1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,
2 @c 2002, 2003 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
7 @chapter Target Description Macros and Functions
8 @cindex machine description macros
9 @cindex target description macros
10 @cindex macros, target description
11 @cindex @file{tm.h} macros
13 In addition to the file @file{@var{machine}.md}, a machine description
14 includes a C header file conventionally given the name
15 @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}.
16 The header file defines numerous macros that convey the information
17 about the target machine that does not fit into the scheme of the
18 @file{.md} file. The file @file{tm.h} should be a link to
19 @file{@var{machine}.h}. The header file @file{config.h} includes
20 @file{tm.h} and most compiler source files include @file{config.h}. The
21 source file defines a variable @code{targetm}, which is a structure
22 containing pointers to functions and data relating to the target
23 machine. @file{@var{machine}.c} should also contain their definitions,
24 if they are not defined elsewhere in GCC, and other functions called
25 through the macros defined in the @file{.h} file.
28 * Target Structure:: The @code{targetm} variable.
29 * Driver:: Controlling how the driver runs the compilation passes.
30 * Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}.
31 * Per-Function Data:: Defining data structures for per-function information.
32 * Storage Layout:: Defining sizes and alignments of data.
33 * Type Layout:: Defining sizes and properties of basic user data types.
34 * Escape Sequences:: Defining the value of target character escape sequences
35 * Registers:: Naming and describing the hardware registers.
36 * Register Classes:: Defining the classes of hardware registers.
37 * Stack and Calling:: Defining which way the stack grows and by how much.
38 * Varargs:: Defining the varargs macros.
39 * Trampolines:: Code set up at run time to enter a nested function.
40 * Library Calls:: Controlling how library routines are implicitly called.
41 * Addressing Modes:: Defining addressing modes valid for memory operands.
42 * Condition Code:: Defining how insns update the condition code.
43 * Costs:: Defining relative costs of different operations.
44 * Scheduling:: Adjusting the behavior of the instruction scheduler.
45 * Sections:: Dividing storage into text, data, and other sections.
46 * PIC:: Macros for position independent code.
47 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
48 * Debugging Info:: Defining the format of debugging output.
49 * Floating Point:: Handling floating point for cross-compilers.
50 * Mode Switching:: Insertion of mode-switching instructions.
51 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
52 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
53 * Misc:: Everything else.
56 @node Target Structure
57 @section The Global @code{targetm} Variable
59 @cindex target functions
61 @deftypevar {struct gcc_target} targetm
62 The target @file{.c} file must define the global @code{targetm} variable
63 which contains pointers to functions and data relating to the target
64 machine. The variable is declared in @file{target.h};
65 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
66 used to initialize the variable, and macros for the default initializers
67 for elements of the structure. The @file{.c} file should override those
68 macros for which the default definition is inappropriate. For example:
71 #include "target-def.h"
73 /* @r{Initialize the GCC target structure.} */
75 #undef TARGET_COMP_TYPE_ATTRIBUTES
76 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
78 struct gcc_target targetm = TARGET_INITIALIZER;
82 Where a macro should be defined in the @file{.c} file in this manner to
83 form part of the @code{targetm} structure, it is documented below as a
84 ``Target Hook'' with a prototype. Many macros will change in future
85 from being defined in the @file{.h} file to being part of the
86 @code{targetm} structure.
89 @section Controlling the Compilation Driver, @file{gcc}
91 @cindex controlling the compilation driver
93 @c prevent bad page break with this line
94 You can control the compilation driver.
97 @findex SWITCH_TAKES_ARG
98 @item SWITCH_TAKES_ARG (@var{char})
99 A C expression which determines whether the option @option{-@var{char}}
100 takes arguments. The value should be the number of arguments that
101 option takes--zero, for many options.
103 By default, this macro is defined as
104 @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options
105 properly. You need not define @code{SWITCH_TAKES_ARG} unless you
106 wish to add additional options which take arguments. Any redefinition
107 should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for
110 @findex WORD_SWITCH_TAKES_ARG
111 @item WORD_SWITCH_TAKES_ARG (@var{name})
112 A C expression which determines whether the option @option{-@var{name}}
113 takes arguments. The value should be the number of arguments that
114 option takes--zero, for many options. This macro rather than
115 @code{SWITCH_TAKES_ARG} is used for multi-character option names.
117 By default, this macro is defined as
118 @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
119 properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
120 wish to add additional options which take arguments. Any redefinition
121 should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
124 @findex SWITCH_CURTAILS_COMPILATION
125 @item SWITCH_CURTAILS_COMPILATION (@var{char})
126 A C expression which determines whether the option @option{-@var{char}}
127 stops compilation before the generation of an executable. The value is
128 boolean, nonzero if the option does stop an executable from being
129 generated, zero otherwise.
131 By default, this macro is defined as
132 @code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard
133 options properly. You need not define
134 @code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional
135 options which affect the generation of an executable. Any redefinition
136 should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check
137 for additional options.
139 @findex SWITCHES_NEED_SPACES
140 @item SWITCHES_NEED_SPACES
141 A string-valued C expression which enumerates the options for which
142 the linker needs a space between the option and its argument.
144 If this macro is not defined, the default value is @code{""}.
146 @findex TARGET_OPTION_TRANSLATE_TABLE
147 @item TARGET_OPTION_TRANSLATE_TABLE
148 If defined, a list of pairs of strings, the first of which is a
149 potential command line target to the @file{gcc} driver program, and the
150 second of which is a space-separated (tabs and other whitespace are not
151 supported) list of options with which to replace the first option. The
152 target defining this list is responsible for assuring that the results
153 are valid. Replacement options may not be the @code{--opt} style, they
154 must be the @code{-opt} style. It is the intention of this macro to
155 provide a mechanism for substitution that affects the multilibs chosen,
156 such as one option that enables many options, some of which select
157 multilibs. Example nonsensical definition, where @code{-malt-abi},
158 @code{-EB}, and @code{-mspoo} cause different multilibs to be chosen:
161 #define TARGET_OPTION_TRANSLATE_TABLE \
162 @{ "-fast", "-march=fast-foo -malt-abi -I/usr/fast-foo" @}, \
163 @{ "-compat", "-EB -malign=4 -mspoo" @}
166 @findex DRIVER_SELF_SPECS
167 @item DRIVER_SELF_SPECS
168 A list of specs for the driver itself. It should be a suitable
169 initializer for an array of strings, with no surrounding braces.
171 The driver applies these specs to its own command line between loading
172 default @file{specs} files (but not command-line specified ones) and
173 choosing the multilib directory or running any subcommands. It
174 applies them in the order given, so each spec can depend on the
175 options added by earlier ones. It is also possible to remove options
176 using @samp{%<@var{option}} in the usual way.
178 This macro can be useful when a port has several interdependent target
179 options. It provides a way of standardizing the command line so
180 that the other specs are easier to write.
182 Do not define this macro if it does not need to do anything.
186 A C string constant that tells the GCC driver program options to
187 pass to CPP@. It can also specify how to translate options you
188 give to GCC into options for GCC to pass to the CPP@.
190 Do not define this macro if it does not need to do anything.
192 @findex CPLUSPLUS_CPP_SPEC
193 @item CPLUSPLUS_CPP_SPEC
194 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
195 than C@. If you do not define this macro, then the value of
196 @code{CPP_SPEC} (if any) will be used instead.
200 A C string constant that tells the GCC driver program options to
201 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
203 It can also specify how to translate options you give to GCC into options
204 for GCC to pass to front ends.
206 Do not define this macro if it does not need to do anything.
210 A C string constant that tells the GCC driver program options to
211 pass to @code{cc1plus}. It can also specify how to translate options you
212 give to GCC into options for GCC to pass to the @code{cc1plus}.
214 Do not define this macro if it does not need to do anything.
215 Note that everything defined in CC1_SPEC is already passed to
216 @code{cc1plus} so there is no need to duplicate the contents of
217 CC1_SPEC in CC1PLUS_SPEC@.
221 A C string constant that tells the GCC driver program options to
222 pass to the assembler. It can also specify how to translate options
223 you give to GCC into options for GCC to pass to the assembler.
224 See the file @file{sun3.h} for an example of this.
226 Do not define this macro if it does not need to do anything.
228 @findex ASM_FINAL_SPEC
230 A C string constant that tells the GCC driver program how to
231 run any programs which cleanup after the normal assembler.
232 Normally, this is not needed. See the file @file{mips.h} for
235 Do not define this macro if it does not need to do anything.
237 @findex AS_NEEDS_DASH_FOR_PIPED_INPUT
238 @item AS_NEEDS_DASH_FOR_PIPED_INPUT
239 Define this macro, with no value, if the driver should give the assembler
240 an argument consisting of a single dash, @option{-}, to instruct it to
241 read from its standard input (which will be a pipe connected to the
242 output of the compiler proper). This argument is given after any
243 @option{-o} option specifying the name of the output file.
245 If you do not define this macro, the assembler is assumed to read its
246 standard input if given no non-option arguments. If your assembler
247 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
248 see @file{mips.h} for instance.
252 A C string constant that tells the GCC driver program options to
253 pass to the linker. It can also specify how to translate options you
254 give to GCC into options for GCC to pass to the linker.
256 Do not define this macro if it does not need to do anything.
260 Another C string constant used much like @code{LINK_SPEC}. The difference
261 between the two is that @code{LIB_SPEC} is used at the end of the
262 command given to the linker.
264 If this macro is not defined, a default is provided that
265 loads the standard C library from the usual place. See @file{gcc.c}.
269 Another C string constant that tells the GCC driver program
270 how and when to place a reference to @file{libgcc.a} into the
271 linker command line. This constant is placed both before and after
272 the value of @code{LIB_SPEC}.
274 If this macro is not defined, the GCC driver provides a default that
275 passes the string @option{-lgcc} to the linker.
277 @findex STARTFILE_SPEC
279 Another C string constant used much like @code{LINK_SPEC}. The
280 difference between the two is that @code{STARTFILE_SPEC} is used at
281 the very beginning of the command given to the linker.
283 If this macro is not defined, a default is provided that loads the
284 standard C startup file from the usual place. See @file{gcc.c}.
288 Another C string constant used much like @code{LINK_SPEC}. The
289 difference between the two is that @code{ENDFILE_SPEC} is used at
290 the very end of the command given to the linker.
292 Do not define this macro if it does not need to do anything.
294 @findex THREAD_MODEL_SPEC
295 @item THREAD_MODEL_SPEC
296 GCC @code{-v} will print the thread model GCC was configured to use.
297 However, this doesn't work on platforms that are multilibbed on thread
298 models, such as AIX 4.3. On such platforms, define
299 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
300 blanks that names one of the recognized thread models. @code{%*}, the
301 default value of this macro, will expand to the value of
302 @code{thread_file} set in @file{config.gcc}.
306 Define this macro to provide additional specifications to put in the
307 @file{specs} file that can be used in various specifications like
310 The definition should be an initializer for an array of structures,
311 containing a string constant, that defines the specification name, and a
312 string constant that provides the specification.
314 Do not define this macro if it does not need to do anything.
316 @code{EXTRA_SPECS} is useful when an architecture contains several
317 related targets, which have various @code{@dots{}_SPECS} which are similar
318 to each other, and the maintainer would like one central place to keep
321 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
322 define either @code{_CALL_SYSV} when the System V calling sequence is
323 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
326 The @file{config/rs6000/rs6000.h} target file defines:
329 #define EXTRA_SPECS \
330 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
332 #define CPP_SYS_DEFAULT ""
335 The @file{config/rs6000/sysv.h} target file defines:
339 "%@{posix: -D_POSIX_SOURCE @} \
340 %@{mcall-sysv: -D_CALL_SYSV @} \
341 %@{!mcall-sysv: %(cpp_sysv_default) @} \
342 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
344 #undef CPP_SYSV_DEFAULT
345 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
348 while the @file{config/rs6000/eabiaix.h} target file defines
349 @code{CPP_SYSV_DEFAULT} as:
352 #undef CPP_SYSV_DEFAULT
353 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
356 @findex LINK_LIBGCC_SPECIAL
357 @item LINK_LIBGCC_SPECIAL
358 Define this macro if the driver program should find the library
359 @file{libgcc.a} itself and should not pass @option{-L} options to the
360 linker. If you do not define this macro, the driver program will pass
361 the argument @option{-lgcc} to tell the linker to do the search and will
362 pass @option{-L} options to it.
364 @findex LINK_LIBGCC_SPECIAL_1
365 @item LINK_LIBGCC_SPECIAL_1
366 Define this macro if the driver program should find the library
367 @file{libgcc.a}. If you do not define this macro, the driver program will pass
368 the argument @option{-lgcc} to tell the linker to do the search.
369 This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does
370 not affect @option{-L} options.
372 @findex LINK_GCC_C_SEQUENCE_SPEC
373 @item LINK_GCC_C_SEQUENCE_SPEC
374 The sequence in which libgcc and libc are specified to the linker.
375 By default this is @code{%G %L %G}.
377 @findex LINK_COMMAND_SPEC
378 @item LINK_COMMAND_SPEC
379 A C string constant giving the complete command line need to execute the
380 linker. When you do this, you will need to update your port each time a
381 change is made to the link command line within @file{gcc.c}. Therefore,
382 define this macro only if you need to completely redefine the command
383 line for invoking the linker and there is no other way to accomplish
384 the effect you need. Overriding this macro may be avoidable by overriding
385 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
387 @findex LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
388 @item LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
389 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
390 directories from linking commands. Do not give it a nonzero value if
391 removing duplicate search directories changes the linker's semantics.
393 @findex MULTILIB_DEFAULTS
394 @item MULTILIB_DEFAULTS
395 Define this macro as a C expression for the initializer of an array of
396 string to tell the driver program which options are defaults for this
397 target and thus do not need to be handled specially when using
398 @code{MULTILIB_OPTIONS}.
400 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
401 the target makefile fragment or if none of the options listed in
402 @code{MULTILIB_OPTIONS} are set by default.
403 @xref{Target Fragment}.
405 @findex RELATIVE_PREFIX_NOT_LINKDIR
406 @item RELATIVE_PREFIX_NOT_LINKDIR
407 Define this macro to tell @command{gcc} that it should only translate
408 a @option{-B} prefix into a @option{-L} linker option if the prefix
409 indicates an absolute file name.
411 @findex STANDARD_EXEC_PREFIX
412 @item STANDARD_EXEC_PREFIX
413 Define this macro as a C string constant if you wish to override the
414 standard choice of @file{/usr/local/lib/gcc-lib/} as the default prefix to
415 try when searching for the executable files of the compiler.
417 @findex MD_EXEC_PREFIX
419 If defined, this macro is an additional prefix to try after
420 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
421 when the @option{-b} option is used, or the compiler is built as a cross
422 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
423 to the list of directories used to find the assembler in @file{configure.in}.
425 @findex STANDARD_STARTFILE_PREFIX
426 @item STANDARD_STARTFILE_PREFIX
427 Define this macro as a C string constant if you wish to override the
428 standard choice of @file{/usr/local/lib/} as the default prefix to
429 try when searching for startup files such as @file{crt0.o}.
431 @findex MD_STARTFILE_PREFIX
432 @item MD_STARTFILE_PREFIX
433 If defined, this macro supplies an additional prefix to try after the
434 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
435 @option{-b} option is used, or when the compiler is built as a cross
438 @findex MD_STARTFILE_PREFIX_1
439 @item MD_STARTFILE_PREFIX_1
440 If defined, this macro supplies yet another prefix to try after the
441 standard prefixes. It is not searched when the @option{-b} option is
442 used, or when the compiler is built as a cross compiler.
444 @findex INIT_ENVIRONMENT
445 @item INIT_ENVIRONMENT
446 Define this macro as a C string constant if you wish to set environment
447 variables for programs called by the driver, such as the assembler and
448 loader. The driver passes the value of this macro to @code{putenv} to
449 initialize the necessary environment variables.
451 @findex LOCAL_INCLUDE_DIR
452 @item LOCAL_INCLUDE_DIR
453 Define this macro as a C string constant if you wish to override the
454 standard choice of @file{/usr/local/include} as the default prefix to
455 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
456 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
458 Cross compilers do not search either @file{/usr/local/include} or its
461 @findex MODIFY_TARGET_NAME
462 @item MODIFY_TARGET_NAME
463 Define this macro if you with to define command-line switches that modify the
466 For each switch, you can include a string to be appended to the first
467 part of the configuration name or a string to be deleted from the
468 configuration name, if present. The definition should be an initializer
469 for an array of structures. Each array element should have three
470 elements: the switch name (a string constant, including the initial
471 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
472 indicate whether the string should be inserted or deleted, and the string
473 to be inserted or deleted (a string constant).
475 For example, on a machine where @samp{64} at the end of the
476 configuration name denotes a 64-bit target and you want the @option{-32}
477 and @option{-64} switches to select between 32- and 64-bit targets, you would
481 #define MODIFY_TARGET_NAME \
482 @{ @{ "-32", DELETE, "64"@}, \
483 @{"-64", ADD, "64"@}@}
487 @findex SYSTEM_INCLUDE_DIR
488 @item SYSTEM_INCLUDE_DIR
489 Define this macro as a C string constant if you wish to specify a
490 system-specific directory to search for header files before the standard
491 directory. @code{SYSTEM_INCLUDE_DIR} comes before
492 @code{STANDARD_INCLUDE_DIR} in the search order.
494 Cross compilers do not use this macro and do not search the directory
497 @findex STANDARD_INCLUDE_DIR
498 @item STANDARD_INCLUDE_DIR
499 Define this macro as a C string constant if you wish to override the
500 standard choice of @file{/usr/include} as the default prefix to
501 try when searching for header files.
503 Cross compilers do not use this macro and do not search either
504 @file{/usr/include} or its replacement.
506 @findex STANDARD_INCLUDE_COMPONENT
507 @item STANDARD_INCLUDE_COMPONENT
508 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
509 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
510 If you do not define this macro, no component is used.
512 @findex INCLUDE_DEFAULTS
513 @item INCLUDE_DEFAULTS
514 Define this macro if you wish to override the entire default search path
515 for include files. For a native compiler, the default search path
516 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
517 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
518 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
519 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
520 and specify private search areas for GCC@. The directory
521 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
523 The definition should be an initializer for an array of structures.
524 Each array element should have four elements: the directory name (a
525 string constant), the component name (also a string constant), a flag
526 for C++-only directories,
527 and a flag showing that the includes in the directory don't need to be
528 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
529 the array with a null element.
531 The component name denotes what GNU package the include file is part of,
532 if any, in all upper-case letters. For example, it might be @samp{GCC}
533 or @samp{BINUTILS}. If the package is part of a vendor-supplied
534 operating system, code the component name as @samp{0}.
536 For example, here is the definition used for VAX/VMS:
539 #define INCLUDE_DEFAULTS \
541 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
542 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
543 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
550 Here is the order of prefixes tried for exec files:
554 Any prefixes specified by the user with @option{-B}.
557 The environment variable @code{GCC_EXEC_PREFIX}, if any.
560 The directories specified by the environment variable @code{COMPILER_PATH}.
563 The macro @code{STANDARD_EXEC_PREFIX}.
566 @file{/usr/lib/gcc/}.
569 The macro @code{MD_EXEC_PREFIX}, if any.
572 Here is the order of prefixes tried for startfiles:
576 Any prefixes specified by the user with @option{-B}.
579 The environment variable @code{GCC_EXEC_PREFIX}, if any.
582 The directories specified by the environment variable @code{LIBRARY_PATH}
583 (or port-specific name; native only, cross compilers do not use this).
586 The macro @code{STANDARD_EXEC_PREFIX}.
589 @file{/usr/lib/gcc/}.
592 The macro @code{MD_EXEC_PREFIX}, if any.
595 The macro @code{MD_STARTFILE_PREFIX}, if any.
598 The macro @code{STANDARD_STARTFILE_PREFIX}.
607 @node Run-time Target
608 @section Run-time Target Specification
609 @cindex run-time target specification
610 @cindex predefined macros
611 @cindex target specifications
613 @c prevent bad page break with this line
614 Here are run-time target specifications.
617 @findex TARGET_CPU_CPP_BUILTINS
618 @item TARGET_CPU_CPP_BUILTINS()
619 This function-like macro expands to a block of code that defines
620 built-in preprocessor macros and assertions for the target cpu, using
621 the functions @code{builtin_define}, @code{builtin_define_std} and
622 @code{builtin_assert}. When the front end
623 calls this macro it provides a trailing semicolon, and since it has
624 finished command line option processing your code can use those
627 @code{builtin_assert} takes a string in the form you pass to the
628 command-line option @option{-A}, such as @code{cpu=mips}, and creates
629 the assertion. @code{builtin_define} takes a string in the form
630 accepted by option @option{-D} and unconditionally defines the macro.
632 @code{builtin_define_std} takes a string representing the name of an
633 object-like macro. If it doesn't lie in the user's namespace,
634 @code{builtin_define_std} defines it unconditionally. Otherwise, it
635 defines a version with two leading underscores, and another version
636 with two leading and trailing underscores, and defines the original
637 only if an ISO standard was not requested on the command line. For
638 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
639 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
640 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
641 defines only @code{_ABI64}.
643 You can also test for the C dialect being compiled. The variable
644 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
645 or @code{clk_objective_c}. Note that if we are preprocessing
646 assembler, this variable will be @code{clk_c} but the function-like
647 macro @code{preprocessing_asm_p()} will return true, so you might want
648 to check for that first. If you need to check for strict ANSI, the
649 variable @code{flag_iso} can be used. The function-like macro
650 @code{preprocessing_trad_p()} can be used to check for traditional
653 With @code{TARGET_OS_CPP_BUILTINS} this macro obsoletes the
654 @code{CPP_PREDEFINES} target macro.
656 @findex TARGET_OS_CPP_BUILTINS
657 @item TARGET_OS_CPP_BUILTINS()
658 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
659 and is used for the target operating system instead.
661 With @code{TARGET_CPU_CPP_BUILTINS} this macro obsoletes the
662 @code{CPP_PREDEFINES} target macro.
664 @findex CPP_PREDEFINES
666 Define this to be a string constant containing @option{-D} options to
667 define the predefined macros that identify this machine and system.
668 These macros will be predefined unless the @option{-ansi} option (or a
669 @option{-std} option for strict ISO C conformance) is specified.
671 In addition, a parallel set of macros are predefined, whose names are
672 made by appending @samp{__} at the beginning and at the end. These
673 @samp{__} macros are permitted by the ISO standard, so they are
674 predefined regardless of whether @option{-ansi} or a @option{-std} option
677 For example, on the Sun, one can use the following value:
680 "-Dmc68000 -Dsun -Dunix"
683 The result is to define the macros @code{__mc68000__}, @code{__sun__}
684 and @code{__unix__} unconditionally, and the macros @code{mc68000},
685 @code{sun} and @code{unix} provided @option{-ansi} is not specified.
687 @findex extern int target_flags
688 @item extern int target_flags;
689 This declaration should be present.
691 @cindex optional hardware or system features
692 @cindex features, optional, in system conventions
694 This series of macros is to allow compiler command arguments to
695 enable or disable the use of optional features of the target machine.
696 For example, one machine description serves both the 68000 and
697 the 68020; a command argument tells the compiler whether it should
698 use 68020-only instructions or not. This command argument works
699 by means of a macro @code{TARGET_68020} that tests a bit in
702 Define a macro @code{TARGET_@var{featurename}} for each such option.
703 Its definition should test a bit in @code{target_flags}. It is
704 recommended that a helper macro @code{TARGET_MASK_@var{featurename}}
705 is defined for each bit-value to test, and used in
706 @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For
710 #define TARGET_MASK_68020 1
711 #define TARGET_68020 (target_flags & TARGET_MASK_68020)
714 One place where these macros are used is in the condition-expressions
715 of instruction patterns. Note how @code{TARGET_68020} appears
716 frequently in the 68000 machine description file, @file{m68k.md}.
717 Another place they are used is in the definitions of the other
718 macros in the @file{@var{machine}.h} file.
720 @findex TARGET_SWITCHES
721 @item TARGET_SWITCHES
722 This macro defines names of command options to set and clear
723 bits in @code{target_flags}. Its definition is an initializer
724 with a subgrouping for each command option.
726 Each subgrouping contains a string constant, that defines the option
727 name, a number, which contains the bits to set in
728 @code{target_flags}, and a second string which is the description
729 displayed by @option{--help}. If the number is negative then the bits specified
730 by the number are cleared instead of being set. If the description
731 string is present but empty, then no help information will be displayed
732 for that option, but it will not count as an undocumented option. The
733 actual option name is made by appending @samp{-m} to the specified name.
734 Non-empty description strings should be marked with @code{N_(@dots{})} for
735 @command{xgettext}. Please do not mark empty strings because the empty
736 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
737 of the message catalog with meta information, not the empty string.
739 In addition to the description for @option{--help},
740 more detailed documentation for each option should be added to
743 One of the subgroupings should have a null string. The number in
744 this grouping is the default value for @code{target_flags}. Any
745 target options act starting with that value.
