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" @}
166 @findex DRIVER_SELF_SPECS
167 @item DRIVER_SELF_SPECS
168 A list of specs for the driver itself. It should be a suitable
169 initializer for an array of strings, with no surrounding braces.
171 The driver applies these specs to its own command line before choosing
172 the multilib directory or running any subcommands. It applies them in
173 the order given, so each spec can depend on the options added by
174 earlier ones. It is also possible to remove options using
175 @samp{%<@var{option}} in the usual way.
177 This macro can be useful when a port has several interdependent target
178 options. It provides a way of standardizing the command line so
179 that the other specs are easier to write.
181 Do not define this macro if it does not need to do anything.
185 A C string constant that tells the GCC driver program options to
186 pass to CPP@. It can also specify how to translate options you
187 give to GCC into options for GCC to pass to the CPP@.
189 Do not define this macro if it does not need to do anything.
191 @findex CPLUSPLUS_CPP_SPEC
192 @item CPLUSPLUS_CPP_SPEC
193 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
194 than C@. If you do not define this macro, then the value of
195 @code{CPP_SPEC} (if any) will be used instead.
199 A C string constant that tells the GCC driver program options to
200 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
202 It can also specify how to translate options you give to GCC into options
203 for GCC to pass to front ends.
205 Do not define this macro if it does not need to do anything.
209 A C string constant that tells the GCC driver program options to
210 pass to @code{cc1plus}. It can also specify how to translate options you
211 give to GCC into options for GCC to pass to the @code{cc1plus}.
213 Do not define this macro if it does not need to do anything.
214 Note that everything defined in CC1_SPEC is already passed to
215 @code{cc1plus} so there is no need to duplicate the contents of
216 CC1_SPEC in CC1PLUS_SPEC@.
220 A C string constant that tells the GCC driver program options to
221 pass to the assembler. It can also specify how to translate options
222 you give to GCC into options for GCC to pass to the assembler.
223 See the file @file{sun3.h} for an example of this.
225 Do not define this macro if it does not need to do anything.
227 @findex ASM_FINAL_SPEC
229 A C string constant that tells the GCC driver program how to
230 run any programs which cleanup after the normal assembler.
231 Normally, this is not needed. See the file @file{mips.h} for
234 Do not define this macro if it does not need to do anything.
238 A C string constant that tells the GCC driver program options to
239 pass to the linker. It can also specify how to translate options you
240 give to GCC into options for GCC to pass to the linker.
242 Do not define this macro if it does not need to do anything.
246 Another C string constant used much like @code{LINK_SPEC}. The difference
247 between the two is that @code{LIB_SPEC} is used at the end of the
248 command given to the linker.
250 If this macro is not defined, a default is provided that
251 loads the standard C library from the usual place. See @file{gcc.c}.
255 Another C string constant that tells the GCC driver program
256 how and when to place a reference to @file{libgcc.a} into the
257 linker command line. This constant is placed both before and after
258 the value of @code{LIB_SPEC}.
260 If this macro is not defined, the GCC driver provides a default that
261 passes the string @option{-lgcc} to the linker.
263 @findex STARTFILE_SPEC
265 Another C string constant used much like @code{LINK_SPEC}. The
266 difference between the two is that @code{STARTFILE_SPEC} is used at
267 the very beginning of the command given to the linker.
269 If this macro is not defined, a default is provided that loads the
270 standard C startup file from the usual place. See @file{gcc.c}.
274 Another C string constant used much like @code{LINK_SPEC}. The
275 difference between the two is that @code{ENDFILE_SPEC} is used at
276 the very end of the command given to the linker.
278 Do not define this macro if it does not need to do anything.
280 @findex THREAD_MODEL_SPEC
281 @item THREAD_MODEL_SPEC
282 GCC @code{-v} will print the thread model GCC was configured to use.
283 However, this doesn't work on platforms that are multilibbed on thread
284 models, such as AIX 4.3. On such platforms, define
285 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
286 blanks that names one of the recognized thread models. @code{%*}, the
287 default value of this macro, will expand to the value of
288 @code{thread_file} set in @file{config.gcc}.
292 Define this macro to provide additional specifications to put in the
293 @file{specs} file that can be used in various specifications like
296 The definition should be an initializer for an array of structures,
297 containing a string constant, that defines the specification name, and a
298 string constant that provides the specification.
300 Do not define this macro if it does not need to do anything.
302 @code{EXTRA_SPECS} is useful when an architecture contains several
303 related targets, which have various @code{@dots{}_SPECS} which are similar
304 to each other, and the maintainer would like one central place to keep
307 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
308 define either @code{_CALL_SYSV} when the System V calling sequence is
309 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
312 The @file{config/rs6000/rs6000.h} target file defines:
315 #define EXTRA_SPECS \
316 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
318 #define CPP_SYS_DEFAULT ""
321 The @file{config/rs6000/sysv.h} target file defines:
325 "%@{posix: -D_POSIX_SOURCE @} \
326 %@{mcall-sysv: -D_CALL_SYSV @} %@{mcall-aix: -D_CALL_AIX @} \
327 %@{!mcall-sysv: %@{!mcall-aix: %(cpp_sysv_default) @}@} \
328 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
330 #undef CPP_SYSV_DEFAULT
331 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
334 while the @file{config/rs6000/eabiaix.h} target file defines
335 @code{CPP_SYSV_DEFAULT} as:
338 #undef CPP_SYSV_DEFAULT
339 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
342 @findex LINK_LIBGCC_SPECIAL
343 @item LINK_LIBGCC_SPECIAL
344 Define this macro if the driver program should find the library
345 @file{libgcc.a} itself and should not pass @option{-L} options to the
346 linker. If you do not define this macro, the driver program will pass
347 the argument @option{-lgcc} to tell the linker to do the search and will
348 pass @option{-L} options to it.
350 @findex LINK_LIBGCC_SPECIAL_1
351 @item LINK_LIBGCC_SPECIAL_1
352 Define this macro if the driver program should find the library
353 @file{libgcc.a}. If you do not define this macro, the driver program will pass
354 the argument @option{-lgcc} to tell the linker to do the search.
355 This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does
356 not affect @option{-L} options.
358 @findex LINK_GCC_C_SEQUENCE_SPEC
359 @item LINK_GCC_C_SEQUENCE_SPEC
360 The sequence in which libgcc and libc are specified to the linker.
361 By default this is @code{%G %L %G}.
363 @findex LINK_COMMAND_SPEC
364 @item LINK_COMMAND_SPEC
365 A C string constant giving the complete command line need to execute the
366 linker. When you do this, you will need to update your port each time a
367 change is made to the link command line within @file{gcc.c}. Therefore,
368 define this macro only if you need to completely redefine the command
369 line for invoking the linker and there is no other way to accomplish
370 the effect you need. Overriding this macro may be avoidable by overriding
371 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
373 @findex LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
374 @item LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
375 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
376 directories from linking commands. Do not give it a nonzero value if
377 removing duplicate search directories changes the linker's semantics.
379 @findex MULTILIB_DEFAULTS
380 @item MULTILIB_DEFAULTS
381 Define this macro as a C expression for the initializer of an array of
382 string to tell the driver program which options are defaults for this
383 target and thus do not need to be handled specially when using
384 @code{MULTILIB_OPTIONS}.
386 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
387 the target makefile fragment or if none of the options listed in
388 @code{MULTILIB_OPTIONS} are set by default.
389 @xref{Target Fragment}.
391 @findex RELATIVE_PREFIX_NOT_LINKDIR
392 @item RELATIVE_PREFIX_NOT_LINKDIR
393 Define this macro to tell @code{gcc} that it should only translate
394 a @option{-B} prefix into a @option{-L} linker option if the prefix
395 indicates an absolute file name.
397 @findex STANDARD_EXEC_PREFIX
398 @item STANDARD_EXEC_PREFIX
399 Define this macro as a C string constant if you wish to override the
400 standard choice of @file{/usr/local/lib/gcc-lib/} as the default prefix to
401 try when searching for the executable files of the compiler.
403 @findex MD_EXEC_PREFIX
405 If defined, this macro is an additional prefix to try after
406 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
407 when the @option{-b} option is used, or the compiler is built as a cross
408 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
409 to the list of directories used to find the assembler in @file{configure.in}.
411 @findex STANDARD_STARTFILE_PREFIX
412 @item STANDARD_STARTFILE_PREFIX
413 Define this macro as a C string constant if you wish to override the
414 standard choice of @file{/usr/local/lib/} as the default prefix to
415 try when searching for startup files such as @file{crt0.o}.
417 @findex MD_STARTFILE_PREFIX
418 @item MD_STARTFILE_PREFIX
419 If defined, this macro supplies an additional prefix to try after the
420 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
421 @option{-b} option is used, or when the compiler is built as a cross
424 @findex MD_STARTFILE_PREFIX_1
425 @item MD_STARTFILE_PREFIX_1
426 If defined, this macro supplies yet another prefix to try after the
427 standard prefixes. It is not searched when the @option{-b} option is
428 used, or when the compiler is built as a cross compiler.
430 @findex INIT_ENVIRONMENT
431 @item INIT_ENVIRONMENT
432 Define this macro as a C string constant if you wish to set environment
433 variables for programs called by the driver, such as the assembler and
434 loader. The driver passes the value of this macro to @code{putenv} to
435 initialize the necessary environment variables.
437 @findex LOCAL_INCLUDE_DIR
438 @item LOCAL_INCLUDE_DIR
439 Define this macro as a C string constant if you wish to override the
440 standard choice of @file{/usr/local/include} as the default prefix to
441 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
442 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
444 Cross compilers do not search either @file{/usr/local/include} or its
447 @findex MODIFY_TARGET_NAME
448 @item MODIFY_TARGET_NAME
449 Define this macro if you with to define command-line switches that modify the
452 For each switch, you can include a string to be appended to the first
453 part of the configuration name or a string to be deleted from the
454 configuration name, if present. The definition should be an initializer
455 for an array of structures. Each array element should have three
456 elements: the switch name (a string constant, including the initial
457 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
458 indicate whether the string should be inserted or deleted, and the string
459 to be inserted or deleted (a string constant).
461 For example, on a machine where @samp{64} at the end of the
462 configuration name denotes a 64-bit target and you want the @option{-32}
463 and @option{-64} switches to select between 32- and 64-bit targets, you would
467 #define MODIFY_TARGET_NAME \
468 @{ @{ "-32", DELETE, "64"@}, \
469 @{"-64", ADD, "64"@}@}
473 @findex SYSTEM_INCLUDE_DIR
474 @item SYSTEM_INCLUDE_DIR
475 Define this macro as a C string constant if you wish to specify a
476 system-specific directory to search for header files before the standard
477 directory. @code{SYSTEM_INCLUDE_DIR} comes before
478 @code{STANDARD_INCLUDE_DIR} in the search order.
480 Cross compilers do not use this macro and do not search the directory
483 @findex STANDARD_INCLUDE_DIR
484 @item STANDARD_INCLUDE_DIR
485 Define this macro as a C string constant if you wish to override the
486 standard choice of @file{/usr/include} as the default prefix to
487 try when searching for header files.
489 Cross compilers do not use this macro and do not search either
490 @file{/usr/include} or its replacement.
492 @findex STANDARD_INCLUDE_COMPONENT
493 @item STANDARD_INCLUDE_COMPONENT
494 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
495 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
496 If you do not define this macro, no component is used.
498 @findex INCLUDE_DEFAULTS
499 @item INCLUDE_DEFAULTS
500 Define this macro if you wish to override the entire default search path
501 for include files. For a native compiler, the default search path
502 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
503 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
504 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
505 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
506 and specify private search areas for GCC@. The directory
507 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
509 The definition should be an initializer for an array of structures.
510 Each array element should have four elements: the directory name (a
511 string constant), the component name (also a string constant), a flag
512 for C++-only directories,
513 and a flag showing that the includes in the directory don't need to be
514 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
515 the array with a null element.
517 The component name denotes what GNU package the include file is part of,
518 if any, in all upper-case letters. For example, it might be @samp{GCC}
519 or @samp{BINUTILS}. If the package is part of a vendor-supplied
520 operating system, code the component name as @samp{0}.
522 For example, here is the definition used for VAX/VMS:
525 #define INCLUDE_DEFAULTS \
527 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
528 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
529 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
536 Here is the order of prefixes tried for exec files:
540 Any prefixes specified by the user with @option{-B}.
543 The environment variable @code{GCC_EXEC_PREFIX}, if any.
546 The directories specified by the environment variable @code{COMPILER_PATH}.
549 The macro @code{STANDARD_EXEC_PREFIX}.
552 @file{/usr/lib/gcc/}.
555 The macro @code{MD_EXEC_PREFIX}, if any.
558 Here is the order of prefixes tried for startfiles:
562 Any prefixes specified by the user with @option{-B}.
565 The environment variable @code{GCC_EXEC_PREFIX}, if any.
568 The directories specified by the environment variable @code{LIBRARY_PATH}
569 (or port-specific name; native only, cross compilers do not use this).
572 The macro @code{STANDARD_EXEC_PREFIX}.
575 @file{/usr/lib/gcc/}.
578 The macro @code{MD_EXEC_PREFIX}, if any.
581 The macro @code{MD_STARTFILE_PREFIX}, if any.
584 The macro @code{STANDARD_STARTFILE_PREFIX}.
593 @node Run-time Target
594 @section Run-time Target Specification
595 @cindex run-time target specification
596 @cindex predefined macros
597 @cindex target specifications
599 @c prevent bad page break with this line
600 Here are run-time target specifications.
603 @findex TARGET_CPU_CPP_BUILTINS
604 @item TARGET_CPU_CPP_BUILTINS()
605 This function-like macro expands to a block of code that defines
606 built-in preprocessor macros and assertions for the target cpu, using
607 the functions @code{builtin_define}, @code{builtin_define_std} and
608 @code{builtin_assert} defined in @file{c-common.c}. When the front end
609 calls this macro it provides a trailing semicolon, and since it has
610 finished command line option processing your code can use those
613 @code{builtin_assert} takes a string in the form you pass to the
614 command-line option @option{-A}, such as @code{cpu=mips}, and creates
615 the assertion. @code{builtin_define} takes a string in the form
616 accepted by option @option{-D} and unconditionally defines the macro.
618 @code{builtin_define_std} takes a string representing the name of an
619 object-like macro. If it doesn't lie in the user's namespace,
620 @code{builtin_define_std} defines it unconditionally. Otherwise, it
621 defines a version with two leading underscores, and another version
622 with two leading and trailing underscores, and defines the original
623 only if an ISO standard was not requested on the command line. For
624 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
625 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
626 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
627 defines only @code{_ABI64}.
629 You can also test for the C dialect being compiled. The variable
630 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
631 or @code{clk_objective_c}. Note that if we are preprocessing
632 assembler, this variable will be @code{clk_c} but the function-like
633 macro @code{preprocessing_asm_p()} will return true, so you might want
634 to check for that first. If you need to check for strict ANSI, the
635 variable @code{flag_iso} can be used. The function-like macro
636 @code{preprocessing_trad_p()} can be used to check for traditional
639 With @code{TARGET_OS_CPP_BUILTINS} this macro obsoletes the
640 @code{CPP_PREDEFINES} target macro.
642 @findex TARGET_OS_CPP_BUILTINS
643 @item TARGET_OS_CPP_BUILTINS()
644 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
645 and is used for the target operating system instead.
647 With @code{TARGET_CPU_CPP_BUILTINS} this macro obsoletes the
648 @code{CPP_PREDEFINES} target macro.
650 @findex CPP_PREDEFINES
652 Define this to be a string constant containing @option{-D} options to
653 define the predefined macros that identify this machine and system.
654 These macros will be predefined unless the @option{-ansi} option (or a
655 @option{-std} option for strict ISO C conformance) is specified.
657 In addition, a parallel set of macros are predefined, whose names are
658 made by appending @samp{__} at the beginning and at the end. These
659 @samp{__} macros are permitted by the ISO standard, so they are
660 predefined regardless of whether @option{-ansi} or a @option{-std} option
663 For example, on the Sun, one can use the following value:
666 "-Dmc68000 -Dsun -Dunix"
669 The result is to define the macros @code{__mc68000__}, @code{__sun__}
670 and @code{__unix__} unconditionally, and the macros @code{mc68000},
671 @code{sun} and @code{unix} provided @option{-ansi} is not specified.
673 @findex extern int target_flags
674 @item extern int target_flags;
675 This declaration should be present.
677 @cindex optional hardware or system features
678 @cindex features, optional, in system conventions
680 This series of macros is to allow compiler command arguments to
681 enable or disable the use of optional features of the target machine.
682 For example, one machine description serves both the 68000 and
683 the 68020; a command argument tells the compiler whether it should
684 use 68020-only instructions or not. This command argument works
685 by means of a macro @code{TARGET_68020} that tests a bit in
688 Define a macro @code{TARGET_@var{featurename}} for each such option.
689 Its definition should test a bit in @code{target_flags}. It is
690 recommended that a helper macro @code{TARGET_MASK_@var{featurename}}
691 is defined for each bit-value to test, and used in
692 @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For
696 #define TARGET_MASK_68020 1
697 #define TARGET_68020 (target_flags & TARGET_MASK_68020)
700 One place where these macros are used is in the condition-expressions
701 of instruction patterns. Note how @code{TARGET_68020} appears
702 frequently in the 68000 machine description file, @file{m68k.md}.
703 Another place they are used is in the definitions of the other
704 macros in the @file{@var{machine}.h} file.
706 @findex TARGET_SWITCHES
707 @item TARGET_SWITCHES
708 This macro defines names of command options to set and clear
709 bits in @code{target_flags}. Its definition is an initializer
710 with a subgrouping for each command option.
712 Each subgrouping contains a string constant, that defines the option
713 name, a number, which contains the bits to set in
714 @code{target_flags}, and a second string which is the description
715 displayed by @option{--help}. If the number is negative then the bits specified
716 by the number are cleared instead of being set. If the description
717 string is present but empty, then no help information will be displayed
718 for that option, but it will not count as an undocumented option. The
719 actual option name is made by appending @samp{-m} to the specified name.
720 Non-empty description strings should be marked with @code{N_(@dots{})} for
721 @command{xgettext}. Please do not mark empty strings because the empty
722 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
723 of the message catalog with meta information, not the empty string.
725 In addition to the description for @option{--help},
726 more detailed documentation for each option should be added to
729 One of the subgroupings should have a null string. The number in
730 this grouping is the default value for @code{target_flags}. Any
731 target options act starting with that value.
733 Here is an example which defines @option{-m68000} and @option{-m68020}
734 with opposite meanings, and picks the latter as the default:
737 #define TARGET_SWITCHES \
738 @{ @{ "68020", TARGET_MASK_68020, "" @}, \
739 @{ "68000", -TARGET_MASK_68020, \
740 N_("Compile for the 68000") @}, \
741 @{ "", TARGET_MASK_68020, "" @}@}
744 @findex TARGET_OPTIONS
746 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
747 options that have values. Its definition is an initializer with a
748 subgrouping for each command option.
750 Each subgrouping contains a string constant, that defines the fixed part
751 of the option name, the address of a variable, and a description string.
752 Non-empty description strings should be marked with @code{N_(@dots{})} for
753 @command{xgettext}. Please do not mark empty strings because the empty
754 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
755 of the message catalog with meta information, not the empty string.
757 The variable, type @code{char *}, is set to the variable part of the
758 given option if the fixed part matches. The actual option name is made
759 by appending @samp{-m} to the specified name. Again, each option should
760 also be documented in @file{invoke.texi}.
762 Here is an example which defines @option{-mshort-data-@var{number}}. If the
763 given option is @option{-mshort-data-512}, the variable @code{m88k_short_data}
764 will be set to the string @code{"512"}.
767 extern char *m88k_short_data;
768 #define TARGET_OPTIONS \
769 @{ @{ "short-data-", &m88k_short_data, \
770 N_("Specify the size of the short data section") @} @}
773 @findex TARGET_VERSION
775 This macro is a C statement to print on @code{stderr} a string
776 describing the particular machine description choice. Every machine
777 description should define @code{TARGET_VERSION}. For example:
781 #define TARGET_VERSION \
782 fprintf (stderr, " (68k, Motorola syntax)");
784 #define TARGET_VERSION \
785 fprintf (stderr, " (68k, MIT syntax)");
789 @findex OVERRIDE_OPTIONS
790 @item OVERRIDE_OPTIONS
791 Sometimes certain combinations of command options do not make sense on
792 a particular target machine. You can define a macro
793 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
794 defined, is executed once just after all the command options have been
797 Don't use this macro to turn on various extra optimizations for
798 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
800 @findex OPTIMIZATION_OPTIONS
801 @item OPTIMIZATION_OPTIONS (@var{level}, @var{size})
802 Some machines may desire to change what optimizations are performed for
803 various optimization levels. This macro, if defined, is executed once
804 just after the optimization level is determined and before the remainder
805 of the command options have been parsed. Values set in this macro are
806 used as the default values for the other command line options.
808 @var{level} is the optimization level specified; 2 if @option{-O2} is
809 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
811 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
813 You should not use this macro to change options that are not
814 machine-specific. These should uniformly selected by the same
815 optimization level on all supported machines. Use this macro to enable
816 machine-specific optimizations.
818 @strong{Do not examine @code{write_symbols} in
819 this macro!} The debugging options are not supposed to alter the
822 @findex CAN_DEBUG_WITHOUT_FP
823 @item CAN_DEBUG_WITHOUT_FP
824 Define this macro if debugging can be performed even without a frame
825 pointer. If this macro is defined, GCC will turn on the
826 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
829 @node Per-Function Data
830 @section Defining data structures for per-function information.
831 @cindex per-function data
832 @cindex data structures
834 If the target needs to store information on a per-function basis, GCC
835 provides a macro and a couple of variables to allow this. Note, just
836 using statics to store the information is a bad idea, since GCC supports
837 nested functions, so you can be halfway through encoding one function
838 when another one comes along.
840 GCC defines a data structure called @code{struct function} which
841 contains all of the data specific to an individual function. This
842 structure contains a field called @code{machine} whose type is
843 @code{struct machine_function *}, which can be used by targets to point
844 to their own specific data.
846 If a target needs per-function specific data it should define the type
847 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
848 This macro should be used to initialize the function pointer
849 @code{init_machine_status}. This pointer is explained below.
851 One typical use of per-function, target specific data is to create an
852 RTX to hold the register containing the function's return address. This
853 RTX can then be used to implement the @code{__builtin_return_address}
854 function, for level 0.
856 Note---earlier implementations of GCC used a single data area to hold
857 all of the per-function information. Thus when processing of a nested
858 function began the old per-function data had to be pushed onto a
859 stack, and when the processing was finished, it had to be popped off the
860 stack. GCC used to provide function pointers called
861 @code{save_machine_status} and @code{restore_machine_status} to handle
862 the saving and restoring of the target specific information. Since the
863 single data area approach is no longer used, these pointers are no
866 The macro and function pointers are described below.
869 @findex INIT_EXPANDERS
871 Macro called to initialize any target specific information. This macro
872 is called once per function, before generation of any RTL has begun.
873 The intention of this macro is to allow the initialization of the
874 function pointers below.
876 @findex init_machine_status
877 @item init_machine_status
878 This is a @code{void (*)(struct function *)} function pointer. If this
879 pointer is non-@code{NULL} it will be called once per function, before function
880 compilation starts, in order to allow the target to perform any target
881 specific initialization of the @code{struct function} structure. It is
882 intended that this would be used to initialize the @code{machine} of
885 @code{struct machine_function} structures are expected to be freed by GC.
886 Generally, any memory that they reference must be allocated by using
887 @code{ggc_alloc}, including the structure itself.
892 @section Storage Layout
893 @cindex storage layout
895 Note that the definitions of the macros in this table which are sizes or
896 alignments measured in bits do not need to be constant. They can be C
897 expressions that refer to static variables, such as the @code{target_flags}.
898 @xref{Run-time Target}.
901 @findex BITS_BIG_ENDIAN
902 @item BITS_BIG_ENDIAN
903 Define this macro to have the value 1 if the most significant bit in a
904 byte has the lowest number; otherwise define it to have the value zero.
905 This means that bit-field instructions count from the most significant
906 bit. If the machine has no bit-field instructions, then this must still
907 be defined, but it doesn't matter which value it is defined to. This
908 macro need not be a constant.
910 This macro does not affect the way structure fields are packed into
911 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
913 @findex BYTES_BIG_ENDIAN
914 @item BYTES_BIG_ENDIAN
915 Define this macro to have the value 1 if the most significant byte in a
916 word has the lowest number. This macro need not be a constant.
918 @findex WORDS_BIG_ENDIAN
919 @item WORDS_BIG_ENDIAN
920 Define this macro to have the value 1 if, in a multiword object, the
921 most significant word has the lowest number. This applies to both
922 memory locations and registers; GCC fundamentally assumes that the
923 order of words in memory is the same as the order in registers. This
924 macro need not be a constant.
926 @findex LIBGCC2_WORDS_BIG_ENDIAN
927 @item LIBGCC2_WORDS_BIG_ENDIAN
928 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
929 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
930 used only when compiling @file{libgcc2.c}. Typically the value will be set
931 based on preprocessor defines.
933 @findex FLOAT_WORDS_BIG_ENDIAN
934 @item FLOAT_WORDS_BIG_ENDIAN
935 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
936 @code{TFmode} floating point numbers are stored in memory with the word
937 containing the sign bit at the lowest address; otherwise define it to
938 have the value 0. This macro need not be a constant.
940 You need not define this macro if the ordering is the same as for
943 @findex BITS_PER_UNIT
945 Define this macro to be the number of bits in an addressable storage
946 unit (byte). If you do not define this macro the default is 8.
948 @findex BITS_PER_WORD
950 Number of bits in a word. If you do not define this macro, the default
951 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
953 @findex MAX_BITS_PER_WORD
954 @item MAX_BITS_PER_WORD
955 Maximum number of bits in a word. If this is undefined, the default is
956 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
957 largest value that @code{BITS_PER_WORD} can have at run-time.
959 @findex UNITS_PER_WORD
961 Number of storage units in a word; normally 4.
963 @findex MIN_UNITS_PER_WORD
964 @item MIN_UNITS_PER_WORD
965 Minimum number of units in a word. If this is undefined, the default is
966 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
967 smallest value that @code{UNITS_PER_WORD} can have at run-time.
971 Width of a pointer, in bits. You must specify a value no wider than the
972 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
973 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
974 a value the default is @code{BITS_PER_WORD}.
976 @findex POINTERS_EXTEND_UNSIGNED
977 @item POINTERS_EXTEND_UNSIGNED
978 A C expression whose value is greater than zero if pointers that need to be
979 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
980 be zero-extended and zero if they are to be sign-extended. If the value
981 is less then zero then there must be an "ptr_extend" instruction that
982 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
984 You need not define this macro if the @code{POINTER_SIZE} is equal
985 to the width of @code{Pmode}.
988 @item PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
989 A macro to update @var{m} and @var{unsignedp} when an object whose type
990 is @var{type} and which has the specified mode and signedness is to be
991 stored in a register. This macro is only called when @var{type} is a
994 On most RISC machines, which only have operations that operate on a full
995 register, define this macro to set @var{m} to @code{word_mode} if
996 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
997 cases, only integer modes should be widened because wider-precision
998 floating-point operations are usually more expensive than their narrower
1001 For most machines, the macro definition does not change @var{unsignedp}.
1002 However, some machines, have instructions that preferentially handle
1003 either signed or unsigned quantities of certain modes. For example, on
1004 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
1005 sign-extend the result to 64 bits. On such machines, set
1006 @var{unsignedp} according to which kind of extension is more efficient.
1008 Do not define this macro if it would never modify @var{m}.
1010 @findex PROMOTE_FUNCTION_ARGS
1011 @item PROMOTE_FUNCTION_ARGS
1012 Define this macro if the promotion described by @code{PROMOTE_MODE}
1013 should also be done for outgoing function arguments.
1015 @findex PROMOTE_FUNCTION_RETURN
1016 @item PROMOTE_FUNCTION_RETURN
1017 Define this macro if the promotion described by @code{PROMOTE_MODE}
1018 should also be done for the return value of functions.
1020 If this macro is defined, @code{FUNCTION_VALUE} must perform the same
1021 promotions done by @code{PROMOTE_MODE}.
1023 @findex PROMOTE_FOR_CALL_ONLY
1024 @item PROMOTE_FOR_CALL_ONLY
1025 Define this macro if the promotion described by @code{PROMOTE_MODE}
1026 should @emph{only} be performed for outgoing function arguments or
1027 function return values, as specified by @code{PROMOTE_FUNCTION_ARGS}
1028 and @code{PROMOTE_FUNCTION_RETURN}, respectively.
1030 @findex PARM_BOUNDARY
1032 Normal alignment required for function parameters on the stack, in
1033 bits. All stack parameters receive at least this much alignment
1034 regardless of data type. On most machines, this is the same as the
1037 @findex STACK_BOUNDARY
1038 @item STACK_BOUNDARY
1039 Define this macro to the minimum alignment enforced by hardware for the
1040 stack pointer on this machine. The definition is a C expression for the
1041 desired alignment (measured in bits). This value is used as a default
1042 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1043 this should be the same as @code{PARM_BOUNDARY}.
1045 @findex PREFERRED_STACK_BOUNDARY
1046 @item PREFERRED_STACK_BOUNDARY
1047 Define this macro if you wish to preserve a certain alignment for the
1048 stack pointer, greater than what the hardware enforces. The definition
1049 is a C expression for the desired alignment (measured in bits). This
1050 macro must evaluate to a value equal to or larger than
1051 @code{STACK_BOUNDARY}.
1053 @findex FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1054 @item FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1055 A C expression that evaluates true if @code{PREFERRED_STACK_BOUNDARY} is
1056 not guaranteed by the runtime and we should emit code to align the stack
1057 at the beginning of @code{main}.
1059 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
1060 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
1061 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
1062 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
1063 be momentarily unaligned while pushing arguments.
1065 @findex FUNCTION_BOUNDARY
1066 @item FUNCTION_BOUNDARY
1067 Alignment required for a function entry point, in bits.
1069 @findex BIGGEST_ALIGNMENT
1070 @item BIGGEST_ALIGNMENT
1071 Biggest alignment that any data type can require on this machine, in bits.
1073 @findex MINIMUM_ATOMIC_ALIGNMENT
1074 @item MINIMUM_ATOMIC_ALIGNMENT
1075 If defined, the smallest alignment, in bits, that can be given to an
1076 object that can be referenced in one operation, without disturbing any
1077 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1078 on machines that don't have byte or half-word store operations.
1080 @findex BIGGEST_FIELD_ALIGNMENT
1081 @item BIGGEST_FIELD_ALIGNMENT
1082 Biggest alignment that any structure or union field can require on this
1083 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1084 structure and union fields only, unless the field alignment has been set
1085 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1087 @findex ADJUST_FIELD_ALIGN
1088 @item ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1089 An expression for the alignment of a structure field @var{field} if the
1090 alignment computed in the usual way (including applying of
1091 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1092 alignment) is @var{computed}. It overrides alignment only if the
1093 field alignment has not been set by the
1094 @code{__attribute__ ((aligned (@var{n})))} construct.
1096 @findex MAX_OFILE_ALIGNMENT
1097 @item MAX_OFILE_ALIGNMENT
1098 Biggest alignment supported by the object file format of this machine.
1099 Use this macro to limit the alignment which can be specified using the
1100 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1101 the default value is @code{BIGGEST_ALIGNMENT}.
1103 @findex DATA_ALIGNMENT
1104 @item DATA_ALIGNMENT (@var{type}, @var{basic-align})
1105 If defined, a C expression to compute the alignment for a variable in
1106 the static store. @var{type} is the data type, and @var{basic-align} is
1107 the alignment that the object would ordinarily have. The value of this
1108 macro is used instead of that alignment to align the object.
1110 If this macro is not defined, then @var{basic-align} is used.
1113 One use of this macro is to increase alignment of medium-size data to
1114 make it all fit in fewer cache lines. Another is to cause character
1115 arrays to be word-aligned so that @code{strcpy} calls that copy
1116 constants to character arrays can be done inline.
1118 @findex CONSTANT_ALIGNMENT
1119 @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1120 If defined, a C expression to compute the alignment given to a constant
1121 that is being placed in memory. @var{constant} is the constant and
1122 @var{basic-align} is the alignment that the object would ordinarily
1123 have. The value of this macro is used instead of that alignment to
1126 If this macro is not defined, then @var{basic-align} is used.
1128 The typical use of this macro is to increase alignment for string
1129 constants to be word aligned so that @code{strcpy} calls that copy
1130 constants can be done inline.
1132 @findex LOCAL_ALIGNMENT
1133 @item LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1134 If defined, a C expression to compute the alignment for a variable in
1135 the local store. @var{type} is the data type, and @var{basic-align} is
1136 the alignment that the object would ordinarily have. The value of this
1137 macro is used instead of that alignment to align the object.
1139 If this macro is not defined, then @var{basic-align} is used.
1141 One use of this macro is to increase alignment of medium-size data to
1142 make it all fit in fewer cache lines.
1144 @findex EMPTY_FIELD_BOUNDARY
1145 @item EMPTY_FIELD_BOUNDARY
1146 Alignment in bits to be given to a structure bit-field that follows an
1147 empty field such as @code{int : 0;}.
1149 Note that @code{PCC_BITFIELD_TYPE_MATTERS} also affects the alignment
1150 that results from an empty field.
1152 @findex STRUCTURE_SIZE_BOUNDARY
1153 @item STRUCTURE_SIZE_BOUNDARY
1154 Number of bits which any structure or union's size must be a multiple of.
1155 Each structure or union's size is rounded up to a multiple of this.
1157 If you do not define this macro, the default is the same as
1158 @code{BITS_PER_UNIT}.
1160 @findex STRICT_ALIGNMENT
1161 @item STRICT_ALIGNMENT
1162 Define this macro to be the value 1 if instructions will fail to work
1163 if given data not on the nominal alignment. If instructions will merely
1164 go slower in that case, define this macro as 0.
1166 @findex PCC_BITFIELD_TYPE_MATTERS
1167 @item PCC_BITFIELD_TYPE_MATTERS
1168 Define this if you wish to imitate the way many other C compilers handle
1169 alignment of bit-fields and the structures that contain them.
1171 The behavior is that the type written for a bit-field (@code{int},
1172 @code{short}, or other integer type) imposes an alignment for the
1173 entire structure, as if the structure really did contain an ordinary
1174 field of that type. In addition, the bit-field is placed within the
1175 structure so that it would fit within such a field, not crossing a
1178 Thus, on most machines, a bit-field whose type is written as @code{int}
1179 would not cross a four-byte boundary, and would force four-byte
1180 alignment for the whole structure. (The alignment used may not be four
1181 bytes; it is controlled by the other alignment parameters.)
1183 If the macro is defined, its definition should be a C expression;
1184 a nonzero value for the expression enables this behavior.
1186 Note that if this macro is not defined, or its value is zero, some
1187 bit-fields may cross more than one alignment boundary. The compiler can
1188 support such references if there are @samp{insv}, @samp{extv}, and
1189 @samp{extzv} insns that can directly reference memory.
1191 The other known way of making bit-fields work is to define
1192 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1193 Then every structure can be accessed with fullwords.
1195 Unless the machine has bit-field instructions or you define
1196 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1197 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1199 If your aim is to make GCC use the same conventions for laying out
1200 bit-fields as are used by another compiler, here is how to investigate
1201 what the other compiler does. Compile and run this program:
1220 printf ("Size of foo1 is %d\n",
1221 sizeof (struct foo1));
1222 printf ("Size of foo2 is %d\n",
1223 sizeof (struct foo2));
1228 If this prints 2 and 5, then the compiler's behavior is what you would
1229 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1231 @findex BITFIELD_NBYTES_LIMITED
1232 @item BITFIELD_NBYTES_LIMITED
1233 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1234 to aligning a bit-field within the structure.
1236 @findex MEMBER_TYPE_FORCES_BLK
1237 @item MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1238 Return 1 if a structure or array containing @var{field} should be accessed using
1241 If @var{field} is the only field in the structure, @var{mode} is its
1242 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1243 case where structures of one field would require the structure's mode to
1244 retain the field's mode.
1246 Normally, this is not needed. See the file @file{c4x.h} for an example
1247 of how to use this macro to prevent a structure having a floating point
1248 field from being accessed in an integer mode.
1250 @findex ROUND_TYPE_SIZE
1251 @item ROUND_TYPE_SIZE (@var{type}, @var{computed}, @var{specified})
1252 Define this macro as an expression for the overall size of a type
1253 (given by @var{type} as a tree node) when the size computed in the
1254 usual way is @var{computed} and the alignment is @var{specified}.
1256 The default is to round @var{computed} up to a multiple of @var{specified}.
1258 @findex ROUND_TYPE_SIZE_UNIT
1259 @item ROUND_TYPE_SIZE_UNIT (@var{type}, @var{computed}, @var{specified})
1260 Similar to @code{ROUND_TYPE_SIZE}, but sizes and alignments are
1261 specified in units (bytes). If you define @code{ROUND_TYPE_SIZE},
1262 you must also define this macro and they must be defined consistently
1265 @findex ROUND_TYPE_ALIGN
1266 @item ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1267 Define this macro as an expression for the alignment of a type (given
1268 by @var{type} as a tree node) if the alignment computed in the usual
1269 way is @var{computed} and the alignment explicitly specified was
1272 The default is to use @var{specified} if it is larger; otherwise, use
1273 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1275 @findex MAX_FIXED_MODE_SIZE
1276 @item MAX_FIXED_MODE_SIZE
1277 An integer expression for the size in bits of the largest integer
1278 machine mode that should actually be used. All integer machine modes of
1279 this size or smaller can be used for structures and unions with the
1280 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1281 (DImode)} is assumed.
1283 @findex VECTOR_MODE_SUPPORTED_P
1284 @item VECTOR_MODE_SUPPORTED_P(@var{mode})
1285 Define this macro to be nonzero if the port is prepared to handle insns
1286 involving vector mode @var{mode}. At the very least, it must have move
1287 patterns for this mode.
1289 @findex STACK_SAVEAREA_MODE
1290 @item STACK_SAVEAREA_MODE (@var{save_level})
1291 If defined, an expression of type @code{enum machine_mode} that
1292 specifies the mode of the save area operand of a
1293 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1294 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1295 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1296 having its mode specified.
1298 You need not define this macro if it always returns @code{Pmode}. You
1299 would most commonly define this macro if the
1300 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1303 @findex STACK_SIZE_MODE
1304 @item STACK_SIZE_MODE
1305 If defined, an expression of type @code{enum machine_mode} that
1306 specifies the mode of the size increment operand of an
1307 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1309 You need not define this macro if it always returns @code{word_mode}.
1310 You would most commonly define this macro if the @code{allocate_stack}
1311 pattern needs to support both a 32- and a 64-bit mode.
1313 @findex TARGET_FLOAT_FORMAT
1314 @item TARGET_FLOAT_FORMAT
1315 A code distinguishing the floating point format of the target machine.
1316 There are five defined values:
1319 @findex IEEE_FLOAT_FORMAT
1320 @item IEEE_FLOAT_FORMAT
1321 This code indicates IEEE floating point. It is the default; there is no
1322 need to define this macro when the format is IEEE@.
1324 @findex VAX_FLOAT_FORMAT
1325 @item VAX_FLOAT_FORMAT
1326 This code indicates the ``F float'' (for @code{float}) and ``D float''
1327 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
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.
1343 formats are actually in use on supported machines, new codes should be
1346 The ordering of the component words of floating point values stored in
1347 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1349 @findex MODE_HAS_NANS
1350 @item MODE_HAS_NANS (@var{mode})
1351 When defined, this macro should be true if @var{mode} has a NaN
1352 representation. The compiler assumes that NaNs are not equal to
1353 anything (including themselves) and that addition, subtraction,
1354 multiplication and division all return NaNs when one operand is
1357 By default, this macro is true if @var{mode} is a floating-point
1358 mode and the target floating-point format is IEEE@.
1360 @findex MODE_HAS_INFINITIES
1361 @item MODE_HAS_INFINITIES (@var{mode})
1362 This macro should be true if @var{mode} can represent infinity. At
1363 present, the compiler uses this macro to decide whether @samp{x - x}
1364 is always defined. By default, the macro is true when @var{mode}
1365 is a floating-point mode and the target format is IEEE@.
1367 @findex MODE_HAS_SIGNED_ZEROS
1368 @item MODE_HAS_SIGNED_ZEROS (@var{mode})
1369 True if @var{mode} distinguishes between positive and negative zero.
1370 The rules are expected to follow the IEEE standard:
1374 @samp{x + x} has the same sign as @samp{x}.
