for
3480 arguments that are passed entirely in registers or that are entirely
3481 pushed on the stack.
3483 On some machines, certain arguments must be passed partially in
3484 registers and partially in memory. On these machines, typically the
3485 first @var{n} words of arguments are passed in registers, and the rest
3486 on the stack. If a multi-word argument (a @code{double} or a
3487 structure) crosses that boundary, its first few words must be passed
3488 in registers and the rest must be pushed. This macro tells the
3489 compiler when this occurs, and how many of the words should go in
3492 @code{FUNCTION_ARG} for these arguments should return the first
3493 register to be used by the caller for this argument; likewise
3494 @code{FUNCTION_INCOMING_ARG}, for the called function.
3496 @findex FUNCTION_ARG_PASS_BY_REFERENCE
3497 @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3498 A C expression that indicates when an argument must be passed by reference.
3499 If nonzero for an argument, a copy of that argument is made in memory and a
3500 pointer to the argument is passed instead of the argument itself.
3501 The pointer is passed in whatever way is appropriate for passing a pointer
3504 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
3505 definition of this macro might be
3507 #define FUNCTION_ARG_PASS_BY_REFERENCE\
3508 (CUM, MODE, TYPE, NAMED) \
3509 MUST_PASS_IN_STACK (MODE, TYPE)
3511 @c this is *still* too long. --mew 5feb93
3513 @findex FUNCTION_ARG_CALLEE_COPIES
3514 @item FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
3515 If defined, a C expression that indicates when it is the called function's
3516 responsibility to make a copy of arguments passed by invisible reference.
3517 Normally, the caller makes a copy and passes the address of the copy to the
3518 routine being called. When @code{FUNCTION_ARG_CALLEE_COPIES} is defined and is
3519 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
3520 ``live'' value. The called function must not modify this value. If it can be
3521 determined that the value won't be modified, it need not make a copy;
3522 otherwise a copy must be made.
3524 @findex FUNCTION_ARG_REG_LITTLE_ENDIAN
3525 @item FUNCTION_ARG_REG_LITTLE_ENDIAN
3526 If defined TRUE on a big-endian system then structure arguments passed
3527 (and returned) in registers are passed in a little-endian manner instead of
3528 the big-endian manner. On the HP-UX IA64 and PA64 platforms structures are
3529 aligned differently then integral values and setting this value to true will
3530 allow for the special handling of structure arguments and return values.
3532 @findex CUMULATIVE_ARGS
3533 @item CUMULATIVE_ARGS
3534 A C type for declaring a variable that is used as the first argument of
3535 @code{FUNCTION_ARG} and other related values. For some target machines,
3536 the type @code{int} suffices and can hold the number of bytes of
3539 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3540 arguments that have been passed on the stack. The compiler has other
3541 variables to keep track of that. For target machines on which all
3542 arguments are passed on the stack, there is no need to store anything in
3543 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3544 should not be empty, so use @code{int}.
3546 @findex INIT_CUMULATIVE_ARGS
3547 @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{indirect})
3548 A C statement (sans semicolon) for initializing the variable @var{cum}
3549 for the state at the beginning of the argument list. The variable has
3550 type @code{CUMULATIVE_ARGS}. The value of @var{fntype} is the tree node
3551 for the data type of the function which will receive the args, or 0
3552 if the args are to a compiler support library function. The value of
3553 @var{indirect} is nonzero when processing an indirect call, for example
3554 a call through a function pointer. The value of @var{indirect} is zero
3555 for a call to an explicitly named function, a library function call, or when
3556 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3559 When processing a call to a compiler support library function,
3560 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3561 contains the name of the function, as a string. @var{libname} is 0 when
3562 an ordinary C function call is being processed. Thus, each time this
3563 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3564 never both of them at once.
3566 @findex INIT_CUMULATIVE_LIBCALL_ARGS
3567 @item INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3568 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3569 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3570 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3571 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3572 0)} is used instead.
3574 @findex INIT_CUMULATIVE_INCOMING_ARGS
3575 @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3576 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3577 finding the arguments for the function being compiled. If this macro is
3578 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3580 The value passed for @var{libname} is always 0, since library routines
3581 with special calling conventions are never compiled with GCC@. The
3582 argument @var{libname} exists for symmetry with
3583 @code{INIT_CUMULATIVE_ARGS}.
3584 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3585 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3587 @findex FUNCTION_ARG_ADVANCE
3588 @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3589 A C statement (sans semicolon) to update the summarizer variable
3590 @var{cum} to advance past an argument in the argument list. The
3591 values @var{mode}, @var{type} and @var{named} describe that argument.
3592 Once this is done, the variable @var{cum} is suitable for analyzing
3593 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
3595 This macro need not do anything if the argument in question was passed
3596 on the stack. The compiler knows how to track the amount of stack space
3597 used for arguments without any special help.
3599 @findex FUNCTION_ARG_PADDING
3600 @item FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3601 If defined, a C expression which determines whether, and in which direction,
3602 to pad out an argument with extra space. The value should be of type
3603 @code{enum direction}: either @code{upward} to pad above the argument,
3604 @code{downward} to pad below, or @code{none} to inhibit padding.
3606 The @emph{amount} of padding is always just enough to reach the next
3607 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3610 This macro has a default definition which is right for most systems.
3611 For little-endian machines, the default is to pad upward. For
3612 big-endian machines, the default is to pad downward for an argument of
3613 constant size shorter than an @code{int}, and upward otherwise.
3615 @findex PAD_VARARGS_DOWN
3616 @item PAD_VARARGS_DOWN
3617 If defined, a C expression which determines whether the default
3618 implementation of va_arg will attempt to pad down before reading the
3619 next argument, if that argument is smaller than its aligned space as
3620 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3621 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3623 @findex FUNCTION_ARG_BOUNDARY
3624 @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3625 If defined, a C expression that gives the alignment boundary, in bits,
3626 of an argument with the specified mode and type. If it is not defined,
3627 @code{PARM_BOUNDARY} is used for all arguments.
3629 @findex FUNCTION_ARG_REGNO_P
3630 @item FUNCTION_ARG_REGNO_P (@var{regno})
3631 A C expression that is nonzero if @var{regno} is the number of a hard
3632 register in which function arguments are sometimes passed. This does
3633 @emph{not} include implicit arguments such as the static chain and
3634 the structure-value address. On many machines, no registers can be
3635 used for this purpose since all function arguments are pushed on the
3638 @findex LOAD_ARGS_REVERSED
3639 @item LOAD_ARGS_REVERSED
3640 If defined, the order in which arguments are loaded into their
3641 respective argument registers is reversed so that the last
3642 argument is loaded first. This macro only affects arguments
3643 passed in registers.
3648 @subsection How Scalar Function Values Are Returned
3649 @cindex return values in registers
3650 @cindex values, returned by functions
3651 @cindex scalars, returned as values
3653 This section discusses the macros that control returning scalars as
3654 values---values that can fit in registers.
3657 @findex FUNCTION_VALUE
3658 @item FUNCTION_VALUE (@var{valtype}, @var{func})
3659 A C expression to create an RTX representing the place where a
3660 function returns a value of data type @var{valtype}. @var{valtype} is
3661 a tree node representing a data type. Write @code{TYPE_MODE
3662 (@var{valtype})} to get the machine mode used to represent that type.
3663 On many machines, only the mode is relevant. (Actually, on most
3664 machines, scalar values are returned in the same place regardless of
3667 The value of the expression is usually a @code{reg} RTX for the hard
3668 register where the return value is stored. The value can also be a
3669 @code{parallel} RTX, if the return value is in multiple places. See
3670 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3672 If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
3673 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3676 If the precise function being called is known, @var{func} is a tree
3677 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3678 pointer. This makes it possible to use a different value-returning
3679 convention for specific functions when all their calls are
3682 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3683 types, because these are returned in another way. See
3684 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3686 @findex FUNCTION_OUTGOING_VALUE
3687 @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3688 Define this macro if the target machine has ``register windows''
3689 so that the register in which a function returns its value is not
3690 the same as the one in which the caller sees the value.
3692 For such machines, @code{FUNCTION_VALUE} computes the register in which
3693 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3694 defined in a similar fashion to tell the function where to put the
3697 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3698 @code{FUNCTION_VALUE} serves both purposes.
3700 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3701 aggregate data types, because these are returned in another way. See
3702 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3704 @findex LIBCALL_VALUE
3705 @item LIBCALL_VALUE (@var{mode})
3706 A C expression to create an RTX representing the place where a library
3707 function returns a value of mode @var{mode}. If the precise function
3708 being called is known, @var{func} is a tree node
3709 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3710 pointer. This makes it possible to use a different value-returning
3711 convention for specific functions when all their calls are
3714 Note that ``library function'' in this context means a compiler
3715 support routine, used to perform arithmetic, whose name is known
3716 specially by the compiler and was not mentioned in the C code being
3719 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3720 data types, because none of the library functions returns such types.
3722 @findex FUNCTION_VALUE_REGNO_P
3723 @item FUNCTION_VALUE_REGNO_P (@var{regno})
3724 A C expression that is nonzero if @var{regno} is the number of a hard
3725 register in which the values of called function may come back.
3727 A register whose use for returning values is limited to serving as the
3728 second of a pair (for a value of type @code{double}, say) need not be
3729 recognized by this macro. So for most machines, this definition
3733 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3736 If the machine has register windows, so that the caller and the called
3737 function use different registers for the return value, this macro
3738 should recognize only the caller's register numbers.
3740 @findex APPLY_RESULT_SIZE
3741 @item APPLY_RESULT_SIZE
3742 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3743 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3744 saving and restoring an arbitrary return value.
3747 @node Aggregate Return
3748 @subsection How Large Values Are Returned
3749 @cindex aggregates as return values
3750 @cindex large return values
3751 @cindex returning aggregate values
3752 @cindex structure value address
3754 When a function value's mode is @code{BLKmode} (and in some other
3755 cases), the value is not returned according to @code{FUNCTION_VALUE}
3756 (@pxref{Scalar Return}). Instead, the caller passes the address of a
3757 block of memory in which the value should be stored. This address
3758 is called the @dfn{structure value address}.
3760 This section describes how to control returning structure values in
3764 @findex RETURN_IN_MEMORY
3765 @item RETURN_IN_MEMORY (@var{type})
3766 A C expression which can inhibit the returning of certain function
3767 values in registers, based on the type of value. A nonzero value says
3768 to return the function value in memory, just as large structures are
3769 always returned. Here @var{type} will be a C expression of type
3770 @code{tree}, representing the data type of the value.
3772 Note that values of mode @code{BLKmode} must be explicitly handled
3773 by this macro. Also, the option @option{-fpcc-struct-return}
3774 takes effect regardless of this macro. On most systems, it is
3775 possible to leave the macro undefined; this causes a default
3776 definition to be used, whose value is the constant 1 for @code{BLKmode}
3777 values, and 0 otherwise.
3779 Do not use this macro to indicate that structures and unions should always
3780 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
3783 @findex DEFAULT_PCC_STRUCT_RETURN
3784 @item DEFAULT_PCC_STRUCT_RETURN
3785 Define this macro to be 1 if all structure and union return values must be
3786 in memory. Since this results in slower code, this should be defined
3787 only if needed for compatibility with other compilers or with an ABI@.
3788 If you define this macro to be 0, then the conventions used for structure
3789 and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
3791 If not defined, this defaults to the value 1.
3793 @findex STRUCT_VALUE_REGNUM
3794 @item STRUCT_VALUE_REGNUM
3795 If the structure value address is passed in a register, then
3796 @code{STRUCT_VALUE_REGNUM} should be the number of that register.
3798 @findex STRUCT_VALUE
3800 If the structure value address is not passed in a register, define
3801 @code{STRUCT_VALUE} as an expression returning an RTX for the place
3802 where the address is passed. If it returns 0, the address is passed as
3803 an ``invisible'' first argument.
3805 @findex STRUCT_VALUE_INCOMING_REGNUM
3806 @item STRUCT_VALUE_INCOMING_REGNUM
3807 On some architectures the place where the structure value address
3808 is found by the called function is not the same place that the
3809 caller put it. This can be due to register windows, or it could
3810 be because the function prologue moves it to a different place.
3812 If the incoming location of the structure value address is in a
3813 register, define this macro as the register number.
3815 @findex STRUCT_VALUE_INCOMING
3816 @item STRUCT_VALUE_INCOMING
3817 If the incoming location is not a register, then you should define
3818 @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the
3819 called function should find the value. If it should find the value on
3820 the stack, define this to create a @code{mem} which refers to the frame
3821 pointer. A definition of 0 means that the address is passed as an
3822 ``invisible'' first argument.
3824 @findex PCC_STATIC_STRUCT_RETURN
3825 @item PCC_STATIC_STRUCT_RETURN
3826 Define this macro if the usual system convention on the target machine
3827 for returning structures and unions is for the called function to return
3828 the address of a static variable containing the value.
3830 Do not define this if the usual system convention is for the caller to
3831 pass an address to the subroutine.
3833 This macro has effect in @option{-fpcc-struct-return} mode, but it does
3834 nothing when you use @option{-freg-struct-return} mode.
3838 @subsection Caller-Saves Register Allocation
3840 If you enable it, GCC can save registers around function calls. This
3841 makes it possible to use call-clobbered registers to hold variables that
3842 must live across calls.
3845 @findex DEFAULT_CALLER_SAVES
3846 @item DEFAULT_CALLER_SAVES
3847 Define this macro if function calls on the target machine do not preserve
3848 any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
3849 for all registers. When defined, this macro enables @option{-fcaller-saves}
3850 by default for all optimization levels. It has no effect for optimization
3851 levels 2 and higher, where @option{-fcaller-saves} is the default.
3853 @findex CALLER_SAVE_PROFITABLE
3854 @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
3855 A C expression to determine whether it is worthwhile to consider placing
3856 a pseudo-register in a call-clobbered hard register and saving and
3857 restoring it around each function call. The expression should be 1 when
3858 this is worth doing, and 0 otherwise.
3860 If you don't define this macro, a default is used which is good on most
3861 machines: @code{4 * @var{calls} < @var{refs}}.
3863 @findex HARD_REGNO_CALLER_SAVE_MODE
3864 @item HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3865 A C expression specifying which mode is required for saving @var{nregs}
3866 of a pseudo-register in call-clobbered hard register @var{regno}. If
3867 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3868 returned. For most machines this macro need not be defined since GCC
3869 will select the smallest suitable mode.
3872 @node Function Entry
3873 @subsection Function Entry and Exit
3874 @cindex function entry and exit
3878 This section describes the macros that output function entry
3879 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3881 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
3882 If defined, a function that outputs the assembler code for entry to a
3883 function. The prologue is responsible for setting up the stack frame,
3884 initializing the frame pointer register, saving registers that must be
3885 saved, and allocating @var{size} additional bytes of storage for the
3886 local variables. @var{size} is an integer. @var{file} is a stdio
3887 stream to which the assembler code should be output.
3889 The label for the beginning of the function need not be output by this
3890 macro. That has already been done when the macro is run.
3892 @findex regs_ever_live
3893 To determine which registers to save, the macro can refer to the array
3894 @code{regs_ever_live}: element @var{r} is nonzero if hard register
3895 @var{r} is used anywhere within the function. This implies the function
3896 prologue should save register @var{r}, provided it is not one of the
3897 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
3898 @code{regs_ever_live}.)
3900 On machines that have ``register windows'', the function entry code does
3901 not save on the stack the registers that are in the windows, even if
3902 they are supposed to be preserved by function calls; instead it takes
3903 appropriate steps to ``push'' the register stack, if any non-call-used
3904 registers are used in the function.
3906 @findex frame_pointer_needed
3907 On machines where functions may or may not have frame-pointers, the
3908 function entry code must vary accordingly; it must set up the frame
3909 pointer if one is wanted, and not otherwise. To determine whether a
3910 frame pointer is in wanted, the macro can refer to the variable
3911 @code{frame_pointer_needed}. The variable's value will be 1 at run
3912 time in a function that needs a frame pointer. @xref{Elimination}.
3914 The function entry code is responsible for allocating any stack space
3915 required for the function. This stack space consists of the regions
3916 listed below. In most cases, these regions are allocated in the
3917 order listed, with the last listed region closest to the top of the
3918 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
3919 the highest address if it is not defined). You can use a different order
3920 for a machine if doing so is more convenient or required for
3921 compatibility reasons. Except in cases where required by standard
3922 or by a debugger, there is no reason why the stack layout used by GCC
3923 need agree with that used by other compilers for a machine.
3926 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
3927 If defined, a function that outputs assembler code at the end of a
3928 prologue. This should be used when the function prologue is being
3929 emitted as RTL, and you have some extra assembler that needs to be
3930 emitted. @xref{prologue instruction pattern}.
3933 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
3934 If defined, a function that outputs assembler code at the start of an
3935 epilogue. This should be used when the function epilogue is being
3936 emitted as RTL, and you have some extra assembler that needs to be
3937 emitted. @xref{epilogue instruction pattern}.
3940 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
3941 If defined, a function that outputs the assembler code for exit from a
3942 function. The epilogue is responsible for restoring the saved
3943 registers and stack pointer to their values when the function was
3944 called, and returning control to the caller. This macro takes the
3945 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
3946 registers to restore are determined from @code{regs_ever_live} and
3947 @code{CALL_USED_REGISTERS} in the same way.
3949 On some machines, there is a single instruction that does all the work
3950 of returning from the function. On these machines, give that
3951 instruction the name @samp{return} and do not define the macro
3952 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
3954 Do not define a pattern named @samp{return} if you want the
3955 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
3956 switches to control whether return instructions or epilogues are used,
3957 define a @samp{return} pattern with a validity condition that tests the
3958 target switches appropriately. If the @samp{return} pattern's validity
3959 condition is false, epilogues will be used.
3961 On machines where functions may or may not have frame-pointers, the
3962 function exit code must vary accordingly. Sometimes the code for these
3963 two cases is completely different. To determine whether a frame pointer
3964 is wanted, the macro can refer to the variable
3965 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
3966 a function that needs a frame pointer.
3968 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
3969 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
3970 The C variable @code{current_function_is_leaf} is nonzero for such a
3971 function. @xref{Leaf Functions}.
3973 On some machines, some functions pop their arguments on exit while
3974 others leave that for the caller to do. For example, the 68020 when
3975 given @option{-mrtd} pops arguments in functions that take a fixed
3976 number of arguments.
3978 @findex current_function_pops_args
3979 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
3980 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
3981 needs to know what was decided. The variable that is called
3982 @code{current_function_pops_args} is the number of bytes of its
3983 arguments that a function should pop. @xref{Scalar Return}.
3984 @c what is the "its arguments" in the above sentence referring to, pray
3985 @c tell? --mew 5feb93
3992 @findex current_function_pretend_args_size
3993 A region of @code{current_function_pretend_args_size} bytes of
3994 uninitialized space just underneath the first argument arriving on the
3995 stack. (This may not be at the very start of the allocated stack region
3996 if the calling sequence has pushed anything else since pushing the stack
3997 arguments. But usually, on such machines, nothing else has been pushed
3998 yet, because the function prologue itself does all the pushing.) This
3999 region is used on machines where an argument may be passed partly in
4000 registers and partly in memory, and, in some cases to support the
4001 features in @code{<varargs.h>} and @code{<stdarg.h>}.
4004 An area of memory used to save certain registers used by the function.
4005 The size of this area, which may also include space for such things as
4006 the return address and pointers to previous stack frames, is
4007 machine-specific and usually depends on which registers have been used
4008 in the function. Machines with register windows often do not require
4012 A region of at least @var{size} bytes, possibly rounded up to an allocation
4013 boundary, to contain the local variables of the function. On some machines,
4014 this region and the save area may occur in the opposite order, with the
4015 save area closer to the top of the stack.
4018 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4019 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4020 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4021 argument lists of the function. @xref{Stack Arguments}.
4024 Normally, it is necessary for the macros
4025 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4026 @code{TARGET_ASM_FUNCTION_EPILOGUE} to treat leaf functions specially.
4027 The C variable @code{current_function_is_leaf} is nonzero for such a
4030 @findex EXIT_IGNORE_STACK
4031 @item EXIT_IGNORE_STACK
4032 Define this macro as a C expression that is nonzero if the return
4033 instruction or the function epilogue ignores the value of the stack
4034 pointer; in other words, if it is safe to delete an instruction to
4035 adjust the stack pointer before a return from the function.
4037 Note that this macro's value is relevant only for functions for which
4038 frame pointers are maintained. It is never safe to delete a final
4039 stack adjustment in a function that has no frame pointer, and the
4040 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4042 @findex EPILOGUE_USES
4043 @item EPILOGUE_USES (@var{regno})
4044 Define this macro as a C expression that is nonzero for registers that are
4045 used by the epilogue or the @samp{return} pattern. The stack and frame
4046 pointer registers are already be assumed to be used as needed.
4049 @item EH_USES (@var{regno})
4050 Define this macro as a C expression that is nonzero for registers that are
4051 used by the exception handling mechanism, and so should be considered live
4052 on entry to an exception edge.
4054 @findex DELAY_SLOTS_FOR_EPILOGUE
4055 @item DELAY_SLOTS_FOR_EPILOGUE
4056 Define this macro if the function epilogue contains delay slots to which
4057 instructions from the rest of the function can be ``moved''. The
4058 definition should be a C expression whose value is an integer
4059 representing the number of delay slots there.
4061 @findex ELIGIBLE_FOR_EPILOGUE_DELAY
4062 @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4063 A C expression that returns 1 if @var{insn} can be placed in delay
4064 slot number @var{n} of the epilogue.
4066 The argument @var{n} is an integer which identifies the delay slot now
4067 being considered (since different slots may have different rules of
4068 eligibility). It is never negative and is always less than the number
4069 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4070 If you reject a particular insn for a given delay slot, in principle, it
4071 may be reconsidered for a subsequent delay slot. Also, other insns may
4072 (at least in principle) be considered for the so far unfilled delay
4075 @findex current_function_epilogue_delay_list
4076 @findex final_scan_insn
4077 The insns accepted to fill the epilogue delay slots are put in an RTL
4078 list made with @code{insn_list} objects, stored in the variable
4079 @code{current_function_epilogue_delay_list}. The insn for the first
4080 delay slot comes first in the list. Your definition of the macro
4081 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4082 outputting the insns in this list, usually by calling
4083 @code{final_scan_insn}.
4085 You need not define this macro if you did not define
4086 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4088 @findex ASM_OUTPUT_MI_THUNK
4089 @item ASM_OUTPUT_MI_THUNK (@var{file}, @var{thunk_fndecl}, @var{delta}, @var{function})
4090 A C compound statement that outputs the assembler code for a thunk
4091 function, used to implement C++ virtual function calls with multiple
4092 inheritance. The thunk acts as a wrapper around a virtual function,
4093 adjusting the implicit object parameter before handing control off to
4096 First, emit code to add the integer @var{delta} to the location that
4097 contains the incoming first argument. Assume that this argument
4098 contains a pointer, and is the one used to pass the @code{this} pointer
4099 in C++. This is the incoming argument @emph{before} the function prologue,
4100 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4101 all other incoming arguments.
4103 After the addition, emit code to jump to @var{function}, which is a
4104 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4105 not touch the return address. Hence returning from @var{FUNCTION} will
4106 return to whoever called the current @samp{thunk}.
4108 The effect must be as if @var{function} had been called directly with
4109 the adjusted first argument. This macro is responsible for emitting all
4110 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4111 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4113 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4114 have already been extracted from it.) It might possibly be useful on
4115 some targets, but probably not.
4117 If you do not define this macro, the target-independent code in the C++
4118 front end will generate a less efficient heavyweight thunk that calls
4119 @var{function} instead of jumping to it. The generic approach does
4120 not support varargs.
4124 @subsection Generating Code for Profiling
4125 @cindex profiling, code generation
4127 These macros will help you generate code for profiling.
4130 @findex FUNCTION_PROFILER
4131 @item FUNCTION_PROFILER (@var{file}, @var{labelno})
4132 A C statement or compound statement to output to @var{file} some
4133 assembler code to call the profiling subroutine @code{mcount}.
4136 The details of how @code{mcount} expects to be called are determined by
4137 your operating system environment, not by GCC@. To figure them out,
4138 compile a small program for profiling using the system's installed C
4139 compiler and look at the assembler code that results.
4141 Older implementations of @code{mcount} expect the address of a counter
4142 variable to be loaded into some register. The name of this variable is
4143 @samp{LP} followed by the number @var{labelno}, so you would generate
4144 the name using @samp{LP%d} in a @code{fprintf}.
4146 @findex PROFILE_HOOK
4148 A C statement or compound statement to output to @var{file} some assembly
4149 code to call the profiling subroutine @code{mcount} even the target does
4150 not support profiling.
4152 @findex NO_PROFILE_COUNTERS
4153 @item NO_PROFILE_COUNTERS
4154 Define this macro if the @code{mcount} subroutine on your system does
4155 not need a counter variable allocated for each function. This is true
4156 for almost all modern implementations. If you define this macro, you
4157 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
4159 @findex PROFILE_BEFORE_PROLOGUE
4160 @item PROFILE_BEFORE_PROLOGUE
4161 Define this macro if the code for function profiling should come before
4162 the function prologue. Normally, the profiling code comes after.