747 Here is an example which defines @option{-m68000} and @option{-m68020}
748 with opposite meanings, and picks the latter as the default:
751 #define TARGET_SWITCHES \
752 @{ @{ "68020", TARGET_MASK_68020, "" @}, \
753 @{ "68000", -TARGET_MASK_68020, \
754 N_("Compile for the 68000") @}, \
755 @{ "", TARGET_MASK_68020, "" @}@}
758 @findex TARGET_OPTIONS
760 This macro is similar to @code{TARGET_SWITCHES} but defines names of command
761 options that have values. Its definition is an initializer with a
762 subgrouping for each command option.
764 Each subgrouping contains a string constant, that defines the option
765 name, the address of a variable, a description string, and a value.
766 Non-empty description strings should be marked with @code{N_(@dots{})}
767 for @command{xgettext}. Please do not mark empty strings because the
768 empty string is reserved by GNU gettext. @code{gettext("")} returns the
769 header entry of the message catalog with meta information, not the empty
772 If the value listed in the table is @code{NULL}, then the variable, type
773 @code{char *}, is set to the variable part of the given option if the
774 fixed part matches. In other words, if the first part of the option
775 matches what's in the table, the variable will be set to point to the
776 rest of the option. This allows the user to specify a value for that
777 option. The actual option name is made by appending @samp{-m} to the
778 specified name. Again, each option should also be documented in
781 If the value listed in the table is non-@code{NULL}, then the option
782 must match the option in the table exactly (with @samp{-m}), and the
783 variable is set to point to the value listed in the table.
785 Here is an example which defines @option{-mshort-data-@var{number}}. If the
786 given option is @option{-mshort-data-512}, the variable @code{m88k_short_data}
787 will be set to the string @code{"512"}.
790 extern char *m88k_short_data;
791 #define TARGET_OPTIONS \
792 @{ @{ "short-data-", &m88k_short_data, \
793 N_("Specify the size of the short data section"), 0 @} @}
796 Here is an variant of the above that allows the user to also specify
797 just @option{-mshort-data} where a default of @code{"64"} is used.
800 extern char *m88k_short_data;
801 #define TARGET_OPTIONS \
802 @{ @{ "short-data-", &m88k_short_data, \
803 N_("Specify the size of the short data section"), 0 @} \
804 @{ "short-data", &m88k_short_data, "", "64" @},
808 Here is an example which defines @option{-mno-alu}, @option{-malu1}, and
809 @option{-malu2} as a three-state switch, along with suitable macros for
810 checking the state of the option (documentation is elided for brevity).
814 char *chip_alu = ""; /* Specify default here. */
817 extern char *chip_alu;
818 #define TARGET_OPTIONS \
819 @{ @{ "no-alu", &chip_alu, "", "" @}, \
820 @{ "alu1", &chip_alu, "", "1" @}, \
821 @{ "alu2", &chip_alu, "", "2" @}, @}
822 #define TARGET_ALU (chip_alu[0] != '\0')
823 #define TARGET_ALU1 (chip_alu[0] == '1')
824 #define TARGET_ALU2 (chip_alu[0] == '2')
827 @findex TARGET_VERSION
829 This macro is a C statement to print on @code{stderr} a string
830 describing the particular machine description choice. Every machine
831 description should define @code{TARGET_VERSION}. For example:
835 #define TARGET_VERSION \
836 fprintf (stderr, " (68k, Motorola syntax)");
838 #define TARGET_VERSION \
839 fprintf (stderr, " (68k, MIT syntax)");
843 @findex OVERRIDE_OPTIONS
844 @item OVERRIDE_OPTIONS
845 Sometimes certain combinations of command options do not make sense on
846 a particular target machine. You can define a macro
847 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
848 defined, is executed once just after all the command options have been
851 Don't use this macro to turn on various extra optimizations for
852 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
854 @findex OPTIMIZATION_OPTIONS
855 @item OPTIMIZATION_OPTIONS (@var{level}, @var{size})
856 Some machines may desire to change what optimizations are performed for
857 various optimization levels. This macro, if defined, is executed once
858 just after the optimization level is determined and before the remainder
859 of the command options have been parsed. Values set in this macro are
860 used as the default values for the other command line options.
862 @var{level} is the optimization level specified; 2 if @option{-O2} is
863 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
865 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
867 You should not use this macro to change options that are not
868 machine-specific. These should uniformly selected by the same
869 optimization level on all supported machines. Use this macro to enable
870 machine-specific optimizations.
872 @strong{Do not examine @code{write_symbols} in
873 this macro!} The debugging options are not supposed to alter the
876 @findex CAN_DEBUG_WITHOUT_FP
877 @item CAN_DEBUG_WITHOUT_FP
878 Define this macro if debugging can be performed even without a frame
879 pointer. If this macro is defined, GCC will turn on the
880 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
883 @node Per-Function Data
884 @section Defining data structures for per-function information.
885 @cindex per-function data
886 @cindex data structures
888 If the target needs to store information on a per-function basis, GCC
889 provides a macro and a couple of variables to allow this. Note, just
890 using statics to store the information is a bad idea, since GCC supports
891 nested functions, so you can be halfway through encoding one function
892 when another one comes along.
894 GCC defines a data structure called @code{struct function} which
895 contains all of the data specific to an individual function. This
896 structure contains a field called @code{machine} whose type is
897 @code{struct machine_function *}, which can be used by targets to point
898 to their own specific data.
900 If a target needs per-function specific data it should define the type
901 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
902 This macro should be used to initialize the function pointer
903 @code{init_machine_status}. This pointer is explained below.
905 One typical use of per-function, target specific data is to create an
906 RTX to hold the register containing the function's return address. This
907 RTX can then be used to implement the @code{__builtin_return_address}
908 function, for level 0.
910 Note---earlier implementations of GCC used a single data area to hold
911 all of the per-function information. Thus when processing of a nested
912 function began the old per-function data had to be pushed onto a
913 stack, and when the processing was finished, it had to be popped off the
914 stack. GCC used to provide function pointers called
915 @code{save_machine_status} and @code{restore_machine_status} to handle
916 the saving and restoring of the target specific information. Since the
917 single data area approach is no longer used, these pointers are no
920 The macro and function pointers are described below.
923 @findex INIT_EXPANDERS
925 Macro called to initialize any target specific information. This macro
926 is called once per function, before generation of any RTL has begun.
927 The intention of this macro is to allow the initialization of the
928 function pointers below.
930 @findex init_machine_status
931 @item init_machine_status
932 This is a @code{void (*)(struct function *)} function pointer. If this
933 pointer is non-@code{NULL} it will be called once per function, before function
934 compilation starts, in order to allow the target to perform any target
935 specific initialization of the @code{struct function} structure. It is
936 intended that this would be used to initialize the @code{machine} of
939 @code{struct machine_function} structures are expected to be freed by GC.
940 Generally, any memory that they reference must be allocated by using
941 @code{ggc_alloc}, including the structure itself.
946 @section Storage Layout
947 @cindex storage layout
949 Note that the definitions of the macros in this table which are sizes or
950 alignments measured in bits do not need to be constant. They can be C
951 expressions that refer to static variables, such as the @code{target_flags}.
952 @xref{Run-time Target}.
955 @findex BITS_BIG_ENDIAN
956 @item BITS_BIG_ENDIAN
957 Define this macro to have the value 1 if the most significant bit in a
958 byte has the lowest number; otherwise define it to have the value zero.
959 This means that bit-field instructions count from the most significant
960 bit. If the machine has no bit-field instructions, then this must still
961 be defined, but it doesn't matter which value it is defined to. This
962 macro need not be a constant.
964 This macro does not affect the way structure fields are packed into
965 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
967 @findex BYTES_BIG_ENDIAN
968 @item BYTES_BIG_ENDIAN
969 Define this macro to have the value 1 if the most significant byte in a
970 word has the lowest number. This macro need not be a constant.
972 @findex WORDS_BIG_ENDIAN
973 @item WORDS_BIG_ENDIAN
974 Define this macro to have the value 1 if, in a multiword object, the
975 most significant word has the lowest number. This applies to both
976 memory locations and registers; GCC fundamentally assumes that the
977 order of words in memory is the same as the order in registers. This
978 macro need not be a constant.
980 @findex LIBGCC2_WORDS_BIG_ENDIAN
981 @item LIBGCC2_WORDS_BIG_ENDIAN
982 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
983 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
984 used only when compiling @file{libgcc2.c}. Typically the value will be set
985 based on preprocessor defines.
987 @findex FLOAT_WORDS_BIG_ENDIAN
988 @item FLOAT_WORDS_BIG_ENDIAN
989 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
990 @code{TFmode} floating point numbers are stored in memory with the word
991 containing the sign bit at the lowest address; otherwise define it to
992 have the value 0. This macro need not be a constant.
994 You need not define this macro if the ordering is the same as for
997 @findex BITS_PER_UNIT
999 Define this macro to be the number of bits in an addressable storage
1000 unit (byte). If you do not define this macro the default is 8.
1002 @findex BITS_PER_WORD
1004 Number of bits in a word. If you do not define this macro, the default
1005 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
1007 @findex MAX_BITS_PER_WORD
1008 @item MAX_BITS_PER_WORD
1009 Maximum number of bits in a word. If this is undefined, the default is
1010 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
1011 largest value that @code{BITS_PER_WORD} can have at run-time.
1013 @findex UNITS_PER_WORD
1014 @item UNITS_PER_WORD
1015 Number of storage units in a word; normally 4.
1017 @findex MIN_UNITS_PER_WORD
1018 @item MIN_UNITS_PER_WORD
1019 Minimum number of units in a word. If this is undefined, the default is
1020 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
1021 smallest value that @code{UNITS_PER_WORD} can have at run-time.
1023 @findex POINTER_SIZE
1025 Width of a pointer, in bits. You must specify a value no wider than the
1026 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
1027 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
1028 a value the default is @code{BITS_PER_WORD}.
1030 @findex POINTERS_EXTEND_UNSIGNED
1031 @item POINTERS_EXTEND_UNSIGNED
1032 A C expression whose value is greater than zero if pointers that need to be
1033 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
1034 be zero-extended and zero if they are to be sign-extended. If the value
1035 is less then zero then there must be an "ptr_extend" instruction that
1036 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
1038 You need not define this macro if the @code{POINTER_SIZE} is equal
1039 to the width of @code{Pmode}.
1041 @findex PROMOTE_MODE
1042 @item PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
1043 A macro to update @var{m} and @var{unsignedp} when an object whose type
1044 is @var{type} and which has the specified mode and signedness is to be
1045 stored in a register. This macro is only called when @var{type} is a
1048 On most RISC machines, which only have operations that operate on a full
1049 register, define this macro to set @var{m} to @code{word_mode} if
1050 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
1051 cases, only integer modes should be widened because wider-precision
1052 floating-point operations are usually more expensive than their narrower
1055 For most machines, the macro definition does not change @var{unsignedp}.
1056 However, some machines, have instructions that preferentially handle
1057 either signed or unsigned quantities of certain modes. For example, on
1058 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
1059 sign-extend the result to 64 bits. On such machines, set
1060 @var{unsignedp} according to which kind of extension is more efficient.
1062 Do not define this macro if it would never modify @var{m}.
1064 @findex PROMOTE_FUNCTION_ARGS
1065 @item PROMOTE_FUNCTION_ARGS
1066 Define this macro if the promotion described by @code{PROMOTE_MODE}
1067 should also be done for outgoing function arguments.
1069 @findex PROMOTE_FUNCTION_RETURN
1070 @item PROMOTE_FUNCTION_RETURN
1071 Define this macro if the promotion described by @code{PROMOTE_MODE}
1072 should also be done for the return value of functions.
1074 If this macro is defined, @code{FUNCTION_VALUE} must perform the same
1075 promotions done by @code{PROMOTE_MODE}.
1077 @findex PROMOTE_FOR_CALL_ONLY
1078 @item PROMOTE_FOR_CALL_ONLY
1079 Define this macro if the promotion described by @code{PROMOTE_MODE}
1080 should @emph{only} be performed for outgoing function arguments or
1081 function return values, as specified by @code{PROMOTE_FUNCTION_ARGS}
1082 and @code{PROMOTE_FUNCTION_RETURN}, respectively.
1084 @findex PARM_BOUNDARY
1086 Normal alignment required for function parameters on the stack, in
1087 bits. All stack parameters receive at least this much alignment
1088 regardless of data type. On most machines, this is the same as the
1091 @findex STACK_BOUNDARY
1092 @item STACK_BOUNDARY
1093 Define this macro to the minimum alignment enforced by hardware for the
1094 stack pointer on this machine. The definition is a C expression for the
1095 desired alignment (measured in bits). This value is used as a default
1096 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1097 this should be the same as @code{PARM_BOUNDARY}.
1099 @findex PREFERRED_STACK_BOUNDARY
1100 @item PREFERRED_STACK_BOUNDARY
1101 Define this macro if you wish to preserve a certain alignment for the
1102 stack pointer, greater than what the hardware enforces. The definition
1103 is a C expression for the desired alignment (measured in bits). This
1104 macro must evaluate to a value equal to or larger than
1105 @code{STACK_BOUNDARY}.
1107 @findex FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1108 @item FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1109 A C expression that evaluates true if @code{PREFERRED_STACK_BOUNDARY} is
1110 not guaranteed by the runtime and we should emit code to align the stack
1111 at the beginning of @code{main}.
1113 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
1114 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
1115 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
1116 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
1117 be momentarily unaligned while pushing arguments.
1119 @findex FUNCTION_BOUNDARY
1120 @item FUNCTION_BOUNDARY
1121 Alignment required for a function entry point, in bits.
1123 @findex BIGGEST_ALIGNMENT
1124 @item BIGGEST_ALIGNMENT
1125 Biggest alignment that any data type can require on this machine, in bits.
1127 @findex MINIMUM_ATOMIC_ALIGNMENT
1128 @item MINIMUM_ATOMIC_ALIGNMENT
1129 If defined, the smallest alignment, in bits, that can be given to an
1130 object that can be referenced in one operation, without disturbing any
1131 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1132 on machines that don't have byte or half-word store operations.
1134 @findex BIGGEST_FIELD_ALIGNMENT
1135 @item BIGGEST_FIELD_ALIGNMENT
1136 Biggest alignment that any structure or union field can require on this
1137 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1138 structure and union fields only, unless the field alignment has been set
1139 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1141 @findex ADJUST_FIELD_ALIGN
1142 @item ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1143 An expression for the alignment of a structure field @var{field} if the
1144 alignment computed in the usual way (including applying of
1145 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1146 alignment) is @var{computed}. It overrides alignment only if the
1147 field alignment has not been set by the
1148 @code{__attribute__ ((aligned (@var{n})))} construct.
1150 @findex MAX_OFILE_ALIGNMENT
1151 @item MAX_OFILE_ALIGNMENT
1152 Biggest alignment supported by the object file format of this machine.
1153 Use this macro to limit the alignment which can be specified using the
1154 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1155 the default value is @code{BIGGEST_ALIGNMENT}.
1157 @findex DATA_ALIGNMENT
1158 @item DATA_ALIGNMENT (@var{type}, @var{basic-align})
1159 If defined, a C expression to compute the alignment for a variable in
1160 the static store. @var{type} is the data type, and @var{basic-align} is
1161 the alignment that the object would ordinarily have. The value of this
1162 macro is used instead of that alignment to align the object.
1164 If this macro is not defined, then @var{basic-align} is used.
1167 One use of this macro is to increase alignment of medium-size data to
1168 make it all fit in fewer cache lines. Another is to cause character
1169 arrays to be word-aligned so that @code{strcpy} calls that copy
1170 constants to character arrays can be done inline.
1172 @findex CONSTANT_ALIGNMENT
1173 @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1174 If defined, a C expression to compute the alignment given to a constant
1175 that is being placed in memory. @var{constant} is the constant and
1176 @var{basic-align} is the alignment that the object would ordinarily
1177 have. The value of this macro is used instead of that alignment to
1180 If this macro is not defined, then @var{basic-align} is used.
1182 The typical use of this macro is to increase alignment for string
1183 constants to be word aligned so that @code{strcpy} calls that copy
1184 constants can be done inline.
1186 @findex LOCAL_ALIGNMENT
1187 @item LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1188 If defined, a C expression to compute the alignment for a variable in
1189 the local store. @var{type} is the data type, and @var{basic-align} is
1190 the alignment that the object would ordinarily have. The value of this
1191 macro is used instead of that alignment to align the object.
1193 If this macro is not defined, then @var{basic-align} is used.
1195 One use of this macro is to increase alignment of medium-size data to
1196 make it all fit in fewer cache lines.
1198 @findex EMPTY_FIELD_BOUNDARY
1199 @item EMPTY_FIELD_BOUNDARY
1200 Alignment in bits to be given to a structure bit-field that follows an
1201 empty field such as @code{int : 0;}.
1203 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1205 @findex STRUCTURE_SIZE_BOUNDARY
1206 @item STRUCTURE_SIZE_BOUNDARY
1207 Number of bits which any structure or union's size must be a multiple of.
1208 Each structure or union's size is rounded up to a multiple of this.
1210 If you do not define this macro, the default is the same as
1211 @code{BITS_PER_UNIT}.
1213 @findex STRICT_ALIGNMENT
1214 @item STRICT_ALIGNMENT
1215 Define this macro to be the value 1 if instructions will fail to work
1216 if given data not on the nominal alignment. If instructions will merely
1217 go slower in that case, define this macro as 0.
1219 @findex PCC_BITFIELD_TYPE_MATTERS
1220 @item PCC_BITFIELD_TYPE_MATTERS
1221 Define this if you wish to imitate the way many other C compilers handle
1222 alignment of bit-fields and the structures that contain them.
1224 The behavior is that the type written for a named bit-field (@code{int},
1225 @code{short}, or other integer type) imposes an alignment for the entire
1226 structure, as if the structure really did contain an ordinary field of
1227 that type. In addition, the bit-field is placed within the structure so
1228 that it would fit within such a field, not crossing a boundary for it.
1230 Thus, on most machines, a named bit-field whose type is written as
1231 @code{int} would not cross a four-byte boundary, and would force
1232 four-byte alignment for the whole structure. (The alignment used may
1233 not be four bytes; it is controlled by the other alignment parameters.)
1235 An unnamed bit-field will not affect the alignment of the containing
1238 If the macro is defined, its definition should be a C expression;
1239 a nonzero value for the expression enables this behavior.
1241 Note that if this macro is not defined, or its value is zero, some
1242 bit-fields may cross more than one alignment boundary. The compiler can
1243 support such references if there are @samp{insv}, @samp{extv}, and
1244 @samp{extzv} insns that can directly reference memory.
1246 The other known way of making bit-fields work is to define
1247 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1248 Then every structure can be accessed with fullwords.
1250 Unless the machine has bit-field instructions or you define
1251 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1252 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1254 If your aim is to make GCC use the same conventions for laying out
1255 bit-fields as are used by another compiler, here is how to investigate
1256 what the other compiler does. Compile and run this program:
1275 printf ("Size of foo1 is %d\n",
1276 sizeof (struct foo1));
1277 printf ("Size of foo2 is %d\n",
1278 sizeof (struct foo2));
1283 If this prints 2 and 5, then the compiler's behavior is what you would
1284 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1286 @findex BITFIELD_NBYTES_LIMITED
1287 @item BITFIELD_NBYTES_LIMITED
1288 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1289 to aligning a bit-field within the structure.
1291 @findex MEMBER_TYPE_FORCES_BLK
1292 @item MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1293 Return 1 if a structure or array containing @var{field} should be accessed using
1296 If @var{field} is the only field in the structure, @var{mode} is its
1297 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1298 case where structures of one field would require the structure's mode to
1299 retain the field's mode.
1301 Normally, this is not needed. See the file @file{c4x.h} for an example
1302 of how to use this macro to prevent a structure having a floating point
1303 field from being accessed in an integer mode.
1305 @findex ROUND_TYPE_SIZE
1306 @item ROUND_TYPE_SIZE (@var{type}, @var{computed}, @var{specified})
1307 Define this macro as an expression for the overall size of a type
1308 (given by @var{type} as a tree node) when the size computed in the
1309 usual way is @var{computed} and the alignment is @var{specified}.
1311 The default is to round @var{computed} up to a multiple of @var{specified}.
1313 @findex ROUND_TYPE_SIZE_UNIT
1314 @item ROUND_TYPE_SIZE_UNIT (@var{type}, @var{computed}, @var{specified})
1315 Similar to @code{ROUND_TYPE_SIZE}, but sizes and alignments are
1316 specified in units (bytes). If you define @code{ROUND_TYPE_SIZE},
1317 you must also define this macro and they must be defined consistently
1320 @findex ROUND_TYPE_ALIGN
1321 @item ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1322 Define this macro as an expression for the alignment of a type (given
1323 by @var{type} as a tree node) if the alignment computed in the usual
1324 way is @var{computed} and the alignment explicitly specified was
1327 The default is to use @var{specified} if it is larger; otherwise, use
1328 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1330 @findex MAX_FIXED_MODE_SIZE
1331 @item MAX_FIXED_MODE_SIZE
1332 An integer expression for the size in bits of the largest integer
1333 machine mode that should actually be used. All integer machine modes of
1334 this size or smaller can be used for structures and unions with the
1335 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1336 (DImode)} is assumed.
1338 @findex VECTOR_MODE_SUPPORTED_P
1339 @item VECTOR_MODE_SUPPORTED_P(@var{mode})
1340 Define this macro to be nonzero if the port is prepared to handle insns
1341 involving vector mode @var{mode}. At the very least, it must have move
1342 patterns for this mode.
1344 @findex STACK_SAVEAREA_MODE
1345 @item STACK_SAVEAREA_MODE (@var{save_level})
1346 If defined, an expression of type @code{enum machine_mode} that
1347 specifies the mode of the save area operand of a
1348 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1349 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1350 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1351 having its mode specified.
1353 You need not define this macro if it always returns @code{Pmode}. You
1354 would most commonly define this macro if the
1355 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1358 @findex STACK_SIZE_MODE
1359 @item STACK_SIZE_MODE
1360 If defined, an expression of type @code{enum machine_mode} that
1361 specifies the mode of the size increment operand of an
1362 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1364 You need not define this macro if it always returns @code{word_mode}.
1365 You would most commonly define this macro if the @code{allocate_stack}
1366 pattern needs to support both a 32- and a 64-bit mode.
1368 @findex TARGET_FLOAT_FORMAT
1369 @item TARGET_FLOAT_FORMAT
1370 A code distinguishing the floating point format of the target machine.
1371 There are five defined values:
1374 @findex IEEE_FLOAT_FORMAT
1375 @item IEEE_FLOAT_FORMAT
1376 This code indicates IEEE floating point. It is the default; there is no
1377 need to define this macro when the format is IEEE@.
1379 @findex VAX_FLOAT_FORMAT
1380 @item VAX_FLOAT_FORMAT
1381 This code indicates the ``F float'' (for @code{float}) and ``D float''
1382 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
1384 @findex IBM_FLOAT_FORMAT
1385 @item IBM_FLOAT_FORMAT
1386 This code indicates the format used on the IBM System/370.
1388 @findex C4X_FLOAT_FORMAT
1389 @item C4X_FLOAT_FORMAT
1390 This code indicates the format used on the TMS320C3x/C4x.
1392 @findex UNKNOWN_FLOAT_FORMAT
1393 @item UNKNOWN_FLOAT_FORMAT
1394 This code indicates any other format.
1398 formats are actually in use on supported machines, new codes should be
1401 The ordering of the component words of floating point values stored in
1402 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1404 @findex MODE_HAS_NANS
1405 @item MODE_HAS_NANS (@var{mode})
1406 When defined, this macro should be true if @var{mode} has a NaN
1407 representation. The compiler assumes that NaNs are not equal to
1408 anything (including themselves) and that addition, subtraction,
1409 multiplication and division all return NaNs when one operand is
1412 By default, this macro is true if @var{mode} is a floating-point
1413 mode and the target floating-point format is IEEE@.
1415 @findex MODE_HAS_INFINITIES
1416 @item MODE_HAS_INFINITIES (@var{mode})
1417 This macro should be true if @var{mode} can represent infinity. At
1418 present, the compiler uses this macro to decide whether @samp{x - x}
1419 is always defined. By default, the macro is true when @var{mode}
1420 is a floating-point mode and the target format is IEEE@.
1422 @findex MODE_HAS_SIGNED_ZEROS
1423 @item MODE_HAS_SIGNED_ZEROS (@var{mode})
1424 True if @var{mode} distinguishes between positive and negative zero.
1425 The rules are expected to follow the IEEE standard:
1429 @samp{x + x} has the same sign as @samp{x}.
1432 If the sum of two values with opposite sign is zero, the result is
1433 positive for all rounding modes expect towards @minus{}infinity, for
1434 which it is negative.
1437 The sign of a product or quotient is negative when exactly one
1438 of the operands is negative.