1377 If the sum of two values with opposite sign is zero, the result is
1378 positive for all rounding modes expect towards @minus{}infinity, for
1379 which it is negative.
1382 The sign of a product or quotient is negative when exactly one
1383 of the operands is negative.
1386 The default definition is true if @var{mode} is a floating-point
1387 mode and the target format is IEEE@.
1389 @findex MODE_HAS_SIGN_DEPENDENT_ROUNDING
1390 @item MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1391 If defined, this macro should be true for @var{mode} if it has at
1392 least one rounding mode in which @samp{x} and @samp{-x} can be
1393 rounded to numbers of different magnitude. Two such modes are
1394 towards @minus{}infinity and towards +infinity.
1396 The default definition of this macro is true if @var{mode} is
1397 a floating-point mode and the target format is IEEE@.
1399 @findex ROUND_TOWARDS_ZERO
1400 @item ROUND_TOWARDS_ZERO
1401 If defined, this macro should be true if the prevailing rounding
1402 mode is towards zero. A true value has the following effects:
1406 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1409 @file{libgcc.a}'s floating-point emulator will round towards zero
1410 rather than towards nearest.
1413 The compiler's floating-point emulator will round towards zero after
1414 doing arithmetic, and when converting from the internal float format to
1418 The macro does not affect the parsing of string literals. When the
1419 primary rounding mode is towards zero, library functions like
1420 @code{strtod} might still round towards nearest, and the compiler's
1421 parser should behave like the target's @code{strtod} where possible.
1423 Not defining this macro is equivalent to returning zero.
1425 @findex LARGEST_EXPONENT_IS_NORMAL
1426 @item LARGEST_EXPONENT_IS_NORMAL (@var{size})
1427 This macro should return true if floats with @var{size}
1428 bits do not have a NaN or infinity representation, but use the largest
1429 exponent for normal numbers instead.
1431 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1432 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1433 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1434 floating-point arithmetic.
1436 The default definition of this macro returns false for all sizes.
1439 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1440 This target hook returns @code{true} if bit-fields in the given
1441 @var{record_type} are to be laid out following the rules of Microsoft
1442 Visual C/C++, namely: (i) a bit-field won't share the same storage
1443 unit with the previous bit-field if their underlying types have
1444 different sizes, and the bit-field will be aligned to the highest
1445 alignment of the underlying types of itself and of the previous
1446 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1447 the whole enclosing structure, even if it is unnamed; except that
1448 (iii) a zero-sized bit-field will be disregarded unless it follows
1449 another bit-field of nonzero size. If this hook returns @code{true},
1450 other macros that control bit-field layout are ignored.
1452 When a bit-field is inserted into a packed record, the whole size
1453 of the underlying type is used by one or more same-size adjacent
1454 bit-fields (that is, if its long:3, 32 bits is used in the record,
1455 and any additional adjacent long bit-fields are packed into the same
1456 chunk of 32 bits. However, if the size changes, a new field of that
1457 size is allocated). In an unpacked record, this is the same as using
1458 alignment, but not equivalent when packing.
1460 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1461 the latter will take precedence. If @samp{__attribute__((packed))} is
1462 used on a single field when MS bit-fields are in use, it will take
1463 precedence for that field, but the alignment of the rest of the structure
1464 may affect its placement.
1468 @section Layout of Source Language Data Types
1470 These macros define the sizes and other characteristics of the standard
1471 basic data types used in programs being compiled. Unlike the macros in
1472 the previous section, these apply to specific features of C and related
1473 languages, rather than to fundamental aspects of storage layout.
1476 @findex INT_TYPE_SIZE
1478 A C expression for the size in bits of the type @code{int} on the
1479 target machine. If you don't define this, the default is one word.
1481 @findex SHORT_TYPE_SIZE
1482 @item SHORT_TYPE_SIZE
1483 A C expression for the size in bits of the type @code{short} on the
1484 target machine. If you don't define this, the default is half a word.
1485 (If this would be less than one storage unit, it is rounded up to one
1488 @findex LONG_TYPE_SIZE
1489 @item LONG_TYPE_SIZE
1490 A C expression for the size in bits of the type @code{long} on the
1491 target machine. If you don't define this, the default is one word.
1493 @findex ADA_LONG_TYPE_SIZE
1494 @item ADA_LONG_TYPE_SIZE
1495 On some machines, the size used for the Ada equivalent of the type
1496 @code{long} by a native Ada compiler differs from that used by C. In
1497 that situation, define this macro to be a C expression to be used for
1498 the size of that type. If you don't define this, the default is the
1499 value of @code{LONG_TYPE_SIZE}.
1501 @findex MAX_LONG_TYPE_SIZE
1502 @item MAX_LONG_TYPE_SIZE
1503 Maximum number for the size in bits of the type @code{long} on the
1504 target machine. If this is undefined, the default is
1505 @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the
1506 largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is
1509 @findex LONG_LONG_TYPE_SIZE
1510 @item LONG_LONG_TYPE_SIZE
1511 A C expression for the size in bits of the type @code{long long} on the
1512 target machine. If you don't define this, the default is two
1513 words. If you want to support GNU Ada on your machine, the value of this
1514 macro must be at least 64.
1516 @findex CHAR_TYPE_SIZE
1517 @item CHAR_TYPE_SIZE
1518 A C expression for the size in bits of the type @code{char} on the
1519 target machine. If you don't define this, the default is
1520 @code{BITS_PER_UNIT}.
1522 @findex BOOL_TYPE_SIZE
1523 @item BOOL_TYPE_SIZE
1524 A C expression for the size in bits of the C++ type @code{bool} and
1525 C99 type @code{_Bool} on the target machine. If you don't define
1526 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1528 @findex FLOAT_TYPE_SIZE
1529 @item FLOAT_TYPE_SIZE
1530 A C expression for the size in bits of the type @code{float} on the
1531 target machine. If you don't define this, the default is one word.
1533 @findex DOUBLE_TYPE_SIZE
1534 @item DOUBLE_TYPE_SIZE
1535 A C expression for the size in bits of the type @code{double} on the
1536 target machine. If you don't define this, the default is two
1539 @findex LONG_DOUBLE_TYPE_SIZE
1540 @item LONG_DOUBLE_TYPE_SIZE
1541 A C expression for the size in bits of the type @code{long double} on
1542 the target machine. If you don't define this, the default is two
1545 @findex MAX_LONG_DOUBLE_TYPE_SIZE
1546 Maximum number for the size in bits of the type @code{long double} on the
1547 target machine. If this is undefined, the default is
1548 @code{LONG_DOUBLE_TYPE_SIZE}. Otherwise, it is the constant value that is
1549 the largest value that @code{LONG_DOUBLE_TYPE_SIZE} can have at run-time.
1550 This is used in @code{cpp}.
1552 @findex TARGET_FLT_EVAL_METHOD
1553 @item TARGET_FLT_EVAL_METHOD
1554 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1555 assuming, if applicable, that the floating-point control word is in its
1556 default state. If you do not define this macro the value of
1557 @code{FLT_EVAL_METHOD} will be zero.
1559 @findex WIDEST_HARDWARE_FP_SIZE
1560 @item WIDEST_HARDWARE_FP_SIZE
1561 A C expression for the size in bits of the widest floating-point format
1562 supported by the hardware. If you define this macro, you must specify a
1563 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1564 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1567 @findex DEFAULT_SIGNED_CHAR
1568 @item DEFAULT_SIGNED_CHAR
1569 An expression whose value is 1 or 0, according to whether the type
1570 @code{char} should be signed or unsigned by default. The user can
1571 always override this default with the options @option{-fsigned-char}
1572 and @option{-funsigned-char}.
1574 @findex DEFAULT_SHORT_ENUMS
1575 @item DEFAULT_SHORT_ENUMS
1576 A C expression to determine whether to give an @code{enum} type
1577 only as many bytes as it takes to represent the range of possible values
1578 of that type. A nonzero value means to do that; a zero value means all
1579 @code{enum} types should be allocated like @code{int}.
1581 If you don't define the macro, the default is 0.
1585 A C expression for a string describing the name of the data type to use
1586 for size values. The typedef name @code{size_t} is defined using the
1587 contents of the string.
1589 The string can contain more than one keyword. If so, separate them with
1590 spaces, and write first any length keyword, then @code{unsigned} if
1591 appropriate, and finally @code{int}. The string must exactly match one
1592 of the data type names defined in the function
1593 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1594 omit @code{int} or change the order---that would cause the compiler to
1597 If you don't define this macro, the default is @code{"long unsigned
1600 @findex PTRDIFF_TYPE
1602 A C expression for a string describing the name of the data type to use
1603 for the result of subtracting two pointers. The typedef name
1604 @code{ptrdiff_t} is defined using the contents of the string. See
1605 @code{SIZE_TYPE} above for more information.
1607 If you don't define this macro, the default is @code{"long int"}.
1611 A C expression for a string describing the name of the data type to use
1612 for wide characters. The typedef name @code{wchar_t} is defined using
1613 the contents of the string. See @code{SIZE_TYPE} above for more
1616 If you don't define this macro, the default is @code{"int"}.
1618 @findex WCHAR_TYPE_SIZE
1619 @item WCHAR_TYPE_SIZE
1620 A C expression for the size in bits of the data type for wide
1621 characters. This is used in @code{cpp}, which cannot make use of
1624 @findex MAX_WCHAR_TYPE_SIZE
1625 @item MAX_WCHAR_TYPE_SIZE
1626 Maximum number for the size in bits of the data type for wide
1627 characters. If this is undefined, the default is
1628 @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1629 largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is
1632 @findex GCOV_TYPE_SIZE
1633 @item GCOV_TYPE_SIZE
1634 A C expression for the size in bits of the type used for gcov counters on the
1635 target machine. If you don't define this, the default is one
1636 @code{LONG_TYPE_SIZE} in case it is greater or equal to 64-bit and
1637 @code{LONG_LONG_TYPE_SIZE} otherwise. You may want to re-define the type to
1638 ensure atomicity for counters in multithreaded programs.
1642 A C expression for a string describing the name of the data type to
1643 use for wide characters passed to @code{printf} and returned from
1644 @code{getwc}. The typedef name @code{wint_t} is defined using the
1645 contents of the string. See @code{SIZE_TYPE} above for more
1648 If you don't define this macro, the default is @code{"unsigned int"}.
1652 A C expression for a string describing the name of the data type that
1653 can represent any value of any standard or extended signed integer type.
1654 The typedef name @code{intmax_t} is defined using the contents of the
1655 string. See @code{SIZE_TYPE} above for more information.
1657 If you don't define this macro, the default is the first of
1658 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1659 much precision as @code{long long int}.
1661 @findex UINTMAX_TYPE
1663 A C expression for a string describing the name of the data type that
1664 can represent any value of any standard or extended unsigned integer
1665 type. The typedef name @code{uintmax_t} is defined using the contents
1666 of the string. See @code{SIZE_TYPE} above for more information.
1668 If you don't define this macro, the default is the first of
1669 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1670 unsigned int"} that has as much precision as @code{long long unsigned
1673 @findex TARGET_PTRMEMFUNC_VBIT_LOCATION
1674 @item TARGET_PTRMEMFUNC_VBIT_LOCATION
1675 The C++ compiler represents a pointer-to-member-function with a struct
1682 ptrdiff_t vtable_index;
1689 The C++ compiler must use one bit to indicate whether the function that
1690 will be called through a pointer-to-member-function is virtual.
1691 Normally, we assume that the low-order bit of a function pointer must
1692 always be zero. Then, by ensuring that the vtable_index is odd, we can
1693 distinguish which variant of the union is in use. But, on some
1694 platforms function pointers can be odd, and so this doesn't work. In
1695 that case, we use the low-order bit of the @code{delta} field, and shift
1696 the remainder of the @code{delta} field to the left.
1698 GCC will automatically make the right selection about where to store
1699 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1700 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1701 set such that functions always start at even addresses, but the lowest
1702 bit of pointers to functions indicate whether the function at that
1703 address is in ARM or Thumb mode. If this is the case of your
1704 architecture, you should define this macro to
1705 @code{ptrmemfunc_vbit_in_delta}.
1707 In general, you should not have to define this macro. On architectures
1708 in which function addresses are always even, according to
1709 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1710 @code{ptrmemfunc_vbit_in_pfn}.
1712 @findex TARGET_VTABLE_USES_DESCRIPTORS
1713 @item TARGET_VTABLE_USES_DESCRIPTORS
1714 Normally, the C++ compiler uses function pointers in vtables. This
1715 macro allows the target to change to use ``function descriptors''
1716 instead. Function descriptors are found on targets for whom a
1717 function pointer is actually a small data structure. Normally the
1718 data structure consists of the actual code address plus a data
1719 pointer to which the function's data is relative.
1721 If vtables are used, the value of this macro should be the number
1722 of words that the function descriptor occupies.
1724 @findex TARGET_VTABLE_ENTRY_ALIGN
1725 @item TARGET_VTABLE_ENTRY_ALIGN
1726 By default, the vtable entries are void pointers, the so the alignment
1727 is the same as pointer alignment. The value of this macro specifies
1728 the alignment of the vtable entry in bits. It should be defined only
1729 when special alignment is necessary. */
1731 @findex TARGET_VTABLE_DATA_ENTRY_DISTANCE
1732 @item TARGET_VTABLE_DATA_ENTRY_DISTANCE
1733 There are a few non-descriptor entries in the vtable at offsets below
1734 zero. If these entries must be padded (say, to preserve the alignment
1735 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1736 of words in each data entry.
1739 @node Escape Sequences
1740 @section Target Character Escape Sequences
1741 @cindex escape sequences
1743 By default, GCC assumes that the C character escape sequences take on
1744 their ASCII values for the target. If this is not correct, you must
1745 explicitly define all of the macros below.
1750 A C constant expression for the integer value for escape sequence
1755 A C constant expression for the integer value of the target escape
1756 character. As an extension, GCC evaluates the escape sequences
1757 @samp{\e} and @samp{\E} to this.
1761 @findex TARGET_NEWLINE
1764 @itemx TARGET_NEWLINE
1765 C constant expressions for the integer values for escape sequences
1766 @samp{\b}, @samp{\t} and @samp{\n}.
1774 C constant expressions for the integer values for escape sequences
1775 @samp{\v}, @samp{\f} and @samp{\r}.
1779 @section Register Usage
1780 @cindex register usage
1782 This section explains how to describe what registers the target machine
1783 has, and how (in general) they can be used.
1785 The description of which registers a specific instruction can use is
1786 done with register classes; see @ref{Register Classes}. For information
1787 on using registers to access a stack frame, see @ref{Frame Registers}.
1788 For passing values in registers, see @ref{Register Arguments}.
1789 For returning values in registers, see @ref{Scalar Return}.
1792 * Register Basics:: Number and kinds of registers.
1793 * Allocation Order:: Order in which registers are allocated.
1794 * Values in Registers:: What kinds of values each reg can hold.
1795 * Leaf Functions:: Renumbering registers for leaf functions.
1796 * Stack Registers:: Handling a register stack such as 80387.
1799 @node Register Basics
1800 @subsection Basic Characteristics of Registers
1802 @c prevent bad page break with this line
1803 Registers have various characteristics.
1806 @findex FIRST_PSEUDO_REGISTER
1807 @item FIRST_PSEUDO_REGISTER
1808 Number of hardware registers known to the compiler. They receive
1809 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1810 pseudo register's number really is assigned the number
1811 @code{FIRST_PSEUDO_REGISTER}.
1813 @item FIXED_REGISTERS
1814 @findex FIXED_REGISTERS
1815 @cindex fixed register
1816 An initializer that says which registers are used for fixed purposes
1817 all throughout the compiled code and are therefore not available for
1818 general allocation. These would include the stack pointer, the frame
1819 pointer (except on machines where that can be used as a general
1820 register when no frame pointer is needed), the program counter on
1821 machines where that is considered one of the addressable registers,
1822 and any other numbered register with a standard use.
1824 This information is expressed as a sequence of numbers, separated by
1825 commas and surrounded by braces. The @var{n}th number is 1 if
1826 register @var{n} is fixed, 0 otherwise.
1828 The table initialized from this macro, and the table initialized by
1829 the following one, may be overridden at run time either automatically,
1830 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1831 the user with the command options @option{-ffixed-@var{reg}},
1832 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1834 @findex CALL_USED_REGISTERS
1835 @item CALL_USED_REGISTERS
1836 @cindex call-used register
1837 @cindex call-clobbered register
1838 @cindex call-saved register
1839 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1840 clobbered (in general) by function calls as well as for fixed
1841 registers. This macro therefore identifies the registers that are not
1842 available for general allocation of values that must live across
1845 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1846 automatically saves it on function entry and restores it on function
1847 exit, if the register is used within the function.
1849 @findex CALL_REALLY_USED_REGISTERS
1850 @item CALL_REALLY_USED_REGISTERS
1851 @cindex call-used register
1852 @cindex call-clobbered register
1853 @cindex call-saved register
1854 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1855 that the entire set of @code{FIXED_REGISTERS} be included.
1856 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1857 This macro is optional. If not specified, it defaults to the value
1858 of @code{CALL_USED_REGISTERS}.
1860 @findex HARD_REGNO_CALL_PART_CLOBBERED
1861 @item HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1862 @cindex call-used register
1863 @cindex call-clobbered register
1864 @cindex call-saved register
1865 A C expression that is nonzero if it is not permissible to store a
1866 value of mode @var{mode} in hard register number @var{regno} across a
1867 call without some part of it being clobbered. For most machines this
1868 macro need not be defined. It is only required for machines that do not
1869 preserve the entire contents of a register across a call.
1871 @findex CONDITIONAL_REGISTER_USAGE
1873 @findex call_used_regs
1874 @item CONDITIONAL_REGISTER_USAGE
1875 Zero or more C statements that may conditionally modify five variables
1876 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1877 @code{reg_names}, and @code{reg_class_contents}, to take into account
1878 any dependence of these register sets on target flags. The first three
1879 of these are of type @code{char []} (interpreted as Boolean vectors).
1880 @code{global_regs} is a @code{const char *[]}, and
1881 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1882 called, @code{fixed_regs}, @code{call_used_regs},
1883 @code{reg_class_contents}, and @code{reg_names} have been initialized
1884 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1885 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1886 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1887 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1888 command options have been applied.
1890 You need not define this macro if it has no work to do.
1892 @cindex disabling certain registers
1893 @cindex controlling register usage
1894 If the usage of an entire class of registers depends on the target
1895 flags, you may indicate this to GCC by using this macro to modify
1896 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1897 registers in the classes which should not be used by GCC@. Also define
1898 the macro @code{REG_CLASS_FROM_LETTER} to return @code{NO_REGS} if it
1899 is called with a letter for a class that shouldn't be used.
1901 (However, if this class is not included in @code{GENERAL_REGS} and all
1902 of the insn patterns whose constraints permit this class are
1903 controlled by target switches, then GCC will automatically avoid using
1904 these registers when the target switches are opposed to them.)
1906 @findex NON_SAVING_SETJMP
1907 @item NON_SAVING_SETJMP
1908 If this macro is defined and has a nonzero value, it means that
1909 @code{setjmp} and related functions fail to save the registers, or that
1910 @code{longjmp} fails to restore them. To compensate, the compiler
1911 avoids putting variables in registers in functions that use
1914 @findex INCOMING_REGNO
1915 @item INCOMING_REGNO (@var{out})
1916 Define this macro if the target machine has register windows. This C
1917 expression returns the register number as seen by the called function
1918 corresponding to the register number @var{out} as seen by the calling
1919 function. Return @var{out} if register number @var{out} is not an
1922 @findex OUTGOING_REGNO
1923 @item OUTGOING_REGNO (@var{in})
1924 Define this macro if the target machine has register windows. This C
1925 expression returns the register number as seen by the calling function
1926 corresponding to the register number @var{in} as seen by the called
1927 function. Return @var{in} if register number @var{in} is not an inbound
1931 @item LOCAL_REGNO (@var{regno})
1932 Define this macro if the target machine has register windows. This C
1933 expression returns true if the register is call-saved but is in the
1934 register window. Unlike most call-saved registers, such registers
1935 need not be explicitly restored on function exit or during non-local
1941 If the program counter has a register number, define this as that
1942 register number. Otherwise, do not define it.
1946 @node Allocation Order
1947 @subsection Order of Allocation of Registers
1948 @cindex order of register allocation
1949 @cindex register allocation order
1951 @c prevent bad page break with this line
1952 Registers are allocated in order.
1955 @findex REG_ALLOC_ORDER
1956 @item REG_ALLOC_ORDER
1957 If defined, an initializer for a vector of integers, containing the
1958 numbers of hard registers in the order in which GCC should prefer
1959 to use them (from most preferred to least).
1961 If this macro is not defined, registers are used lowest numbered first
1962 (all else being equal).
1964 One use of this macro is on machines where the highest numbered
1965 registers must always be saved and the save-multiple-registers
1966 instruction supports only sequences of consecutive registers. On such
1967 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1968 the highest numbered allocable register first.
1970 @findex ORDER_REGS_FOR_LOCAL_ALLOC
1971 @item ORDER_REGS_FOR_LOCAL_ALLOC
1972 A C statement (sans semicolon) to choose the order in which to allocate
1973 hard registers for pseudo-registers local to a basic block.
1975 Store the desired register order in the array @code{reg_alloc_order}.
1976 Element 0 should be the register to allocate first; element 1, the next
1977 register; and so on.
1979 The macro body should not assume anything about the contents of
1980 @code{reg_alloc_order} before execution of the macro.
1982 On most machines, it is not necessary to define this macro.
1985 @node Values in Registers
1986 @subsection How Values Fit in Registers
1988 This section discusses the macros that describe which kinds of values
1989 (specifically, which machine modes) each register can hold, and how many
1990 consecutive registers are needed for a given mode.
1993 @findex HARD_REGNO_NREGS
1994 @item HARD_REGNO_NREGS (@var{regno}, @var{mode})
1995 A C expression for the number of consecutive hard registers, starting
1996 at register number @var{regno}, required to hold a value of mode
1999 On a machine where all registers are exactly one word, a suitable
2000 definition of this macro is
2003 #define HARD_REGNO_NREGS(REGNO, MODE) \
2004 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2008 @findex HARD_REGNO_MODE_OK
2009 @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2010 A C expression that is nonzero if it is permissible to store a value
2011 of mode @var{mode} in hard register number @var{regno} (or in several
2012 registers starting with that one). For a machine where all registers
2013 are equivalent, a suitable definition is
2016 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2019 You need not include code to check for the numbers of fixed registers,
2020 because the allocation mechanism considers them to be always occupied.
2022 @cindex register pairs
2023 On some machines, double-precision values must be kept in even/odd
2024 register pairs. You can implement that by defining this macro to reject
2025 odd register numbers for such modes.
2027 The minimum requirement for a mode to be OK in a register is that the
2028 @samp{mov@var{mode}} instruction pattern support moves between the
2029 register and other hard register in the same class and that moving a
2030 value into the register and back out not alter it.
2032 Since the same instruction used to move @code{word_mode} will work for
2033 all narrower integer modes, it is not necessary on any machine for
2034 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2035 you define patterns @samp{movhi}, etc., to take advantage of this. This
2036 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2037 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2040 Many machines have special registers for floating point arithmetic.
2041 Often people assume that floating point machine modes are allowed only
2042 in floating point registers. This is not true. Any registers that
2043 can hold integers can safely @emph{hold} a floating point machine
2044 mode, whether or not floating arithmetic can be done on it in those
2045 registers. Integer move instructions can be used to move the values.
2047 On some machines, though, the converse is true: fixed-point machine
2048 modes may not go in floating registers. This is true if the floating
2049 registers normalize any value stored in them, because storing a
2050 non-floating value there would garble it. In this case,
2051 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2052 floating registers. But if the floating registers do not automatically
2053 normalize, if you can store any bit pattern in one and retrieve it
2054 unchanged without a trap, then any machine mode may go in a floating
2055 register, so you can define this macro to say so.
2057 The primary significance of special floating registers is rather that
2058 they are the registers acceptable in floating point arithmetic
2059 instructions. However, this is of no concern to
2060 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2061 constraints for those instructions.
2063 On some machines, the floating registers are especially slow to access,
2064 so that it is better to store a value in a stack frame than in such a
2065 register if floating point arithmetic is not being done. As long as the
2066 floating registers are not in class @code{GENERAL_REGS}, they will not
2067 be used unless some pattern's constraint asks for one.
2069 @findex MODES_TIEABLE_P
2070 @item MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2071 A C expression that is nonzero if a value of mode
2072 @var{mode1} is accessible in mode @var{mode2} without copying.
2074 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2075 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2076 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2077 should be nonzero. If they differ for any @var{r}, you should define
2078 this macro to return zero unless some other mechanism ensures the
2079 accessibility of the value in a narrower mode.
2081 You should define this macro to return nonzero in as many cases as
2082 possible since doing so will allow GCC to perform better register
2085 @findex AVOID_CCMODE_COPIES
2086 @item AVOID_CCMODE_COPIES
2087 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2088 registers. You should only define this macro if support for copying to/from
2089 @code{CCmode} is incomplete.
2092 @node Leaf Functions
2093 @subsection Handling Leaf Functions
2095 @cindex leaf functions
2096 @cindex functions, leaf
2097 On some machines, a leaf function (i.e., one which makes no calls) can run
2098 more efficiently if it does not make its own register window. Often this
2099 means it is required to receive its arguments in the registers where they
2100 are passed by the caller, instead of the registers where they would
2103 The special treatment for leaf functions generally applies only when
2104 other conditions are met; for example, often they may use only those
2105 registers for its own variables and temporaries. We use the term ``leaf
2106 function'' to mean a function that is suitable for this special
2107 handling, so that functions with no calls are not necessarily ``leaf
2110 GCC assigns register numbers before it knows whether the function is
2111 suitable for leaf function treatment. So it needs to renumber the
2112 registers in order to output a leaf function. The following macros
2116 @findex LEAF_REGISTERS
2117 @item LEAF_REGISTERS
2118 Name of a char vector, indexed by hard register number, which
2119 contains 1 for a register that is allowable in a candidate for leaf
2122 If leaf function treatment involves renumbering the registers, then the
2123 registers marked here should be the ones before renumbering---those that
2124 GCC would ordinarily allocate. The registers which will actually be
2125 used in the assembler code, after renumbering, should not be marked with 1
2128 Define this macro only if the target machine offers a way to optimize
2129 the treatment of leaf functions.
2131 @findex LEAF_REG_REMAP
2132 @item LEAF_REG_REMAP (@var{regno})
2133 A C expression whose value is the register number to which @var{regno}
2134 should be renumbered, when a function is treated as a leaf function.
2136 If @var{regno} is a register number which should not appear in a leaf
2137 function before renumbering, then the expression should yield @minus{}1, which
2138 will cause the compiler to abort.
2140 Define this macro only if the target machine offers a way to optimize the
2141 treatment of leaf functions, and registers need to be renumbered to do
2145 @findex current_function_is_leaf
2146 @findex current_function_uses_only_leaf_regs
2147 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2148 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2149 specially. They can test the C variable @code{current_function_is_leaf}
2150 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2151 set prior to local register allocation and is valid for the remaining
2152 compiler passes. They can also test the C variable
2153 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2154 functions which only use leaf registers.
2155 @code{current_function_uses_only_leaf_regs} is valid after reload and is
2156 only useful if @code{LEAF_REGISTERS} is defined.
2157 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2158 @c of the next paragraph?! --mew 2feb93
2160 @node Stack Registers
2161 @subsection Registers That Form a Stack
2163 There are special features to handle computers where some of the
2164 ``registers'' form a stack, as in the 80387 coprocessor for the 80386.
2165 Stack registers are normally written by pushing onto the stack, and are
2166 numbered relative to the top of the stack.
2168 Currently, GCC can only handle one group of stack-like registers, and
2169 they must be consecutively numbered.
2174 Define this if the machine has any stack-like registers.
2176 @findex FIRST_STACK_REG
2177 @item FIRST_STACK_REG
2178 The number of the first stack-like register. This one is the top
2181 @findex LAST_STACK_REG
2182 @item LAST_STACK_REG
2183 The number of the last stack-like register. This one is the bottom of
2187 @node Register Classes
2188 @section Register Classes
2189 @cindex register class definitions
2190 @cindex class definitions, register
2192 On many machines, the numbered registers are not all equivalent.
2193 For example, certain registers may not be allowed for indexed addressing;
2194 certain registers may not be allowed in some instructions. These machine
2195 restrictions are described to the compiler using @dfn{register classes}.
2197 You define a number of register classes, giving each one a name and saying
2198 which of the registers belong to it. Then you can specify register classes
2199 that are allowed as operands to particular instruction patterns.
2203 In general, each register will belong to several classes. In fact, one
2204 class must be named @code{ALL_REGS} and contain all the registers. Another
2205 class must be named @code{NO_REGS} and contain no registers. Often the
2206 union of two classes will be another class; however, this is not required.
2208 @findex GENERAL_REGS
2209 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2210 terribly special about the name, but the operand constraint letters
2211 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2212 the same as @code{ALL_REGS}, just define it as a macro which expands
2215 Order the classes so that if class @var{x} is contained in class @var{y}
2216 then @var{x} has a lower class number than @var{y}.
2218 The way classes other than @code{GENERAL_REGS} are specified in operand
2219 constraints is through machine-dependent operand constraint letters.
2220 You can define such letters to correspond to various classes, then use
2221 them in operand constraints.
2223 You should define a class for the union of two classes whenever some
2224 instruction allows both classes. For example, if an instruction allows
2225 either a floating point (coprocessor) register or a general register for a
2226 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2227 which includes both of them. Otherwise you will get suboptimal code.
2229 You must also specify certain redundant information about the register
2230 classes: for each class, which classes contain it and which ones are
2231 contained in it; for each pair of classes, the largest class contained
2234 When a value occupying several consecutive registers is expected in a
2235 certain class, all the registers used must belong to that class.
2236 Therefore, register classes cannot be used to enforce a requirement for
2237 a register pair to start with an even-numbered register. The way to
2238 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2240 Register classes used for input-operands of bitwise-and or shift
2241 instructions have a special requirement: each such class must have, for
2242 each fixed-point machine mode, a subclass whose registers can transfer that
2243 mode to or from memory. For example, on some machines, the operations for
2244 single-byte values (@code{QImode}) are limited to certain registers. When
2245 this is so, each register class that is used in a bitwise-and or shift
2246 instruction must have a subclass consisting of registers from which
2247 single-byte values can be loaded or stored. This is so that
2248 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2251 @findex enum reg_class
2252 @item enum reg_class
2253 An enumeral type that must be defined with all the register class names
2254 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
2255 must be the last register class, followed by one more enumeral value,
2256 @code{LIM_REG_CLASSES}, which is not a register class but rather
2257 tells how many classes there are.
2259 Each register class has a number, which is the value of casting
2260 the class name to type @code{int}. The number serves as an index
2261 in many of the tables described below.
2263 @findex N_REG_CLASSES
2265 The number of distinct register classes, defined as follows:
2268 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2271 @findex REG_CLASS_NAMES
2272 @item REG_CLASS_NAMES
2273 An initializer containing the names of the register classes as C string
2274 constants. These names are used in writing some of the debugging dumps.
2276 @findex REG_CLASS_CONTENTS
2277 @item REG_CLASS_CONTENTS
2278 An initializer containing the contents of the register classes, as integers
2279 which are bit masks. The @var{n}th integer specifies the contents of class
2280 @var{n}. The way the integer @var{mask} is interpreted is that
2281 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2283 When the machine has more than 32 registers, an integer does not suffice.
2284 Then the integers are replaced by sub-initializers, braced groupings containing
2285 several integers. Each sub-initializer must be suitable as an initializer
2286 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2287 In this situation, the first integer in each sub-initializer corresponds to
2288 registers 0 through 31, the second integer to registers 32 through 63, and
2291 @findex REGNO_REG_CLASS
2292 @item REGNO_REG_CLASS (@var{regno})
2293 A C expression whose value is a register class containing hard register
2294 @var{regno}. In general there is more than one such class; choose a class
2295 which is @dfn{minimal}, meaning that no smaller class also contains the
2298 @findex BASE_REG_CLASS
2299 @item BASE_REG_CLASS
2300 A macro whose definition is the name of the class to which a valid
2301 base register must belong. A base register is one used in an address
2302 which is the register value plus a displacement.
2304 @findex MODE_BASE_REG_CLASS
2305 @item MODE_BASE_REG_CLASS (@var{mode})
2306 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2307 the selection of a base register in a mode depenedent manner. If
2308 @var{mode} is VOIDmode then it should return the same value as
2309 @code{BASE_REG_CLASS}.
2311 @findex INDEX_REG_CLASS
2312 @item INDEX_REG_CLASS
2313 A macro whose definition is the name of the class to which a valid
2314 index register must belong. An index register is one used in an
2315 address where its value is either multiplied by a scale factor or
2316 added to another register (as well as added to a displacement).
2318 @findex REG_CLASS_FROM_LETTER
2319 @item REG_CLASS_FROM_LETTER (@var{char})
2320 A C expression which defines the machine-dependent operand constraint
2321 letters for register classes. If @var{char} is such a letter, the
2322 value should be the register class corresponding to it. Otherwise,
2323 the value should be @code{NO_REGS}. The register letter @samp{r},
2324 corresponding to class @code{GENERAL_REGS}, will not be passed
2325 to this macro; you do not need to handle it.
2327 @findex REGNO_OK_FOR_BASE_P
2328 @item REGNO_OK_FOR_BASE_P (@var{num})
2329 A C expression which is nonzero if register number @var{num} is
2330 suitable for use as a base register in operand addresses. It may be
2331 either a suitable hard register or a pseudo register that has been
2332 allocated such a hard register.
2334 @findex REGNO_MODE_OK_FOR_BASE_P
2335 @item REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2336 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2337 that expression may examine the mode of the memory reference in
2338 @var{mode}. You should define this macro if the mode of the memory
2339 reference affects whether a register may be used as a base register. If
2340 you define this macro, the compiler will use it instead of
2341 @code{REGNO_OK_FOR_BASE_P}.
2343 @findex REGNO_OK_FOR_INDEX_P
2344 @item REGNO_OK_FOR_INDEX_P (@var{num})
2345 A C expression which is nonzero if register number @var{num} is
2346 suitable for use as an index register in operand addresses. It may be
2347 either a suitable hard register or a pseudo register that has been
2348 allocated such a hard register.
2350 The difference between an index register and a base register is that
2351 the index register may be scaled. If an address involves the sum of
2352 two registers, neither one of them scaled, then either one may be
2353 labeled the ``base'' and the other the ``index''; but whichever
2354 labeling is used must fit the machine's constraints of which registers
2355 may serve in each capacity. The compiler will try both labelings,
2356 looking for one that is valid, and will reload one or both registers
2357 only if neither labeling works.
2359 @findex PREFERRED_RELOAD_CLASS
2360 @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2361 A C expression that places additional restrictions on the register class
2362 to use when it is necessary to copy value @var{x} into a register in class
2363 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2364 another, smaller class. On many machines, the following definition is
2368 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2371 Sometimes returning a more restrictive class makes better code. For
2372 example, on the 68000, when @var{x} is an integer constant that is in range
2373 for a @samp{moveq} instruction, the value of this macro is always
2374 @code{DATA_REGS} as long as @var{class} includes the data registers.
2375 Requiring a data register guarantees that a @samp{moveq} will be used.
2377 If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
2378 you can force @var{x} into a memory constant. This is useful on
2379 certain machines where immediate floating values cannot be loaded into
2380 certain kinds of registers.
2382 @findex PREFERRED_OUTPUT_RELOAD_CLASS
2383 @item PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2384 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2385 input reloads. If you don't define this macro, the default is to use
2386 @var{class}, unchanged.
2388 @findex LIMIT_RELOAD_CLASS
2389 @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2390 A C expression that places additional restrictions on the register class
2391 to use when it is necessary to be able to hold a value of mode
2392 @var{mode} in a reload register for which class @var{class} would
2395 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2396 there are certain modes that simply can't go in certain reload classes.
2398 The value is a register class; perhaps @var{class}, or perhaps another,
2401 Don't define this macro unless the target machine has limitations which
2402 require the macro to do something nontrivial.
2404 @findex SECONDARY_RELOAD_CLASS
2405 @findex SECONDARY_INPUT_RELOAD_CLASS
2406 @findex SECONDARY_OUTPUT_RELOAD_CLASS
2407 @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2408 @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2409 @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2410 Many machines have some registers that cannot be copied directly to or
2411 from memory or even from other types of registers. An example is the
2412 @samp{MQ} register, which on most machines, can only be copied to or
2413 from general registers, but not memory. Some machines allow copying all
2414 registers to and from memory, but require a scratch register for stores
2415 to some memory locations (e.g., those with symbolic address on the RT,
2416 and those with certain symbolic address on the SPARC when compiling
2417 PIC)@. In some cases, both an intermediate and a scratch register are
2420 You should define these macros to indicate to the reload phase that it may
2421 need to allocate at least one register for a reload in addition to the
2422 register to contain the data. Specifically, if copying @var{x} to a
2423 register @var{class} in @var{mode} requires an intermediate register,
2424 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2425 largest register class all of whose registers can be used as
2426 intermediate registers or scratch registers.
2428 If copying a register @var{class} in @var{mode} to @var{x} requires an
2429 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2430 should be defined to return the largest register class required. If the
2431 requirements for input and output reloads are the same, the macro
2432 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2435 The values returned by these macros are often @code{GENERAL_REGS}.
2436 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2437 can be directly copied to or from a register of @var{class} in
2438 @var{mode} without requiring a scratch register. Do not define this
2439 macro if it would always return @code{NO_REGS}.
2441 If a scratch register is required (either with or without an
2442 intermediate register), you should define patterns for
2443 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2444 (@pxref{Standard Names}. These patterns, which will normally be
2445 implemented with a @code{define_expand}, should be similar to the
2446 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2449 Define constraints for the reload register and scratch register that
2450 contain a single register class. If the original reload register (whose
2451 class is @var{class}) can meet the constraint given in the pattern, the
2452 value returned by these macros is used for the class of the scratch
2453 register. Otherwise, two additional reload registers are required.
2454 Their classes are obtained from the constraints in the insn pattern.
2456 @var{x} might be a pseudo-register or a @code{subreg} of a
2457 pseudo-register, which could either be in a hard register or in memory.
2458 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2459 in memory and the hard register number if it is in a register.
2461 These macros should not be used in the case where a particular class of
2462 registers can only be copied to memory and not to another class of
2463 registers. In that case, secondary reload registers are not needed and
2464 would not be helpful. Instead, a stack location must be used to perform
2465 the copy and the @code{mov@var{m}} pattern should use memory as an
2466 intermediate storage. This case often occurs between floating-point and
2469 @findex SECONDARY_MEMORY_NEEDED
2470 @item SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2471 Certain machines have the property that some registers cannot be copied
2472 to some other registers without using memory. Define this macro on
2473 those machines to be a C expression that is nonzero if objects of mode
2474 @var{m} in registers of @var{class1} can only be copied to registers of
2475 class @var{class2} by storing a register of @var{class1} into memory
2476 and loading that memory location into a register of @var{class2}.
2478 Do not define this macro if its value would always be zero.
2480 @findex SECONDARY_MEMORY_NEEDED_RTX
2481 @item SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2482 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2483 allocates a stack slot for a memory location needed for register copies.
2484 If this macro is defined, the compiler instead uses the memory location
2485 defined by this macro.
2487 Do not define this macro if you do not define
2488 @code{SECONDARY_MEMORY_NEEDED}.
2490 @findex SECONDARY_MEMORY_NEEDED_MODE
2491 @item SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2492 When the compiler needs a secondary memory location to copy between two
2493 registers of mode @var{mode}, it normally allocates sufficient memory to
2494 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2495 load operations in a mode that many bits wide and whose class is the
2496 same as that of @var{mode}.
2498 This is right thing to do on most machines because it ensures that all
2499 bits of the register are copied and prevents accesses to the registers
2500 in a narrower mode, which some machines prohibit for floating-point
2503 However, this default behavior is not correct on some machines, such as
2504 the DEC Alpha, that store short integers in floating-point registers
2505 differently than in integer registers. On those machines, the default
2506 widening will not work correctly and you must define this macro to
2507 suppress that widening in some cases. See the file @file{alpha.h} for
2510 Do not define this macro if you do not define
2511 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2512 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2514 @findex SMALL_REGISTER_CLASSES
2515 @item SMALL_REGISTER_CLASSES
2516 On some machines, it is risky to let hard registers live across arbitrary
2517 insns. Typically, these machines have instructions that require values
2518 to be in specific registers (like an accumulator), and reload will fail
2519 if the required hard register is used for another purpose across such an
2522 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2523 value on these machines. When this macro has a nonzero value, the
2524 compiler will try to minimize the lifetime of hard registers.