4166 @subsection Permitting tail calls
4170 @findex FUNCTION_OK_FOR_SIBCALL
4171 @item FUNCTION_OK_FOR_SIBCALL (@var{decl})
4172 A C expression that evaluates to true if it is ok to perform a sibling
4173 call to @var{decl} from the current function.
4175 It is not uncommon for limitations of calling conventions to prevent
4176 tail calls to functions outside the current unit of translation, or
4177 during PIC compilation. Use this macro to enforce these restrictions,
4178 as the @code{sibcall} md pattern can not fail, or fall over to a
4183 @section Implementing the Varargs Macros
4184 @cindex varargs implementation
4186 GCC comes with an implementation of @code{<varargs.h>} and
4187 @code{<stdarg.h>} that work without change on machines that pass arguments
4188 on the stack. Other machines require their own implementations of
4189 varargs, and the two machine independent header files must have
4190 conditionals to include it.
4192 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4193 the calling convention for @code{va_start}. The traditional
4194 implementation takes just one argument, which is the variable in which
4195 to store the argument pointer. The ISO implementation of
4196 @code{va_start} takes an additional second argument. The user is
4197 supposed to write the last named argument of the function here.
4199 However, @code{va_start} should not use this argument. The way to find
4200 the end of the named arguments is with the built-in functions described
4204 @findex __builtin_saveregs
4205 @item __builtin_saveregs ()
4206 Use this built-in function to save the argument registers in memory so
4207 that the varargs mechanism can access them. Both ISO and traditional
4208 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4209 you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
4211 On some machines, @code{__builtin_saveregs} is open-coded under the
4212 control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
4213 it calls a routine written in assembler language, found in
4216 Code generated for the call to @code{__builtin_saveregs} appears at the
4217 beginning of the function, as opposed to where the call to
4218 @code{__builtin_saveregs} is written, regardless of what the code is.
4219 This is because the registers must be saved before the function starts
4220 to use them for its own purposes.
4221 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4224 @findex __builtin_args_info
4225 @item __builtin_args_info (@var{category})
4226 Use this built-in function to find the first anonymous arguments in
4229 In general, a machine may have several categories of registers used for
4230 arguments, each for a particular category of data types. (For example,
4231 on some machines, floating-point registers are used for floating-point
4232 arguments while other arguments are passed in the general registers.)
4233 To make non-varargs functions use the proper calling convention, you
4234 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4235 registers in each category have been used so far
4237 @code{__builtin_args_info} accesses the same data structure of type
4238 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4239 with it, with @var{category} specifying which word to access. Thus, the
4240 value indicates the first unused register in a given category.
4242 Normally, you would use @code{__builtin_args_info} in the implementation
4243 of @code{va_start}, accessing each category just once and storing the
4244 value in the @code{va_list} object. This is because @code{va_list} will
4245 have to update the values, and there is no way to alter the
4246 values accessed by @code{__builtin_args_info}.
4248 @findex __builtin_next_arg
4249 @item __builtin_next_arg (@var{lastarg})
4250 This is the equivalent of @code{__builtin_args_info}, for stack
4251 arguments. It returns the address of the first anonymous stack
4252 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4253 returns the address of the location above the first anonymous stack
4254 argument. Use it in @code{va_start} to initialize the pointer for
4255 fetching arguments from the stack. Also use it in @code{va_start} to
4256 verify that the second parameter @var{lastarg} is the last named argument
4257 of the current function.
4259 @findex __builtin_classify_type
4260 @item __builtin_classify_type (@var{object})
4261 Since each machine has its own conventions for which data types are
4262 passed in which kind of register, your implementation of @code{va_arg}
4263 has to embody these conventions. The easiest way to categorize the
4264 specified data type is to use @code{__builtin_classify_type} together
4265 with @code{sizeof} and @code{__alignof__}.
4267 @code{__builtin_classify_type} ignores the value of @var{object},
4268 considering only its data type. It returns an integer describing what
4269 kind of type that is---integer, floating, pointer, structure, and so on.
4271 The file @file{typeclass.h} defines an enumeration that you can use to
4272 interpret the values of @code{__builtin_classify_type}.
4275 These machine description macros help implement varargs:
4278 @findex EXPAND_BUILTIN_SAVEREGS
4279 @item EXPAND_BUILTIN_SAVEREGS ()
4280 If defined, is a C expression that produces the machine-specific code
4281 for a call to @code{__builtin_saveregs}. This code will be moved to the
4282 very beginning of the function, before any parameter access are made.
4283 The return value of this function should be an RTX that contains the
4284 value to use as the return of @code{__builtin_saveregs}.
4286 @findex SETUP_INCOMING_VARARGS
4287 @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type}, @var{pretend_args_size}, @var{second_time})
4288 This macro offers an alternative to using @code{__builtin_saveregs} and
4289 defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the
4290 anonymous register arguments into the stack so that all the arguments
4291 appear to have been passed consecutively on the stack. Once this is
4292 done, you can use the standard implementation of varargs that works for
4293 machines that pass all their arguments on the stack.
4295 The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data
4296 structure, containing the values that are obtained after processing the
4297 named arguments. The arguments @var{mode} and @var{type} describe the
4298 last named argument---its machine mode and its data type as a tree node.
4300 The macro implementation should do two things: first, push onto the
4301 stack all the argument registers @emph{not} used for the named
4302 arguments, and second, store the size of the data thus pushed into the
4303 @code{int}-valued variable whose name is supplied as the argument
4304 @var{pretend_args_size}. The value that you store here will serve as
4305 additional offset for setting up the stack frame.
4307 Because you must generate code to push the anonymous arguments at
4308 compile time without knowing their data types,
4309 @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just
4310 a single category of argument register and use it uniformly for all data
4313 If the argument @var{second_time} is nonzero, it means that the
4314 arguments of the function are being analyzed for the second time. This
4315 happens for an inline function, which is not actually compiled until the
4316 end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should
4317 not generate any instructions in this case.
4319 @findex STRICT_ARGUMENT_NAMING
4320 @item STRICT_ARGUMENT_NAMING
4321 Define this macro to be a nonzero value if the location where a function
4322 argument is passed depends on whether or not it is a named argument.
4324 This macro controls how the @var{named} argument to @code{FUNCTION_ARG}
4325 is set for varargs and stdarg functions. If this macro returns a
4326 nonzero value, the @var{named} argument is always true for named
4327 arguments, and false for unnamed arguments. If it returns a value of
4328 zero, but @code{SETUP_INCOMING_VARARGS} is defined, then all arguments
4329 are treated as named. Otherwise, all named arguments except the last
4330 are treated as named.
4332 You need not define this macro if it always returns zero.
4334 @findex PRETEND_OUTGOING_VARARGS_NAMED
4335 @item PRETEND_OUTGOING_VARARGS_NAMED
4336 If you need to conditionally change ABIs so that one works with
4337 @code{SETUP_INCOMING_VARARGS}, but the other works like neither
4338 @code{SETUP_INCOMING_VARARGS} nor @code{STRICT_ARGUMENT_NAMING} was
4339 defined, then define this macro to return nonzero if
4340 @code{SETUP_INCOMING_VARARGS} is used, zero otherwise.
4341 Otherwise, you should not define this macro.
4345 @section Trampolines for Nested Functions
4346 @cindex trampolines for nested functions
4347 @cindex nested functions, trampolines for
4349 A @dfn{trampoline} is a small piece of code that is created at run time
4350 when the address of a nested function is taken. It normally resides on
4351 the stack, in the stack frame of the containing function. These macros
4352 tell GCC how to generate code to allocate and initialize a
4355 The instructions in the trampoline must do two things: load a constant
4356 address into the static chain register, and jump to the real address of
4357 the nested function. On CISC machines such as the m68k, this requires
4358 two instructions, a move immediate and a jump. Then the two addresses
4359 exist in the trampoline as word-long immediate operands. On RISC
4360 machines, it is often necessary to load each address into a register in
4361 two parts. Then pieces of each address form separate immediate
4364 The code generated to initialize the trampoline must store the variable
4365 parts---the static chain value and the function address---into the
4366 immediate operands of the instructions. On a CISC machine, this is
4367 simply a matter of copying each address to a memory reference at the
4368 proper offset from the start of the trampoline. On a RISC machine, it
4369 may be necessary to take out pieces of the address and store them
4373 @findex TRAMPOLINE_TEMPLATE
4374 @item TRAMPOLINE_TEMPLATE (@var{file})
4375 A C statement to output, on the stream @var{file}, assembler code for a
4376 block of data that contains the constant parts of a trampoline. This
4377 code should not include a label---the label is taken care of
4380 If you do not define this macro, it means no template is needed
4381 for the target. Do not define this macro on systems where the block move
4382 code to copy the trampoline into place would be larger than the code
4383 to generate it on the spot.
4385 @findex TRAMPOLINE_SECTION
4386 @item TRAMPOLINE_SECTION
4387 The name of a subroutine to switch to the section in which the
4388 trampoline template is to be placed (@pxref{Sections}). The default is
4389 a value of @samp{readonly_data_section}, which places the trampoline in
4390 the section containing read-only data.
4392 @findex TRAMPOLINE_SIZE
4393 @item TRAMPOLINE_SIZE
4394 A C expression for the size in bytes of the trampoline, as an integer.
4396 @findex TRAMPOLINE_ALIGNMENT
4397 @item TRAMPOLINE_ALIGNMENT
4398 Alignment required for trampolines, in bits.
4400 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4401 is used for aligning trampolines.
4403 @findex INITIALIZE_TRAMPOLINE
4404 @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4405 A C statement to initialize the variable parts of a trampoline.
4406 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4407 an RTX for the address of the nested function; @var{static_chain} is an
4408 RTX for the static chain value that should be passed to the function
4411 @findex TRAMPOLINE_ADJUST_ADDRESS
4412 @item TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4413 A C statement that should perform any machine-specific adjustment in
4414 the address of the trampoline. Its argument contains the address that
4415 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4416 used for a function call should be different from the address in which
4417 the template was stored, the different address should be assigned to
4418 @var{addr}. If this macro is not defined, @var{addr} will be used for
4421 @findex ALLOCATE_TRAMPOLINE
4422 @item ALLOCATE_TRAMPOLINE (@var{fp})
4423 A C expression to allocate run-time space for a trampoline. The
4424 expression value should be an RTX representing a memory reference to the
4425 space for the trampoline.
4427 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4428 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4429 If this macro is not defined, by default the trampoline is allocated as
4430 a stack slot. This default is right for most machines. The exceptions
4431 are machines where it is impossible to execute instructions in the stack
4432 area. On such machines, you may have to implement a separate stack,
4433 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4434 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4436 @var{fp} points to a data structure, a @code{struct function}, which
4437 describes the compilation status of the immediate containing function of
4438 the function which the trampoline is for. Normally (when
4439 @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the
4440 trampoline is in the stack frame of this containing function. Other
4441 allocation strategies probably must do something analogous with this
4445 Implementing trampolines is difficult on many machines because they have
4446 separate instruction and data caches. Writing into a stack location
4447 fails to clear the memory in the instruction cache, so when the program
4448 jumps to that location, it executes the old contents.
4450 Here are two possible solutions. One is to clear the relevant parts of
4451 the instruction cache whenever a trampoline is set up. The other is to
4452 make all trampolines identical, by having them jump to a standard
4453 subroutine. The former technique makes trampoline execution faster; the
4454 latter makes initialization faster.
4456 To clear the instruction cache when a trampoline is initialized, define
4457 the following macros which describe the shape of the cache.
4460 @findex INSN_CACHE_SIZE
4461 @item INSN_CACHE_SIZE
4462 The total size in bytes of the cache.
4464 @findex INSN_CACHE_LINE_WIDTH
4465 @item INSN_CACHE_LINE_WIDTH
4466 The length in bytes of each cache line. The cache is divided into cache
4467 lines which are disjoint slots, each holding a contiguous chunk of data
4468 fetched from memory. Each time data is brought into the cache, an
4469 entire line is read at once. The data loaded into a cache line is
4470 always aligned on a boundary equal to the line size.
4472 @findex INSN_CACHE_DEPTH
4473 @item INSN_CACHE_DEPTH
4474 The number of alternative cache lines that can hold any particular memory
4478 Alternatively, if the machine has system calls or instructions to clear
4479 the instruction cache directly, you can define the following macro.
4482 @findex CLEAR_INSN_CACHE
4483 @item CLEAR_INSN_CACHE (@var{beg}, @var{end})
4484 If defined, expands to a C expression clearing the @emph{instruction
4485 cache} in the specified interval. If it is not defined, and the macro
4486 @code{INSN_CACHE_SIZE} is defined, some generic code is generated to clear the
4487 cache. The definition of this macro would typically be a series of
4488 @code{asm} statements. Both @var{beg} and @var{end} are both pointer
4492 To use a standard subroutine, define the following macro. In addition,
4493 you must make sure that the instructions in a trampoline fill an entire
4494 cache line with identical instructions, or else ensure that the
4495 beginning of the trampoline code is always aligned at the same point in
4496 its cache line. Look in @file{m68k.h} as a guide.
4499 @findex TRANSFER_FROM_TRAMPOLINE
4500 @item TRANSFER_FROM_TRAMPOLINE
4501 Define this macro if trampolines need a special subroutine to do their
4502 work. The macro should expand to a series of @code{asm} statements
4503 which will be compiled with GCC@. They go in a library function named
4504 @code{__transfer_from_trampoline}.
4506 If you need to avoid executing the ordinary prologue code of a compiled
4507 C function when you jump to the subroutine, you can do so by placing a
4508 special label of your own in the assembler code. Use one @code{asm}
4509 statement to generate an assembler label, and another to make the label
4510 global. Then trampolines can use that label to jump directly to your
4511 special assembler code.
4515 @section Implicit Calls to Library Routines
4516 @cindex library subroutine names
4517 @cindex @file{libgcc.a}
4519 @c prevent bad page break with this line
4520 Here is an explanation of implicit calls to library routines.
4523 @findex MULSI3_LIBCALL
4524 @item MULSI3_LIBCALL
4525 A C string constant giving the name of the function to call for
4526 multiplication of one signed full-word by another. If you do not
4527 define this macro, the default name is used, which is @code{__mulsi3},
4528 a function defined in @file{libgcc.a}.
4530 @findex DIVSI3_LIBCALL
4531 @item DIVSI3_LIBCALL
4532 A C string constant giving the name of the function to call for
4533 division of one signed full-word by another. If you do not define
4534 this macro, the default name is used, which is @code{__divsi3}, a
4535 function defined in @file{libgcc.a}.
4537 @findex UDIVSI3_LIBCALL
4538 @item UDIVSI3_LIBCALL
4539 A C string constant giving the name of the function to call for
4540 division of one unsigned full-word by another. If you do not define
4541 this macro, the default name is used, which is @code{__udivsi3}, a
4542 function defined in @file{libgcc.a}.
4544 @findex MODSI3_LIBCALL
4545 @item MODSI3_LIBCALL
4546 A C string constant giving the name of the function to call for the
4547 remainder in division of one signed full-word by another. If you do
4548 not define this macro, the default name is used, which is
4549 @code{__modsi3}, a function defined in @file{libgcc.a}.
4551 @findex UMODSI3_LIBCALL
4552 @item UMODSI3_LIBCALL
4553 A C string constant giving the name of the function to call for the
4554 remainder in division of one unsigned full-word by another. If you do
4555 not define this macro, the default name is used, which is
4556 @code{__umodsi3}, a function defined in @file{libgcc.a}.
4558 @findex MULDI3_LIBCALL
4559 @item MULDI3_LIBCALL
4560 A C string constant giving the name of the function to call for
4561 multiplication of one signed double-word by another. If you do not
4562 define this macro, the default name is used, which is @code{__muldi3},
4563 a function defined in @file{libgcc.a}.
4565 @findex DIVDI3_LIBCALL
4566 @item DIVDI3_LIBCALL
4567 A C string constant giving the name of the function to call for
4568 division of one signed double-word by another. If you do not define
4569 this macro, the default name is used, which is @code{__divdi3}, a
4570 function defined in @file{libgcc.a}.
4572 @findex UDIVDI3_LIBCALL
4573 @item UDIVDI3_LIBCALL
4574 A C string constant giving the name of the function to call for
4575 division of one unsigned full-word by another. If you do not define
4576 this macro, the default name is used, which is @code{__udivdi3}, a
4577 function defined in @file{libgcc.a}.
4579 @findex MODDI3_LIBCALL
4580 @item MODDI3_LIBCALL
4581 A C string constant giving the name of the function to call for the
4582 remainder in division of one signed double-word by another. If you do
4583 not define this macro, the default name is used, which is
4584 @code{__moddi3}, a function defined in @file{libgcc.a}.
4586 @findex UMODDI3_LIBCALL
4587 @item UMODDI3_LIBCALL
4588 A C string constant giving the name of the function to call for the
4589 remainder in division of one unsigned full-word by another. If you do
4590 not define this macro, the default name is used, which is
4591 @code{__umoddi3}, a function defined in @file{libgcc.a}.
4593 @findex INIT_TARGET_OPTABS
4594 @item INIT_TARGET_OPTABS
4595 Define this macro as a C statement that declares additional library
4596 routines renames existing ones. @code{init_optabs} calls this macro after
4597 initializing all the normal library routines.
4599 @findex FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4600 @item FLOAT_LIB_COMPARE_RETURNS_BOOL
4601 Define this macro as a C statement that returns nonzero if a call to
4602 the floating point comparison library function will return a boolean
4603 value that indicates the result of the comparison. It should return
4604 zero if one of gcc's own libgcc functions is called.
4606 Most ports don't need to define this macro.
4609 @cindex @code{EDOM}, implicit usage
4611 The value of @code{EDOM} on the target machine, as a C integer constant
4612 expression. If you don't define this macro, GCC does not attempt to
4613 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4614 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4617 If you do not define @code{TARGET_EDOM}, then compiled code reports
4618 domain errors by calling the library function and letting it report the
4619 error. If mathematical functions on your system use @code{matherr} when
4620 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4621 that @code{matherr} is used normally.
4623 @findex GEN_ERRNO_RTX
4624 @cindex @code{errno}, implicit usage
4626 Define this macro as a C expression to create an rtl expression that
4627 refers to the global ``variable'' @code{errno}. (On certain systems,
4628 @code{errno} may not actually be a variable.) If you don't define this
4629 macro, a reasonable default is used.
4631 @findex TARGET_MEM_FUNCTIONS
4632 @cindex @code{bcopy}, implicit usage
4633 @cindex @code{memcpy}, implicit usage
4634 @cindex @code{memmove}, implicit usage
4635 @cindex @code{bzero}, implicit usage
4636 @cindex @code{memset}, implicit usage
4637 @item TARGET_MEM_FUNCTIONS
4638 Define this macro if GCC should generate calls to the ISO C
4639 (and System V) library functions @code{memcpy}, @code{memmove} and
4640 @code{memset} rather than the BSD functions @code{bcopy} and @code{bzero}.
4642 @findex LIBGCC_NEEDS_DOUBLE
4643 @item LIBGCC_NEEDS_DOUBLE
4644 Define this macro if @code{float} arguments cannot be passed to library
4645 routines (so they must be converted to @code{double}). This macro
4646 affects both how library calls are generated and how the library
4647 routines in @file{libgcc.a} accept their arguments. It is useful on
4648 machines where floating and fixed point arguments are passed
4649 differently, such as the i860.
4651 @findex NEXT_OBJC_RUNTIME
4652 @item NEXT_OBJC_RUNTIME
4653 Define this macro to generate code for Objective-C message sending using
4654 the calling convention of the NeXT system. This calling convention
4655 involves passing the object, the selector and the method arguments all
4656 at once to the method-lookup library function.
4658 The default calling convention passes just the object and the selector
4659 to the lookup function, which returns a pointer to the method.
4662 @node Addressing Modes
4663 @section Addressing Modes
4664 @cindex addressing modes
4666 @c prevent bad page break with this line
4667 This is about addressing modes.
4670 @findex HAVE_PRE_INCREMENT
4671 @findex HAVE_PRE_DECREMENT
4672 @findex HAVE_POST_INCREMENT
4673 @findex HAVE_POST_DECREMENT
4674 @item HAVE_PRE_INCREMENT
4675 @itemx HAVE_PRE_DECREMENT
4676 @itemx HAVE_POST_INCREMENT
4677 @itemx HAVE_POST_DECREMENT
4678 A C expression that is nonzero if the machine supports pre-increment,
4679 pre-decrement, post-increment, or post-decrement addressing respectively.
4681 @findex HAVE_POST_MODIFY_DISP
4682 @findex HAVE_PRE_MODIFY_DISP
4683 @item HAVE_PRE_MODIFY_DISP
4684 @itemx HAVE_POST_MODIFY_DISP
4685 A C expression that is nonzero if the machine supports pre- or
4686 post-address side-effect generation involving constants other than
4687 the size of the memory operand.
4689 @findex HAVE_POST_MODIFY_REG
4690 @findex HAVE_PRE_MODIFY_REG
4691 @item HAVE_PRE_MODIFY_REG
4692 @itemx HAVE_POST_MODIFY_REG
4693 A C expression that is nonzero if the machine supports pre- or
4694 post-address side-effect generation involving a register displacement.
4696 @findex CONSTANT_ADDRESS_P
4697 @item CONSTANT_ADDRESS_P (@var{x})
4698 A C expression that is 1 if the RTX @var{x} is a constant which
4699 is a valid address. On most machines, this can be defined as
4700 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4701 in which constant addresses are supported.
4704 @code{CONSTANT_P} accepts integer-values expressions whose values are
4705 not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and
4706 @code{high} expressions and @code{const} arithmetic expressions, in
4707 addition to @code{const_int} and @code{const_double} expressions.
4709 @findex MAX_REGS_PER_ADDRESS
4710 @item MAX_REGS_PER_ADDRESS
4711 A number, the maximum number of registers that can appear in a valid
4712 memory address. Note that it is up to you to specify a value equal to
4713 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4716 @findex GO_IF_LEGITIMATE_ADDRESS
4717 @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4718 A C compound statement with a conditional @code{goto @var{label};}
4719 executed if @var{x} (an RTX) is a legitimate memory address on the
4720 target machine for a memory operand of mode @var{mode}.
4722 It usually pays to define several simpler macros to serve as
4723 subroutines for this one. Otherwise it may be too complicated to
4726 This macro must exist in two variants: a strict variant and a
4727 non-strict one. The strict variant is used in the reload pass. It
4728 must be defined so that any pseudo-register that has not been
4729 allocated a hard register is considered a memory reference. In
4730 contexts where some kind of register is required, a pseudo-register
4731 with no hard register must be rejected.
4733 The non-strict variant is used in other passes. It must be defined to
4734 accept all pseudo-registers in every context where some kind of
4735 register is required.
4737 @findex REG_OK_STRICT
4738 Compiler source files that want to use the strict variant of this
4739 macro define the macro @code{REG_OK_STRICT}. You should use an
4740 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4741 in that case and the non-strict variant otherwise.
4743 Subroutines to check for acceptable registers for various purposes (one
4744 for base registers, one for index registers, and so on) are typically
4745 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4746 Then only these subroutine macros need have two variants; the higher
4747 levels of macros may be the same whether strict or not.
4749 Normally, constant addresses which are the sum of a @code{symbol_ref}
4750 and an integer are stored inside a @code{const} RTX to mark them as
4751 constant. Therefore, there is no need to recognize such sums
4752 specifically as legitimate addresses. Normally you would simply
4753 recognize any @code{const} as legitimate.
4755 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4756 sums that are not marked with @code{const}. It assumes that a naked
4757 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4758 naked constant sums as illegitimate addresses, so that none of them will
4759 be given to @code{PRINT_OPERAND_ADDRESS}.
4761 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
4762 On some machines, whether a symbolic address is legitimate depends on
4763 the section that the address refers to. On these machines, define the
4764 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
4765 into the @code{symbol_ref}, and then check for it here. When you see a
4766 @code{const}, you will have to look inside it to find the
4767 @code{symbol_ref} in order to determine the section. @xref{Assembler
4770 @findex saveable_obstack
4771 The best way to modify the name string is by adding text to the
4772 beginning, with suitable punctuation to prevent any ambiguity. Allocate
4773 the new name in @code{saveable_obstack}. You will have to modify
4774 @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and
4775 output the name accordingly, and define @code{TARGET_STRIP_NAME_ENCODING}
4776 to access the original name string.
4778 You can check the information stored here into the @code{symbol_ref} in
4779 the definitions of the macros @code{GO_IF_LEGITIMATE_ADDRESS} and
4780 @code{PRINT_OPERAND_ADDRESS}.
4782 @findex REG_OK_FOR_BASE_P
4783 @item REG_OK_FOR_BASE_P (@var{x})
4784 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4785 RTX) is valid for use as a base register. For hard registers, it
4786 should always accept those which the hardware permits and reject the
4787 others. Whether the macro accepts or rejects pseudo registers must be
4788 controlled by @code{REG_OK_STRICT} as described above. This usually
4789 requires two variant definitions, of which @code{REG_OK_STRICT}
4790 controls the one actually used.
4792 @findex REG_MODE_OK_FOR_BASE_P
4793 @item REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4794 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4795 that expression may examine the mode of the memory reference in
4796 @var{mode}. You should define this macro if the mode of the memory
4797 reference affects whether a register may be used as a base register. If
4798 you define this macro, the compiler will use it instead of
4799 @code{REG_OK_FOR_BASE_P}.
4801 @findex REG_OK_FOR_INDEX_P
4802 @item REG_OK_FOR_INDEX_P (@var{x})
4803 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4804 RTX) is valid for use as an index register.