1441 The default definition is true if @var{mode} is a floating-point
1442 mode and the target format is IEEE@.
1444 @findex MODE_HAS_SIGN_DEPENDENT_ROUNDING
1445 @item MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1446 If defined, this macro should be true for @var{mode} if it has at
1447 least one rounding mode in which @samp{x} and @samp{-x} can be
1448 rounded to numbers of different magnitude. Two such modes are
1449 towards @minus{}infinity and towards +infinity.
1451 The default definition of this macro is true if @var{mode} is
1452 a floating-point mode and the target format is IEEE@.
1454 @findex ROUND_TOWARDS_ZERO
1455 @item ROUND_TOWARDS_ZERO
1456 If defined, this macro should be true if the prevailing rounding
1457 mode is towards zero. A true value has the following effects:
1461 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1464 @file{libgcc.a}'s floating-point emulator will round towards zero
1465 rather than towards nearest.
1468 The compiler's floating-point emulator will round towards zero after
1469 doing arithmetic, and when converting from the internal float format to
1473 The macro does not affect the parsing of string literals. When the
1474 primary rounding mode is towards zero, library functions like
1475 @code{strtod} might still round towards nearest, and the compiler's
1476 parser should behave like the target's @code{strtod} where possible.
1478 Not defining this macro is equivalent to returning zero.
1480 @findex LARGEST_EXPONENT_IS_NORMAL
1481 @item LARGEST_EXPONENT_IS_NORMAL (@var{size})
1482 This macro should return true if floats with @var{size}
1483 bits do not have a NaN or infinity representation, but use the largest
1484 exponent for normal numbers instead.
1486 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1487 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1488 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1489 floating-point arithmetic.
1491 The default definition of this macro returns false for all sizes.
1494 @deftypefn {Target Hook} bool TARGET_VECTOR_OPAQUE_P (tree @var{type})
1495 This target hook should return @code{true} a vector is opaque. That
1496 is, if no cast is needed when copying a vector value of type
1497 @var{type} into another vector lvalue of the same size. Vector opaque
1498 types cannot be initialized. The default is that there are no such
1502 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1503 This target hook returns @code{true} if bit-fields in the given
1504 @var{record_type} are to be laid out following the rules of Microsoft
1505 Visual C/C++, namely: (i) a bit-field won't share the same storage
1506 unit with the previous bit-field if their underlying types have
1507 different sizes, and the bit-field will be aligned to the highest
1508 alignment of the underlying types of itself and of the previous
1509 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1510 the whole enclosing structure, even if it is unnamed; except that
1511 (iii) a zero-sized bit-field will be disregarded unless it follows
1512 another bit-field of nonzero size. If this hook returns @code{true},
1513 other macros that control bit-field layout are ignored.
1515 When a bit-field is inserted into a packed record, the whole size
1516 of the underlying type is used by one or more same-size adjacent
1517 bit-fields (that is, if its long:3, 32 bits is used in the record,
1518 and any additional adjacent long bit-fields are packed into the same
1519 chunk of 32 bits. However, if the size changes, a new field of that
1520 size is allocated). In an unpacked record, this is the same as using
1521 alignment, but not equivalent when packing.
1523 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1524 the latter will take precedence. If @samp{__attribute__((packed))} is
1525 used on a single field when MS bit-fields are in use, it will take
1526 precedence for that field, but the alignment of the rest of the structure
1527 may affect its placement.
1531 @section Layout of Source Language Data Types
1533 These macros define the sizes and other characteristics of the standard
1534 basic data types used in programs being compiled. Unlike the macros in
1535 the previous section, these apply to specific features of C and related
1536 languages, rather than to fundamental aspects of storage layout.
1539 @findex INT_TYPE_SIZE
1541 A C expression for the size in bits of the type @code{int} on the
1542 target machine. If you don't define this, the default is one word.
1544 @findex SHORT_TYPE_SIZE
1545 @item SHORT_TYPE_SIZE
1546 A C expression for the size in bits of the type @code{short} on the
1547 target machine. If you don't define this, the default is half a word.
1548 (If this would be less than one storage unit, it is rounded up to one
1551 @findex LONG_TYPE_SIZE
1552 @item LONG_TYPE_SIZE
1553 A C expression for the size in bits of the type @code{long} on the
1554 target machine. If you don't define this, the default is one word.
1556 @findex ADA_LONG_TYPE_SIZE
1557 @item ADA_LONG_TYPE_SIZE
1558 On some machines, the size used for the Ada equivalent of the type
1559 @code{long} by a native Ada compiler differs from that used by C. In
1560 that situation, define this macro to be a C expression to be used for
1561 the size of that type. If you don't define this, the default is the
1562 value of @code{LONG_TYPE_SIZE}.
1564 @findex MAX_LONG_TYPE_SIZE
1565 @item MAX_LONG_TYPE_SIZE
1566 Maximum number for the size in bits of the type @code{long} on the
1567 target machine. If this is undefined, the default is
1568 @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the
1569 largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is
1572 @findex LONG_LONG_TYPE_SIZE
1573 @item LONG_LONG_TYPE_SIZE
1574 A C expression for the size in bits of the type @code{long long} on the
1575 target machine. If you don't define this, the default is two
1576 words. If you want to support GNU Ada on your machine, the value of this
1577 macro must be at least 64.
1579 @findex CHAR_TYPE_SIZE
1580 @item CHAR_TYPE_SIZE
1581 A C expression for the size in bits of the type @code{char} on the
1582 target machine. If you don't define this, the default is
1583 @code{BITS_PER_UNIT}.
1585 @findex BOOL_TYPE_SIZE
1586 @item BOOL_TYPE_SIZE
1587 A C expression for the size in bits of the C++ type @code{bool} and
1588 C99 type @code{_Bool} on the target machine. If you don't define
1589 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1591 @findex FLOAT_TYPE_SIZE
1592 @item FLOAT_TYPE_SIZE
1593 A C expression for the size in bits of the type @code{float} on the
1594 target machine. If you don't define this, the default is one word.
1596 @findex DOUBLE_TYPE_SIZE
1597 @item DOUBLE_TYPE_SIZE
1598 A C expression for the size in bits of the type @code{double} on the
1599 target machine. If you don't define this, the default is two
1602 @findex LONG_DOUBLE_TYPE_SIZE
1603 @item LONG_DOUBLE_TYPE_SIZE
1604 A C expression for the size in bits of the type @code{long double} on
1605 the target machine. If you don't define this, the default is two
1608 @findex MAX_LONG_DOUBLE_TYPE_SIZE
1609 Maximum number for the size in bits of the type @code{long double} on the
1610 target machine. If this is undefined, the default is
1611 @code{LONG_DOUBLE_TYPE_SIZE}. Otherwise, it is the constant value that is
1612 the largest value that @code{LONG_DOUBLE_TYPE_SIZE} can have at run-time.
1613 This is used in @code{cpp}.
1615 @findex TARGET_FLT_EVAL_METHOD
1616 @item TARGET_FLT_EVAL_METHOD
1617 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1618 assuming, if applicable, that the floating-point control word is in its
1619 default state. If you do not define this macro the value of
1620 @code{FLT_EVAL_METHOD} will be zero.
1622 @findex WIDEST_HARDWARE_FP_SIZE
1623 @item WIDEST_HARDWARE_FP_SIZE
1624 A C expression for the size in bits of the widest floating-point format
1625 supported by the hardware. If you define this macro, you must specify a
1626 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1627 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1630 @findex DEFAULT_SIGNED_CHAR
1631 @item DEFAULT_SIGNED_CHAR
1632 An expression whose value is 1 or 0, according to whether the type
1633 @code{char} should be signed or unsigned by default. The user can
1634 always override this default with the options @option{-fsigned-char}
1635 and @option{-funsigned-char}.
1637 @findex DEFAULT_SHORT_ENUMS
1638 @item DEFAULT_SHORT_ENUMS
1639 A C expression to determine whether to give an @code{enum} type
1640 only as many bytes as it takes to represent the range of possible values
1641 of that type. A nonzero value means to do that; a zero value means all
1642 @code{enum} types should be allocated like @code{int}.
1644 If you don't define the macro, the default is 0.
1648 A C expression for a string describing the name of the data type to use
1649 for size values. The typedef name @code{size_t} is defined using the
1650 contents of the string.
1652 The string can contain more than one keyword. If so, separate them with
1653 spaces, and write first any length keyword, then @code{unsigned} if
1654 appropriate, and finally @code{int}. The string must exactly match one
1655 of the data type names defined in the function
1656 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1657 omit @code{int} or change the order---that would cause the compiler to
1660 If you don't define this macro, the default is @code{"long unsigned
1663 @findex PTRDIFF_TYPE
1665 A C expression for a string describing the name of the data type to use
1666 for the result of subtracting two pointers. The typedef name
1667 @code{ptrdiff_t} is defined using the contents of the string. See
1668 @code{SIZE_TYPE} above for more information.
1670 If you don't define this macro, the default is @code{"long int"}.
1674 A C expression for a string describing the name of the data type to use
1675 for wide characters. The typedef name @code{wchar_t} is defined using
1676 the contents of the string. See @code{SIZE_TYPE} above for more
1679 If you don't define this macro, the default is @code{"int"}.
1681 @findex WCHAR_TYPE_SIZE
1682 @item WCHAR_TYPE_SIZE
1683 A C expression for the size in bits of the data type for wide
1684 characters. This is used in @code{cpp}, which cannot make use of
1687 @findex MAX_WCHAR_TYPE_SIZE
1688 @item MAX_WCHAR_TYPE_SIZE
1689 Maximum number for the size in bits of the data type for wide
1690 characters. If this is undefined, the default is
1691 @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1692 largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is
1695 @findex GCOV_TYPE_SIZE
1696 @item GCOV_TYPE_SIZE
1697 A C expression for the size in bits of the type used for gcov counters on the
1698 target machine. If you don't define this, the default is one
1699 @code{LONG_TYPE_SIZE} in case it is greater or equal to 64-bit and
1700 @code{LONG_LONG_TYPE_SIZE} otherwise. You may want to re-define the type to
1701 ensure atomicity for counters in multithreaded programs.
1705 A C expression for a string describing the name of the data type to
1706 use for wide characters passed to @code{printf} and returned from
1707 @code{getwc}. The typedef name @code{wint_t} is defined using the
1708 contents of the string. See @code{SIZE_TYPE} above for more
1711 If you don't define this macro, the default is @code{"unsigned int"}.
1715 A C expression for a string describing the name of the data type that
1716 can represent any value of any standard or extended signed integer type.
1717 The typedef name @code{intmax_t} is defined using the contents of the
1718 string. See @code{SIZE_TYPE} above for more information.
1720 If you don't define this macro, the default is the first of
1721 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1722 much precision as @code{long long int}.
1724 @findex UINTMAX_TYPE
1726 A C expression for a string describing the name of the data type that
1727 can represent any value of any standard or extended unsigned integer
1728 type. The typedef name @code{uintmax_t} is defined using the contents
1729 of the string. See @code{SIZE_TYPE} above for more information.
1731 If you don't define this macro, the default is the first of
1732 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1733 unsigned int"} that has as much precision as @code{long long unsigned
1736 @findex TARGET_PTRMEMFUNC_VBIT_LOCATION
1737 @item TARGET_PTRMEMFUNC_VBIT_LOCATION
1738 The C++ compiler represents a pointer-to-member-function with a struct
1745 ptrdiff_t vtable_index;
1752 The C++ compiler must use one bit to indicate whether the function that
1753 will be called through a pointer-to-member-function is virtual.
1754 Normally, we assume that the low-order bit of a function pointer must
1755 always be zero. Then, by ensuring that the vtable_index is odd, we can
1756 distinguish which variant of the union is in use. But, on some
1757 platforms function pointers can be odd, and so this doesn't work. In
1758 that case, we use the low-order bit of the @code{delta} field, and shift
1759 the remainder of the @code{delta} field to the left.
1761 GCC will automatically make the right selection about where to store
1762 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1763 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1764 set such that functions always start at even addresses, but the lowest
1765 bit of pointers to functions indicate whether the function at that
1766 address is in ARM or Thumb mode. If this is the case of your
1767 architecture, you should define this macro to
1768 @code{ptrmemfunc_vbit_in_delta}.
1770 In general, you should not have to define this macro. On architectures
1771 in which function addresses are always even, according to
1772 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1773 @code{ptrmemfunc_vbit_in_pfn}.
1775 @findex TARGET_VTABLE_USES_DESCRIPTORS
1776 @item TARGET_VTABLE_USES_DESCRIPTORS
1777 Normally, the C++ compiler uses function pointers in vtables. This
1778 macro allows the target to change to use ``function descriptors''
1779 instead. Function descriptors are found on targets for whom a
1780 function pointer is actually a small data structure. Normally the
1781 data structure consists of the actual code address plus a data
1782 pointer to which the function's data is relative.
1784 If vtables are used, the value of this macro should be the number
1785 of words that the function descriptor occupies.
1787 @findex TARGET_VTABLE_ENTRY_ALIGN
1788 @item TARGET_VTABLE_ENTRY_ALIGN
1789 By default, the vtable entries are void pointers, the so the alignment
1790 is the same as pointer alignment. The value of this macro specifies
1791 the alignment of the vtable entry in bits. It should be defined only
1792 when special alignment is necessary. */
1794 @findex TARGET_VTABLE_DATA_ENTRY_DISTANCE
1795 @item TARGET_VTABLE_DATA_ENTRY_DISTANCE
1796 There are a few non-descriptor entries in the vtable at offsets below
1797 zero. If these entries must be padded (say, to preserve the alignment
1798 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1799 of words in each data entry.
1802 @node Escape Sequences
1803 @section Target Character Escape Sequences
1804 @cindex escape sequences
1806 By default, GCC assumes that the C character escape sequences take on
1807 their ASCII values for the target. If this is not correct, you must
1808 explicitly define all of the macros below.
1813 A C constant expression for the integer value for escape sequence
1818 A C constant expression for the integer value of the target escape
1819 character. As an extension, GCC evaluates the escape sequences
1820 @samp{\e} and @samp{\E} to this.
1824 @findex TARGET_NEWLINE
1827 @itemx TARGET_NEWLINE
1828 C constant expressions for the integer values for escape sequences
1829 @samp{\b}, @samp{\t} and @samp{\n}.
1837 C constant expressions for the integer values for escape sequences
1838 @samp{\v}, @samp{\f} and @samp{\r}.
1842 @section Register Usage
1843 @cindex register usage
1845 This section explains how to describe what registers the target machine
1846 has, and how (in general) they can be used.
1848 The description of which registers a specific instruction can use is
1849 done with register classes; see @ref{Register Classes}. For information
1850 on using registers to access a stack frame, see @ref{Frame Registers}.
1851 For passing values in registers, see @ref{Register Arguments}.
1852 For returning values in registers, see @ref{Scalar Return}.
1855 * Register Basics:: Number and kinds of registers.
1856 * Allocation Order:: Order in which registers are allocated.
1857 * Values in Registers:: What kinds of values each reg can hold.
1858 * Leaf Functions:: Renumbering registers for leaf functions.
1859 * Stack Registers:: Handling a register stack such as 80387.
1862 @node Register Basics
1863 @subsection Basic Characteristics of Registers
1865 @c prevent bad page break with this line
1866 Registers have various characteristics.
1869 @findex FIRST_PSEUDO_REGISTER
1870 @item FIRST_PSEUDO_REGISTER
1871 Number of hardware registers known to the compiler. They receive
1872 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1873 pseudo register's number really is assigned the number
1874 @code{FIRST_PSEUDO_REGISTER}.
1876 @item FIXED_REGISTERS
1877 @findex FIXED_REGISTERS
1878 @cindex fixed register
1879 An initializer that says which registers are used for fixed purposes
1880 all throughout the compiled code and are therefore not available for
1881 general allocation. These would include the stack pointer, the frame
1882 pointer (except on machines where that can be used as a general
1883 register when no frame pointer is needed), the program counter on
1884 machines where that is considered one of the addressable registers,
1885 and any other numbered register with a standard use.
1887 This information is expressed as a sequence of numbers, separated by
1888 commas and surrounded by braces. The @var{n}th number is 1 if
1889 register @var{n} is fixed, 0 otherwise.
1891 The table initialized from this macro, and the table initialized by
1892 the following one, may be overridden at run time either automatically,
1893 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1894 the user with the command options @option{-ffixed-@var{reg}},
1895 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1897 @findex CALL_USED_REGISTERS
1898 @item CALL_USED_REGISTERS
1899 @cindex call-used register
1900 @cindex call-clobbered register
1901 @cindex call-saved register
1902 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1903 clobbered (in general) by function calls as well as for fixed
1904 registers. This macro therefore identifies the registers that are not
1905 available for general allocation of values that must live across
1908 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1909 automatically saves it on function entry and restores it on function
1910 exit, if the register is used within the function.
1912 @findex CALL_REALLY_USED_REGISTERS
1913 @item CALL_REALLY_USED_REGISTERS
1914 @cindex call-used register
1915 @cindex call-clobbered register
1916 @cindex call-saved register
1917 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1918 that the entire set of @code{FIXED_REGISTERS} be included.
1919 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1920 This macro is optional. If not specified, it defaults to the value
1921 of @code{CALL_USED_REGISTERS}.
1923 @findex HARD_REGNO_CALL_PART_CLOBBERED
1924 @item HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1925 @cindex call-used register
1926 @cindex call-clobbered register
1927 @cindex call-saved register
1928 A C expression that is nonzero if it is not permissible to store a
1929 value of mode @var{mode} in hard register number @var{regno} across a
1930 call without some part of it being clobbered. For most machines this
1931 macro need not be defined. It is only required for machines that do not
1932 preserve the entire contents of a register across a call.
1934 @findex CONDITIONAL_REGISTER_USAGE
1936 @findex call_used_regs
1937 @item CONDITIONAL_REGISTER_USAGE
1938 Zero or more C statements that may conditionally modify five variables
1939 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1940 @code{reg_names}, and @code{reg_class_contents}, to take into account
1941 any dependence of these register sets on target flags. The first three
1942 of these are of type @code{char []} (interpreted as Boolean vectors).
1943 @code{global_regs} is a @code{const char *[]}, and
1944 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1945 called, @code{fixed_regs}, @code{call_used_regs},
1946 @code{reg_class_contents}, and @code{reg_names} have been initialized
1947 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1948 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1949 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1950 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1951 command options have been applied.
1953 You need not define this macro if it has no work to do.
1955 @cindex disabling certain registers
1956 @cindex controlling register usage
1957 If the usage of an entire class of registers depends on the target
1958 flags, you may indicate this to GCC by using this macro to modify
1959 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1960 registers in the classes which should not be used by GCC@. Also define
1961 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1962 to return @code{NO_REGS} if it
1963 is called with a letter for a class that shouldn't be used.
1965 (However, if this class is not included in @code{GENERAL_REGS} and all
1966 of the insn patterns whose constraints permit this class are
1967 controlled by target switches, then GCC will automatically avoid using
1968 these registers when the target switches are opposed to them.)
1970 @findex NON_SAVING_SETJMP
1971 @item NON_SAVING_SETJMP
1972 If this macro is defined and has a nonzero value, it means that
1973 @code{setjmp} and related functions fail to save the registers, or that
1974 @code{longjmp} fails to restore them. To compensate, the compiler
1975 avoids putting variables in registers in functions that use
1978 @findex INCOMING_REGNO
1979 @item INCOMING_REGNO (@var{out})
1980 Define this macro if the target machine has register windows. This C
1981 expression returns the register number as seen by the called function
1982 corresponding to the register number @var{out} as seen by the calling
1983 function. Return @var{out} if register number @var{out} is not an
1986 @findex OUTGOING_REGNO
1987 @item OUTGOING_REGNO (@var{in})
1988 Define this macro if the target machine has register windows. This C
1989 expression returns the register number as seen by the calling function
1990 corresponding to the register number @var{in} as seen by the called
1991 function. Return @var{in} if register number @var{in} is not an inbound
1995 @item LOCAL_REGNO (@var{regno})
1996 Define this macro if the target machine has register windows. This C
1997 expression returns true if the register is call-saved but is in the
1998 register window. Unlike most call-saved registers, such registers
1999 need not be explicitly restored on function exit or during non-local
2005 If the program counter has a register number, define this as that
2006 register number. Otherwise, do not define it.
2010 @node Allocation Order
2011 @subsection Order of Allocation of Registers
2012 @cindex order of register allocation
2013 @cindex register allocation order
2015 @c prevent bad page break with this line
2016 Registers are allocated in order.
2019 @findex REG_ALLOC_ORDER
2020 @item REG_ALLOC_ORDER
2021 If defined, an initializer for a vector of integers, containing the
2022 numbers of hard registers in the order in which GCC should prefer
2023 to use them (from most preferred to least).
2025 If this macro is not defined, registers are used lowest numbered first
2026 (all else being equal).
2028 One use of this macro is on machines where the highest numbered
2029 registers must always be saved and the save-multiple-registers
2030 instruction supports only sequences of consecutive registers. On such
2031 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
2032 the highest numbered allocable register first.
2034 @findex ORDER_REGS_FOR_LOCAL_ALLOC
2035 @item ORDER_REGS_FOR_LOCAL_ALLOC
2036 A C statement (sans semicolon) to choose the order in which to allocate
2037 hard registers for pseudo-registers local to a basic block.
2039 Store the desired register order in the array @code{reg_alloc_order}.
2040 Element 0 should be the register to allocate first; element 1, the next
2041 register; and so on.
2043 The macro body should not assume anything about the contents of
2044 @code{reg_alloc_order} before execution of the macro.
2046 On most machines, it is not necessary to define this macro.
2049 @node Values in Registers
2050 @subsection How Values Fit in Registers
2052 This section discusses the macros that describe which kinds of values
2053 (specifically, which machine modes) each register can hold, and how many
2054 consecutive registers are needed for a given mode.
2057 @findex HARD_REGNO_NREGS
2058 @item HARD_REGNO_NREGS (@var{regno}, @var{mode})
2059 A C expression for the number of consecutive hard registers, starting
2060 at register number @var{regno}, required to hold a value of mode
2063 On a machine where all registers are exactly one word, a suitable
2064 definition of this macro is
2067 #define HARD_REGNO_NREGS(REGNO, MODE) \
2068 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2072 @findex HARD_REGNO_MODE_OK
2073 @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2074 A C expression that is nonzero if it is permissible to store a value
2075 of mode @var{mode} in hard register number @var{regno} (or in several
2076 registers starting with that one). For a machine where all registers
2077 are equivalent, a suitable definition is
2080 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2083 You need not include code to check for the numbers of fixed registers,
2084 because the allocation mechanism considers them to be always occupied.
2086 @cindex register pairs
2087 On some machines, double-precision values must be kept in even/odd
2088 register pairs. You can implement that by defining this macro to reject
2089 odd register numbers for such modes.
2091 The minimum requirement for a mode to be OK in a register is that the
2092 @samp{mov@var{mode}} instruction pattern support moves between the
2093 register and other hard register in the same class and that moving a
2094 value into the register and back out not alter it.
2096 Since the same instruction used to move @code{word_mode} will work for
2097 all narrower integer modes, it is not necessary on any machine for
2098 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2099 you define patterns @samp{movhi}, etc., to take advantage of this. This
2100 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2101 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2104 Many machines have special registers for floating point arithmetic.
2105 Often people assume that floating point machine modes are allowed only
2106 in floating point registers. This is not true. Any registers that
2107 can hold integers can safely @emph{hold} a floating point machine
2108 mode, whether or not floating arithmetic can be done on it in those
2109 registers. Integer move instructions can be used to move the values.
2111 On some machines, though, the converse is true: fixed-point machine
2112 modes may not go in floating registers. This is true if the floating
2113 registers normalize any value stored in them, because storing a
2114 non-floating value there would garble it. In this case,
2115 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2116 floating registers. But if the floating registers do not automatically
2117 normalize, if you can store any bit pattern in one and retrieve it
2118 unchanged without a trap, then any machine mode may go in a floating
2119 register, so you can define this macro to say so.
2121 The primary significance of special floating registers is rather that
2122 they are the registers acceptable in floating point arithmetic
2123 instructions. However, this is of no concern to
2124 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2125 constraints for those instructions.
2127 On some machines, the floating registers are especially slow to access,
2128 so that it is better to store a value in a stack frame than in such a
2129 register if floating point arithmetic is not being done. As long as the
2130 floating registers are not in class @code{GENERAL_REGS}, they will not
2131 be used unless some pattern's constraint asks for one.
2133 @findex MODES_TIEABLE_P
2134 @item MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2135 A C expression that is nonzero if a value of mode
2136 @var{mode1} is accessible in mode @var{mode2} without copying.
2138 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2139 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2140 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2141 should be nonzero. If they differ for any @var{r}, you should define
2142 this macro to return zero unless some other mechanism ensures the
2143 accessibility of the value in a narrower mode.