2526 It is always safe to define this macro with a nonzero value, but if you
2527 unnecessarily define it, you will reduce the amount of optimizations
2528 that can be performed in some cases. If you do not define this macro
2529 with a nonzero value when it is required, the compiler will run out of
2530 spill registers and print a fatal error message. For most machines, you
2531 should not define this macro at all.
2533 @findex CLASS_LIKELY_SPILLED_P
2534 @item CLASS_LIKELY_SPILLED_P (@var{class})
2535 A C expression whose value is nonzero if pseudos that have been assigned
2536 to registers of class @var{class} would likely be spilled because
2537 registers of @var{class} are needed for spill registers.
2539 The default value of this macro returns 1 if @var{class} has exactly one
2540 register and zero otherwise. On most machines, this default should be
2541 used. Only define this macro to some other expression if pseudos
2542 allocated by @file{local-alloc.c} end up in memory because their hard
2543 registers were needed for spill registers. If this macro returns nonzero
2544 for those classes, those pseudos will only be allocated by
2545 @file{global.c}, which knows how to reallocate the pseudo to another
2546 register. If there would not be another register available for
2547 reallocation, you should not change the definition of this macro since
2548 the only effect of such a definition would be to slow down register
2551 @findex CLASS_MAX_NREGS
2552 @item CLASS_MAX_NREGS (@var{class}, @var{mode})
2553 A C expression for the maximum number of consecutive registers
2554 of class @var{class} needed to hold a value of mode @var{mode}.
2556 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2557 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2558 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2559 @var{mode})} for all @var{regno} values in the class @var{class}.
2561 This macro helps control the handling of multiple-word values
2564 @item CLASS_CANNOT_CHANGE_MODE
2565 If defined, a C expression for a class that contains registers for
2566 which the compiler may not change modes arbitrarily.
2568 @item CLASS_CANNOT_CHANGE_MODE_P(@var{from}, @var{to})
2569 A C expression that is true if, for a register in
2570 @code{CLASS_CANNOT_CHANGE_MODE}, the requested mode punning is invalid.
2572 For the example, loading 32-bit integer or floating-point objects into
2573 floating-point registers on the Alpha extends them to 64 bits.
2574 Therefore loading a 64-bit object and then storing it as a 32-bit object
2575 does not store the low-order 32 bits, as would be the case for a normal
2576 register. Therefore, @file{alpha.h} defines @code{CLASS_CANNOT_CHANGE_MODE}
2577 as @code{FLOAT_REGS} and @code{CLASS_CANNOT_CHANGE_MODE_P} restricts
2578 mode changes to same-size modes.
2580 Compare this to IA-64, which extends floating-point values to 82-bits,
2581 and stores 64-bit integers in a different format than 64-bit doubles.
2582 Therefore @code{CLASS_CANNOT_CHANGE_MODE_P} is always true.
2585 Three other special macros describe which operands fit which constraint
2589 @findex CONST_OK_FOR_LETTER_P
2590 @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2591 A C expression that defines the machine-dependent operand constraint
2592 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2593 particular ranges of integer values. If @var{c} is one of those
2594 letters, the expression should check that @var{value}, an integer, is in
2595 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2596 not one of those letters, the value should be 0 regardless of
2599 @findex CONST_DOUBLE_OK_FOR_LETTER_P
2600 @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2601 A C expression that defines the machine-dependent operand constraint
2602 letters that specify particular ranges of @code{const_double} values
2603 (@samp{G} or @samp{H}).
2605 If @var{c} is one of those letters, the expression should check that
2606 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2607 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2608 letters, the value should be 0 regardless of @var{value}.
2610 @code{const_double} is used for all floating-point constants and for
2611 @code{DImode} fixed-point constants. A given letter can accept either
2612 or both kinds of values. It can use @code{GET_MODE} to distinguish
2613 between these kinds.
2615 @findex EXTRA_CONSTRAINT
2616 @item EXTRA_CONSTRAINT (@var{value}, @var{c})
2617 A C expression that defines the optional machine-dependent constraint
2618 letters that can be used to segregate specific types of operands, usually
2619 memory references, for the target machine. Any letter that is not
2620 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER}
2621 may be used. Normally this macro will not be defined.
2623 If it is required for a particular target machine, it should return 1
2624 if @var{value} corresponds to the operand type represented by the
2625 constraint letter @var{c}. If @var{c} is not defined as an extra
2626 constraint, the value returned should be 0 regardless of @var{value}.
2628 For example, on the ROMP, load instructions cannot have their output
2629 in r0 if the memory reference contains a symbolic address. Constraint
2630 letter @samp{Q} is defined as representing a memory address that does
2631 @emph{not} contain a symbolic address. An alternative is specified with
2632 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2633 alternative specifies @samp{m} on the input and a register class that
2634 does not include r0 on the output.
2636 @findex EXTRA_MEMORY_CONSTRAINT
2637 @item EXTRA_MEMORY_CONSTRAINT (@var{c})
2638 A C expression that defines the optional machine-dependent constraint
2639 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2640 be treated like memory constraints by the reload pass.
2642 It should return 1 if the operand type represented by the constraint
2643 letter @var{c} comprises a subset of all memory references including
2644 all those whose address is simply a base register. This allows the reload
2645 pass to reload an operand, if it does not directly correspond to the operand
2646 type of @var{c}, by copying its address into a base register.
2648 For example, on the S/390, some instructions do not accept arbitrary
2649 memory references, but only those that do not make use of an index
2650 register. The constraint letter @samp{Q} is defined via
2651 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2652 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2653 a @samp{Q} constraint can handle any memory operand, because the
2654 reload pass knows it can be reloaded by copying the memory address
2655 into a base register if required. This is analogous to the way
2656 a @samp{o} constraint can handle any memory operand.
2658 @findex EXTRA_ADDRESS_CONSTRAINT
2659 @item EXTRA_ADDRESS_CONSTRAINT (@var{c})
2660 A C expression that defines the optional machine-dependent constraint
2661 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2662 be treated like address constraints by the reload pass.
2664 It should return 1 if the operand type represented by the constraint
2665 letter @var{c} comprises a subset of all memory addresses including
2666 all those that consist of just a base register. This allows the reload
2667 pass to reload an operand, if it does not directly correspond to the operand
2668 type of @var{c}, by copying it into a base register.
2670 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2671 be used with the @code{address_operand} predicate. It is treated
2672 analogously to the @samp{p} constraint.
2675 @node Stack and Calling
2676 @section Stack Layout and Calling Conventions
2677 @cindex calling conventions
2679 @c prevent bad page break with this line
2680 This describes the stack layout and calling conventions.
2684 * Exception Handling::
2689 * Register Arguments::
2691 * Aggregate Return::
2699 @subsection Basic Stack Layout
2700 @cindex stack frame layout
2701 @cindex frame layout
2703 @c prevent bad page break with this line
2704 Here is the basic stack layout.
2707 @findex STACK_GROWS_DOWNWARD
2708 @item STACK_GROWS_DOWNWARD
2709 Define this macro if pushing a word onto the stack moves the stack
2710 pointer to a smaller address.
2712 When we say, ``define this macro if @dots{},'' it means that the
2713 compiler checks this macro only with @code{#ifdef} so the precise
2714 definition used does not matter.
2716 @findex STACK_PUSH_CODE
2717 @item STACK_PUSH_CODE
2719 This macro defines the operation used when something is pushed
2720 on the stack. In RTL, a push operation will be
2721 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2723 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2724 and @code{POST_INC}. Which of these is correct depends on
2725 the stack direction and on whether the stack pointer points
2726 to the last item on the stack or whether it points to the
2727 space for the next item on the stack.
2729 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2730 defined, which is almost always right, and @code{PRE_INC} otherwise,
2731 which is often wrong.
2733 @findex FRAME_GROWS_DOWNWARD
2734 @item FRAME_GROWS_DOWNWARD
2735 Define this macro if the addresses of local variable slots are at negative
2736 offsets from the frame pointer.
2738 @findex ARGS_GROW_DOWNWARD
2739 @item ARGS_GROW_DOWNWARD
2740 Define this macro if successive arguments to a function occupy decreasing
2741 addresses on the stack.
2743 @findex STARTING_FRAME_OFFSET
2744 @item STARTING_FRAME_OFFSET
2745 Offset from the frame pointer to the first local variable slot to be allocated.
2747 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2748 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2749 Otherwise, it is found by adding the length of the first slot to the
2750 value @code{STARTING_FRAME_OFFSET}.
2751 @c i'm not sure if the above is still correct.. had to change it to get
2752 @c rid of an overfull. --mew 2feb93
2754 @findex STACK_POINTER_OFFSET
2755 @item STACK_POINTER_OFFSET
2756 Offset from the stack pointer register to the first location at which
2757 outgoing arguments are placed. If not specified, the default value of
2758 zero is used. This is the proper value for most machines.
2760 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2761 the first location at which outgoing arguments are placed.
2763 @findex FIRST_PARM_OFFSET
2764 @item FIRST_PARM_OFFSET (@var{fundecl})
2765 Offset from the argument pointer register to the first argument's
2766 address. On some machines it may depend on the data type of the
2769 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2770 the first argument's address.
2772 @findex STACK_DYNAMIC_OFFSET
2773 @item STACK_DYNAMIC_OFFSET (@var{fundecl})
2774 Offset from the stack pointer register to an item dynamically allocated
2775 on the stack, e.g., by @code{alloca}.
2777 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2778 length of the outgoing arguments. The default is correct for most
2779 machines. See @file{function.c} for details.
2781 @findex DYNAMIC_CHAIN_ADDRESS
2782 @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2783 A C expression whose value is RTL representing the address in a stack
2784 frame where the pointer to the caller's frame is stored. Assume that
2785 @var{frameaddr} is an RTL expression for the address of the stack frame
2788 If you don't define this macro, the default is to return the value
2789 of @var{frameaddr}---that is, the stack frame address is also the
2790 address of the stack word that points to the previous frame.
2792 @findex SETUP_FRAME_ADDRESSES
2793 @item SETUP_FRAME_ADDRESSES
2794 If defined, a C expression that produces the machine-specific code to
2795 setup the stack so that arbitrary frames can be accessed. For example,
2796 on the SPARC, we must flush all of the register windows to the stack
2797 before we can access arbitrary stack frames. You will seldom need to
2800 @findex BUILTIN_SETJMP_FRAME_VALUE
2801 @item BUILTIN_SETJMP_FRAME_VALUE
2802 If defined, a C expression that contains an rtx that is used to store
2803 the address of the current frame into the built in @code{setjmp} buffer.
2804 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2805 machines. One reason you may need to define this macro is if
2806 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2808 @findex RETURN_ADDR_RTX
2809 @item RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2810 A C expression whose value is RTL representing the value of the return
2811 address for the frame @var{count} steps up from the current frame, after
2812 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2813 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2814 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2816 The value of the expression must always be the correct address when
2817 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2818 determine the return address of other frames.
2820 @findex RETURN_ADDR_IN_PREVIOUS_FRAME
2821 @item RETURN_ADDR_IN_PREVIOUS_FRAME
2822 Define this if the return address of a particular stack frame is accessed
2823 from the frame pointer of the previous stack frame.
2825 @findex INCOMING_RETURN_ADDR_RTX
2826 @item INCOMING_RETURN_ADDR_RTX
2827 A C expression whose value is RTL representing the location of the
2828 incoming return address at the beginning of any function, before the
2829 prologue. This RTL is either a @code{REG}, indicating that the return
2830 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2833 You only need to define this macro if you want to support call frame
2834 debugging information like that provided by DWARF 2.
2836 If this RTL is a @code{REG}, you should also define
2837 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2839 @findex INCOMING_FRAME_SP_OFFSET
2840 @item INCOMING_FRAME_SP_OFFSET
2841 A C expression whose value is an integer giving the offset, in bytes,
2842 from the value of the stack pointer register to the top of the stack
2843 frame at the beginning of any function, before the prologue. The top of
2844 the frame is defined to be the value of the stack pointer in the
2845 previous frame, just before the call instruction.
2847 You only need to define this macro if you want to support call frame
2848 debugging information like that provided by DWARF 2.
2850 @findex ARG_POINTER_CFA_OFFSET
2851 @item ARG_POINTER_CFA_OFFSET (@var{fundecl})
2852 A C expression whose value is an integer giving the offset, in bytes,
2853 from the argument pointer to the canonical frame address (cfa). The
2854 final value should coincide with that calculated by
2855 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2856 during virtual register instantiation.
2858 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
2859 which is correct for most machines; in general, the arguments are found
2860 immediately before the stack frame. Note that this is not the case on
2861 some targets that save registers into the caller's frame, such as SPARC
2862 and rs6000, and so such targets need to define this macro.
2864 You only need to define this macro if the default is incorrect, and you
2865 want to support call frame debugging information like that provided by
2870 Define this macro if the stack size for the target is very small. This
2871 has the effect of disabling gcc's built-in @samp{alloca}, though
2872 @samp{__builtin_alloca} is not affected.
2875 @node Exception Handling
2876 @subsection Exception Handling Support
2877 @cindex exception handling
2880 @findex EH_RETURN_DATA_REGNO
2881 @item EH_RETURN_DATA_REGNO (@var{N})
2882 A C expression whose value is the @var{N}th register number used for
2883 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2884 @var{N} registers are usable.
2886 The exception handling library routines communicate with the exception
2887 handlers via a set of agreed upon registers. Ideally these registers
2888 should be call-clobbered; it is possible to use call-saved registers,
2889 but may negatively impact code size. The target must support at least
2890 2 data registers, but should define 4 if there are enough free registers.
2892 You must define this macro if you want to support call frame exception
2893 handling like that provided by DWARF 2.
2895 @findex EH_RETURN_STACKADJ_RTX
2896 @item EH_RETURN_STACKADJ_RTX
2897 A C expression whose value is RTL representing a location in which
2898 to store a stack adjustment to be applied before function return.
2899 This is used to unwind the stack to an exception handler's call frame.
2900 It will be assigned zero on code paths that return normally.
2902 Typically this is a call-clobbered hard register that is otherwise
2903 untouched by the epilogue, but could also be a stack slot.
2905 You must define this macro if you want to support call frame exception
2906 handling like that provided by DWARF 2.
2908 @findex EH_RETURN_HANDLER_RTX
2909 @item EH_RETURN_HANDLER_RTX
2910 A C expression whose value is RTL representing a location in which
2911 to store the address of an exception handler to which we should
2912 return. It will not be assigned on code paths that return normally.
2914 Typically this is the location in the call frame at which the normal
2915 return address is stored. For targets that return by popping an
2916 address off the stack, this might be a memory address just below
2917 the @emph{target} call frame rather than inside the current call
2918 frame. @code{EH_RETURN_STACKADJ_RTX} will have already been assigned,
2919 so it may be used to calculate the location of the target call frame.
2921 Some targets have more complex requirements than storing to an
2922 address calculable during initial code generation. In that case
2923 the @code{eh_return} instruction pattern should be used instead.
2925 If you want to support call frame exception handling, you must
2926 define either this macro or the @code{eh_return} instruction pattern.
2928 @findex ASM_PREFERRED_EH_DATA_FORMAT
2929 @item ASM_PREFERRED_EH_DATA_FORMAT(@var{code}, @var{global})
2930 This macro chooses the encoding of pointers embedded in the exception
2931 handling sections. If at all possible, this should be defined such
2932 that the exception handling section will not require dynamic relocations,
2933 and so may be read-only.
2935 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
2936 @var{global} is true if the symbol may be affected by dynamic relocations.
2937 The macro should return a combination of the @code{DW_EH_PE_*} defines
2938 as found in @file{dwarf2.h}.
2940 If this macro is not defined, pointers will not be encoded but
2941 represented directly.
2943 @findex ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX
2944 @item ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX(@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
2945 This macro allows the target to emit whatever special magic is required
2946 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
2947 Generic code takes care of pc-relative and indirect encodings; this must
2948 be defined if the target uses text-relative or data-relative encodings.
2950 This is a C statement that branches to @var{done} if the format was
2951 handled. @var{encoding} is the format chosen, @var{size} is the number
2952 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
2955 @findex MD_FALLBACK_FRAME_STATE_FOR
2956 @item MD_FALLBACK_FRAME_STATE_FOR(@var{context}, @var{fs}, @var{success})
2957 This macro allows the target to add cpu and operating system specific
2958 code to the call-frame unwinder for use when there is no unwind data
2959 available. The most common reason to implement this macro is to unwind
2960 through signal frames.
2962 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
2963 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
2964 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
2965 for the address of the code being executed and @code{context->cfa} for
2966 the stack pointer value. If the frame can be decoded, the register save
2967 addresses should be updated in @var{fs} and the macro should branch to
2968 @var{success}. If the frame cannot be decoded, the macro should do
2972 @node Stack Checking
2973 @subsection Specifying How Stack Checking is Done
2975 GCC will check that stack references are within the boundaries of
2976 the stack, if the @option{-fstack-check} is specified, in one of three ways:
2980 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
2981 will assume that you have arranged for stack checking to be done at
2982 appropriate places in the configuration files, e.g., in
2983 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
2987 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
2988 called @code{check_stack} in your @file{md} file, GCC will call that
2989 pattern with one argument which is the address to compare the stack
2990 value against. You must arrange for this pattern to report an error if
2991 the stack pointer is out of range.
2994 If neither of the above are true, GCC will generate code to periodically
2995 ``probe'' the stack pointer using the values of the macros defined below.
2998 Normally, you will use the default values of these macros, so GCC
2999 will use the third approach.
3002 @findex STACK_CHECK_BUILTIN
3003 @item STACK_CHECK_BUILTIN
3004 A nonzero value if stack checking is done by the configuration files in a
3005 machine-dependent manner. You should define this macro if stack checking
3006 is require by the ABI of your machine or if you would like to have to stack
3007 checking in some more efficient way than GCC's portable approach.
3008 The default value of this macro is zero.
3010 @findex STACK_CHECK_PROBE_INTERVAL
3011 @item STACK_CHECK_PROBE_INTERVAL
3012 An integer representing the interval at which GCC must generate stack
3013 probe instructions. You will normally define this macro to be no larger
3014 than the size of the ``guard pages'' at the end of a stack area. The
3015 default value of 4096 is suitable for most systems.
3017 @findex STACK_CHECK_PROBE_LOAD
3018 @item STACK_CHECK_PROBE_LOAD
3019 A integer which is nonzero if GCC should perform the stack probe
3020 as a load instruction and zero if GCC should use a store instruction.
3021 The default is zero, which is the most efficient choice on most systems.
3023 @findex STACK_CHECK_PROTECT
3024 @item STACK_CHECK_PROTECT
3025 The number of bytes of stack needed to recover from a stack overflow,
3026 for languages where such a recovery is supported. The default value of
3027 75 words should be adequate for most machines.
3029 @findex STACK_CHECK_MAX_FRAME_SIZE
3030 @item STACK_CHECK_MAX_FRAME_SIZE
3031 The maximum size of a stack frame, in bytes. GCC will generate probe
3032 instructions in non-leaf functions to ensure at least this many bytes of
3033 stack are available. If a stack frame is larger than this size, stack
3034 checking will not be reliable and GCC will issue a warning. The
3035 default is chosen so that GCC only generates one instruction on most
3036 systems. You should normally not change the default value of this macro.
3038 @findex STACK_CHECK_FIXED_FRAME_SIZE
3039 @item STACK_CHECK_FIXED_FRAME_SIZE
3040 GCC uses this value to generate the above warning message. It
3041 represents the amount of fixed frame used by a function, not including
3042 space for any callee-saved registers, temporaries and user variables.
3043 You need only specify an upper bound for this amount and will normally
3044 use the default of four words.
3046 @findex STACK_CHECK_MAX_VAR_SIZE
3047 @item STACK_CHECK_MAX_VAR_SIZE
3048 The maximum size, in bytes, of an object that GCC will place in the
3049 fixed area of the stack frame when the user specifies
3050 @option{-fstack-check}.
3051 GCC computed the default from the values of the above macros and you will
3052 normally not need to override that default.
3056 @node Frame Registers
3057 @subsection Registers That Address the Stack Frame
3059 @c prevent bad page break with this line
3060 This discusses registers that address the stack frame.
3063 @findex STACK_POINTER_REGNUM
3064 @item STACK_POINTER_REGNUM
3065 The register number of the stack pointer register, which must also be a
3066 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3067 the hardware determines which register this is.
3069 @findex FRAME_POINTER_REGNUM
3070 @item FRAME_POINTER_REGNUM
3071 The register number of the frame pointer register, which is used to
3072 access automatic variables in the stack frame. On some machines, the
3073 hardware determines which register this is. On other machines, you can
3074 choose any register you wish for this purpose.
3076 @findex HARD_FRAME_POINTER_REGNUM
3077 @item HARD_FRAME_POINTER_REGNUM
3078 On some machines the offset between the frame pointer and starting
3079 offset of the automatic variables is not known until after register
3080 allocation has been done (for example, because the saved registers are
3081 between these two locations). On those machines, define
3082 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3083 be used internally until the offset is known, and define
3084 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3085 used for the frame pointer.
3087 You should define this macro only in the very rare circumstances when it
3088 is not possible to calculate the offset between the frame pointer and
3089 the automatic variables until after register allocation has been
3090 completed. When this macro is defined, you must also indicate in your
3091 definition of @code{ELIMINABLE_REGS} how to eliminate
3092 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3093 or @code{STACK_POINTER_REGNUM}.
3095 Do not define this macro if it would be the same as
3096 @code{FRAME_POINTER_REGNUM}.
3098 @findex ARG_POINTER_REGNUM
3099 @item ARG_POINTER_REGNUM
3100 The register number of the arg pointer register, which is used to access
3101 the function's argument list. On some machines, this is the same as the
3102 frame pointer register. On some machines, the hardware determines which
3103 register this is. On other machines, you can choose any register you
3104 wish for this purpose. If this is not the same register as the frame
3105 pointer register, then you must mark it as a fixed register according to
3106 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3107 (@pxref{Elimination}).
3109 @findex RETURN_ADDRESS_POINTER_REGNUM
3110 @item RETURN_ADDRESS_POINTER_REGNUM
3111 The register number of the return address pointer register, which is used to
3112 access the current function's return address from the stack. On some
3113 machines, the return address is not at a fixed offset from the frame
3114 pointer or stack pointer or argument pointer. This register can be defined
3115 to point to the return address on the stack, and then be converted by
3116 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3118 Do not define this macro unless there is no other way to get the return
3119 address from the stack.
3121 @findex STATIC_CHAIN_REGNUM
3122 @findex STATIC_CHAIN_INCOMING_REGNUM
3123 @item STATIC_CHAIN_REGNUM
3124 @itemx STATIC_CHAIN_INCOMING_REGNUM
3125 Register numbers used for passing a function's static chain pointer. If
3126 register windows are used, the register number as seen by the called
3127 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3128 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3129 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3132 The static chain register need not be a fixed register.
3134 If the static chain is passed in memory, these macros should not be
3135 defined; instead, the next two macros should be defined.
3137 @findex STATIC_CHAIN
3138 @findex STATIC_CHAIN_INCOMING
3140 @itemx STATIC_CHAIN_INCOMING
3141 If the static chain is passed in memory, these macros provide rtx giving
3142 @code{mem} expressions that denote where they are stored.
3143 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3144 as seen by the calling and called functions, respectively. Often the former
3145 will be at an offset from the stack pointer and the latter at an offset from
3148 @findex stack_pointer_rtx
3149 @findex frame_pointer_rtx
3150 @findex arg_pointer_rtx
3151 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3152 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3153 macros and should be used to refer to those items.
3155 If the static chain is passed in a register, the two previous macros should
3158 @findex DWARF_FRAME_REGISTERS
3159 @item DWARF_FRAME_REGISTERS
3160 This macro specifies the maximum number of hard registers that can be
3161 saved in a call frame. This is used to size data structures used in
3162 DWARF2 exception handling.
3164 Prior to GCC 3.0, this macro was needed in order to establish a stable
3165 exception handling ABI in the face of adding new hard registers for ISA
3166 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3167 in the number of hard registers. Nevertheless, this macro can still be
3168 used to reduce the runtime memory requirements of the exception handling
3169 routines, which can be substantial if the ISA contains a lot of
3170 registers that are not call-saved.
3172 If this macro is not defined, it defaults to
3173 @code{FIRST_PSEUDO_REGISTER}.
3175 @findex PRE_GCC3_DWARF_FRAME_REGISTERS
3176 @item PRE_GCC3_DWARF_FRAME_REGISTERS
3178 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3179 for backward compatibility in pre GCC 3.0 compiled code.
3181 If this macro is not defined, it defaults to
3182 @code{DWARF_FRAME_REGISTERS}.
3187 @subsection Eliminating Frame Pointer and Arg Pointer
3189 @c prevent bad page break with this line
3190 This is about eliminating the frame pointer and arg pointer.
3193 @findex FRAME_POINTER_REQUIRED
3194 @item FRAME_POINTER_REQUIRED
3195 A C expression which is nonzero if a function must have and use a frame
3196 pointer. This expression is evaluated in the reload pass. If its value is
3197 nonzero the function will have a frame pointer.
3199 The expression can in principle examine the current function and decide
3200 according to the facts, but on most machines the constant 0 or the
3201 constant 1 suffices. Use 0 when the machine allows code to be generated
3202 with no frame pointer, and doing so saves some time or space. Use 1
3203 when there is no possible advantage to avoiding a frame pointer.
3205 In certain cases, the compiler does not know how to produce valid code
3206 without a frame pointer. The compiler recognizes those cases and
3207 automatically gives the function a frame pointer regardless of what
3208 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3211 In a function that does not require a frame pointer, the frame pointer
3212 register can be allocated for ordinary usage, unless you mark it as a
3213 fixed register. See @code{FIXED_REGISTERS} for more information.
3215 @findex INITIAL_FRAME_POINTER_OFFSET
3216 @findex get_frame_size
3217 @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3218 A C statement to store in the variable @var{depth-var} the difference
3219 between the frame pointer and the stack pointer values immediately after
3220 the function prologue. The value would be computed from information
3221 such as the result of @code{get_frame_size ()} and the tables of
3222 registers @code{regs_ever_live} and @code{call_used_regs}.
3224 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3225 need not be defined. Otherwise, it must be defined even if
3226 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3227 case, you may set @var{depth-var} to anything.
3229 @findex ELIMINABLE_REGS
3230 @item ELIMINABLE_REGS
3231 If defined, this macro specifies a table of register pairs used to
3232 eliminate unneeded registers that point into the stack frame. If it is not
3233 defined, the only elimination attempted by the compiler is to replace
3234 references to the frame pointer with references to the stack pointer.
3236 The definition of this macro is a list of structure initializations, each
3237 of which specifies an original and replacement register.
3239 On some machines, the position of the argument pointer is not known until
3240 the compilation is completed. In such a case, a separate hard register
3241 must be used for the argument pointer. This register can be eliminated by
3242 replacing it with either the frame pointer or the argument pointer,
3243 depending on whether or not the frame pointer has been eliminated.
3245 In this case, you might specify:
3247 #define ELIMINABLE_REGS \
3248 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3249 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3250 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3253 Note that the elimination of the argument pointer with the stack pointer is
3254 specified first since that is the preferred elimination.
3256 @findex CAN_ELIMINATE
3257 @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3258 A C expression that returns nonzero if the compiler is allowed to try
3259 to replace register number @var{from-reg} with register number
3260 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3261 is defined, and will usually be the constant 1, since most of the cases
3262 preventing register elimination are things that the compiler already
3265 @findex INITIAL_ELIMINATION_OFFSET
3266 @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3267 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3268 specifies the initial difference between the specified pair of
3269 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3273 @node Stack Arguments
3274 @subsection Passing Function Arguments on the Stack
3275 @cindex arguments on stack
3276 @cindex stack arguments
3278 The macros in this section control how arguments are passed
3279 on the stack. See the following section for other macros that
3280 control passing certain arguments in registers.
3283 @findex PROMOTE_PROTOTYPES
3284 @item PROMOTE_PROTOTYPES
3285 A C expression whose value is nonzero if an argument declared in
3286 a prototype as an integral type smaller than @code{int} should
3287 actually be passed as an @code{int}. In addition to avoiding
3288 errors in certain cases of mismatch, it also makes for better
3289 code on certain machines. If the macro is not defined in target
3290 header files, it defaults to 0.
3294 A C expression. If nonzero, push insns will be used to pass
3296 If the target machine does not have a push instruction, set it to zero.
3297 That directs GCC to use an alternate strategy: to
3298 allocate the entire argument block and then store the arguments into
3299 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3300 On some machines, the definition
3302 @findex PUSH_ROUNDING
3303 @item PUSH_ROUNDING (@var{npushed})
3304 A C expression that is the number of bytes actually pushed onto the
3305 stack when an instruction attempts to push @var{npushed} bytes.
3307 On some machines, the definition
3310 #define PUSH_ROUNDING(BYTES) (BYTES)
3314 will suffice. But on other machines, instructions that appear
3315 to push one byte actually push two bytes in an attempt to maintain
3316 alignment. Then the definition should be
3319 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3322 @findex ACCUMULATE_OUTGOING_ARGS
3323 @findex current_function_outgoing_args_size
3324 @item ACCUMULATE_OUTGOING_ARGS
3325 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3326 will be computed and placed into the variable
3327 @code{current_function_outgoing_args_size}. No space will be pushed
3328 onto the stack for each call; instead, the function prologue should
3329 increase the stack frame size by this amount.
3331 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3334 @findex REG_PARM_STACK_SPACE
3335 @item REG_PARM_STACK_SPACE (@var{fndecl})
3336 Define this macro if functions should assume that stack space has been
3337 allocated for arguments even when their values are passed in
3340 The value of this macro is the size, in bytes, of the area reserved for
3341 arguments passed in registers for the function represented by @var{fndecl},
3342 which can be zero if GCC is calling a library function.
3344 This space can be allocated by the caller, or be a part of the
3345 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3347 @c above is overfull. not sure what to do. --mew 5feb93 did
3348 @c something, not sure if it looks good. --mew 10feb93
3350 @findex MAYBE_REG_PARM_STACK_SPACE
3351 @findex FINAL_REG_PARM_STACK_SPACE
3352 @item MAYBE_REG_PARM_STACK_SPACE
3353 @itemx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
3354 Define these macros in addition to the one above if functions might
3355 allocate stack space for arguments even when their values are passed
3356 in registers. These should be used when the stack space allocated
3357 for arguments in registers is not a simple constant independent of the
3358 function declaration.
3360 The value of the first macro is the size, in bytes, of the area that
3361 we should initially assume would be reserved for arguments passed in registers.
3363 The value of the second macro is the actual size, in bytes, of the area
3364 that will be reserved for arguments passed in registers. This takes two
3365 arguments: an integer representing the number of bytes of fixed sized
3366 arguments on the stack, and a tree representing the number of bytes of
3367 variable sized arguments on the stack.
3369 When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
3370 called for libcall functions, the current function, or for a function
3371 being called when it is known that such stack space must be allocated.
3372 In each case this value can be easily computed.
3374 When deciding whether a called function needs such stack space, and how
3375 much space to reserve, GCC uses these two macros instead of
3376 @code{REG_PARM_STACK_SPACE}.
3378 @findex OUTGOING_REG_PARM_STACK_SPACE
3379 @item OUTGOING_REG_PARM_STACK_SPACE
3380 Define this if it is the responsibility of the caller to allocate the area
3381 reserved for arguments passed in registers.
3383 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3384 whether the space for these arguments counts in the value of
3385 @code{current_function_outgoing_args_size}.
3387 @findex STACK_PARMS_IN_REG_PARM_AREA
3388 @item STACK_PARMS_IN_REG_PARM_AREA
3389 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3390 stack parameters don't skip the area specified by it.
3391 @c i changed this, makes more sens and it should have taken care of the
3392 @c overfull.. not as specific, tho. --mew 5feb93
3394 Normally, when a parameter is not passed in registers, it is placed on the
3395 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3396 suppresses this behavior and causes the parameter to be passed on the
3397 stack in its natural location.
3399 @findex RETURN_POPS_ARGS
3400 @item RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3401 A C expression that should indicate the number of bytes of its own
3402 arguments that a function pops on returning, or 0 if the
3403 function pops no arguments and the caller must therefore pop them all
3404 after the function returns.
3406 @var{fundecl} is a C variable whose value is a tree node that describes
3407 the function in question. Normally it is a node of type
3408 @code{FUNCTION_DECL} that describes the declaration of the function.
3409 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3411 @var{funtype} is a C variable whose value is a tree node that
3412 describes the function in question. Normally it is a node of type
3413 @code{FUNCTION_TYPE} that describes the data type of the function.
3414 From this it is possible to obtain the data types of the value and
3415 arguments (if known).
3417 When a call to a library function is being considered, @var{fundecl}
3418 will contain an identifier node for the library function. Thus, if
3419 you need to distinguish among various library functions, you can do so
3420 by their names. Note that ``library function'' in this context means
3421 a function used to perform arithmetic, whose name is known specially
3422 in the compiler and was not mentioned in the C code being compiled.
3424 @var{stack-size} is the number of bytes of arguments passed on the
3425 stack. If a variable number of bytes is passed, it is zero, and
3426 argument popping will always be the responsibility of the calling function.
3428 On the VAX, all functions always pop their arguments, so the definition
3429 of this macro is @var{stack-size}. On the 68000, using the standard
3430 calling convention, no functions pop their arguments, so the value of
3431 the macro is always 0 in this case. But an alternative calling
3432 convention is available in which functions that take a fixed number of
3433 arguments pop them but other functions (such as @code{printf}) pop
3434 nothing (the caller pops all). When this convention is in use,
3435 @var{funtype} is examined to determine whether a function takes a fixed
3436 number of arguments.
3438 @findex CALL_POPS_ARGS
3439 @item CALL_POPS_ARGS (@var{cum})
3440 A C expression that should indicate the number of bytes a call sequence
3441 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3442 when compiling a function call.
3444 @var{cum} is the variable in which all arguments to the called function
3445 have been accumulated.
3447 On certain architectures, such as the SH5, a call trampoline is used
3448 that pops certain registers off the stack, depending on the arguments
3449 that have been passed to the function. Since this is a property of the
3450 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3455 @node Register Arguments
3456 @subsection Passing Arguments in Registers
3457 @cindex arguments in registers
3458 @cindex registers arguments
3460 This section describes the macros which let you control how various
3461 types of arguments are passed in registers or how they are arranged in
3465 @findex FUNCTION_ARG
3466 @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3467 A C expression that controls whether a function argument is passed
3468 in a register, and which register.
3470 The arguments are @var{cum}, which summarizes all the previous
3471 arguments; @var{mode}, the machine mode of the argument; @var{type},
3472 the data type of the argument as a tree node or 0 if that is not known
3473 (which happens for C support library functions); and @var{named},
3474 which is 1 for an ordinary argument and 0 for nameless arguments that
3475 correspond to @samp{@dots{}} in the called function's prototype.
3476 @var{type} can be an incomplete type if a syntax error has previously
3479 The value of the expression is usually either a @code{reg} RTX for the
3480 hard register in which to pass the argument, or zero to pass the
3481 argument on the stack.
3483 For machines like the VAX and 68000, where normally all arguments are
3484 pushed, zero suffices as a definition.
3486 The value of the expression can also be a @code{parallel} RTX@. This is
3487 used when an argument is passed in multiple locations. The mode of the
3488 of the @code{parallel} should be the mode of the entire argument. The
3489 @code{parallel} holds any number of @code{expr_list} pairs; each one
3490 describes where part of the argument is passed. In each
3491 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3492 register in which to pass this part of the argument, and the mode of the
3493 register RTX indicates how large this part of the argument is. The
3494 second operand of the @code{expr_list} is a @code{const_int} which gives
3495 the offset in bytes into the entire argument of where this part starts.
3496 As a special exception the first @code{expr_list} in the @code{parallel}
3497 RTX may have a first operand of zero. This indicates that the entire
3498 argument is also stored on the stack.
3500 The last time this macro is called, it is called with @code{MODE ==
3501 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3502 pattern as operands 2 and 3 respectively.
3504 @cindex @file{stdarg.h} and register arguments
3505 The usual way to make the ISO library @file{stdarg.h} work on a machine
3506 where some arguments are usually passed in registers, is to cause
3507 nameless arguments to be passed on the stack instead. This is done
3508 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3510 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3511 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3512 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
3513 in the definition of this macro to determine if this argument is of a
3514 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3515 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3516 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3517 defined, the argument will be computed in the stack and then loaded into
3520 @findex MUST_PASS_IN_STACK
3521 @item MUST_PASS_IN_STACK (@var{mode}, @var{type})
3522 Define as a C expression that evaluates to nonzero if we do not know how
3523 to pass TYPE solely in registers. The file @file{expr.h} defines a
3524 definition that is usually appropriate, refer to @file{expr.h} for additional
3527 @findex FUNCTION_INCOMING_ARG
3528 @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3529 Define this macro if the target machine has ``register windows'', so
3530 that the register in which a function sees an arguments is not
3531 necessarily the same as the one in which the caller passed the
3534 For such machines, @code{FUNCTION_ARG} computes the register in which
3535 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3536 be defined in a similar fashion to tell the function being called
3537 where the arguments will arrive.
3539 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3540 serves both purposes.
3542 @findex FUNCTION_ARG_PARTIAL_NREGS
3543 @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
3544 A C expression for the number of words, at the beginning of an
3545 argument, that must be put in registers. The value must be zero for
3546 arguments that are passed entirely in registers or that are entirely
3547 pushed on the stack.
3549 On some machines, certain arguments must be passed partially in
3550 registers and partially in memory. On these machines, typically the
3551 first @var{n} words of arguments are passed in registers, and the rest
3552 on the stack. If a multi-word argument (a @code{double} or a
3553 structure) crosses that boundary, its first few words must be passed
3554 in registers and the rest must be pushed. This macro tells the
3555 compiler when this occurs, and how many of the words should go in
3558 @code{FUNCTION_ARG} for these arguments should return the first
3559 register to be used by the caller for this argument; likewise
3560 @code{FUNCTION_INCOMING_ARG}, for the called function.
3562 @findex FUNCTION_ARG_PASS_BY_REFERENCE
3563 @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3564 A C expression that indicates when an argument must be passed by reference.
3565 If nonzero for an argument, a copy of that argument is made in memory and a
3566 pointer to the argument is passed instead of the argument itself.
3567 The pointer is passed in whatever way is appropriate for passing a pointer
3570 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
3571 definition of this macro might be
3573 #define FUNCTION_ARG_PASS_BY_REFERENCE\
3574 (CUM, MODE, TYPE, NAMED) \
3575 MUST_PASS_IN_STACK (MODE, TYPE)
3577 @c this is *still* too long. --mew 5feb93
3579 @findex FUNCTION_ARG_CALLEE_COPIES
3580 @item FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
3581 If defined, a C expression that indicates when it is the called function's
3582 responsibility to make a copy of arguments passed by invisible reference.
3583 Normally, the caller makes a copy and passes the address of the copy to the
3584 routine being called. When @code{FUNCTION_ARG_CALLEE_COPIES} is defined and is
3585 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
3586 ``live'' value. The called function must not modify this value. If it can be
3587 determined that the value won't be modified, it need not make a copy;
3588 otherwise a copy must be made.
3590 @findex FUNCTION_ARG_REG_LITTLE_ENDIAN
3591 @item FUNCTION_ARG_REG_LITTLE_ENDIAN
3592 If defined TRUE on a big-endian system then structure arguments passed
3593 (and returned) in registers are passed in a little-endian manner instead of
3594 the big-endian manner. On the HP-UX IA64 and PA64 platforms structures are
3595 aligned differently then integral values and setting this value to true will
3596 allow for the special handling of structure arguments and return values.
3598 @findex CUMULATIVE_ARGS
3599 @item CUMULATIVE_ARGS
3600 A C type for declaring a variable that is used as the first argument of
3601 @code{FUNCTION_ARG} and other related values. For some target machines,
3602 the type @code{int} suffices and can hold the number of bytes of
3605 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3606 arguments that have been passed on the stack. The compiler has other
3607 variables to keep track of that. For target machines on which all
3608 arguments are passed on the stack, there is no need to store anything in
3609 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3610 should not be empty, so use @code{int}.