4806 The difference between an index register and a base register is that
4807 the index register may be scaled. If an address involves the sum of
4808 two registers, neither one of them scaled, then either one may be
4809 labeled the ``base'' and the other the ``index''; but whichever
4810 labeling is used must fit the machine's constraints of which registers
4811 may serve in each capacity. The compiler will try both labelings,
4812 looking for one that is valid, and will reload one or both registers
4813 only if neither labeling works.
4815 @findex FIND_BASE_TERM
4816 @item FIND_BASE_TERM (@var{x})
4817 A C expression to determine the base term of address @var{x}.
4818 This macro is used in only one place: `find_base_term' in alias.c.
4820 It is always safe for this macro to not be defined. It exists so
4821 that alias analysis can understand machine-dependent addresses.
4823 The typical use of this macro is to handle addresses containing
4824 a label_ref or symbol_ref within an UNSPEC@.
4826 @findex LEGITIMIZE_ADDRESS
4827 @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4828 A C compound statement that attempts to replace @var{x} with a valid
4829 memory address for an operand of mode @var{mode}. @var{win} will be a
4830 C statement label elsewhere in the code; the macro definition may use
4833 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4837 to avoid further processing if the address has become legitimate.
4839 @findex break_out_memory_refs
4840 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4841 and @var{oldx} will be the operand that was given to that function to produce
4844 The code generated by this macro should not alter the substructure of
4845 @var{x}. If it transforms @var{x} into a more legitimate form, it
4846 should assign @var{x} (which will always be a C variable) a new value.
4848 It is not necessary for this macro to come up with a legitimate
4849 address. The compiler has standard ways of doing so in all cases. In
4850 fact, it is safe for this macro to do nothing. But often a
4851 machine-dependent strategy can generate better code.
4853 @findex LEGITIMIZE_RELOAD_ADDRESS
4854 @item LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4855 A C compound statement that attempts to replace @var{x}, which is an address
4856 that needs reloading, with a valid memory address for an operand of mode
4857 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4858 It is not necessary to define this macro, but it might be useful for
4859 performance reasons.
4861 For example, on the i386, it is sometimes possible to use a single
4862 reload register instead of two by reloading a sum of two pseudo
4863 registers into a register. On the other hand, for number of RISC
4864 processors offsets are limited so that often an intermediate address
4865 needs to be generated in order to address a stack slot. By defining
4866 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
4867 generated for adjacent some stack slots can be made identical, and thus
4870 @emph{Note}: This macro should be used with caution. It is necessary
4871 to know something of how reload works in order to effectively use this,
4872 and it is quite easy to produce macros that build in too much knowledge
4873 of reload internals.
4875 @emph{Note}: This macro must be able to reload an address created by a
4876 previous invocation of this macro. If it fails to handle such addresses
4877 then the compiler may generate incorrect code or abort.
4880 The macro definition should use @code{push_reload} to indicate parts that
4881 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4882 suitable to be passed unaltered to @code{push_reload}.
4884 The code generated by this macro must not alter the substructure of
4885 @var{x}. If it transforms @var{x} into a more legitimate form, it
4886 should assign @var{x} (which will always be a C variable) a new value.
4887 This also applies to parts that you change indirectly by calling
4890 @findex strict_memory_address_p
4891 The macro definition may use @code{strict_memory_address_p} to test if
4892 the address has become legitimate.
4895 If you want to change only a part of @var{x}, one standard way of doing
4896 this is to use @code{copy_rtx}. Note, however, that is unshares only a
4897 single level of rtl. Thus, if the part to be changed is not at the
4898 top level, you'll need to replace first the top level.
4899 It is not necessary for this macro to come up with a legitimate
4900 address; but often a machine-dependent strategy can generate better code.
4902 @findex GO_IF_MODE_DEPENDENT_ADDRESS
4903 @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
4904 A C statement or compound statement with a conditional @code{goto
4905 @var{label};} executed if memory address @var{x} (an RTX) can have
4906 different meanings depending on the machine mode of the memory
4907 reference it is used for or if the address is valid for some modes
4910 Autoincrement and autodecrement addresses typically have mode-dependent
4911 effects because the amount of the increment or decrement is the size
4912 of the operand being addressed. Some machines have other mode-dependent
4913 addresses. Many RISC machines have no mode-dependent addresses.
4915 You may assume that @var{addr} is a valid address for the machine.
4917 @findex LEGITIMATE_CONSTANT_P
4918 @item LEGITIMATE_CONSTANT_P (@var{x})
4919 A C expression that is nonzero if @var{x} is a legitimate constant for
4920 an immediate operand on the target machine. You can assume that
4921 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
4922 @samp{1} is a suitable definition for this macro on machines where
4923 anything @code{CONSTANT_P} is valid.
4926 @node Condition Code
4927 @section Condition Code Status
4928 @cindex condition code status
4930 @c prevent bad page break with this line
4931 This describes the condition code status.
4934 The file @file{conditions.h} defines a variable @code{cc_status} to
4935 describe how the condition code was computed (in case the interpretation of
4936 the condition code depends on the instruction that it was set by). This
4937 variable contains the RTL expressions on which the condition code is
4938 currently based, and several standard flags.
4940 Sometimes additional machine-specific flags must be defined in the machine
4941 description header file. It can also add additional machine-specific
4942 information by defining @code{CC_STATUS_MDEP}.
4945 @findex CC_STATUS_MDEP
4946 @item CC_STATUS_MDEP
4947 C code for a data type which is used for declaring the @code{mdep}
4948 component of @code{cc_status}. It defaults to @code{int}.
4950 This macro is not used on machines that do not use @code{cc0}.
4952 @findex CC_STATUS_MDEP_INIT
4953 @item CC_STATUS_MDEP_INIT
4954 A C expression to initialize the @code{mdep} field to ``empty''.
4955 The default definition does nothing, since most machines don't use
4956 the field anyway. If you want to use the field, you should probably
4957 define this macro to initialize it.
4959 This macro is not used on machines that do not use @code{cc0}.
4961 @findex NOTICE_UPDATE_CC
4962 @item NOTICE_UPDATE_CC (@var{exp}, @var{insn})
4963 A C compound statement to set the components of @code{cc_status}
4964 appropriately for an insn @var{insn} whose body is @var{exp}. It is
4965 this macro's responsibility to recognize insns that set the condition
4966 code as a byproduct of other activity as well as those that explicitly
4969 This macro is not used on machines that do not use @code{cc0}.
4971 If there are insns that do not set the condition code but do alter
4972 other machine registers, this macro must check to see whether they
4973 invalidate the expressions that the condition code is recorded as
4974 reflecting. For example, on the 68000, insns that store in address
4975 registers do not set the condition code, which means that usually
4976 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
4977 insns. But suppose that the previous insn set the condition code
4978 based on location @samp{a4@@(102)} and the current insn stores a new
4979 value in @samp{a4}. Although the condition code is not changed by
4980 this, it will no longer be true that it reflects the contents of
4981 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
4982 @code{cc_status} in this case to say that nothing is known about the
4983 condition code value.
4985 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
4986 with the results of peephole optimization: insns whose patterns are
4987 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
4988 constants which are just the operands. The RTL structure of these
4989 insns is not sufficient to indicate what the insns actually do. What
4990 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
4991 @code{CC_STATUS_INIT}.
4993 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
4994 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
4995 @samp{cc}. This avoids having detailed information about patterns in
4996 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
4998 @findex EXTRA_CC_MODES
4999 @item EXTRA_CC_MODES
5000 A list of additional modes for condition code values in registers
5001 (@pxref{Jump Patterns}). This macro should expand to a sequence of
5002 calls of the macro @code{CC} separated by white space. @code{CC} takes
5003 two arguments. The first is the enumeration name of the mode, which
5004 should begin with @samp{CC} and end with @samp{mode}. The second is a C
5005 string giving the printable name of the mode; it should be the same as
5006 the first argument, but with the trailing @samp{mode} removed.
5008 You should only define this macro if additional modes are required.
5010 A sample definition of @code{EXTRA_CC_MODES} is:
5012 #define EXTRA_CC_MODES \
5013 CC(CC_NOOVmode, "CC_NOOV") \
5014 CC(CCFPmode, "CCFP") \
5015 CC(CCFPEmode, "CCFPE")
5018 @findex SELECT_CC_MODE
5019 @item SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5020 Returns a mode from class @code{MODE_CC} to be used when comparison
5021 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5022 example, on the Sparc, @code{SELECT_CC_MODE} is defined as (see
5023 @pxref{Jump Patterns} for a description of the reason for this
5027 #define SELECT_CC_MODE(OP,X,Y) \
5028 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5029 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5030 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5031 || GET_CODE (X) == NEG) \
5032 ? CC_NOOVmode : CCmode))
5035 You need not define this macro if @code{EXTRA_CC_MODES} is not defined.
5037 @findex CANONICALIZE_COMPARISON
5038 @item CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5039 On some machines not all possible comparisons are defined, but you can
5040 convert an invalid comparison into a valid one. For example, the Alpha
5041 does not have a @code{GT} comparison, but you can use an @code{LT}
5042 comparison instead and swap the order of the operands.
5044 On such machines, define this macro to be a C statement to do any
5045 required conversions. @var{code} is the initial comparison code
5046 and @var{op0} and @var{op1} are the left and right operands of the
5047 comparison, respectively. You should modify @var{code}, @var{op0}, and
5048 @var{op1} as required.
5050 GCC will not assume that the comparison resulting from this macro is
5051 valid but will see if the resulting insn matches a pattern in the
5054 You need not define this macro if it would never change the comparison
5057 @findex REVERSIBLE_CC_MODE
5058 @item REVERSIBLE_CC_MODE (@var{mode})
5059 A C expression whose value is one if it is always safe to reverse a
5060 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5061 can ever return @var{mode} for a floating-point inequality comparison,
5062 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5064 You need not define this macro if it would always returns zero or if the
5065 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5066 For example, here is the definition used on the Sparc, where floating-point
5067 inequality comparisons are always given @code{CCFPEmode}:
5070 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5073 @findex REVERSE_CONDITION (@var{code}, @var{mode})
5074 A C expression whose value is reversed condition code of the @var{code} for
5075 comparison done in CC_MODE @var{mode}. The macro is used only in case
5076 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5077 machine has some non-standard way how to reverse certain conditionals. For
5078 instance in case all floating point conditions are non-trapping, compiler may
5079 freely convert unordered compares to ordered one. Then definition may look
5083 #define REVERSE_CONDITION(CODE, MODE) \
5084 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5085 : reverse_condition_maybe_unordered (CODE))
5088 @findex REVERSE_CONDEXEC_PREDICATES_P
5089 @item REVERSE_CONDEXEC_PREDICATES_P (@var{code1}, @var{code2})
5090 A C expression that returns true if the conditional execution predicate
5091 @var{code1} is the inverse of @var{code2} and vice versa. Define this to
5092 return 0 if the target has conditional execution predicates that cannot be
5093 reversed safely. If no expansion is specified, this macro is defined as
5097 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5098 ((x) == reverse_condition (y))
5104 @section Describing Relative Costs of Operations
5105 @cindex costs of instructions
5106 @cindex relative costs
5107 @cindex speed of instructions
5109 These macros let you describe the relative speed of various operations
5110 on the target machine.
5114 @item CONST_COSTS (@var{x}, @var{code}, @var{outer_code})
5115 A part of a C @code{switch} statement that describes the relative costs
5116 of constant RTL expressions. It must contain @code{case} labels for
5117 expression codes @code{const_int}, @code{const}, @code{symbol_ref},
5118 @code{label_ref} and @code{const_double}. Each case must ultimately
5119 reach a @code{return} statement to return the relative cost of the use
5120 of that kind of constant value in an expression. The cost may depend on
5121 the precise value of the constant, which is available for examination in
5122 @var{x}, and the rtx code of the expression in which it is contained,
5123 found in @var{outer_code}.
5125 @var{code} is the expression code---redundant, since it can be
5126 obtained with @code{GET_CODE (@var{x})}.
5129 @findex COSTS_N_INSNS
5130 @item RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
5131 Like @code{CONST_COSTS} but applies to nonconstant RTL expressions.
5132 This can be used, for example, to indicate how costly a multiply
5133 instruction is. In writing this macro, you can use the construct
5134 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
5135 instructions. @var{outer_code} is the code of the expression in which
5136 @var{x} is contained.
5138 This macro is optional; do not define it if the default cost assumptions
5139 are adequate for the target machine.
5141 @findex DEFAULT_RTX_COSTS
5142 @item DEFAULT_RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
5143 This macro, if defined, is called for any case not handled by the
5144 @code{RTX_COSTS} or @code{CONST_COSTS} macros. This eliminates the need
5145 to put case labels into the macro, but the code, or any functions it
5146 calls, must assume that the RTL in @var{x} could be of any type that has
5147 not already been handled. The arguments are the same as for
5148 @code{RTX_COSTS}, and the macro should execute a return statement giving
5149 the cost of any RTL expressions that it can handle. The default cost
5150 calculation is used for any RTL for which this macro does not return a
5153 This macro is optional; do not define it if the default cost assumptions
5154 are adequate for the target machine.
5156 @findex ADDRESS_COST
5157 @item ADDRESS_COST (@var{address})
5158 An expression giving the cost of an addressing mode that contains
5159 @var{address}. If not defined, the cost is computed from
5160 the @var{address} expression and the @code{CONST_COSTS} values.
5162 For most CISC machines, the default cost is a good approximation of the
5163 true cost of the addressing mode. However, on RISC machines, all
5164 instructions normally have the same length and execution time. Hence
5165 all addresses will have equal costs.
5167 In cases where more than one form of an address is known, the form with
5168 the lowest cost will be used. If multiple forms have the same, lowest,
5169 cost, the one that is the most complex will be used.
5171 For example, suppose an address that is equal to the sum of a register
5172 and a constant is used twice in the same basic block. When this macro
5173 is not defined, the address will be computed in a register and memory
5174 references will be indirect through that register. On machines where
5175 the cost of the addressing mode containing the sum is no higher than
5176 that of a simple indirect reference, this will produce an additional
5177 instruction and possibly require an additional register. Proper
5178 specification of this macro eliminates this overhead for such machines.
5180 Similar use of this macro is made in strength reduction of loops.
5182 @var{address} need not be valid as an address. In such a case, the cost
5183 is not relevant and can be any value; invalid addresses need not be
5184 assigned a different cost.
5186 On machines where an address involving more than one register is as
5187 cheap as an address computation involving only one register, defining
5188 @code{ADDRESS_COST} to reflect this can cause two registers to be live
5189 over a region of code where only one would have been if
5190 @code{ADDRESS_COST} were not defined in that manner. This effect should
5191 be considered in the definition of this macro. Equivalent costs should
5192 probably only be given to addresses with different numbers of registers
5193 on machines with lots of registers.
5195 This macro will normally either not be defined or be defined as a
5198 @findex REGISTER_MOVE_COST
5199 @item REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5200 A C expression for the cost of moving data of mode @var{mode} from a
5201 register in class @var{from} to one in class @var{to}. The classes are
5202 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5203 value of 2 is the default; other values are interpreted relative to
5206 It is not required that the cost always equal 2 when @var{from} is the
5207 same as @var{to}; on some machines it is expensive to move between
5208 registers if they are not general registers.
5210 If reload sees an insn consisting of a single @code{set} between two
5211 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5212 classes returns a value of 2, reload does not check to ensure that the
5213 constraints of the insn are met. Setting a cost of other than 2 will
5214 allow reload to verify that the constraints are met. You should do this
5215 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5217 @findex MEMORY_MOVE_COST
5218 @item MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5219 A C expression for the cost of moving data of mode @var{mode} between a
5220 register of class @var{class} and memory; @var{in} is zero if the value
5221 is to be written to memory, nonzero if it is to be read in. This cost
5222 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5223 registers and memory is more expensive than between two registers, you
5224 should define this macro to express the relative cost.
5226 If you do not define this macro, GCC uses a default cost of 4 plus
5227 the cost of copying via a secondary reload register, if one is
5228 needed. If your machine requires a secondary reload register to copy
5229 between memory and a register of @var{class} but the reload mechanism is
5230 more complex than copying via an intermediate, define this macro to
5231 reflect the actual cost of the move.
5233 GCC defines the function @code{memory_move_secondary_cost} if
5234 secondary reloads are needed. It computes the costs due to copying via
5235 a secondary register. If your machine copies from memory using a
5236 secondary register in the conventional way but the default base value of
5237 4 is not correct for your machine, define this macro to add some other
5238 value to the result of that function. The arguments to that function
5239 are the same as to this macro.
5243 A C expression for the cost of a branch instruction. A value of 1 is
5244 the default; other values are interpreted relative to that.
5247 Here are additional macros which do not specify precise relative costs,
5248 but only that certain actions are more expensive than GCC would
5252 @findex SLOW_BYTE_ACCESS
5253 @item SLOW_BYTE_ACCESS
5254 Define this macro as a C expression which is nonzero if accessing less
5255 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5256 faster than accessing a word of memory, i.e., if such access
5257 require more than one instruction or if there is no difference in cost
5258 between byte and (aligned) word loads.
5260 When this macro is not defined, the compiler will access a field by
5261 finding the smallest containing object; when it is defined, a fullword
5262 load will be used if alignment permits. Unless bytes accesses are
5263 faster than word accesses, using word accesses is preferable since it
5264 may eliminate subsequent memory access if subsequent accesses occur to
5265 other fields in the same word of the structure, but to different bytes.
5267 @findex SLOW_UNALIGNED_ACCESS
5268 @item SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5269 Define this macro to be the value 1 if memory accesses described by the
5270 @var{mode} and @var{alignment} parameters have a cost many times greater
5271 than aligned accesses, for example if they are emulated in a trap
5274 When this macro is nonzero, the compiler will act as if
5275 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5276 moves. This can cause significantly more instructions to be produced.
5277 Therefore, do not set this macro nonzero if unaligned accesses only add a
5278 cycle or two to the time for a memory access.
5280 If the value of this macro is always zero, it need not be defined. If
5281 this macro is defined, it should produce a nonzero value when
5282 @code{STRICT_ALIGNMENT} is nonzero.
5284 @findex DONT_REDUCE_ADDR
5285 @item DONT_REDUCE_ADDR
5286 Define this macro to inhibit strength reduction of memory addresses.
5287 (On some machines, such strength reduction seems to do harm rather
5292 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5293 which a sequence of insns should be generated instead of a
5294 string move insn or a library call. Increasing the value will always
5295 make code faster, but eventually incurs high cost in increased code size.
5297 Note that on machines where the corresponding move insn is a
5298 @code{define_expand} that emits a sequence of insns, this macro counts
5299 the number of such sequences.
5301 If you don't define this, a reasonable default is used.
5303 @findex MOVE_BY_PIECES_P
5304 @item MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5305 A C expression used to determine whether @code{move_by_pieces} will be used to
5306 copy a chunk of memory, or whether some other block move mechanism
5307 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5308 than @code{MOVE_RATIO}.
5310 @findex MOVE_MAX_PIECES
5311 @item MOVE_MAX_PIECES
5312 A C expression used by @code{move_by_pieces} to determine the largest unit
5313 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5315 @findex USE_LOAD_POST_INCREMENT
5316 @item USE_LOAD_POST_INCREMENT (@var{mode})
5317 A C expression used to determine whether a load postincrement is a good
5318 thing to use for a given mode. Defaults to the value of
5319 @code{HAVE_POST_INCREMENT}.
5321 @findex USE_LOAD_POST_DECREMENT
5322 @item USE_LOAD_POST_DECREMENT (@var{mode})
5323 A C expression used to determine whether a load postdecrement is a good
5324 thing to use for a given mode. Defaults to the value of
5325 @code{HAVE_POST_DECREMENT}.
5327 @findex USE_LOAD_PRE_INCREMENT
5328 @item USE_LOAD_PRE_INCREMENT (@var{mode})
5329 A C expression used to determine whether a load preincrement is a good
5330 thing to use for a given mode. Defaults to the value of
5331 @code{HAVE_PRE_INCREMENT}.
5333 @findex USE_LOAD_PRE_DECREMENT
5334 @item USE_LOAD_PRE_DECREMENT (@var{mode})
5335 A C expression used to determine whether a load predecrement is a good
5336 thing to use for a given mode. Defaults to the value of
5337 @code{HAVE_PRE_DECREMENT}.
5339 @findex USE_STORE_POST_INCREMENT
5340 @item USE_STORE_POST_INCREMENT (@var{mode})
5341 A C expression used to determine whether a store postincrement is a good
5342 thing to use for a given mode. Defaults to the value of
5343 @code{HAVE_POST_INCREMENT}.
5345 @findex USE_STORE_POST_DECREMENT
5346 @item USE_STORE_POST_DECREMENT (@var{mode})
5347 A C expression used to determine whether a store postdecrement is a good
5348 thing to use for a given mode. Defaults to the value of
5349 @code{HAVE_POST_DECREMENT}.
5351 @findex USE_STORE_PRE_INCREMENT
5352 @item USE_STORE_PRE_INCREMENT (@var{mode})
5353 This macro is used to determine whether a store preincrement is a good
5354 thing to use for a given mode. Defaults to the value of
5355 @code{HAVE_PRE_INCREMENT}.
5357 @findex USE_STORE_PRE_DECREMENT
5358 @item USE_STORE_PRE_DECREMENT (@var{mode})
5359 This macro is used to determine whether a store predecrement is a good
5360 thing to use for a given mode. Defaults to the value of
5361 @code{HAVE_PRE_DECREMENT}.
5363 @findex NO_FUNCTION_CSE
5364 @item NO_FUNCTION_CSE
5365 Define this macro if it is as good or better to call a constant
5366 function address than to call an address kept in a register.
5368 @findex NO_RECURSIVE_FUNCTION_CSE
5369 @item NO_RECURSIVE_FUNCTION_CSE
5370 Define this macro if it is as good or better for a function to call
5371 itself with an explicit address than to call an address kept in a
5376 @section Adjusting the Instruction Scheduler
5378 The instruction scheduler may need a fair amount of machine-specific
5379 adjustment in order to produce good code. GCC provides several target
5380 hooks for this purpose. It is usually enough to define just a few of
5381 them: try the first ones in this list first.
5383 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
5384 This hook returns the maximum number of instructions that can ever
5385 issue at the same time on the target machine. The default is one.
5386 Although the insn scheduler can define itself the possibility of issue
5387 an insn on the same cycle, the value can serve as an additional
5388 constraint to issue insns on the same simulated processor cycle (see
5389 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
5390 This value must be constant over the entire compilation. If you need
5391 it to vary depending on what the instructions are, you must use
5392 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
5394 You could use the value of macro @samp{MAX_DFA_ISSUE_RATE} to return
5395 the value of the hook @samp{TARGET_SCHED_ISSUE_RATE} for the automaton
5396 based pipeline interface.
5399 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
5400 This hook is executed by the scheduler after it has scheduled an insn
5401 from the ready list. It should return the number of insns which can
5402 still be issued in the current cycle. Normally this is
5403 @samp{@w{@var{more} - 1}}. You should define this hook if some insns
5404 take more machine resources than others, so that fewer insns can follow
5405 them in the same cycle. @var{file} is either a null pointer, or a stdio
5406 stream to write any debug output to. @var{verbose} is the verbose level
5407 provided by @option{-fsched-verbose-@var{n}}. @var{insn} is the
5408 instruction that was scheduled.
5411 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
5412 This function corrects the value of @var{cost} based on the
5413 relationship between @var{insn} and @var{dep_insn} through the
5414 dependence @var{link}. It should return the new value. The default
5415 is to make no adjustment to @var{cost}. This can be used for example
5416 to specify to the scheduler using the traditional pipeline description
5417 that an output- or anti-dependence does not incur the same cost as a
5418 data-dependence. If the scheduler using the automaton based pipeline
5419 description, the cost of anti-dependence is zero and the cost of
5420 output-dependence is maximum of one and the difference of latency
5421 times of the first and the second insns. If these values are not
5422 acceptable, you could use the hook to modify them too. See also
5423 @pxref{Automaton pipeline description}.
5426 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
5427 This hook adjusts the integer scheduling priority @var{priority} of
5428 @var{insn}. It should return the new priority. Reduce the priority to
5429 execute @var{insn} earlier, increase the priority to execute @var{insn}
5430 later. Do not define this hook if you do not need to adjust the
5431 scheduling priorities of insns.
5434 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
5435 This hook is executed by the scheduler after it has scheduled the ready
5436 list, to allow the machine description to reorder it (for example to
5437 combine two small instructions together on @samp{VLIW} machines).
5438 @var{file} is either a null pointer, or a stdio stream to write any
5439 debug output to. @var{verbose} is the verbose level provided by
5440 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
5441 list of instructions that are ready to be scheduled. @var{n_readyp} is
5442 a pointer to the number of elements in the ready list. The scheduler
5443 reads the ready list in reverse order, starting with
5444 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
5445 is the timer tick of the scheduler. You may modify the ready list and
5446 the number of ready insns. The return value is the number of insns that
5447 can issue this cycle; normally this is just @code{issue_rate}. See also
5448 @samp{TARGET_SCHED_REORDER2}.
5451 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
5452 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
5453 function is called whenever the scheduler starts a new cycle. This one
5454 is called once per iteration over a cycle, immediately after
5455 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
5456 return the number of insns to be scheduled in the same cycle. Defining
5457 this hook can be useful if there are frequent situations where
5458 scheduling one insn causes other insns to become ready in the same
5459 cycle. These other insns can then be taken into account properly.