2145 You should define this macro to return nonzero in as many cases as
2146 possible since doing so will allow GCC to perform better register
2149 @findex AVOID_CCMODE_COPIES
2150 @item AVOID_CCMODE_COPIES
2151 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2152 registers. You should only define this macro if support for copying to/from
2153 @code{CCmode} is incomplete.
2156 @node Leaf Functions
2157 @subsection Handling Leaf Functions
2159 @cindex leaf functions
2160 @cindex functions, leaf
2161 On some machines, a leaf function (i.e., one which makes no calls) can run
2162 more efficiently if it does not make its own register window. Often this
2163 means it is required to receive its arguments in the registers where they
2164 are passed by the caller, instead of the registers where they would
2167 The special treatment for leaf functions generally applies only when
2168 other conditions are met; for example, often they may use only those
2169 registers for its own variables and temporaries. We use the term ``leaf
2170 function'' to mean a function that is suitable for this special
2171 handling, so that functions with no calls are not necessarily ``leaf
2174 GCC assigns register numbers before it knows whether the function is
2175 suitable for leaf function treatment. So it needs to renumber the
2176 registers in order to output a leaf function. The following macros
2180 @findex LEAF_REGISTERS
2181 @item LEAF_REGISTERS
2182 Name of a char vector, indexed by hard register number, which
2183 contains 1 for a register that is allowable in a candidate for leaf
2186 If leaf function treatment involves renumbering the registers, then the
2187 registers marked here should be the ones before renumbering---those that
2188 GCC would ordinarily allocate. The registers which will actually be
2189 used in the assembler code, after renumbering, should not be marked with 1
2192 Define this macro only if the target machine offers a way to optimize
2193 the treatment of leaf functions.
2195 @findex LEAF_REG_REMAP
2196 @item LEAF_REG_REMAP (@var{regno})
2197 A C expression whose value is the register number to which @var{regno}
2198 should be renumbered, when a function is treated as a leaf function.
2200 If @var{regno} is a register number which should not appear in a leaf
2201 function before renumbering, then the expression should yield @minus{}1, which
2202 will cause the compiler to abort.
2204 Define this macro only if the target machine offers a way to optimize the
2205 treatment of leaf functions, and registers need to be renumbered to do
2209 @findex current_function_is_leaf
2210 @findex current_function_uses_only_leaf_regs
2211 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2212 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2213 specially. They can test the C variable @code{current_function_is_leaf}
2214 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2215 set prior to local register allocation and is valid for the remaining
2216 compiler passes. They can also test the C variable
2217 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2218 functions which only use leaf registers.
2219 @code{current_function_uses_only_leaf_regs} is valid after reload and is
2220 only useful if @code{LEAF_REGISTERS} is defined.
2221 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2222 @c of the next paragraph?! --mew 2feb93
2224 @node Stack Registers
2225 @subsection Registers That Form a Stack
2227 There are special features to handle computers where some of the
2228 ``registers'' form a stack, as in the 80387 coprocessor for the 80386.
2229 Stack registers are normally written by pushing onto the stack, and are
2230 numbered relative to the top of the stack.
2232 Currently, GCC can only handle one group of stack-like registers, and
2233 they must be consecutively numbered.
2238 Define this if the machine has any stack-like registers.
2240 @findex FIRST_STACK_REG
2241 @item FIRST_STACK_REG
2242 The number of the first stack-like register. This one is the top
2245 @findex LAST_STACK_REG
2246 @item LAST_STACK_REG
2247 The number of the last stack-like register. This one is the bottom of
2251 @node Register Classes
2252 @section Register Classes
2253 @cindex register class definitions
2254 @cindex class definitions, register
2256 On many machines, the numbered registers are not all equivalent.
2257 For example, certain registers may not be allowed for indexed addressing;
2258 certain registers may not be allowed in some instructions. These machine
2259 restrictions are described to the compiler using @dfn{register classes}.
2261 You define a number of register classes, giving each one a name and saying
2262 which of the registers belong to it. Then you can specify register classes
2263 that are allowed as operands to particular instruction patterns.
2267 In general, each register will belong to several classes. In fact, one
2268 class must be named @code{ALL_REGS} and contain all the registers. Another
2269 class must be named @code{NO_REGS} and contain no registers. Often the
2270 union of two classes will be another class; however, this is not required.
2272 @findex GENERAL_REGS
2273 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2274 terribly special about the name, but the operand constraint letters
2275 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2276 the same as @code{ALL_REGS}, just define it as a macro which expands
2279 Order the classes so that if class @var{x} is contained in class @var{y}
2280 then @var{x} has a lower class number than @var{y}.
2282 The way classes other than @code{GENERAL_REGS} are specified in operand
2283 constraints is through machine-dependent operand constraint letters.
2284 You can define such letters to correspond to various classes, then use
2285 them in operand constraints.
2287 You should define a class for the union of two classes whenever some
2288 instruction allows both classes. For example, if an instruction allows
2289 either a floating point (coprocessor) register or a general register for a
2290 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2291 which includes both of them. Otherwise you will get suboptimal code.
2293 You must also specify certain redundant information about the register
2294 classes: for each class, which classes contain it and which ones are
2295 contained in it; for each pair of classes, the largest class contained
2298 When a value occupying several consecutive registers is expected in a
2299 certain class, all the registers used must belong to that class.
2300 Therefore, register classes cannot be used to enforce a requirement for
2301 a register pair to start with an even-numbered register. The way to
2302 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2304 Register classes used for input-operands of bitwise-and or shift
2305 instructions have a special requirement: each such class must have, for
2306 each fixed-point machine mode, a subclass whose registers can transfer that
2307 mode to or from memory. For example, on some machines, the operations for
2308 single-byte values (@code{QImode}) are limited to certain registers. When
2309 this is so, each register class that is used in a bitwise-and or shift
2310 instruction must have a subclass consisting of registers from which
2311 single-byte values can be loaded or stored. This is so that
2312 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2315 @findex enum reg_class
2316 @item enum reg_class
2317 An enumeral type that must be defined with all the register class names
2318 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
2319 must be the last register class, followed by one more enumeral value,
2320 @code{LIM_REG_CLASSES}, which is not a register class but rather
2321 tells how many classes there are.
2323 Each register class has a number, which is the value of casting
2324 the class name to type @code{int}. The number serves as an index
2325 in many of the tables described below.
2327 @findex N_REG_CLASSES
2329 The number of distinct register classes, defined as follows:
2332 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2335 @findex REG_CLASS_NAMES
2336 @item REG_CLASS_NAMES
2337 An initializer containing the names of the register classes as C string
2338 constants. These names are used in writing some of the debugging dumps.
2340 @findex REG_CLASS_CONTENTS
2341 @item REG_CLASS_CONTENTS
2342 An initializer containing the contents of the register classes, as integers
2343 which are bit masks. The @var{n}th integer specifies the contents of class
2344 @var{n}. The way the integer @var{mask} is interpreted is that
2345 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2347 When the machine has more than 32 registers, an integer does not suffice.
2348 Then the integers are replaced by sub-initializers, braced groupings containing
2349 several integers. Each sub-initializer must be suitable as an initializer
2350 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2351 In this situation, the first integer in each sub-initializer corresponds to
2352 registers 0 through 31, the second integer to registers 32 through 63, and
2355 @findex REGNO_REG_CLASS
2356 @item REGNO_REG_CLASS (@var{regno})
2357 A C expression whose value is a register class containing hard register
2358 @var{regno}. In general there is more than one such class; choose a class
2359 which is @dfn{minimal}, meaning that no smaller class also contains the
2362 @findex BASE_REG_CLASS
2363 @item BASE_REG_CLASS
2364 A macro whose definition is the name of the class to which a valid
2365 base register must belong. A base register is one used in an address
2366 which is the register value plus a displacement.
2368 @findex MODE_BASE_REG_CLASS
2369 @item MODE_BASE_REG_CLASS (@var{mode})
2370 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2371 the selection of a base register in a mode dependent manner. If
2372 @var{mode} is VOIDmode then it should return the same value as
2373 @code{BASE_REG_CLASS}.
2375 @findex INDEX_REG_CLASS
2376 @item INDEX_REG_CLASS
2377 A macro whose definition is the name of the class to which a valid
2378 index register must belong. An index register is one used in an
2379 address where its value is either multiplied by a scale factor or
2380 added to another register (as well as added to a displacement).
2382 @findex CONSTRAINT_LEN
2383 @item CONSTRAINT_LEN (@var{char}, @var{str})
2384 For the constraint at the start of @var{str}, which starts with the letter
2385 @var{c}, return the length. This allows you to have register class /
2386 constant / extra constraints that are longer than a single letter;
2387 you don't need to define this macro if you can do with single-letter
2388 constraints only. The definition of this macro should use
2389 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2390 to handle specially.
2391 There are some sanity checks in genoutput.c that check the constraint lengths
2392 for the md file, so you can also use this macro to help you while you are
2393 transitioning from a byzantine single-letter-constraint scheme: when you
2394 return a negative length for a constraint you want to re-use, genoutput
2395 will complain about every instance where it is used in the md file.
2397 @findex REG_CLASS_FROM_LETTER
2398 @item REG_CLASS_FROM_LETTER (@var{char})
2399 A C expression which defines the machine-dependent operand constraint
2400 letters for register classes. If @var{char} is such a letter, the
2401 value should be the register class corresponding to it. Otherwise,
2402 the value should be @code{NO_REGS}. The register letter @samp{r},
2403 corresponding to class @code{GENERAL_REGS}, will not be passed
2404 to this macro; you do not need to handle it.
2406 @findex REG_CLASS_FROM_CONSTRAINT
2407 @item REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2408 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2409 passed in @var{str}, so that you can use suffixes to distinguish between
2412 @findex REGNO_OK_FOR_BASE_P
2413 @item REGNO_OK_FOR_BASE_P (@var{num})
2414 A C expression which is nonzero if register number @var{num} is
2415 suitable for use as a base register in operand addresses. It may be
2416 either a suitable hard register or a pseudo register that has been
2417 allocated such a hard register.
2419 @findex REGNO_MODE_OK_FOR_BASE_P
2420 @item REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2421 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2422 that expression may examine the mode of the memory reference in
2423 @var{mode}. You should define this macro if the mode of the memory
2424 reference affects whether a register may be used as a base register. If
2425 you define this macro, the compiler will use it instead of
2426 @code{REGNO_OK_FOR_BASE_P}.
2428 @findex REGNO_OK_FOR_INDEX_P
2429 @item REGNO_OK_FOR_INDEX_P (@var{num})
2430 A C expression which is nonzero if register number @var{num} is
2431 suitable for use as an index register in operand addresses. It may be
2432 either a suitable hard register or a pseudo register that has been
2433 allocated such a hard register.
2435 The difference between an index register and a base register is that
2436 the index register may be scaled. If an address involves the sum of
2437 two registers, neither one of them scaled, then either one may be
2438 labeled the ``base'' and the other the ``index''; but whichever
2439 labeling is used must fit the machine's constraints of which registers
2440 may serve in each capacity. The compiler will try both labelings,
2441 looking for one that is valid, and will reload one or both registers
2442 only if neither labeling works.
2444 @findex PREFERRED_RELOAD_CLASS
2445 @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2446 A C expression that places additional restrictions on the register class
2447 to use when it is necessary to copy value @var{x} into a register in class
2448 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2449 another, smaller class. On many machines, the following definition is
2453 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2456 Sometimes returning a more restrictive class makes better code. For
2457 example, on the 68000, when @var{x} is an integer constant that is in range
2458 for a @samp{moveq} instruction, the value of this macro is always
2459 @code{DATA_REGS} as long as @var{class} includes the data registers.
2460 Requiring a data register guarantees that a @samp{moveq} will be used.
2462 If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
2463 you can force @var{x} into a memory constant. This is useful on
2464 certain machines where immediate floating values cannot be loaded into
2465 certain kinds of registers.
2467 @findex PREFERRED_OUTPUT_RELOAD_CLASS
2468 @item PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2469 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2470 input reloads. If you don't define this macro, the default is to use
2471 @var{class}, unchanged.
2473 @findex LIMIT_RELOAD_CLASS
2474 @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2475 A C expression that places additional restrictions on the register class
2476 to use when it is necessary to be able to hold a value of mode
2477 @var{mode} in a reload register for which class @var{class} would
2480 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2481 there are certain modes that simply can't go in certain reload classes.
2483 The value is a register class; perhaps @var{class}, or perhaps another,
2486 Don't define this macro unless the target machine has limitations which
2487 require the macro to do something nontrivial.
2489 @findex SECONDARY_RELOAD_CLASS
2490 @findex SECONDARY_INPUT_RELOAD_CLASS
2491 @findex SECONDARY_OUTPUT_RELOAD_CLASS
2492 @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2493 @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2494 @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2495 Many machines have some registers that cannot be copied directly to or
2496 from memory or even from other types of registers. An example is the
2497 @samp{MQ} register, which on most machines, can only be copied to or
2498 from general registers, but not memory. Some machines allow copying all
2499 registers to and from memory, but require a scratch register for stores
2500 to some memory locations (e.g., those with symbolic address on the RT,
2501 and those with certain symbolic address on the SPARC when compiling
2502 PIC)@. In some cases, both an intermediate and a scratch register are
2505 You should define these macros to indicate to the reload phase that it may
2506 need to allocate at least one register for a reload in addition to the
2507 register to contain the data. Specifically, if copying @var{x} to a
2508 register @var{class} in @var{mode} requires an intermediate register,
2509 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2510 largest register class all of whose registers can be used as
2511 intermediate registers or scratch registers.
2513 If copying a register @var{class} in @var{mode} to @var{x} requires an
2514 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2515 should be defined to return the largest register class required. If the
2516 requirements for input and output reloads are the same, the macro
2517 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2520 The values returned by these macros are often @code{GENERAL_REGS}.
2521 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2522 can be directly copied to or from a register of @var{class} in
2523 @var{mode} without requiring a scratch register. Do not define this
2524 macro if it would always return @code{NO_REGS}.
2526 If a scratch register is required (either with or without an
2527 intermediate register), you should define patterns for
2528 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2529 (@pxref{Standard Names}. These patterns, which will normally be
2530 implemented with a @code{define_expand}, should be similar to the
2531 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2534 Define constraints for the reload register and scratch register that
2535 contain a single register class. If the original reload register (whose
2536 class is @var{class}) can meet the constraint given in the pattern, the
2537 value returned by these macros is used for the class of the scratch
2538 register. Otherwise, two additional reload registers are required.
2539 Their classes are obtained from the constraints in the insn pattern.
2541 @var{x} might be a pseudo-register or a @code{subreg} of a
2542 pseudo-register, which could either be in a hard register or in memory.
2543 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2544 in memory and the hard register number if it is in a register.
2546 These macros should not be used in the case where a particular class of
2547 registers can only be copied to memory and not to another class of
2548 registers. In that case, secondary reload registers are not needed and
2549 would not be helpful. Instead, a stack location must be used to perform
2550 the copy and the @code{mov@var{m}} pattern should use memory as an
2551 intermediate storage. This case often occurs between floating-point and
2554 @findex SECONDARY_MEMORY_NEEDED
2555 @item SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2556 Certain machines have the property that some registers cannot be copied
2557 to some other registers without using memory. Define this macro on
2558 those machines to be a C expression that is nonzero if objects of mode
2559 @var{m} in registers of @var{class1} can only be copied to registers of
2560 class @var{class2} by storing a register of @var{class1} into memory
2561 and loading that memory location into a register of @var{class2}.
2563 Do not define this macro if its value would always be zero.
2565 @findex SECONDARY_MEMORY_NEEDED_RTX
2566 @item SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2567 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2568 allocates a stack slot for a memory location needed for register copies.
2569 If this macro is defined, the compiler instead uses the memory location
2570 defined by this macro.
2572 Do not define this macro if you do not define
2573 @code{SECONDARY_MEMORY_NEEDED}.
2575 @findex SECONDARY_MEMORY_NEEDED_MODE
2576 @item SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2577 When the compiler needs a secondary memory location to copy between two
2578 registers of mode @var{mode}, it normally allocates sufficient memory to
2579 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2580 load operations in a mode that many bits wide and whose class is the
2581 same as that of @var{mode}.
2583 This is right thing to do on most machines because it ensures that all
2584 bits of the register are copied and prevents accesses to the registers
2585 in a narrower mode, which some machines prohibit for floating-point
2588 However, this default behavior is not correct on some machines, such as
2589 the DEC Alpha, that store short integers in floating-point registers
2590 differently than in integer registers. On those machines, the default
2591 widening will not work correctly and you must define this macro to
2592 suppress that widening in some cases. See the file @file{alpha.h} for
2595 Do not define this macro if you do not define
2596 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2597 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2599 @findex SMALL_REGISTER_CLASSES
2600 @item SMALL_REGISTER_CLASSES
2601 On some machines, it is risky to let hard registers live across arbitrary
2602 insns. Typically, these machines have instructions that require values
2603 to be in specific registers (like an accumulator), and reload will fail
2604 if the required hard register is used for another purpose across such an
2607 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2608 value on these machines. When this macro has a nonzero value, the
2609 compiler will try to minimize the lifetime of hard registers.
2611 It is always safe to define this macro with a nonzero value, but if you
2612 unnecessarily define it, you will reduce the amount of optimizations
2613 that can be performed in some cases. If you do not define this macro
2614 with a nonzero value when it is required, the compiler will run out of
2615 spill registers and print a fatal error message. For most machines, you
2616 should not define this macro at all.
2618 @findex CLASS_LIKELY_SPILLED_P
2619 @item CLASS_LIKELY_SPILLED_P (@var{class})
2620 A C expression whose value is nonzero if pseudos that have been assigned
2621 to registers of class @var{class} would likely be spilled because
2622 registers of @var{class} are needed for spill registers.
2624 The default value of this macro returns 1 if @var{class} has exactly one
2625 register and zero otherwise. On most machines, this default should be
2626 used. Only define this macro to some other expression if pseudos
2627 allocated by @file{local-alloc.c} end up in memory because their hard
2628 registers were needed for spill registers. If this macro returns nonzero
2629 for those classes, those pseudos will only be allocated by
2630 @file{global.c}, which knows how to reallocate the pseudo to another
2631 register. If there would not be another register available for
2632 reallocation, you should not change the definition of this macro since
2633 the only effect of such a definition would be to slow down register
2636 @findex CLASS_MAX_NREGS
2637 @item CLASS_MAX_NREGS (@var{class}, @var{mode})
2638 A C expression for the maximum number of consecutive registers
2639 of class @var{class} needed to hold a value of mode @var{mode}.
2641 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2642 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2643 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2644 @var{mode})} for all @var{regno} values in the class @var{class}.
2646 This macro helps control the handling of multiple-word values
2649 @item CANNOT_CHANGE_MODE_CLASS(@var{from}, @var{to}, @var{class})
2650 If defined, a C expression that returns nonzero for a @var{class} for which
2651 a change from mode @var{from} to mode @var{to} is invalid.
2653 For the example, loading 32-bit integer or floating-point objects into
2654 floating-point registers on the Alpha extends them to 64 bits.
2655 Therefore loading a 64-bit object and then storing it as a 32-bit object
2656 does not store the low-order 32 bits, as would be the case for a normal
2657 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2661 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2662 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2663 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2667 Three other special macros describe which operands fit which constraint
2671 @findex CONST_OK_FOR_LETTER_P
2672 @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2673 A C expression that defines the machine-dependent operand constraint
2674 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2675 particular ranges of integer values. If @var{c} is one of those
2676 letters, the expression should check that @var{value}, an integer, is in
2677 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2678 not one of those letters, the value should be 0 regardless of
2681 @findex CONST_OK_FOR_CONSTRAINT_P
2682 @item CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2683 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2684 string passed in @var{str}, so that you can use suffixes to distinguish
2685 between different variants.
2687 @findex CONST_DOUBLE_OK_FOR_LETTER_P
2688 @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2689 A C expression that defines the machine-dependent operand constraint
2690 letters that specify particular ranges of @code{const_double} values
2691 (@samp{G} or @samp{H}).
2693 If @var{c} is one of those letters, the expression should check that
2694 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2695 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2696 letters, the value should be 0 regardless of @var{value}.
2698 @code{const_double} is used for all floating-point constants and for
2699 @code{DImode} fixed-point constants. A given letter can accept either
2700 or both kinds of values. It can use @code{GET_MODE} to distinguish
2701 between these kinds.
2703 @findex CONST_DOUBLE_OK_FOR_CONSTRAINT_P
2704 @item CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2705 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2706 string passed in @var{str}, so that you can use suffixes to distinguish
2707 between different variants.
2709 @findex EXTRA_CONSTRAINT
2710 @item EXTRA_CONSTRAINT (@var{value}, @var{c})
2711 A C expression that defines the optional machine-dependent constraint
2712 letters that can be used to segregate specific types of operands, usually
2713 memory references, for the target machine. Any letter that is not
2714 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2715 @code{REG_CLASS_FROM_CONSTRAINT}
2716 may be used. Normally this macro will not be defined.
2718 If it is required for a particular target machine, it should return 1
2719 if @var{value} corresponds to the operand type represented by the
2720 constraint letter @var{c}. If @var{c} is not defined as an extra
2721 constraint, the value returned should be 0 regardless of @var{value}.
2723 For example, on the ROMP, load instructions cannot have their output
2724 in r0 if the memory reference contains a symbolic address. Constraint
2725 letter @samp{Q} is defined as representing a memory address that does
2726 @emph{not} contain a symbolic address. An alternative is specified with
2727 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2728 alternative specifies @samp{m} on the input and a register class that
2729 does not include r0 on the output.
2731 @findex EXTRA_CONSTRAINT_STR
2732 @item EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2733 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2734 in @var{str}, so that you can use suffixes to distinguish between different
2737 @findex EXTRA_MEMORY_CONSTRAINT
2738 @item EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2739 A C expression that defines the optional machine-dependent constraint
2740 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2741 be treated like memory constraints by the reload pass.
2743 It should return 1 if the operand type represented by the constraint
2744 at the start of @var{str}, the first letter of which is the letter @var{c},
2745 comprises a subset of all memory references including
2746 all those whose address is simply a base register. This allows the reload
2747 pass to reload an operand, if it does not directly correspond to the operand
2748 type of @var{c}, by copying its address into a base register.
2750 For example, on the S/390, some instructions do not accept arbitrary
2751 memory references, but only those that do not make use of an index
2752 register. The constraint letter @samp{Q} is defined via
2753 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2754 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2755 a @samp{Q} constraint can handle any memory operand, because the
2756 reload pass knows it can be reloaded by copying the memory address
2757 into a base register if required. This is analogous to the way
2758 a @samp{o} constraint can handle any memory operand.
2760 @findex EXTRA_ADDRESS_CONSTRAINT
2761 @item EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2762 A C expression that defines the optional machine-dependent constraint
2763 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2764 @code{EXTRA_CONSTRAINT_STR}, that should
2765 be treated like address constraints by the reload pass.
2767 It should return 1 if the operand type represented by the constraint
2768 at the start of @var{str}, which starts with the letter @var{c}, comprises
2769 a subset of all memory addresses including
2770 all those that consist of just a base register. This allows the reload
2771 pass to reload an operand, if it does not directly correspond to the operand
2772 type of @var{str}, by copying it into a base register.
2774 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2775 be used with the @code{address_operand} predicate. It is treated
2776 analogously to the @samp{p} constraint.
2779 @node Stack and Calling
2780 @section Stack Layout and Calling Conventions
2781 @cindex calling conventions
2783 @c prevent bad page break with this line
2784 This describes the stack layout and calling conventions.
2788 * Exception Handling::
2793 * Register Arguments::
2795 * Aggregate Return::
2803 @subsection Basic Stack Layout
2804 @cindex stack frame layout
2805 @cindex frame layout
2807 @c prevent bad page break with this line
2808 Here is the basic stack layout.
2811 @findex STACK_GROWS_DOWNWARD
2812 @item STACK_GROWS_DOWNWARD
2813 Define this macro if pushing a word onto the stack moves the stack
2814 pointer to a smaller address.
2816 When we say, ``define this macro if @dots{},'' it means that the
2817 compiler checks this macro only with @code{#ifdef} so the precise
2818 definition used does not matter.
2820 @findex STACK_PUSH_CODE
2821 @item STACK_PUSH_CODE
2823 This macro defines the operation used when something is pushed
2824 on the stack. In RTL, a push operation will be
2825 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2827 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2828 and @code{POST_INC}. Which of these is correct depends on
2829 the stack direction and on whether the stack pointer points
2830 to the last item on the stack or whether it points to the
2831 space for the next item on the stack.
2833 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2834 defined, which is almost always right, and @code{PRE_INC} otherwise,
2835 which is often wrong.
2837 @findex FRAME_GROWS_DOWNWARD
2838 @item FRAME_GROWS_DOWNWARD
2839 Define this macro if the addresses of local variable slots are at negative
2840 offsets from the frame pointer.