3612 @findex INIT_CUMULATIVE_ARGS
3613 @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{indirect})
3614 A C statement (sans semicolon) for initializing the variable @var{cum}
3615 for the state at the beginning of the argument list. The variable has
3616 type @code{CUMULATIVE_ARGS}. The value of @var{fntype} is the tree node
3617 for the data type of the function which will receive the args, or 0
3618 if the args are to a compiler support library function. The value of
3619 @var{indirect} is nonzero when processing an indirect call, for example
3620 a call through a function pointer. The value of @var{indirect} is zero
3621 for a call to an explicitly named function, a library function call, or when
3622 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3625 When processing a call to a compiler support library function,
3626 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3627 contains the name of the function, as a string. @var{libname} is 0 when
3628 an ordinary C function call is being processed. Thus, each time this
3629 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3630 never both of them at once.
3632 @findex INIT_CUMULATIVE_LIBCALL_ARGS
3633 @item INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3634 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3635 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3636 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3637 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3638 0)} is used instead.
3640 @findex INIT_CUMULATIVE_INCOMING_ARGS
3641 @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3642 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3643 finding the arguments for the function being compiled. If this macro is
3644 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3646 The value passed for @var{libname} is always 0, since library routines
3647 with special calling conventions are never compiled with GCC@. The
3648 argument @var{libname} exists for symmetry with
3649 @code{INIT_CUMULATIVE_ARGS}.
3650 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3651 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3653 @findex FUNCTION_ARG_ADVANCE
3654 @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3655 A C statement (sans semicolon) to update the summarizer variable
3656 @var{cum} to advance past an argument in the argument list. The
3657 values @var{mode}, @var{type} and @var{named} describe that argument.
3658 Once this is done, the variable @var{cum} is suitable for analyzing
3659 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
3661 This macro need not do anything if the argument in question was passed
3662 on the stack. The compiler knows how to track the amount of stack space
3663 used for arguments without any special help.
3665 @findex FUNCTION_ARG_PADDING
3666 @item FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3667 If defined, a C expression which determines whether, and in which direction,
3668 to pad out an argument with extra space. The value should be of type
3669 @code{enum direction}: either @code{upward} to pad above the argument,
3670 @code{downward} to pad below, or @code{none} to inhibit padding.
3672 The @emph{amount} of padding is always just enough to reach the next
3673 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3676 This macro has a default definition which is right for most systems.
3677 For little-endian machines, the default is to pad upward. For
3678 big-endian machines, the default is to pad downward for an argument of
3679 constant size shorter than an @code{int}, and upward otherwise.
3681 @findex PAD_VARARGS_DOWN
3682 @item PAD_VARARGS_DOWN
3683 If defined, a C expression which determines whether the default
3684 implementation of va_arg will attempt to pad down before reading the
3685 next argument, if that argument is smaller than its aligned space as
3686 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3687 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3689 @findex FUNCTION_ARG_BOUNDARY
3690 @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3691 If defined, a C expression that gives the alignment boundary, in bits,
3692 of an argument with the specified mode and type. If it is not defined,
3693 @code{PARM_BOUNDARY} is used for all arguments.
3695 @findex FUNCTION_ARG_REGNO_P
3696 @item FUNCTION_ARG_REGNO_P (@var{regno})
3697 A C expression that is nonzero if @var{regno} is the number of a hard
3698 register in which function arguments are sometimes passed. This does
3699 @emph{not} include implicit arguments such as the static chain and
3700 the structure-value address. On many machines, no registers can be
3701 used for this purpose since all function arguments are pushed on the
3704 @findex LOAD_ARGS_REVERSED
3705 @item LOAD_ARGS_REVERSED
3706 If defined, the order in which arguments are loaded into their
3707 respective argument registers is reversed so that the last
3708 argument is loaded first. This macro only affects arguments
3709 passed in registers.
3714 @subsection How Scalar Function Values Are Returned
3715 @cindex return values in registers
3716 @cindex values, returned by functions
3717 @cindex scalars, returned as values
3719 This section discusses the macros that control returning scalars as
3720 values---values that can fit in registers.
3723 @findex FUNCTION_VALUE
3724 @item FUNCTION_VALUE (@var{valtype}, @var{func})
3725 A C expression to create an RTX representing the place where a
3726 function returns a value of data type @var{valtype}. @var{valtype} is
3727 a tree node representing a data type. Write @code{TYPE_MODE
3728 (@var{valtype})} to get the machine mode used to represent that type.
3729 On many machines, only the mode is relevant. (Actually, on most
3730 machines, scalar values are returned in the same place regardless of
3733 The value of the expression is usually a @code{reg} RTX for the hard
3734 register where the return value is stored. The value can also be a
3735 @code{parallel} RTX, if the return value is in multiple places. See
3736 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3738 If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
3739 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3742 If the precise function being called is known, @var{func} is a tree
3743 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3744 pointer. This makes it possible to use a different value-returning
3745 convention for specific functions when all their calls are
3748 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3749 types, because these are returned in another way. See
3750 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3752 @findex FUNCTION_OUTGOING_VALUE
3753 @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3754 Define this macro if the target machine has ``register windows''
3755 so that the register in which a function returns its value is not
3756 the same as the one in which the caller sees the value.
3758 For such machines, @code{FUNCTION_VALUE} computes the register in which
3759 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3760 defined in a similar fashion to tell the function where to put the
3763 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3764 @code{FUNCTION_VALUE} serves both purposes.
3766 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3767 aggregate data types, because these are returned in another way. See
3768 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3770 @findex LIBCALL_VALUE
3771 @item LIBCALL_VALUE (@var{mode})
3772 A C expression to create an RTX representing the place where a library
3773 function returns a value of mode @var{mode}. If the precise function
3774 being called is known, @var{func} is a tree node
3775 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3776 pointer. This makes it possible to use a different value-returning
3777 convention for specific functions when all their calls are
3780 Note that ``library function'' in this context means a compiler
3781 support routine, used to perform arithmetic, whose name is known
3782 specially by the compiler and was not mentioned in the C code being
3785 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3786 data types, because none of the library functions returns such types.
3788 @findex FUNCTION_VALUE_REGNO_P
3789 @item FUNCTION_VALUE_REGNO_P (@var{regno})
3790 A C expression that is nonzero if @var{regno} is the number of a hard
3791 register in which the values of called function may come back.
3793 A register whose use for returning values is limited to serving as the
3794 second of a pair (for a value of type @code{double}, say) need not be
3795 recognized by this macro. So for most machines, this definition
3799 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3802 If the machine has register windows, so that the caller and the called
3803 function use different registers for the return value, this macro
3804 should recognize only the caller's register numbers.
3806 @findex APPLY_RESULT_SIZE
3807 @item APPLY_RESULT_SIZE
3808 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3809 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3810 saving and restoring an arbitrary return value.
3813 @node Aggregate Return
3814 @subsection How Large Values Are Returned
3815 @cindex aggregates as return values
3816 @cindex large return values
3817 @cindex returning aggregate values
3818 @cindex structure value address
3820 When a function value's mode is @code{BLKmode} (and in some other
3821 cases), the value is not returned according to @code{FUNCTION_VALUE}
3822 (@pxref{Scalar Return}). Instead, the caller passes the address of a
3823 block of memory in which the value should be stored. This address
3824 is called the @dfn{structure value address}.
3826 This section describes how to control returning structure values in
3830 @findex RETURN_IN_MEMORY
3831 @item RETURN_IN_MEMORY (@var{type})
3832 A C expression which can inhibit the returning of certain function
3833 values in registers, based on the type of value. A nonzero value says
3834 to return the function value in memory, just as large structures are
3835 always returned. Here @var{type} will be a C expression of type
3836 @code{tree}, representing the data type of the value.
3838 Note that values of mode @code{BLKmode} must be explicitly handled
3839 by this macro. Also, the option @option{-fpcc-struct-return}
3840 takes effect regardless of this macro. On most systems, it is
3841 possible to leave the macro undefined; this causes a default
3842 definition to be used, whose value is the constant 1 for @code{BLKmode}
3843 values, and 0 otherwise.
3845 Do not use this macro to indicate that structures and unions should always
3846 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
3849 @findex DEFAULT_PCC_STRUCT_RETURN
3850 @item DEFAULT_PCC_STRUCT_RETURN
3851 Define this macro to be 1 if all structure and union return values must be
3852 in memory. Since this results in slower code, this should be defined
3853 only if needed for compatibility with other compilers or with an ABI@.
3854 If you define this macro to be 0, then the conventions used for structure
3855 and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
3857 If not defined, this defaults to the value 1.
3859 @findex STRUCT_VALUE_REGNUM
3860 @item STRUCT_VALUE_REGNUM
3861 If the structure value address is passed in a register, then
3862 @code{STRUCT_VALUE_REGNUM} should be the number of that register.
3864 @findex STRUCT_VALUE
3866 If the structure value address is not passed in a register, define
3867 @code{STRUCT_VALUE} as an expression returning an RTX for the place
3868 where the address is passed. If it returns 0, the address is passed as
3869 an ``invisible'' first argument.
3871 @findex STRUCT_VALUE_INCOMING_REGNUM
3872 @item STRUCT_VALUE_INCOMING_REGNUM
3873 On some architectures the place where the structure value address
3874 is found by the called function is not the same place that the
3875 caller put it. This can be due to register windows, or it could
3876 be because the function prologue moves it to a different place.
3878 If the incoming location of the structure value address is in a
3879 register, define this macro as the register number.
3881 @findex STRUCT_VALUE_INCOMING
3882 @item STRUCT_VALUE_INCOMING
3883 If the incoming location is not a register, then you should define
3884 @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the
3885 called function should find the value. If it should find the value on
3886 the stack, define this to create a @code{mem} which refers to the frame
3887 pointer. A definition of 0 means that the address is passed as an
3888 ``invisible'' first argument.
3890 @findex PCC_STATIC_STRUCT_RETURN
3891 @item PCC_STATIC_STRUCT_RETURN
3892 Define this macro if the usual system convention on the target machine
3893 for returning structures and unions is for the called function to return
3894 the address of a static variable containing the value.
3896 Do not define this if the usual system convention is for the caller to
3897 pass an address to the subroutine.
3899 This macro has effect in @option{-fpcc-struct-return} mode, but it does
3900 nothing when you use @option{-freg-struct-return} mode.
3904 @subsection Caller-Saves Register Allocation
3906 If you enable it, GCC can save registers around function calls. This
3907 makes it possible to use call-clobbered registers to hold variables that
3908 must live across calls.
3911 @findex DEFAULT_CALLER_SAVES
3912 @item DEFAULT_CALLER_SAVES
3913 Define this macro if function calls on the target machine do not preserve
3914 any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
3915 for all registers. When defined, this macro enables @option{-fcaller-saves}
3916 by default for all optimization levels. It has no effect for optimization
3917 levels 2 and higher, where @option{-fcaller-saves} is the default.
3919 @findex CALLER_SAVE_PROFITABLE
3920 @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
3921 A C expression to determine whether it is worthwhile to consider placing
3922 a pseudo-register in a call-clobbered hard register and saving and
3923 restoring it around each function call. The expression should be 1 when
3924 this is worth doing, and 0 otherwise.
3926 If you don't define this macro, a default is used which is good on most
3927 machines: @code{4 * @var{calls} < @var{refs}}.
3929 @findex HARD_REGNO_CALLER_SAVE_MODE
3930 @item HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3931 A C expression specifying which mode is required for saving @var{nregs}
3932 of a pseudo-register in call-clobbered hard register @var{regno}. If
3933 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3934 returned. For most machines this macro need not be defined since GCC
3935 will select the smallest suitable mode.
3938 @node Function Entry
3939 @subsection Function Entry and Exit
3940 @cindex function entry and exit
3944 This section describes the macros that output function entry
3945 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3947 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
3948 If defined, a function that outputs the assembler code for entry to a
3949 function. The prologue is responsible for setting up the stack frame,
3950 initializing the frame pointer register, saving registers that must be
3951 saved, and allocating @var{size} additional bytes of storage for the
3952 local variables. @var{size} is an integer. @var{file} is a stdio
3953 stream to which the assembler code should be output.
3955 The label for the beginning of the function need not be output by this
3956 macro. That has already been done when the macro is run.
3958 @findex regs_ever_live
3959 To determine which registers to save, the macro can refer to the array
3960 @code{regs_ever_live}: element @var{r} is nonzero if hard register
3961 @var{r} is used anywhere within the function. This implies the function
3962 prologue should save register @var{r}, provided it is not one of the
3963 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
3964 @code{regs_ever_live}.)
3966 On machines that have ``register windows'', the function entry code does
3967 not save on the stack the registers that are in the windows, even if
3968 they are supposed to be preserved by function calls; instead it takes
3969 appropriate steps to ``push'' the register stack, if any non-call-used
3970 registers are used in the function.
3972 @findex frame_pointer_needed
3973 On machines where functions may or may not have frame-pointers, the
3974 function entry code must vary accordingly; it must set up the frame
3975 pointer if one is wanted, and not otherwise. To determine whether a
3976 frame pointer is in wanted, the macro can refer to the variable
3977 @code{frame_pointer_needed}. The variable's value will be 1 at run
3978 time in a function that needs a frame pointer. @xref{Elimination}.
3980 The function entry code is responsible for allocating any stack space
3981 required for the function. This stack space consists of the regions
3982 listed below. In most cases, these regions are allocated in the
3983 order listed, with the last listed region closest to the top of the
3984 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
3985 the highest address if it is not defined). You can use a different order
3986 for a machine if doing so is more convenient or required for
3987 compatibility reasons. Except in cases where required by standard
3988 or by a debugger, there is no reason why the stack layout used by GCC
3989 need agree with that used by other compilers for a machine.
3992 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
3993 If defined, a function that outputs assembler code at the end of a
3994 prologue. This should be used when the function prologue is being
3995 emitted as RTL, and you have some extra assembler that needs to be
3996 emitted. @xref{prologue instruction pattern}.
3999 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4000 If defined, a function that outputs assembler code at the start of an
4001 epilogue. This should be used when the function epilogue is being
4002 emitted as RTL, and you have some extra assembler that needs to be
4003 emitted. @xref{epilogue instruction pattern}.
4006 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4007 If defined, a function that outputs the assembler code for exit from a
4008 function. The epilogue is responsible for restoring the saved
4009 registers and stack pointer to their values when the function was
4010 called, and returning control to the caller. This macro takes the
4011 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4012 registers to restore are determined from @code{regs_ever_live} and
4013 @code{CALL_USED_REGISTERS} in the same way.
4015 On some machines, there is a single instruction that does all the work
4016 of returning from the function. On these machines, give that
4017 instruction the name @samp{return} and do not define the macro
4018 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4020 Do not define a pattern named @samp{return} if you want the
4021 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4022 switches to control whether return instructions or epilogues are used,
4023 define a @samp{return} pattern with a validity condition that tests the
4024 target switches appropriately. If the @samp{return} pattern's validity
4025 condition is false, epilogues will be used.
4027 On machines where functions may or may not have frame-pointers, the
4028 function exit code must vary accordingly. Sometimes the code for these
4029 two cases is completely different. To determine whether a frame pointer
4030 is wanted, the macro can refer to the variable
4031 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4032 a function that needs a frame pointer.
4034 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4035 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4036 The C variable @code{current_function_is_leaf} is nonzero for such a
4037 function. @xref{Leaf Functions}.
4039 On some machines, some functions pop their arguments on exit while
4040 others leave that for the caller to do. For example, the 68020 when
4041 given @option{-mrtd} pops arguments in functions that take a fixed
4042 number of arguments.
4044 @findex current_function_pops_args
4045 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4046 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4047 needs to know what was decided. The variable that is called
4048 @code{current_function_pops_args} is the number of bytes of its
4049 arguments that a function should pop. @xref{Scalar Return}.
4050 @c what is the "its arguments" in the above sentence referring to, pray
4051 @c tell? --mew 5feb93
4058 @findex current_function_pretend_args_size
4059 A region of @code{current_function_pretend_args_size} bytes of
4060 uninitialized space just underneath the first argument arriving on the
4061 stack. (This may not be at the very start of the allocated stack region
4062 if the calling sequence has pushed anything else since pushing the stack
4063 arguments. But usually, on such machines, nothing else has been pushed
4064 yet, because the function prologue itself does all the pushing.) This
4065 region is used on machines where an argument may be passed partly in
4066 registers and partly in memory, and, in some cases to support the
4067 features in @code{<stdarg.h>}.
4070 An area of memory used to save certain registers used by the function.
4071 The size of this area, which may also include space for such things as
4072 the return address and pointers to previous stack frames, is
4073 machine-specific and usually depends on which registers have been used
4074 in the function. Machines with register windows often do not require
4078 A region of at least @var{size} bytes, possibly rounded up to an allocation
4079 boundary, to contain the local variables of the function. On some machines,
4080 this region and the save area may occur in the opposite order, with the
4081 save area closer to the top of the stack.
4084 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4085 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4086 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4087 argument lists of the function. @xref{Stack Arguments}.
4090 Normally, it is necessary for the macros
4091 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4092 @code{TARGET_ASM_FUNCTION_EPILOGUE} to treat leaf functions specially.
4093 The C variable @code{current_function_is_leaf} is nonzero for such a
4096 @findex EXIT_IGNORE_STACK
4097 @item EXIT_IGNORE_STACK
4098 Define this macro as a C expression that is nonzero if the return
4099 instruction or the function epilogue ignores the value of the stack
4100 pointer; in other words, if it is safe to delete an instruction to
4101 adjust the stack pointer before a return from the function.
4103 Note that this macro's value is relevant only for functions for which
4104 frame pointers are maintained. It is never safe to delete a final
4105 stack adjustment in a function that has no frame pointer, and the
4106 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4108 @findex EPILOGUE_USES
4109 @item EPILOGUE_USES (@var{regno})
4110 Define this macro as a C expression that is nonzero for registers that are
4111 used by the epilogue or the @samp{return} pattern. The stack and frame
4112 pointer registers are already be assumed to be used as needed.
4115 @item EH_USES (@var{regno})
4116 Define this macro as a C expression that is nonzero for registers that are
4117 used by the exception handling mechanism, and so should be considered live
4118 on entry to an exception edge.
4120 @findex DELAY_SLOTS_FOR_EPILOGUE
4121 @item DELAY_SLOTS_FOR_EPILOGUE
4122 Define this macro if the function epilogue contains delay slots to which
4123 instructions from the rest of the function can be ``moved''. The
4124 definition should be a C expression whose value is an integer
4125 representing the number of delay slots there.
4127 @findex ELIGIBLE_FOR_EPILOGUE_DELAY
4128 @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4129 A C expression that returns 1 if @var{insn} can be placed in delay
4130 slot number @var{n} of the epilogue.
4132 The argument @var{n} is an integer which identifies the delay slot now
4133 being considered (since different slots may have different rules of
4134 eligibility). It is never negative and is always less than the number
4135 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4136 If you reject a particular insn for a given delay slot, in principle, it
4137 may be reconsidered for a subsequent delay slot. Also, other insns may
4138 (at least in principle) be considered for the so far unfilled delay
4141 @findex current_function_epilogue_delay_list
4142 @findex final_scan_insn
4143 The insns accepted to fill the epilogue delay slots are put in an RTL
4144 list made with @code{insn_list} objects, stored in the variable
4145 @code{current_function_epilogue_delay_list}. The insn for the first
4146 delay slot comes first in the list. Your definition of the macro
4147 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4148 outputting the insns in this list, usually by calling
4149 @code{final_scan_insn}.
4151 You need not define this macro if you did not define
4152 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4154 @findex ASM_OUTPUT_MI_THUNK
4155 @item ASM_OUTPUT_MI_THUNK (@var{file}, @var{thunk_fndecl}, @var{delta}, @var{function})
4156 A C compound statement that outputs the assembler code for a thunk
4157 function, used to implement C++ virtual function calls with multiple
4158 inheritance. The thunk acts as a wrapper around a virtual function,
4159 adjusting the implicit object parameter before handing control off to
4162 First, emit code to add the integer @var{delta} to the location that
4163 contains the incoming first argument. Assume that this argument
4164 contains a pointer, and is the one used to pass the @code{this} pointer
4165 in C++. This is the incoming argument @emph{before} the function prologue,
4166 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4167 all other incoming arguments.
4169 After the addition, emit code to jump to @var{function}, which is a
4170 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4171 not touch the return address. Hence returning from @var{FUNCTION} will
4172 return to whoever called the current @samp{thunk}.
4174 The effect must be as if @var{function} had been called directly with
4175 the adjusted first argument. This macro is responsible for emitting all
4176 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4177 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4179 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4180 have already been extracted from it.) It might possibly be useful on
4181 some targets, but probably not.
4183 If you do not define this macro, the target-independent code in the C++
4184 front end will generate a less efficient heavyweight thunk that calls
4185 @var{function} instead of jumping to it. The generic approach does
4186 not support varargs.
4190 @subsection Generating Code for Profiling
4191 @cindex profiling, code generation
4193 These macros will help you generate code for profiling.
4196 @findex FUNCTION_PROFILER
4197 @item FUNCTION_PROFILER (@var{file}, @var{labelno})
4198 A C statement or compound statement to output to @var{file} some
4199 assembler code to call the profiling subroutine @code{mcount}.
4202 The details of how @code{mcount} expects to be called are determined by
4203 your operating system environment, not by GCC@. To figure them out,
4204 compile a small program for profiling using the system's installed C
4205 compiler and look at the assembler code that results.
4207 Older implementations of @code{mcount} expect the address of a counter
4208 variable to be loaded into some register. The name of this variable is
4209 @samp{LP} followed by the number @var{labelno}, so you would generate
4210 the name using @samp{LP%d} in a @code{fprintf}.
4212 @findex PROFILE_HOOK
4214 A C statement or compound statement to output to @var{file} some assembly
4215 code to call the profiling subroutine @code{mcount} even the target does
4216 not support profiling.
4218 @findex NO_PROFILE_COUNTERS
4219 @item NO_PROFILE_COUNTERS
4220 Define this macro if the @code{mcount} subroutine on your system does
4221 not need a counter variable allocated for each function. This is true
4222 for almost all modern implementations. If you define this macro, you
4223 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
4225 @findex PROFILE_BEFORE_PROLOGUE
4226 @item PROFILE_BEFORE_PROLOGUE
4227 Define this macro if the code for function profiling should come before
4228 the function prologue. Normally, the profiling code comes after.
4232 @subsection Permitting tail calls
4236 @findex FUNCTION_OK_FOR_SIBCALL
4237 @item FUNCTION_OK_FOR_SIBCALL (@var{decl})
4238 A C expression that evaluates to true if it is ok to perform a sibling
4239 call to @var{decl} from the current function.
4241 It is not uncommon for limitations of calling conventions to prevent
4242 tail calls to functions outside the current unit of translation, or
4243 during PIC compilation. Use this macro to enforce these restrictions,
4244 as the @code{sibcall} md pattern can not fail, or fall over to a
4249 @section Implementing the Varargs Macros
4250 @cindex varargs implementation
4252 GCC comes with an implementation of @code{<varargs.h>} and
4253 @code{<stdarg.h>} that work without change on machines that pass arguments
4254 on the stack. Other machines require their own implementations of
4255 varargs, and the two machine independent header files must have
4256 conditionals to include it.
4258 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4259 the calling convention for @code{va_start}. The traditional
4260 implementation takes just one argument, which is the variable in which
4261 to store the argument pointer. The ISO implementation of
4262 @code{va_start} takes an additional second argument. The user is
4263 supposed to write the last named argument of the function here.
4265 However, @code{va_start} should not use this argument. The way to find
4266 the end of the named arguments is with the built-in functions described
4270 @findex __builtin_saveregs
4271 @item __builtin_saveregs ()
4272 Use this built-in function to save the argument registers in memory so
4273 that the varargs mechanism can access them. Both ISO and traditional
4274 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4275 you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
4277 On some machines, @code{__builtin_saveregs} is open-coded under the
4278 control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
4279 it calls a routine written in assembler language, found in
4282 Code generated for the call to @code{__builtin_saveregs} appears at the
4283 beginning of the function, as opposed to where the call to
4284 @code{__builtin_saveregs} is written, regardless of what the code is.
4285 This is because the registers must be saved before the function starts
4286 to use them for its own purposes.
4287 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4290 @findex __builtin_args_info
4291 @item __builtin_args_info (@var{category})
4292 Use this built-in function to find the first anonymous arguments in
4295 In general, a machine may have several categories of registers used for
4296 arguments, each for a particular category of data types. (For example,
4297 on some machines, floating-point registers are used for floating-point
4298 arguments while other arguments are passed in the general registers.)
4299 To make non-varargs functions use the proper calling convention, you
4300 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4301 registers in each category have been used so far
4303 @code{__builtin_args_info} accesses the same data structure of type
4304 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4305 with it, with @var{category} specifying which word to access. Thus, the
4306 value indicates the first unused register in a given category.
4308 Normally, you would use @code{__builtin_args_info} in the implementation
4309 of @code{va_start}, accessing each category just once and storing the
4310 value in the @code{va_list} object. This is because @code{va_list} will
4311 have to update the values, and there is no way to alter the
4312 values accessed by @code{__builtin_args_info}.
4314 @findex __builtin_next_arg
4315 @item __builtin_next_arg (@var{lastarg})
4316 This is the equivalent of @code{__builtin_args_info}, for stack
4317 arguments. It returns the address of the first anonymous stack
4318 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4319 returns the address of the location above the first anonymous stack
4320 argument. Use it in @code{va_start} to initialize the pointer for
4321 fetching arguments from the stack. Also use it in @code{va_start} to
4322 verify that the second parameter @var{lastarg} is the last named argument
4323 of the current function.
4325 @findex __builtin_classify_type
4326 @item __builtin_classify_type (@var{object})
4327 Since each machine has its own conventions for which data types are
4328 passed in which kind of register, your implementation of @code{va_arg}
4329 has to embody these conventions. The easiest way to categorize the
4330 specified data type is to use @code{__builtin_classify_type} together
4331 with @code{sizeof} and @code{__alignof__}.
4333 @code{__builtin_classify_type} ignores the value of @var{object},
4334 considering only its data type. It returns an integer describing what
4335 kind of type that is---integer, floating, pointer, structure, and so on.
4337 The file @file{typeclass.h} defines an enumeration that you can use to
4338 interpret the values of @code{__builtin_classify_type}.
4341 These machine description macros help implement varargs:
4344 @findex EXPAND_BUILTIN_SAVEREGS
4345 @item EXPAND_BUILTIN_SAVEREGS ()
4346 If defined, is a C expression that produces the machine-specific code
4347 for a call to @code{__builtin_saveregs}. This code will be moved to the
4348 very beginning of the function, before any parameter access are made.
4349 The return value of this function should be an RTX that contains the
4350 value to use as the return of @code{__builtin_saveregs}.
4352 @findex SETUP_INCOMING_VARARGS
4353 @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type}, @var{pretend_args_size}, @var{second_time})
4354 This macro offers an alternative to using @code{__builtin_saveregs} and
4355 defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the
4356 anonymous register arguments into the stack so that all the arguments
4357 appear to have been passed consecutively on the stack. Once this is
4358 done, you can use the standard implementation of varargs that works for
4359 machines that pass all their arguments on the stack.
4361 The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data
4362 structure, containing the values that are obtained after processing the
4363 named arguments. The arguments @var{mode} and @var{type} describe the
4364 last named argument---its machine mode and its data type as a tree node.
4366 The macro implementation should do two things: first, push onto the
4367 stack all the argument registers @emph{not} used for the named
4368 arguments, and second, store the size of the data thus pushed into the
4369 @code{int}-valued variable whose name is supplied as the argument
4370 @var{pretend_args_size}. The value that you store here will serve as
4371 additional offset for setting up the stack frame.
4373 Because you must generate code to push the anonymous arguments at
4374 compile time without knowing their data types,
4375 @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just
4376 a single category of argument register and use it uniformly for all data
4379 If the argument @var{second_time} is nonzero, it means that the
4380 arguments of the function are being analyzed for the second time. This
4381 happens for an inline function, which is not actually compiled until the
4382 end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should
4383 not generate any instructions in this case.
4385 @findex STRICT_ARGUMENT_NAMING
4386 @item STRICT_ARGUMENT_NAMING
4387 Define this macro to be a nonzero value if the location where a function
4388 argument is passed depends on whether or not it is a named argument.
4390 This macro controls how the @var{named} argument to @code{FUNCTION_ARG}
4391 is set for varargs and stdarg functions. If this macro returns a
4392 nonzero value, the @var{named} argument is always true for named
4393 arguments, and false for unnamed arguments. If it returns a value of
4394 zero, but @code{SETUP_INCOMING_VARARGS} is defined, then all arguments
4395 are treated as named. Otherwise, all named arguments except the last
4396 are treated as named.
4398 You need not define this macro if it always returns zero.
4400 @findex PRETEND_OUTGOING_VARARGS_NAMED
4401 @item PRETEND_OUTGOING_VARARGS_NAMED
4402 If you need to conditionally change ABIs so that one works with
4403 @code{SETUP_INCOMING_VARARGS}, but the other works like neither
4404 @code{SETUP_INCOMING_VARARGS} nor @code{STRICT_ARGUMENT_NAMING} was
4405 defined, then define this macro to return nonzero if
4406 @code{SETUP_INCOMING_VARARGS} is used, zero otherwise.
4407 Otherwise, you should not define this macro.
4411 @section Trampolines for Nested Functions
4412 @cindex trampolines for nested functions
4413 @cindex nested functions, trampolines for
4415 A @dfn{trampoline} is a small piece of code that is created at run time
4416 when the address of a nested function is taken. It normally resides on
4417 the stack, in the stack frame of the containing function. These macros
4418 tell GCC how to generate code to allocate and initialize a
4421 The instructions in the trampoline must do two things: load a constant
4422 address into the static chain register, and jump to the real address of
4423 the nested function. On CISC machines such as the m68k, this requires
4424 two instructions, a move immediate and a jump. Then the two addresses
4425 exist in the trampoline as word-long immediate operands. On RISC
4426 machines, it is often necessary to load each address into a register in
4427 two parts. Then pieces of each address form separate immediate
4430 The code generated to initialize the trampoline must store the variable
4431 parts---the static chain value and the function address---into the
4432 immediate operands of the instructions. On a CISC machine, this is
4433 simply a matter of copying each address to a memory reference at the
4434 proper offset from the start of the trampoline. On a RISC machine, it
4435 may be necessary to take out pieces of the address and store them
4439 @findex TRAMPOLINE_TEMPLATE
4440 @item TRAMPOLINE_TEMPLATE (@var{file})
4441 A C statement to output, on the stream @var{file}, assembler code for a
4442 block of data that contains the constant parts of a trampoline. This
4443 code should not include a label---the label is taken care of
4446 If you do not define this macro, it means no template is needed
4447 for the target. Do not define this macro on systems where the block move
4448 code to copy the trampoline into place would be larger than the code
4449 to generate it on the spot.
4451 @findex TRAMPOLINE_SECTION
4452 @item TRAMPOLINE_SECTION
4453 The name of a subroutine to switch to the section in which the
4454 trampoline template is to be placed (@pxref{Sections}). The default is
4455 a value of @samp{readonly_data_section}, which places the trampoline in
4456 the section containing read-only data.
4458 @findex TRAMPOLINE_SIZE
4459 @item TRAMPOLINE_SIZE
4460 A C expression for the size in bytes of the trampoline, as an integer.
4462 @findex TRAMPOLINE_ALIGNMENT
4463 @item TRAMPOLINE_ALIGNMENT
4464 Alignment required for trampolines, in bits.
4466 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4467 is used for aligning trampolines.
4469 @findex INITIALIZE_TRAMPOLINE
4470 @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4471 A C statement to initialize the variable parts of a trampoline.
4472 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4473 an RTX for the address of the nested function; @var{static_chain} is an
4474 RTX for the static chain value that should be passed to the function
4477 @findex TRAMPOLINE_ADJUST_ADDRESS
4478 @item TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4479 A C statement that should perform any machine-specific adjustment in
4480 the address of the trampoline. Its argument contains the address that
4481 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4482 used for a function call should be different from the address in which
4483 the template was stored, the different address should be assigned to
4484 @var{addr}. If this macro is not defined, @var{addr} will be used for
4487 @findex ALLOCATE_TRAMPOLINE
4488 @item ALLOCATE_TRAMPOLINE (@var{fp})
4489 A C expression to allocate run-time space for a trampoline. The
4490 expression value should be an RTX representing a memory reference to the
4491 space for the trampoline.
4493 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4494 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4495 If this macro is not defined, by default the trampoline is allocated as
4496 a stack slot. This default is right for most machines. The exceptions
4497 are machines where it is impossible to execute instructions in the stack
4498 area. On such machines, you may have to implement a separate stack,
4499 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4500 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4502 @var{fp} points to a data structure, a @code{struct function}, which
4503 describes the compilation status of the immediate containing function of
4504 the function which the trampoline is for. Normally (when
4505 @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the
4506 trampoline is in the stack frame of this containing function. Other
4507 allocation strategies probably must do something analogous with this
4511 Implementing trampolines is difficult on many machines because they have
4512 separate instruction and data caches. Writing into a stack location
4513 fails to clear the memory in the instruction cache, so when the program
4514 jumps to that location, it executes the old contents.
4516 Here are two possible solutions. One is to clear the relevant parts of
4517 the instruction cache whenever a trampoline is set up. The other is to
4518 make all trampolines identical, by having them jump to a standard
4519 subroutine. The former technique makes trampoline execution faster; the
4520 latter makes initialization faster.
4522 To clear the instruction cache when a trampoline is initialized, define
4523 the following macros which describe the shape of the cache.
4526 @findex INSN_CACHE_SIZE
4527 @item INSN_CACHE_SIZE
4528 The total size in bytes of the cache.
4530 @findex INSN_CACHE_LINE_WIDTH
4531 @item INSN_CACHE_LINE_WIDTH
4532 The length in bytes of each cache line. The cache is divided into cache
4533 lines which are disjoint slots, each holding a contiguous chunk of data
4534 fetched from memory. Each time data is brought into the cache, an
4535 entire line is read at once. The data loaded into a cache line is
4536 always aligned on a boundary equal to the line size.
4538 @findex INSN_CACHE_DEPTH
4539 @item INSN_CACHE_DEPTH
4540 The number of alternative cache lines that can hold any particular memory
4544 Alternatively, if the machine has system calls or instructions to clear
4545 the instruction cache directly, you can define the following macro.
4548 @findex CLEAR_INSN_CACHE
4549 @item CLEAR_INSN_CACHE (@var{beg}, @var{end})
4550 If defined, expands to a C expression clearing the @emph{instruction
4551 cache} in the specified interval. If it is not defined, and the macro
4552 @code{INSN_CACHE_SIZE} is defined, some generic code is generated to clear the
4553 cache. The definition of this macro would typically be a series of
4554 @code{asm} statements. Both @var{beg} and @var{end} are both pointer
4558 To use a standard subroutine, define the following macro. In addition,
4559 you must make sure that the instructions in a trampoline fill an entire
4560 cache line with identical instructions, or else ensure that the
4561 beginning of the trampoline code is always aligned at the same point in
4562 its cache line. Look in @file{m68k.h} as a guide.
4565 @findex TRANSFER_FROM_TRAMPOLINE
4566 @item TRANSFER_FROM_TRAMPOLINE
4567 Define this macro if trampolines need a special subroutine to do their
4568 work. The macro should expand to a series of @code{asm} statements
4569 which will be compiled with GCC@. They go in a library function named
4570 @code{__transfer_from_trampoline}.
4572 If you need to avoid executing the ordinary prologue code of a compiled
4573 C function when you jump to the subroutine, you can do so by placing a
4574 special label of your own in the assembler code. Use one @code{asm}
4575 statement to generate an assembler label, and another to make the label
4576 global. Then trampolines can use that label to jump directly to your
4577 special assembler code.
4581 @section Implicit Calls to Library Routines
4582 @cindex library subroutine names
4583 @cindex @file{libgcc.a}
4585 @c prevent bad page break with this line
4586 Here is an explanation of implicit calls to library routines.
4589 @findex MULSI3_LIBCALL
4590 @item MULSI3_LIBCALL
4591 A C string constant giving the name of the function to call for
4592 multiplication of one signed full-word by another. If you do not
4593 define this macro, the default name is used, which is @code{__mulsi3},
4594 a function defined in @file{libgcc.a}.
4596 @findex DIVSI3_LIBCALL
4597 @item DIVSI3_LIBCALL
4598 A C string constant giving the name of the function to call for
4599 division of one signed full-word by another. If you do not define
4600 this macro, the default name is used, which is @code{__divsi3}, a
4601 function defined in @file{libgcc.a}.
4603 @findex UDIVSI3_LIBCALL
4604 @item UDIVSI3_LIBCALL
4605 A C string constant giving the name of the function to call for
4606 division of one unsigned full-word by another. If you do not define
4607 this macro, the default name is used, which is @code{__udivsi3}, a
4608 function defined in @file{libgcc.a}.
4610 @findex MODSI3_LIBCALL
4611 @item MODSI3_LIBCALL
4612 A C string constant giving the name of the function to call for the
4613 remainder in division of one signed full-word by another. If you do
4614 not define this macro, the default name is used, which is
4615 @code{__modsi3}, a function defined in @file{libgcc.a}.
4617 @findex UMODSI3_LIBCALL
4618 @item UMODSI3_LIBCALL
4619 A C string constant giving the name of the function to call for the
4620 remainder in division of one unsigned full-word by another. If you do
4621 not define this macro, the default name is used, which is
4622 @code{__umodsi3}, a function defined in @file{libgcc.a}.
4624 @findex MULDI3_LIBCALL
4625 @item MULDI3_LIBCALL
4626 A C string constant giving the name of the function to call for
4627 multiplication of one signed double-word by another. If you do not
4628 define this macro, the default name is used, which is @code{__muldi3},
4629 a function defined in @file{libgcc.a}.
4631 @findex DIVDI3_LIBCALL
4632 @item DIVDI3_LIBCALL
4633 A C string constant giving the name of the function to call for
4634 division of one signed double-word by another. If you do not define
4635 this macro, the default name is used, which is @code{__divdi3}, a
4636 function defined in @file{libgcc.a}.
4638 @findex UDIVDI3_LIBCALL
4639 @item UDIVDI3_LIBCALL
4640 A C string constant giving the name of the function to call for
4641 division of one unsigned full-word by another. If you do not define
4642 this macro, the default name is used, which is @code{__udivdi3}, a
4643 function defined in @file{libgcc.a}.
4645 @findex MODDI3_LIBCALL
4646 @item MODDI3_LIBCALL
4647 A C string constant giving the name of the function to call for the
4648 remainder in division of one signed double-word by another. If you do
4649 not define this macro, the default name is used, which is
4650 @code{__moddi3}, a function defined in @file{libgcc.a}.
4652 @findex UMODDI3_LIBCALL
4653 @item UMODDI3_LIBCALL
4654 A C string constant giving the name of the function to call for the
4655 remainder in division of one unsigned full-word by another. If you do
4656 not define this macro, the default name is used, which is
4657 @code{__umoddi3}, a function defined in @file{libgcc.a}.
4659 @findex DECLARE_LIBRARY_RENAMES
4660 @item DECLARE_LIBRARY_RENAMES
4661 This macro, if defined, should expand to a piece of C code that will get
4662 expanded when compiling functions for libgcc.a. It can be used to
4663 provide alternate names for gcc's internal library functions if there
4664 are ABI-mandated names that the compiler should provide.
4666 @findex INIT_TARGET_OPTABS
4667 @item INIT_TARGET_OPTABS
4668 Define this macro as a C statement that declares additional library
4669 routines renames existing ones. @code{init_optabs} calls this macro after
4670 initializing all the normal library routines.
4672 @findex FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4673 @item FLOAT_LIB_COMPARE_RETURNS_BOOL
4674 Define this macro as a C statement that returns nonzero if a call to
4675 the floating point comparison library function will return a boolean
4676 value that indicates the result of the comparison. It should return
4677 zero if one of gcc's own libgcc functions is called.
4679 Most ports don't need to define this macro.
4682 @cindex @code{EDOM}, implicit usage
4684 The value of @code{EDOM} on the target machine, as a C integer constant
4685 expression. If you don't define this macro, GCC does not attempt to
4686 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4687 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4690 If you do not define @code{TARGET_EDOM}, then compiled code reports
4691 domain errors by calling the library function and letting it report the
4692 error. If mathematical functions on your system use @code{matherr} when
4693 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4694 that @code{matherr} is used normally.
4696 @findex GEN_ERRNO_RTX
4697 @cindex @code{errno}, implicit usage
4699 Define this macro as a C expression to create an rtl expression that
4700 refers to the global ``variable'' @code{errno}. (On certain systems,
4701 @code{errno} may not actually be a variable.) If you don't define this
4702 macro, a reasonable default is used.