5462 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
5463 This hook is executed by the scheduler at the beginning of each block of
5464 instructions that are to be scheduled. @var{file} is either a null
5465 pointer, or a stdio stream to write any debug output to. @var{verbose}
5466 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5467 @var{max_ready} is the maximum number of insns in the current scheduling
5468 region that can be live at the same time. This can be used to allocate
5469 scratch space if it is needed, e.g. by @samp{TARGET_SCHED_REORDER}.
5472 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
5473 This hook is executed by the scheduler at the end of each block of
5474 instructions that are to be scheduled. It can be used to perform
5475 cleanup of any actions done by the other scheduling hooks. @var{file}
5476 is either a null pointer, or a stdio stream to write any debug output
5477 to. @var{verbose} is the verbose level provided by
5478 @option{-fsched-verbose-@var{n}}.
5481 @deftypefn {Target Hook} int TARGET_SCHED_USE_DFA_PIPELINE_INTERFACE (void)
5482 This hook is called many times during insn scheduling. If the hook
5483 returns nonzero, the automaton based pipeline description is used for
5484 insn scheduling. Otherwise the traditional pipeline description is
5485 used. The default is usage of the traditional pipeline description.
5487 You should also remember that to simplify the insn scheduler sources
5488 an empty traditional pipeline description interface is generated even
5489 if there is no a traditional pipeline description in the @file{.md}
5490 file. The same is true for the automaton based pipeline description.
5491 That means that you should be accurate in defining the hook.
5494 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
5495 The hook returns an RTL insn. The automaton state used in the
5496 pipeline hazard recognizer is changed as if the insn were scheduled
5497 when the new simulated processor cycle starts. Usage of the hook may
5498 simplify the automaton pipeline description for some @acronym{VLIW}
5499 processors. If the hook is defined, it is used only for the automaton
5500 based pipeline description. The default is not to change the state
5501 when the new simulated processor cycle starts.
5504 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
5505 The hook can be used to initialize data used by the previous hook.
5508 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
5509 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
5510 to changed the state as if the insn were scheduled when the new
5511 simulated processor cycle finishes.
5514 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
5515 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
5516 used to initialize data used by the previous hook.
5519 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
5520 This hook controls better choosing an insn from the ready insn queue
5521 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
5522 chooses the first insn from the queue. If the hook returns a positive
5523 value, an additional scheduler code tries all permutations of
5524 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
5525 subsequent ready insns to choose an insn whose issue will result in
5526 maximal number of issued insns on the same cycle. For the
5527 @acronym{VLIW} processor, the code could actually solve the problem of
5528 packing simple insns into the @acronym{VLIW} insn. Of course, if the
5529 rules of @acronym{VLIW} packing are described in the automaton.
5531 This code also could be used for superscalar @acronym{RISC}
5532 processors. Let us consider a superscalar @acronym{RISC} processor
5533 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
5534 @var{B}, some insns can be executed only in pipelines @var{B} or
5535 @var{C}, and one insn can be executed in pipeline @var{B}. The
5536 processor may issue the 1st insn into @var{A} and the 2nd one into
5537 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
5538 until the next cycle. If the scheduler issues the 3rd insn the first,
5539 the processor could issue all 3 insns per cycle.
5541 Actually this code demonstrates advantages of the automaton based
5542 pipeline hazard recognizer. We try quickly and easy many insn
5543 schedules to choose the best one.
5545 The default is no multipass scheduling.
5548 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_BUBBLES (void)
5549 The @acronym{DFA}-based scheduler could take the insertion of nop
5550 operations for better insn scheduling into account. It can be done
5551 only if the multi-pass insn scheduling works (see hook
5552 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD}).
5554 Let us consider a @acronym{VLIW} processor insn with 3 slots. Each
5555 insn can be placed only in one of the three slots. We have 3 ready
5556 insns @var{A}, @var{B}, and @var{C}. @var{A} and @var{C} can be
5557 placed only in the 1st slot, @var{B} can be placed only in the 3rd
5558 slot. We described the automaton which does not permit empty slot
5559 gaps between insns (usually such description is simpler). Without
5560 this code the scheduler would place each insn in 3 separate
5561 @acronym{VLIW} insns. If the scheduler places a nop insn into the 2nd
5562 slot, it could place the 3 insns into 2 @acronym{VLIW} insns. What is
5563 the nop insn is returned by hook @samp{TARGET_SCHED_DFA_BUBBLE}. Hook
5564 @samp{TARGET_SCHED_INIT_DFA_BUBBLES} can be used to initialize or
5565 create the nop insns.
5567 You should remember that the scheduler does not insert the nop insns.
5568 It is not wise because of the following optimizations. The scheduler
5569 only considers such possibility to improve the result schedule. The
5570 nop insns should be inserted lately, e.g. on the final phase.
5573 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_BUBBLE (int @var{index})
5574 This hook @samp{FIRST_CYCLE_MULTIPASS_SCHEDULING} is used to insert
5575 nop operations for better insn scheduling when @acronym{DFA}-based
5576 scheduler makes multipass insn scheduling (see also description of
5577 hook @samp{TARGET_SCHED_INIT_DFA_BUBBLES}). This hook
5578 returns a nop insn with given @var{index}. The indexes start with
5579 zero. The hook should return @code{NULL} if there are no more nop
5580 insns with indexes greater than given index.
5583 Macros in the following table are generated by the program
5584 @file{genattr} and can be useful for writing the hooks.
5587 @findex TRADITIONAL_PIPELINE_INTERFACE
5588 @item TRADITIONAL_PIPELINE_INTERFACE
5589 The macro definition is generated if there is a traditional pipeline
5590 description in @file{.md} file. You should also remember that to
5591 simplify the insn scheduler sources an empty traditional pipeline
5592 description interface is generated even if there is no a traditional
5593 pipeline description in the @file{.md} file. The macro can be used to
5594 distinguish the two types of the traditional interface.
5596 @findex DFA_PIPELINE_INTERFACE
5597 @item DFA_PIPELINE_INTERFACE
5598 The macro definition is generated if there is an automaton pipeline
5599 description in @file{.md} file. You should also remember that to
5600 simplify the insn scheduler sources an empty automaton pipeline
5601 description interface is generated even if there is no an automaton
5602 pipeline description in the @file{.md} file. The macro can be used to
5603 distinguish the two types of the automaton interface.
5605 @findex MAX_DFA_ISSUE_RATE
5606 @item MAX_DFA_ISSUE_RATE
5607 The macro definition is generated in the automaton based pipeline
5608 description interface. Its value is calculated from the automaton
5609 based pipeline description and is equal to maximal number of all insns
5610 described in constructions @samp{define_insn_reservation} which can be
5611 issued on the same processor cycle.
5616 @section Dividing the Output into Sections (Texts, Data, @dots{})
5617 @c the above section title is WAY too long. maybe cut the part between
5618 @c the (...)? --mew 10feb93
5620 An object file is divided into sections containing different types of
5621 data. In the most common case, there are three sections: the @dfn{text
5622 section}, which holds instructions and read-only data; the @dfn{data
5623 section}, which holds initialized writable data; and the @dfn{bss
5624 section}, which holds uninitialized data. Some systems have other kinds
5627 The compiler must tell the assembler when to switch sections. These
5628 macros control what commands to output to tell the assembler this. You
5629 can also define additional sections.
5632 @findex TEXT_SECTION_ASM_OP
5633 @item TEXT_SECTION_ASM_OP
5634 A C expression whose value is a string, including spacing, containing the
5635 assembler operation that should precede instructions and read-only data.
5636 Normally @code{"\t.text"} is right.
5638 @findex TEXT_SECTION
5640 A C statement that switches to the default section containing instructions.
5641 Normally this is not needed, as simply defining @code{TEXT_SECTION_ASM_OP}
5642 is enough. The MIPS port uses this to sort all functions after all data
5645 @findex HOT_TEXT_SECTION_NAME
5646 @item HOT_TEXT_SECTION_NAME
5647 If defined, a C string constant for the name of the section containing most
5648 frequently executed functions of the program. If not defined, GCC will provide
5649 a default definition if the target supports named sections.
5651 @findex UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5652 @item UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5653 If defined, a C string constant for the name of the section containing unlikely
5654 executed functions in the program.
5656 @findex DATA_SECTION_ASM_OP
5657 @item DATA_SECTION_ASM_OP
5658 A C expression whose value is a string, including spacing, containing the
5659 assembler operation to identify the following data as writable initialized
5660 data. Normally @code{"\t.data"} is right.
5662 @findex READONLY_DATA_SECTION_ASM_OP
5663 @item READONLY_DATA_SECTION_ASM_OP
5664 A C expression whose value is a string, including spacing, containing the
5665 assembler operation to identify the following data as read-only initialized
5668 @findex READONLY_DATA_SECTION
5669 @item READONLY_DATA_SECTION
5670 A macro naming a function to call to switch to the proper section for
5671 read-only data. The default is to use @code{READONLY_DATA_SECTION_ASM_OP}
5672 if defined, else fall back to @code{text_section}.
5674 The most common definition will be @code{data_section}, if the target
5675 does not have a special read-only data section, and does not put data
5676 in the text section.
5678 @findex SHARED_SECTION_ASM_OP
5679 @item SHARED_SECTION_ASM_OP
5680 If defined, a C expression whose value is a string, including spacing,
5681 containing the assembler operation to identify the following data as
5682 shared data. If not defined, @code{DATA_SECTION_ASM_OP} will be used.
5684 @findex BSS_SECTION_ASM_OP
5685 @item BSS_SECTION_ASM_OP
5686 If defined, a C expression whose value is a string, including spacing,
5687 containing the assembler operation to identify the following data as
5688 uninitialized global data. If not defined, and neither
5689 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5690 uninitialized global data will be output in the data section if
5691 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5694 @findex SHARED_BSS_SECTION_ASM_OP
5695 @item SHARED_BSS_SECTION_ASM_OP
5696 If defined, a C expression whose value is a string, including spacing,
5697 containing the assembler operation to identify the following data as
5698 uninitialized global shared data. If not defined, and
5699 @code{BSS_SECTION_ASM_OP} is, the latter will be used.
5701 @findex INIT_SECTION_ASM_OP
5702 @item INIT_SECTION_ASM_OP
5703 If defined, a C expression whose value is a string, including spacing,
5704 containing the assembler operation to identify the following data as
5705 initialization code. If not defined, GCC will assume such a section does
5708 @findex FINI_SECTION_ASM_OP
5709 @item FINI_SECTION_ASM_OP
5710 If defined, a C expression whose value is a string, including spacing,
5711 containing the assembler operation to identify the following data as
5712 finalization code. If not defined, GCC will assume such a section does
5715 @findex CRT_CALL_STATIC_FUNCTION
5716 @item CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
5717 If defined, an ASM statement that switches to a different section
5718 via @var{section_op}, calls @var{function}, and switches back to
5719 the text section. This is used in @file{crtstuff.c} if
5720 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
5721 to initialization and finalization functions from the init and fini
5722 sections. By default, this macro uses a simple function call. Some
5723 ports need hand-crafted assembly code to avoid dependencies on
5724 registers initialized in the function prologue or to ensure that
5725 constant pools don't end up too far way in the text section.
5727 @findex FORCE_CODE_SECTION_ALIGN
5728 @item FORCE_CODE_SECTION_ALIGN
5729 If defined, an ASM statement that aligns a code section to some
5730 arbitrary boundary. This is used to force all fragments of the
5731 @code{.init} and @code{.fini} sections to have to same alignment
5732 and thus prevent the linker from having to add any padding.
5734 @findex EXTRA_SECTIONS
5737 @item EXTRA_SECTIONS
5738 A list of names for sections other than the standard two, which are
5739 @code{in_text} and @code{in_data}. You need not define this macro
5740 on a system with no other sections (that GCC needs to use).
5742 @findex EXTRA_SECTION_FUNCTIONS
5743 @findex text_section
5744 @findex data_section
5745 @item EXTRA_SECTION_FUNCTIONS
5746 One or more functions to be defined in @file{varasm.c}. These
5747 functions should do jobs analogous to those of @code{text_section} and
5748 @code{data_section}, for your additional sections. Do not define this
5749 macro if you do not define @code{EXTRA_SECTIONS}.
5751 @findex JUMP_TABLES_IN_TEXT_SECTION
5752 @item JUMP_TABLES_IN_TEXT_SECTION
5753 Define this macro to be an expression with a nonzero value if jump
5754 tables (for @code{tablejump} insns) should be output in the text
5755 section, along with the assembler instructions. Otherwise, the
5756 readonly data section is used.
5758 This macro is irrelevant if there is no separate readonly data section.
5761 @deftypefn {Target Hook} void TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
5762 Switches to the appropriate section for output of @var{exp}. You can
5763 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
5764 some sort. @var{reloc} indicates whether the initial value of @var{exp}
5765 requires link-time relocations. Bit 0 is set when variable contains
5766 local relocations only, while bit 1 is set for global relocations.
5767 Select the section by calling @code{data_section} or one of the
5768 alternatives for other sections. @var{align} is the constant alignment
5771 The default version of this function takes care of putting read-only
5772 variables in @code{readonly_data_section}.
5775 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
5776 Build up a unique section name, expressed as a @code{STRING_CST} node,
5777 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
5778 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
5779 the initial value of @var{exp} requires link-time relocations.
5781 The default version of this function appends the symbol name to the
5782 ELF section name that would normally be used for the symbol. For
5783 example, the function @code{foo} would be placed in @code{.text.foo}.
5784 Whatever the actual target object format, this is often good enough.
5787 @deftypefn {Target Hook} void TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
5788 Switches to the appropriate section for output of constant pool entry
5789 @var{x} in @var{mode}. You can assume that @var{x} is some kind of
5790 constant in RTL@. The argument @var{mode} is redundant except in the
5791 case of a @code{const_int} rtx. Select the section by calling
5792 @code{readonly_data_section} or one of the alternatives for other
5793 sections. @var{align} is the constant alignment in bits.
5795 The default version of this function takes care of putting symbolic
5796 constants in @code{flag_pic} mode in @code{data_section} and everything
5797 else in @code{readonly_data_section}.
5800 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, int @var{new_decl_p})
5801 Define this hook if references to a symbol or a constant must be
5802 treated differently depending on something about the variable or
5803 function named by the symbol (such as what section it is in).
5805 The hook is executed under two circumstances. One is immediately after
5806 the rtl for @var{decl} that represents a variable or a function has been
5807 created and stored in @code{DECL_RTL(@var{decl})}. The value of the rtl
5808 will be a @code{mem} whose address is a @code{symbol_ref}. The other is
5809 immediately after the rtl for @var{decl} that represents a constant has
5810 been created and stored in @code{TREE_CST_RTL (@var{decl})}. The macro
5811 is called once for each distinct constant in a source file.
5813 The @var{new_decl_p} argument will be true if this is the first time
5814 that @code{ENCODE_SECTION_INFO} has been invoked on this decl. It will
5815 be false for subsequent invocations, which will happen for duplicate
5816 declarations. Whether or not anything must be done for the duplicate
5817 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
5819 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
5820 The usual thing for this hook to do is to record a flag in the
5821 @code{symbol_ref} (such as @code{SYMBOL_REF_FLAG}) or to store a
5822 modified name string in the @code{symbol_ref} (if one bit is not
5823 enough information).
5826 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
5827 Decode @var{name} and return the real name part, sans
5828 the characters that @code{TARGET_ENCODE_SECTION_INFO}
5832 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
5833 Returns true if @var{exp} should be placed into a ``small data'' section.
5834 The default version of this hook always returns false.
5837 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
5838 Returns true if @var{exp} names an object for which name resolution
5839 rules must resolve to the current ``module'' (dynamic shared library
5840 or executable image).
5842 The default version of this hook implements the name resolution rules
5843 for ELF, which has a looser model of global name binding than other
5844 currently supported object file formats.
5848 @section Position Independent Code
5849 @cindex position independent code
5852 This section describes macros that help implement generation of position
5853 independent code. Simply defining these macros is not enough to
5854 generate valid PIC; you must also add support to the macros
5855 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
5856 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
5857 @samp{movsi} to do something appropriate when the source operand
5858 contains a symbolic address. You may also need to alter the handling of
5859 switch statements so that they use relative addresses.
5860 @c i rearranged the order of the macros above to try to force one of
5861 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
5864 @findex PIC_OFFSET_TABLE_REGNUM
5865 @item PIC_OFFSET_TABLE_REGNUM
5866 The register number of the register used to address a table of static
5867 data addresses in memory. In some cases this register is defined by a
5868 processor's ``application binary interface'' (ABI)@. When this macro
5869 is defined, RTL is generated for this register once, as with the stack
5870 pointer and frame pointer registers. If this macro is not defined, it
5871 is up to the machine-dependent files to allocate such a register (if
5872 necessary). Note that this register must be fixed when in use (e.g.@:
5873 when @code{flag_pic} is true).
5875 @findex PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5876 @item PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5877 Define this macro if the register defined by
5878 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
5879 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
5881 @findex FINALIZE_PIC
5883 By generating position-independent code, when two different programs (A
5884 and B) share a common library (libC.a), the text of the library can be
5885 shared whether or not the library is linked at the same address for both
5886 programs. In some of these environments, position-independent code
5887 requires not only the use of different addressing modes, but also
5888 special code to enable the use of these addressing modes.
5890 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
5891 codes once the function is being compiled into assembly code, but not
5892 before. (It is not done before, because in the case of compiling an
5893 inline function, it would lead to multiple PIC prologues being
5894 included in functions which used inline functions and were compiled to
5897 @findex LEGITIMATE_PIC_OPERAND_P
5898 @item LEGITIMATE_PIC_OPERAND_P (@var{x})
5899 A C expression that is nonzero if @var{x} is a legitimate immediate
5900 operand on the target machine when generating position independent code.
5901 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
5902 check this. You can also assume @var{flag_pic} is true, so you need not
5903 check it either. You need not define this macro if all constants
5904 (including @code{SYMBOL_REF}) can be immediate operands when generating
5905 position independent code.
5908 @node Assembler Format
5909 @section Defining the Output Assembler Language
5911 This section describes macros whose principal purpose is to describe how
5912 to write instructions in assembler language---rather than what the
5916 * File Framework:: Structural information for the assembler file.
5917 * Data Output:: Output of constants (numbers, strings, addresses).
5918 * Uninitialized Data:: Output of uninitialized variables.
5919 * Label Output:: Output and generation of labels.
5920 * Initialization:: General principles of initialization
5921 and termination routines.
5922 * Macros for Initialization::
5923 Specific macros that control the handling of
5924 initialization and termination routines.
5925 * Instruction Output:: Output of actual instructions.
5926 * Dispatch Tables:: Output of jump tables.
5927 * Exception Region Output:: Output of exception region code.
5928 * Alignment Output:: Pseudo ops for alignment and skipping data.
5931 @node File Framework
5932 @subsection The Overall Framework of an Assembler File
5933 @cindex assembler format
5934 @cindex output of assembler code
5936 @c prevent bad page break with this line
5937 This describes the overall framework of an assembler file.
5940 @findex ASM_FILE_START
5941 @item ASM_FILE_START (@var{stream})
5942 A C expression which outputs to the stdio stream @var{stream}
5943 some appropriate text to go at the start of an assembler file.
5945 Normally this macro is defined to output a line containing
5946 @samp{#NO_APP}, which is a comment that has no effect on most
5947 assemblers but tells the GNU assembler that it can save time by not
5948 checking for certain assembler constructs.
5950 On systems that use SDB, it is necessary to output certain commands;
5951 see @file{attasm.h}.
5953 @findex ASM_FILE_END
5954 @item ASM_FILE_END (@var{stream})
5955 A C expression which outputs to the stdio stream @var{stream}
5956 some appropriate text to go at the end of an assembler file.
5958 If this macro is not defined, the default is to output nothing
5959 special at the end of the file. Most systems don't require any
5962 On systems that use SDB, it is necessary to output certain commands;
5963 see @file{attasm.h}.
5965 @findex ASM_COMMENT_START
5966 @item ASM_COMMENT_START
5967 A C string constant describing how to begin a comment in the target
5968 assembler language. The compiler assumes that the comment will end at
5969 the end of the line.
5973 A C string constant for text to be output before each @code{asm}
5974 statement or group of consecutive ones. Normally this is
5975 @code{"#APP"}, which is a comment that has no effect on most
5976 assemblers but tells the GNU assembler that it must check the lines
5977 that follow for all valid assembler constructs.
5981 A C string constant for text to be output after each @code{asm}
5982 statement or group of consecutive ones. Normally this is
5983 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
5984 time-saving assumptions that are valid for ordinary compiler output.
5986 @findex ASM_OUTPUT_SOURCE_FILENAME
5987 @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
5988 A C statement to output COFF information or DWARF debugging information
5989 which indicates that filename @var{name} is the current source file to
5990 the stdio stream @var{stream}.
5992 This macro need not be defined if the standard form of output
5993 for the file format in use is appropriate.
5995 @findex OUTPUT_QUOTED_STRING
5996 @item OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
5997 A C statement to output the string @var{string} to the stdio stream
5998 @var{stream}. If you do not call the function @code{output_quoted_string}
5999 in your config files, GCC will only call it to output filenames to
6000 the assembler source. So you can use it to canonicalize the format
6001 of the filename using this macro.
6003 @findex ASM_OUTPUT_SOURCE_LINE
6004 @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
6005 A C statement to output DBX or SDB debugging information before code
6006 for line number @var{line} of the current source file to the
6007 stdio stream @var{stream}.
6009 This macro need not be defined if the standard form of debugging
6010 information for the debugger in use is appropriate.
6012 @findex ASM_OUTPUT_IDENT
6013 @item ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6014 A C statement to output something to the assembler file to handle a
6015 @samp{#ident} directive containing the text @var{string}. If this
6016 macro is not defined, nothing is output for a @samp{#ident} directive.
6018 @findex OBJC_PROLOGUE
6020 A C statement to output any assembler statements which are required to
6021 precede any Objective-C object definitions or message sending. The
6022 statement is executed only when compiling an Objective-C program.
6025 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6026 Output assembly directives to switch to section @var{name}. The section
6027 should have attributes as specified by @var{flags}, which is a bit mask
6028 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6029 is nonzero, it contains an alignment in bytes to be used for the section,
6030 otherwise some target default should be used. Only targets that must
6031 specify an alignment within the section directive need pay attention to
6032 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6035 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6036 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6039 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6040 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6041 based on a variable or function decl, a section name, and whether or not the
6042 declaration's initializer may contain runtime relocations. @var{decl} may be
6043 null, in which case read-write data should be assumed.
6045 The default version if this function handles choosing code vs data,
6046 read-only vs read-write data, and @code{flag_pic}. You should only
6047 need to override this if your target has special flags that might be
6048 set via @code{__attribute__}.
6053 @subsection Output of Data
6056 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6057 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6058 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6059 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6060 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6061 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6062 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6063 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6064 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6065 These hooks specify assembly directives for creating certain kinds
6066 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6067 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6068 aligned two-byte object, and so on. Any of the hooks may be
6069 @code{NULL}, indicating that no suitable directive is available.
6071 The compiler will print these strings at the start of a new line,
6072 followed immediately by the object's initial value. In most cases,
6073 the string should contain a tab, a pseudo-op, and then another tab.
6076 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6077 The @code{assemble_integer} function uses this hook to output an
6078 integer object. @var{x} is the object's value, @var{size} is its size
6079 in bytes and @var{aligned_p} indicates whether it is aligned. The
6080 function should return @code{true} if it was able to output the
6081 object. If it returns false, @code{assemble_integer} will try to
6082 split the object into smaller parts.
6084 The default implementation of this hook will use the
6085 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6086 when the relevant string is @code{NULL}.
6090 @findex OUTPUT_ADDR_CONST_EXTRA
6091 @item OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6092 A C statement to recognize @var{rtx} patterns that
6093 @code{output_addr_const} can't deal with, and output assembly code to
6094 @var{stream} corresponding to the pattern @var{x}. This may be used to
6095 allow machine-dependent @code{UNSPEC}s to appear within constants.
6097 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6098 @code{goto fail}, so that a standard error message is printed. If it
6099 prints an error message itself, by calling, for example,
6100 @code{output_operand_lossage}, it may just complete normally.
6102 @findex ASM_OUTPUT_ASCII
6103 @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6104 A C statement to output to the stdio stream @var{stream} an assembler
6105 instruction to assemble a string constant containing the @var{len}
6106 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6107 @code{char *} and @var{len} a C expression of type @code{int}.
6109 If the assembler has a @code{.ascii} pseudo-op as found in the
6110 Berkeley Unix assembler, do not define the macro
6111 @code{ASM_OUTPUT_ASCII}.
6113 @findex ASM_OUTPUT_FDESC
6114 @item ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6115 A C statement to output word @var{n} of a function descriptor for
6116 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6117 is defined, and is otherwise unused.
6119 @findex CONSTANT_POOL_BEFORE_FUNCTION
6120 @item CONSTANT_POOL_BEFORE_FUNCTION
6121 You may define this macro as a C expression. You should define the
6122 expression to have a nonzero value if GCC should output the constant
6123 pool for a function before the code for the function, or a zero value if
6124 GCC should output the constant pool after the function. If you do
6125 not define this macro, the usual case, GCC will output the constant
6126 pool before the function.