2842 @findex ARGS_GROW_DOWNWARD
2843 @item ARGS_GROW_DOWNWARD
2844 Define this macro if successive arguments to a function occupy decreasing
2845 addresses on the stack.
2847 @findex STARTING_FRAME_OFFSET
2848 @item STARTING_FRAME_OFFSET
2849 Offset from the frame pointer to the first local variable slot to be allocated.
2851 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2852 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2853 Otherwise, it is found by adding the length of the first slot to the
2854 value @code{STARTING_FRAME_OFFSET}.
2855 @c i'm not sure if the above is still correct.. had to change it to get
2856 @c rid of an overfull. --mew 2feb93
2858 @findex STACK_ALIGNMENT_NEEDED
2859 @item STACK_ALIGNMENT_NEEDED
2860 Define to zero to disable final alignment of the stack during reload.
2861 The non-zero default for this macro is suitable for most ports.
2863 On ports where @code{STARTING_FRAME_OFFSET} is non-zero or where there
2864 is a register save block following the local block that doesn't require
2865 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2866 stack alignment and do it in the backend.
2868 @findex STACK_POINTER_OFFSET
2869 @item STACK_POINTER_OFFSET
2870 Offset from the stack pointer register to the first location at which
2871 outgoing arguments are placed. If not specified, the default value of
2872 zero is used. This is the proper value for most machines.
2874 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2875 the first location at which outgoing arguments are placed.
2877 @findex FIRST_PARM_OFFSET
2878 @item FIRST_PARM_OFFSET (@var{fundecl})
2879 Offset from the argument pointer register to the first argument's
2880 address. On some machines it may depend on the data type of the
2883 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2884 the first argument's address.
2886 @findex STACK_DYNAMIC_OFFSET
2887 @item STACK_DYNAMIC_OFFSET (@var{fundecl})
2888 Offset from the stack pointer register to an item dynamically allocated
2889 on the stack, e.g., by @code{alloca}.
2891 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2892 length of the outgoing arguments. The default is correct for most
2893 machines. See @file{function.c} for details.
2895 @findex DYNAMIC_CHAIN_ADDRESS
2896 @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2897 A C expression whose value is RTL representing the address in a stack
2898 frame where the pointer to the caller's frame is stored. Assume that
2899 @var{frameaddr} is an RTL expression for the address of the stack frame
2902 If you don't define this macro, the default is to return the value
2903 of @var{frameaddr}---that is, the stack frame address is also the
2904 address of the stack word that points to the previous frame.
2906 @findex SETUP_FRAME_ADDRESSES
2907 @item SETUP_FRAME_ADDRESSES
2908 If defined, a C expression that produces the machine-specific code to
2909 setup the stack so that arbitrary frames can be accessed. For example,
2910 on the SPARC, we must flush all of the register windows to the stack
2911 before we can access arbitrary stack frames. You will seldom need to
2914 @findex BUILTIN_SETJMP_FRAME_VALUE
2915 @item BUILTIN_SETJMP_FRAME_VALUE
2916 If defined, a C expression that contains an rtx that is used to store
2917 the address of the current frame into the built in @code{setjmp} buffer.
2918 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2919 machines. One reason you may need to define this macro is if
2920 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2922 @findex RETURN_ADDR_RTX
2923 @item RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2924 A C expression whose value is RTL representing the value of the return
2925 address for the frame @var{count} steps up from the current frame, after
2926 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2927 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2928 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2930 The value of the expression must always be the correct address when
2931 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2932 determine the return address of other frames.
2934 @findex RETURN_ADDR_IN_PREVIOUS_FRAME
2935 @item RETURN_ADDR_IN_PREVIOUS_FRAME
2936 Define this if the return address of a particular stack frame is accessed
2937 from the frame pointer of the previous stack frame.
2939 @findex INCOMING_RETURN_ADDR_RTX
2940 @item INCOMING_RETURN_ADDR_RTX
2941 A C expression whose value is RTL representing the location of the
2942 incoming return address at the beginning of any function, before the
2943 prologue. This RTL is either a @code{REG}, indicating that the return
2944 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2947 You only need to define this macro if you want to support call frame
2948 debugging information like that provided by DWARF 2.
2950 If this RTL is a @code{REG}, you should also define
2951 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2953 @findex INCOMING_FRAME_SP_OFFSET
2954 @item INCOMING_FRAME_SP_OFFSET
2955 A C expression whose value is an integer giving the offset, in bytes,
2956 from the value of the stack pointer register to the top of the stack
2957 frame at the beginning of any function, before the prologue. The top of
2958 the frame is defined to be the value of the stack pointer in the
2959 previous frame, just before the call instruction.
2961 You only need to define this macro if you want to support call frame
2962 debugging information like that provided by DWARF 2.
2964 @findex ARG_POINTER_CFA_OFFSET
2965 @item ARG_POINTER_CFA_OFFSET (@var{fundecl})
2966 A C expression whose value is an integer giving the offset, in bytes,
2967 from the argument pointer to the canonical frame address (cfa). The
2968 final value should coincide with that calculated by
2969 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2970 during virtual register instantiation.
2972 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
2973 which is correct for most machines; in general, the arguments are found
2974 immediately before the stack frame. Note that this is not the case on
2975 some targets that save registers into the caller's frame, such as SPARC
2976 and rs6000, and so such targets need to define this macro.
2978 You only need to define this macro if the default is incorrect, and you
2979 want to support call frame debugging information like that provided by
2984 Define this macro if the stack size for the target is very small. This
2985 has the effect of disabling gcc's built-in @samp{alloca}, though
2986 @samp{__builtin_alloca} is not affected.
2989 @node Exception Handling
2990 @subsection Exception Handling Support
2991 @cindex exception handling
2994 @findex EH_RETURN_DATA_REGNO
2995 @item EH_RETURN_DATA_REGNO (@var{N})
2996 A C expression whose value is the @var{N}th register number used for
2997 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2998 @var{N} registers are usable.
3000 The exception handling library routines communicate with the exception
3001 handlers via a set of agreed upon registers. Ideally these registers
3002 should be call-clobbered; it is possible to use call-saved registers,
3003 but may negatively impact code size. The target must support at least
3004 2 data registers, but should define 4 if there are enough free registers.
3006 You must define this macro if you want to support call frame exception
3007 handling like that provided by DWARF 2.
3009 @findex EH_RETURN_STACKADJ_RTX
3010 @item EH_RETURN_STACKADJ_RTX
3011 A C expression whose value is RTL representing a location in which
3012 to store a stack adjustment to be applied before function return.
3013 This is used to unwind the stack to an exception handler's call frame.
3014 It will be assigned zero on code paths that return normally.
3016 Typically this is a call-clobbered hard register that is otherwise
3017 untouched by the epilogue, but could also be a stack slot.
3019 You must define this macro if you want to support call frame exception
3020 handling like that provided by DWARF 2.
3022 @findex EH_RETURN_HANDLER_RTX
3023 @item EH_RETURN_HANDLER_RTX
3024 A C expression whose value is RTL representing a location in which
3025 to store the address of an exception handler to which we should
3026 return. It will not be assigned on code paths that return normally.
3028 Typically this is the location in the call frame at which the normal
3029 return address is stored. For targets that return by popping an
3030 address off the stack, this might be a memory address just below
3031 the @emph{target} call frame rather than inside the current call
3032 frame. @code{EH_RETURN_STACKADJ_RTX} will have already been assigned,
3033 so it may be used to calculate the location of the target call frame.
3035 Some targets have more complex requirements than storing to an
3036 address calculable during initial code generation. In that case
3037 the @code{eh_return} instruction pattern should be used instead.
3039 If you want to support call frame exception handling, you must
3040 define either this macro or the @code{eh_return} instruction pattern.
3042 @findex ASM_PREFERRED_EH_DATA_FORMAT
3043 @item ASM_PREFERRED_EH_DATA_FORMAT(@var{code}, @var{global})
3044 This macro chooses the encoding of pointers embedded in the exception
3045 handling sections. If at all possible, this should be defined such
3046 that the exception handling section will not require dynamic relocations,
3047 and so may be read-only.
3049 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3050 @var{global} is true if the symbol may be affected by dynamic relocations.
3051 The macro should return a combination of the @code{DW_EH_PE_*} defines
3052 as found in @file{dwarf2.h}.
3054 If this macro is not defined, pointers will not be encoded but
3055 represented directly.
3057 @findex ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX
3058 @item ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX(@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3059 This macro allows the target to emit whatever special magic is required
3060 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3061 Generic code takes care of pc-relative and indirect encodings; this must
3062 be defined if the target uses text-relative or data-relative encodings.
3064 This is a C statement that branches to @var{done} if the format was
3065 handled. @var{encoding} is the format chosen, @var{size} is the number
3066 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3069 @findex MD_FALLBACK_FRAME_STATE_FOR
3070 @item MD_FALLBACK_FRAME_STATE_FOR(@var{context}, @var{fs}, @var{success})
3071 This macro allows the target to add cpu and operating system specific
3072 code to the call-frame unwinder for use when there is no unwind data
3073 available. The most common reason to implement this macro is to unwind
3074 through signal frames.
3076 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
3077 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3078 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3079 for the address of the code being executed and @code{context->cfa} for
3080 the stack pointer value. If the frame can be decoded, the register save
3081 addresses should be updated in @var{fs} and the macro should branch to
3082 @var{success}. If the frame cannot be decoded, the macro should do
3086 @node Stack Checking
3087 @subsection Specifying How Stack Checking is Done
3089 GCC will check that stack references are within the boundaries of
3090 the stack, if the @option{-fstack-check} is specified, in one of three ways:
3094 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3095 will assume that you have arranged for stack checking to be done at
3096 appropriate places in the configuration files, e.g., in
3097 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
3101 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
3102 called @code{check_stack} in your @file{md} file, GCC will call that
3103 pattern with one argument which is the address to compare the stack
3104 value against. You must arrange for this pattern to report an error if
3105 the stack pointer is out of range.
3108 If neither of the above are true, GCC will generate code to periodically
3109 ``probe'' the stack pointer using the values of the macros defined below.
3112 Normally, you will use the default values of these macros, so GCC
3113 will use the third approach.
3116 @findex STACK_CHECK_BUILTIN
3117 @item STACK_CHECK_BUILTIN
3118 A nonzero value if stack checking is done by the configuration files in a
3119 machine-dependent manner. You should define this macro if stack checking
3120 is require by the ABI of your machine or if you would like to have to stack
3121 checking in some more efficient way than GCC's portable approach.
3122 The default value of this macro is zero.
3124 @findex STACK_CHECK_PROBE_INTERVAL
3125 @item STACK_CHECK_PROBE_INTERVAL
3126 An integer representing the interval at which GCC must generate stack
3127 probe instructions. You will normally define this macro to be no larger
3128 than the size of the ``guard pages'' at the end of a stack area. The
3129 default value of 4096 is suitable for most systems.
3131 @findex STACK_CHECK_PROBE_LOAD
3132 @item STACK_CHECK_PROBE_LOAD
3133 A integer which is nonzero if GCC should perform the stack probe
3134 as a load instruction and zero if GCC should use a store instruction.
3135 The default is zero, which is the most efficient choice on most systems.
3137 @findex STACK_CHECK_PROTECT
3138 @item STACK_CHECK_PROTECT
3139 The number of bytes of stack needed to recover from a stack overflow,
3140 for languages where such a recovery is supported. The default value of
3141 75 words should be adequate for most machines.
3143 @findex STACK_CHECK_MAX_FRAME_SIZE
3144 @item STACK_CHECK_MAX_FRAME_SIZE
3145 The maximum size of a stack frame, in bytes. GCC will generate probe
3146 instructions in non-leaf functions to ensure at least this many bytes of
3147 stack are available. If a stack frame is larger than this size, stack
3148 checking will not be reliable and GCC will issue a warning. The
3149 default is chosen so that GCC only generates one instruction on most
3150 systems. You should normally not change the default value of this macro.
3152 @findex STACK_CHECK_FIXED_FRAME_SIZE
3153 @item STACK_CHECK_FIXED_FRAME_SIZE
3154 GCC uses this value to generate the above warning message. It
3155 represents the amount of fixed frame used by a function, not including
3156 space for any callee-saved registers, temporaries and user variables.
3157 You need only specify an upper bound for this amount and will normally
3158 use the default of four words.
3160 @findex STACK_CHECK_MAX_VAR_SIZE
3161 @item STACK_CHECK_MAX_VAR_SIZE
3162 The maximum size, in bytes, of an object that GCC will place in the
3163 fixed area of the stack frame when the user specifies
3164 @option{-fstack-check}.
3165 GCC computed the default from the values of the above macros and you will
3166 normally not need to override that default.
3170 @node Frame Registers
3171 @subsection Registers That Address the Stack Frame
3173 @c prevent bad page break with this line
3174 This discusses registers that address the stack frame.
3177 @findex STACK_POINTER_REGNUM
3178 @item STACK_POINTER_REGNUM
3179 The register number of the stack pointer register, which must also be a
3180 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3181 the hardware determines which register this is.
3183 @findex FRAME_POINTER_REGNUM
3184 @item FRAME_POINTER_REGNUM
3185 The register number of the frame pointer register, which is used to
3186 access automatic variables in the stack frame. On some machines, the
3187 hardware determines which register this is. On other machines, you can
3188 choose any register you wish for this purpose.
3190 @findex HARD_FRAME_POINTER_REGNUM
3191 @item HARD_FRAME_POINTER_REGNUM
3192 On some machines the offset between the frame pointer and starting
3193 offset of the automatic variables is not known until after register
3194 allocation has been done (for example, because the saved registers are
3195 between these two locations). On those machines, define
3196 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3197 be used internally until the offset is known, and define
3198 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3199 used for the frame pointer.
3201 You should define this macro only in the very rare circumstances when it
3202 is not possible to calculate the offset between the frame pointer and
3203 the automatic variables until after register allocation has been
3204 completed. When this macro is defined, you must also indicate in your
3205 definition of @code{ELIMINABLE_REGS} how to eliminate
3206 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3207 or @code{STACK_POINTER_REGNUM}.
3209 Do not define this macro if it would be the same as
3210 @code{FRAME_POINTER_REGNUM}.
3212 @findex ARG_POINTER_REGNUM
3213 @item ARG_POINTER_REGNUM
3214 The register number of the arg pointer register, which is used to access
3215 the function's argument list. On some machines, this is the same as the
3216 frame pointer register. On some machines, the hardware determines which
3217 register this is. On other machines, you can choose any register you
3218 wish for this purpose. If this is not the same register as the frame
3219 pointer register, then you must mark it as a fixed register according to
3220 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3221 (@pxref{Elimination}).
3223 @findex RETURN_ADDRESS_POINTER_REGNUM
3224 @item RETURN_ADDRESS_POINTER_REGNUM
3225 The register number of the return address pointer register, which is used to
3226 access the current function's return address from the stack. On some
3227 machines, the return address is not at a fixed offset from the frame
3228 pointer or stack pointer or argument pointer. This register can be defined
3229 to point to the return address on the stack, and then be converted by
3230 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3232 Do not define this macro unless there is no other way to get the return
3233 address from the stack.
3235 @findex STATIC_CHAIN_REGNUM
3236 @findex STATIC_CHAIN_INCOMING_REGNUM
3237 @item STATIC_CHAIN_REGNUM
3238 @itemx STATIC_CHAIN_INCOMING_REGNUM
3239 Register numbers used for passing a function's static chain pointer. If
3240 register windows are used, the register number as seen by the called
3241 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3242 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3243 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3246 The static chain register need not be a fixed register.
3248 If the static chain is passed in memory, these macros should not be
3249 defined; instead, the next two macros should be defined.
3251 @findex STATIC_CHAIN
3252 @findex STATIC_CHAIN_INCOMING
3254 @itemx STATIC_CHAIN_INCOMING
3255 If the static chain is passed in memory, these macros provide rtx giving
3256 @code{mem} expressions that denote where they are stored.
3257 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3258 as seen by the calling and called functions, respectively. Often the former
3259 will be at an offset from the stack pointer and the latter at an offset from
3262 @findex stack_pointer_rtx
3263 @findex frame_pointer_rtx
3264 @findex arg_pointer_rtx
3265 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3266 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3267 macros and should be used to refer to those items.
3269 If the static chain is passed in a register, the two previous macros should
3272 @findex DWARF_FRAME_REGISTERS
3273 @item DWARF_FRAME_REGISTERS
3274 This macro specifies the maximum number of hard registers that can be
3275 saved in a call frame. This is used to size data structures used in
3276 DWARF2 exception handling.
3278 Prior to GCC 3.0, this macro was needed in order to establish a stable
3279 exception handling ABI in the face of adding new hard registers for ISA
3280 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3281 in the number of hard registers. Nevertheless, this macro can still be
3282 used to reduce the runtime memory requirements of the exception handling
3283 routines, which can be substantial if the ISA contains a lot of
3284 registers that are not call-saved.
3286 If this macro is not defined, it defaults to
3287 @code{FIRST_PSEUDO_REGISTER}.
3289 @findex PRE_GCC3_DWARF_FRAME_REGISTERS
3290 @item PRE_GCC3_DWARF_FRAME_REGISTERS
3292 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3293 for backward compatibility in pre GCC 3.0 compiled code.
3295 If this macro is not defined, it defaults to
3296 @code{DWARF_FRAME_REGISTERS}.
3298 @findex DWARF_REG_TO_UNWIND_COLUMN
3299 @item DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3301 Define this macro if the target's representation for dwarf registers
3302 is different than the internal representation for unwind column.
3303 Given a dwarf register, this macro should return the interal unwind
3304 column number to use instead.
3306 See the PowerPC's SPE target for an example.
3311 @subsection Eliminating Frame Pointer and Arg Pointer
3313 @c prevent bad page break with this line
3314 This is about eliminating the frame pointer and arg pointer.
3317 @findex FRAME_POINTER_REQUIRED
3318 @item FRAME_POINTER_REQUIRED
3319 A C expression which is nonzero if a function must have and use a frame
3320 pointer. This expression is evaluated in the reload pass. If its value is
3321 nonzero the function will have a frame pointer.
3323 The expression can in principle examine the current function and decide
3324 according to the facts, but on most machines the constant 0 or the
3325 constant 1 suffices. Use 0 when the machine allows code to be generated
3326 with no frame pointer, and doing so saves some time or space. Use 1
3327 when there is no possible advantage to avoiding a frame pointer.
3329 In certain cases, the compiler does not know how to produce valid code
3330 without a frame pointer. The compiler recognizes those cases and
3331 automatically gives the function a frame pointer regardless of what
3332 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3335 In a function that does not require a frame pointer, the frame pointer
3336 register can be allocated for ordinary usage, unless you mark it as a
3337 fixed register. See @code{FIXED_REGISTERS} for more information.
3339 @findex INITIAL_FRAME_POINTER_OFFSET
3340 @findex get_frame_size
3341 @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3342 A C statement to store in the variable @var{depth-var} the difference
3343 between the frame pointer and the stack pointer values immediately after
3344 the function prologue. The value would be computed from information
3345 such as the result of @code{get_frame_size ()} and the tables of
3346 registers @code{regs_ever_live} and @code{call_used_regs}.
3348 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3349 need not be defined. Otherwise, it must be defined even if
3350 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3351 case, you may set @var{depth-var} to anything.
3353 @findex ELIMINABLE_REGS
3354 @item ELIMINABLE_REGS
3355 If defined, this macro specifies a table of register pairs used to
3356 eliminate unneeded registers that point into the stack frame. If it is not
3357 defined, the only elimination attempted by the compiler is to replace
3358 references to the frame pointer with references to the stack pointer.
3360 The definition of this macro is a list of structure initializations, each
3361 of which specifies an original and replacement register.
3363 On some machines, the position of the argument pointer is not known until
3364 the compilation is completed. In such a case, a separate hard register
3365 must be used for the argument pointer. This register can be eliminated by
3366 replacing it with either the frame pointer or the argument pointer,
3367 depending on whether or not the frame pointer has been eliminated.
3369 In this case, you might specify:
3371 #define ELIMINABLE_REGS \
3372 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3373 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3374 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3377 Note that the elimination of the argument pointer with the stack pointer is
3378 specified first since that is the preferred elimination.
3380 @findex CAN_ELIMINATE
3381 @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3382 A C expression that returns nonzero if the compiler is allowed to try
3383 to replace register number @var{from-reg} with register number
3384 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3385 is defined, and will usually be the constant 1, since most of the cases
3386 preventing register elimination are things that the compiler already
3389 @findex INITIAL_ELIMINATION_OFFSET
3390 @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3391 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3392 specifies the initial difference between the specified pair of
3393 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3397 @node Stack Arguments
3398 @subsection Passing Function Arguments on the Stack
3399 @cindex arguments on stack
3400 @cindex stack arguments
3402 The macros in this section control how arguments are passed
3403 on the stack. See the following section for other macros that
3404 control passing certain arguments in registers.
3407 @findex PROMOTE_PROTOTYPES
3408 @item PROMOTE_PROTOTYPES
3409 A C expression whose value is nonzero if an argument declared in
3410 a prototype as an integral type smaller than @code{int} should
3411 actually be passed as an @code{int}. In addition to avoiding
3412 errors in certain cases of mismatch, it also makes for better
3413 code on certain machines. If the macro is not defined in target
3414 header files, it defaults to 0.
3418 A C expression. If nonzero, push insns will be used to pass
3420 If the target machine does not have a push instruction, set it to zero.
3421 That directs GCC to use an alternate strategy: to
3422 allocate the entire argument block and then store the arguments into
3423 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3425 @findex PUSH_ROUNDING
3426 @item PUSH_ROUNDING (@var{npushed})
3427 A C expression that is the number of bytes actually pushed onto the
3428 stack when an instruction attempts to push @var{npushed} bytes.
3430 On some machines, the definition
3433 #define PUSH_ROUNDING(BYTES) (BYTES)
3437 will suffice. But on other machines, instructions that appear
3438 to push one byte actually push two bytes in an attempt to maintain
3439 alignment. Then the definition should be
3442 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3445 @findex ACCUMULATE_OUTGOING_ARGS
3446 @findex current_function_outgoing_args_size
3447 @item ACCUMULATE_OUTGOING_ARGS
3448 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3449 will be computed and placed into the variable
3450 @code{current_function_outgoing_args_size}. No space will be pushed
3451 onto the stack for each call; instead, the function prologue should
3452 increase the stack frame size by this amount.
3454 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3457 @findex REG_PARM_STACK_SPACE
3458 @item REG_PARM_STACK_SPACE (@var{fndecl})
3459 Define this macro if functions should assume that stack space has been
3460 allocated for arguments even when their values are passed in
3463 The value of this macro is the size, in bytes, of the area reserved for
3464 arguments passed in registers for the function represented by @var{fndecl},
3465 which can be zero if GCC is calling a library function.
3467 This space can be allocated by the caller, or be a part of the
3468 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3470 @c above is overfull. not sure what to do. --mew 5feb93 did
3471 @c something, not sure if it looks good. --mew 10feb93
3473 @findex MAYBE_REG_PARM_STACK_SPACE
3474 @findex FINAL_REG_PARM_STACK_SPACE
3475 @item MAYBE_REG_PARM_STACK_SPACE
3476 @itemx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
3477 Define these macros in addition to the one above if functions might
3478 allocate stack space for arguments even when their values are passed
3479 in registers. These should be used when the stack space allocated
3480 for arguments in registers is not a simple constant independent of the
3481 function declaration.
3483 The value of the first macro is the size, in bytes, of the area that
3484 we should initially assume would be reserved for arguments passed in registers.
3486 The value of the second macro is the actual size, in bytes, of the area
3487 that will be reserved for arguments passed in registers. This takes two
3488 arguments: an integer representing the number of bytes of fixed sized
3489 arguments on the stack, and a tree representing the number of bytes of
3490 variable sized arguments on the stack.
3492 When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
3493 called for libcall functions, the current function, or for a function
3494 being called when it is known that such stack space must be allocated.
3495 In each case this value can be easily computed.
3497 When deciding whether a called function needs such stack space, and how
3498 much space to reserve, GCC uses these two macros instead of
3499 @code{REG_PARM_STACK_SPACE}.
3501 @findex OUTGOING_REG_PARM_STACK_SPACE
3502 @item OUTGOING_REG_PARM_STACK_SPACE
3503 Define this if it is the responsibility of the caller to allocate the area
3504 reserved for arguments passed in registers.
3506 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3507 whether the space for these arguments counts in the value of
3508 @code{current_function_outgoing_args_size}.
3510 @findex STACK_PARMS_IN_REG_PARM_AREA
3511 @item STACK_PARMS_IN_REG_PARM_AREA
3512 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3513 stack parameters don't skip the area specified by it.
3514 @c i changed this, makes more sens and it should have taken care of the
3515 @c overfull.. not as specific, tho. --mew 5feb93
3517 Normally, when a parameter is not passed in registers, it is placed on the
3518 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3519 suppresses this behavior and causes the parameter to be passed on the
3520 stack in its natural location.