4704 @findex TARGET_MEM_FUNCTIONS
4705 @cindex @code{bcopy}, implicit usage
4706 @cindex @code{memcpy}, implicit usage
4707 @cindex @code{memmove}, implicit usage
4708 @cindex @code{bzero}, implicit usage
4709 @cindex @code{memset}, implicit usage
4710 @item TARGET_MEM_FUNCTIONS
4711 Define this macro if GCC should generate calls to the ISO C
4712 (and System V) library functions @code{memcpy}, @code{memmove} and
4713 @code{memset} rather than the BSD functions @code{bcopy} and @code{bzero}.
4715 @findex LIBGCC_NEEDS_DOUBLE
4716 @item LIBGCC_NEEDS_DOUBLE
4717 Define this macro if @code{float} arguments cannot be passed to library
4718 routines (so they must be converted to @code{double}). This macro
4719 affects both how library calls are generated and how the library
4720 routines in @file{libgcc.a} accept their arguments. It is useful on
4721 machines where floating and fixed point arguments are passed
4722 differently, such as the i860.
4724 @findex NEXT_OBJC_RUNTIME
4725 @item NEXT_OBJC_RUNTIME
4726 Define this macro to generate code for Objective-C message sending using
4727 the calling convention of the NeXT system. This calling convention
4728 involves passing the object, the selector and the method arguments all
4729 at once to the method-lookup library function.
4731 The default calling convention passes just the object and the selector
4732 to the lookup function, which returns a pointer to the method.
4735 @node Addressing Modes
4736 @section Addressing Modes
4737 @cindex addressing modes
4739 @c prevent bad page break with this line
4740 This is about addressing modes.
4743 @findex HAVE_PRE_INCREMENT
4744 @findex HAVE_PRE_DECREMENT
4745 @findex HAVE_POST_INCREMENT
4746 @findex HAVE_POST_DECREMENT
4747 @item HAVE_PRE_INCREMENT
4748 @itemx HAVE_PRE_DECREMENT
4749 @itemx HAVE_POST_INCREMENT
4750 @itemx HAVE_POST_DECREMENT
4751 A C expression that is nonzero if the machine supports pre-increment,
4752 pre-decrement, post-increment, or post-decrement addressing respectively.
4754 @findex HAVE_POST_MODIFY_DISP
4755 @findex HAVE_PRE_MODIFY_DISP
4756 @item HAVE_PRE_MODIFY_DISP
4757 @itemx HAVE_POST_MODIFY_DISP
4758 A C expression that is nonzero if the machine supports pre- or
4759 post-address side-effect generation involving constants other than
4760 the size of the memory operand.
4762 @findex HAVE_POST_MODIFY_REG
4763 @findex HAVE_PRE_MODIFY_REG
4764 @item HAVE_PRE_MODIFY_REG
4765 @itemx HAVE_POST_MODIFY_REG
4766 A C expression that is nonzero if the machine supports pre- or
4767 post-address side-effect generation involving a register displacement.
4769 @findex CONSTANT_ADDRESS_P
4770 @item CONSTANT_ADDRESS_P (@var{x})
4771 A C expression that is 1 if the RTX @var{x} is a constant which
4772 is a valid address. On most machines, this can be defined as
4773 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4774 in which constant addresses are supported.
4777 @code{CONSTANT_P} accepts integer-values expressions whose values are
4778 not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and
4779 @code{high} expressions and @code{const} arithmetic expressions, in
4780 addition to @code{const_int} and @code{const_double} expressions.
4782 @findex MAX_REGS_PER_ADDRESS
4783 @item MAX_REGS_PER_ADDRESS
4784 A number, the maximum number of registers that can appear in a valid
4785 memory address. Note that it is up to you to specify a value equal to
4786 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4789 @findex GO_IF_LEGITIMATE_ADDRESS
4790 @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4791 A C compound statement with a conditional @code{goto @var{label};}
4792 executed if @var{x} (an RTX) is a legitimate memory address on the
4793 target machine for a memory operand of mode @var{mode}.
4795 It usually pays to define several simpler macros to serve as
4796 subroutines for this one. Otherwise it may be too complicated to
4799 This macro must exist in two variants: a strict variant and a
4800 non-strict one. The strict variant is used in the reload pass. It
4801 must be defined so that any pseudo-register that has not been
4802 allocated a hard register is considered a memory reference. In
4803 contexts where some kind of register is required, a pseudo-register
4804 with no hard register must be rejected.
4806 The non-strict variant is used in other passes. It must be defined to
4807 accept all pseudo-registers in every context where some kind of
4808 register is required.
4810 @findex REG_OK_STRICT
4811 Compiler source files that want to use the strict variant of this
4812 macro define the macro @code{REG_OK_STRICT}. You should use an
4813 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4814 in that case and the non-strict variant otherwise.
4816 Subroutines to check for acceptable registers for various purposes (one
4817 for base registers, one for index registers, and so on) are typically
4818 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4819 Then only these subroutine macros need have two variants; the higher
4820 levels of macros may be the same whether strict or not.
4822 Normally, constant addresses which are the sum of a @code{symbol_ref}
4823 and an integer are stored inside a @code{const} RTX to mark them as
4824 constant. Therefore, there is no need to recognize such sums
4825 specifically as legitimate addresses. Normally you would simply
4826 recognize any @code{const} as legitimate.
4828 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4829 sums that are not marked with @code{const}. It assumes that a naked
4830 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4831 naked constant sums as illegitimate addresses, so that none of them will
4832 be given to @code{PRINT_OPERAND_ADDRESS}.
4834 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
4835 On some machines, whether a symbolic address is legitimate depends on
4836 the section that the address refers to. On these machines, define the
4837 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
4838 into the @code{symbol_ref}, and then check for it here. When you see a
4839 @code{const}, you will have to look inside it to find the
4840 @code{symbol_ref} in order to determine the section. @xref{Assembler
4843 @findex saveable_obstack
4844 The best way to modify the name string is by adding text to the
4845 beginning, with suitable punctuation to prevent any ambiguity. Allocate
4846 the new name in @code{saveable_obstack}. You will have to modify
4847 @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and
4848 output the name accordingly, and define @code{TARGET_STRIP_NAME_ENCODING}
4849 to access the original name string.
4851 You can check the information stored here into the @code{symbol_ref} in
4852 the definitions of the macros @code{GO_IF_LEGITIMATE_ADDRESS} and
4853 @code{PRINT_OPERAND_ADDRESS}.
4855 @findex REG_OK_FOR_BASE_P
4856 @item REG_OK_FOR_BASE_P (@var{x})
4857 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4858 RTX) is valid for use as a base register. For hard registers, it
4859 should always accept those which the hardware permits and reject the
4860 others. Whether the macro accepts or rejects pseudo registers must be
4861 controlled by @code{REG_OK_STRICT} as described above. This usually
4862 requires two variant definitions, of which @code{REG_OK_STRICT}
4863 controls the one actually used.
4865 @findex REG_MODE_OK_FOR_BASE_P
4866 @item REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4867 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4868 that expression may examine the mode of the memory reference in
4869 @var{mode}. You should define this macro if the mode of the memory
4870 reference affects whether a register may be used as a base register. If
4871 you define this macro, the compiler will use it instead of
4872 @code{REG_OK_FOR_BASE_P}.
4874 @findex REG_OK_FOR_INDEX_P
4875 @item REG_OK_FOR_INDEX_P (@var{x})
4876 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4877 RTX) is valid for use as an index register.
4879 The difference between an index register and a base register is that
4880 the index register may be scaled. If an address involves the sum of
4881 two registers, neither one of them scaled, then either one may be
4882 labeled the ``base'' and the other the ``index''; but whichever
4883 labeling is used must fit the machine's constraints of which registers
4884 may serve in each capacity. The compiler will try both labelings,
4885 looking for one that is valid, and will reload one or both registers
4886 only if neither labeling works.
4888 @findex FIND_BASE_TERM
4889 @item FIND_BASE_TERM (@var{x})
4890 A C expression to determine the base term of address @var{x}.
4891 This macro is used in only one place: `find_base_term' in alias.c.
4893 It is always safe for this macro to not be defined. It exists so
4894 that alias analysis can understand machine-dependent addresses.
4896 The typical use of this macro is to handle addresses containing
4897 a label_ref or symbol_ref within an UNSPEC@.
4899 @findex LEGITIMIZE_ADDRESS
4900 @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4901 A C compound statement that attempts to replace @var{x} with a valid
4902 memory address for an operand of mode @var{mode}. @var{win} will be a
4903 C statement label elsewhere in the code; the macro definition may use
4906 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4910 to avoid further processing if the address has become legitimate.
4912 @findex break_out_memory_refs
4913 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4914 and @var{oldx} will be the operand that was given to that function to produce
4917 The code generated by this macro should not alter the substructure of
4918 @var{x}. If it transforms @var{x} into a more legitimate form, it
4919 should assign @var{x} (which will always be a C variable) a new value.
4921 It is not necessary for this macro to come up with a legitimate
4922 address. The compiler has standard ways of doing so in all cases. In
4923 fact, it is safe for this macro to do nothing. But often a
4924 machine-dependent strategy can generate better code.
4926 @findex LEGITIMIZE_RELOAD_ADDRESS
4927 @item LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4928 A C compound statement that attempts to replace @var{x}, which is an address
4929 that needs reloading, with a valid memory address for an operand of mode
4930 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4931 It is not necessary to define this macro, but it might be useful for
4932 performance reasons.
4934 For example, on the i386, it is sometimes possible to use a single
4935 reload register instead of two by reloading a sum of two pseudo
4936 registers into a register. On the other hand, for number of RISC
4937 processors offsets are limited so that often an intermediate address
4938 needs to be generated in order to address a stack slot. By defining
4939 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
4940 generated for adjacent some stack slots can be made identical, and thus
4943 @emph{Note}: This macro should be used with caution. It is necessary
4944 to know something of how reload works in order to effectively use this,
4945 and it is quite easy to produce macros that build in too much knowledge
4946 of reload internals.
4948 @emph{Note}: This macro must be able to reload an address created by a
4949 previous invocation of this macro. If it fails to handle such addresses
4950 then the compiler may generate incorrect code or abort.
4953 The macro definition should use @code{push_reload} to indicate parts that
4954 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4955 suitable to be passed unaltered to @code{push_reload}.
4957 The code generated by this macro must not alter the substructure of
4958 @var{x}. If it transforms @var{x} into a more legitimate form, it
4959 should assign @var{x} (which will always be a C variable) a new value.
4960 This also applies to parts that you change indirectly by calling
4963 @findex strict_memory_address_p
4964 The macro definition may use @code{strict_memory_address_p} to test if
4965 the address has become legitimate.
4968 If you want to change only a part of @var{x}, one standard way of doing
4969 this is to use @code{copy_rtx}. Note, however, that is unshares only a
4970 single level of rtl. Thus, if the part to be changed is not at the
4971 top level, you'll need to replace first the top level.
4972 It is not necessary for this macro to come up with a legitimate
4973 address; but often a machine-dependent strategy can generate better code.
4975 @findex GO_IF_MODE_DEPENDENT_ADDRESS
4976 @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
4977 A C statement or compound statement with a conditional @code{goto
4978 @var{label};} executed if memory address @var{x} (an RTX) can have
4979 different meanings depending on the machine mode of the memory
4980 reference it is used for or if the address is valid for some modes
4983 Autoincrement and autodecrement addresses typically have mode-dependent
4984 effects because the amount of the increment or decrement is the size
4985 of the operand being addressed. Some machines have other mode-dependent
4986 addresses. Many RISC machines have no mode-dependent addresses.
4988 You may assume that @var{addr} is a valid address for the machine.
4990 @findex LEGITIMATE_CONSTANT_P
4991 @item LEGITIMATE_CONSTANT_P (@var{x})
4992 A C expression that is nonzero if @var{x} is a legitimate constant for
4993 an immediate operand on the target machine. You can assume that
4994 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
4995 @samp{1} is a suitable definition for this macro on machines where
4996 anything @code{CONSTANT_P} is valid.
4999 @node Condition Code
5000 @section Condition Code Status
5001 @cindex condition code status
5003 @c prevent bad page break with this line
5004 This describes the condition code status.
5007 The file @file{conditions.h} defines a variable @code{cc_status} to
5008 describe how the condition code was computed (in case the interpretation of
5009 the condition code depends on the instruction that it was set by). This
5010 variable contains the RTL expressions on which the condition code is
5011 currently based, and several standard flags.
5013 Sometimes additional machine-specific flags must be defined in the machine
5014 description header file. It can also add additional machine-specific
5015 information by defining @code{CC_STATUS_MDEP}.
5018 @findex CC_STATUS_MDEP
5019 @item CC_STATUS_MDEP
5020 C code for a data type which is used for declaring the @code{mdep}
5021 component of @code{cc_status}. It defaults to @code{int}.
5023 This macro is not used on machines that do not use @code{cc0}.
5025 @findex CC_STATUS_MDEP_INIT
5026 @item CC_STATUS_MDEP_INIT
5027 A C expression to initialize the @code{mdep} field to ``empty''.
5028 The default definition does nothing, since most machines don't use
5029 the field anyway. If you want to use the field, you should probably
5030 define this macro to initialize it.
5032 This macro is not used on machines that do not use @code{cc0}.
5034 @findex NOTICE_UPDATE_CC
5035 @item NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5036 A C compound statement to set the components of @code{cc_status}
5037 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5038 this macro's responsibility to recognize insns that set the condition
5039 code as a byproduct of other activity as well as those that explicitly
5042 This macro is not used on machines that do not use @code{cc0}.
5044 If there are insns that do not set the condition code but do alter
5045 other machine registers, this macro must check to see whether they
5046 invalidate the expressions that the condition code is recorded as
5047 reflecting. For example, on the 68000, insns that store in address
5048 registers do not set the condition code, which means that usually
5049 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5050 insns. But suppose that the previous insn set the condition code
5051 based on location @samp{a4@@(102)} and the current insn stores a new
5052 value in @samp{a4}. Although the condition code is not changed by
5053 this, it will no longer be true that it reflects the contents of
5054 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5055 @code{cc_status} in this case to say that nothing is known about the
5056 condition code value.
5058 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5059 with the results of peephole optimization: insns whose patterns are
5060 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5061 constants which are just the operands. The RTL structure of these
5062 insns is not sufficient to indicate what the insns actually do. What
5063 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5064 @code{CC_STATUS_INIT}.
5066 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5067 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5068 @samp{cc}. This avoids having detailed information about patterns in
5069 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5071 @findex EXTRA_CC_MODES
5072 @item EXTRA_CC_MODES
5073 Condition codes are represented in registers by machine modes of class
5074 @code{MODE_CC}. By default, there is just one mode, @code{CCmode}, with
5075 this class. If you need more such modes, create a file named
5076 @file{@var{machine}-modes.def} in your @file{config/@var{machine}}
5077 directory (@pxref{Back End, , Anatomy of a Target Back End}), containing
5078 a list of these modes. Each entry in the list should be a call to the
5079 macro @code{CC}. This macro takes one argument, which is the name of
5080 the mode: it should begin with @samp{CC}. Do not put quotation marks
5081 around the name, or include the trailing @samp{mode}; these are
5082 automatically added. There should not be anything else in the file
5085 A sample @file{@var{machine}-modes.def} file might look like this:
5088 CC (CC_NOOV) /* @r{Comparison only valid if there was no overflow.} */
5089 CC (CCFP) /* @r{Floating point comparison that cannot trap.} */
5090 CC (CCFPE) /* @r{Floating point comparison that may trap.} */
5093 When you create this file, the macro @code{EXTRA_CC_MODES} is
5094 automatically defined by @command{configure}, with value @samp{1}.
5096 @findex SELECT_CC_MODE
5097 @item SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5098 Returns a mode from class @code{MODE_CC} to be used when comparison
5099 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5100 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5101 @pxref{Jump Patterns} for a description of the reason for this
5105 #define SELECT_CC_MODE(OP,X,Y) \
5106 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5107 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5108 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5109 || GET_CODE (X) == NEG) \
5110 ? CC_NOOVmode : CCmode))
5113 You need not define this macro if @code{EXTRA_CC_MODES} is not defined.
5115 @findex CANONICALIZE_COMPARISON
5116 @item CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5117 On some machines not all possible comparisons are defined, but you can
5118 convert an invalid comparison into a valid one. For example, the Alpha
5119 does not have a @code{GT} comparison, but you can use an @code{LT}
5120 comparison instead and swap the order of the operands.
5122 On such machines, define this macro to be a C statement to do any
5123 required conversions. @var{code} is the initial comparison code
5124 and @var{op0} and @var{op1} are the left and right operands of the
5125 comparison, respectively. You should modify @var{code}, @var{op0}, and
5126 @var{op1} as required.
5128 GCC will not assume that the comparison resulting from this macro is
5129 valid but will see if the resulting insn matches a pattern in the
5132 You need not define this macro if it would never change the comparison
5135 @findex REVERSIBLE_CC_MODE
5136 @item REVERSIBLE_CC_MODE (@var{mode})
5137 A C expression whose value is one if it is always safe to reverse a
5138 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5139 can ever return @var{mode} for a floating-point inequality comparison,
5140 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5142 You need not define this macro if it would always returns zero or if the
5143 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5144 For example, here is the definition used on the SPARC, where floating-point
5145 inequality comparisons are always given @code{CCFPEmode}:
5148 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5151 @findex REVERSE_CONDITION (@var{code}, @var{mode})
5152 A C expression whose value is reversed condition code of the @var{code} for
5153 comparison done in CC_MODE @var{mode}. The macro is used only in case
5154 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5155 machine has some non-standard way how to reverse certain conditionals. For
5156 instance in case all floating point conditions are non-trapping, compiler may
5157 freely convert unordered compares to ordered one. Then definition may look
5161 #define REVERSE_CONDITION(CODE, MODE) \
5162 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5163 : reverse_condition_maybe_unordered (CODE))
5166 @findex REVERSE_CONDEXEC_PREDICATES_P
5167 @item REVERSE_CONDEXEC_PREDICATES_P (@var{code1}, @var{code2})
5168 A C expression that returns true if the conditional execution predicate
5169 @var{code1} is the inverse of @var{code2} and vice versa. Define this to
5170 return 0 if the target has conditional execution predicates that cannot be
5171 reversed safely. If no expansion is specified, this macro is defined as
5175 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5176 ((x) == reverse_condition (y))
5182 @section Describing Relative Costs of Operations
5183 @cindex costs of instructions
5184 @cindex relative costs
5185 @cindex speed of instructions
5187 These macros let you describe the relative speed of various operations
5188 on the target machine.
5192 @item CONST_COSTS (@var{x}, @var{code}, @var{outer_code})
5193 A part of a C @code{switch} statement that describes the relative costs
5194 of constant RTL expressions. It must contain @code{case} labels for
5195 expression codes @code{const_int}, @code{const}, @code{symbol_ref},
5196 @code{label_ref} and @code{const_double}. Each case must ultimately
5197 reach a @code{return} statement to return the relative cost of the use
5198 of that kind of constant value in an expression. The cost may depend on
5199 the precise value of the constant, which is available for examination in
5200 @var{x}, and the rtx code of the expression in which it is contained,
5201 found in @var{outer_code}.
5203 @var{code} is the expression code---redundant, since it can be
5204 obtained with @code{GET_CODE (@var{x})}.
5207 @findex COSTS_N_INSNS
5208 @item RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
5209 Like @code{CONST_COSTS} but applies to nonconstant RTL expressions.
5210 This can be used, for example, to indicate how costly a multiply
5211 instruction is. In writing this macro, you can use the construct
5212 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
5213 instructions. @var{outer_code} is the code of the expression in which
5214 @var{x} is contained.
5216 This macro is optional; do not define it if the default cost assumptions
5217 are adequate for the target machine.
5219 @findex DEFAULT_RTX_COSTS
5220 @item DEFAULT_RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
5221 This macro, if defined, is called for any case not handled by the
5222 @code{RTX_COSTS} or @code{CONST_COSTS} macros. This eliminates the need
5223 to put case labels into the macro, but the code, or any functions it
5224 calls, must assume that the RTL in @var{x} could be of any type that has
5225 not already been handled. The arguments are the same as for
5226 @code{RTX_COSTS}, and the macro should execute a return statement giving
5227 the cost of any RTL expressions that it can handle. The default cost
5228 calculation is used for any RTL for which this macro does not return a
5231 This macro is optional; do not define it if the default cost assumptions
5232 are adequate for the target machine.
5234 @findex ADDRESS_COST
5235 @item ADDRESS_COST (@var{address})
5236 An expression giving the cost of an addressing mode that contains
5237 @var{address}. If not defined, the cost is computed from
5238 the @var{address} expression and the @code{CONST_COSTS} values.
5240 For most CISC machines, the default cost is a good approximation of the
5241 true cost of the addressing mode. However, on RISC machines, all
5242 instructions normally have the same length and execution time. Hence
5243 all addresses will have equal costs.
5245 In cases where more than one form of an address is known, the form with
5246 the lowest cost will be used. If multiple forms have the same, lowest,
5247 cost, the one that is the most complex will be used.
5249 For example, suppose an address that is equal to the sum of a register
5250 and a constant is used twice in the same basic block. When this macro
5251 is not defined, the address will be computed in a register and memory
5252 references will be indirect through that register. On machines where
5253 the cost of the addressing mode containing the sum is no higher than
5254 that of a simple indirect reference, this will produce an additional
5255 instruction and possibly require an additional register. Proper
5256 specification of this macro eliminates this overhead for such machines.
5258 Similar use of this macro is made in strength reduction of loops.
5260 @var{address} need not be valid as an address. In such a case, the cost
5261 is not relevant and can be any value; invalid addresses need not be
5262 assigned a different cost.
5264 On machines where an address involving more than one register is as
5265 cheap as an address computation involving only one register, defining
5266 @code{ADDRESS_COST} to reflect this can cause two registers to be live
5267 over a region of code where only one would have been if
5268 @code{ADDRESS_COST} were not defined in that manner. This effect should
5269 be considered in the definition of this macro. Equivalent costs should
5270 probably only be given to addresses with different numbers of registers
5271 on machines with lots of registers.
5273 This macro will normally either not be defined or be defined as a
5276 @findex REGISTER_MOVE_COST
5277 @item REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5278 A C expression for the cost of moving data of mode @var{mode} from a
5279 register in class @var{from} to one in class @var{to}. The classes are
5280 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5281 value of 2 is the default; other values are interpreted relative to
5284 It is not required that the cost always equal 2 when @var{from} is the
5285 same as @var{to}; on some machines it is expensive to move between
5286 registers if they are not general registers.
5288 If reload sees an insn consisting of a single @code{set} between two
5289 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5290 classes returns a value of 2, reload does not check to ensure that the
5291 constraints of the insn are met. Setting a cost of other than 2 will
5292 allow reload to verify that the constraints are met. You should do this
5293 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5295 @findex MEMORY_MOVE_COST
5296 @item MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5297 A C expression for the cost of moving data of mode @var{mode} between a
5298 register of class @var{class} and memory; @var{in} is zero if the value
5299 is to be written to memory, nonzero if it is to be read in. This cost
5300 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5301 registers and memory is more expensive than between two registers, you
5302 should define this macro to express the relative cost.
5304 If you do not define this macro, GCC uses a default cost of 4 plus
5305 the cost of copying via a secondary reload register, if one is
5306 needed. If your machine requires a secondary reload register to copy
5307 between memory and a register of @var{class} but the reload mechanism is
5308 more complex than copying via an intermediate, define this macro to
5309 reflect the actual cost of the move.
5311 GCC defines the function @code{memory_move_secondary_cost} if
5312 secondary reloads are needed. It computes the costs due to copying via
5313 a secondary register. If your machine copies from memory using a
5314 secondary register in the conventional way but the default base value of
5315 4 is not correct for your machine, define this macro to add some other
5316 value to the result of that function. The arguments to that function
5317 are the same as to this macro.
5321 A C expression for the cost of a branch instruction. A value of 1 is
5322 the default; other values are interpreted relative to that.
5325 Here are additional macros which do not specify precise relative costs,
5326 but only that certain actions are more expensive than GCC would
5330 @findex SLOW_BYTE_ACCESS
5331 @item SLOW_BYTE_ACCESS
5332 Define this macro as a C expression which is nonzero if accessing less
5333 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5334 faster than accessing a word of memory, i.e., if such access
5335 require more than one instruction or if there is no difference in cost
5336 between byte and (aligned) word loads.
5338 When this macro is not defined, the compiler will access a field by
5339 finding the smallest containing object; when it is defined, a fullword
5340 load will be used if alignment permits. Unless bytes accesses are
5341 faster than word accesses, using word accesses is preferable since it
5342 may eliminate subsequent memory access if subsequent accesses occur to
5343 other fields in the same word of the structure, but to different bytes.
5345 @findex SLOW_UNALIGNED_ACCESS
5346 @item SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5347 Define this macro to be the value 1 if memory accesses described by the
5348 @var{mode} and @var{alignment} parameters have a cost many times greater
5349 than aligned accesses, for example if they are emulated in a trap
5352 When this macro is nonzero, the compiler will act as if
5353 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5354 moves. This can cause significantly more instructions to be produced.
5355 Therefore, do not set this macro nonzero if unaligned accesses only add a
5356 cycle or two to the time for a memory access.
5358 If the value of this macro is always zero, it need not be defined. If
5359 this macro is defined, it should produce a nonzero value when
5360 @code{STRICT_ALIGNMENT} is nonzero.
5362 @findex DONT_REDUCE_ADDR
5363 @item DONT_REDUCE_ADDR
5364 Define this macro to inhibit strength reduction of memory addresses.
5365 (On some machines, such strength reduction seems to do harm rather
5370 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5371 which a sequence of insns should be generated instead of a
5372 string move insn or a library call. Increasing the value will always
5373 make code faster, but eventually incurs high cost in increased code size.
5375 Note that on machines where the corresponding move insn is a
5376 @code{define_expand} that emits a sequence of insns, this macro counts
5377 the number of such sequences.
5379 If you don't define this, a reasonable default is used.
5381 @findex MOVE_BY_PIECES_P
5382 @item MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5383 A C expression used to determine whether @code{move_by_pieces} will be used to
5384 copy a chunk of memory, or whether some other block move mechanism
5385 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5386 than @code{MOVE_RATIO}.
5388 @findex MOVE_MAX_PIECES
5389 @item MOVE_MAX_PIECES
5390 A C expression used by @code{move_by_pieces} to determine the largest unit
5391 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5395 The threshold of number of scalar move insns, @emph{below} which a sequence
5396 of insns should be generated to clear memory instead of a string clear insn
5397 or a library call. Increasing the value will always make code faster, but
5398 eventually incurs high cost in increased code size.
5400 If you don't define this, a reasonable default is used.
5402 @findex CLEAR_BY_PIECES_P
5403 @item CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5404 A C expression used to determine whether @code{clear_by_pieces} will be used
5405 to clear a chunk of memory, or whether some other block clear mechanism
5406 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5407 than @code{CLEAR_RATIO}.
5409 @findex USE_LOAD_POST_INCREMENT
5410 @item USE_LOAD_POST_INCREMENT (@var{mode})
5411 A C expression used to determine whether a load postincrement is a good
5412 thing to use for a given mode. Defaults to the value of
5413 @code{HAVE_POST_INCREMENT}.
5415 @findex USE_LOAD_POST_DECREMENT
5416 @item USE_LOAD_POST_DECREMENT (@var{mode})
5417 A C expression used to determine whether a load postdecrement is a good
5418 thing to use for a given mode. Defaults to the value of
5419 @code{HAVE_POST_DECREMENT}.
5421 @findex USE_LOAD_PRE_INCREMENT
5422 @item USE_LOAD_PRE_INCREMENT (@var{mode})
5423 A C expression used to determine whether a load preincrement is a good
5424 thing to use for a given mode. Defaults to the value of
5425 @code{HAVE_PRE_INCREMENT}.
5427 @findex USE_LOAD_PRE_DECREMENT
5428 @item USE_LOAD_PRE_DECREMENT (@var{mode})
5429 A C expression used to determine whether a load predecrement is a good
5430 thing to use for a given mode. Defaults to the value of
5431 @code{HAVE_PRE_DECREMENT}.
5433 @findex USE_STORE_POST_INCREMENT
5434 @item USE_STORE_POST_INCREMENT (@var{mode})
5435 A C expression used to determine whether a store postincrement is a good
5436 thing to use for a given mode. Defaults to the value of
5437 @code{HAVE_POST_INCREMENT}.
5439 @findex USE_STORE_POST_DECREMENT
5440 @item USE_STORE_POST_DECREMENT (@var{mode})
5441 A C expression used to determine whether a store postdecrement is a good
5442 thing to use for a given mode. Defaults to the value of
5443 @code{HAVE_POST_DECREMENT}.
5445 @findex USE_STORE_PRE_INCREMENT
5446 @item USE_STORE_PRE_INCREMENT (@var{mode})
5447 This macro is used to determine whether a store preincrement is a good
5448 thing to use for a given mode. Defaults to the value of
5449 @code{HAVE_PRE_INCREMENT}.
5451 @findex USE_STORE_PRE_DECREMENT
5452 @item USE_STORE_PRE_DECREMENT (@var{mode})
5453 This macro is used to determine whether a store predecrement is a good
5454 thing to use for a given mode. Defaults to the value of
5455 @code{HAVE_PRE_DECREMENT}.
5457 @findex NO_FUNCTION_CSE
5458 @item NO_FUNCTION_CSE
5459 Define this macro if it is as good or better to call a constant
5460 function address than to call an address kept in a register.
5462 @findex NO_RECURSIVE_FUNCTION_CSE
5463 @item NO_RECURSIVE_FUNCTION_CSE
5464 Define this macro if it is as good or better for a function to call
5465 itself with an explicit address than to call an address kept in a
5470 @section Adjusting the Instruction Scheduler
5472 The instruction scheduler may need a fair amount of machine-specific
5473 adjustment in order to produce good code. GCC provides several target
5474 hooks for this purpose. It is usually enough to define just a few of
5475 them: try the first ones in this list first.
5477 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
5478 This hook returns the maximum number of instructions that can ever
5479 issue at the same time on the target machine. The default is one.
5480 Although the insn scheduler can define itself the possibility of issue
5481 an insn on the same cycle, the value can serve as an additional
5482 constraint to issue insns on the same simulated processor cycle (see
5483 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
5484 This value must be constant over the entire compilation. If you need
5485 it to vary depending on what the instructions are, you must use
5486 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
5488 For the automaton based pipeline interface, you could define this hook
5489 to return the value of the macro @code{MAX_DFA_ISSUE_RATE}.
5492 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
5493 This hook is executed by the scheduler after it has scheduled an insn
5494 from the ready list. It should return the number of insns which can
5495 still be issued in the current cycle. Normally this is
5496 @samp{@w{@var{more} - 1}}. You should define this hook if some insns
5497 take more machine resources than others, so that fewer insns can follow
5498 them in the same cycle. @var{file} is either a null pointer, or a stdio
5499 stream to write any debug output to. @var{verbose} is the verbose level
5500 provided by @option{-fsched-verbose-@var{n}}. @var{insn} is the
5501 instruction that was scheduled.
5504 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
5505 This function corrects the value of @var{cost} based on the
5506 relationship between @var{insn} and @var{dep_insn} through the
5507 dependence @var{link}. It should return the new value. The default
5508 is to make no adjustment to @var{cost}. This can be used for example
5509 to specify to the scheduler using the traditional pipeline description
5510 that an output- or anti-dependence does not incur the same cost as a
5511 data-dependence. If the scheduler using the automaton based pipeline
5512 description, the cost of anti-dependence is zero and the cost of
5513 output-dependence is maximum of one and the difference of latency
5514 times of the first and the second insns. If these values are not
5515 acceptable, you could use the hook to modify them too. See also
5516 @pxref{Automaton pipeline description}.
5519 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
5520 This hook adjusts the integer scheduling priority @var{priority} of
5521 @var{insn}. It should return the new priority. Reduce the priority to
5522 execute @var{insn} earlier, increase the priority to execute @var{insn}
5523 later. Do not define this hook if you do not need to adjust the
5524 scheduling priorities of insns.
5527 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
5528 This hook is executed by the scheduler after it has scheduled the ready
5529 list, to allow the machine description to reorder it (for example to
5530 combine two small instructions together on @samp{VLIW} machines).
5531 @var{file} is either a null pointer, or a stdio stream to write any
5532 debug output to. @var{verbose} is the verbose level provided by
5533 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
5534 list of instructions that are ready to be scheduled. @var{n_readyp} is
5535 a pointer to the number of elements in the ready list. The scheduler
5536 reads the ready list in reverse order, starting with
5537 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
5538 is the timer tick of the scheduler. You may modify the ready list and
5539 the number of ready insns. The return value is the number of insns that
5540 can issue this cycle; normally this is just @code{issue_rate}. See also
5541 @samp{TARGET_SCHED_REORDER2}.
5544 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
5545 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
5546 function is called whenever the scheduler starts a new cycle. This one
5547 is called once per iteration over a cycle, immediately after
5548 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
5549 return the number of insns to be scheduled in the same cycle. Defining
5550 this hook can be useful if there are frequent situations where
5551 scheduling one insn causes other insns to become ready in the same
5552 cycle. These other insns can then be taken into account properly.
5555 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
5556 This hook is executed by the scheduler at the beginning of each block of
5557 instructions that are to be scheduled. @var{file} is either a null
5558 pointer, or a stdio stream to write any debug output to. @var{verbose}
5559 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5560 @var{max_ready} is the maximum number of insns in the current scheduling
5561 region that can be live at the same time. This can be used to allocate
5562 scratch space if it is needed, e.g. by @samp{TARGET_SCHED_REORDER}.
5565 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
5566 This hook is executed by the scheduler at the end of each block of
5567 instructions that are to be scheduled. It can be used to perform
5568 cleanup of any actions done by the other scheduling hooks. @var{file}
5569 is either a null pointer, or a stdio stream to write any debug output
5570 to. @var{verbose} is the verbose level provided by
5571 @option{-fsched-verbose-@var{n}}.
5574 @deftypefn {Target Hook} int TARGET_SCHED_USE_DFA_PIPELINE_INTERFACE (void)
5575 This hook is called many times during insn scheduling. If the hook
5576 returns nonzero, the automaton based pipeline description is used for
5577 insn scheduling. Otherwise the traditional pipeline description is
5578 used. The default is usage of the traditional pipeline description.
5580 You should also remember that to simplify the insn scheduler sources
5581 an empty traditional pipeline description interface is generated even
5582 if there is no a traditional pipeline description in the @file{.md}
5583 file. The same is true for the automaton based pipeline description.
5584 That means that you should be accurate in defining the hook.
5587 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
5588 The hook returns an RTL insn. The automaton state used in the
5589 pipeline hazard recognizer is changed as if the insn were scheduled
5590 when the new simulated processor cycle starts. Usage of the hook may
5591 simplify the automaton pipeline description for some @acronym{VLIW}
5592 processors. If the hook is defined, it is used only for the automaton
5593 based pipeline description. The default is not to change the state
5594 when the new simulated processor cycle starts.
5597 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
5598 The hook can be used to initialize data used by the previous hook.
5601 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
5602 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
5603 to changed the state as if the insn were scheduled when the new
5604 simulated processor cycle finishes.
5607 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
5608 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
5609 used to initialize data used by the previous hook.
5612 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
5613 This hook controls better choosing an insn from the ready insn queue
5614 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
5615 chooses the first insn from the queue. If the hook returns a positive
5616 value, an additional scheduler code tries all permutations of
5617 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
5618 subsequent ready insns to choose an insn whose issue will result in
5619 maximal number of issued insns on the same cycle. For the
5620 @acronym{VLIW} processor, the code could actually solve the problem of
5621 packing simple insns into the @acronym{VLIW} insn. Of course, if the
5622 rules of @acronym{VLIW} packing are described in the automaton.
5624 This code also could be used for superscalar @acronym{RISC}
5625 processors. Let us consider a superscalar @acronym{RISC} processor
5626 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
5627 @var{B}, some insns can be executed only in pipelines @var{B} or
5628 @var{C}, and one insn can be executed in pipeline @var{B}. The
5629 processor may issue the 1st insn into @var{A} and the 2nd one into
5630 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
5631 until the next cycle. If the scheduler issues the 3rd insn the first,
5632 the processor could issue all 3 insns per cycle.
5634 Actually this code demonstrates advantages of the automaton based
5635 pipeline hazard recognizer. We try quickly and easy many insn
5636 schedules to choose the best one.
5638 The default is no multipass scheduling.
5641 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_BUBBLES (void)
5642 The @acronym{DFA}-based scheduler could take the insertion of nop
5643 operations for better insn scheduling into account. It can be done
5644 only if the multi-pass insn scheduling works (see hook
5645 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD}).
5647 Let us consider a @acronym{VLIW} processor insn with 3 slots. Each
5648 insn can be placed only in one of the three slots. We have 3 ready
5649 insns @var{A}, @var{B}, and @var{C}. @var{A} and @var{C} can be
5650 placed only in the 1st slot, @var{B} can be placed only in the 3rd
5651 slot. We described the automaton which does not permit empty slot
5652 gaps between insns (usually such description is simpler). Without
5653 this code the scheduler would place each insn in 3 separate
5654 @acronym{VLIW} insns. If the scheduler places a nop insn into the 2nd
5655 slot, it could place the 3 insns into 2 @acronym{VLIW} insns. What is
5656 the nop insn is returned by hook @samp{TARGET_SCHED_DFA_BUBBLE}. Hook
5657 @samp{TARGET_SCHED_INIT_DFA_BUBBLES} can be used to initialize or
5658 create the nop insns.
5660 You should remember that the scheduler does not insert the nop insns.
5661 It is not wise because of the following optimizations. The scheduler
5662 only considers such possibility to improve the result schedule. The
5663 nop insns should be inserted lately, e.g. on the final phase.
5666 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_BUBBLE (int @var{index})
5667 This hook @samp{FIRST_CYCLE_MULTIPASS_SCHEDULING} is used to insert
5668 nop operations for better insn scheduling when @acronym{DFA}-based
5669 scheduler makes multipass insn scheduling (see also description of
5670 hook @samp{TARGET_SCHED_INIT_DFA_BUBBLES}). This hook
5671 returns a nop insn with given @var{index}. The indexes start with
5672 zero. The hook should return @code{NULL} if there are no more nop
5673 insns with indexes greater than given index.
5676 Macros in the following table are generated by the program
5677 @file{genattr} and can be useful for writing the hooks.
5680 @findex TRADITIONAL_PIPELINE_INTERFACE
5681 @item TRADITIONAL_PIPELINE_INTERFACE
5682 The macro definition is generated if there is a traditional pipeline
5683 description in @file{.md} file. You should also remember that to
5684 simplify the insn scheduler sources an empty traditional pipeline
5685 description interface is generated even if there is no a traditional
5686 pipeline description in the @file{.md} file. The macro can be used to
5687 distinguish the two types of the traditional interface.
5689 @findex DFA_PIPELINE_INTERFACE
5690 @item DFA_PIPELINE_INTERFACE
5691 The macro definition is generated if there is an automaton pipeline
5692 description in @file{.md} file. You should also remember that to
5693 simplify the insn scheduler sources an empty automaton pipeline
5694 description interface is generated even if there is no an automaton
5695 pipeline description in the @file{.md} file. The macro can be used to
5696 distinguish the two types of the automaton interface.
5698 @findex MAX_DFA_ISSUE_RATE
5699 @item MAX_DFA_ISSUE_RATE
5700 The macro definition is generated in the automaton based pipeline
5701 description interface. Its value is calculated from the automaton
5702 based pipeline description and is equal to maximal number of all insns
5703 described in constructions @samp{define_insn_reservation} which can be
5704 issued on the same processor cycle.
5709 @section Dividing the Output into Sections (Texts, Data, @dots{})
5710 @c the above section title is WAY too long. maybe cut the part between
5711 @c the (...)? --mew 10feb93
5713 An object file is divided into sections containing different types of
5714 data. In the most common case, there are three sections: the @dfn{text
5715 section}, which holds instructions and read-only data; the @dfn{data
5716 section}, which holds initialized writable data; and the @dfn{bss
5717 section}, which holds uninitialized data. Some systems have other kinds
5720 The compiler must tell the assembler when to switch sections. These
5721 macros control what commands to output to tell the assembler this. You
5722 can also define additional sections.