6128 @findex ASM_OUTPUT_POOL_PROLOGUE
6129 @item ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
6130 A C statement to output assembler commands to define the start of the
6131 constant pool for a function. @var{funname} is a string giving
6132 the name of the function. Should the return type of the function
6133 be required, it can be obtained via @var{fundecl}. @var{size}
6134 is the size, in bytes, of the constant pool that will be written
6135 immediately after this call.
6137 If no constant-pool prefix is required, the usual case, this macro need
6140 @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY
6141 @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
6142 A C statement (with or without semicolon) to output a constant in the
6143 constant pool, if it needs special treatment. (This macro need not do
6144 anything for RTL expressions that can be output normally.)
6146 The argument @var{file} is the standard I/O stream to output the
6147 assembler code on. @var{x} is the RTL expression for the constant to
6148 output, and @var{mode} is the machine mode (in case @var{x} is a
6149 @samp{const_int}). @var{align} is the required alignment for the value
6150 @var{x}; you should output an assembler directive to force this much
6153 The argument @var{labelno} is a number to use in an internal label for
6154 the address of this pool entry. The definition of this macro is
6155 responsible for outputting the label definition at the proper place.
6156 Here is how to do this:
6159 ASM_OUTPUT_INTERNAL_LABEL (@var{file}, "LC", @var{labelno});
6162 When you output a pool entry specially, you should end with a
6163 @code{goto} to the label @var{jumpto}. This will prevent the same pool
6164 entry from being output a second time in the usual manner.
6166 You need not define this macro if it would do nothing.
6168 @findex CONSTANT_AFTER_FUNCTION_P
6169 @item CONSTANT_AFTER_FUNCTION_P (@var{exp})
6170 Define this macro as a C expression which is nonzero if the constant
6171 @var{exp}, of type @code{tree}, should be output after the code for a
6172 function. The compiler will normally output all constants before the
6173 function; you need not define this macro if this is OK@.
6175 @findex ASM_OUTPUT_POOL_EPILOGUE
6176 @item ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
6177 A C statement to output assembler commands to at the end of the constant
6178 pool for a function. @var{funname} is a string giving the name of the
6179 function. Should the return type of the function be required, you can
6180 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
6181 constant pool that GCC wrote immediately before this call.
6183 If no constant-pool epilogue is required, the usual case, you need not
6186 @findex IS_ASM_LOGICAL_LINE_SEPARATOR
6187 @item IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
6188 Define this macro as a C expression which is nonzero if @var{C} is
6189 used as a logical line separator by the assembler.
6191 If you do not define this macro, the default is that only
6192 the character @samp{;} is treated as a logical line separator.
6195 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
6196 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
6197 These target hooks are C string constants, describing the syntax in the
6198 assembler for grouping arithmetic expressions. If not overridden, they
6199 default to normal parentheses, which is correct for most assemblers.
6202 These macros are provided by @file{real.h} for writing the definitions
6203 of @code{ASM_OUTPUT_DOUBLE} and the like:
6206 @item REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
6207 @itemx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
6208 @itemx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
6209 @findex REAL_VALUE_TO_TARGET_SINGLE
6210 @findex REAL_VALUE_TO_TARGET_DOUBLE
6211 @findex REAL_VALUE_TO_TARGET_LONG_DOUBLE
6212 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
6213 floating point representation, and store its bit pattern in the variable
6214 @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE}, this variable should
6215 be a simple @code{long int}. For the others, it should be an array of
6216 @code{long int}. The number of elements in this array is determined by
6217 the size of the desired target floating point data type: 32 bits of it
6218 go in each @code{long int} array element. Each array element holds 32
6219 bits of the result, even if @code{long int} is wider than 32 bits on the
6222 The array element values are designed so that you can print them out
6223 using @code{fprintf} in the order they should appear in the target
6226 @item REAL_VALUE_TO_DECIMAL (@var{x}, @var{format}, @var{string})
6227 @findex REAL_VALUE_TO_DECIMAL
6228 This macro converts @var{x}, of type @code{REAL_VALUE_TYPE}, to a
6229 decimal number and stores it as a string into @var{string}.
6230 You must pass, as @var{string}, the address of a long enough block
6231 of space to hold the result.
6233 The argument @var{format} is a @code{printf}-specification that serves
6234 as a suggestion for how to format the output string.
6237 @node Uninitialized Data
6238 @subsection Output of Uninitialized Variables
6240 Each of the macros in this section is used to do the whole job of
6241 outputting a single uninitialized variable.
6244 @findex ASM_OUTPUT_COMMON
6245 @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6246 A C statement (sans semicolon) to output to the stdio stream
6247 @var{stream} the assembler definition of a common-label named
6248 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6249 is the size rounded up to whatever alignment the caller wants.
6251 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6252 output the name itself; before and after that, output the additional
6253 assembler syntax for defining the name, and a newline.
6255 This macro controls how the assembler definitions of uninitialized
6256 common global variables are output.
6258 @findex ASM_OUTPUT_ALIGNED_COMMON
6259 @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
6260 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
6261 separate, explicit argument. If you define this macro, it is used in
6262 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
6263 handling the required alignment of the variable. The alignment is specified
6264 as the number of bits.
6266 @findex ASM_OUTPUT_ALIGNED_DECL_COMMON
6267 @item ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6268 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
6269 variable to be output, if there is one, or @code{NULL_TREE} if there
6270 is no corresponding variable. If you define this macro, GCC will use it
6271 in place of both @code{ASM_OUTPUT_COMMON} and
6272 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
6273 the variable's decl in order to chose what to output.
6275 @findex ASM_OUTPUT_SHARED_COMMON
6276 @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6277 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
6278 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
6281 @findex ASM_OUTPUT_BSS
6282 @item ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6283 A C statement (sans semicolon) to output to the stdio stream
6284 @var{stream} the assembler definition of uninitialized global @var{decl} named
6285 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6286 is the size rounded up to whatever alignment the caller wants.
6288 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
6289 defining this macro. If unable, use the expression
6290 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
6291 before and after that, output the additional assembler syntax for defining
6292 the name, and a newline.
6294 This macro controls how the assembler definitions of uninitialized global
6295 variables are output. This macro exists to properly support languages like
6296 C++ which do not have @code{common} data. However, this macro currently
6297 is not defined for all targets. If this macro and
6298 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
6299 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
6300 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
6302 @findex ASM_OUTPUT_ALIGNED_BSS
6303 @item ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6304 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
6305 separate, explicit argument. If you define this macro, it is used in
6306 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
6307 handling the required alignment of the variable. The alignment is specified
6308 as the number of bits.
6310 Try to use function @code{asm_output_aligned_bss} defined in file
6311 @file{varasm.c} when defining this macro.
6313 @findex ASM_OUTPUT_SHARED_BSS
6314 @item ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6315 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
6316 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
6319 @findex ASM_OUTPUT_LOCAL
6320 @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6321 A C statement (sans semicolon) to output to the stdio stream
6322 @var{stream} the assembler definition of a local-common-label named
6323 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6324 is the size rounded up to whatever alignment the caller wants.
6326 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6327 output the name itself; before and after that, output the additional
6328 assembler syntax for defining the name, and a newline.
6330 This macro controls how the assembler definitions of uninitialized
6331 static variables are output.
6333 @findex ASM_OUTPUT_ALIGNED_LOCAL
6334 @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
6335 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
6336 separate, explicit argument. If you define this macro, it is used in
6337 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
6338 handling the required alignment of the variable. The alignment is specified
6339 as the number of bits.
6341 @findex ASM_OUTPUT_ALIGNED_DECL_LOCAL
6342 @item ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6343 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
6344 variable to be output, if there is one, or @code{NULL_TREE} if there
6345 is no corresponding variable. If you define this macro, GCC will use it
6346 in place of both @code{ASM_OUTPUT_DECL} and
6347 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
6348 the variable's decl in order to chose what to output.
6350 @findex ASM_OUTPUT_SHARED_LOCAL
6351 @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6352 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
6353 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
6358 @subsection Output and Generation of Labels
6360 @c prevent bad page break with this line
6361 This is about outputting labels.
6364 @findex ASM_OUTPUT_LABEL
6365 @findex assemble_name
6366 @item ASM_OUTPUT_LABEL (@var{stream}, @var{name})
6367 A C statement (sans semicolon) to output to the stdio stream
6368 @var{stream} the assembler definition of a label named @var{name}.
6369 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6370 output the name itself; before and after that, output the additional
6371 assembler syntax for defining the name, and a newline.
6373 @findex ASM_DECLARE_FUNCTION_NAME
6374 @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
6375 A C statement (sans semicolon) to output to the stdio stream
6376 @var{stream} any text necessary for declaring the name @var{name} of a
6377 function which is being defined. This macro is responsible for
6378 outputting the label definition (perhaps using
6379 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
6380 @code{FUNCTION_DECL} tree node representing the function.
6382 If this macro is not defined, then the function name is defined in the
6383 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6385 @findex ASM_DECLARE_FUNCTION_SIZE
6386 @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
6387 A C statement (sans semicolon) to output to the stdio stream
6388 @var{stream} any text necessary for declaring the size of a function
6389 which is being defined. The argument @var{name} is the name of the
6390 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
6391 representing the function.
6393 If this macro is not defined, then the function size is not defined.
6395 @findex ASM_DECLARE_OBJECT_NAME
6396 @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6397 A C statement (sans semicolon) to output to the stdio stream
6398 @var{stream} any text necessary for declaring the name @var{name} of an
6399 initialized variable which is being defined. This macro must output the
6400 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6401 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6403 If this macro is not defined, then the variable name is defined in the
6404 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6406 @findex ASM_DECLARE_REGISTER_GLOBAL
6407 @item ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6408 A C statement (sans semicolon) to output to the stdio stream
6409 @var{stream} any text necessary for claiming a register @var{regno}
6410 for a global variable @var{decl} with name @var{name}.
6412 If you don't define this macro, that is equivalent to defining it to do
6415 @findex ASM_FINISH_DECLARE_OBJECT
6416 @item ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6417 A C statement (sans semicolon) to finish up declaring a variable name
6418 once the compiler has processed its initializer fully and thus has had a
6419 chance to determine the size of an array when controlled by an
6420 initializer. This is used on systems where it's necessary to declare
6421 something about the size of the object.
6423 If you don't define this macro, that is equivalent to defining it to do
6426 @findex ASM_GLOBALIZE_LABEL
6427 @item ASM_GLOBALIZE_LABEL (@var{stream}, @var{name})
6428 A C statement (sans semicolon) to output to the stdio stream
6429 @var{stream} some commands that will make the label @var{name} global;
6430 that is, available for reference from other files. Use the expression
6431 @code{assemble_name (@var{stream}, @var{name})} to output the name
6432 itself; before and after that, output the additional assembler syntax
6433 for making that name global, and a newline.
6435 @findex ASM_WEAKEN_LABEL
6436 @item ASM_WEAKEN_LABEL (@var{stream}, @var{name})
6437 A C statement (sans semicolon) to output to the stdio stream
6438 @var{stream} some commands that will make the label @var{name} weak;
6439 that is, available for reference from other files but only used if
6440 no other definition is available. Use the expression
6441 @code{assemble_name (@var{stream}, @var{name})} to output the name
6442 itself; before and after that, output the additional assembler syntax
6443 for making that name weak, and a newline.
6445 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
6446 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
6449 @findex ASM_WEAKEN_DECL
6450 @item ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
6451 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
6452 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
6453 or variable decl. If @var{value} is not @code{NULL}, this C statement
6454 should output to the stdio stream @var{stream} assembler code which
6455 defines (equates) the weak symbol @var{name} to have the value
6456 @var{value}. If @var{value} is @code{NULL}, it should output commands
6457 to make @var{name} weak.
6459 @findex SUPPORTS_WEAK
6461 A C expression which evaluates to true if the target supports weak symbols.
6463 If you don't define this macro, @file{defaults.h} provides a default
6464 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
6465 is defined, the default definition is @samp{1}; otherwise, it is
6466 @samp{0}. Define this macro if you want to control weak symbol support
6467 with a compiler flag such as @option{-melf}.
6469 @findex MAKE_DECL_ONE_ONLY (@var{decl})
6470 @item MAKE_DECL_ONE_ONLY
6471 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
6472 public symbol such that extra copies in multiple translation units will
6473 be discarded by the linker. Define this macro if your object file
6474 format provides support for this concept, such as the @samp{COMDAT}
6475 section flags in the Microsoft Windows PE/COFF format, and this support
6476 requires changes to @var{decl}, such as putting it in a separate section.
6478 @findex SUPPORTS_ONE_ONLY
6479 @item SUPPORTS_ONE_ONLY
6480 A C expression which evaluates to true if the target supports one-only
6483 If you don't define this macro, @file{varasm.c} provides a default
6484 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6485 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6486 you want to control one-only symbol support with a compiler flag, or if
6487 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6488 be emitted as one-only.
6490 @findex ASM_OUTPUT_EXTERNAL
6491 @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6492 A C statement (sans semicolon) to output to the stdio stream
6493 @var{stream} any text necessary for declaring the name of an external
6494 symbol named @var{name} which is referenced in this compilation but
6495 not defined. The value of @var{decl} is the tree node for the
6498 This macro need not be defined if it does not need to output anything.
6499 The GNU assembler and most Unix assemblers don't require anything.
6501 @findex ASM_OUTPUT_EXTERNAL_LIBCALL
6502 @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref})
6503 A C statement (sans semicolon) to output on @var{stream} an assembler
6504 pseudo-op to declare a library function name external. The name of the
6505 library function is given by @var{symref}, which has type @code{rtx} and
6506 is a @code{symbol_ref}.
6508 This macro need not be defined if it does not need to output anything.
6509 The GNU assembler and most Unix assemblers don't require anything.
6511 @findex ASM_OUTPUT_LABELREF
6512 @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6513 A C statement (sans semicolon) to output to the stdio stream
6514 @var{stream} a reference in assembler syntax to a label named
6515 @var{name}. This should add @samp{_} to the front of the name, if that
6516 is customary on your operating system, as it is in most Berkeley Unix
6517 systems. This macro is used in @code{assemble_name}.
6519 @findex ASM_OUTPUT_SYMBOL_REF
6520 @item ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6521 A C statement (sans semicolon) to output a reference to
6522 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
6523 will be used to output the name of the symbol. This macro may be used
6524 to modify the way a symbol is referenced depending on information
6525 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
6527 @findex ASM_OUTPUT_LABEL_REF
6528 @item ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
6529 A C statement (sans semicolon) to output a reference to @var{buf}, the
6530 result of ASM_GENERATE_INTERNAL_LABEL. If not defined,
6531 @code{assemble_name} will be used to output the name of the symbol.
6532 This macro is not used by @code{output_asm_label}, or the @code{%l}
6533 specifier that calls it; the intention is that this macro should be set
6534 when it is necessary to output a label differently when its address
6537 @findex ASM_OUTPUT_INTERNAL_LABEL
6538 @item ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{prefix}, @var{num})
6539 A C statement to output to the stdio stream @var{stream} a label whose
6540 name is made from the string @var{prefix} and the number @var{num}.
6542 It is absolutely essential that these labels be distinct from the labels
6543 used for user-level functions and variables. Otherwise, certain programs
6544 will have name conflicts with internal labels.
6546 It is desirable to exclude internal labels from the symbol table of the
6547 object file. Most assemblers have a naming convention for labels that
6548 should be excluded; on many systems, the letter @samp{L} at the
6549 beginning of a label has this effect. You should find out what
6550 convention your system uses, and follow it.
6552 The usual definition of this macro is as follows:
6555 fprintf (@var{stream}, "L%s%d:\n", @var{prefix}, @var{num})
6558 @findex ASM_OUTPUT_DEBUG_LABEL
6559 @item ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
6560 A C statement to output to the stdio stream @var{stream} a debug info
6561 label whose name is made from the string @var{prefix} and the number
6562 @var{num}. This is useful for VLIW targets, where debug info labels
6563 may need to be treated differently than branch target labels. On some
6564 systems, branch target labels must be at the beginning of instruction
6565 bundles, but debug info labels can occur in the middle of instruction
6568 If this macro is not defined, then @code{ASM_OUTPUT_INTERNAL_LABEL} will be
6571 @findex ASM_OUTPUT_ALTERNATE_LABEL_NAME
6572 @item ASM_OUTPUT_ALTERNATE_LABEL_NAME (@var{stream}, @var{string})
6573 A C statement to output to the stdio stream @var{stream} the string
6576 The default definition of this macro is as follows:
6579 fprintf (@var{stream}, "%s:\n", LABEL_ALTERNATE_NAME (INSN))
6582 @findex ASM_GENERATE_INTERNAL_LABEL
6583 @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
6584 A C statement to store into the string @var{string} a label whose name
6585 is made from the string @var{prefix} and the number @var{num}.
6587 This string, when output subsequently by @code{assemble_name}, should
6588 produce the output that @code{ASM_OUTPUT_INTERNAL_LABEL} would produce
6589 with the same @var{prefix} and @var{num}.
6591 If the string begins with @samp{*}, then @code{assemble_name} will
6592 output the rest of the string unchanged. It is often convenient for
6593 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
6594 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
6595 to output the string, and may change it. (Of course,
6596 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
6597 you should know what it does on your machine.)
6599 @findex ASM_FORMAT_PRIVATE_NAME
6600 @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
6601 A C expression to assign to @var{outvar} (which is a variable of type
6602 @code{char *}) a newly allocated string made from the string
6603 @var{name} and the number @var{number}, with some suitable punctuation
6604 added. Use @code{alloca} to get space for the string.
6606 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
6607 produce an assembler label for an internal static variable whose name is
6608 @var{name}. Therefore, the string must be such as to result in valid
6609 assembler code. The argument @var{number} is different each time this
6610 macro is executed; it prevents conflicts between similarly-named
6611 internal static variables in different scopes.
6613 Ideally this string should not be a valid C identifier, to prevent any
6614 conflict with the user's own symbols. Most assemblers allow periods
6615 or percent signs in assembler symbols; putting at least one of these
6616 between the name and the number will suffice.
6618 @findex ASM_OUTPUT_DEF
6619 @item ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
6620 A C statement to output to the stdio stream @var{stream} assembler code
6621 which defines (equates) the symbol @var{name} to have the value @var{value}.
6624 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6625 correct for most systems.
6627 @findex ASM_OUTPUT_DEF_FROM_DECLS
6628 @item ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
6629 A C statement to output to the stdio stream @var{stream} assembler code
6630 which defines (equates) the symbol whose tree node is @var{decl_of_name}
6631 to have the value of the tree node @var{decl_of_value}. This macro will
6632 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
6633 the tree nodes are available.
6635 @findex ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL
6636 @item ASM_OUTPUT_DEFINE_LABEL_DIFFERENCE_SYMBOL (@var{stream}, @var{symbol}, @var{high}, @var{low})
6637 A C statement to output to the stdio stream @var{stream} assembler code
6638 which defines (equates) the symbol @var{symbol} to have a value equal to
6639 the difference of the two symbols @var{high} and @var{low},
6640 i.e.@: @var{high} minus @var{low}. GCC guarantees that the symbols @var{high}
6641 and @var{low} are already known by the assembler so that the difference
6642 resolves into a constant.
6645 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6646 correct for most systems.
6648 @findex ASM_OUTPUT_WEAK_ALIAS
6649 @item ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
6650 A C statement to output to the stdio stream @var{stream} assembler code
6651 which defines (equates) the weak symbol @var{name} to have the value
6652 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
6653 an undefined weak symbol.
6655 Define this macro if the target only supports weak aliases; define
6656 @code{ASM_OUTPUT_DEF} instead if possible.
6658 @findex OBJC_GEN_METHOD_LABEL
6659 @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
6660 Define this macro to override the default assembler names used for
6661 Objective-C methods.
6663 The default name is a unique method number followed by the name of the
6664 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
6665 the category is also included in the assembler name (e.g.@:
6668 These names are safe on most systems, but make debugging difficult since
6669 the method's selector is not present in the name. Therefore, particular
6670 systems define other ways of computing names.
6672 @var{buf} is an expression of type @code{char *} which gives you a
6673 buffer in which to store the name; its length is as long as
6674 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
6675 50 characters extra.
6677 The argument @var{is_inst} specifies whether the method is an instance
6678 method or a class method; @var{class_name} is the name of the class;
6679 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
6680 in a category); and @var{sel_name} is the name of the selector.
6682 On systems where the assembler can handle quoted names, you can use this
6683 macro to provide more human-readable names.
6685 @findex ASM_DECLARE_CLASS_REFERENCE
6686 @item ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
6687 A C statement (sans semicolon) to output to the stdio stream
6688 @var{stream} commands to declare that the label @var{name} is an
6689 Objective-C class reference. This is only needed for targets whose
6690 linkers have special support for NeXT-style runtimes.
6692 @findex ASM_DECLARE_UNRESOLVED_REFERENCE
6693 @item ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
6694 A C statement (sans semicolon) to output to the stdio stream
6695 @var{stream} commands to declare that the label @var{name} is an
6696 unresolved Objective-C class reference. This is only needed for targets
6697 whose linkers have special support for NeXT-style runtimes.
6700 @node Initialization
6701 @subsection How Initialization Functions Are Handled
6702 @cindex initialization routines
6703 @cindex termination routines
6704 @cindex constructors, output of
6705 @cindex destructors, output of
6707 The compiled code for certain languages includes @dfn{constructors}
6708 (also called @dfn{initialization routines})---functions to initialize
6709 data in the program when the program is started. These functions need
6710 to be called before the program is ``started''---that is to say, before
6711 @code{main} is called.
6713 Compiling some languages generates @dfn{destructors} (also called
6714 @dfn{termination routines}) that should be called when the program
6717 To make the initialization and termination functions work, the compiler
6718 must output something in the assembler code to cause those functions to
6719 be called at the appropriate time. When you port the compiler to a new
6720 system, you need to specify how to do this.
6722 There are two major ways that GCC currently supports the execution of
6723 initialization and termination functions. Each way has two variants.
6724 Much of the structure is common to all four variations.
6726 @findex __CTOR_LIST__
6727 @findex __DTOR_LIST__
6728 The linker must build two lists of these functions---a list of
6729 initialization functions, called @code{__CTOR_LIST__}, and a list of
6730 termination functions, called @code{__DTOR_LIST__}.
6732 Each list always begins with an ignored function pointer (which may hold
6733 0, @minus{}1, or a count of the function pointers after it, depending on
6734 the environment). This is followed by a series of zero or more function
6735 pointers to constructors (or destructors), followed by a function
6736 pointer containing zero.
6738 Depending on the operating system and its executable file format, either
6739 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
6740 time and exit time. Constructors are called in reverse order of the
6741 list; destructors in forward order.
6743 The best way to handle static constructors works only for object file
6744 formats which provide arbitrarily-named sections. A section is set
6745 aside for a list of constructors, and another for a list of destructors.
6746 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
6747 object file that defines an initialization function also puts a word in
6748 the constructor section to point to that function. The linker
6749 accumulates all these words into one contiguous @samp{.ctors} section.
6750 Termination functions are handled similarly.
6752 This method will be chosen as the default by @file{target-def.h} if
6753 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
6754 support arbitrary sections, but does support special designated
6755 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
6756 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
6758 When arbitrary sections are available, there are two variants, depending
6759 upon how the code in @file{crtstuff.c} is called. On systems that
6760 support a @dfn{.init} section which is executed at program startup,
6761 parts of @file{crtstuff.c} are compiled into that section. The
6762 program is linked by the @code{gcc} driver like this:
6765 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
6768 The prologue of a function (@code{__init}) appears in the @code{.init}
6769 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
6770 for the function @code{__fini} in the @dfn{.fini} section. Normally these
6771 files are provided by the operating system or by the GNU C library, but
6772 are provided by GCC for a few targets.
6774 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
6775 compiled from @file{crtstuff.c}. They contain, among other things, code
6776 fragments within the @code{.init} and @code{.fini} sections that branch
6777 to routines in the @code{.text} section. The linker will pull all parts
6778 of a section together, which results in a complete @code{__init} function
6779 that invokes the routines we need at startup.
6781 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
6784 If no init section is available, when GCC compiles any function called
6785 @code{main} (or more accurately, any function designated as a program
6786 entry point by the language front end calling @code{expand_main_function}),
6787 it inserts a procedure call to @code{__main} as the first executable code
6788 after the function prologue. The @code{__main} function is defined
6789 in @file{libgcc2.c} and runs the global constructors.
6791 In file formats that don't support arbitrary sections, there are again
6792 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
6793 and an `a.out' format must be used. In this case,
6794 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
6795 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
6796 and with the address of the void function containing the initialization
6797 code as its value. The GNU linker recognizes this as a request to add
6798 the value to a @dfn{set}; the values are accumulated, and are eventually
6799 placed in the executable as a vector in the format described above, with
6800 a leading (ignored) count and a trailing zero element.
6801 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
6802 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
6803 the compilation of @code{main} to call @code{__main} as above, starting
6804 the initialization process.
6806 The last variant uses neither arbitrary sections nor the GNU linker.
6807 This is preferable when you want to do dynamic linking and when using
6808 file formats which the GNU linker does not support, such as `ECOFF'@. In
6809 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
6810 termination functions are recognized simply by their names. This requires
6811 an extra program in the linkage step, called @command{collect2}. This program
6812 pretends to be the linker, for use with GCC; it does its job by running
6813 the ordinary linker, but also arranges to include the vectors of
6814 initialization and termination functions. These functions are called
6815 via @code{__main} as described above. In order to use this method,
6816 @code{use_collect2} must be defined in the target in @file{config.gcc}.