3522 @findex RETURN_POPS_ARGS
3523 @item RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3524 A C expression that should indicate the number of bytes of its own
3525 arguments that a function pops on returning, or 0 if the
3526 function pops no arguments and the caller must therefore pop them all
3527 after the function returns.
3529 @var{fundecl} is a C variable whose value is a tree node that describes
3530 the function in question. Normally it is a node of type
3531 @code{FUNCTION_DECL} that describes the declaration of the function.
3532 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3534 @var{funtype} is a C variable whose value is a tree node that
3535 describes the function in question. Normally it is a node of type
3536 @code{FUNCTION_TYPE} that describes the data type of the function.
3537 From this it is possible to obtain the data types of the value and
3538 arguments (if known).
3540 When a call to a library function is being considered, @var{fundecl}
3541 will contain an identifier node for the library function. Thus, if
3542 you need to distinguish among various library functions, you can do so
3543 by their names. Note that ``library function'' in this context means
3544 a function used to perform arithmetic, whose name is known specially
3545 in the compiler and was not mentioned in the C code being compiled.
3547 @var{stack-size} is the number of bytes of arguments passed on the
3548 stack. If a variable number of bytes is passed, it is zero, and
3549 argument popping will always be the responsibility of the calling function.
3551 On the VAX, all functions always pop their arguments, so the definition
3552 of this macro is @var{stack-size}. On the 68000, using the standard
3553 calling convention, no functions pop their arguments, so the value of
3554 the macro is always 0 in this case. But an alternative calling
3555 convention is available in which functions that take a fixed number of
3556 arguments pop them but other functions (such as @code{printf}) pop
3557 nothing (the caller pops all). When this convention is in use,
3558 @var{funtype} is examined to determine whether a function takes a fixed
3559 number of arguments.
3561 @findex CALL_POPS_ARGS
3562 @item CALL_POPS_ARGS (@var{cum})
3563 A C expression that should indicate the number of bytes a call sequence
3564 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3565 when compiling a function call.
3567 @var{cum} is the variable in which all arguments to the called function
3568 have been accumulated.
3570 On certain architectures, such as the SH5, a call trampoline is used
3571 that pops certain registers off the stack, depending on the arguments
3572 that have been passed to the function. Since this is a property of the
3573 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3578 @node Register Arguments
3579 @subsection Passing Arguments in Registers
3580 @cindex arguments in registers
3581 @cindex registers arguments
3583 This section describes the macros which let you control how various
3584 types of arguments are passed in registers or how they are arranged in
3588 @findex FUNCTION_ARG
3589 @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3590 A C expression that controls whether a function argument is passed
3591 in a register, and which register.
3593 The arguments are @var{cum}, which summarizes all the previous
3594 arguments; @var{mode}, the machine mode of the argument; @var{type},
3595 the data type of the argument as a tree node or 0 if that is not known
3596 (which happens for C support library functions); and @var{named},
3597 which is 1 for an ordinary argument and 0 for nameless arguments that
3598 correspond to @samp{@dots{}} in the called function's prototype.
3599 @var{type} can be an incomplete type if a syntax error has previously
3602 The value of the expression is usually either a @code{reg} RTX for the
3603 hard register in which to pass the argument, or zero to pass the
3604 argument on the stack.
3606 For machines like the VAX and 68000, where normally all arguments are
3607 pushed, zero suffices as a definition.
3609 The value of the expression can also be a @code{parallel} RTX@. This is
3610 used when an argument is passed in multiple locations. The mode of the
3611 of the @code{parallel} should be the mode of the entire argument. The
3612 @code{parallel} holds any number of @code{expr_list} pairs; each one
3613 describes where part of the argument is passed. In each
3614 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3615 register in which to pass this part of the argument, and the mode of the
3616 register RTX indicates how large this part of the argument is. The
3617 second operand of the @code{expr_list} is a @code{const_int} which gives
3618 the offset in bytes into the entire argument of where this part starts.
3619 As a special exception the first @code{expr_list} in the @code{parallel}
3620 RTX may have a first operand of zero. This indicates that the entire
3621 argument is also stored on the stack.
3623 The last time this macro is called, it is called with @code{MODE ==
3624 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3625 pattern as operands 2 and 3 respectively.
3627 @cindex @file{stdarg.h} and register arguments
3628 The usual way to make the ISO library @file{stdarg.h} work on a machine
3629 where some arguments are usually passed in registers, is to cause
3630 nameless arguments to be passed on the stack instead. This is done
3631 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3633 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3634 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3635 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
3636 in the definition of this macro to determine if this argument is of a
3637 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3638 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3639 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3640 defined, the argument will be computed in the stack and then loaded into
3643 @findex MUST_PASS_IN_STACK
3644 @item MUST_PASS_IN_STACK (@var{mode}, @var{type})
3645 Define as a C expression that evaluates to nonzero if we do not know how
3646 to pass TYPE solely in registers. The file @file{expr.h} defines a
3647 definition that is usually appropriate, refer to @file{expr.h} for additional
3650 @findex FUNCTION_INCOMING_ARG
3651 @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3652 Define this macro if the target machine has ``register windows'', so
3653 that the register in which a function sees an arguments is not
3654 necessarily the same as the one in which the caller passed the
3657 For such machines, @code{FUNCTION_ARG} computes the register in which
3658 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3659 be defined in a similar fashion to tell the function being called
3660 where the arguments will arrive.
3662 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3663 serves both purposes.
3665 @findex FUNCTION_ARG_PARTIAL_NREGS
3666 @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
3667 A C expression for the number of words, at the beginning of an
3668 argument, that must be put in registers. The value must be zero for
3669 arguments that are passed entirely in registers or that are entirely
3670 pushed on the stack.
3672 On some machines, certain arguments must be passed partially in
3673 registers and partially in memory. On these machines, typically the
3674 first @var{n} words of arguments are passed in registers, and the rest
3675 on the stack. If a multi-word argument (a @code{double} or a
3676 structure) crosses that boundary, its first few words must be passed
3677 in registers and the rest must be pushed. This macro tells the
3678 compiler when this occurs, and how many of the words should go in
3681 @code{FUNCTION_ARG} for these arguments should return the first
3682 register to be used by the caller for this argument; likewise
3683 @code{FUNCTION_INCOMING_ARG}, for the called function.
3685 @findex FUNCTION_ARG_PASS_BY_REFERENCE
3686 @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3687 A C expression that indicates when an argument must be passed by reference.
3688 If nonzero for an argument, a copy of that argument is made in memory and a
3689 pointer to the argument is passed instead of the argument itself.
3690 The pointer is passed in whatever way is appropriate for passing a pointer
3693 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
3694 definition of this macro might be
3696 #define FUNCTION_ARG_PASS_BY_REFERENCE\
3697 (CUM, MODE, TYPE, NAMED) \
3698 MUST_PASS_IN_STACK (MODE, TYPE)
3700 @c this is *still* too long. --mew 5feb93
3702 @findex FUNCTION_ARG_CALLEE_COPIES
3703 @item FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
3704 If defined, a C expression that indicates when it is the called function's
3705 responsibility to make a copy of arguments passed by invisible reference.
3706 Normally, the caller makes a copy and passes the address of the copy to the
3707 routine being called. When @code{FUNCTION_ARG_CALLEE_COPIES} is defined and is
3708 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
3709 ``live'' value. The called function must not modify this value. If it can be
3710 determined that the value won't be modified, it need not make a copy;
3711 otherwise a copy must be made.
3713 @findex CUMULATIVE_ARGS
3714 @item CUMULATIVE_ARGS
3715 A C type for declaring a variable that is used as the first argument of
3716 @code{FUNCTION_ARG} and other related values. For some target machines,
3717 the type @code{int} suffices and can hold the number of bytes of
3720 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3721 arguments that have been passed on the stack. The compiler has other
3722 variables to keep track of that. For target machines on which all
3723 arguments are passed on the stack, there is no need to store anything in
3724 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3725 should not be empty, so use @code{int}.
3727 @findex INIT_CUMULATIVE_ARGS
3728 @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname},
3729 @var{fndecl}) A C statement (sans semicolon) for initializing the variable
3730 @var{cum} for the state at the beginning of the argument list. The variable
3731 has type @code{CUMULATIVE_ARGS}. The value of @var{fntype} is the tree node
3732 for the data type of the function which will receive the args, or 0 if the args
3733 are to a compiler support library function. For direct calls that are not
3734 libcalls, @var{fndecl} contain the declaration node of the function.
3735 @var{fndecl} is also set when @code{INIT_CUMULATIVE_ARGS} is used to find
3736 arguments for the function being compiled.
3738 When processing a call to a compiler support library function,
3739 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3740 contains the name of the function, as a string. @var{libname} is 0 when
3741 an ordinary C function call is being processed. Thus, each time this
3742 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3743 never both of them at once.
3745 @findex INIT_CUMULATIVE_LIBCALL_ARGS
3746 @item INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3747 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3748 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3749 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3750 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3751 0)} is used instead.
3753 @findex INIT_CUMULATIVE_INCOMING_ARGS
3754 @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3755 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3756 finding the arguments for the function being compiled. If this macro is
3757 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3759 The value passed for @var{libname} is always 0, since library routines
3760 with special calling conventions are never compiled with GCC@. The
3761 argument @var{libname} exists for symmetry with
3762 @code{INIT_CUMULATIVE_ARGS}.
3763 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3764 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3766 @findex FUNCTION_ARG_ADVANCE
3767 @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3768 A C statement (sans semicolon) to update the summarizer variable
3769 @var{cum} to advance past an argument in the argument list. The
3770 values @var{mode}, @var{type} and @var{named} describe that argument.
3771 Once this is done, the variable @var{cum} is suitable for analyzing
3772 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
3774 This macro need not do anything if the argument in question was passed
3775 on the stack. The compiler knows how to track the amount of stack space
3776 used for arguments without any special help.
3778 @findex FUNCTION_ARG_PADDING
3779 @item FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3780 If defined, a C expression which determines whether, and in which direction,
3781 to pad out an argument with extra space. The value should be of type
3782 @code{enum direction}: either @code{upward} to pad above the argument,
3783 @code{downward} to pad below, or @code{none} to inhibit padding.
3785 The @emph{amount} of padding is always just enough to reach the next
3786 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3789 This macro has a default definition which is right for most systems.
3790 For little-endian machines, the default is to pad upward. For
3791 big-endian machines, the default is to pad downward for an argument of
3792 constant size shorter than an @code{int}, and upward otherwise.
3794 @findex PAD_VARARGS_DOWN
3795 @item PAD_VARARGS_DOWN
3796 If defined, a C expression which determines whether the default
3797 implementation of va_arg will attempt to pad down before reading the
3798 next argument, if that argument is smaller than its aligned space as
3799 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3800 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3802 @findex FUNCTION_ARG_BOUNDARY
3803 @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3804 If defined, a C expression that gives the alignment boundary, in bits,
3805 of an argument with the specified mode and type. If it is not defined,
3806 @code{PARM_BOUNDARY} is used for all arguments.
3808 @findex FUNCTION_ARG_REGNO_P
3809 @item FUNCTION_ARG_REGNO_P (@var{regno})
3810 A C expression that is nonzero if @var{regno} is the number of a hard
3811 register in which function arguments are sometimes passed. This does
3812 @emph{not} include implicit arguments such as the static chain and
3813 the structure-value address. On many machines, no registers can be
3814 used for this purpose since all function arguments are pushed on the
3817 @findex LOAD_ARGS_REVERSED
3818 @item LOAD_ARGS_REVERSED
3819 If defined, the order in which arguments are loaded into their
3820 respective argument registers is reversed so that the last
3821 argument is loaded first. This macro only affects arguments
3822 passed in registers.
3827 @subsection How Scalar Function Values Are Returned
3828 @cindex return values in registers
3829 @cindex values, returned by functions
3830 @cindex scalars, returned as values
3832 This section discusses the macros that control returning scalars as
3833 values---values that can fit in registers.
3836 @findex FUNCTION_VALUE
3837 @item FUNCTION_VALUE (@var{valtype}, @var{func})
3838 A C expression to create an RTX representing the place where a
3839 function returns a value of data type @var{valtype}. @var{valtype} is
3840 a tree node representing a data type. Write @code{TYPE_MODE
3841 (@var{valtype})} to get the machine mode used to represent that type.
3842 On many machines, only the mode is relevant. (Actually, on most
3843 machines, scalar values are returned in the same place regardless of
3846 The value of the expression is usually a @code{reg} RTX for the hard
3847 register where the return value is stored. The value can also be a
3848 @code{parallel} RTX, if the return value is in multiple places. See
3849 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3851 If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
3852 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3855 If the precise function being called is known, @var{func} is a tree
3856 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3857 pointer. This makes it possible to use a different value-returning
3858 convention for specific functions when all their calls are
3861 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3862 types, because these are returned in another way. See
3863 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3865 @findex FUNCTION_OUTGOING_VALUE
3866 @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3867 Define this macro if the target machine has ``register windows''
3868 so that the register in which a function returns its value is not
3869 the same as the one in which the caller sees the value.
3871 For such machines, @code{FUNCTION_VALUE} computes the register in which
3872 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3873 defined in a similar fashion to tell the function where to put the
3876 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3877 @code{FUNCTION_VALUE} serves both purposes.
3879 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3880 aggregate data types, because these are returned in another way. See
3881 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3883 @findex LIBCALL_VALUE
3884 @item LIBCALL_VALUE (@var{mode})
3885 A C expression to create an RTX representing the place where a library
3886 function returns a value of mode @var{mode}. If the precise function
3887 being called is known, @var{func} is a tree node
3888 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3889 pointer. This makes it possible to use a different value-returning
3890 convention for specific functions when all their calls are
3893 Note that ``library function'' in this context means a compiler
3894 support routine, used to perform arithmetic, whose name is known
3895 specially by the compiler and was not mentioned in the C code being
3898 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3899 data types, because none of the library functions returns such types.
3901 @findex FUNCTION_VALUE_REGNO_P
3902 @item FUNCTION_VALUE_REGNO_P (@var{regno})
3903 A C expression that is nonzero if @var{regno} is the number of a hard
3904 register in which the values of called function may come back.
3906 A register whose use for returning values is limited to serving as the
3907 second of a pair (for a value of type @code{double}, say) need not be
3908 recognized by this macro. So for most machines, this definition
3912 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3915 If the machine has register windows, so that the caller and the called
3916 function use different registers for the return value, this macro
3917 should recognize only the caller's register numbers.
3919 @findex APPLY_RESULT_SIZE
3920 @item APPLY_RESULT_SIZE
3921 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3922 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3923 saving and restoring an arbitrary return value.
3926 @node Aggregate Return
3927 @subsection How Large Values Are Returned
3928 @cindex aggregates as return values
3929 @cindex large return values
3930 @cindex returning aggregate values
3931 @cindex structure value address
3933 When a function value's mode is @code{BLKmode} (and in some other
3934 cases), the value is not returned according to @code{FUNCTION_VALUE}
3935 (@pxref{Scalar Return}). Instead, the caller passes the address of a
3936 block of memory in which the value should be stored. This address
3937 is called the @dfn{structure value address}.
3939 This section describes how to control returning structure values in
3943 @findex RETURN_IN_MEMORY
3944 @item RETURN_IN_MEMORY (@var{type})
3945 A C expression which can inhibit the returning of certain function
3946 values in registers, based on the type of value. A nonzero value says
3947 to return the function value in memory, just as large structures are
3948 always returned. Here @var{type} will be a C expression of type
3949 @code{tree}, representing the data type of the value.
3951 Note that values of mode @code{BLKmode} must be explicitly handled
3952 by this macro. Also, the option @option{-fpcc-struct-return}
3953 takes effect regardless of this macro. On most systems, it is
3954 possible to leave the macro undefined; this causes a default
3955 definition to be used, whose value is the constant 1 for @code{BLKmode}
3956 values, and 0 otherwise.
3958 Do not use this macro to indicate that structures and unions should always
3959 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
3962 @findex DEFAULT_PCC_STRUCT_RETURN
3963 @item DEFAULT_PCC_STRUCT_RETURN
3964 Define this macro to be 1 if all structure and union return values must be
3965 in memory. Since this results in slower code, this should be defined
3966 only if needed for compatibility with other compilers or with an ABI@.
3967 If you define this macro to be 0, then the conventions used for structure
3968 and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
3970 If not defined, this defaults to the value 1.
3972 @findex STRUCT_VALUE_REGNUM
3973 @item STRUCT_VALUE_REGNUM
3974 If the structure value address is passed in a register, then
3975 @code{STRUCT_VALUE_REGNUM} should be the number of that register.
3977 @findex STRUCT_VALUE
3979 If the structure value address is not passed in a register, define
3980 @code{STRUCT_VALUE} as an expression returning an RTX for the place
3981 where the address is passed. If it returns 0, the address is passed as
3982 an ``invisible'' first argument.
3984 @findex STRUCT_VALUE_INCOMING_REGNUM
3985 @item STRUCT_VALUE_INCOMING_REGNUM
3986 On some architectures the place where the structure value address
3987 is found by the called function is not the same place that the
3988 caller put it. This can be due to register windows, or it could
3989 be because the function prologue moves it to a different place.
3991 If the incoming location of the structure value address is in a
3992 register, define this macro as the register number.
3994 @findex STRUCT_VALUE_INCOMING
3995 @item STRUCT_VALUE_INCOMING
3996 If the incoming location is not a register, then you should define
3997 @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the
3998 called function should find the value. If it should find the value on
3999 the stack, define this to create a @code{mem} which refers to the frame
4000 pointer. A definition of 0 means that the address is passed as an
4001 ``invisible'' first argument.
4003 @findex PCC_STATIC_STRUCT_RETURN
4004 @item PCC_STATIC_STRUCT_RETURN
4005 Define this macro if the usual system convention on the target machine
4006 for returning structures and unions is for the called function to return
4007 the address of a static variable containing the value.
4009 Do not define this if the usual system convention is for the caller to
4010 pass an address to the subroutine.
4012 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4013 nothing when you use @option{-freg-struct-return} mode.
4017 @subsection Caller-Saves Register Allocation
4019 If you enable it, GCC can save registers around function calls. This
4020 makes it possible to use call-clobbered registers to hold variables that
4021 must live across calls.
4024 @findex DEFAULT_CALLER_SAVES
4025 @item DEFAULT_CALLER_SAVES
4026 Define this macro if function calls on the target machine do not preserve
4027 any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
4028 for all registers. When defined, this macro enables @option{-fcaller-saves}
4029 by default for all optimization levels. It has no effect for optimization
4030 levels 2 and higher, where @option{-fcaller-saves} is the default.
4032 @findex CALLER_SAVE_PROFITABLE
4033 @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4034 A C expression to determine whether it is worthwhile to consider placing
4035 a pseudo-register in a call-clobbered hard register and saving and
4036 restoring it around each function call. The expression should be 1 when
4037 this is worth doing, and 0 otherwise.
4039 If you don't define this macro, a default is used which is good on most
4040 machines: @code{4 * @var{calls} < @var{refs}}.
4042 @findex HARD_REGNO_CALLER_SAVE_MODE
4043 @item HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4044 A C expression specifying which mode is required for saving @var{nregs}
4045 of a pseudo-register in call-clobbered hard register @var{regno}. If
4046 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4047 returned. For most machines this macro need not be defined since GCC
4048 will select the smallest suitable mode.
4051 @node Function Entry
4052 @subsection Function Entry and Exit
4053 @cindex function entry and exit
4057 This section describes the macros that output function entry
4058 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4060 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4061 If defined, a function that outputs the assembler code for entry to a
4062 function. The prologue is responsible for setting up the stack frame,
4063 initializing the frame pointer register, saving registers that must be
4064 saved, and allocating @var{size} additional bytes of storage for the
4065 local variables. @var{size} is an integer. @var{file} is a stdio
4066 stream to which the assembler code should be output.
4068 The label for the beginning of the function need not be output by this
4069 macro. That has already been done when the macro is run.
4071 @findex regs_ever_live
4072 To determine which registers to save, the macro can refer to the array
4073 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4074 @var{r} is used anywhere within the function. This implies the function
4075 prologue should save register @var{r}, provided it is not one of the
4076 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4077 @code{regs_ever_live}.)
4079 On machines that have ``register windows'', the function entry code does
4080 not save on the stack the registers that are in the windows, even if
4081 they are supposed to be preserved by function calls; instead it takes
4082 appropriate steps to ``push'' the register stack, if any non-call-used
4083 registers are used in the function.
4085 @findex frame_pointer_needed
4086 On machines where functions may or may not have frame-pointers, the
4087 function entry code must vary accordingly; it must set up the frame
4088 pointer if one is wanted, and not otherwise. To determine whether a
4089 frame pointer is in wanted, the macro can refer to the variable
4090 @code{frame_pointer_needed}. The variable's value will be 1 at run
4091 time in a function that needs a frame pointer. @xref{Elimination}.
4093 The function entry code is responsible for allocating any stack space
4094 required for the function. This stack space consists of the regions
4095 listed below. In most cases, these regions are allocated in the
4096 order listed, with the last listed region closest to the top of the
4097 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4098 the highest address if it is not defined). You can use a different order
4099 for a machine if doing so is more convenient or required for
4100 compatibility reasons. Except in cases where required by standard
4101 or by a debugger, there is no reason why the stack layout used by GCC
4102 need agree with that used by other compilers for a machine.
4105 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4106 If defined, a function that outputs assembler code at the end of a
4107 prologue. This should be used when the function prologue is being
4108 emitted as RTL, and you have some extra assembler that needs to be
4109 emitted. @xref{prologue instruction pattern}.
4112 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4113 If defined, a function that outputs assembler code at the start of an
4114 epilogue. This should be used when the function epilogue is being
4115 emitted as RTL, and you have some extra assembler that needs to be
4116 emitted. @xref{epilogue instruction pattern}.
4119 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4120 If defined, a function that outputs the assembler code for exit from a
4121 function. The epilogue is responsible for restoring the saved
4122 registers and stack pointer to their values when the function was
4123 called, and returning control to the caller. This macro takes the
4124 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4125 registers to restore are determined from @code{regs_ever_live} and
4126 @code{CALL_USED_REGISTERS} in the same way.
4128 On some machines, there is a single instruction that does all the work
4129 of returning from the function. On these machines, give that
4130 instruction the name @samp{return} and do not define the macro
4131 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4133 Do not define a pattern named @samp{return} if you want the
4134 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4135 switches to control whether return instructions or epilogues are used,
4136 define a @samp{return} pattern with a validity condition that tests the
4137 target switches appropriately. If the @samp{return} pattern's validity
4138 condition is false, epilogues will be used.
4140 On machines where functions may or may not have frame-pointers, the
4141 function exit code must vary accordingly. Sometimes the code for these
4142 two cases is completely different. To determine whether a frame pointer
4143 is wanted, the macro can refer to the variable
4144 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4145 a function that needs a frame pointer.
4147 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4148 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4149 The C variable @code{current_function_is_leaf} is nonzero for such a
4150 function. @xref{Leaf Functions}.
4152 On some machines, some functions pop their arguments on exit while
4153 others leave that for the caller to do. For example, the 68020 when
4154 given @option{-mrtd} pops arguments in functions that take a fixed
4155 number of arguments.
4157 @findex current_function_pops_args
4158 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4159 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4160 needs to know what was decided. The variable that is called
4161 @code{current_function_pops_args} is the number of bytes of its
4162 arguments that a function should pop. @xref{Scalar Return}.
4163 @c what is the "its arguments" in the above sentence referring to, pray
4164 @c tell? --mew 5feb93
4171 @findex current_function_pretend_args_size
4172 A region of @code{current_function_pretend_args_size} bytes of
4173 uninitialized space just underneath the first argument arriving on the
4174 stack. (This may not be at the very start of the allocated stack region
4175 if the calling sequence has pushed anything else since pushing the stack
4176 arguments. But usually, on such machines, nothing else has been pushed
4177 yet, because the function prologue itself does all the pushing.) This
4178 region is used on machines where an argument may be passed partly in
4179 registers and partly in memory, and, in some cases to support the
4180 features in @code{<stdarg.h>}.
4183 An area of memory used to save certain registers used by the function.
4184 The size of this area, which may also include space for such things as
4185 the return address and pointers to previous stack frames, is
4186 machine-specific and usually depends on which registers have been used
4187 in the function. Machines with register windows often do not require
4191 A region of at least @var{size} bytes, possibly rounded up to an allocation
4192 boundary, to contain the local variables of the function. On some machines,
4193 this region and the save area may occur in the opposite order, with the
4194 save area closer to the top of the stack.
4197 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4198 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4199 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4200 argument lists of the function. @xref{Stack Arguments}.
4203 Normally, it is necessary for the macros
4204 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4205 @code{TARGET_ASM_FUNCTION_EPILOGUE} to treat leaf functions specially.