5725 @findex TEXT_SECTION_ASM_OP
5726 @item TEXT_SECTION_ASM_OP
5727 A C expression whose value is a string, including spacing, containing the
5728 assembler operation that should precede instructions and read-only data.
5729 Normally @code{"\t.text"} is right.
5731 @findex TEXT_SECTION
5733 A C statement that switches to the default section containing instructions.
5734 Normally this is not needed, as simply defining @code{TEXT_SECTION_ASM_OP}
5735 is enough. The MIPS port uses this to sort all functions after all data
5738 @findex HOT_TEXT_SECTION_NAME
5739 @item HOT_TEXT_SECTION_NAME
5740 If defined, a C string constant for the name of the section containing most
5741 frequently executed functions of the program. If not defined, GCC will provide
5742 a default definition if the target supports named sections.
5744 @findex UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5745 @item UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5746 If defined, a C string constant for the name of the section containing unlikely
5747 executed functions in the program.
5749 @findex DATA_SECTION_ASM_OP
5750 @item DATA_SECTION_ASM_OP
5751 A C expression whose value is a string, including spacing, containing the
5752 assembler operation to identify the following data as writable initialized
5753 data. Normally @code{"\t.data"} is right.
5755 @findex READONLY_DATA_SECTION_ASM_OP
5756 @item READONLY_DATA_SECTION_ASM_OP
5757 A C expression whose value is a string, including spacing, containing the
5758 assembler operation to identify the following data as read-only initialized
5761 @findex READONLY_DATA_SECTION
5762 @item READONLY_DATA_SECTION
5763 A macro naming a function to call to switch to the proper section for
5764 read-only data. The default is to use @code{READONLY_DATA_SECTION_ASM_OP}
5765 if defined, else fall back to @code{text_section}.
5767 The most common definition will be @code{data_section}, if the target
5768 does not have a special read-only data section, and does not put data
5769 in the text section.
5771 @findex SHARED_SECTION_ASM_OP
5772 @item SHARED_SECTION_ASM_OP
5773 If defined, a C expression whose value is a string, including spacing,
5774 containing the assembler operation to identify the following data as
5775 shared data. If not defined, @code{DATA_SECTION_ASM_OP} will be used.
5777 @findex BSS_SECTION_ASM_OP
5778 @item BSS_SECTION_ASM_OP
5779 If defined, a C expression whose value is a string, including spacing,
5780 containing the assembler operation to identify the following data as
5781 uninitialized global data. If not defined, and neither
5782 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5783 uninitialized global data will be output in the data section if
5784 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5787 @findex SHARED_BSS_SECTION_ASM_OP
5788 @item SHARED_BSS_SECTION_ASM_OP
5789 If defined, a C expression whose value is a string, including spacing,
5790 containing the assembler operation to identify the following data as
5791 uninitialized global shared data. If not defined, and
5792 @code{BSS_SECTION_ASM_OP} is, the latter will be used.
5794 @findex INIT_SECTION_ASM_OP
5795 @item INIT_SECTION_ASM_OP
5796 If defined, a C expression whose value is a string, including spacing,
5797 containing the assembler operation to identify the following data as
5798 initialization code. If not defined, GCC will assume such a section does
5801 @findex FINI_SECTION_ASM_OP
5802 @item FINI_SECTION_ASM_OP
5803 If defined, a C expression whose value is a string, including spacing,
5804 containing the assembler operation to identify the following data as
5805 finalization code. If not defined, GCC will assume such a section does
5808 @findex CRT_CALL_STATIC_FUNCTION
5809 @item CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
5810 If defined, an ASM statement that switches to a different section
5811 via @var{section_op}, calls @var{function}, and switches back to
5812 the text section. This is used in @file{crtstuff.c} if
5813 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
5814 to initialization and finalization functions from the init and fini
5815 sections. By default, this macro uses a simple function call. Some
5816 ports need hand-crafted assembly code to avoid dependencies on
5817 registers initialized in the function prologue or to ensure that
5818 constant pools don't end up too far way in the text section.
5820 @findex FORCE_CODE_SECTION_ALIGN
5821 @item FORCE_CODE_SECTION_ALIGN
5822 If defined, an ASM statement that aligns a code section to some
5823 arbitrary boundary. This is used to force all fragments of the
5824 @code{.init} and @code{.fini} sections to have to same alignment
5825 and thus prevent the linker from having to add any padding.
5827 @findex EXTRA_SECTIONS
5830 @item EXTRA_SECTIONS
5831 A list of names for sections other than the standard two, which are
5832 @code{in_text} and @code{in_data}. You need not define this macro
5833 on a system with no other sections (that GCC needs to use).
5835 @findex EXTRA_SECTION_FUNCTIONS
5836 @findex text_section
5837 @findex data_section
5838 @item EXTRA_SECTION_FUNCTIONS
5839 One or more functions to be defined in @file{varasm.c}. These
5840 functions should do jobs analogous to those of @code{text_section} and
5841 @code{data_section}, for your additional sections. Do not define this
5842 macro if you do not define @code{EXTRA_SECTIONS}.
5844 @findex JUMP_TABLES_IN_TEXT_SECTION
5845 @item JUMP_TABLES_IN_TEXT_SECTION
5846 Define this macro to be an expression with a nonzero value if jump
5847 tables (for @code{tablejump} insns) should be output in the text
5848 section, along with the assembler instructions. Otherwise, the
5849 readonly data section is used.
5851 This macro is irrelevant if there is no separate readonly data section.
5854 @deftypefn {Target Hook} void TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
5855 Switches to the appropriate section for output of @var{exp}. You can
5856 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
5857 some sort. @var{reloc} indicates whether the initial value of @var{exp}
5858 requires link-time relocations. Bit 0 is set when variable contains
5859 local relocations only, while bit 1 is set for global relocations.
5860 Select the section by calling @code{data_section} or one of the
5861 alternatives for other sections. @var{align} is the constant alignment
5864 The default version of this function takes care of putting read-only
5865 variables in @code{readonly_data_section}.
5868 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
5869 Build up a unique section name, expressed as a @code{STRING_CST} node,
5870 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
5871 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
5872 the initial value of @var{exp} requires link-time relocations.
5874 The default version of this function appends the symbol name to the
5875 ELF section name that would normally be used for the symbol. For
5876 example, the function @code{foo} would be placed in @code{.text.foo}.
5877 Whatever the actual target object format, this is often good enough.
5880 @deftypefn {Target Hook} void TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
5881 Switches to the appropriate section for output of constant pool entry
5882 @var{x} in @var{mode}. You can assume that @var{x} is some kind of
5883 constant in RTL@. The argument @var{mode} is redundant except in the
5884 case of a @code{const_int} rtx. Select the section by calling
5885 @code{readonly_data_section} or one of the alternatives for other
5886 sections. @var{align} is the constant alignment in bits.
5888 The default version of this function takes care of putting symbolic
5889 constants in @code{flag_pic} mode in @code{data_section} and everything
5890 else in @code{readonly_data_section}.
5893 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, int @var{new_decl_p})
5894 Define this hook if references to a symbol or a constant must be
5895 treated differently depending on something about the variable or
5896 function named by the symbol (such as what section it is in).
5898 The hook is executed under two circumstances. One is immediately after
5899 the rtl for @var{decl} that represents a variable or a function has been
5900 created and stored in @code{DECL_RTL(@var{decl})}. The value of the rtl
5901 will be a @code{mem} whose address is a @code{symbol_ref}. The other is
5902 immediately after the rtl for @var{decl} that represents a constant has
5903 been created and stored in @code{TREE_CST_RTL (@var{decl})}. The macro
5904 is called once for each distinct constant in a source file.
5906 The @var{new_decl_p} argument will be true if this is the first time
5907 that @code{ENCODE_SECTION_INFO} has been invoked on this decl. It will
5908 be false for subsequent invocations, which will happen for duplicate
5909 declarations. Whether or not anything must be done for the duplicate
5910 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
5912 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
5913 The usual thing for this hook to do is to record a flag in the
5914 @code{symbol_ref} (such as @code{SYMBOL_REF_FLAG}) or to store a
5915 modified name string in the @code{symbol_ref} (if one bit is not
5916 enough information).
5919 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
5920 Decode @var{name} and return the real name part, sans
5921 the characters that @code{TARGET_ENCODE_SECTION_INFO}
5925 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
5926 Returns true if @var{exp} should be placed into a ``small data'' section.
5927 The default version of this hook always returns false.
5930 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
5931 Contains the value true if the target places read-only
5932 ``small data'' into a separate section. The default value is false.
5935 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
5936 Returns true if @var{exp} names an object for which name resolution
5937 rules must resolve to the current ``module'' (dynamic shared library
5938 or executable image).
5940 The default version of this hook implements the name resolution rules
5941 for ELF, which has a looser model of global name binding than other
5942 currently supported object file formats.
5945 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
5946 Contains the value true if the target supports thread-local storage.
5947 The default value is false.
5952 @section Position Independent Code
5953 @cindex position independent code
5956 This section describes macros that help implement generation of position
5957 independent code. Simply defining these macros is not enough to
5958 generate valid PIC; you must also add support to the macros
5959 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
5960 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
5961 @samp{movsi} to do something appropriate when the source operand
5962 contains a symbolic address. You may also need to alter the handling of
5963 switch statements so that they use relative addresses.
5964 @c i rearranged the order of the macros above to try to force one of
5965 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
5968 @findex PIC_OFFSET_TABLE_REGNUM
5969 @item PIC_OFFSET_TABLE_REGNUM
5970 The register number of the register used to address a table of static
5971 data addresses in memory. In some cases this register is defined by a
5972 processor's ``application binary interface'' (ABI)@. When this macro
5973 is defined, RTL is generated for this register once, as with the stack
5974 pointer and frame pointer registers. If this macro is not defined, it
5975 is up to the machine-dependent files to allocate such a register (if
5976 necessary). Note that this register must be fixed when in use (e.g.@:
5977 when @code{flag_pic} is true).
5979 @findex PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5980 @item PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5981 Define this macro if the register defined by
5982 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
5983 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
5985 @findex FINALIZE_PIC
5987 By generating position-independent code, when two different programs (A
5988 and B) share a common library (libC.a), the text of the library can be
5989 shared whether or not the library is linked at the same address for both
5990 programs. In some of these environments, position-independent code
5991 requires not only the use of different addressing modes, but also
5992 special code to enable the use of these addressing modes.
5994 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
5995 codes once the function is being compiled into assembly code, but not
5996 before. (It is not done before, because in the case of compiling an
5997 inline function, it would lead to multiple PIC prologues being
5998 included in functions which used inline functions and were compiled to
6001 @findex LEGITIMATE_PIC_OPERAND_P
6002 @item LEGITIMATE_PIC_OPERAND_P (@var{x})
6003 A C expression that is nonzero if @var{x} is a legitimate immediate
6004 operand on the target machine when generating position independent code.
6005 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6006 check this. You can also assume @var{flag_pic} is true, so you need not
6007 check it either. You need not define this macro if all constants
6008 (including @code{SYMBOL_REF}) can be immediate operands when generating
6009 position independent code.
6012 @node Assembler Format
6013 @section Defining the Output Assembler Language
6015 This section describes macros whose principal purpose is to describe how
6016 to write instructions in assembler language---rather than what the
6020 * File Framework:: Structural information for the assembler file.
6021 * Data Output:: Output of constants (numbers, strings, addresses).
6022 * Uninitialized Data:: Output of uninitialized variables.
6023 * Label Output:: Output and generation of labels.
6024 * Initialization:: General principles of initialization
6025 and termination routines.
6026 * Macros for Initialization::
6027 Specific macros that control the handling of
6028 initialization and termination routines.
6029 * Instruction Output:: Output of actual instructions.
6030 * Dispatch Tables:: Output of jump tables.
6031 * Exception Region Output:: Output of exception region code.
6032 * Alignment Output:: Pseudo ops for alignment and skipping data.
6035 @node File Framework
6036 @subsection The Overall Framework of an Assembler File
6037 @cindex assembler format
6038 @cindex output of assembler code
6040 @c prevent bad page break with this line
6041 This describes the overall framework of an assembler file.
6044 @findex ASM_FILE_START
6045 @item ASM_FILE_START (@var{stream})
6046 A C expression which outputs to the stdio stream @var{stream}
6047 some appropriate text to go at the start of an assembler file.
6049 Normally this macro is defined to output a line containing
6050 @samp{#NO_APP}, which is a comment that has no effect on most
6051 assemblers but tells the GNU assembler that it can save time by not
6052 checking for certain assembler constructs.
6054 On systems that use SDB, it is necessary to output certain commands;
6055 see @file{attasm.h}.
6057 @findex ASM_FILE_END
6058 @item ASM_FILE_END (@var{stream})
6059 A C expression which outputs to the stdio stream @var{stream}
6060 some appropriate text to go at the end of an assembler file.
6062 If this macro is not defined, the default is to output nothing
6063 special at the end of the file. Most systems don't require any
6066 On systems that use SDB, it is necessary to output certain commands;
6067 see @file{attasm.h}.
6069 @findex ASM_COMMENT_START
6070 @item ASM_COMMENT_START
6071 A C string constant describing how to begin a comment in the target
6072 assembler language. The compiler assumes that the comment will end at
6073 the end of the line.
6077 A C string constant for text to be output before each @code{asm}
6078 statement or group of consecutive ones. Normally this is
6079 @code{"#APP"}, which is a comment that has no effect on most
6080 assemblers but tells the GNU assembler that it must check the lines
6081 that follow for all valid assembler constructs.
6085 A C string constant for text to be output after each @code{asm}
6086 statement or group of consecutive ones. Normally this is
6087 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6088 time-saving assumptions that are valid for ordinary compiler output.
6090 @findex ASM_OUTPUT_SOURCE_FILENAME
6091 @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6092 A C statement to output COFF information or DWARF debugging information
6093 which indicates that filename @var{name} is the current source file to
6094 the stdio stream @var{stream}.
6096 This macro need not be defined if the standard form of output
6097 for the file format in use is appropriate.
6099 @findex OUTPUT_QUOTED_STRING
6100 @item OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6101 A C statement to output the string @var{string} to the stdio stream
6102 @var{stream}. If you do not call the function @code{output_quoted_string}
6103 in your config files, GCC will only call it to output filenames to
6104 the assembler source. So you can use it to canonicalize the format
6105 of the filename using this macro.
6107 @findex ASM_OUTPUT_SOURCE_LINE
6108 @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
6109 A C statement to output DBX or SDB debugging information before code
6110 for line number @var{line} of the current source file to the
6111 stdio stream @var{stream}.
6113 This macro need not be defined if the standard form of debugging
6114 information for the debugger in use is appropriate.
6116 @findex ASM_OUTPUT_IDENT
6117 @item ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6118 A C statement to output something to the assembler file to handle a
6119 @samp{#ident} directive containing the text @var{string}. If this
6120 macro is not defined, nothing is output for a @samp{#ident} directive.
6122 @findex OBJC_PROLOGUE
6124 A C statement to output any assembler statements which are required to
6125 precede any Objective-C object definitions or message sending. The
6126 statement is executed only when compiling an Objective-C program.
6129 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6130 Output assembly directives to switch to section @var{name}. The section
6131 should have attributes as specified by @var{flags}, which is a bit mask
6132 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6133 is nonzero, it contains an alignment in bytes to be used for the section,
6134 otherwise some target default should be used. Only targets that must
6135 specify an alignment within the section directive need pay attention to
6136 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6139 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6140 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6143 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6144 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6145 based on a variable or function decl, a section name, and whether or not the
6146 declaration's initializer may contain runtime relocations. @var{decl} may be
6147 null, in which case read-write data should be assumed.
6149 The default version if this function handles choosing code vs data,
6150 read-only vs read-write data, and @code{flag_pic}. You should only
6151 need to override this if your target has special flags that might be
6152 set via @code{__attribute__}.
6157 @subsection Output of Data
6160 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6161 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6162 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6163 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6164 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6165 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6166 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6167 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6168 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6169 These hooks specify assembly directives for creating certain kinds
6170 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6171 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6172 aligned two-byte object, and so on. Any of the hooks may be
6173 @code{NULL}, indicating that no suitable directive is available.
6175 The compiler will print these strings at the start of a new line,
6176 followed immediately by the object's initial value. In most cases,
6177 the string should contain a tab, a pseudo-op, and then another tab.
6180 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6181 The @code{assemble_integer} function uses this hook to output an
6182 integer object. @var{x} is the object's value, @var{size} is its size
6183 in bytes and @var{aligned_p} indicates whether it is aligned. The
6184 function should return @code{true} if it was able to output the
6185 object. If it returns false, @code{assemble_integer} will try to
6186 split the object into smaller parts.
6188 The default implementation of this hook will use the
6189 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6190 when the relevant string is @code{NULL}.
6194 @findex OUTPUT_ADDR_CONST_EXTRA
6195 @item OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6196 A C statement to recognize @var{rtx} patterns that
6197 @code{output_addr_const} can't deal with, and output assembly code to
6198 @var{stream} corresponding to the pattern @var{x}. This may be used to
6199 allow machine-dependent @code{UNSPEC}s to appear within constants.
6201 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6202 @code{goto fail}, so that a standard error message is printed. If it
6203 prints an error message itself, by calling, for example,
6204 @code{output_operand_lossage}, it may just complete normally.
6206 @findex ASM_OUTPUT_ASCII
6207 @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6208 A C statement to output to the stdio stream @var{stream} an assembler
6209 instruction to assemble a string constant containing the @var{len}
6210 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6211 @code{char *} and @var{len} a C expression of type @code{int}.
6213 If the assembler has a @code{.ascii} pseudo-op as found in the
6214 Berkeley Unix assembler, do not define the macro
6215 @code{ASM_OUTPUT_ASCII}.
6217 @findex ASM_OUTPUT_FDESC
6218 @item ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6219 A C statement to output word @var{n} of a function descriptor for
6220 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6221 is defined, and is otherwise unused.
6223 @findex CONSTANT_POOL_BEFORE_FUNCTION
6224 @item CONSTANT_POOL_BEFORE_FUNCTION
6225 You may define this macro as a C expression. You should define the
6226 expression to have a nonzero value if GCC should output the constant
6227 pool for a function before the code for the function, or a zero value if
6228 GCC should output the constant pool after the function. If you do
6229 not define this macro, the usual case, GCC will output the constant
6230 pool before the function.
6232 @findex ASM_OUTPUT_POOL_PROLOGUE
6233 @item ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
6234 A C statement to output assembler commands to define the start of the
6235 constant pool for a function. @var{funname} is a string giving
6236 the name of the function. Should the return type of the function
6237 be required, it can be obtained via @var{fundecl}. @var{size}
6238 is the size, in bytes, of the constant pool that will be written
6239 immediately after this call.
6241 If no constant-pool prefix is required, the usual case, this macro need
6244 @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY
6245 @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
6246 A C statement (with or without semicolon) to output a constant in the
6247 constant pool, if it needs special treatment. (This macro need not do
6248 anything for RTL expressions that can be output normally.)
6250 The argument @var{file} is the standard I/O stream to output the
6251 assembler code on. @var{x} is the RTL expression for the constant to
6252 output, and @var{mode} is the machine mode (in case @var{x} is a
6253 @samp{const_int}). @var{align} is the required alignment for the value
6254 @var{x}; you should output an assembler directive to force this much
6257 The argument @var{labelno} is a number to use in an internal label for
6258 the address of this pool entry. The definition of this macro is
6259 responsible for outputting the label definition at the proper place.
6260 Here is how to do this:
6263 ASM_OUTPUT_INTERNAL_LABEL (@var{file}, "LC", @var{labelno});
6266 When you output a pool entry specially, you should end with a
6267 @code{goto} to the label @var{jumpto}. This will prevent the same pool
6268 entry from being output a second time in the usual manner.
6270 You need not define this macro if it would do nothing.
6272 @findex CONSTANT_AFTER_FUNCTION_P
6273 @item CONSTANT_AFTER_FUNCTION_P (@var{exp})
6274 Define this macro as a C expression which is nonzero if the constant
6275 @var{exp}, of type @code{tree}, should be output after the code for a
6276 function. The compiler will normally output all constants before the
6277 function; you need not define this macro if this is OK@.
6279 @findex ASM_OUTPUT_POOL_EPILOGUE
6280 @item ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
6281 A C statement to output assembler commands to at the end of the constant
6282 pool for a function. @var{funname} is a string giving the name of the
6283 function. Should the return type of the function be required, you can
6284 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
6285 constant pool that GCC wrote immediately before this call.
6287 If no constant-pool epilogue is required, the usual case, you need not
6290 @findex IS_ASM_LOGICAL_LINE_SEPARATOR
6291 @item IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
6292 Define this macro as a C expression which is nonzero if @var{C} is
6293 used as a logical line separator by the assembler.
6295 If you do not define this macro, the default is that only
6296 the character @samp{;} is treated as a logical line separator.
6299 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
6300 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
6301 These target hooks are C string constants, describing the syntax in the
6302 assembler for grouping arithmetic expressions. If not overridden, they
6303 default to normal parentheses, which is correct for most assemblers.
6306 These macros are provided by @file{real.h} for writing the definitions
6307 of @code{ASM_OUTPUT_DOUBLE} and the like:
6310 @item REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
6311 @itemx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
6312 @itemx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
6313 @findex REAL_VALUE_TO_TARGET_SINGLE
6314 @findex REAL_VALUE_TO_TARGET_DOUBLE
6315 @findex REAL_VALUE_TO_TARGET_LONG_DOUBLE
6316 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
6317 floating point representation, and store its bit pattern in the variable
6318 @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE}, this variable should
6319 be a simple @code{long int}. For the others, it should be an array of
6320 @code{long int}. The number of elements in this array is determined by
6321 the size of the desired target floating point data type: 32 bits of it
6322 go in each @code{long int} array element. Each array element holds 32
6323 bits of the result, even if @code{long int} is wider than 32 bits on the
6326 The array element values are designed so that you can print them out
6327 using @code{fprintf} in the order they should appear in the target
6331 @node Uninitialized Data
6332 @subsection Output of Uninitialized Variables
6334 Each of the macros in this section is used to do the whole job of
6335 outputting a single uninitialized variable.
6338 @findex ASM_OUTPUT_COMMON
6339 @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6340 A C statement (sans semicolon) to output to the stdio stream
6341 @var{stream} the assembler definition of a common-label named
6342 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6343 is the size rounded up to whatever alignment the caller wants.
6345 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6346 output the name itself; before and after that, output the additional
6347 assembler syntax for defining the name, and a newline.
6349 This macro controls how the assembler definitions of uninitialized
6350 common global variables are output.
6352 @findex ASM_OUTPUT_ALIGNED_COMMON
6353 @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
6354 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
6355 separate, explicit argument. If you define this macro, it is used in
6356 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
6357 handling the required alignment of the variable. The alignment is specified
6358 as the number of bits.
6360 @findex ASM_OUTPUT_ALIGNED_DECL_COMMON
6361 @item ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6362 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
6363 variable to be output, if there is one, or @code{NULL_TREE} if there
6364 is no corresponding variable. If you define this macro, GCC will use it
6365 in place of both @code{ASM_OUTPUT_COMMON} and
6366 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
6367 the variable's decl in order to chose what to output.
6369 @findex ASM_OUTPUT_SHARED_COMMON
6370 @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6371 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
6372 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
6375 @findex ASM_OUTPUT_BSS
6376 @item ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6377 A C statement (sans semicolon) to output to the stdio stream
6378 @var{stream} the assembler definition of uninitialized global @var{decl} named
6379 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6380 is the size rounded up to whatever alignment the caller wants.
6382 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
6383 defining this macro. If unable, use the expression
6384 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
6385 before and after that, output the additional assembler syntax for defining
6386 the name, and a newline.
6388 This macro controls how the assembler definitions of uninitialized global
6389 variables are output. This macro exists to properly support languages like
6390 C++ which do not have @code{common} data. However, this macro currently
6391 is not defined for all targets. If this macro and
6392 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
6393 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
6394 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
6396 @findex ASM_OUTPUT_ALIGNED_BSS
6397 @item ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6398 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
6399 separate, explicit argument. If you define this macro, it is used in
6400 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
6401 handling the required alignment of the variable. The alignment is specified
6402 as the number of bits.
6404 Try to use function @code{asm_output_aligned_bss} defined in file
6405 @file{varasm.c} when defining this macro.
6407 @findex ASM_OUTPUT_SHARED_BSS
6408 @item ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6409 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
6410 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
6413 @findex ASM_OUTPUT_LOCAL
6414 @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6415 A C statement (sans semicolon) to output to the stdio stream
6416 @var{stream} the assembler definition of a local-common-label named
6417 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6418 is the size rounded up to whatever alignment the caller wants.
6420 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6421 output the name itself; before and after that, output the additional
6422 assembler syntax for defining the name, and a newline.
6424 This macro controls how the assembler definitions of uninitialized
6425 static variables are output.
6427 @findex ASM_OUTPUT_ALIGNED_LOCAL
6428 @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
6429 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
6430 separate, explicit argument. If you define this macro, it is used in
6431 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
6432 handling the required alignment of the variable. The alignment is specified
6433 as the number of bits.
6435 @findex ASM_OUTPUT_ALIGNED_DECL_LOCAL
6436 @item ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6437 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
6438 variable to be output, if there is one, or @code{NULL_TREE} if there
6439 is no corresponding variable. If you define this macro, GCC will use it
6440 in place of both @code{ASM_OUTPUT_DECL} and
6441 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
6442 the variable's decl in order to chose what to output.
6444 @findex ASM_OUTPUT_SHARED_LOCAL
6445 @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6446 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
6447 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
6452 @subsection Output and Generation of Labels
6454 @c prevent bad page break with this line
6455 This is about outputting labels.
6458 @findex ASM_OUTPUT_LABEL
6459 @findex assemble_name
6460 @item ASM_OUTPUT_LABEL (@var{stream}, @var{name})
6461 A C statement (sans semicolon) to output to the stdio stream
6462 @var{stream} the assembler definition of a label named @var{name}.
6463 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6464 output the name itself; before and after that, output the additional
6465 assembler syntax for defining the name, and a newline. A default
6466 definition of this macro is provided which is correct for most systems.
6470 A C string containing the appropriate assembler directive to specify the
6471 size of a symbol, without any arguments. On systems that use ELF, the
6472 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
6473 systems, the default is not to define this macro.
6475 Define this macro only if it is correct to use the default definitions
6476 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
6477 for your system. If you need your own custom definitions of those
6478 macros, or if you do not need explicit symbol sizes at all, do not
6481 @findex ASM_OUTPUT_SIZE_DIRECTIVE
6482 @item ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
6483 A C statement (sans semicolon) to output to the stdio stream
6484 @var{stream} a directive telling the assembler that the size of the
6485 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
6486 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6489 @findex ASM_OUTPUT_MEASURED_SIZE
6490 @item ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
6491 A C statement (sans semicolon) to output to the stdio stream
6492 @var{stream} a directive telling the assembler to calculate the size of
6493 the symbol @var{name} by subtracting its address from the current
6496 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6497 provided. The default assumes that the assembler recognizes a special
6498 @samp{.} symbol as referring to the current address, and can calculate
6499 the difference between this and another symbol. If your assembler does
6500 not recognize @samp{.} or cannot do calculations with it, you will need
6501 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
6505 A C string containing the appropriate assembler directive to specify the
6506 type of a symbol, without any arguments. On systems that use ELF, the
6507 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
6508 systems, the default is not to define this macro.
6510 Define this macro only if it is correct to use the default definition of
6511 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6512 custom definition of this macro, or if you do not need explicit symbol
6513 types at all, do not define this macro.
6515 @findex TYPE_OPERAND_FMT
6516 @item TYPE_OPERAND_FMT
6517 A C string which specifies (using @code{printf} syntax) the format of
6518 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
6519 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
6520 the default is not to define this macro.
6522 Define this macro only if it is correct to use the default definition of
6523 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6524 custom definition of this macro, or if you do not need explicit symbol
6525 types at all, do not define this macro.
6527 @findex ASM_OUTPUT_TYPE_DIRECTIVE
6528 @item ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
6529 A C statement (sans semicolon) to output to the stdio stream
6530 @var{stream} a directive telling the assembler that the type of the
6531 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
6532 that string is always either @samp{"function"} or @samp{"object"}, but
6533 you should not count on this.
6535 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
6536 definition of this macro is provided.
6538 @findex ASM_DECLARE_FUNCTION_NAME
6539 @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
6540 A C statement (sans semicolon) to output to the stdio stream
6541 @var{stream} any text necessary for declaring the name @var{name} of a
6542 function which is being defined. This macro is responsible for
6543 outputting the label definition (perhaps using
6544 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
6545 @code{FUNCTION_DECL} tree node representing the function.
6547 If this macro is not defined, then the function name is defined in the
6548 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6550 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6553 @findex ASM_DECLARE_FUNCTION_SIZE
6554 @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
6555 A C statement (sans semicolon) to output to the stdio stream
6556 @var{stream} any text necessary for declaring the size of a function
6557 which is being defined. The argument @var{name} is the name of the
6558 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
6559 representing the function.
6561 If this macro is not defined, then the function size is not defined.
6563 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
6566 @findex ASM_DECLARE_OBJECT_NAME
6567 @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6568 A C statement (sans semicolon) to output to the stdio stream
6569 @var{stream} any text necessary for declaring the name @var{name} of an
6570 initialized variable which is being defined. This macro must output the
6571 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6572 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6574 If this macro is not defined, then the variable name is defined in the
6575 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6577 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
6578 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
6580 @findex ASM_DECLARE_REGISTER_GLOBAL
6581 @item ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6582 A C statement (sans semicolon) to output to the stdio stream
6583 @var{stream} any text necessary for claiming a register @var{regno}
6584 for a global variable @var{decl} with name @var{name}.
6586 If you don't define this macro, that is equivalent to defining it to do
6589 @findex ASM_FINISH_DECLARE_OBJECT
6590 @item ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6591 A C statement (sans semicolon) to finish up declaring a variable name
6592 once the compiler has processed its initializer fully and thus has had a
6593 chance to determine the size of an array when controlled by an
6594 initializer. This is used on systems where it's necessary to declare
6595 something about the size of the object.
6597 If you don't define this macro, that is equivalent to defining it to do
6600 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
6601 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
6604 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
6605 This target hook is a function to output to the stdio stream
6606 @var{stream} some commands that will make the label @var{name} global;
6607 that is, available for reference from other files.
6609 The default implementation relies on a proper definition of
6610 @code{GLOBAL_ASM_OP}.
6614 @findex ASM_WEAKEN_LABEL
6615 @item ASM_WEAKEN_LABEL (@var{stream}, @var{name})
6616 A C statement (sans semicolon) to output to the stdio stream
6617 @var{stream} some commands that will make the label @var{name} weak;
6618 that is, available for reference from other files but only used if
6619 no other definition is available. Use the expression
6620 @code{assemble_name (@var{stream}, @var{name})} to output the name
6621 itself; before and after that, output the additional assembler syntax
6622 for making that name weak, and a newline.
6624 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
6625 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
6628 @findex ASM_WEAKEN_DECL
6629 @item ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
6630 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
6631 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
6632 or variable decl. If @var{value} is not @code{NULL}, this C statement
6633 should output to the stdio stream @var{stream} assembler code which
6634 defines (equates) the weak symbol @var{name} to have the value
6635 @var{value}. If @var{value} is @code{NULL}, it should output commands
6636 to make @var{name} weak.
6638 @findex SUPPORTS_WEAK
6640 A C expression which evaluates to true if the target supports weak symbols.
6642 If you don't define this macro, @file{defaults.h} provides a default
6643 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
6644 is defined, the default definition is @samp{1}; otherwise, it is
6645 @samp{0}. Define this macro if you want to control weak symbol support
6646 with a compiler flag such as @option{-melf}.
6648 @findex MAKE_DECL_ONE_ONLY (@var{decl})
6649 @item MAKE_DECL_ONE_ONLY
6650 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
6651 public symbol such that extra copies in multiple translation units will
6652 be discarded by the linker. Define this macro if your object file
6653 format provides support for this concept, such as the @samp{COMDAT}
6654 section flags in the Microsoft Windows PE/COFF format, and this support
6655 requires changes to @var{decl}, such as putting it in a separate section.
6657 @findex SUPPORTS_ONE_ONLY
6658 @item SUPPORTS_ONE_ONLY
6659 A C expression which evaluates to true if the target supports one-only
6662 If you don't define this macro, @file{varasm.c} provides a default
6663 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6664 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6665 you want to control one-only symbol support with a compiler flag, or if
6666 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6667 be emitted as one-only.
6669 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
6670 This target hook is a function to output to @var{asm_out_file} some
6671 commands that will make the symbol(s) associated with @var{decl} have
6672 hidden, protected or internal visibility as specified by @var{visibility}.
6675 @findex ASM_OUTPUT_EXTERNAL
6676 @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6677 A C statement (sans semicolon) to output to the stdio stream
6678 @var{stream} any text necessary for declaring the name of an external
6679 symbol named @var{name} which is referenced in this compilation but
6680 not defined. The value of @var{decl} is the tree node for the
6683 This macro need not be defined if it does not need to output anything.
6684 The GNU assembler and most Unix assemblers don't require anything.
6686 @findex ASM_OUTPUT_EXTERNAL_LIBCALL
6687 @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref})
6688 A C statement (sans semicolon) to output on @var{stream} an assembler
6689 pseudo-op to declare a library function name external. The name of the
6690 library function is given by @var{symref}, which has type @code{rtx} and
6691 is a @code{symbol_ref}.
6693 This macro need not be defined if it does not need to output anything.
6694 The GNU assembler and most Unix assemblers don't require anything.
6696 @findex ASM_OUTPUT_LABELREF
6697 @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6698 A C statement (sans semicolon) to output to the stdio stream
6699 @var{stream} a reference in assembler syntax to a label named
6700 @var{name}. This should add @samp{_} to the front of the name, if that
6701 is customary on your operating system, as it is in most Berkeley Unix
6702 systems. This macro is used in @code{assemble_name}.
6704 @findex ASM_OUTPUT_SYMBOL_REF
6705 @item ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6706 A C statement (sans semicolon) to output a reference to
6707 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
6708 will be used to output the name of the symbol. This macro may be used
6709 to modify the way a symbol is referenced depending on information
6710 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
6712 @findex ASM_OUTPUT_LABEL_REF
6713 @item ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
6714 A C statement (sans semicolon) to output a reference to @var{buf}, the
6715 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
6716 @code{assemble_name} will be used to output the name of the symbol.
6717 This macro is not used by @code{output_asm_label}, or the @code{%l}
6718 specifier that calls it; the intention is that this macro should be set
6719 when it is necessary to output a label differently when its address is
6722 @findex ASM_OUTPUT_INTERNAL_LABEL
6723 @item ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{prefix}, @var{num})
6724 A C statement to output to the stdio stream @var{stream} a label whose
6725 name is made from the string @var{prefix} and the number @var{num}.
6727 It is absolutely essential that these labels be distinct from the labels
6728 used for user-level functions and variables. Otherwise, certain programs
6729 will have name conflicts with internal labels.
6731 It is desirable to exclude internal labels from the symbol table of the
6732 object file. Most assemblers have a naming convention for labels that
6733 should be excluded; on many systems, the letter @samp{L} at the
6734 beginning of a label has this effect. You should find out what
6735 convention your system uses, and follow it.
6737 The usual definition of this macro is as follows:
6740 fprintf (@var{stream}, "L%s%d:\n", @var{prefix}, @var{num})
6743 @findex ASM_OUTPUT_DEBUG_LABEL
6744 @item ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
6745 A C statement to output to the stdio stream @var{stream} a debug info
6746 label whose name is made from the string @var{prefix} and the number
6747 @var{num}. This is useful for VLIW targets, where debug info labels
6748 may need to be treated differently than branch target labels. On some
6749 systems, branch target labels must be at the beginning of instruction
6750 bundles, but debug info labels can occur in the middle of instruction
6753 If this macro is not defined, then @code{ASM_OUTPUT_INTERNAL_LABEL} will be
6756 @findex ASM_GENERATE_INTERNAL_LABEL
6757 @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
6758 A C statement to store into the string @var{string} a label whose name
6759 is made from the string @var{prefix} and the number @var{num}.
6761 This string, when output subsequently by @code{assemble_name}, should
6762 produce the output that @code{ASM_OUTPUT_INTERNAL_LABEL} would produce
6763 with the same @var{prefix} and @var{num}.
6765 If the string begins with @samp{*}, then @code{assemble_name} will
6766 output the rest of the string unchanged. It is often convenient for
6767 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
6768 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
6769 to output the string, and may change it. (Of course,
6770 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
6771 you should know what it does on your machine.)
6773 @findex ASM_FORMAT_PRIVATE_NAME
6774 @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
6775 A C expression to assign to @var{outvar} (which is a variable of type
6776 @code{char *}) a newly allocated string made from the string
6777 @var{name} and the number @var{number}, with some suitable punctuation
6778 added. Use @code{alloca} to get space for the string.
6780 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
6781 produce an assembler label for an internal static variable whose name is
6782 @var{name}. Therefore, the string must be such as to result in valid
6783 assembler code. The argument @var{number} is different each time this
6784 macro is executed; it prevents conflicts between similarly-named
6785 internal static variables in different scopes.
6787 Ideally this string should not be a valid C identifier, to prevent any
6788 conflict with the user's own symbols. Most assemblers allow periods
6789 or percent signs in assembler symbols; putting at least one of these
6790 between the name and the number will suffice.
6792 @findex ASM_OUTPUT_DEF
6793 @item ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
6794 A C statement to output to the stdio stream @var{stream} assembler code
6795 which defines (equates) the symbol @var{name} to have the value @var{value}.
6798 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6799 correct for most systems.
6801 @findex ASM_OUTPUT_DEF_FROM_DECLS
6802 @item ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
6803 A C statement to output to the stdio stream @var{stream} assembler code
6804 which defines (equates) the symbol whose tree node is @var{decl_of_name}
6805 to have the value of the tree node @var{decl_of_value}. This macro will
6806 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
6807 the tree nodes are available.
6810 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6811 correct for most systems.
6813 @findex ASM_OUTPUT_WEAK_ALIAS
6814 @item ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
6815 A C statement to output to the stdio stream @var{stream} assembler code
6816 which defines (equates) the weak symbol @var{name} to have the value
6817 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
6818 an undefined weak symbol.
6820 Define this macro if the target only supports weak aliases; define
6821 @code{ASM_OUTPUT_DEF} instead if possible.
6823 @findex OBJC_GEN_METHOD_LABEL
6824 @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
6825 Define this macro to override the default assembler names used for
6826 Objective-C methods.
6828 The default name is a unique method number followed by the name of the
6829 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
6830 the category is also included in the assembler name (e.g.@:
6833 These names are safe on most systems, but make debugging difficult since
6834 the method's selector is not present in the name. Therefore, particular
6835 systems define other ways of computing names.
6837 @var{buf} is an expression of type @code{char *} which gives you a
6838 buffer in which to store the name; its length is as long as
6839 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
6840 50 characters extra.
6842 The argument @var{is_inst} specifies whether the method is an instance
6843 method or a class method; @var{class_name} is the name of the class;
6844 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
6845 in a category); and @var{sel_name} is the name of the selector.
6847 On systems where the assembler can handle quoted names, you can use this
6848 macro to provide more human-readable names.
6850 @findex ASM_DECLARE_CLASS_REFERENCE
6851 @item ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
6852 A C statement (sans semicolon) to output to the stdio stream
6853 @var{stream} commands to declare that the label @var{name} is an
6854 Objective-C class reference. This is only needed for targets whose
6855 linkers have special support for NeXT-style runtimes.
6857 @findex ASM_DECLARE_UNRESOLVED_REFERENCE
6858 @item ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
6859 A C statement (sans semicolon) to output to the stdio stream
6860 @var{stream} commands to declare that the label @var{name} is an
6861 unresolved Objective-C class reference. This is only needed for targets
6862 whose linkers have special support for NeXT-style runtimes.
6865 @node Initialization
6866 @subsection How Initialization Functions Are Handled
6867 @cindex initialization routines
6868 @cindex termination routines
6869 @cindex constructors, output of
6870 @cindex destructors, output of
6872 The compiled code for certain languages includes @dfn{constructors}
6873 (also called @dfn{initialization routines})---functions to initialize
6874 data in the program when the program is started. These functions need
6875 to be called before the program is ``started''---that is to say, before
6876 @code{main} is called.
6878 Compiling some languages generates @dfn{destructors} (also called
6879 @dfn{termination routines}) that should be called when the program
6882 To make the initialization and termination functions work, the compiler
6883 must output something in the assembler code to cause those functions to
6884 be called at the appropriate time. When you port the compiler to a new
6885 system, you need to specify how to do this.