6819 The following section describes the specific macros that control and
6820 customize the handling of initialization and termination functions.
6823 @node Macros for Initialization
6824 @subsection Macros Controlling Initialization Routines
6826 Here are the macros that control how the compiler handles initialization
6827 and termination functions:
6830 @findex INIT_SECTION_ASM_OP
6831 @item INIT_SECTION_ASM_OP
6832 If defined, a C string constant, including spacing, for the assembler
6833 operation to identify the following data as initialization code. If not
6834 defined, GCC will assume such a section does not exist. When you are
6835 using special sections for initialization and termination functions, this
6836 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
6837 run the initialization functions.
6839 @item HAS_INIT_SECTION
6840 @findex HAS_INIT_SECTION
6841 If defined, @code{main} will not call @code{__main} as described above.
6842 This macro should be defined for systems that control start-up code
6843 on a symbol-by-symbol basis, such as OSF/1, and should not
6844 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
6846 @item LD_INIT_SWITCH
6847 @findex LD_INIT_SWITCH
6848 If defined, a C string constant for a switch that tells the linker that
6849 the following symbol is an initialization routine.
6851 @item LD_FINI_SWITCH
6852 @findex LD_FINI_SWITCH
6853 If defined, a C string constant for a switch that tells the linker that
6854 the following symbol is a finalization routine.
6856 @item COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
6857 If defined, a C statement that will write a function that can be
6858 automatically called when a shared library is loaded. The function
6859 should call @var{func}, which takes no arguments. If not defined, and
6860 the object format requires an explicit initialization function, then a
6861 function called @code{_GLOBAL__DI} will be generated.
6863 This function and the following one are used by collect2 when linking a
6864 shared library that needs constructors or destructors, or has DWARF2
6865 exception tables embedded in the code.
6867 @item COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
6868 If defined, a C statement that will write a function that can be
6869 automatically called when a shared library is unloaded. The function
6870 should call @var{func}, which takes no arguments. If not defined, and
6871 the object format requires an explicit finalization function, then a
6872 function called @code{_GLOBAL__DD} will be generated.
6875 @findex INVOKE__main
6876 If defined, @code{main} will call @code{__main} despite the presence of
6877 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
6878 where the init section is not actually run automatically, but is still
6879 useful for collecting the lists of constructors and destructors.
6881 @item SUPPORTS_INIT_PRIORITY
6882 @findex SUPPORTS_INIT_PRIORITY
6883 If nonzero, the C++ @code{init_priority} attribute is supported and the
6884 compiler should emit instructions to control the order of initialization
6885 of objects. If zero, the compiler will issue an error message upon
6886 encountering an @code{init_priority} attribute.
6889 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
6890 This value is true if the target supports some ``native'' method of
6891 collecting constructors and destructors to be run at startup and exit.
6892 It is false if we must use @command{collect2}.
6895 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
6896 If defined, a function that outputs assembler code to arrange to call
6897 the function referenced by @var{symbol} at initialization time.
6899 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
6900 no arguments and with no return value. If the target supports initialization
6901 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
6902 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
6904 If this macro is not defined by the target, a suitable default will
6905 be chosen if (1) the target supports arbitrary section names, (2) the
6906 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
6910 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
6911 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
6912 functions rather than initialization functions.
6915 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
6916 generated for the generated object file will have static linkage.
6918 If your system uses @command{collect2} as the means of processing
6919 constructors, then that program normally uses @command{nm} to scan
6920 an object file for constructor functions to be called.
6922 On certain kinds of systems, you can define these macros to make
6923 @command{collect2} work faster (and, in some cases, make it work at all):
6926 @findex OBJECT_FORMAT_COFF
6927 @item OBJECT_FORMAT_COFF
6928 Define this macro if the system uses COFF (Common Object File Format)
6929 object files, so that @command{collect2} can assume this format and scan
6930 object files directly for dynamic constructor/destructor functions.
6932 @findex OBJECT_FORMAT_ROSE
6933 @item OBJECT_FORMAT_ROSE
6934 Define this macro if the system uses ROSE format object files, so that
6935 @command{collect2} can assume this format and scan object files directly
6936 for dynamic constructor/destructor functions.
6938 These macros are effective only in a native compiler; @command{collect2} as
6939 part of a cross compiler always uses @command{nm} for the target machine.
6941 @findex REAL_NM_FILE_NAME
6942 @item REAL_NM_FILE_NAME
6943 Define this macro as a C string constant containing the file name to use
6944 to execute @command{nm}. The default is to search the path normally for
6947 If your system supports shared libraries and has a program to list the
6948 dynamic dependencies of a given library or executable, you can define
6949 these macros to enable support for running initialization and
6950 termination functions in shared libraries:
6954 Define this macro to a C string constant containing the name of the program
6955 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
6957 @findex PARSE_LDD_OUTPUT
6958 @item PARSE_LDD_OUTPUT (@var{ptr})
6959 Define this macro to be C code that extracts filenames from the output
6960 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
6961 of type @code{char *} that points to the beginning of a line of output
6962 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
6963 code must advance @var{ptr} to the beginning of the filename on that
6964 line. Otherwise, it must set @var{ptr} to @code{NULL}.
6967 @node Instruction Output
6968 @subsection Output of Assembler Instructions
6970 @c prevent bad page break with this line
6971 This describes assembler instruction output.
6974 @findex REGISTER_NAMES
6975 @item REGISTER_NAMES
6976 A C initializer containing the assembler's names for the machine
6977 registers, each one as a C string constant. This is what translates
6978 register numbers in the compiler into assembler language.
6980 @findex ADDITIONAL_REGISTER_NAMES
6981 @item ADDITIONAL_REGISTER_NAMES
6982 If defined, a C initializer for an array of structures containing a name
6983 and a register number. This macro defines additional names for hard
6984 registers, thus allowing the @code{asm} option in declarations to refer
6985 to registers using alternate names.
6987 @findex ASM_OUTPUT_OPCODE
6988 @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
6989 Define this macro if you are using an unusual assembler that
6990 requires different names for the machine instructions.
6992 The definition is a C statement or statements which output an
6993 assembler instruction opcode to the stdio stream @var{stream}. The
6994 macro-operand @var{ptr} is a variable of type @code{char *} which
6995 points to the opcode name in its ``internal'' form---the form that is
6996 written in the machine description. The definition should output the
6997 opcode name to @var{stream}, performing any translation you desire, and
6998 increment the variable @var{ptr} to point at the end of the opcode
6999 so that it will not be output twice.
7001 In fact, your macro definition may process less than the entire opcode
7002 name, or more than the opcode name; but if you want to process text
7003 that includes @samp{%}-sequences to substitute operands, you must take
7004 care of the substitution yourself. Just be sure to increment
7005 @var{ptr} over whatever text should not be output normally.
7007 @findex recog_data.operand
7008 If you need to look at the operand values, they can be found as the
7009 elements of @code{recog_data.operand}.
7011 If the macro definition does nothing, the instruction is output
7014 @findex FINAL_PRESCAN_INSN
7015 @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7016 If defined, a C statement to be executed just prior to the output of
7017 assembler code for @var{insn}, to modify the extracted operands so
7018 they will be output differently.
7020 Here the argument @var{opvec} is the vector containing the operands
7021 extracted from @var{insn}, and @var{noperands} is the number of
7022 elements of the vector which contain meaningful data for this insn.
7023 The contents of this vector are what will be used to convert the insn
7024 template into assembler code, so you can change the assembler output
7025 by changing the contents of the vector.
7027 This macro is useful when various assembler syntaxes share a single
7028 file of instruction patterns; by defining this macro differently, you
7029 can cause a large class of instructions to be output differently (such
7030 as with rearranged operands). Naturally, variations in assembler
7031 syntax affecting individual insn patterns ought to be handled by
7032 writing conditional output routines in those patterns.
7034 If this macro is not defined, it is equivalent to a null statement.
7036 @findex FINAL_PRESCAN_LABEL
7037 @item FINAL_PRESCAN_LABEL
7038 If defined, @code{FINAL_PRESCAN_INSN} will be called on each
7039 @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and
7040 @var{noperands} will be zero.
7042 @findex PRINT_OPERAND
7043 @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
7044 A C compound statement to output to stdio stream @var{stream} the
7045 assembler syntax for an instruction operand @var{x}. @var{x} is an
7048 @var{code} is a value that can be used to specify one of several ways
7049 of printing the operand. It is used when identical operands must be
7050 printed differently depending on the context. @var{code} comes from
7051 the @samp{%} specification that was used to request printing of the
7052 operand. If the specification was just @samp{%@var{digit}} then
7053 @var{code} is 0; if the specification was @samp{%@var{ltr}
7054 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
7057 If @var{x} is a register, this macro should print the register's name.
7058 The names can be found in an array @code{reg_names} whose type is
7059 @code{char *[]}. @code{reg_names} is initialized from
7060 @code{REGISTER_NAMES}.
7062 When the machine description has a specification @samp{%@var{punct}}
7063 (a @samp{%} followed by a punctuation character), this macro is called
7064 with a null pointer for @var{x} and the punctuation character for
7067 @findex PRINT_OPERAND_PUNCT_VALID_P
7068 @item PRINT_OPERAND_PUNCT_VALID_P (@var{code})
7069 A C expression which evaluates to true if @var{code} is a valid
7070 punctuation character for use in the @code{PRINT_OPERAND} macro. If
7071 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
7072 punctuation characters (except for the standard one, @samp{%}) are used
7075 @findex PRINT_OPERAND_ADDRESS
7076 @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
7077 A C compound statement to output to stdio stream @var{stream} the
7078 assembler syntax for an instruction operand that is a memory reference
7079 whose address is @var{x}. @var{x} is an RTL expression.
7081 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
7082 On some machines, the syntax for a symbolic address depends on the
7083 section that the address refers to. On these machines, define the hook
7084 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
7085 @code{symbol_ref}, and then check for it here. @xref{Assembler Format}.
7087 @findex DBR_OUTPUT_SEQEND
7088 @findex dbr_sequence_length
7089 @item DBR_OUTPUT_SEQEND(@var{file})
7090 A C statement, to be executed after all slot-filler instructions have
7091 been output. If necessary, call @code{dbr_sequence_length} to
7092 determine the number of slots filled in a sequence (zero if not
7093 currently outputting a sequence), to decide how many no-ops to output,
7096 Don't define this macro if it has nothing to do, but it is helpful in
7097 reading assembly output if the extent of the delay sequence is made
7098 explicit (e.g.@: with white space).
7100 @findex final_sequence
7101 Note that output routines for instructions with delay slots must be
7102 prepared to deal with not being output as part of a sequence
7103 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
7104 found.) The variable @code{final_sequence} is null when not
7105 processing a sequence, otherwise it contains the @code{sequence} rtx
7108 @findex REGISTER_PREFIX
7109 @findex LOCAL_LABEL_PREFIX
7110 @findex USER_LABEL_PREFIX
7111 @findex IMMEDIATE_PREFIX
7113 @item REGISTER_PREFIX
7114 @itemx LOCAL_LABEL_PREFIX
7115 @itemx USER_LABEL_PREFIX
7116 @itemx IMMEDIATE_PREFIX
7117 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
7118 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
7119 @file{final.c}). These are useful when a single @file{md} file must
7120 support multiple assembler formats. In that case, the various @file{tm.h}
7121 files can define these macros differently.
7123 @item ASM_FPRINTF_EXTENSIONS(@var{file}, @var{argptr}, @var{format})
7124 @findex ASM_FPRINTF_EXTENSIONS
7125 If defined this macro should expand to a series of @code{case}
7126 statements which will be parsed inside the @code{switch} statement of
7127 the @code{asm_fprintf} function. This allows targets to define extra
7128 printf formats which may useful when generating their assembler
7129 statements. Note that upper case letters are reserved for future
7130 generic extensions to asm_fprintf, and so are not available to target
7131 specific code. The output file is given by the parameter @var{file}.
7132 The varargs input pointer is @var{argptr} and the rest of the format
7133 string, starting the character after the one that is being switched
7134 upon, is pointed to by @var{format}.
7136 @findex ASSEMBLER_DIALECT
7137 @item ASSEMBLER_DIALECT
7138 If your target supports multiple dialects of assembler language (such as
7139 different opcodes), define this macro as a C expression that gives the
7140 numeric index of the assembler language dialect to use, with zero as the
7143 If this macro is defined, you may use constructs of the form
7145 @samp{@{option0|option1|option2@dots{}@}}
7148 in the output templates of patterns (@pxref{Output Template}) or in the
7149 first argument of @code{asm_fprintf}. This construct outputs
7150 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
7151 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
7152 within these strings retain their usual meaning. If there are fewer
7153 alternatives within the braces than the value of
7154 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
7156 If you do not define this macro, the characters @samp{@{}, @samp{|} and
7157 @samp{@}} do not have any special meaning when used in templates or
7158 operands to @code{asm_fprintf}.
7160 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
7161 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
7162 the variations in assembler language syntax with that mechanism. Define
7163 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
7164 if the syntax variant are larger and involve such things as different
7165 opcodes or operand order.
7167 @findex ASM_OUTPUT_REG_PUSH
7168 @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
7169 A C expression to output to @var{stream} some assembler code
7170 which will push hard register number @var{regno} onto the stack.
7171 The code need not be optimal, since this macro is used only when
7174 @findex ASM_OUTPUT_REG_POP
7175 @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
7176 A C expression to output to @var{stream} some assembler code
7177 which will pop hard register number @var{regno} off of the stack.
7178 The code need not be optimal, since this macro is used only when
7182 @node Dispatch Tables
7183 @subsection Output of Dispatch Tables
7185 @c prevent bad page break with this line
7186 This concerns dispatch tables.
7189 @cindex dispatch table
7190 @findex ASM_OUTPUT_ADDR_DIFF_ELT
7191 @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
7192 A C statement to output to the stdio stream @var{stream} an assembler
7193 pseudo-instruction to generate a difference between two labels.
7194 @var{value} and @var{rel} are the numbers of two internal labels. The
7195 definitions of these labels are output using
7196 @code{ASM_OUTPUT_INTERNAL_LABEL}, and they must be printed in the same
7197 way here. For example,
7200 fprintf (@var{stream}, "\t.word L%d-L%d\n",
7201 @var{value}, @var{rel})
7204 You must provide this macro on machines where the addresses in a
7205 dispatch table are relative to the table's own address. If defined, GCC
7206 will also use this macro on all machines when producing PIC@.
7207 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
7208 mode and flags can be read.
7210 @findex ASM_OUTPUT_ADDR_VEC_ELT
7211 @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
7212 This macro should be provided on machines where the addresses
7213 in a dispatch table are absolute.
7215 The definition should be a C statement to output to the stdio stream
7216 @var{stream} an assembler pseudo-instruction to generate a reference to
7217 a label. @var{value} is the number of an internal label whose
7218 definition is output using @code{ASM_OUTPUT_INTERNAL_LABEL}.
7222 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
7225 @findex ASM_OUTPUT_CASE_LABEL
7226 @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
7227 Define this if the label before a jump-table needs to be output
7228 specially. The first three arguments are the same as for
7229 @code{ASM_OUTPUT_INTERNAL_LABEL}; the fourth argument is the
7230 jump-table which follows (a @code{jump_insn} containing an
7231 @code{addr_vec} or @code{addr_diff_vec}).
7233 This feature is used on system V to output a @code{swbeg} statement
7236 If this macro is not defined, these labels are output with
7237 @code{ASM_OUTPUT_INTERNAL_LABEL}.
7239 @findex ASM_OUTPUT_CASE_END
7240 @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
7241 Define this if something special must be output at the end of a
7242 jump-table. The definition should be a C statement to be executed
7243 after the assembler code for the table is written. It should write
7244 the appropriate code to stdio stream @var{stream}. The argument
7245 @var{table} is the jump-table insn, and @var{num} is the label-number
7246 of the preceding label.
7248 If this macro is not defined, nothing special is output at the end of
7252 @node Exception Region Output
7253 @subsection Assembler Commands for Exception Regions
7255 @c prevent bad page break with this line
7257 This describes commands marking the start and the end of an exception
7261 @findex EH_FRAME_SECTION_NAME
7262 @item EH_FRAME_SECTION_NAME
7263 If defined, a C string constant for the name of the section containing
7264 exception handling frame unwind information. If not defined, GCC will
7265 provide a default definition if the target supports named sections.
7266 @file{crtstuff.c} uses this macro to switch to the appropriate section.
7268 You should define this symbol if your target supports DWARF 2 frame
7269 unwind information and the default definition does not work.
7271 @findex EH_FRAME_IN_DATA_SECTION
7272 @item EH_FRAME_IN_DATA_SECTION
7273 If defined, DWARF 2 frame unwind information will be placed in the
7274 data section even though the target supports named sections. This
7275 might be necessary, for instance, if the system linker does garbage
7276 collection and sections cannot be marked as not to be collected.
7278 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
7281 @findex MASK_RETURN_ADDR
7282 @item MASK_RETURN_ADDR
7283 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
7284 that it does not contain any extraneous set bits in it.
7286 @findex DWARF2_UNWIND_INFO
7287 @item DWARF2_UNWIND_INFO
7288 Define this macro to 0 if your target supports DWARF 2 frame unwind
7289 information, but it does not yet work with exception handling.
7290 Otherwise, if your target supports this information (if it defines
7291 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
7292 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
7295 If this macro is defined to 1, the DWARF 2 unwinder will be the default
7296 exception handling mechanism; otherwise, @code{setjmp}/@code{longjmp} will be used by
7299 If this macro is defined to anything, the DWARF 2 unwinder will be used
7300 instead of inline unwinders and @code{__unwind_function} in the non-@code{setjmp} case.
7302 @findex DWARF_CIE_DATA_ALIGNMENT
7303 @item DWARF_CIE_DATA_ALIGNMENT
7304 This macro need only be defined if the target might save registers in the
7305 function prologue at an offset to the stack pointer that is not aligned to
7306 @code{UNITS_PER_WORD}. The definition should be the negative minimum
7307 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
7308 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
7309 the target supports DWARF 2 frame unwind information.
7313 @deftypefn {Target Hook} void TARGET_ASM_EXCEPTION_SECTION ()
7314 If defined, a function that switches to the section in which the main
7315 exception table is to be placed (@pxref{Sections}). The default is a
7316 function that switches to a section named @code{.gcc_except_table} on
7317 machines that support named sections via
7318 @code{TARGET_ASM_NAMED_SECTION}, otherwise if @option{-fpic} or
7319 @option{-fPIC} is in effect, the @code{data_section}, otherwise the
7320 @code{readonly_data_section}.
7323 @deftypefn {Target Hook} void TARGET_ASM_EH_FRAME_SECTION ()
7324 If defined, a function that switches to the section in which the DWARF 2
7325 frame unwind information to be placed (@pxref{Sections}). The default
7326 is a function that outputs a standard GAS section directive, if
7327 @code{EH_FRAME_SECTION_NAME} is defined, or else a data section
7328 directive followed by a synthetic label.
7331 @node Alignment Output
7332 @subsection Assembler Commands for Alignment
7334 @c prevent bad page break with this line
7335 This describes commands for alignment.
7339 @item JUMP_ALIGN (@var{label})
7340 The alignment (log base 2) to put in front of @var{label}, which is
7341 a common destination of jumps and has no fallthru incoming edge.
7343 This macro need not be defined if you don't want any special alignment
7344 to be done at such a time. Most machine descriptions do not currently
7347 Unless it's necessary to inspect the @var{label} parameter, it is better
7348 to set the variable @var{align_jumps} in the target's
7349 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7350 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
7352 @findex LABEL_ALIGN_AFTER_BARRIER
7353 @item LABEL_ALIGN_AFTER_BARRIER (@var{label})
7354 The alignment (log base 2) to put in front of @var{label}, which follows
7357 This macro need not be defined if you don't want any special alignment
7358 to be done at such a time. Most machine descriptions do not currently
7361 @findex LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7362 @item LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7363 The maximum number of bytes to skip when applying
7364 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
7365 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7368 @item LOOP_ALIGN (@var{label})
7369 The alignment (log base 2) to put in front of @var{label}, which follows
7370 a @code{NOTE_INSN_LOOP_BEG} note.
7372 This macro need not be defined if you don't want any special alignment
7373 to be done at such a time. Most machine descriptions do not currently
7376 Unless it's necessary to inspect the @var{label} parameter, it is better
7377 to set the variable @code{align_loops} in the target's
7378 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7379 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
7381 @findex LOOP_ALIGN_MAX_SKIP
7382 @item LOOP_ALIGN_MAX_SKIP
7383 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
7384 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7387 @item LABEL_ALIGN (@var{label})
7388 The alignment (log base 2) to put in front of @var{label}.
7389 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
7390 the maximum of the specified values is used.
7392 Unless it's necessary to inspect the @var{label} parameter, it is better
7393 to set the variable @code{align_labels} in the target's
7394 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7395 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
7397 @findex LABEL_ALIGN_MAX_SKIP
7398 @item LABEL_ALIGN_MAX_SKIP
7399 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
7400 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7402 @findex ASM_OUTPUT_SKIP
7403 @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
7404 A C statement to output to the stdio stream @var{stream} an assembler
7405 instruction to advance the location counter by @var{nbytes} bytes.
7406 Those bytes should be zero when loaded. @var{nbytes} will be a C
7407 expression of type @code{int}.
7409 @findex ASM_NO_SKIP_IN_TEXT
7410 @item ASM_NO_SKIP_IN_TEXT
7411 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
7412 text section because it fails to put zeros in the bytes that are skipped.
7413 This is true on many Unix systems, where the pseudo--op to skip bytes
7414 produces no-op instructions rather than zeros when used in the text
7417 @findex ASM_OUTPUT_ALIGN
7418 @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
7419 A C statement to output to the stdio stream @var{stream} an assembler
7420 command to advance the location counter to a multiple of 2 to the
7421 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
7423 @findex ASM_OUTPUT_MAX_SKIP_ALIGN
7424 @item ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
7425 A C statement to output to the stdio stream @var{stream} an assembler
7426 command to advance the location counter to a multiple of 2 to the
7427 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
7428 satisfy the alignment request. @var{power} and @var{max_skip} will be
7429 a C expression of type @code{int}.
7433 @node Debugging Info
7434 @section Controlling Debugging Information Format
7436 @c prevent bad page break with this line
7437 This describes how to specify debugging information.
7440 * All Debuggers:: Macros that affect all debugging formats uniformly.
7441 * DBX Options:: Macros enabling specific options in DBX format.
7442 * DBX Hooks:: Hook macros for varying DBX format.
7443 * File Names and DBX:: Macros controlling output of file names in DBX format.
7444 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
7445 * VMS Debug:: Macros for VMS debug format.
7449 @subsection Macros Affecting All Debugging Formats
7451 @c prevent bad page break with this line
7452 These macros affect all debugging formats.
7455 @findex DBX_REGISTER_NUMBER
7456 @item DBX_REGISTER_NUMBER (@var{regno})
7457 A C expression that returns the DBX register number for the compiler
7458 register number @var{regno}. In the default macro provided, the value
7459 of this expression will be @var{regno} itself. But sometimes there are
7460 some registers that the compiler knows about and DBX does not, or vice
7461 versa. In such cases, some register may need to have one number in the
7462 compiler and another for DBX@.
7464 If two registers have consecutive numbers inside GCC, and they can be
7465 used as a pair to hold a multiword value, then they @emph{must} have
7466 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
7467 Otherwise, debuggers will be unable to access such a pair, because they
7468 expect register pairs to be consecutive in their own numbering scheme.
7470 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
7471 does not preserve register pairs, then what you must do instead is
7472 redefine the actual register numbering scheme.
7474 @findex DEBUGGER_AUTO_OFFSET
7475 @item DEBUGGER_AUTO_OFFSET (@var{x})
7476 A C expression that returns the integer offset value for an automatic
7477 variable having address @var{x} (an RTL expression). The default
7478 computation assumes that @var{x} is based on the frame-pointer and
7479 gives the offset from the frame-pointer. This is required for targets
7480 that produce debugging output for DBX or COFF-style debugging output
7481 for SDB and allow the frame-pointer to be eliminated when the
7482 @option{-g} options is used.
7484 @findex DEBUGGER_ARG_OFFSET
7485 @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
7486 A C expression that returns the integer offset value for an argument
7487 having address @var{x} (an RTL expression). The nominal offset is
7490 @findex PREFERRED_DEBUGGING_TYPE
7491 @item PREFERRED_DEBUGGING_TYPE
7492 A C expression that returns the type of debugging output GCC should
7493 produce when the user specifies just @option{-g}. Define
7494 this if you have arranged for GCC to support more than one format of
7495 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7496 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
7497 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
7499 When the user specifies @option{-ggdb}, GCC normally also uses the
7500 value of this macro to select the debugging output format, but with two
7501 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined and
7502 @code{LINKER_DOES_NOT_WORK_WITH_DWARF2} is not defined, GCC uses the
7503 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7504 defined, GCC uses @code{DBX_DEBUG}.