4206 The C variable @code{current_function_is_leaf} is nonzero for such a
4209 @findex EXIT_IGNORE_STACK
4210 @item EXIT_IGNORE_STACK
4211 Define this macro as a C expression that is nonzero if the return
4212 instruction or the function epilogue ignores the value of the stack
4213 pointer; in other words, if it is safe to delete an instruction to
4214 adjust the stack pointer before a return from the function.
4216 Note that this macro's value is relevant only for functions for which
4217 frame pointers are maintained. It is never safe to delete a final
4218 stack adjustment in a function that has no frame pointer, and the
4219 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4221 @findex EPILOGUE_USES
4222 @item EPILOGUE_USES (@var{regno})
4223 Define this macro as a C expression that is nonzero for registers that are
4224 used by the epilogue or the @samp{return} pattern. The stack and frame
4225 pointer registers are already be assumed to be used as needed.
4228 @item EH_USES (@var{regno})
4229 Define this macro as a C expression that is nonzero for registers that are
4230 used by the exception handling mechanism, and so should be considered live
4231 on entry to an exception edge.
4233 @findex DELAY_SLOTS_FOR_EPILOGUE
4234 @item DELAY_SLOTS_FOR_EPILOGUE
4235 Define this macro if the function epilogue contains delay slots to which
4236 instructions from the rest of the function can be ``moved''. The
4237 definition should be a C expression whose value is an integer
4238 representing the number of delay slots there.
4240 @findex ELIGIBLE_FOR_EPILOGUE_DELAY
4241 @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4242 A C expression that returns 1 if @var{insn} can be placed in delay
4243 slot number @var{n} of the epilogue.
4245 The argument @var{n} is an integer which identifies the delay slot now
4246 being considered (since different slots may have different rules of
4247 eligibility). It is never negative and is always less than the number
4248 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4249 If you reject a particular insn for a given delay slot, in principle, it
4250 may be reconsidered for a subsequent delay slot. Also, other insns may
4251 (at least in principle) be considered for the so far unfilled delay
4254 @findex current_function_epilogue_delay_list
4255 @findex final_scan_insn
4256 The insns accepted to fill the epilogue delay slots are put in an RTL
4257 list made with @code{insn_list} objects, stored in the variable
4258 @code{current_function_epilogue_delay_list}. The insn for the first
4259 delay slot comes first in the list. Your definition of the macro
4260 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4261 outputting the insns in this list, usually by calling
4262 @code{final_scan_insn}.
4264 You need not define this macro if you did not define
4265 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4269 @findex TARGET_ASM_OUTPUT_MI_THUNK
4270 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, tree @var{function})
4271 A function that outputs the assembler code for a thunk
4272 function, used to implement C++ virtual function calls with multiple
4273 inheritance. The thunk acts as a wrapper around a virtual function,
4274 adjusting the implicit object parameter before handing control off to
4277 First, emit code to add the integer @var{delta} to the location that
4278 contains the incoming first argument. Assume that this argument
4279 contains a pointer, and is the one used to pass the @code{this} pointer
4280 in C++. This is the incoming argument @emph{before} the function prologue,
4281 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4282 all other incoming arguments.
4284 After the addition, emit code to jump to @var{function}, which is a
4285 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4286 not touch the return address. Hence returning from @var{FUNCTION} will
4287 return to whoever called the current @samp{thunk}.
4289 The effect must be as if @var{function} had been called directly with
4290 the adjusted first argument. This macro is responsible for emitting all
4291 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4292 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4294 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4295 have already been extracted from it.) It might possibly be useful on
4296 some targets, but probably not.
4298 If you do not define this macro, the target-independent code in the C++
4299 front end will generate a less efficient heavyweight thunk that calls
4300 @var{function} instead of jumping to it. The generic approach does
4301 not support varargs.
4304 @findex TARGET_ASM_OUTPUT_MI_VCALL_THUNK
4305 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_VCALL_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, int @var{vcall_offset}, tree @var{function})
4306 A function like @code{TARGET_ASM_OUTPUT_MI_THUNK}, except that if
4307 @var{vcall_offset} is nonzero, an additional adjustment should be made
4308 after adding @code{delta}. In particular, if @var{p} is the
4309 adjusted pointer, the following adjustment should be made:
4312 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4316 If this function is defined, it will always be used in place of
4317 @code{TARGET_ASM_OUTPUT_MI_THUNK}.
4322 @subsection Generating Code for Profiling
4323 @cindex profiling, code generation
4325 These macros will help you generate code for profiling.
4328 @findex FUNCTION_PROFILER
4329 @item FUNCTION_PROFILER (@var{file}, @var{labelno})
4330 A C statement or compound statement to output to @var{file} some
4331 assembler code to call the profiling subroutine @code{mcount}.
4334 The details of how @code{mcount} expects to be called are determined by
4335 your operating system environment, not by GCC@. To figure them out,
4336 compile a small program for profiling using the system's installed C
4337 compiler and look at the assembler code that results.
4339 Older implementations of @code{mcount} expect the address of a counter
4340 variable to be loaded into some register. The name of this variable is
4341 @samp{LP} followed by the number @var{labelno}, so you would generate
4342 the name using @samp{LP%d} in a @code{fprintf}.
4344 @findex PROFILE_HOOK
4346 A C statement or compound statement to output to @var{file} some assembly
4347 code to call the profiling subroutine @code{mcount} even the target does
4348 not support profiling.
4350 @findex NO_PROFILE_COUNTERS
4351 @item NO_PROFILE_COUNTERS
4352 Define this macro if the @code{mcount} subroutine on your system does
4353 not need a counter variable allocated for each function. This is true
4354 for almost all modern implementations. If you define this macro, you
4355 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
4357 @findex PROFILE_BEFORE_PROLOGUE
4358 @item PROFILE_BEFORE_PROLOGUE
4359 Define this macro if the code for function profiling should come before
4360 the function prologue. Normally, the profiling code comes after.
4364 @subsection Permitting tail calls
4367 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4368 True if it is ok to do sibling call optimization for the specified
4369 call expression @var{exp}. @var{decl} will be the called function,
4370 or @code{NULL} if this is an indirect call.
4372 It is not uncommon for limitations of calling conventions to prevent
4373 tail calls to functions outside the current unit of translation, or
4374 during PIC compilation. The hook is used to enforce these restrictions,
4375 as the @code{sibcall} md pattern can not fail, or fall over to a
4376 ``normal'' call. The criteria for successful sibling call optimization
4377 may vary greatly between different architectures.
4381 @section Implementing the Varargs Macros
4382 @cindex varargs implementation
4384 GCC comes with an implementation of @code{<varargs.h>} and
4385 @code{<stdarg.h>} that work without change on machines that pass arguments
4386 on the stack. Other machines require their own implementations of
4387 varargs, and the two machine independent header files must have
4388 conditionals to include it.
4390 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4391 the calling convention for @code{va_start}. The traditional
4392 implementation takes just one argument, which is the variable in which
4393 to store the argument pointer. The ISO implementation of
4394 @code{va_start} takes an additional second argument. The user is
4395 supposed to write the last named argument of the function here.
4397 However, @code{va_start} should not use this argument. The way to find
4398 the end of the named arguments is with the built-in functions described
4402 @findex __builtin_saveregs
4403 @item __builtin_saveregs ()
4404 Use this built-in function to save the argument registers in memory so
4405 that the varargs mechanism can access them. Both ISO and traditional
4406 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4407 you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
4409 On some machines, @code{__builtin_saveregs} is open-coded under the
4410 control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
4411 it calls a routine written in assembler language, found in
4414 Code generated for the call to @code{__builtin_saveregs} appears at the
4415 beginning of the function, as opposed to where the call to
4416 @code{__builtin_saveregs} is written, regardless of what the code is.
4417 This is because the registers must be saved before the function starts
4418 to use them for its own purposes.
4419 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4422 @findex __builtin_args_info
4423 @item __builtin_args_info (@var{category})
4424 Use this built-in function to find the first anonymous arguments in
4427 In general, a machine may have several categories of registers used for
4428 arguments, each for a particular category of data types. (For example,
4429 on some machines, floating-point registers are used for floating-point
4430 arguments while other arguments are passed in the general registers.)
4431 To make non-varargs functions use the proper calling convention, you
4432 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4433 registers in each category have been used so far
4435 @code{__builtin_args_info} accesses the same data structure of type
4436 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4437 with it, with @var{category} specifying which word to access. Thus, the
4438 value indicates the first unused register in a given category.
4440 Normally, you would use @code{__builtin_args_info} in the implementation
4441 of @code{va_start}, accessing each category just once and storing the
4442 value in the @code{va_list} object. This is because @code{va_list} will
4443 have to update the values, and there is no way to alter the
4444 values accessed by @code{__builtin_args_info}.
4446 @findex __builtin_next_arg
4447 @item __builtin_next_arg (@var{lastarg})
4448 This is the equivalent of @code{__builtin_args_info}, for stack
4449 arguments. It returns the address of the first anonymous stack
4450 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4451 returns the address of the location above the first anonymous stack
4452 argument. Use it in @code{va_start} to initialize the pointer for
4453 fetching arguments from the stack. Also use it in @code{va_start} to
4454 verify that the second parameter @var{lastarg} is the last named argument
4455 of the current function.
4457 @findex __builtin_classify_type
4458 @item __builtin_classify_type (@var{object})
4459 Since each machine has its own conventions for which data types are
4460 passed in which kind of register, your implementation of @code{va_arg}
4461 has to embody these conventions. The easiest way to categorize the
4462 specified data type is to use @code{__builtin_classify_type} together
4463 with @code{sizeof} and @code{__alignof__}.
4465 @code{__builtin_classify_type} ignores the value of @var{object},
4466 considering only its data type. It returns an integer describing what
4467 kind of type that is---integer, floating, pointer, structure, and so on.
4469 The file @file{typeclass.h} defines an enumeration that you can use to
4470 interpret the values of @code{__builtin_classify_type}.
4473 These machine description macros help implement varargs:
4476 @findex EXPAND_BUILTIN_SAVEREGS
4477 @item EXPAND_BUILTIN_SAVEREGS ()
4478 If defined, is a C expression that produces the machine-specific code
4479 for a call to @code{__builtin_saveregs}. This code will be moved to the
4480 very beginning of the function, before any parameter access are made.
4481 The return value of this function should be an RTX that contains the
4482 value to use as the return of @code{__builtin_saveregs}.
4484 @findex SETUP_INCOMING_VARARGS
4485 @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type}, @var{pretend_args_size}, @var{second_time})
4486 This macro offers an alternative to using @code{__builtin_saveregs} and
4487 defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the
4488 anonymous register arguments into the stack so that all the arguments
4489 appear to have been passed consecutively on the stack. Once this is
4490 done, you can use the standard implementation of varargs that works for
4491 machines that pass all their arguments on the stack.
4493 The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data
4494 structure, containing the values that are obtained after processing the
4495 named arguments. The arguments @var{mode} and @var{type} describe the
4496 last named argument---its machine mode and its data type as a tree node.
4498 The macro implementation should do two things: first, push onto the
4499 stack all the argument registers @emph{not} used for the named
4500 arguments, and second, store the size of the data thus pushed into the
4501 @code{int}-valued variable whose name is supplied as the argument
4502 @var{pretend_args_size}. The value that you store here will serve as
4503 additional offset for setting up the stack frame.
4505 Because you must generate code to push the anonymous arguments at
4506 compile time without knowing their data types,
4507 @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just
4508 a single category of argument register and use it uniformly for all data
4511 If the argument @var{second_time} is nonzero, it means that the
4512 arguments of the function are being analyzed for the second time. This
4513 happens for an inline function, which is not actually compiled until the
4514 end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should
4515 not generate any instructions in this case.
4517 @findex STRICT_ARGUMENT_NAMING
4518 @item STRICT_ARGUMENT_NAMING
4519 Define this macro to be a nonzero value if the location where a function
4520 argument is passed depends on whether or not it is a named argument.
4522 This macro controls how the @var{named} argument to @code{FUNCTION_ARG}
4523 is set for varargs and stdarg functions. If this macro returns a
4524 nonzero value, the @var{named} argument is always true for named
4525 arguments, and false for unnamed arguments. If it returns a value of
4526 zero, but @code{SETUP_INCOMING_VARARGS} is defined, then all arguments
4527 are treated as named. Otherwise, all named arguments except the last
4528 are treated as named.
4530 You need not define this macro if it always returns zero.
4532 @findex PRETEND_OUTGOING_VARARGS_NAMED
4533 @item PRETEND_OUTGOING_VARARGS_NAMED
4534 If you need to conditionally change ABIs so that one works with
4535 @code{SETUP_INCOMING_VARARGS}, but the other works like neither
4536 @code{SETUP_INCOMING_VARARGS} nor @code{STRICT_ARGUMENT_NAMING} was
4537 defined, then define this macro to return nonzero if
4538 @code{SETUP_INCOMING_VARARGS} is used, zero otherwise.
4539 Otherwise, you should not define this macro.
4543 @section Trampolines for Nested Functions
4544 @cindex trampolines for nested functions
4545 @cindex nested functions, trampolines for
4547 A @dfn{trampoline} is a small piece of code that is created at run time
4548 when the address of a nested function is taken. It normally resides on
4549 the stack, in the stack frame of the containing function. These macros
4550 tell GCC how to generate code to allocate and initialize a
4553 The instructions in the trampoline must do two things: load a constant
4554 address into the static chain register, and jump to the real address of
4555 the nested function. On CISC machines such as the m68k, this requires
4556 two instructions, a move immediate and a jump. Then the two addresses
4557 exist in the trampoline as word-long immediate operands. On RISC
4558 machines, it is often necessary to load each address into a register in
4559 two parts. Then pieces of each address form separate immediate
4562 The code generated to initialize the trampoline must store the variable
4563 parts---the static chain value and the function address---into the
4564 immediate operands of the instructions. On a CISC machine, this is
4565 simply a matter of copying each address to a memory reference at the
4566 proper offset from the start of the trampoline. On a RISC machine, it
4567 may be necessary to take out pieces of the address and store them
4571 @findex TRAMPOLINE_TEMPLATE
4572 @item TRAMPOLINE_TEMPLATE (@var{file})
4573 A C statement to output, on the stream @var{file}, assembler code for a
4574 block of data that contains the constant parts of a trampoline. This
4575 code should not include a label---the label is taken care of
4578 If you do not define this macro, it means no template is needed
4579 for the target. Do not define this macro on systems where the block move
4580 code to copy the trampoline into place would be larger than the code
4581 to generate it on the spot.
4583 @findex TRAMPOLINE_SECTION
4584 @item TRAMPOLINE_SECTION
4585 The name of a subroutine to switch to the section in which the
4586 trampoline template is to be placed (@pxref{Sections}). The default is
4587 a value of @samp{readonly_data_section}, which places the trampoline in
4588 the section containing read-only data.
4590 @findex TRAMPOLINE_SIZE
4591 @item TRAMPOLINE_SIZE
4592 A C expression for the size in bytes of the trampoline, as an integer.
4594 @findex TRAMPOLINE_ALIGNMENT
4595 @item TRAMPOLINE_ALIGNMENT
4596 Alignment required for trampolines, in bits.
4598 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4599 is used for aligning trampolines.
4601 @findex INITIALIZE_TRAMPOLINE
4602 @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4603 A C statement to initialize the variable parts of a trampoline.
4604 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4605 an RTX for the address of the nested function; @var{static_chain} is an
4606 RTX for the static chain value that should be passed to the function
4609 @findex TRAMPOLINE_ADJUST_ADDRESS
4610 @item TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4611 A C statement that should perform any machine-specific adjustment in
4612 the address of the trampoline. Its argument contains the address that
4613 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4614 used for a function call should be different from the address in which
4615 the template was stored, the different address should be assigned to
4616 @var{addr}. If this macro is not defined, @var{addr} will be used for
4619 @findex ALLOCATE_TRAMPOLINE
4620 @item ALLOCATE_TRAMPOLINE (@var{fp})
4621 A C expression to allocate run-time space for a trampoline. The
4622 expression value should be an RTX representing a memory reference to the
4623 space for the trampoline.
4625 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4626 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4627 If this macro is not defined, by default the trampoline is allocated as
4628 a stack slot. This default is right for most machines. The exceptions
4629 are machines where it is impossible to execute instructions in the stack
4630 area. On such machines, you may have to implement a separate stack,
4631 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4632 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4634 @var{fp} points to a data structure, a @code{struct function}, which
4635 describes the compilation status of the immediate containing function of
4636 the function which the trampoline is for. Normally (when
4637 @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the
4638 trampoline is in the stack frame of this containing function. Other
4639 allocation strategies probably must do something analogous with this
4643 Implementing trampolines is difficult on many machines because they have
4644 separate instruction and data caches. Writing into a stack location
4645 fails to clear the memory in the instruction cache, so when the program
4646 jumps to that location, it executes the old contents.
4648 Here are two possible solutions. One is to clear the relevant parts of
4649 the instruction cache whenever a trampoline is set up. The other is to
4650 make all trampolines identical, by having them jump to a standard
4651 subroutine. The former technique makes trampoline execution faster; the
4652 latter makes initialization faster.
4654 To clear the instruction cache when a trampoline is initialized, define
4655 the following macros which describe the shape of the cache.
4658 @findex INSN_CACHE_SIZE
4659 @item INSN_CACHE_SIZE
4660 The total size in bytes of the cache.
4662 @findex INSN_CACHE_LINE_WIDTH
4663 @item INSN_CACHE_LINE_WIDTH
4664 The length in bytes of each cache line. The cache is divided into cache
4665 lines which are disjoint slots, each holding a contiguous chunk of data
4666 fetched from memory. Each time data is brought into the cache, an
4667 entire line is read at once. The data loaded into a cache line is
4668 always aligned on a boundary equal to the line size.
4670 @findex INSN_CACHE_DEPTH
4671 @item INSN_CACHE_DEPTH
4672 The number of alternative cache lines that can hold any particular memory
4676 Alternatively, if the machine has system calls or instructions to clear
4677 the instruction cache directly, you can define the following macro.
4680 @findex CLEAR_INSN_CACHE
4681 @item CLEAR_INSN_CACHE (@var{beg}, @var{end})
4682 If defined, expands to a C expression clearing the @emph{instruction
4683 cache} in the specified interval. If it is not defined, and the macro
4684 @code{INSN_CACHE_SIZE} is defined, some generic code is generated to clear the
4685 cache. The definition of this macro would typically be a series of
4686 @code{asm} statements. Both @var{beg} and @var{end} are both pointer
4690 To use a standard subroutine, define the following macro. In addition,
4691 you must make sure that the instructions in a trampoline fill an entire
4692 cache line with identical instructions, or else ensure that the
4693 beginning of the trampoline code is always aligned at the same point in
4694 its cache line. Look in @file{m68k.h} as a guide.
4697 @findex TRANSFER_FROM_TRAMPOLINE
4698 @item TRANSFER_FROM_TRAMPOLINE
4699 Define this macro if trampolines need a special subroutine to do their
4700 work. The macro should expand to a series of @code{asm} statements
4701 which will be compiled with GCC@. They go in a library function named
4702 @code{__transfer_from_trampoline}.
4704 If you need to avoid executing the ordinary prologue code of a compiled
4705 C function when you jump to the subroutine, you can do so by placing a
4706 special label of your own in the assembler code. Use one @code{asm}
4707 statement to generate an assembler label, and another to make the label
4708 global. Then trampolines can use that label to jump directly to your
4709 special assembler code.
4713 @section Implicit Calls to Library Routines
4714 @cindex library subroutine names
4715 @cindex @file{libgcc.a}
4717 @c prevent bad page break with this line
4718 Here is an explanation of implicit calls to library routines.
4721 @findex MULSI3_LIBCALL
4722 @item MULSI3_LIBCALL
4723 A C string constant giving the name of the function to call for
4724 multiplication of one signed full-word by another. If you do not
4725 define this macro, the default name is used, which is @code{__mulsi3},
4726 a function defined in @file{libgcc.a}.
4728 @findex DIVSI3_LIBCALL
4729 @item DIVSI3_LIBCALL
4730 A C string constant giving the name of the function to call for
4731 division of one signed full-word by another. If you do not define
4732 this macro, the default name is used, which is @code{__divsi3}, a
4733 function defined in @file{libgcc.a}.
4735 @findex UDIVSI3_LIBCALL
4736 @item UDIVSI3_LIBCALL
4737 A C string constant giving the name of the function to call for
4738 division of one unsigned full-word by another. If you do not define
4739 this macro, the default name is used, which is @code{__udivsi3}, a
4740 function defined in @file{libgcc.a}.
4742 @findex MODSI3_LIBCALL
4743 @item MODSI3_LIBCALL
4744 A C string constant giving the name of the function to call for the
4745 remainder in division of one signed full-word by another. If you do
4746 not define this macro, the default name is used, which is
4747 @code{__modsi3}, a function defined in @file{libgcc.a}.
4749 @findex UMODSI3_LIBCALL
4750 @item UMODSI3_LIBCALL
4751 A C string constant giving the name of the function to call for the
4752 remainder in division of one unsigned full-word by another. If you do
4753 not define this macro, the default name is used, which is
4754 @code{__umodsi3}, a function defined in @file{libgcc.a}.
4756 @findex MULDI3_LIBCALL
4757 @item MULDI3_LIBCALL
4758 A C string constant giving the name of the function to call for
4759 multiplication of one signed double-word by another. If you do not
4760 define this macro, the default name is used, which is @code{__muldi3},
4761 a function defined in @file{libgcc.a}.
4763 @findex DIVDI3_LIBCALL
4764 @item DIVDI3_LIBCALL
4765 A C string constant giving the name of the function to call for
4766 division of one signed double-word by another. If you do not define
4767 this macro, the default name is used, which is @code{__divdi3}, a
4768 function defined in @file{libgcc.a}.
4770 @findex UDIVDI3_LIBCALL
4771 @item UDIVDI3_LIBCALL
4772 A C string constant giving the name of the function to call for
4773 division of one unsigned full-word by another. If you do not define
4774 this macro, the default name is used, which is @code{__udivdi3}, a
4775 function defined in @file{libgcc.a}.
4777 @findex MODDI3_LIBCALL
4778 @item MODDI3_LIBCALL
4779 A C string constant giving the name of the function to call for the
4780 remainder in division of one signed double-word by another. If you do
4781 not define this macro, the default name is used, which is
4782 @code{__moddi3}, a function defined in @file{libgcc.a}.
4784 @findex UMODDI3_LIBCALL
4785 @item UMODDI3_LIBCALL
4786 A C string constant giving the name of the function to call for the
4787 remainder in division of one unsigned full-word by another. If you do
4788 not define this macro, the default name is used, which is
4789 @code{__umoddi3}, a function defined in @file{libgcc.a}.
4791 @findex DECLARE_LIBRARY_RENAMES
4792 @item DECLARE_LIBRARY_RENAMES
4793 This macro, if defined, should expand to a piece of C code that will get
4794 expanded when compiling functions for libgcc.a. It can be used to
4795 provide alternate names for gcc's internal library functions if there
4796 are ABI-mandated names that the compiler should provide.
4798 @findex INIT_TARGET_OPTABS
4799 @item INIT_TARGET_OPTABS
4800 Define this macro as a C statement that declares additional library
4801 routines renames existing ones. @code{init_optabs} calls this macro after
4802 initializing all the normal library routines.
4804 @findex FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4805 @item FLOAT_LIB_COMPARE_RETURNS_BOOL
4806 Define this macro as a C statement that returns nonzero if a call to
4807 the floating point comparison library function will return a boolean
4808 value that indicates the result of the comparison. It should return
4809 zero if one of gcc's own libgcc functions is called.
4811 Most ports don't need to define this macro.
4814 @cindex @code{EDOM}, implicit usage
4816 The value of @code{EDOM} on the target machine, as a C integer constant
4817 expression. If you don't define this macro, GCC does not attempt to
4818 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4819 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4822 If you do not define @code{TARGET_EDOM}, then compiled code reports
4823 domain errors by calling the library function and letting it report the
4824 error. If mathematical functions on your system use @code{matherr} when
4825 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4826 that @code{matherr} is used normally.
4828 @findex GEN_ERRNO_RTX
4829 @cindex @code{errno}, implicit usage
4831 Define this macro as a C expression to create an rtl expression that
4832 refers to the global ``variable'' @code{errno}. (On certain systems,
4833 @code{errno} may not actually be a variable.) If you don't define this
4834 macro, a reasonable default is used.
4836 @findex TARGET_MEM_FUNCTIONS
4837 @cindex @code{bcopy}, implicit usage
4838 @cindex @code{memcpy}, implicit usage
4839 @cindex @code{memmove}, implicit usage
4840 @cindex @code{bzero}, implicit usage
4841 @cindex @code{memset}, implicit usage
4842 @item TARGET_MEM_FUNCTIONS
4843 Define this macro if GCC should generate calls to the ISO C
4844 (and System V) library functions @code{memcpy}, @code{memmove} and
4845 @code{memset} rather than the BSD functions @code{bcopy} and @code{bzero}.