6887 There are two major ways that GCC currently supports the execution of
6888 initialization and termination functions. Each way has two variants.
6889 Much of the structure is common to all four variations.
6891 @findex __CTOR_LIST__
6892 @findex __DTOR_LIST__
6893 The linker must build two lists of these functions---a list of
6894 initialization functions, called @code{__CTOR_LIST__}, and a list of
6895 termination functions, called @code{__DTOR_LIST__}.
6897 Each list always begins with an ignored function pointer (which may hold
6898 0, @minus{}1, or a count of the function pointers after it, depending on
6899 the environment). This is followed by a series of zero or more function
6900 pointers to constructors (or destructors), followed by a function
6901 pointer containing zero.
6903 Depending on the operating system and its executable file format, either
6904 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
6905 time and exit time. Constructors are called in reverse order of the
6906 list; destructors in forward order.
6908 The best way to handle static constructors works only for object file
6909 formats which provide arbitrarily-named sections. A section is set
6910 aside for a list of constructors, and another for a list of destructors.
6911 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
6912 object file that defines an initialization function also puts a word in
6913 the constructor section to point to that function. The linker
6914 accumulates all these words into one contiguous @samp{.ctors} section.
6915 Termination functions are handled similarly.
6917 This method will be chosen as the default by @file{target-def.h} if
6918 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
6919 support arbitrary sections, but does support special designated
6920 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
6921 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
6923 When arbitrary sections are available, there are two variants, depending
6924 upon how the code in @file{crtstuff.c} is called. On systems that
6925 support a @dfn{.init} section which is executed at program startup,
6926 parts of @file{crtstuff.c} are compiled into that section. The
6927 program is linked by the @code{gcc} driver like this:
6930 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
6933 The prologue of a function (@code{__init}) appears in the @code{.init}
6934 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
6935 for the function @code{__fini} in the @dfn{.fini} section. Normally these
6936 files are provided by the operating system or by the GNU C library, but
6937 are provided by GCC for a few targets.
6939 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
6940 compiled from @file{crtstuff.c}. They contain, among other things, code
6941 fragments within the @code{.init} and @code{.fini} sections that branch
6942 to routines in the @code{.text} section. The linker will pull all parts
6943 of a section together, which results in a complete @code{__init} function
6944 that invokes the routines we need at startup.
6946 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
6949 If no init section is available, when GCC compiles any function called
6950 @code{main} (or more accurately, any function designated as a program
6951 entry point by the language front end calling @code{expand_main_function}),
6952 it inserts a procedure call to @code{__main} as the first executable code
6953 after the function prologue. The @code{__main} function is defined
6954 in @file{libgcc2.c} and runs the global constructors.
6956 In file formats that don't support arbitrary sections, there are again
6957 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
6958 and an `a.out' format must be used. In this case,
6959 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
6960 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
6961 and with the address of the void function containing the initialization
6962 code as its value. The GNU linker recognizes this as a request to add
6963 the value to a @dfn{set}; the values are accumulated, and are eventually
6964 placed in the executable as a vector in the format described above, with
6965 a leading (ignored) count and a trailing zero element.
6966 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
6967 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
6968 the compilation of @code{main} to call @code{__main} as above, starting
6969 the initialization process.
6971 The last variant uses neither arbitrary sections nor the GNU linker.
6972 This is preferable when you want to do dynamic linking and when using
6973 file formats which the GNU linker does not support, such as `ECOFF'@. In
6974 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
6975 termination functions are recognized simply by their names. This requires
6976 an extra program in the linkage step, called @command{collect2}. This program
6977 pretends to be the linker, for use with GCC; it does its job by running
6978 the ordinary linker, but also arranges to include the vectors of
6979 initialization and termination functions. These functions are called
6980 via @code{__main} as described above. In order to use this method,
6981 @code{use_collect2} must be defined in the target in @file{config.gcc}.
6984 The following section describes the specific macros that control and
6985 customize the handling of initialization and termination functions.
6988 @node Macros for Initialization
6989 @subsection Macros Controlling Initialization Routines
6991 Here are the macros that control how the compiler handles initialization
6992 and termination functions:
6995 @findex INIT_SECTION_ASM_OP
6996 @item INIT_SECTION_ASM_OP
6997 If defined, a C string constant, including spacing, for the assembler
6998 operation to identify the following data as initialization code. If not
6999 defined, GCC will assume such a section does not exist. When you are
7000 using special sections for initialization and termination functions, this
7001 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7002 run the initialization functions.
7004 @item HAS_INIT_SECTION
7005 @findex HAS_INIT_SECTION
7006 If defined, @code{main} will not call @code{__main} as described above.
7007 This macro should be defined for systems that control start-up code
7008 on a symbol-by-symbol basis, such as OSF/1, and should not
7009 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7011 @item LD_INIT_SWITCH
7012 @findex LD_INIT_SWITCH
7013 If defined, a C string constant for a switch that tells the linker that
7014 the following symbol is an initialization routine.
7016 @item LD_FINI_SWITCH
7017 @findex LD_FINI_SWITCH
7018 If defined, a C string constant for a switch that tells the linker that
7019 the following symbol is a finalization routine.
7021 @item COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7022 If defined, a C statement that will write a function that can be
7023 automatically called when a shared library is loaded. The function
7024 should call @var{func}, which takes no arguments. If not defined, and
7025 the object format requires an explicit initialization function, then a
7026 function called @code{_GLOBAL__DI} will be generated.
7028 This function and the following one are used by collect2 when linking a
7029 shared library that needs constructors or destructors, or has DWARF2
7030 exception tables embedded in the code.
7032 @item COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7033 If defined, a C statement that will write a function that can be
7034 automatically called when a shared library is unloaded. The function
7035 should call @var{func}, which takes no arguments. If not defined, and
7036 the object format requires an explicit finalization function, then a
7037 function called @code{_GLOBAL__DD} will be generated.
7040 @findex INVOKE__main
7041 If defined, @code{main} will call @code{__main} despite the presence of
7042 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7043 where the init section is not actually run automatically, but is still
7044 useful for collecting the lists of constructors and destructors.
7046 @item SUPPORTS_INIT_PRIORITY
7047 @findex SUPPORTS_INIT_PRIORITY
7048 If nonzero, the C++ @code{init_priority} attribute is supported and the
7049 compiler should emit instructions to control the order of initialization
7050 of objects. If zero, the compiler will issue an error message upon
7051 encountering an @code{init_priority} attribute.
7054 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7055 This value is true if the target supports some ``native'' method of
7056 collecting constructors and destructors to be run at startup and exit.
7057 It is false if we must use @command{collect2}.
7060 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7061 If defined, a function that outputs assembler code to arrange to call
7062 the function referenced by @var{symbol} at initialization time.
7064 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7065 no arguments and with no return value. If the target supports initialization
7066 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7067 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7069 If this macro is not defined by the target, a suitable default will
7070 be chosen if (1) the target supports arbitrary section names, (2) the
7071 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7075 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7076 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7077 functions rather than initialization functions.
7080 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7081 generated for the generated object file will have static linkage.
7083 If your system uses @command{collect2} as the means of processing
7084 constructors, then that program normally uses @command{nm} to scan
7085 an object file for constructor functions to be called.
7087 On certain kinds of systems, you can define these macros to make
7088 @command{collect2} work faster (and, in some cases, make it work at all):
7091 @findex OBJECT_FORMAT_COFF
7092 @item OBJECT_FORMAT_COFF
7093 Define this macro if the system uses COFF (Common Object File Format)
7094 object files, so that @command{collect2} can assume this format and scan
7095 object files directly for dynamic constructor/destructor functions.
7097 @findex OBJECT_FORMAT_ROSE
7098 @item OBJECT_FORMAT_ROSE
7099 Define this macro if the system uses ROSE format object files, so that
7100 @command{collect2} can assume this format and scan object files directly
7101 for dynamic constructor/destructor functions.
7103 These macros are effective only in a native compiler; @command{collect2} as
7104 part of a cross compiler always uses @command{nm} for the target machine.
7106 @findex REAL_NM_FILE_NAME
7107 @item REAL_NM_FILE_NAME
7108 Define this macro as a C string constant containing the file name to use
7109 to execute @command{nm}. The default is to search the path normally for
7112 If your system supports shared libraries and has a program to list the
7113 dynamic dependencies of a given library or executable, you can define
7114 these macros to enable support for running initialization and
7115 termination functions in shared libraries:
7119 Define this macro to a C string constant containing the name of the program
7120 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7122 @findex PARSE_LDD_OUTPUT
7123 @item PARSE_LDD_OUTPUT (@var{ptr})
7124 Define this macro to be C code that extracts filenames from the output
7125 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7126 of type @code{char *} that points to the beginning of a line of output
7127 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7128 code must advance @var{ptr} to the beginning of the filename on that
7129 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7132 @node Instruction Output
7133 @subsection Output of Assembler Instructions
7135 @c prevent bad page break with this line
7136 This describes assembler instruction output.
7139 @findex REGISTER_NAMES
7140 @item REGISTER_NAMES
7141 A C initializer containing the assembler's names for the machine
7142 registers, each one as a C string constant. This is what translates
7143 register numbers in the compiler into assembler language.
7145 @findex ADDITIONAL_REGISTER_NAMES
7146 @item ADDITIONAL_REGISTER_NAMES
7147 If defined, a C initializer for an array of structures containing a name
7148 and a register number. This macro defines additional names for hard
7149 registers, thus allowing the @code{asm} option in declarations to refer
7150 to registers using alternate names.
7152 @findex ASM_OUTPUT_OPCODE
7153 @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7154 Define this macro if you are using an unusual assembler that
7155 requires different names for the machine instructions.
7157 The definition is a C statement or statements which output an
7158 assembler instruction opcode to the stdio stream @var{stream}. The
7159 macro-operand @var{ptr} is a variable of type @code{char *} which
7160 points to the opcode name in its ``internal'' form---the form that is
7161 written in the machine description. The definition should output the
7162 opcode name to @var{stream}, performing any translation you desire, and
7163 increment the variable @var{ptr} to point at the end of the opcode
7164 so that it will not be output twice.
7166 In fact, your macro definition may process less than the entire opcode
7167 name, or more than the opcode name; but if you want to process text
7168 that includes @samp{%}-sequences to substitute operands, you must take
7169 care of the substitution yourself. Just be sure to increment
7170 @var{ptr} over whatever text should not be output normally.
7172 @findex recog_data.operand
7173 If you need to look at the operand values, they can be found as the
7174 elements of @code{recog_data.operand}.
7176 If the macro definition does nothing, the instruction is output
7179 @findex FINAL_PRESCAN_INSN
7180 @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7181 If defined, a C statement to be executed just prior to the output of
7182 assembler code for @var{insn}, to modify the extracted operands so
7183 they will be output differently.
7185 Here the argument @var{opvec} is the vector containing the operands
7186 extracted from @var{insn}, and @var{noperands} is the number of
7187 elements of the vector which contain meaningful data for this insn.
7188 The contents of this vector are what will be used to convert the insn
7189 template into assembler code, so you can change the assembler output
7190 by changing the contents of the vector.
7192 This macro is useful when various assembler syntaxes share a single
7193 file of instruction patterns; by defining this macro differently, you
7194 can cause a large class of instructions to be output differently (such
7195 as with rearranged operands). Naturally, variations in assembler
7196 syntax affecting individual insn patterns ought to be handled by
7197 writing conditional output routines in those patterns.
7199 If this macro is not defined, it is equivalent to a null statement.
7201 @findex FINAL_PRESCAN_LABEL
7202 @item FINAL_PRESCAN_LABEL
7203 If defined, @code{FINAL_PRESCAN_INSN} will be called on each
7204 @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and
7205 @var{noperands} will be zero.
7207 @findex PRINT_OPERAND
7208 @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
7209 A C compound statement to output to stdio stream @var{stream} the
7210 assembler syntax for an instruction operand @var{x}. @var{x} is an
7213 @var{code} is a value that can be used to specify one of several ways
7214 of printing the operand. It is used when identical operands must be
7215 printed differently depending on the context. @var{code} comes from
7216 the @samp{%} specification that was used to request printing of the
7217 operand. If the specification was just @samp{%@var{digit}} then
7218 @var{code} is 0; if the specification was @samp{%@var{ltr}
7219 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
7222 If @var{x} is a register, this macro should print the register's name.
7223 The names can be found in an array @code{reg_names} whose type is
7224 @code{char *[]}. @code{reg_names} is initialized from
7225 @code{REGISTER_NAMES}.
7227 When the machine description has a specification @samp{%@var{punct}}
7228 (a @samp{%} followed by a punctuation character), this macro is called
7229 with a null pointer for @var{x} and the punctuation character for
7232 @findex PRINT_OPERAND_PUNCT_VALID_P
7233 @item PRINT_OPERAND_PUNCT_VALID_P (@var{code})
7234 A C expression which evaluates to true if @var{code} is a valid
7235 punctuation character for use in the @code{PRINT_OPERAND} macro. If
7236 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
7237 punctuation characters (except for the standard one, @samp{%}) are used
7240 @findex PRINT_OPERAND_ADDRESS
7241 @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
7242 A C compound statement to output to stdio stream @var{stream} the
7243 assembler syntax for an instruction operand that is a memory reference
7244 whose address is @var{x}. @var{x} is an RTL expression.
7246 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
7247 On some machines, the syntax for a symbolic address depends on the
7248 section that the address refers to. On these machines, define the hook
7249 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
7250 @code{symbol_ref}, and then check for it here. @xref{Assembler Format}.
7252 @findex DBR_OUTPUT_SEQEND
7253 @findex dbr_sequence_length
7254 @item DBR_OUTPUT_SEQEND(@var{file})
7255 A C statement, to be executed after all slot-filler instructions have
7256 been output. If necessary, call @code{dbr_sequence_length} to
7257 determine the number of slots filled in a sequence (zero if not
7258 currently outputting a sequence), to decide how many no-ops to output,
7261 Don't define this macro if it has nothing to do, but it is helpful in
7262 reading assembly output if the extent of the delay sequence is made
7263 explicit (e.g.@: with white space).
7265 @findex final_sequence
7266 Note that output routines for instructions with delay slots must be
7267 prepared to deal with not being output as part of a sequence
7268 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
7269 found.) The variable @code{final_sequence} is null when not
7270 processing a sequence, otherwise it contains the @code{sequence} rtx
7273 @findex REGISTER_PREFIX
7274 @findex LOCAL_LABEL_PREFIX
7275 @findex USER_LABEL_PREFIX
7276 @findex IMMEDIATE_PREFIX
7278 @item REGISTER_PREFIX
7279 @itemx LOCAL_LABEL_PREFIX
7280 @itemx USER_LABEL_PREFIX
7281 @itemx IMMEDIATE_PREFIX
7282 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
7283 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
7284 @file{final.c}). These are useful when a single @file{md} file must
7285 support multiple assembler formats. In that case, the various @file{tm.h}
7286 files can define these macros differently.
7288 @item ASM_FPRINTF_EXTENSIONS(@var{file}, @var{argptr}, @var{format})
7289 @findex ASM_FPRINTF_EXTENSIONS
7290 If defined this macro should expand to a series of @code{case}
7291 statements which will be parsed inside the @code{switch} statement of
7292 the @code{asm_fprintf} function. This allows targets to define extra
7293 printf formats which may useful when generating their assembler
7294 statements. Note that upper case letters are reserved for future
7295 generic extensions to asm_fprintf, and so are not available to target
7296 specific code. The output file is given by the parameter @var{file}.
7297 The varargs input pointer is @var{argptr} and the rest of the format
7298 string, starting the character after the one that is being switched
7299 upon, is pointed to by @var{format}.
7301 @findex ASSEMBLER_DIALECT
7302 @item ASSEMBLER_DIALECT
7303 If your target supports multiple dialects of assembler language (such as
7304 different opcodes), define this macro as a C expression that gives the
7305 numeric index of the assembler language dialect to use, with zero as the
7308 If this macro is defined, you may use constructs of the form
7310 @samp{@{option0|option1|option2@dots{}@}}
7313 in the output templates of patterns (@pxref{Output Template}) or in the
7314 first argument of @code{asm_fprintf}. This construct outputs
7315 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
7316 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
7317 within these strings retain their usual meaning. If there are fewer
7318 alternatives within the braces than the value of
7319 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
7321 If you do not define this macro, the characters @samp{@{}, @samp{|} and
7322 @samp{@}} do not have any special meaning when used in templates or
7323 operands to @code{asm_fprintf}.
7325 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
7326 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
7327 the variations in assembler language syntax with that mechanism. Define
7328 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
7329 if the syntax variant are larger and involve such things as different
7330 opcodes or operand order.
7332 @findex ASM_OUTPUT_REG_PUSH
7333 @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
7334 A C expression to output to @var{stream} some assembler code
7335 which will push hard register number @var{regno} onto the stack.
7336 The code need not be optimal, since this macro is used only when
7339 @findex ASM_OUTPUT_REG_POP
7340 @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
7341 A C expression to output to @var{stream} some assembler code
7342 which will pop hard register number @var{regno} off of the stack.
7343 The code need not be optimal, since this macro is used only when
7347 @node Dispatch Tables
7348 @subsection Output of Dispatch Tables
7350 @c prevent bad page break with this line
7351 This concerns dispatch tables.
7354 @cindex dispatch table
7355 @findex ASM_OUTPUT_ADDR_DIFF_ELT
7356 @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
7357 A C statement to output to the stdio stream @var{stream} an assembler
7358 pseudo-instruction to generate a difference between two labels.
7359 @var{value} and @var{rel} are the numbers of two internal labels. The
7360 definitions of these labels are output using
7361 @code{ASM_OUTPUT_INTERNAL_LABEL}, and they must be printed in the same
7362 way here. For example,
7365 fprintf (@var{stream}, "\t.word L%d-L%d\n",
7366 @var{value}, @var{rel})
7369 You must provide this macro on machines where the addresses in a
7370 dispatch table are relative to the table's own address. If defined, GCC
7371 will also use this macro on all machines when producing PIC@.
7372 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
7373 mode and flags can be read.
7375 @findex ASM_OUTPUT_ADDR_VEC_ELT
7376 @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
7377 This macro should be provided on machines where the addresses
7378 in a dispatch table are absolute.
7380 The definition should be a C statement to output to the stdio stream
7381 @var{stream} an assembler pseudo-instruction to generate a reference to
7382 a label. @var{value} is the number of an internal label whose
7383 definition is output using @code{ASM_OUTPUT_INTERNAL_LABEL}.
7387 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
7390 @findex ASM_OUTPUT_CASE_LABEL
7391 @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
7392 Define this if the label before a jump-table needs to be output
7393 specially. The first three arguments are the same as for
7394 @code{ASM_OUTPUT_INTERNAL_LABEL}; the fourth argument is the
7395 jump-table which follows (a @code{jump_insn} containing an
7396 @code{addr_vec} or @code{addr_diff_vec}).
7398 This feature is used on system V to output a @code{swbeg} statement
7401 If this macro is not defined, these labels are output with
7402 @code{ASM_OUTPUT_INTERNAL_LABEL}.
7404 @findex ASM_OUTPUT_CASE_END
7405 @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
7406 Define this if something special must be output at the end of a
7407 jump-table. The definition should be a C statement to be executed
7408 after the assembler code for the table is written. It should write
7409 the appropriate code to stdio stream @var{stream}. The argument
7410 @var{table} is the jump-table insn, and @var{num} is the label-number
7411 of the preceding label.
7413 If this macro is not defined, nothing special is output at the end of
7417 @node Exception Region Output
7418 @subsection Assembler Commands for Exception Regions
7420 @c prevent bad page break with this line
7422 This describes commands marking the start and the end of an exception
7426 @findex EH_FRAME_SECTION_NAME
7427 @item EH_FRAME_SECTION_NAME
7428 If defined, a C string constant for the name of the section containing
7429 exception handling frame unwind information. If not defined, GCC will
7430 provide a default definition if the target supports named sections.
7431 @file{crtstuff.c} uses this macro to switch to the appropriate section.
7433 You should define this symbol if your target supports DWARF 2 frame
7434 unwind information and the default definition does not work.
7436 @findex EH_FRAME_IN_DATA_SECTION
7437 @item EH_FRAME_IN_DATA_SECTION
7438 If defined, DWARF 2 frame unwind information will be placed in the
7439 data section even though the target supports named sections. This
7440 might be necessary, for instance, if the system linker does garbage
7441 collection and sections cannot be marked as not to be collected.
7443 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
7446 @findex MASK_RETURN_ADDR
7447 @item MASK_RETURN_ADDR
7448 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
7449 that it does not contain any extraneous set bits in it.
7451 @findex DWARF2_UNWIND_INFO
7452 @item DWARF2_UNWIND_INFO
7453 Define this macro to 0 if your target supports DWARF 2 frame unwind
7454 information, but it does not yet work with exception handling.
7455 Otherwise, if your target supports this information (if it defines
7456 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
7457 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
7460 If this macro is defined to 1, the DWARF 2 unwinder will be the default
7461 exception handling mechanism; otherwise, @code{setjmp}/@code{longjmp} will be used by
7464 If this macro is defined to anything, the DWARF 2 unwinder will be used
7465 instead of inline unwinders and @code{__unwind_function} in the non-@code{setjmp} case.
7467 @findex DWARF_CIE_DATA_ALIGNMENT
7468 @item DWARF_CIE_DATA_ALIGNMENT
7469 This macro need only be defined if the target might save registers in the
7470 function prologue at an offset to the stack pointer that is not aligned to
7471 @code{UNITS_PER_WORD}. The definition should be the negative minimum
7472 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
7473 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
7474 the target supports DWARF 2 frame unwind information.
7478 @deftypefn {Target Hook} void TARGET_ASM_EXCEPTION_SECTION ()
7479 If defined, a function that switches to the section in which the main
7480 exception table is to be placed (@pxref{Sections}). The default is a
7481 function that switches to a section named @code{.gcc_except_table} on
7482 machines that support named sections via
7483 @code{TARGET_ASM_NAMED_SECTION}, otherwise if @option{-fpic} or
7484 @option{-fPIC} is in effect, the @code{data_section}, otherwise the
7485 @code{readonly_data_section}.
7488 @deftypefn {Target Hook} void TARGET_ASM_EH_FRAME_SECTION ()
7489 If defined, a function that switches to the section in which the DWARF 2
7490 frame unwind information to be placed (@pxref{Sections}). The default
7491 is a function that outputs a standard GAS section directive, if
7492 @code{EH_FRAME_SECTION_NAME} is defined, or else a data section
7493 directive followed by a synthetic label.
7496 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
7497 Contains the value true if the target should add a zero word onto the
7498 end of a Dwarf-2 frame info section when used for exception handling.
7499 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
7503 @node Alignment Output
7504 @subsection Assembler Commands for Alignment
7506 @c prevent bad page break with this line
7507 This describes commands for alignment.
7511 @item JUMP_ALIGN (@var{label})
7512 The alignment (log base 2) to put in front of @var{label}, which is
7513 a common destination of jumps and has no fallthru incoming edge.
7515 This macro need not be defined if you don't want any special alignment
7516 to be done at such a time. Most machine descriptions do not currently
7519 Unless it's necessary to inspect the @var{label} parameter, it is better
7520 to set the variable @var{align_jumps} in the target's
7521 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7522 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
7524 @findex LABEL_ALIGN_AFTER_BARRIER
7525 @item LABEL_ALIGN_AFTER_BARRIER (@var{label})
7526 The alignment (log base 2) to put in front of @var{label}, which follows
7529 This macro need not be defined if you don't want any special alignment
7530 to be done at such a time. Most machine descriptions do not currently
7533 @findex LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7534 @item LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7535 The maximum number of bytes to skip when applying
7536 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
7537 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7540 @item LOOP_ALIGN (@var{label})
7541 The alignment (log base 2) to put in front of @var{label}, which follows
7542 a @code{NOTE_INSN_LOOP_BEG} note.
7544 This macro need not be defined if you don't want any special alignment
7545 to be done at such a time. Most machine descriptions do not currently
7548 Unless it's necessary to inspect the @var{label} parameter, it is better
7549 to set the variable @code{align_loops} in the target's
7550 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7551 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
7553 @findex LOOP_ALIGN_MAX_SKIP
7554 @item LOOP_ALIGN_MAX_SKIP
7555 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
7556 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7559 @item LABEL_ALIGN (@var{label})
7560 The alignment (log base 2) to put in front of @var{label}.
7561 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
7562 the maximum of the specified values is used.
7564 Unless it's necessary to inspect the @var{label} parameter, it is better
7565 to set the variable @code{align_labels} in the target's
7566 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7567 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
7569 @findex LABEL_ALIGN_MAX_SKIP
7570 @item LABEL_ALIGN_MAX_SKIP
7571 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
7572 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7574 @findex ASM_OUTPUT_SKIP
7575 @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
7576 A C statement to output to the stdio stream @var{stream} an assembler
7577 instruction to advance the location counter by @var{nbytes} bytes.
7578 Those bytes should be zero when loaded. @var{nbytes} will be a C
7579 expression of type @code{int}.
7581 @findex ASM_NO_SKIP_IN_TEXT
7582 @item ASM_NO_SKIP_IN_TEXT
7583 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
7584 text section because it fails to put zeros in the bytes that are skipped.
7585 This is true on many Unix systems, where the pseudo--op to skip bytes
7586 produces no-op instructions rather than zeros when used in the text
7589 @findex ASM_OUTPUT_ALIGN
7590 @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
7591 A C statement to output to the stdio stream @var{stream} an assembler
7592 command to advance the location counter to a multiple of 2 to the
7593 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
7595 @findex ASM_OUTPUT_ALIGN_WITH_NOP
7596 @item ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
7597 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
7598 for padding, if necessary.
7600 @findex ASM_OUTPUT_MAX_SKIP_ALIGN
7601 @item ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
7602 A C statement to output to the stdio stream @var{stream} an assembler
7603 command to advance the location counter to a multiple of 2 to the
7604 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
7605 satisfy the alignment request. @var{power} and @var{max_skip} will be
7606 a C expression of type @code{int}.
7610 @node Debugging Info
7611 @section Controlling Debugging Information Format
7613 @c prevent bad page break with this line
7614 This describes how to specify debugging information.
7617 * All Debuggers:: Macros that affect all debugging formats uniformly.
7618 * DBX Options:: Macros enabling specific options in DBX format.
7619 * DBX Hooks:: Hook macros for varying DBX format.
7620 * File Names and DBX:: Macros controlling output of file names in DBX format.
7621 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
7622 * VMS Debug:: Macros for VMS debug format.
7626 @subsection Macros Affecting All Debugging Formats
7628 @c prevent bad page break with this line
7629 These macros affect all debugging formats.
7632 @findex DBX_REGISTER_NUMBER
7633 @item DBX_REGISTER_NUMBER (@var{regno})
7634 A C expression that returns the DBX register number for the compiler
7635 register number @var{regno}. In the default macro provided, the value
7636 of this expression will be @var{regno} itself. But sometimes there are
7637 some registers that the compiler knows about and DBX does not, or vice
7638 versa. In such cases, some register may need to have one number in the
7639 compiler and another for DBX@.
7641 If two registers have consecutive numbers inside GCC, and they can be
7642 used as a pair to hold a multiword value, then they @emph{must} have
7643 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
7644 Otherwise, debuggers will be unable to access such a pair, because they
7645 expect register pairs to be consecutive in their own numbering scheme.
7647 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
7648 does not preserve register pairs, then what you must do instead is
7649 redefine the actual register numbering scheme.
7651 @findex DEBUGGER_AUTO_OFFSET
7652 @item DEBUGGER_AUTO_OFFSET (@var{x})
7653 A C expression that returns the integer offset value for an automatic
7654 variable having address @var{x} (an RTL expression). The default
7655 computation assumes that @var{x} is based on the frame-pointer and
7656 gives the offset from the frame-pointer. This is required for targets
7657 that produce debugging output for DBX or COFF-style debugging output
7658 for SDB and allow the frame-pointer to be eliminated when the
7659 @option{-g} options is used.
7661 @findex DEBUGGER_ARG_OFFSET
7662 @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
7663 A C expression that returns the integer offset value for an argument
7664 having address @var{x} (an RTL expression). The nominal offset is
7667 @findex PREFERRED_DEBUGGING_TYPE
7668 @item PREFERRED_DEBUGGING_TYPE
7669 A C expression that returns the type of debugging output GCC should
7670 produce when the user specifies just @option{-g}. Define
7671 this if you have arranged for GCC to support more than one format of
7672 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7673 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
7674 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
7676 When the user specifies @option{-ggdb}, GCC normally also uses the
7677 value of this macro to select the debugging output format, but with two
7678 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined and
7679 @code{LINKER_DOES_NOT_WORK_WITH_DWARF2} is not defined, GCC uses the
7680 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7681 defined, GCC uses @code{DBX_DEBUG}.
7683 The value of this macro only affects the default debugging output; the
7684 user can always get a specific type of output by using @option{-gstabs},
7685 @option{-gcoff}, @option{-gdwarf-1}, @option{-gdwarf-2}, @option{-gxcoff},
7690 @subsection Specific Options for DBX Output
7692 @c prevent bad page break with this line
7693 These are specific options for DBX output.
7696 @findex DBX_DEBUGGING_INFO
7697 @item DBX_DEBUGGING_INFO
7698 Define this macro if GCC should produce debugging output for DBX
7699 in response to the @option{-g} option.
7701 @findex XCOFF_DEBUGGING_INFO
7702 @item XCOFF_DEBUGGING_INFO
7703 Define this macro if GCC should produce XCOFF format debugging output
7704 in response to the @option{-g} option. This is a variant of DBX format.
7706 @findex DEFAULT_GDB_EXTENSIONS
7707 @item DEFAULT_GDB_EXTENSIONS
7708 Define this macro to control whether GCC should by default generate
7709 GDB's extended version of DBX debugging information (assuming DBX-format
7710 debugging information is enabled at all). If you don't define the
7711 macro, the default is 1: always generate the extended information
7712 if there is any occasion to.
7714 @findex DEBUG_SYMS_TEXT
7715 @item DEBUG_SYMS_TEXT
7716 Define this macro if all @code{.stabs} commands should be output while
7717 in the text section.
7719 @findex ASM_STABS_OP
7721 A C string constant, including spacing, naming the assembler pseudo op to
7722 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
7723 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
7724 applies only to DBX debugging information format.
7726 @findex ASM_STABD_OP
7728 A C string constant, including spacing, naming the assembler pseudo op to
7729 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
7730 value is the current location. If you don't define this macro,
7731 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
7734 @findex ASM_STABN_OP
7736 A C string constant, including spacing, naming the assembler pseudo op to
7737 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
7738 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
7739 macro applies only to DBX debugging information format.
7741 @findex DBX_NO_XREFS
7743 Define this macro if DBX on your system does not support the construct
7744 @samp{xs@var{tagname}}. On some systems, this construct is used to
7745 describe a forward reference to a structure named @var{tagname}.
7746 On other systems, this construct is not supported at all.
7748 @findex DBX_CONTIN_LENGTH
7749 @item DBX_CONTIN_LENGTH
7750 A symbol name in DBX-format debugging information is normally
7751 continued (split into two separate @code{.stabs} directives) when it
7752 exceeds a certain length (by default, 80 characters). On some
7753 operating systems, DBX requires this splitting; on others, splitting
7754 must not be done. You can inhibit splitting by defining this macro
7755 with the value zero. You can override the default splitting-length by
7756 defining this macro as an expression for the length you desire.
7758 @findex DBX_CONTIN_CHAR
7759 @item DBX_CONTIN_CHAR
7760 Normally continuation is indicated by adding a @samp{\} character to
7761 the end of a @code{.stabs} string when a continuation follows. To use
7762 a different character instead, define this macro as a character
7763 constant for the character you want to use. Do not define this macro
7764 if backslash is correct for your system.
7766 @findex DBX_STATIC_STAB_DATA_SECTION
7767 @item DBX_STATIC_STAB_DATA_SECTION
7768 Define this macro if it is necessary to go to the data section before
7769 outputting the @samp{.stabs} pseudo-op for a non-global static
7772 @findex DBX_TYPE_DECL_STABS_CODE
7773 @item DBX_TYPE_DECL_STABS_CODE
7774 The value to use in the ``code'' field of the @code{.stabs} directive
7775 for a typedef. The default is @code{N_LSYM}.
7777 @findex DBX_STATIC_CONST_VAR_CODE
7778 @item DBX_STATIC_CONST_VAR_CODE
7779 The value to use in the ``code'' field of the @code{.stabs} directive
7780 for a static variable located in the text section. DBX format does not
7781 provide any ``right'' way to do this. The default is @code{N_FUN}.
7783 @findex DBX_REGPARM_STABS_CODE
7784 @item DBX_REGPARM_STABS_CODE
7785 The value to use in the ``code'' field of the @code{.stabs} directive
7786 for a parameter passed in registers. DBX format does not provide any
7787 ``right'' way to do this. The default is @code{N_RSYM}.
7789 @findex DBX_REGPARM_STABS_LETTER
7790 @item DBX_REGPARM_STABS_LETTER
7791 The letter to use in DBX symbol data to identify a symbol as a parameter
7792 passed in registers. DBX format does not customarily provide any way to
7793 do this. The default is @code{'P'}.
7795 @findex DBX_MEMPARM_STABS_LETTER
7796 @item DBX_MEMPARM_STABS_LETTER
7797 The letter to use in DBX symbol data to identify a symbol as a stack
7798 parameter. The default is @code{'p'}.
7800 @findex DBX_FUNCTION_FIRST
7801 @item DBX_FUNCTION_FIRST
7802 Define this macro if the DBX information for a function and its
7803 arguments should precede the assembler code for the function. Normally,
7804 in DBX format, the debugging information entirely follows the assembler
7807 @findex DBX_LBRAC_FIRST
7808 @item DBX_LBRAC_FIRST
7809 Define this macro if the @code{N_LBRAC} symbol for a block should
7810 precede the debugging information for variables and functions defined in
7811 that block. Normally, in DBX format, the @code{N_LBRAC} symbol comes
7814 @findex DBX_BLOCKS_FUNCTION_RELATIVE
7815 @item DBX_BLOCKS_FUNCTION_RELATIVE
7816 Define this macro if the value of a symbol describing the scope of a
7817 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
7818 of the enclosing function. Normally, GCC uses an absolute address.
7820 @findex DBX_USE_BINCL
7822 Define this macro if GCC should generate @code{N_BINCL} and
7823 @code{N_EINCL} stabs for included header files, as on Sun systems. This
7824 macro also directs GCC to output a type number as a pair of a file
7825 number and a type number within the file. Normally, GCC does not
7826 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
7827 number for a type number.
7831 @subsection Open-Ended Hooks for DBX Format
7833 @c prevent bad page break with this line
7834 These are hooks for DBX format.
7837 @findex DBX_OUTPUT_LBRAC
7838 @item DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
7839 Define this macro to say how to output to @var{stream} the debugging
7840 information for the start of a scope level for variable names. The
7841 argument @var{name} is the name of an assembler symbol (for use with
7842 @code{assemble_name}) whose value is the address where the scope begins.
7844 @findex DBX_OUTPUT_RBRAC
7845 @item DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
7846 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
7848 @findex DBX_OUTPUT_NFUN
7849 @item DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
7850 Define this macro if the target machine requires special handling to
7851 output an @code{N_FUN} entry for the function @var{decl}.
7853 @findex DBX_OUTPUT_ENUM
7854 @item DBX_OUTPUT_ENUM (@var{stream}, @var{type})
7855 Define this macro if the target machine requires special handling to
7856 output an enumeration type. The definition should be a C statement
7857 (sans semicolon) to output the appropriate information to @var{stream}
7858 for the type @var{type}.
7860 @findex DBX_OUTPUT_FUNCTION_END
7861 @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
7862 Define this macro if the target machine requires special output at the
7863 end of the debugging information for a function. The definition should
7864 be a C statement (sans semicolon) to output the appropriate information
7865 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
7868 @findex DBX_OUTPUT_STANDARD_TYPES
7869 @item DBX_OUTPUT_STANDARD_TYPES (@var{syms})
7870 Define this macro if you need to control the order of output of the
7871 standard data types at the beginning of compilation. The argument
7872 @var{syms} is a @code{tree} which is a chain of all the predefined
7873 global symbols, including names of data types.
7875 Normally, DBX output starts with definitions of the types for integers
7876 and characters, followed by all the other predefined types of the
7877 particular language in no particular order.
7879 On some machines, it is necessary to output different particular types
7880 first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
7881 those symbols in the necessary order. Any predefined types that you
7882 don't explicitly output will be output afterward in no particular order.
7884 Be careful not to define this macro so that it works only for C@. There
7885 are no global variables to access most of the built-in types, because
7886 another language may have another set of types. The way to output a
7887 particular type is to look through @var{syms} to see if you can find it.
7893 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7894 if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
7896 dbxout_symbol (decl);
7902 This does nothing if the expected type does not exist.
7904 See the function @code{init_decl_processing} in @file{c-decl.c} to find
7905 the names to use for all the built-in C types.
7907 Here is another way of finding a particular type:
7909 @c this is still overfull. --mew 10feb93
7913 for (decl = syms; decl; decl = TREE_CHAIN (decl))
7914 if (TREE_CODE (decl) == TYPE_DECL
7915 && (TREE_CODE (TREE_TYPE (decl))
7917 && TYPE_PRECISION (TREE_TYPE (decl)) == 16
7918 && TYPE_UNSIGNED (TREE_TYPE (decl)))
7920 /* @r{This must be @code{unsigned short}.} */
7921 dbxout_symbol (decl);
7927 @findex NO_DBX_FUNCTION_END
7928 @item NO_DBX_FUNCTION_END
7929 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
7930 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
7931 On those machines, define this macro to turn this feature off without
7932 disturbing the rest of the gdb extensions.
7936 @node File Names and DBX
7937 @subsection File Names in DBX Format
7939 @c prevent bad page break with this line
7940 This describes file names in DBX format.
7943 @findex DBX_WORKING_DIRECTORY
7944 @item DBX_WORKING_DIRECTORY
7945 Define this if DBX wants to have the current directory recorded in each
7948 Note that the working directory is always recorded if GDB extensions are
7951 @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME
7952 @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
7953 A C statement to output DBX debugging information to the stdio stream
7954 @var{stream} which indicates that file @var{name} is the main source
7955 file---the file specified as the input file for compilation.
7956 This macro is called only once, at the beginning of compilation.
7958 This macro need not be defined if the standard form of output
7959 for DBX debugging information is appropriate.
7961 @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY
7962 @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
7963 A C statement to output DBX debugging information to the stdio stream
7964 @var{stream} which indicates that the current directory during
7965 compilation is named @var{name}.
7967 This macro need not be defined if the standard form of output
7968 for DBX debugging information is appropriate.
7970 @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END
7971 @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
7972 A C statement to output DBX debugging information at the end of
7973 compilation of the main source file @var{name}.
7975 If you don't define this macro, nothing special is output at the end
7976 of compilation, which is correct for most machines.
7978 @findex DBX_OUTPUT_SOURCE_FILENAME
7979 @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7980 A C statement to output DBX debugging information to the stdio stream
7981 @var{stream} which indicates that file @var{name} is the current source
7982 file. This output is generated each time input shifts to a different
7983 source file as a result of @samp{#include}, the end of an included file,
7984 or a @samp{#line} command.
7986 This macro need not be defined if the standard form of output
7987 for DBX debugging information is appropriate.
7992 @subsection Macros for SDB and DWARF Output
7994 @c prevent bad page break with this line
7995 Here are macros for SDB and DWARF output.
7998 @findex SDB_DEBUGGING_INFO
7999 @item SDB_DEBUGGING_INFO
8000 Define this macro if GCC should produce COFF-style debugging output
8001 for SDB in response to the @option{-g} option.
8003 @findex DWARF_DEBUGGING_INFO
8004 @item DWARF_DEBUGGING_INFO
8005 Define this macro if GCC should produce dwarf format debugging output
8006 in response to the @option{-g} option.
8008 @findex DWARF2_DEBUGGING_INFO
8009 @item DWARF2_DEBUGGING_INFO
8010 Define this macro if GCC should produce dwarf version 2 format
8011 debugging output in response to the @option{-g} option.
8013 To support optional call frame debugging information, you must also
8014 define @code{INCOMING_RETURN_ADDR_RTX} and either set
8015 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
8016 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
8017 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
8019 @findex DWARF2_FRAME_INFO
8020 @item DWARF2_FRAME_INFO
8021 Define this macro to a nonzero value if GCC should always output
8022 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
8023 (@pxref{Exception Region Output} is nonzero, GCC will output this
8024 information not matter how you define @code{DWARF2_FRAME_INFO}.