7506 The value of this macro only affects the default debugging output; the
7507 user can always get a specific type of output by using @option{-gstabs},
7508 @option{-gcoff}, @option{-gdwarf-1}, @option{-gdwarf-2}, @option{-gxcoff},
7513 @subsection Specific Options for DBX Output
7515 @c prevent bad page break with this line
7516 These are specific options for DBX output.
7519 @findex DBX_DEBUGGING_INFO
7520 @item DBX_DEBUGGING_INFO
7521 Define this macro if GCC should produce debugging output for DBX
7522 in response to the @option{-g} option.
7524 @findex XCOFF_DEBUGGING_INFO
7525 @item XCOFF_DEBUGGING_INFO
7526 Define this macro if GCC should produce XCOFF format debugging output
7527 in response to the @option{-g} option. This is a variant of DBX format.
7529 @findex DEFAULT_GDB_EXTENSIONS
7530 @item DEFAULT_GDB_EXTENSIONS
7531 Define this macro to control whether GCC should by default generate
7532 GDB's extended version of DBX debugging information (assuming DBX-format
7533 debugging information is enabled at all). If you don't define the
7534 macro, the default is 1: always generate the extended information
7535 if there is any occasion to.
7537 @findex DEBUG_SYMS_TEXT
7538 @item DEBUG_SYMS_TEXT
7539 Define this macro if all @code{.stabs} commands should be output while
7540 in the text section.
7542 @findex ASM_STABS_OP
7544 A C string constant, including spacing, naming the assembler pseudo op to
7545 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
7546 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
7547 applies only to DBX debugging information format.
7549 @findex ASM_STABD_OP
7551 A C string constant, including spacing, naming the assembler pseudo op to
7552 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
7553 value is the current location. If you don't define this macro,
7554 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
7557 @findex ASM_STABN_OP
7559 A C string constant, including spacing, naming the assembler pseudo op to
7560 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
7561 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
7562 macro applies only to DBX debugging information format.
7564 @findex DBX_NO_XREFS
7566 Define this macro if DBX on your system does not support the construct
7567 @samp{xs@var{tagname}}. On some systems, this construct is used to
7568 describe a forward reference to a structure named @var{tagname}.
7569 On other systems, this construct is not supported at all.
7571 @findex DBX_CONTIN_LENGTH
7572 @item DBX_CONTIN_LENGTH
7573 A symbol name in DBX-format debugging information is normally
7574 continued (split into two separate @code{.stabs} directives) when it
7575 exceeds a certain length (by default, 80 characters). On some
7576 operating systems, DBX requires this splitting; on others, splitting
7577 must not be done. You can inhibit splitting by defining this macro
7578 with the value zero. You can override the default splitting-length by
7579 defining this macro as an expression for the length you desire.
7581 @findex DBX_CONTIN_CHAR
7582 @item DBX_CONTIN_CHAR
7583 Normally continuation is indicated by adding a @samp{\} character to
7584 the end of a @code{.stabs} string when a continuation follows. To use
7585 a different character instead, define this macro as a character
7586 constant for the character you want to use. Do not define this macro
7587 if backslash is correct for your system.
7589 @findex DBX_STATIC_STAB_DATA_SECTION
7590 @item DBX_STATIC_STAB_DATA_SECTION
7591 Define this macro if it is necessary to go to the data section before
7592 outputting the @samp{.stabs} pseudo-op for a non-global static
7595 @findex DBX_TYPE_DECL_STABS_CODE
7596 @item DBX_TYPE_DECL_STABS_CODE
7597 The value to use in the ``code'' field of the @code{.stabs} directive
7598 for a typedef. The default is @code{N_LSYM}.
7600 @findex DBX_STATIC_CONST_VAR_CODE
7601 @item DBX_STATIC_CONST_VAR_CODE
7602 The value to use in the ``code'' field of the @code{.stabs} directive
7603 for a static variable located in the text section. DBX format does not
7604 provide any ``right'' way to do this. The default is @code{N_FUN}.
7606 @findex DBX_REGPARM_STABS_CODE
7607 @item DBX_REGPARM_STABS_CODE
7608 The value to use in the ``code'' field of the @code{.stabs} directive
7609 for a parameter passed in registers. DBX format does not provide any
7610 ``right'' way to do this. The default is @code{N_RSYM}.
7612 @findex DBX_REGPARM_STABS_LETTER
7613 @item DBX_REGPARM_STABS_LETTER
7614 The letter to use in DBX symbol data to identify a symbol as a parameter
7615 passed in registers. DBX format does not customarily provide any way to
7616 do this. The default is @code{'P'}.
7618 @findex DBX_MEMPARM_STABS_LETTER
7619 @item DBX_MEMPARM_STABS_LETTER
7620 The letter to use in DBX symbol data to identify a symbol as a stack
7621 parameter. The default is @code{'p'}.
7623 @findex DBX_FUNCTION_FIRST
7624 @item DBX_FUNCTION_FIRST
7625 Define this macro if the DBX information for a function and its
7626 arguments should precede the assembler code for the function. Normally,
7627 in DBX format, the debugging information entirely follows the assembler
7630 @findex DBX_LBRAC_FIRST
7631 @item DBX_LBRAC_FIRST
7632 Define this macro if the @code{N_LBRAC} symbol for a block should
7633 precede the debugging information for variables and functions defined in
7634 that block. Normally, in DBX format, the @code{N_LBRAC} symbol comes
7637 @findex DBX_BLOCKS_FUNCTION_RELATIVE
7638 @item DBX_BLOCKS_FUNCTION_RELATIVE
7639 Define this macro if the value of a symbol describing the scope of a
7640 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
7641 of the enclosing function. Normally, GCC uses an absolute address.
7643 @findex DBX_USE_BINCL
7645 Define this macro if GCC should generate @code{N_BINCL} and
7646 @code{N_EINCL} stabs for included header files, as on Sun systems. This
7647 macro also directs GCC to output a type number as a pair of a file
7648 number and a type number within the file. Normally, GCC does not
7649 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
7650 number for a type number.
7654 @subsection Open-Ended Hooks for DBX Format
7656 @c prevent bad page break with this line
7657 These are hooks for DBX format.
7660 @findex DBX_OUTPUT_LBRAC
7661 @item DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
7662 Define this macro to say how to output to @var{stream} the debugging
7663 information for the start of a scope level for variable names. The
7664 argument @var{name} is the name of an assembler symbol (for use with
7665 @code{assemble_name}) whose value is the address where the scope begins.
7667 @findex DBX_OUTPUT_RBRAC
7668 @item DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
7669 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
7671 @findex DBX_OUTPUT_ENUM
7672 @item DBX_OUTPUT_ENUM (@var{stream}, @var{type})
7673 Define this macro if the target machine requires special handling to
7674 output an enumeration type. The definition should be a C statement
7675 (sans semicolon) to output the appropriate information to @var{stream}
7676 for the type @var{type}.
7678 @findex DBX_OUTPUT_FUNCTION_END
7679 @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
7680 Define this macro if the target machine requires special output at the
7681 end of the debugging information for a function. The definition should
7682 be a C statement (sans semicolon) to output the appropriate information
7683 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
7686 @findex DBX_OUTPUT_STANDARD_TYPES
7687 @item DBX_OUTPUT_STANDARD_TYPES (@var{syms})
7688 Define this macro if you need to control the order of output of the
7689 standard data types at the beginning of compilation. The argument
7690 @var{syms} is a @code{tree} which is a chain of all the predefined
7691 global symbols, including names of data types.
7693 Normally, DBX output starts with definitions of the types for integers
7694 and characters, followed by all the other predefined types of the
7695 particular language in no particular order.
7697 On some machines, it is necessary to output different particular types
7698 first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
7699 those symbols in the necessary order. Any predefined types that you
7700 don't explicitly output will be output afterward in no particular order.
7702 Be careful not to define this macro so that it works only for C@. There
7703 are no global variables to access most of the built-in types, because
7704 another language may have another set of types. The way to output a
7705 particular type is to look through @var{syms} to see if you can find it.
7711 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7712 if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
7714 dbxout_symbol (decl);
7720 This does nothing if the expected type does not exist.
7722 See the function @code{init_decl_processing} in @file{c-decl.c} to find
7723 the names to use for all the built-in C types.
7725 Here is another way of finding a particular type:
7727 @c this is still overfull. --mew 10feb93
7731 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7732 if (TREE_CODE (decl) == TYPE_DECL
7733 && (TREE_CODE (TREE_TYPE (decl))
7735 && TYPE_PRECISION (TREE_TYPE (decl)) == 16
7736 && TYPE_UNSIGNED (TREE_TYPE (decl)))
7738 /* @r{This must be @code{unsigned short}.} */
7739 dbxout_symbol (decl);
7745 @findex NO_DBX_FUNCTION_END
7746 @item NO_DBX_FUNCTION_END
7747 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
7748 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
7749 On those machines, define this macro to turn this feature off without
7750 disturbing the rest of the gdb extensions.
7754 @node File Names and DBX
7755 @subsection File Names in DBX Format
7757 @c prevent bad page break with this line
7758 This describes file names in DBX format.
7761 @findex DBX_WORKING_DIRECTORY
7762 @item DBX_WORKING_DIRECTORY
7763 Define this if DBX wants to have the current directory recorded in each
7766 Note that the working directory is always recorded if GDB extensions are
7769 @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME
7770 @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
7771 A C statement to output DBX debugging information to the stdio stream
7772 @var{stream} which indicates that file @var{name} is the main source
7773 file---the file specified as the input file for compilation.
7774 This macro is called only once, at the beginning of compilation.
7776 This macro need not be defined if the standard form of output
7777 for DBX debugging information is appropriate.
7779 @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY
7780 @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
7781 A C statement to output DBX debugging information to the stdio stream
7782 @var{stream} which indicates that the current directory during
7783 compilation is named @var{name}.
7785 This macro need not be defined if the standard form of output
7786 for DBX debugging information is appropriate.
7788 @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END
7789 @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
7790 A C statement to output DBX debugging information at the end of
7791 compilation of the main source file @var{name}.
7793 If you don't define this macro, nothing special is output at the end
7794 of compilation, which is correct for most machines.
7796 @findex DBX_OUTPUT_SOURCE_FILENAME
7797 @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7798 A C statement to output DBX debugging information to the stdio stream
7799 @var{stream} which indicates that file @var{name} is the current source
7800 file. This output is generated each time input shifts to a different
7801 source file as a result of @samp{#include}, the end of an included file,
7802 or a @samp{#line} command.
7804 This macro need not be defined if the standard form of output
7805 for DBX debugging information is appropriate.
7810 @subsection Macros for SDB and DWARF Output
7812 @c prevent bad page break with this line
7813 Here are macros for SDB and DWARF output.
7816 @findex SDB_DEBUGGING_INFO
7817 @item SDB_DEBUGGING_INFO
7818 Define this macro if GCC should produce COFF-style debugging output
7819 for SDB in response to the @option{-g} option.
7821 @findex DWARF_DEBUGGING_INFO
7822 @item DWARF_DEBUGGING_INFO
7823 Define this macro if GCC should produce dwarf format debugging output
7824 in response to the @option{-g} option.
7826 @findex DWARF2_DEBUGGING_INFO
7827 @item DWARF2_DEBUGGING_INFO
7828 Define this macro if GCC should produce dwarf version 2 format
7829 debugging output in response to the @option{-g} option.
7831 To support optional call frame debugging information, you must also
7832 define @code{INCOMING_RETURN_ADDR_RTX} and either set
7833 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
7834 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
7835 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
7837 @findex DWARF2_FRAME_INFO
7838 @item DWARF2_FRAME_INFO
7839 Define this macro to a nonzero value if GCC should always output
7840 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
7841 (@pxref{Exception Region Output} is nonzero, GCC will output this
7842 information not matter how you define @code{DWARF2_FRAME_INFO}.
7844 @findex LINKER_DOES_NOT_WORK_WITH_DWARF2
7845 @item LINKER_DOES_NOT_WORK_WITH_DWARF2
7846 Define this macro if the linker does not work with Dwarf version 2.
7847 Normally, if the user specifies only @option{-ggdb} GCC will use Dwarf
7848 version 2 if available; this macro disables this. See the description
7849 of the @code{PREFERRED_DEBUGGING_TYPE} macro for more details.
7851 @findex DWARF2_GENERATE_TEXT_SECTION_LABEL
7852 @item DWARF2_GENERATE_TEXT_SECTION_LABEL
7853 By default, the Dwarf 2 debugging information generator will generate a
7854 label to mark the beginning of the text section. If it is better simply
7855 to use the name of the text section itself, rather than an explicit label,
7856 to indicate the beginning of the text section, define this macro to zero.
7858 @findex DWARF2_ASM_LINE_DEBUG_INFO
7859 @item DWARF2_ASM_LINE_DEBUG_INFO
7860 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
7861 line debug info sections. This will result in much more compact line number
7862 tables, and hence is desirable if it works.
7864 @findex PUT_SDB_@dots{}
7865 @item PUT_SDB_@dots{}
7866 Define these macros to override the assembler syntax for the special
7867 SDB assembler directives. See @file{sdbout.c} for a list of these
7868 macros and their arguments. If the standard syntax is used, you need
7869 not define them yourself.
7873 Some assemblers do not support a semicolon as a delimiter, even between
7874 SDB assembler directives. In that case, define this macro to be the
7875 delimiter to use (usually @samp{\n}). It is not necessary to define
7876 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
7879 @findex SDB_GENERATE_FAKE
7880 @item SDB_GENERATE_FAKE
7881 Define this macro to override the usual method of constructing a dummy
7882 name for anonymous structure and union types. See @file{sdbout.c} for
7885 @findex SDB_ALLOW_UNKNOWN_REFERENCES
7886 @item SDB_ALLOW_UNKNOWN_REFERENCES
7887 Define this macro to allow references to unknown structure,
7888 union, or enumeration tags to be emitted. Standard COFF does not
7889 allow handling of unknown references, MIPS ECOFF has support for
7892 @findex SDB_ALLOW_FORWARD_REFERENCES
7893 @item SDB_ALLOW_FORWARD_REFERENCES
7894 Define this macro to allow references to structure, union, or
7895 enumeration tags that have not yet been seen to be handled. Some
7896 assemblers choke if forward tags are used, while some require it.
7901 @subsection Macros for VMS Debug Format
7903 @c prevent bad page break with this line
7904 Here are macros for VMS debug format.
7907 @findex VMS_DEBUGGING_INFO
7908 @item VMS_DEBUGGING_INFO
7909 Define this macro if GCC should produce debugging output for VMS
7910 in response to the @option{-g} option. The default behavior for VMS
7911 is to generate minimal debug info for a traceback in the absence of
7912 @option{-g} unless explicitly overridden with @option{-g0}. This
7913 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
7914 @code{OVERRIDE_OPTIONS}.
7917 @node Floating Point
7918 @section Cross Compilation and Floating Point
7919 @cindex cross compilation and floating point
7920 @cindex floating point and cross compilation
7922 While all modern machines use twos-complement representation for integers,
7923 there are a variety of representations for floating point numbers. This
7924 means that in a cross-compiler the representation of floating point numbers
7925 in the compiled program may be different from that used in the machine
7926 doing the compilation.
7928 Because different representation systems may offer different amounts of
7929 range and precision, all floating point constants must be represented in
7930 the target machine's format. Therefore, the cross compiler cannot
7931 safely use the host machine's floating point arithmetic; it must emulate
7932 the target's arithmetic. To ensure consistency, GCC always uses
7933 emulation to work with floating point values, even when the host and
7934 target floating point formats are identical.
7936 The following macros are provided by @file{real.h} for the compiler to
7937 use. All parts of the compiler which generate or optimize
7938 floating-point calculations must use these macros. They may evaluate
7939 their operands more than once, so operands must not have side effects.
7941 @defmac REAL_VALUE_TYPE
7942 The C data type to be used to hold a floating point value in the target
7943 machine's format. Typically this is a @code{struct} containing an
7944 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
7948 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
7949 Compares for equality the two values, @var{x} and @var{y}. If the target
7950 floating point format supports negative zeroes and/or NaNs,
7951 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
7952 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
7955 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
7956 Tests whether @var{x} is less than @var{y}.
7960 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_LDEXP (REAL_VALUE_TYPE @var{x}, int @var{scale})
7961 Multiplies @var{x} by 2 raised to the power @var{scale}.
7964 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
7965 Truncates @var{x} to a signed integer, rounding toward zero.
7968 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
7969 Truncates @var{x} to an unsigned integer, rounding toward zero. If
7970 @var{x} is negative, returns zero.
7973 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_RNDZINT (REAL_VALUE_TYPE @var{x})
7974 Rounds the target-machine floating point value @var{x} towards zero to an
7975 integer value, but leaves it represented as a floating point number.
7978 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_UNSIGNED_RNDZINT (REAL_VALUE_TYPE @var{x})
7979 Rounds the target-machine floating point value @var{x} towards zero to an
7980 unsigned integer value, but leaves it represented as a floating point
7981 number. If @var{x} is negative, returns (positive) zero.
7984 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
7985 Converts @var{string} into a floating point number in the target machine's
7986 representation for mode @var{mode}. This routine can handle both
7987 decimal and hexadecimal floating point constants, using the syntax
7988 defined by the C language for both.
7991 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
7992 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
7995 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
7996 Determines whether @var{x} represents infinity (positive or negative).
7999 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8000 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8003 @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})
8004 Calculates an arithmetic operation on the two floating point values
8005 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8008 The operation to be performed is specified by @var{code}. Only the
8009 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8010 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8012 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8013 target's floating point format cannot represent infinity, it will call
8014 @code{abort}. Callers should check for this situation first, using
8015 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8018 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8019 Returns the negative of the floating point value @var{x}.
8022 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8023 Returns the absolute value of @var{x}.
8026 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8027 Truncates the floating point value @var{x} to fit in @var{mode}. The
8028 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8029 appropriate bit pattern to be output asa floating constant whose
8030 precision accords with mode @var{mode}.
8033 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8034 Converts a floating point value @var{x} into a double-precision integer
8035 which is then stored into @var{low} and @var{high}. If the value is not
8036 integral, it is truncated.
8039 @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})
8040 @findex REAL_VALUE_FROM_INT
8041 Converts a double-precision integer found in @var{low} and @var{high},
8042 into a floating point value which is then stored into @var{x}. The
8043 value is truncated to fit in mode @var{mode}.
8046 @node Mode Switching
8047 @section Mode Switching Instructions
8048 @cindex mode switching
8049 The following macros control mode switching optimizations:
8052 @findex OPTIMIZE_MODE_SWITCHING
8053 @item OPTIMIZE_MODE_SWITCHING (@var{entity})
8054 Define this macro if the port needs extra instructions inserted for mode
8055 switching in an optimizing compilation.
8057 For an example, the SH4 can perform both single and double precision
8058 floating point operations, but to perform a single precision operation,
8059 the FPSCR PR bit has to be cleared, while for a double precision
8060 operation, this bit has to be set. Changing the PR bit requires a general
8061 purpose register as a scratch register, hence these FPSCR sets have to
8062 be inserted before reload, i.e.@: you can't put this into instruction emitting
8063 or @code{MACHINE_DEPENDENT_REORG}.
8065 You can have multiple entities that are mode-switched, and select at run time
8066 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
8067 return nonzero for any @var{entity} that needs mode-switching.
8068 If you define this macro, you also have to define
8069 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
8070 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
8071 @code{NORMAL_MODE} is optional.
8073 @findex NUM_MODES_FOR_MODE_SWITCHING
8074 @item NUM_MODES_FOR_MODE_SWITCHING
8075 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
8076 initializer for an array of integers. Each initializer element
8077 N refers to an entity that needs mode switching, and specifies the number
8078 of different modes that might need to be set for this entity.
8079 The position of the initializer in the initializer - starting counting at
8080 zero - determines the integer that is used to refer to the mode-switched
8082 In macros that take mode arguments / yield a mode result, modes are
8083 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
8084 switch is needed / supplied.
8087 @item MODE_NEEDED (@var{entity}, @var{insn})
8088 @var{entity} is an integer specifying a mode-switched entity. If
8089 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
8090 return an integer value not larger than the corresponding element in
8091 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
8092 be switched into prior to the execution of @var{insn}.
8095 @item NORMAL_MODE (@var{entity})
8096 If this macro is defined, it is evaluated for every @var{entity} that needs
8097 mode switching. It should evaluate to an integer, which is a mode that
8098 @var{entity} is assumed to be switched to at function entry and exit.
8100 @findex MODE_PRIORITY_TO_MODE
8101 @item MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
8102 This macro specifies the order in which modes for @var{entity} are processed.
8103 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
8104 lowest. The value of the macro should be an integer designating a mode
8105 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
8106 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
8107 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
8109 @findex EMIT_MODE_SET
8110 @item EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
8111 Generate one or more insns to set @var{entity} to @var{mode}.
8112 @var{hard_reg_live} is the set of hard registers live at the point where
8113 the insn(s) are to be inserted.
8116 @node Target Attributes
8117 @section Defining target-specific uses of @code{__attribute__}
8118 @cindex target attributes
8119 @cindex machine attributes
8120 @cindex attributes, target-specific
8122 Target-specific attributes may be defined for functions, data and types.
8123 These are described using the following target hooks; they also need to
8124 be documented in @file{extend.texi}.
8126 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
8127 If defined, this target hook points to an array of @samp{struct
8128 attribute_spec} (defined in @file{tree.h}) specifying the machine
8129 specific attributes for this target and some of the restrictions on the
8130 entities to which these attributes are applied and the arguments they
8134 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8135 If defined, this target hook is a function which returns zero if the attributes on
8136 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8137 and two if they are nearly compatible (which causes a warning to be
8138 generated). If this is not defined, machine-specific attributes are
8139 supposed always to be compatible.
8142 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
8143 If defined, this target hook is a function which assigns default attributes to
8144 newly defined @var{type}.
8147 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8148 Define this target hook if the merging of type attributes needs special
8149 handling. If defined, the result is a list of the combined
8150 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
8151 that @code{comptypes} has already been called and returned 1. This
8152 function may call @code{merge_attributes} to handle machine-independent
8156 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
8157 Define this target hook if the merging of decl attributes needs special
8158 handling. If defined, the result is a list of the combined
8159 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
8160 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
8161 when this is needed are when one attribute overrides another, or when an
8162 attribute is nullified by a subsequent definition. This function may
8163 call @code{merge_attributes} to handle machine-independent merging.
8165 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
8166 If the only target-specific handling you require is @samp{dllimport} for
8167 Windows targets, you should define the macro
8168 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. This links in a function
8169 called @code{merge_dllimport_decl_attributes} which can then be defined
8170 as the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. This is done
8171 in @file{i386/cygwin.h} and @file{i386/i386.c}, for example.
8174 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
8175 Define this target hook if you want to be able to add attributes to a decl
8176 when it is being created. This is normally useful for back ends which
8177 wish to implement a pragma by using the attributes which correspond to
8178 the pragma's effect. The @var{node} argument is the decl which is being
8179 created. The @var{attr_ptr} argument is a pointer to the attribute list
8180 for this decl. The list itself should not be modified, since it may be
8181 shared with other decls, but attributes may be chained on the head of
8182 the list and @code{*@var{attr_ptr}} modified to point to the new
8183 attributes, or a copy of the list may be made if further changes are
8187 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
8189 This target hook returns @code{true} if it is ok to inline @var{fndecl}
8190 into the current function, despite its having target-specific
8191 attributes, @code{false} otherwise. By default, if a function has a
8192 target specific attribute attached to it, it will not be inlined.
8195 @node MIPS Coprocessors
8196 @section Defining coprocessor specifics for MIPS targets.
8197 @cindex MIPS coprocessor-definition macros
8199 The MIPS specification allows MIPS implementations to have as many as 4
8200 coprocessors, each with as many as 32 private registers. gcc supports
8201 accessing these registers and transferring values between the registers
8202 and memory using asm-ized variables. For example:
8205 register unsigned int cp0count asm ("c0r1");
8211 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
8212 names may be added as described below, or the default names may be
8213 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
8215 Coprocessor registers are assumed to be epilogue-used; sets to them will
8216 be preserved even if it does not appear that the register is used again
8217 later in the function.
8219 Another note: according to the MIPS spec, coprocessor 1 (if present) is
8220 the FPU. One accesses COP1 registers through standard mips
8221 floating-point support; they are not included in this mechanism.
8223 There is one macro used in defining the MIPS coprocessor interface which
8224 you may want to override in subtargets; it is described below.
8228 @item ALL_COP_ADDITIONAL_REGISTER_NAMES
8229 @findex ALL_COP_ADDITIONAL_REGISTER_NAMES
8230 A comma-separated list (with leading comma) of pairs describing the
8231 alternate names of coprocessor registers. The format of each entry should be
8233 @{ @var{alternatename}, @var{register_number}@}
8240 @section Miscellaneous Parameters
8241 @cindex parameters, miscellaneous
8243 @c prevent bad page break with this line
8244 Here are several miscellaneous parameters.
8247 @item PREDICATE_CODES
8248 @findex PREDICATE_CODES
8249 Define this if you have defined special-purpose predicates in the file
8250 @file{@var{machine}.c}. This macro is called within an initializer of an
8251 array of structures. The first field in the structure is the name of a
8252 predicate and the second field is an array of rtl codes. For each
8253 predicate, list all rtl codes that can be in expressions matched by the
8254 predicate. The list should have a trailing comma. Here is an example
8255 of two entries in the list for a typical RISC machine:
8258 #define PREDICATE_CODES \
8259 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
8260 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
8263 Defining this macro does not affect the generated code (however,
8264 incorrect definitions that omit an rtl code that may be matched by the
8265 predicate can cause the compiler to malfunction). Instead, it allows
8266 the table built by @file{genrecog} to be more compact and efficient,
8267 thus speeding up the compiler. The most important predicates to include
8268 in the list specified by this macro are those used in the most insn
8271 For each predicate function named in @code{PREDICATE_CODES}, a
8272 declaration will be generated in @file{insn-codes.h}.