8026 @findex LINKER_DOES_NOT_WORK_WITH_DWARF2
8027 @item LINKER_DOES_NOT_WORK_WITH_DWARF2
8028 Define this macro if the linker does not work with Dwarf version 2.
8029 Normally, if the user specifies only @option{-ggdb} GCC will use Dwarf
8030 version 2 if available; this macro disables this. See the description
8031 of the @code{PREFERRED_DEBUGGING_TYPE} macro for more details.
8033 @findex DWARF2_GENERATE_TEXT_SECTION_LABEL
8034 @item DWARF2_GENERATE_TEXT_SECTION_LABEL
8035 By default, the Dwarf 2 debugging information generator will generate a
8036 label to mark the beginning of the text section. If it is better simply
8037 to use the name of the text section itself, rather than an explicit label,
8038 to indicate the beginning of the text section, define this macro to zero.
8040 @findex DWARF2_ASM_LINE_DEBUG_INFO
8041 @item DWARF2_ASM_LINE_DEBUG_INFO
8042 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8043 line debug info sections. This will result in much more compact line number
8044 tables, and hence is desirable if it works.
8046 @findex ASM_OUTPUT_DWARF_DELTA
8047 @item ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8048 A C statement to issue assembly directives that create a difference
8049 between the two given labels, using an integer of the given size.
8051 @findex ASM_OUTPUT_DWARF_OFFSET
8052 @item ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label})
8053 A C statement to issue assembly directives that create a
8054 section-relative reference to the given label, using an integer of the
8057 @findex ASM_OUTPUT_DWARF_PCREL
8058 @item ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8059 A C statement to issue assembly directives that create a self-relative
8060 reference to the given label, using an integer of the given size.
8062 @findex PUT_SDB_@dots{}
8063 @item PUT_SDB_@dots{}
8064 Define these macros to override the assembler syntax for the special
8065 SDB assembler directives. See @file{sdbout.c} for a list of these
8066 macros and their arguments. If the standard syntax is used, you need
8067 not define them yourself.
8071 Some assemblers do not support a semicolon as a delimiter, even between
8072 SDB assembler directives. In that case, define this macro to be the
8073 delimiter to use (usually @samp{\n}). It is not necessary to define
8074 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8077 @findex SDB_GENERATE_FAKE
8078 @item SDB_GENERATE_FAKE
8079 Define this macro to override the usual method of constructing a dummy
8080 name for anonymous structure and union types. See @file{sdbout.c} for
8083 @findex SDB_ALLOW_UNKNOWN_REFERENCES
8084 @item SDB_ALLOW_UNKNOWN_REFERENCES
8085 Define this macro to allow references to unknown structure,
8086 union, or enumeration tags to be emitted. Standard COFF does not
8087 allow handling of unknown references, MIPS ECOFF has support for
8090 @findex SDB_ALLOW_FORWARD_REFERENCES
8091 @item SDB_ALLOW_FORWARD_REFERENCES
8092 Define this macro to allow references to structure, union, or
8093 enumeration tags that have not yet been seen to be handled. Some
8094 assemblers choke if forward tags are used, while some require it.
8099 @subsection Macros for VMS Debug Format
8101 @c prevent bad page break with this line
8102 Here are macros for VMS debug format.
8105 @findex VMS_DEBUGGING_INFO
8106 @item VMS_DEBUGGING_INFO
8107 Define this macro if GCC should produce debugging output for VMS
8108 in response to the @option{-g} option. The default behavior for VMS
8109 is to generate minimal debug info for a traceback in the absence of
8110 @option{-g} unless explicitly overridden with @option{-g0}. This
8111 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8112 @code{OVERRIDE_OPTIONS}.
8115 @node Floating Point
8116 @section Cross Compilation and Floating Point
8117 @cindex cross compilation and floating point
8118 @cindex floating point and cross compilation
8120 While all modern machines use twos-complement representation for integers,
8121 there are a variety of representations for floating point numbers. This
8122 means that in a cross-compiler the representation of floating point numbers
8123 in the compiled program may be different from that used in the machine
8124 doing the compilation.
8126 Because different representation systems may offer different amounts of
8127 range and precision, all floating point constants must be represented in
8128 the target machine's format. Therefore, the cross compiler cannot
8129 safely use the host machine's floating point arithmetic; it must emulate
8130 the target's arithmetic. To ensure consistency, GCC always uses
8131 emulation to work with floating point values, even when the host and
8132 target floating point formats are identical.
8134 The following macros are provided by @file{real.h} for the compiler to
8135 use. All parts of the compiler which generate or optimize
8136 floating-point calculations must use these macros. They may evaluate
8137 their operands more than once, so operands must not have side effects.
8139 @defmac REAL_VALUE_TYPE
8140 The C data type to be used to hold a floating point value in the target
8141 machine's format. Typically this is a @code{struct} containing an
8142 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8146 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8147 Compares for equality the two values, @var{x} and @var{y}. If the target
8148 floating point format supports negative zeroes and/or NaNs,
8149 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8150 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8153 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8154 Tests whether @var{x} is less than @var{y}.
8157 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8158 Truncates @var{x} to a signed integer, rounding toward zero.
8161 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8162 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8163 @var{x} is negative, returns zero.
8166 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8167 Converts @var{string} into a floating point number in the target machine's
8168 representation for mode @var{mode}. This routine can handle both
8169 decimal and hexadecimal floating point constants, using the syntax
8170 defined by the C language for both.
8173 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8174 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8177 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8178 Determines whether @var{x} represents infinity (positive or negative).
8181 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8182 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8185 @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})
8186 Calculates an arithmetic operation on the two floating point values
8187 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8190 The operation to be performed is specified by @var{code}. Only the
8191 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8192 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8194 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8195 target's floating point format cannot represent infinity, it will call
8196 @code{abort}. Callers should check for this situation first, using
8197 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8200 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8201 Returns the negative of the floating point value @var{x}.
8204 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8205 Returns the absolute value of @var{x}.
8208 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8209 Truncates the floating point value @var{x} to fit in @var{mode}. The
8210 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8211 appropriate bit pattern to be output asa floating constant whose
8212 precision accords with mode @var{mode}.
8215 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8216 Converts a floating point value @var{x} into a double-precision integer
8217 which is then stored into @var{low} and @var{high}. If the value is not
8218 integral, it is truncated.
8221 @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})
8222 @findex REAL_VALUE_FROM_INT
8223 Converts a double-precision integer found in @var{low} and @var{high},
8224 into a floating point value which is then stored into @var{x}. The
8225 value is truncated to fit in mode @var{mode}.
8228 @node Mode Switching
8229 @section Mode Switching Instructions
8230 @cindex mode switching
8231 The following macros control mode switching optimizations:
8234 @findex OPTIMIZE_MODE_SWITCHING
8235 @item OPTIMIZE_MODE_SWITCHING (@var{entity})
8236 Define this macro if the port needs extra instructions inserted for mode
8237 switching in an optimizing compilation.
8239 For an example, the SH4 can perform both single and double precision
8240 floating point operations, but to perform a single precision operation,
8241 the FPSCR PR bit has to be cleared, while for a double precision
8242 operation, this bit has to be set. Changing the PR bit requires a general
8243 purpose register as a scratch register, hence these FPSCR sets have to
8244 be inserted before reload, i.e.@: you can't put this into instruction emitting
8245 or @code{MACHINE_DEPENDENT_REORG}.
8247 You can have multiple entities that are mode-switched, and select at run time
8248 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
8249 return nonzero for any @var{entity} that needs mode-switching.
8250 If you define this macro, you also have to define
8251 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
8252 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
8253 @code{NORMAL_MODE} is optional.
8255 @findex NUM_MODES_FOR_MODE_SWITCHING
8256 @item NUM_MODES_FOR_MODE_SWITCHING
8257 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
8258 initializer for an array of integers. Each initializer element
8259 N refers to an entity that needs mode switching, and specifies the number
8260 of different modes that might need to be set for this entity.
8261 The position of the initializer in the initializer - starting counting at
8262 zero - determines the integer that is used to refer to the mode-switched
8264 In macros that take mode arguments / yield a mode result, modes are
8265 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
8266 switch is needed / supplied.
8269 @item MODE_NEEDED (@var{entity}, @var{insn})
8270 @var{entity} is an integer specifying a mode-switched entity. If
8271 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
8272 return an integer value not larger than the corresponding element in
8273 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
8274 be switched into prior to the execution of @var{insn}.
8277 @item NORMAL_MODE (@var{entity})
8278 If this macro is defined, it is evaluated for every @var{entity} that needs
8279 mode switching. It should evaluate to an integer, which is a mode that
8280 @var{entity} is assumed to be switched to at function entry and exit.
8282 @findex MODE_PRIORITY_TO_MODE
8283 @item MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
8284 This macro specifies the order in which modes for @var{entity} are processed.
8285 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
8286 lowest. The value of the macro should be an integer designating a mode
8287 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
8288 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
8289 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
8291 @findex EMIT_MODE_SET
8292 @item EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
8293 Generate one or more insns to set @var{entity} to @var{mode}.
8294 @var{hard_reg_live} is the set of hard registers live at the point where
8295 the insn(s) are to be inserted.
8298 @node Target Attributes
8299 @section Defining target-specific uses of @code{__attribute__}
8300 @cindex target attributes
8301 @cindex machine attributes
8302 @cindex attributes, target-specific
8304 Target-specific attributes may be defined for functions, data and types.
8305 These are described using the following target hooks; they also need to
8306 be documented in @file{extend.texi}.
8308 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
8309 If defined, this target hook points to an array of @samp{struct
8310 attribute_spec} (defined in @file{tree.h}) specifying the machine
8311 specific attributes for this target and some of the restrictions on the
8312 entities to which these attributes are applied and the arguments they
8316 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8317 If defined, this target hook is a function which returns zero if the attributes on
8318 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8319 and two if they are nearly compatible (which causes a warning to be
8320 generated). If this is not defined, machine-specific attributes are
8321 supposed always to be compatible.
8324 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
8325 If defined, this target hook is a function which assigns default attributes to
8326 newly defined @var{type}.
8329 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8330 Define this target hook if the merging of type attributes needs special
8331 handling. If defined, the result is a list of the combined
8332 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
8333 that @code{comptypes} has already been called and returned 1. This
8334 function may call @code{merge_attributes} to handle machine-independent
8338 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
8339 Define this target hook if the merging of decl attributes needs special
8340 handling. If defined, the result is a list of the combined
8341 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
8342 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
8343 when this is needed are when one attribute overrides another, or when an
8344 attribute is nullified by a subsequent definition. This function may
8345 call @code{merge_attributes} to handle machine-independent merging.
8347 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
8348 If the only target-specific handling you require is @samp{dllimport} for
8349 Windows targets, you should define the macro
8350 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. This links in a function
8351 called @code{merge_dllimport_decl_attributes} which can then be defined
8352 as the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. This is done
8353 in @file{i386/cygwin.h} and @file{i386/i386.c}, for example.
8356 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
8357 Define this target hook if you want to be able to add attributes to a decl
8358 when it is being created. This is normally useful for back ends which
8359 wish to implement a pragma by using the attributes which correspond to
8360 the pragma's effect. The @var{node} argument is the decl which is being
8361 created. The @var{attr_ptr} argument is a pointer to the attribute list
8362 for this decl. The list itself should not be modified, since it may be
8363 shared with other decls, but attributes may be chained on the head of
8364 the list and @code{*@var{attr_ptr}} modified to point to the new
8365 attributes, or a copy of the list may be made if further changes are
8369 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
8371 This target hook returns @code{true} if it is ok to inline @var{fndecl}
8372 into the current function, despite its having target-specific
8373 attributes, @code{false} otherwise. By default, if a function has a
8374 target specific attribute attached to it, it will not be inlined.
8377 @node MIPS Coprocessors
8378 @section Defining coprocessor specifics for MIPS targets.
8379 @cindex MIPS coprocessor-definition macros
8381 The MIPS specification allows MIPS implementations to have as many as 4
8382 coprocessors, each with as many as 32 private registers. gcc supports
8383 accessing these registers and transferring values between the registers
8384 and memory using asm-ized variables. For example:
8387 register unsigned int cp0count asm ("c0r1");
8393 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
8394 names may be added as described below, or the default names may be
8395 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
8397 Coprocessor registers are assumed to be epilogue-used; sets to them will
8398 be preserved even if it does not appear that the register is used again
8399 later in the function.
8401 Another note: according to the MIPS spec, coprocessor 1 (if present) is
8402 the FPU. One accesses COP1 registers through standard mips
8403 floating-point support; they are not included in this mechanism.
8405 There is one macro used in defining the MIPS coprocessor interface which
8406 you may want to override in subtargets; it is described below.
8410 @item ALL_COP_ADDITIONAL_REGISTER_NAMES
8411 @findex ALL_COP_ADDITIONAL_REGISTER_NAMES
8412 A comma-separated list (with leading comma) of pairs describing the
8413 alternate names of coprocessor registers. The format of each entry should be
8415 @{ @var{alternatename}, @var{register_number}@}
8422 @section Miscellaneous Parameters
8423 @cindex parameters, miscellaneous
8425 @c prevent bad page break with this line
8426 Here are several miscellaneous parameters.
8429 @item PREDICATE_CODES
8430 @findex PREDICATE_CODES
8431 Define this if you have defined special-purpose predicates in the file
8432 @file{@var{machine}.c}. This macro is called within an initializer of an
8433 array of structures. The first field in the structure is the name of a
8434 predicate and the second field is an array of rtl codes. For each
8435 predicate, list all rtl codes that can be in expressions matched by the
8436 predicate. The list should have a trailing comma. Here is an example
8437 of two entries in the list for a typical RISC machine:
8440 #define PREDICATE_CODES \
8441 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
8442 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
8445 Defining this macro does not affect the generated code (however,
8446 incorrect definitions that omit an rtl code that may be matched by the
8447 predicate can cause the compiler to malfunction). Instead, it allows
8448 the table built by @file{genrecog} to be more compact and efficient,
8449 thus speeding up the compiler. The most important predicates to include
8450 in the list specified by this macro are those used in the most insn
8453 For each predicate function named in @code{PREDICATE_CODES}, a
8454 declaration will be generated in @file{insn-codes.h}.
8456 @item SPECIAL_MODE_PREDICATES
8457 @findex SPECIAL_MODE_PREDICATES
8458 Define this if you have special predicates that know special things
8459 about modes. Genrecog will warn about certain forms of
8460 @code{match_operand} without a mode; if the operand predicate is
8461 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
8464 Here is an example from the IA-32 port (@code{ext_register_operand}
8465 specially checks for @code{HImode} or @code{SImode} in preparation
8466 for a byte extraction from @code{%ah} etc.).
8469 #define SPECIAL_MODE_PREDICATES \
8470 "ext_register_operand",
8473 @findex CASE_VECTOR_MODE
8474 @item CASE_VECTOR_MODE
8475 An alias for a machine mode name. This is the machine mode that
8476 elements of a jump-table should have.
8478 @findex CASE_VECTOR_SHORTEN_MODE
8479 @item CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
8480 Optional: return the preferred mode for an @code{addr_diff_vec}
8481 when the minimum and maximum offset are known. If you define this,
8482 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
8483 To make this work, you also have to define @code{INSN_ALIGN} and
8484 make the alignment for @code{addr_diff_vec} explicit.
8485 The @var{body} argument is provided so that the offset_unsigned and scale
8486 flags can be updated.
8488 @findex CASE_VECTOR_PC_RELATIVE
8489 @item CASE_VECTOR_PC_RELATIVE
8490 Define this macro to be a C expression to indicate when jump-tables
8491 should contain relative addresses. If jump-tables never contain
8492 relative addresses, then you need not define this macro.
8494 @findex CASE_DROPS_THROUGH
8495 @item CASE_DROPS_THROUGH
8496 Define this if control falls through a @code{case} insn when the index
8497 value is out of range. This means the specified default-label is
8498 actually ignored by the @code{case} insn proper.
8500 @findex CASE_VALUES_THRESHOLD
8501 @item CASE_VALUES_THRESHOLD
8502 Define this to be the smallest number of different values for which it
8503 is best to use a jump-table instead of a tree of conditional branches.
8504 The default is four for machines with a @code{casesi} instruction and
8505 five otherwise. This is best for most machines.
8507 @findex WORD_REGISTER_OPERATIONS
8508 @item WORD_REGISTER_OPERATIONS
8509 Define this macro if operations between registers with integral mode
8510 smaller than a word are always performed on the entire register.
8511 Most RISC machines have this property and most CISC machines do not.
8513 @findex LOAD_EXTEND_OP
8514 @item LOAD_EXTEND_OP (@var{mode})
8515 Define this macro to be a C expression indicating when insns that read
8516 memory in @var{mode}, an integral mode narrower than a word, set the
8517 bits outside of @var{mode} to be either the sign-extension or the
8518 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
8519 of @var{mode} for which the
8520 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
8521 @code{NIL} for other modes.
8523 This macro is not called with @var{mode} non-integral or with a width
8524 greater than or equal to @code{BITS_PER_WORD}, so you may return any
8525 value in this case. Do not define this macro if it would always return
8526 @code{NIL}. On machines where this macro is defined, you will normally
8527 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
8529 @findex SHORT_IMMEDIATES_SIGN_EXTEND
8530 @item SHORT_IMMEDIATES_SIGN_EXTEND
8531 Define this macro if loading short immediate values into registers sign
8534 @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC
8535 @item FIXUNS_TRUNC_LIKE_FIX_TRUNC
8536 Define this macro if the same instructions that convert a floating
8537 point number to a signed fixed point number also convert validly to an
8542 The maximum number of bytes that a single instruction can move quickly
8543 between memory and registers or between two memory locations.
8545 @findex MAX_MOVE_MAX
8547 The maximum number of bytes that a single instruction can move quickly
8548 between memory and registers or between two memory locations. If this
8549 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
8550 constant value that is the largest value that @code{MOVE_MAX} can have
8553 @findex SHIFT_COUNT_TRUNCATED
8554 @item SHIFT_COUNT_TRUNCATED
8555 A C expression that is nonzero if on this machine the number of bits
8556 actually used for the count of a shift operation is equal to the number
8557 of bits needed to represent the size of the object being shifted. When
8558 this macro is nonzero, the compiler will assume that it is safe to omit
8559 a sign-extend, zero-extend, and certain bitwise `and' instructions that
8560 truncates the count of a shift operation. On machines that have
8561 instructions that act on bit-fields at variable positions, which may
8562 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
8563 also enables deletion of truncations of the values that serve as
8564 arguments to bit-field instructions.
8566 If both types of instructions truncate the count (for shifts) and
8567 position (for bit-field operations), or if no variable-position bit-field
8568 instructions exist, you should define this macro.
8570 However, on some machines, such as the 80386 and the 680x0, truncation
8571 only applies to shift operations and not the (real or pretended)
8572 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
8573 such machines. Instead, add patterns to the @file{md} file that include
8574 the implied truncation of the shift instructions.
8576 You need not define this macro if it would always have the value of zero.
8578 @findex TRULY_NOOP_TRUNCATION
8579 @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
8580 A C expression which is nonzero if on this machine it is safe to
8581 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
8582 bits (where @var{outprec} is smaller than @var{inprec}) by merely
8583 operating on it as if it had only @var{outprec} bits.
8585 On many machines, this expression can be 1.
8587 @c rearranged this, removed the phrase "it is reported that". this was
8588 @c to fix an overfull hbox. --mew 10feb93
8589 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
8590 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
8591 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
8592 such cases may improve things.
8594 @findex STORE_FLAG_VALUE
8595 @item STORE_FLAG_VALUE
8596 A C expression describing the value returned by a comparison operator
8597 with an integral mode and stored by a store-flag instruction
8598 (@samp{s@var{cond}}) when the condition is true. This description must
8599 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
8600 comparison operators whose results have a @code{MODE_INT} mode.
8602 A value of 1 or @minus{}1 means that the instruction implementing the
8603 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
8604 and 0 when the comparison is false. Otherwise, the value indicates
8605 which bits of the result are guaranteed to be 1 when the comparison is
8606 true. This value is interpreted in the mode of the comparison
8607 operation, which is given by the mode of the first operand in the
8608 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
8609 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
8612 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
8613 generate code that depends only on the specified bits. It can also
8614 replace comparison operators with equivalent operations if they cause
8615 the required bits to be set, even if the remaining bits are undefined.
8616 For example, on a machine whose comparison operators return an
8617 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
8618 @samp{0x80000000}, saying that just the sign bit is relevant, the
8622 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
8629 (ashift:SI @var{x} (const_int @var{n}))
8633 where @var{n} is the appropriate shift count to move the bit being
8634 tested into the sign bit.
8636 There is no way to describe a machine that always sets the low-order bit
8637 for a true value, but does not guarantee the value of any other bits,
8638 but we do not know of any machine that has such an instruction. If you
8639 are trying to port GCC to such a machine, include an instruction to
8640 perform a logical-and of the result with 1 in the pattern for the
8641 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
8643 Often, a machine will have multiple instructions that obtain a value
8644 from a comparison (or the condition codes). Here are rules to guide the
8645 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
8650 Use the shortest sequence that yields a valid definition for
8651 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
8652 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
8653 comparison operators to do so because there may be opportunities to
8654 combine the normalization with other operations.
8657 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
8658 slightly preferred on machines with expensive jumps and 1 preferred on
8662 As a second choice, choose a value of @samp{0x80000001} if instructions
8663 exist that set both the sign and low-order bits but do not define the
8667 Otherwise, use a value of @samp{0x80000000}.
8670 Many machines can produce both the value chosen for
8671 @code{STORE_FLAG_VALUE} and its negation in the same number of
8672 instructions. On those machines, you should also define a pattern for
8673 those cases, e.g., one matching
8676 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
8679 Some machines can also perform @code{and} or @code{plus} operations on
8680 condition code values with less instructions than the corresponding
8681 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
8682 machines, define the appropriate patterns. Use the names @code{incscc}
8683 and @code{decscc}, respectively, for the patterns which perform
8684 @code{plus} or @code{minus} operations on condition code values. See
8685 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
8686 find such instruction sequences on other machines.
8688 You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
8691 @findex FLOAT_STORE_FLAG_VALUE
8692 @item FLOAT_STORE_FLAG_VALUE (@var{mode})
8693 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
8694 returned when comparison operators with floating-point results are true.
8695 Define this macro on machine that have comparison operations that return
8696 floating-point values. If there are no such operations, do not define
8701 An alias for the machine mode for pointers. On most machines, define
8702 this to be the integer mode corresponding to the width of a hardware
8703 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
8704 On some machines you must define this to be one of the partial integer
8705 modes, such as @code{PSImode}.
8707 The width of @code{Pmode} must be at least as large as the value of
8708 @code{POINTER_SIZE}. If it is not equal, you must define the macro
8709 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
8712 @findex FUNCTION_MODE
8714 An alias for the machine mode used for memory references to functions
8715 being called, in @code{call} RTL expressions. On most machines this
8716 should be @code{QImode}.
8718 @findex INTEGRATE_THRESHOLD
8719 @item INTEGRATE_THRESHOLD (@var{decl})
8720 A C expression for the maximum number of instructions above which the
8721 function @var{decl} should not be inlined. @var{decl} is a
8722 @code{FUNCTION_DECL} node.
8724 The default definition of this macro is 64 plus 8 times the number of
8725 arguments that the function accepts. Some people think a larger
8726 threshold should be used on RISC machines.
8728 @findex STDC_0_IN_SYSTEM_HEADERS
8729 @item STDC_0_IN_SYSTEM_HEADERS
8730 In normal operation, the preprocessor expands @code{__STDC__} to the
8731 constant 1, to signify that GCC conforms to ISO Standard C@. On some
8732 hosts, like Solaris, the system compiler uses a different convention,
8733 where @code{__STDC__} is normally 0, but is 1 if the user specifies
8734 strict conformance to the C Standard.
8736 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
8737 convention when processing system header files, but when processing user
8738 files @code{__STDC__} will always expand to 1.
8740 @findex NO_IMPLICIT_EXTERN_C
8741 @item NO_IMPLICIT_EXTERN_C
8742 Define this macro if the system header files support C++ as well as C@.
8743 This macro inhibits the usual method of using system header files in
8744 C++, which is to pretend that the file's contents are enclosed in
8745 @samp{extern "C" @{@dots{}@}}.
8747 @findex HANDLE_PRAGMA
8748 @item HANDLE_PRAGMA (@var{getc}, @var{ungetc}, @var{name})
8749 This macro is no longer supported. You must use
8750 @code{REGISTER_TARGET_PRAGMAS} instead.
8752 @findex REGISTER_TARGET_PRAGMAS
8755 @item REGISTER_TARGET_PRAGMAS (@var{pfile})
8756 Define this macro if you want to implement any target-specific pragmas.
8757 If defined, it is a C expression which makes a series of calls to
8758 @code{cpp_register_pragma} for each pragma, with @var{pfile} passed as
8759 the first argument to to these functions. The macro may also do any
8760 setup required for the pragmas.
8762 The primary reason to define this macro is to provide compatibility with
8763 other compilers for the same target. In general, we discourage
8764 definition of target-specific pragmas for GCC@.
8766 If the pragma can be implemented by attributes then you should consider
8767 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
8769 Preprocessor macros that appear on pragma lines are not expanded. All
8770 @samp{#pragma} directives that do not match any registered pragma are
8771 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
8773 @deftypefun void cpp_register_pragma (cpp_reader *@var{pfile}, const char *@var{space}, const char *@var{name}, void (*@var{callback}) (cpp_reader *))
8775 Each call to @code{cpp_register_pragma} establishes one pragma. The
8776 @var{callback} routine will be called when the preprocessor encounters a
8780 #pragma [@var{space}] @var{name} @dots{}
8783 @var{space} is the case-sensitive namespace of the pragma, or
8784 @code{NULL} to put the pragma in the global namespace. The callback
8785 routine receives @var{pfile} as its first argument, which can be passed
8786 on to cpplib's functions if necessary. You can lex tokens after the
8787 @var{name} by calling @code{c_lex}. Tokens that are not read by the
8788 callback will be silently ignored. The end of the line is indicated by
8789 a token of type @code{CPP_EOF}.
8791 For an example use of this routine, see @file{c4x.h} and the callback
8792 routines defined in @file{c4x-c.c}.
8794 Note that the use of @code{c_lex} is specific to the C and C++
8795 compilers. It will not work in the Java or Fortran compilers, or any
8796 other language compilers for that matter. Thus if @code{c_lex} is going
8797 to be called from target-specific code, it must only be done so when
8798 building the C and C++ compilers. This can be done by defining the
8799 variables @code{c_target_objs} and @code{cxx_target_objs} in the
8800 target entry in the @file{config.gcc} file. These variables should name
8801 the target-specific, language-specific object file which contains the
8802 code that uses @code{c_lex}. Note it will also be necessary to add a
8803 rule to the makefile fragment pointed to by @code{tmake_file} that shows
8804 how to build this object file.
8807 @findex HANDLE_SYSV_PRAGMA
8810 @item HANDLE_SYSV_PRAGMA
8811 Define this macro (to a value of 1) if you want the System V style
8812 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
8813 [=<value>]} to be supported by gcc.
8815 The pack pragma specifies the maximum alignment (in bytes) of fields
8816 within a structure, in much the same way as the @samp{__aligned__} and
8817 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
8818 the behavior to the default.
8820 A subtlety for Microsoft Visual C/C++ style bit-field packing
8821 (e.g. -mms-bitfields) for targets that support it:
8822 When a bit-field is inserted into a packed record, the whole size
8823 of the underlying type is used by one or more same-size adjacent
8824 bit-fields (that is, if its long:3, 32 bits is used in the record,
8825 and any additional adjacent long bit-fields are packed into the same
8826 chunk of 32 bits. However, if the size changes, a new field of that
8829 If both MS bit-fields and @samp{__attribute__((packed))} are used,
8830 the latter will take precedence. If @samp{__attribute__((packed))} is
8831 used on a single field when MS bit-fields are in use, it will take
8832 precedence for that field, but the alignment of the rest of the structure
8833 may affect its placement.
8835 The weak pragma only works if @code{SUPPORTS_WEAK} and
8836 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
8837 of specifically named weak labels, optionally with a value.
8839 @findex HANDLE_PRAGMA_PACK_PUSH_POP
8842 @item HANDLE_PRAGMA_PACK_PUSH_POP
8843 Define this macro (to a value of 1) if you want to support the Win32
8844 style pragmas @samp{#pragma pack(push,@var{n})} and @samp{#pragma
8845 pack(pop)}. The @samp{pack(push,@var{n})} pragma specifies the maximum alignment
8846 (in bytes) of fields within a structure, in much the same way as the
8847 @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
8848 pack value of zero resets the behavior to the default. Successive
8849 invocations of this pragma cause the previous values to be stacked, so
8850 that invocations of @samp{#pragma pack(pop)} will return to the previous
8853 @findex DOLLARS_IN_IDENTIFIERS
8854 @item DOLLARS_IN_IDENTIFIERS
8855 Define this macro to control use of the character @samp{$} in identifier
8856 names. 0 means @samp{$} is not allowed by default; 1 means it is allowed.
8857 1 is the default; there is no need to define this macro in that case.
8858 This macro controls the compiler proper; it does not affect the preprocessor.
8860 @findex NO_DOLLAR_IN_LABEL
8861 @item NO_DOLLAR_IN_LABEL
8862 Define this macro if the assembler does not accept the character
8863 @samp{$} in label names. By default constructors and destructors in
8864 G++ have @samp{$} in the identifiers. If this macro is defined,
8865 @samp{.} is used instead.
8867 @findex NO_DOT_IN_LABEL
8868 @item NO_DOT_IN_LABEL
8869 Define this macro if the assembler does not accept the character
8870 @samp{.} in label names. By default constructors and destructors in G++
8871 have names that use @samp{.}. If this macro is defined, these names
8872 are rewritten to avoid @samp{.}.
8874 @findex DEFAULT_MAIN_RETURN
8875 @item DEFAULT_MAIN_RETURN
8876 Define this macro if the target system expects every program's @code{main}
8877 function to return a standard ``success'' value by default (if no other
8878 value is explicitly returned).
8880 The definition should be a C statement (sans semicolon) to generate the
8881 appropriate rtl instructions. It is used only when compiling the end of
8886 Define this if the target system lacks the function @code{atexit}
8887 from the ISO C standard. If this macro is defined, a default definition
8888 will be provided to support C++. If @code{ON_EXIT} is not defined,
8889 a default @code{exit} function will also be provided.
8893 Define this macro if the target has another way to implement atexit
8894 functionality without replacing @code{exit}. For instance, SunOS 4 has
8895 a similar @code{on_exit} library function.
8897 The definition should be a functional macro which can be used just like
8898 the @code{atexit} function.
8902 Define this if your @code{exit} function needs to do something
8903 besides calling an external function @code{_cleanup} before
8904 terminating with @code{_exit}. The @code{EXIT_BODY} macro is
8905 only needed if @code{NEED_ATEXIT} is defined and @code{ON_EXIT} is not
8908 @findex INSN_SETS_ARE_DELAYED
8909 @item INSN_SETS_ARE_DELAYED (@var{insn})
8910 Define this macro as a C expression that is nonzero if it is safe for the
8911 delay slot scheduler to place instructions in the delay slot of @var{insn},
8912 even if they appear to use a resource set or clobbered in @var{insn}.
8913 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
8914 every @code{call_insn} has this behavior. On machines where some @code{insn}
8915 or @code{jump_insn} is really a function call and hence has this behavior,
8916 you should define this macro.
8918 You need not define this macro if it would always return zero.
8920 @findex INSN_REFERENCES_ARE_DELAYED
8921 @item INSN_REFERENCES_ARE_DELAYED (@var{insn})
8922 Define this macro as a C expression that is nonzero if it is safe for the
8923 delay slot scheduler to place instructions in the delay slot of @var{insn},
8924 even if they appear to set or clobber a resource referenced in @var{insn}.
8925 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
8926 some @code{insn} or @code{jump_insn} is really a function call and its operands
8927 are registers whose use is actually in the subroutine it calls, you should
8928 define this macro. Doing so allows the delay slot scheduler to move
8929 instructions which copy arguments into the argument registers into the delay
8932 You need not define this macro if it would always return zero.
8934 @findex MACHINE_DEPENDENT_REORG
8935 @item MACHINE_DEPENDENT_REORG (@var{insn})
8936 In rare cases, correct code generation requires extra machine
8937 dependent processing between the second jump optimization pass and
8938 delayed branch scheduling. On those machines, define this macro as a C
8939 statement to act on the code starting at @var{insn}.
8941 @findex MULTIPLE_SYMBOL_SPACES
8942 @item MULTIPLE_SYMBOL_SPACES
8943 Define this macro if in some cases global symbols from one translation
8944 unit may not be bound to undefined symbols in another translation unit
8945 without user intervention. For instance, under Microsoft Windows
8946 symbols must be explicitly imported from shared libraries (DLLs).
8948 @findex MD_ASM_CLOBBERS
8949 @item MD_ASM_CLOBBERS (@var{clobbers})
8950 A C statement that adds to @var{clobbers} @code{STRING_CST} trees for
8951 any hard regs the port wishes to automatically clobber for all asms.
8953 @findex MAX_INTEGER_COMPUTATION_MODE
8954 @item MAX_INTEGER_COMPUTATION_MODE
8955 Define this to the largest integer machine mode which can be used for
8956 operations other than load, store and copy operations.
8958 You need only define this macro if the target holds values larger than
8959 @code{word_mode} in general purpose registers. Most targets should not define
8962 @findex MATH_LIBRARY
8964 Define this macro as a C string constant for the linker argument to link
8965 in the system math library, or @samp{""} if the target does not have a
8966 separate math library.
8968 You need only define this macro if the default of @samp{"-lm"} is wrong.
8970 @findex LIBRARY_PATH_ENV
8971 @item LIBRARY_PATH_ENV
8972 Define this macro as a C string constant for the environment variable that
8973 specifies where the linker should look for libraries.
8975 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
8978 @findex TARGET_HAS_F_SETLKW
8979 @item TARGET_HAS_F_SETLKW
8980 Define this macro if the target supports file locking with fcntl / F_SETLKW@.
8981 Note that this functionality is part of POSIX@.
8982 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
8983 to use file locking when exiting a program, which avoids race conditions
8984 if the program has forked.
8986 @findex MAX_CONDITIONAL_EXECUTE
8987 @item MAX_CONDITIONAL_EXECUTE
8989 A C expression for the maximum number of instructions to execute via
8990 conditional execution instructions instead of a branch. A value of
8991 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
8992 1 if it does use cc0.
8994 @findex IFCVT_MODIFY_TESTS
8995 @item IFCVT_MODIFY_TESTS(@var{ce_info}, @var{true_expr}, @var{false_expr})
8996 Used if the target needs to perform machine-dependent modifications on the
8997 conditionals used for turning basic blocks into conditionally executed code.
8998 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
8999 contains information about the currently processed blocks. @var{true_expr}
9000 and @var{false_expr} are the tests that are used for converting the
9001 then-block and the else-block, respectively. Set either @var{true_expr} or
9002 @var{false_expr} to a null pointer if the tests cannot be converted.
9004 @findex IFCVT_MODIFY_MULTIPLE_TESTS
9005 @item IFCVT_MODIFY_MULTIPLE_TESTS(@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
9006 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
9007 if-statements into conditions combined by @code{and} and @code{or} operations.
9008 @var{bb} contains the basic block that contains the test that is currently
9009 being processed and about to be turned into a condition.
9011 @findex IFCVT_MODIFY_INSN
9012 @item IFCVT_MODIFY_INSN(@var{ce_info}, @var{pattern}, @var{insn})
9013 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
9014 be converted to conditional execution format. @var{ce_info} points to
9015 a data structure, @code{struct ce_if_block}, which contains information
9016 about the currently processed blocks.
9018 @findex IFCVT_MODIFY_FINAL
9019 @item IFCVT_MODIFY_FINAL(@var{ce_info})
9020 A C expression to perform any final machine dependent modifications in
9021 converting code to conditional execution. The involved basic blocks
9022 can be found in the @code{struct ce_if_block} structure that is pointed
9023 to by @var{ce_info}.
9025 @findex IFCVT_MODIFY_CANCEL
9026 @item IFCVT_MODIFY_CANCEL(@var{ce_info})
9027 A C expression to cancel any machine dependent modifications in
9028 converting code to conditional execution. The involved basic blocks
9029 can be found in the @code{struct ce_if_block} structure that is pointed
9030 to by @var{ce_info}.
9032 @findex IFCVT_INIT_EXTRA_FIELDS
9033 @item IFCVT_INIT_EXTRA_FIELDS(@var{ce_info})
9034 A C expression to initialize any extra fields in a @code{struct ce_if_block}
9035 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
9037 @findex IFCVT_EXTRA_FIELDS
9038 @item IFCVT_EXTRA_FIELDS
9039 If defined, it should expand to a set of field declarations that will be
9040 added to the @code{struct ce_if_block} structure. These should be intialized
9041 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
9045 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
9046 Define this hook if you have any machine-specific built-in functions
9047 that need to be defined. It should be a function that performs the
9050 Machine specific built-in functions can be useful to expand special machine
9051 instructions that would otherwise not normally be generated because
9052 they have no equivalent in the source language (for example, SIMD vector
9053 instructions or prefetch instructions).
9055 To create a built-in function, call the function @code{builtin_function}
9056 which is defined by the language front end. You can use any type nodes set
9057 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
9058 only language front ends that use those two functions will call
9059 @samp{TARGET_INIT_BUILTINS}.
9062 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
9064 Expand a call to a machine specific built-in function that was set up by
9065 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
9066 function call; the result should go to @var{target} if that is
9067 convenient, and have mode @var{mode} if that is convenient.
9068 @var{subtarget} may be used as the target for computing one of
9069 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
9070 ignored. This function should return the result of the call to the
9075 @findex MD_CAN_REDIRECT_BRANCH
9076 @item MD_CAN_REDIRECT_BRANCH(@var{branch1}, @var{branch2})
9078 Take a branch insn in @var{branch1} and another in @var{branch2}.
9079 Return true if redirecting @var{branch1} to the destination of
9080 @var{branch2} is possible.
9082 On some targets, branches may have a limited range. Optimizing the
9083 filling of delay slots can result in branches being redirected, and this
9084 may in turn cause a branch offset to overflow.
9086 @findex ALLOCATE_INITIAL_VALUE
9087 @item ALLOCATE_INITIAL_VALUE(@var{hard_reg})
9089 When the initial value of a hard register has been copied in a pseudo
9090 register, it is often not necessary to actually allocate another register
9091 to this pseudo register, because the original hard register or a stack slot
9092 it has been saved into can be used. @code{ALLOCATE_INITIAL_VALUE}, if
9093 defined, is called at the start of register allocation once for each
9094 hard register that had its initial value copied by using
9095 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
9096 Possible values are @code{NULL_RTX}, if you don't want
9097 to do any special allocation, a @code{REG} rtx---that would typically be
9098 the hard register itself, if it is known not to be clobbered---or a
9100 If you are returning a @code{MEM}, this is only a hint for the allocator;
9101 it might decide to use another register anyways.
9102 You may use @code{current_function_leaf_function} in the definition of the
9103 macro, functions that use @code{REG_N_SETS}, to determine if the hard
9104 register in question will not be clobbered.
9106 @findex TARGET_OBJECT_SUFFIX
9107 @item TARGET_OBJECT_SUFFIX
9108 Define this macro to be a C string representing the suffix for object
9109 files on your target machine. If you do not define this macro, GCC will
9110 use @samp{.o} as the suffix for object files.
9112 @findex TARGET_EXECUTABLE_SUFFIX
9113 @item TARGET_EXECUTABLE_SUFFIX
9114 Define this macro to be a C string representing the suffix to be
9115 automatically added to executable files on your target machine. If you
9116 do not define this macro, GCC will use the null string as the suffix for
9119 @findex COLLECT_EXPORT_LIST
9120 @item COLLECT_EXPORT_LIST
9121 If defined, @code{collect2} will scan the individual object files
9122 specified on its command line and create an export list for the linker.
9123 Define this macro for systems like AIX, where the linker discards
9124 object files that are not referenced from @code{main} and uses export
9129 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
9130 This target hook returns @code{true} past the point in which new jump
9131 instructions could be created. On machines that require a register for
9132 every jump such as the SHmedia ISA of SH5, this point would typically be
9133 reload, so this target hook should be defined to a function such as:
9137 cannot_modify_jumps_past_reload_p ()
9139 return (reload_completed || reload_in_progress);