8274 @item SPECIAL_MODE_PREDICATES
8275 @findex SPECIAL_MODE_PREDICATES
8276 Define this if you have special predicates that know special things
8277 about modes. Genrecog will warn about certain forms of
8278 @code{match_operand} without a mode; if the operand predicate is
8279 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
8282 Here is an example from the IA-32 port (@code{ext_register_operand}
8283 specially checks for @code{HImode} or @code{SImode} in preparation
8284 for a byte extraction from @code{%ah} etc.).
8287 #define SPECIAL_MODE_PREDICATES \
8288 "ext_register_operand",
8291 @findex CASE_VECTOR_MODE
8292 @item CASE_VECTOR_MODE
8293 An alias for a machine mode name. This is the machine mode that
8294 elements of a jump-table should have.
8296 @findex CASE_VECTOR_SHORTEN_MODE
8297 @item CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
8298 Optional: return the preferred mode for an @code{addr_diff_vec}
8299 when the minimum and maximum offset are known. If you define this,
8300 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
8301 To make this work, you also have to define INSN_ALIGN and
8302 make the alignment for @code{addr_diff_vec} explicit.
8303 The @var{body} argument is provided so that the offset_unsigned and scale
8304 flags can be updated.
8306 @findex CASE_VECTOR_PC_RELATIVE
8307 @item CASE_VECTOR_PC_RELATIVE
8308 Define this macro to be a C expression to indicate when jump-tables
8309 should contain relative addresses. If jump-tables never contain
8310 relative addresses, then you need not define this macro.
8312 @findex CASE_DROPS_THROUGH
8313 @item CASE_DROPS_THROUGH
8314 Define this if control falls through a @code{case} insn when the index
8315 value is out of range. This means the specified default-label is
8316 actually ignored by the @code{case} insn proper.
8318 @findex CASE_VALUES_THRESHOLD
8319 @item CASE_VALUES_THRESHOLD
8320 Define this to be the smallest number of different values for which it
8321 is best to use a jump-table instead of a tree of conditional branches.
8322 The default is four for machines with a @code{casesi} instruction and
8323 five otherwise. This is best for most machines.
8325 @findex WORD_REGISTER_OPERATIONS
8326 @item WORD_REGISTER_OPERATIONS
8327 Define this macro if operations between registers with integral mode
8328 smaller than a word are always performed on the entire register.
8329 Most RISC machines have this property and most CISC machines do not.
8331 @findex LOAD_EXTEND_OP
8332 @item LOAD_EXTEND_OP (@var{mode})
8333 Define this macro to be a C expression indicating when insns that read
8334 memory in @var{mode}, an integral mode narrower than a word, set the
8335 bits outside of @var{mode} to be either the sign-extension or the
8336 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
8337 of @var{mode} for which the
8338 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
8339 @code{NIL} for other modes.
8341 This macro is not called with @var{mode} non-integral or with a width
8342 greater than or equal to @code{BITS_PER_WORD}, so you may return any
8343 value in this case. Do not define this macro if it would always return
8344 @code{NIL}. On machines where this macro is defined, you will normally
8345 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
8347 @findex SHORT_IMMEDIATES_SIGN_EXTEND
8348 @item SHORT_IMMEDIATES_SIGN_EXTEND
8349 Define this macro if loading short immediate values into registers sign
8352 @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC
8353 @item FIXUNS_TRUNC_LIKE_FIX_TRUNC
8354 Define this macro if the same instructions that convert a floating
8355 point number to a signed fixed point number also convert validly to an
8360 The maximum number of bytes that a single instruction can move quickly
8361 between memory and registers or between two memory locations.
8363 @findex MAX_MOVE_MAX
8365 The maximum number of bytes that a single instruction can move quickly
8366 between memory and registers or between two memory locations. If this
8367 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
8368 constant value that is the largest value that @code{MOVE_MAX} can have
8371 @findex SHIFT_COUNT_TRUNCATED
8372 @item SHIFT_COUNT_TRUNCATED
8373 A C expression that is nonzero if on this machine the number of bits
8374 actually used for the count of a shift operation is equal to the number
8375 of bits needed to represent the size of the object being shifted. When
8376 this macro is nonzero, the compiler will assume that it is safe to omit
8377 a sign-extend, zero-extend, and certain bitwise `and' instructions that
8378 truncates the count of a shift operation. On machines that have
8379 instructions that act on bit-fields at variable positions, which may
8380 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
8381 also enables deletion of truncations of the values that serve as
8382 arguments to bit-field instructions.
8384 If both types of instructions truncate the count (for shifts) and
8385 position (for bit-field operations), or if no variable-position bit-field
8386 instructions exist, you should define this macro.
8388 However, on some machines, such as the 80386 and the 680x0, truncation
8389 only applies to shift operations and not the (real or pretended)
8390 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
8391 such machines. Instead, add patterns to the @file{md} file that include
8392 the implied truncation of the shift instructions.
8394 You need not define this macro if it would always have the value of zero.
8396 @findex TRULY_NOOP_TRUNCATION
8397 @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
8398 A C expression which is nonzero if on this machine it is safe to
8399 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
8400 bits (where @var{outprec} is smaller than @var{inprec}) by merely
8401 operating on it as if it had only @var{outprec} bits.
8403 On many machines, this expression can be 1.
8405 @c rearranged this, removed the phrase "it is reported that". this was
8406 @c to fix an overfull hbox. --mew 10feb93
8407 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
8408 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
8409 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
8410 such cases may improve things.
8412 @findex STORE_FLAG_VALUE
8413 @item STORE_FLAG_VALUE
8414 A C expression describing the value returned by a comparison operator
8415 with an integral mode and stored by a store-flag instruction
8416 (@samp{s@var{cond}}) when the condition is true. This description must
8417 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
8418 comparison operators whose results have a @code{MODE_INT} mode.
8420 A value of 1 or @minus{}1 means that the instruction implementing the
8421 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
8422 and 0 when the comparison is false. Otherwise, the value indicates
8423 which bits of the result are guaranteed to be 1 when the comparison is
8424 true. This value is interpreted in the mode of the comparison
8425 operation, which is given by the mode of the first operand in the
8426 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
8427 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
8430 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
8431 generate code that depends only on the specified bits. It can also
8432 replace comparison operators with equivalent operations if they cause
8433 the required bits to be set, even if the remaining bits are undefined.
8434 For example, on a machine whose comparison operators return an
8435 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
8436 @samp{0x80000000}, saying that just the sign bit is relevant, the
8440 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
8447 (ashift:SI @var{x} (const_int @var{n}))
8451 where @var{n} is the appropriate shift count to move the bit being
8452 tested into the sign bit.
8454 There is no way to describe a machine that always sets the low-order bit
8455 for a true value, but does not guarantee the value of any other bits,
8456 but we do not know of any machine that has such an instruction. If you
8457 are trying to port GCC to such a machine, include an instruction to
8458 perform a logical-and of the result with 1 in the pattern for the
8459 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
8461 Often, a machine will have multiple instructions that obtain a value
8462 from a comparison (or the condition codes). Here are rules to guide the
8463 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
8468 Use the shortest sequence that yields a valid definition for
8469 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
8470 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
8471 comparison operators to do so because there may be opportunities to
8472 combine the normalization with other operations.
8475 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
8476 slightly preferred on machines with expensive jumps and 1 preferred on
8480 As a second choice, choose a value of @samp{0x80000001} if instructions
8481 exist that set both the sign and low-order bits but do not define the
8485 Otherwise, use a value of @samp{0x80000000}.
8488 Many machines can produce both the value chosen for
8489 @code{STORE_FLAG_VALUE} and its negation in the same number of
8490 instructions. On those machines, you should also define a pattern for
8491 those cases, e.g., one matching
8494 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
8497 Some machines can also perform @code{and} or @code{plus} operations on
8498 condition code values with less instructions than the corresponding
8499 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
8500 machines, define the appropriate patterns. Use the names @code{incscc}
8501 and @code{decscc}, respectively, for the patterns which perform
8502 @code{plus} or @code{minus} operations on condition code values. See
8503 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
8504 find such instruction sequences on other machines.
8506 You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
8509 @findex FLOAT_STORE_FLAG_VALUE
8510 @item FLOAT_STORE_FLAG_VALUE (@var{mode})
8511 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
8512 returned when comparison operators with floating-point results are true.
8513 Define this macro on machine that have comparison operations that return
8514 floating-point values. If there are no such operations, do not define
8519 An alias for the machine mode for pointers. On most machines, define
8520 this to be the integer mode corresponding to the width of a hardware
8521 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
8522 On some machines you must define this to be one of the partial integer
8523 modes, such as @code{PSImode}.
8525 The width of @code{Pmode} must be at least as large as the value of
8526 @code{POINTER_SIZE}. If it is not equal, you must define the macro
8527 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
8530 @findex FUNCTION_MODE
8532 An alias for the machine mode used for memory references to functions
8533 being called, in @code{call} RTL expressions. On most machines this
8534 should be @code{QImode}.
8536 @findex INTEGRATE_THRESHOLD
8537 @item INTEGRATE_THRESHOLD (@var{decl})
8538 A C expression for the maximum number of instructions above which the
8539 function @var{decl} should not be inlined. @var{decl} is a
8540 @code{FUNCTION_DECL} node.
8542 The default definition of this macro is 64 plus 8 times the number of
8543 arguments that the function accepts. Some people think a larger
8544 threshold should be used on RISC machines.
8546 @findex STDC_0_IN_SYSTEM_HEADERS
8547 @item STDC_0_IN_SYSTEM_HEADERS
8548 In normal operation, the preprocessor expands @code{__STDC__} to the
8549 constant 1, to signify that GCC conforms to ISO Standard C@. On some
8550 hosts, like Solaris, the system compiler uses a different convention,
8551 where @code{__STDC__} is normally 0, but is 1 if the user specifies
8552 strict conformance to the C Standard.
8554 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
8555 convention when processing system header files, but when processing user
8556 files @code{__STDC__} will always expand to 1.
8558 @findex SCCS_DIRECTIVE
8559 @item SCCS_DIRECTIVE
8560 Define this if the preprocessor should ignore @code{#sccs} directives
8561 and print no error message.
8563 @findex NO_IMPLICIT_EXTERN_C
8564 @item NO_IMPLICIT_EXTERN_C
8565 Define this macro if the system header files support C++ as well as C@.
8566 This macro inhibits the usual method of using system header files in
8567 C++, which is to pretend that the file's contents are enclosed in
8568 @samp{extern "C" @{@dots{}@}}.
8570 @findex HANDLE_PRAGMA
8571 @item HANDLE_PRAGMA (@var{getc}, @var{ungetc}, @var{name})
8572 This macro is no longer supported. You must use
8573 @code{REGISTER_TARGET_PRAGMAS} instead.
8575 @findex REGISTER_TARGET_PRAGMAS
8578 @item REGISTER_TARGET_PRAGMAS (@var{pfile})
8579 Define this macro if you want to implement any target-specific pragmas.
8580 If defined, it is a C expression which makes a series of calls to
8581 @code{cpp_register_pragma} for each pragma, with @var{pfile} passed as
8582 the first argument to to these functions. The macro may also do any
8583 setup required for the pragmas.
8585 The primary reason to define this macro is to provide compatibility with
8586 other compilers for the same target. In general, we discourage
8587 definition of target-specific pragmas for GCC@.
8589 If the pragma can be implemented by attributes then you should consider
8590 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
8592 Preprocessor macros that appear on pragma lines are not expanded. All
8593 @samp{#pragma} directives that do not match any registered pragma are
8594 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
8596 @deftypefun void cpp_register_pragma (cpp_reader *@var{pfile}, const char *@var{space}, const char *@var{name}, void (*@var{callback}) (cpp_reader *))
8598 Each call to @code{cpp_register_pragma} establishes one pragma. The
8599 @var{callback} routine will be called when the preprocessor encounters a
8603 #pragma [@var{space}] @var{name} @dots{}
8606 @var{space} is the case-sensitive namespace of the pragma, or
8607 @code{NULL} to put the pragma in the global namespace. The callback
8608 routine receives @var{pfile} as its first argument, which can be passed
8609 on to cpplib's functions if necessary. You can lex tokens after the
8610 @var{name} by calling @code{c_lex}. Tokens that are not read by the
8611 callback will be silently ignored. The end of the line is indicated by
8612 a token of type @code{CPP_EOF}.
8614 For an example use of this routine, see @file{c4x.h} and the callback
8615 routines defined in @file{c4x-c.c}.
8617 Note that the use of @code{c_lex} is specific to the C and C++
8618 compilers. It will not work in the Java or Fortran compilers, or any
8619 other language compilers for that matter. Thus if @code{c_lex} is going
8620 to be called from target-specific code, it must only be done so when
8621 building the C and C++ compilers. This can be done by defining the
8622 variables @code{c_target_objs} and @code{cxx_target_objs} in the
8623 target entry in the @file{config.gcc} file. These variables should name
8624 the target-specific, language-specific object file which contains the
8625 code that uses @code{c_lex}. Note it will also be necessary to add a
8626 rule to the makefile fragment pointed to by @code{tmake_file} that shows
8627 how to build this object file.
8630 @findex HANDLE_SYSV_PRAGMA
8633 @item HANDLE_SYSV_PRAGMA
8634 Define this macro (to a value of 1) if you want the System V style
8635 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
8636 [=<value>]} to be supported by gcc.
8638 The pack pragma specifies the maximum alignment (in bytes) of fields
8639 within a structure, in much the same way as the @samp{__aligned__} and
8640 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
8641 the behavior to the default.
8643 The weak pragma only works if @code{SUPPORTS_WEAK} and
8644 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
8645 of specifically named weak labels, optionally with a value.
8647 @findex HANDLE_PRAGMA_PACK_PUSH_POP
8650 @item HANDLE_PRAGMA_PACK_PUSH_POP
8651 Define this macro (to a value of 1) if you want to support the Win32
8652 style pragmas @samp{#pragma pack(push,@var{n})} and @samp{#pragma
8653 pack(pop)}. The @samp{pack(push,@var{n})} pragma specifies the maximum alignment
8654 (in bytes) of fields within a structure, in much the same way as the
8655 @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
8656 pack value of zero resets the behavior to the default. Successive
8657 invocations of this pragma cause the previous values to be stacked, so
8658 that invocations of @samp{#pragma pack(pop)} will return to the previous
8661 @findex DOLLARS_IN_IDENTIFIERS
8662 @item DOLLARS_IN_IDENTIFIERS
8663 Define this macro to control use of the character @samp{$} in identifier
8664 names. 0 means @samp{$} is not allowed by default; 1 means it is allowed.
8665 1 is the default; there is no need to define this macro in that case.
8666 This macro controls the compiler proper; it does not affect the preprocessor.
8668 @findex NO_DOLLAR_IN_LABEL
8669 @item NO_DOLLAR_IN_LABEL
8670 Define this macro if the assembler does not accept the character
8671 @samp{$} in label names. By default constructors and destructors in
8672 G++ have @samp{$} in the identifiers. If this macro is defined,
8673 @samp{.} is used instead.
8675 @findex NO_DOT_IN_LABEL
8676 @item NO_DOT_IN_LABEL
8677 Define this macro if the assembler does not accept the character
8678 @samp{.} in label names. By default constructors and destructors in G++
8679 have names that use @samp{.}. If this macro is defined, these names
8680 are rewritten to avoid @samp{.}.
8682 @findex DEFAULT_MAIN_RETURN
8683 @item DEFAULT_MAIN_RETURN
8684 Define this macro if the target system expects every program's @code{main}
8685 function to return a standard ``success'' value by default (if no other
8686 value is explicitly returned).
8688 The definition should be a C statement (sans semicolon) to generate the
8689 appropriate rtl instructions. It is used only when compiling the end of
8694 Define this if the target system lacks the function @code{atexit}
8695 from the ISO C standard. If this macro is defined, a default definition
8696 will be provided to support C++. If @code{ON_EXIT} is not defined,
8697 a default @code{exit} function will also be provided.
8701 Define this macro if the target has another way to implement atexit
8702 functionality without replacing @code{exit}. For instance, SunOS 4 has
8703 a similar @code{on_exit} library function.
8705 The definition should be a functional macro which can be used just like
8706 the @code{atexit} function.
8710 Define this if your @code{exit} function needs to do something
8711 besides calling an external function @code{_cleanup} before
8712 terminating with @code{_exit}. The @code{EXIT_BODY} macro is
8713 only needed if @code{NEED_ATEXIT} is defined and @code{ON_EXIT} is not
8716 @findex INSN_SETS_ARE_DELAYED
8717 @item INSN_SETS_ARE_DELAYED (@var{insn})
8718 Define this macro as a C expression that is nonzero if it is safe for the
8719 delay slot scheduler to place instructions in the delay slot of @var{insn},
8720 even if they appear to use a resource set or clobbered in @var{insn}.
8721 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
8722 every @code{call_insn} has this behavior. On machines where some @code{insn}
8723 or @code{jump_insn} is really a function call and hence has this behavior,
8724 you should define this macro.
8726 You need not define this macro if it would always return zero.
8728 @findex INSN_REFERENCES_ARE_DELAYED
8729 @item INSN_REFERENCES_ARE_DELAYED (@var{insn})
8730 Define this macro as a C expression that is nonzero if it is safe for the
8731 delay slot scheduler to place instructions in the delay slot of @var{insn},
8732 even if they appear to set or clobber a resource referenced in @var{insn}.
8733 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
8734 some @code{insn} or @code{jump_insn} is really a function call and its operands
8735 are registers whose use is actually in the subroutine it calls, you should
8736 define this macro. Doing so allows the delay slot scheduler to move
8737 instructions which copy arguments into the argument registers into the delay
8740 You need not define this macro if it would always return zero.
8742 @findex MACHINE_DEPENDENT_REORG
8743 @item MACHINE_DEPENDENT_REORG (@var{insn})
8744 In rare cases, correct code generation requires extra machine
8745 dependent processing between the second jump optimization pass and
8746 delayed branch scheduling. On those machines, define this macro as a C
8747 statement to act on the code starting at @var{insn}.
8749 @findex MULTIPLE_SYMBOL_SPACES
8750 @item MULTIPLE_SYMBOL_SPACES
8751 Define this macro if in some cases global symbols from one translation
8752 unit may not be bound to undefined symbols in another translation unit
8753 without user intervention. For instance, under Microsoft Windows
8754 symbols must be explicitly imported from shared libraries (DLLs).
8756 @findex MD_ASM_CLOBBERS
8757 @item MD_ASM_CLOBBERS (@var{clobbers})
8758 A C statement that adds to @var{clobbers} @code{STRING_CST} trees for
8759 any hard regs the port wishes to automatically clobber for all asms.
8761 @findex MAX_INTEGER_COMPUTATION_MODE
8762 @item MAX_INTEGER_COMPUTATION_MODE
8763 Define this to the largest integer machine mode which can be used for
8764 operations other than load, store and copy operations.
8766 You need only define this macro if the target holds values larger than
8767 @code{word_mode} in general purpose registers. Most targets should not define
8770 @findex MATH_LIBRARY
8772 Define this macro as a C string constant for the linker argument to link
8773 in the system math library, or @samp{""} if the target does not have a
8774 separate math library.
8776 You need only define this macro if the default of @samp{"-lm"} is wrong.
8778 @findex LIBRARY_PATH_ENV
8779 @item LIBRARY_PATH_ENV
8780 Define this macro as a C string constant for the environment variable that
8781 specifies where the linker should look for libraries.
8783 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
8786 @findex TARGET_HAS_F_SETLKW
8787 @item TARGET_HAS_F_SETLKW
8788 Define this macro if the target supports file locking with fcntl / F_SETLKW@.
8789 Note that this functionality is part of POSIX@.
8790 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
8791 to use file locking when exiting a program, which avoids race conditions
8792 if the program has forked.
8794 @findex MAX_CONDITIONAL_EXECUTE
8795 @item MAX_CONDITIONAL_EXECUTE
8797 A C expression for the maximum number of instructions to execute via
8798 conditional execution instructions instead of a branch. A value of
8799 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
8800 1 if it does use cc0.
8802 @findex IFCVT_MODIFY_TESTS
8803 @item IFCVT_MODIFY_TESTS
8804 A C expression to modify the tests in @code{TRUE_EXPR}, and
8805 @code{FALSE_EXPR} for use in converting insns in @code{TEST_BB},
8806 @code{THEN_BB}, @code{ELSE_BB}, and @code{JOIN_BB} basic blocks to
8807 conditional execution. Set either @code{TRUE_EXPR} or @code{FALSE_EXPR}
8808 to a null pointer if the tests cannot be converted.
8810 @findex IFCVT_MODIFY_INSN
8811 @item IFCVT_MODIFY_INSN
8812 A C expression to modify the @code{PATTERN} of an @code{INSN} that is to
8813 be converted to conditional execution format.
8815 @findex IFCVT_MODIFY_FINAL
8816 @item IFCVT_MODIFY_FINAL
8817 A C expression to perform any final machine dependent modifications in
8818 converting code to conditional execution in the basic blocks
8819 @code{TEST_BB}, @code{THEN_BB}, @code{ELSE_BB}, and @code{JOIN_BB}.
8821 @findex IFCVT_MODIFY_CANCEL
8822 @item IFCVT_MODIFY_CANCEL
8823 A C expression to cancel any machine dependent modifications in
8824 converting code to conditional execution in the basic blocks
8825 @code{TEST_BB}, @code{THEN_BB}, @code{ELSE_BB}, and @code{JOIN_BB}.
8828 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
8829 Define this hook if you have any machine-specific built-in functions
8830 that need to be defined. It should be a function that performs the
8833 Machine specific built-in functions can be useful to expand special machine
8834 instructions that would otherwise not normally be generated because
8835 they have no equivalent in the source language (for example, SIMD vector
8836 instructions or prefetch instructions).
8838 To create a built-in function, call the function @code{builtin_function}
8839 which is defined by the language front end. You can use any type nodes set
8840 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
8841 only language front ends that use those two functions will call
8842 @samp{TARGET_INIT_BUILTINS}.
8845 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
8847 Expand a call to a machine specific built-in function that was set up by
8848 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
8849 function call; the result should go to @var{target} if that is
8850 convenient, and have mode @var{mode} if that is convenient.
8851 @var{subtarget} may be used as the target for computing one of
8852 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
8853 ignored. This function should return the result of the call to the
8858 @findex MD_CAN_REDIRECT_BRANCH
8859 @item MD_CAN_REDIRECT_BRANCH(@var{branch1}, @var{branch2})
8861 Take a branch insn in @var{branch1} and another in @var{branch2}.
8862 Return true if redirecting @var{branch1} to the destination of
8863 @var{branch2} is possible.
8865 On some targets, branches may have a limited range. Optimizing the
8866 filling of delay slots can result in branches being redirected, and this
8867 may in turn cause a branch offset to overflow.
8869 @findex ALLOCATE_INITIAL_VALUE
8870 @item ALLOCATE_INITIAL_VALUE(@var{hard_reg})
8872 When the initial value of a hard register has been copied in a pseudo
8873 register, it is often not necessary to actually allocate another register
8874 to this pseudo register, because the original hard register or a stack slot
8875 it has been saved into can be used. @code{ALLOCATE_INITIAL_VALUE}, if
8876 defined, is called at the start of register allocation once for each
8877 hard register that had its initial value copied by using
8878 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
8879 Possible values are @code{NULL_RTX}, if you don't want
8880 to do any special allocation, a @code{REG} rtx---that would typically be
8881 the hard register itself, if it is known not to be clobbered---or a
8883 If you are returning a @code{MEM}, this is only a hint for the allocator;
8884 it might decide to use another register anyways.
8885 You may use @code{current_function_leaf_function} in the definition of the
8886 macro, functions that use @code{REG_N_SETS}, to determine if the hard
8887 register in question will not be clobbered.
8889 @findex TARGET_OBJECT_SUFFIX
8890 @item TARGET_OBJECT_SUFFIX
8891 Define this macro to be a C string representing the suffix for object
8892 files on your target machine. If you do not define this macro, GCC will
8893 use @samp{.o} as the suffix for object files.
8895 @findex TARGET_EXECUTABLE_SUFFIX
8896 @item TARGET_EXECUTABLE_SUFFIX
8897 Define this macro to be a C string representing the suffix to be
8898 automatically added to executable files on your target machine. If you
8899 do not define this macro, GCC will use the null string as the suffix for
8902 @findex COLLECT_EXPORT_LIST
8903 @item COLLECT_EXPORT_LIST
8904 If defined, @code{collect2} will scan the individual object files
8905 specified on its command line and create an export list for the linker.
8906 Define this macro for systems like AIX, where the linker discards
8907 object files that are not referenced from @code{main} and uses export
8912 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
8913 This target hook returns @code{true} past the point in which new jump
8914 instructions could be created. On machines that require a register for
8915 every jump such as the SHmedia ISA of SH5, this point would typically be
8916 reload, so this target hook should be defined to a function such as:
8920 cannot_modify_jumps_past_reload_p ()
8922 return (reload_completed || reload_in_progress);