4847 @findex TARGET_C99_FUNCTIONS
4848 @cindex C99 math functions, implicit usage
4849 @item TARGET_C99_FUNCTIONS
4850 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
4851 @code{sinf} and similarly for other functions defined by C99 standard. The
4852 default is nonzero that should be proper value for most modern systems, however
4853 number of existing systems lacks support for these functions in the runtime so
4854 they needs this macro to be redefined to 0.
4856 @findex LIBGCC_NEEDS_DOUBLE
4857 @item LIBGCC_NEEDS_DOUBLE
4858 Define this macro if @code{float} arguments cannot be passed to library
4859 routines (so they must be converted to @code{double}). This macro
4860 affects both how library calls are generated and how the library
4861 routines in @file{libgcc.a} accept their arguments. It is useful on
4862 machines where floating and fixed point arguments are passed
4863 differently, such as the i860.
4865 @findex NEXT_OBJC_RUNTIME
4866 @item NEXT_OBJC_RUNTIME
4867 Define this macro to generate code for Objective-C message sending using
4868 the calling convention of the NeXT system. This calling convention
4869 involves passing the object, the selector and the method arguments all
4870 at once to the method-lookup library function.
4872 The default calling convention passes just the object and the selector
4873 to the lookup function, which returns a pointer to the method.
4876 @node Addressing Modes
4877 @section Addressing Modes
4878 @cindex addressing modes
4880 @c prevent bad page break with this line
4881 This is about addressing modes.
4884 @findex HAVE_PRE_INCREMENT
4885 @findex HAVE_PRE_DECREMENT
4886 @findex HAVE_POST_INCREMENT
4887 @findex HAVE_POST_DECREMENT
4888 @item HAVE_PRE_INCREMENT
4889 @itemx HAVE_PRE_DECREMENT
4890 @itemx HAVE_POST_INCREMENT
4891 @itemx HAVE_POST_DECREMENT
4892 A C expression that is nonzero if the machine supports pre-increment,
4893 pre-decrement, post-increment, or post-decrement addressing respectively.
4895 @findex HAVE_POST_MODIFY_DISP
4896 @findex HAVE_PRE_MODIFY_DISP
4897 @item HAVE_PRE_MODIFY_DISP
4898 @itemx HAVE_POST_MODIFY_DISP
4899 A C expression that is nonzero if the machine supports pre- or
4900 post-address side-effect generation involving constants other than
4901 the size of the memory operand.
4903 @findex HAVE_POST_MODIFY_REG
4904 @findex HAVE_PRE_MODIFY_REG
4905 @item HAVE_PRE_MODIFY_REG
4906 @itemx HAVE_POST_MODIFY_REG
4907 A C expression that is nonzero if the machine supports pre- or
4908 post-address side-effect generation involving a register displacement.
4910 @findex CONSTANT_ADDRESS_P
4911 @item CONSTANT_ADDRESS_P (@var{x})
4912 A C expression that is 1 if the RTX @var{x} is a constant which
4913 is a valid address. On most machines, this can be defined as
4914 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4915 in which constant addresses are supported.
4918 @code{CONSTANT_P} accepts integer-values expressions whose values are
4919 not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and
4920 @code{high} expressions and @code{const} arithmetic expressions, in
4921 addition to @code{const_int} and @code{const_double} expressions.
4923 @findex MAX_REGS_PER_ADDRESS
4924 @item MAX_REGS_PER_ADDRESS
4925 A number, the maximum number of registers that can appear in a valid
4926 memory address. Note that it is up to you to specify a value equal to
4927 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4930 @findex GO_IF_LEGITIMATE_ADDRESS
4931 @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4932 A C compound statement with a conditional @code{goto @var{label};}
4933 executed if @var{x} (an RTX) is a legitimate memory address on the
4934 target machine for a memory operand of mode @var{mode}.
4936 It usually pays to define several simpler macros to serve as
4937 subroutines for this one. Otherwise it may be too complicated to
4940 This macro must exist in two variants: a strict variant and a
4941 non-strict one. The strict variant is used in the reload pass. It
4942 must be defined so that any pseudo-register that has not been
4943 allocated a hard register is considered a memory reference. In
4944 contexts where some kind of register is required, a pseudo-register
4945 with no hard register must be rejected.
4947 The non-strict variant is used in other passes. It must be defined to
4948 accept all pseudo-registers in every context where some kind of
4949 register is required.
4951 @findex REG_OK_STRICT
4952 Compiler source files that want to use the strict variant of this
4953 macro define the macro @code{REG_OK_STRICT}. You should use an
4954 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4955 in that case and the non-strict variant otherwise.
4957 Subroutines to check for acceptable registers for various purposes (one
4958 for base registers, one for index registers, and so on) are typically
4959 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4960 Then only these subroutine macros need have two variants; the higher
4961 levels of macros may be the same whether strict or not.
4963 Normally, constant addresses which are the sum of a @code{symbol_ref}
4964 and an integer are stored inside a @code{const} RTX to mark them as
4965 constant. Therefore, there is no need to recognize such sums
4966 specifically as legitimate addresses. Normally you would simply
4967 recognize any @code{const} as legitimate.
4969 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4970 sums that are not marked with @code{const}. It assumes that a naked
4971 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4972 naked constant sums as illegitimate addresses, so that none of them will
4973 be given to @code{PRINT_OPERAND_ADDRESS}.
4975 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
4976 On some machines, whether a symbolic address is legitimate depends on
4977 the section that the address refers to. On these machines, define the
4978 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
4979 into the @code{symbol_ref}, and then check for it here. When you see a
4980 @code{const}, you will have to look inside it to find the
4981 @code{symbol_ref} in order to determine the section. @xref{Assembler
4984 @findex saveable_obstack
4985 The best way to modify the name string is by adding text to the
4986 beginning, with suitable punctuation to prevent any ambiguity. Allocate
4987 the new name in @code{saveable_obstack}. You will have to modify
4988 @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and
4989 output the name accordingly, and define @code{TARGET_STRIP_NAME_ENCODING}
4990 to access the original name string.
4992 You can check the information stored here into the @code{symbol_ref} in
4993 the definitions of the macros @code{GO_IF_LEGITIMATE_ADDRESS} and
4994 @code{PRINT_OPERAND_ADDRESS}.
4996 @findex REG_OK_FOR_BASE_P
4997 @item REG_OK_FOR_BASE_P (@var{x})
4998 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4999 RTX) is valid for use as a base register. For hard registers, it
5000 should always accept those which the hardware permits and reject the
5001 others. Whether the macro accepts or rejects pseudo registers must be
5002 controlled by @code{REG_OK_STRICT} as described above. This usually
5003 requires two variant definitions, of which @code{REG_OK_STRICT}
5004 controls the one actually used.
5006 @findex REG_MODE_OK_FOR_BASE_P
5007 @item REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
5008 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
5009 that expression may examine the mode of the memory reference in
5010 @var{mode}. You should define this macro if the mode of the memory
5011 reference affects whether a register may be used as a base register. If
5012 you define this macro, the compiler will use it instead of
5013 @code{REG_OK_FOR_BASE_P}.
5015 @findex REG_OK_FOR_INDEX_P
5016 @item REG_OK_FOR_INDEX_P (@var{x})
5017 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
5018 RTX) is valid for use as an index register.
5020 The difference between an index register and a base register is that
5021 the index register may be scaled. If an address involves the sum of
5022 two registers, neither one of them scaled, then either one may be
5023 labeled the ``base'' and the other the ``index''; but whichever
5024 labeling is used must fit the machine's constraints of which registers
5025 may serve in each capacity. The compiler will try both labelings,
5026 looking for one that is valid, and will reload one or both registers
5027 only if neither labeling works.
5029 @findex FIND_BASE_TERM
5030 @item FIND_BASE_TERM (@var{x})
5031 A C expression to determine the base term of address @var{x}.
5032 This macro is used in only one place: `find_base_term' in alias.c.
5034 It is always safe for this macro to not be defined. It exists so
5035 that alias analysis can understand machine-dependent addresses.
5037 The typical use of this macro is to handle addresses containing
5038 a label_ref or symbol_ref within an UNSPEC@.
5040 @findex LEGITIMIZE_ADDRESS
5041 @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
5042 A C compound statement that attempts to replace @var{x} with a valid
5043 memory address for an operand of mode @var{mode}. @var{win} will be a
5044 C statement label elsewhere in the code; the macro definition may use
5047 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
5051 to avoid further processing if the address has become legitimate.
5053 @findex break_out_memory_refs
5054 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5055 and @var{oldx} will be the operand that was given to that function to produce
5058 The code generated by this macro should not alter the substructure of
5059 @var{x}. If it transforms @var{x} into a more legitimate form, it
5060 should assign @var{x} (which will always be a C variable) a new value.
5062 It is not necessary for this macro to come up with a legitimate
5063 address. The compiler has standard ways of doing so in all cases. In
5064 fact, it is safe for this macro to do nothing. But often a
5065 machine-dependent strategy can generate better code.
5067 @findex LEGITIMIZE_RELOAD_ADDRESS
5068 @item LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5069 A C compound statement that attempts to replace @var{x}, which is an address
5070 that needs reloading, with a valid memory address for an operand of mode
5071 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5072 It is not necessary to define this macro, but it might be useful for
5073 performance reasons.
5075 For example, on the i386, it is sometimes possible to use a single
5076 reload register instead of two by reloading a sum of two pseudo
5077 registers into a register. On the other hand, for number of RISC
5078 processors offsets are limited so that often an intermediate address
5079 needs to be generated in order to address a stack slot. By defining
5080 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5081 generated for adjacent some stack slots can be made identical, and thus
5084 @emph{Note}: This macro should be used with caution. It is necessary
5085 to know something of how reload works in order to effectively use this,
5086 and it is quite easy to produce macros that build in too much knowledge
5087 of reload internals.
5089 @emph{Note}: This macro must be able to reload an address created by a
5090 previous invocation of this macro. If it fails to handle such addresses
5091 then the compiler may generate incorrect code or abort.
5094 The macro definition should use @code{push_reload} to indicate parts that
5095 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5096 suitable to be passed unaltered to @code{push_reload}.
5098 The code generated by this macro must not alter the substructure of
5099 @var{x}. If it transforms @var{x} into a more legitimate form, it
5100 should assign @var{x} (which will always be a C variable) a new value.
5101 This also applies to parts that you change indirectly by calling
5104 @findex strict_memory_address_p
5105 The macro definition may use @code{strict_memory_address_p} to test if
5106 the address has become legitimate.
5109 If you want to change only a part of @var{x}, one standard way of doing
5110 this is to use @code{copy_rtx}. Note, however, that is unshares only a
5111 single level of rtl. Thus, if the part to be changed is not at the
5112 top level, you'll need to replace first the top level.
5113 It is not necessary for this macro to come up with a legitimate
5114 address; but often a machine-dependent strategy can generate better code.
5116 @findex GO_IF_MODE_DEPENDENT_ADDRESS
5117 @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5118 A C statement or compound statement with a conditional @code{goto
5119 @var{label};} executed if memory address @var{x} (an RTX) can have
5120 different meanings depending on the machine mode of the memory
5121 reference it is used for or if the address is valid for some modes
5124 Autoincrement and autodecrement addresses typically have mode-dependent
5125 effects because the amount of the increment or decrement is the size
5126 of the operand being addressed. Some machines have other mode-dependent
5127 addresses. Many RISC machines have no mode-dependent addresses.
5129 You may assume that @var{addr} is a valid address for the machine.
5131 @findex LEGITIMATE_CONSTANT_P
5132 @item LEGITIMATE_CONSTANT_P (@var{x})
5133 A C expression that is nonzero if @var{x} is a legitimate constant for
5134 an immediate operand on the target machine. You can assume that
5135 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5136 @samp{1} is a suitable definition for this macro on machines where
5137 anything @code{CONSTANT_P} is valid.
5140 @node Condition Code
5141 @section Condition Code Status
5142 @cindex condition code status
5144 @c prevent bad page break with this line
5145 This describes the condition code status.
5148 The file @file{conditions.h} defines a variable @code{cc_status} to
5149 describe how the condition code was computed (in case the interpretation of
5150 the condition code depends on the instruction that it was set by). This
5151 variable contains the RTL expressions on which the condition code is
5152 currently based, and several standard flags.
5154 Sometimes additional machine-specific flags must be defined in the machine
5155 description header file. It can also add additional machine-specific
5156 information by defining @code{CC_STATUS_MDEP}.
5159 @findex CC_STATUS_MDEP
5160 @item CC_STATUS_MDEP
5161 C code for a data type which is used for declaring the @code{mdep}
5162 component of @code{cc_status}. It defaults to @code{int}.
5164 This macro is not used on machines that do not use @code{cc0}.
5166 @findex CC_STATUS_MDEP_INIT
5167 @item CC_STATUS_MDEP_INIT
5168 A C expression to initialize the @code{mdep} field to ``empty''.
5169 The default definition does nothing, since most machines don't use
5170 the field anyway. If you want to use the field, you should probably
5171 define this macro to initialize it.
5173 This macro is not used on machines that do not use @code{cc0}.
5175 @findex NOTICE_UPDATE_CC
5176 @item NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5177 A C compound statement to set the components of @code{cc_status}
5178 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5179 this macro's responsibility to recognize insns that set the condition
5180 code as a byproduct of other activity as well as those that explicitly
5183 This macro is not used on machines that do not use @code{cc0}.
5185 If there are insns that do not set the condition code but do alter
5186 other machine registers, this macro must check to see whether they
5187 invalidate the expressions that the condition code is recorded as
5188 reflecting. For example, on the 68000, insns that store in address
5189 registers do not set the condition code, which means that usually
5190 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5191 insns. But suppose that the previous insn set the condition code
5192 based on location @samp{a4@@(102)} and the current insn stores a new
5193 value in @samp{a4}. Although the condition code is not changed by
5194 this, it will no longer be true that it reflects the contents of
5195 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5196 @code{cc_status} in this case to say that nothing is known about the
5197 condition code value.
5199 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5200 with the results of peephole optimization: insns whose patterns are
5201 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5202 constants which are just the operands. The RTL structure of these
5203 insns is not sufficient to indicate what the insns actually do. What
5204 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5205 @code{CC_STATUS_INIT}.
5207 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5208 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5209 @samp{cc}. This avoids having detailed information about patterns in
5210 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5212 @findex EXTRA_CC_MODES
5213 @item EXTRA_CC_MODES
5214 Condition codes are represented in registers by machine modes of class
5215 @code{MODE_CC}. By default, there is just one mode, @code{CCmode}, with
5216 this class. If you need more such modes, create a file named
5217 @file{@var{machine}-modes.def} in your @file{config/@var{machine}}
5218 directory (@pxref{Back End, , Anatomy of a Target Back End}), containing
5219 a list of these modes. Each entry in the list should be a call to the
5220 macro @code{CC}. This macro takes one argument, which is the name of
5221 the mode: it should begin with @samp{CC}. Do not put quotation marks
5222 around the name, or include the trailing @samp{mode}; these are
5223 automatically added. There should not be anything else in the file
5226 A sample @file{@var{machine}-modes.def} file might look like this:
5229 CC (CC_NOOV) /* @r{Comparison only valid if there was no overflow.} */
5230 CC (CCFP) /* @r{Floating point comparison that cannot trap.} */
5231 CC (CCFPE) /* @r{Floating point comparison that may trap.} */
5234 When you create this file, the macro @code{EXTRA_CC_MODES} is
5235 automatically defined by @command{configure}, with value @samp{1}.
5237 @findex SELECT_CC_MODE
5238 @item SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5239 Returns a mode from class @code{MODE_CC} to be used when comparison
5240 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5241 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5242 @pxref{Jump Patterns} for a description of the reason for this
5246 #define SELECT_CC_MODE(OP,X,Y) \
5247 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5248 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5249 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5250 || GET_CODE (X) == NEG) \
5251 ? CC_NOOVmode : CCmode))
5254 You need not define this macro if @code{EXTRA_CC_MODES} is not defined.
5256 @findex CANONICALIZE_COMPARISON
5257 @item CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5258 On some machines not all possible comparisons are defined, but you can
5259 convert an invalid comparison into a valid one. For example, the Alpha
5260 does not have a @code{GT} comparison, but you can use an @code{LT}
5261 comparison instead and swap the order of the operands.
5263 On such machines, define this macro to be a C statement to do any
5264 required conversions. @var{code} is the initial comparison code
5265 and @var{op0} and @var{op1} are the left and right operands of the
5266 comparison, respectively. You should modify @var{code}, @var{op0}, and
5267 @var{op1} as required.
5269 GCC will not assume that the comparison resulting from this macro is
5270 valid but will see if the resulting insn matches a pattern in the
5273 You need not define this macro if it would never change the comparison
5276 @findex REVERSIBLE_CC_MODE
5277 @item REVERSIBLE_CC_MODE (@var{mode})
5278 A C expression whose value is one if it is always safe to reverse a
5279 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5280 can ever return @var{mode} for a floating-point inequality comparison,
5281 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5283 You need not define this macro if it would always returns zero or if the
5284 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5285 For example, here is the definition used on the SPARC, where floating-point
5286 inequality comparisons are always given @code{CCFPEmode}:
5289 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5292 @findex REVERSE_CONDITION (@var{code}, @var{mode})
5293 A C expression whose value is reversed condition code of the @var{code} for
5294 comparison done in CC_MODE @var{mode}. The macro is used only in case
5295 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5296 machine has some non-standard way how to reverse certain conditionals. For
5297 instance in case all floating point conditions are non-trapping, compiler may
5298 freely convert unordered compares to ordered one. Then definition may look
5302 #define REVERSE_CONDITION(CODE, MODE) \
5303 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5304 : reverse_condition_maybe_unordered (CODE))
5307 @findex REVERSE_CONDEXEC_PREDICATES_P
5308 @item REVERSE_CONDEXEC_PREDICATES_P (@var{code1}, @var{code2})
5309 A C expression that returns true if the conditional execution predicate
5310 @var{code1} is the inverse of @var{code2} and vice versa. Define this to
5311 return 0 if the target has conditional execution predicates that cannot be
5312 reversed safely. If no expansion is specified, this macro is defined as
5316 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5317 ((x) == reverse_condition (y))
5323 @section Describing Relative Costs of Operations
5324 @cindex costs of instructions
5325 @cindex relative costs
5326 @cindex speed of instructions
5328 These macros let you describe the relative speed of various operations
5329 on the target machine.
5332 @findex REGISTER_MOVE_COST
5333 @item REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5334 A C expression for the cost of moving data of mode @var{mode} from a
5335 register in class @var{from} to one in class @var{to}. The classes are
5336 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5337 value of 2 is the default; other values are interpreted relative to
5340 It is not required that the cost always equal 2 when @var{from} is the
5341 same as @var{to}; on some machines it is expensive to move between
5342 registers if they are not general registers.
5344 If reload sees an insn consisting of a single @code{set} between two
5345 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5346 classes returns a value of 2, reload does not check to ensure that the
5347 constraints of the insn are met. Setting a cost of other than 2 will
5348 allow reload to verify that the constraints are met. You should do this
5349 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5351 @findex MEMORY_MOVE_COST
5352 @item MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5353 A C expression for the cost of moving data of mode @var{mode} between a
5354 register of class @var{class} and memory; @var{in} is zero if the value
5355 is to be written to memory, nonzero if it is to be read in. This cost
5356 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5357 registers and memory is more expensive than between two registers, you
5358 should define this macro to express the relative cost.
5360 If you do not define this macro, GCC uses a default cost of 4 plus
5361 the cost of copying via a secondary reload register, if one is
5362 needed. If your machine requires a secondary reload register to copy
5363 between memory and a register of @var{class} but the reload mechanism is
5364 more complex than copying via an intermediate, define this macro to
5365 reflect the actual cost of the move.
5367 GCC defines the function @code{memory_move_secondary_cost} if
5368 secondary reloads are needed. It computes the costs due to copying via
5369 a secondary register. If your machine copies from memory using a
5370 secondary register in the conventional way but the default base value of
5371 4 is not correct for your machine, define this macro to add some other
5372 value to the result of that function. The arguments to that function
5373 are the same as to this macro.
5377 A C expression for the cost of a branch instruction. A value of 1 is
5378 the default; other values are interpreted relative to that.
5381 Here are additional macros which do not specify precise relative costs,
5382 but only that certain actions are more expensive than GCC would
5386 @findex SLOW_BYTE_ACCESS
5387 @item SLOW_BYTE_ACCESS
5388 Define this macro as a C expression which is nonzero if accessing less
5389 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5390 faster than accessing a word of memory, i.e., if such access
5391 require more than one instruction or if there is no difference in cost
5392 between byte and (aligned) word loads.
5394 When this macro is not defined, the compiler will access a field by
5395 finding the smallest containing object; when it is defined, a fullword
5396 load will be used if alignment permits. Unless bytes accesses are
5397 faster than word accesses, using word accesses is preferable since it
5398 may eliminate subsequent memory access if subsequent accesses occur to
5399 other fields in the same word of the structure, but to different bytes.
5401 @findex SLOW_UNALIGNED_ACCESS
5402 @item SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5403 Define this macro to be the value 1 if memory accesses described by the
5404 @var{mode} and @var{alignment} parameters have a cost many times greater
5405 than aligned accesses, for example if they are emulated in a trap
5408 When this macro is nonzero, the compiler will act as if
5409 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5410 moves. This can cause significantly more instructions to be produced.
5411 Therefore, do not set this macro nonzero if unaligned accesses only add a
5412 cycle or two to the time for a memory access.
5414 If the value of this macro is always zero, it need not be defined. If
5415 this macro is defined, it should produce a nonzero value when
5416 @code{STRICT_ALIGNMENT} is nonzero.
5418 @findex DONT_REDUCE_ADDR
5419 @item DONT_REDUCE_ADDR
5420 Define this macro to inhibit strength reduction of memory addresses.
5421 (On some machines, such strength reduction seems to do harm rather
5426 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5427 which a sequence of insns should be generated instead of a
5428 string move insn or a library call. Increasing the value will always
5429 make code faster, but eventually incurs high cost in increased code size.
5431 Note that on machines where the corresponding move insn is a
5432 @code{define_expand} that emits a sequence of insns, this macro counts
5433 the number of such sequences.
5435 If you don't define this, a reasonable default is used.
5437 @findex MOVE_BY_PIECES_P
5438 @item MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5439 A C expression used to determine whether @code{move_by_pieces} will be used to
5440 copy a chunk of memory, or whether some other block move mechanism
5441 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5442 than @code{MOVE_RATIO}.
5444 @findex MOVE_MAX_PIECES
5445 @item MOVE_MAX_PIECES
5446 A C expression used by @code{move_by_pieces} to determine the largest unit
5447 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5451 The threshold of number of scalar move insns, @emph{below} which a sequence
5452 of insns should be generated to clear memory instead of a string clear insn
5453 or a library call. Increasing the value will always make code faster, but
5454 eventually incurs high cost in increased code size.
5456 If you don't define this, a reasonable default is used.
5458 @findex CLEAR_BY_PIECES_P
5459 @item CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5460 A C expression used to determine whether @code{clear_by_pieces} will be used
5461 to clear a chunk of memory, or whether some other block clear mechanism
5462 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5463 than @code{CLEAR_RATIO}.
5465 @findex STORE_BY_PIECES_P
5466 @item STORE_BY_PIECES_P (@var{size}, @var{alignment})
5467 A C expression used to determine whether @code{store_by_pieces} will be
5468 used to set a chunk of memory to a constant value, or whether some other
5469 mechanism will be used. Used by @code{__builtin_memset} when storing
5470 values other than constant zero and by @code{__builtin_strcpy} when
5471 when called with a constant source string.
5472 Defaults to @code{MOVE_BY_PIECES_P}.
5474 @findex USE_LOAD_POST_INCREMENT
5475 @item USE_LOAD_POST_INCREMENT (@var{mode})
5476 A C expression used to determine whether a load postincrement is a good
5477 thing to use for a given mode. Defaults to the value of
5478 @code{HAVE_POST_INCREMENT}.
5480 @findex USE_LOAD_POST_DECREMENT
5481 @item USE_LOAD_POST_DECREMENT (@var{mode})
5482 A C expression used to determine whether a load postdecrement is a good
5483 thing to use for a given mode. Defaults to the value of
5484 @code{HAVE_POST_DECREMENT}.
5486 @findex USE_LOAD_PRE_INCREMENT
5487 @item USE_LOAD_PRE_INCREMENT (@var{mode})
5488 A C expression used to determine whether a load preincrement is a good
5489 thing to use for a given mode. Defaults to the value of
5490 @code{HAVE_PRE_INCREMENT}.
5492 @findex USE_LOAD_PRE_DECREMENT
5493 @item USE_LOAD_PRE_DECREMENT (@var{mode})
5494 A C expression used to determine whether a load predecrement is a good
5495 thing to use for a given mode. Defaults to the value of
5496 @code{HAVE_PRE_DECREMENT}.
5498 @findex USE_STORE_POST_INCREMENT
5499 @item USE_STORE_POST_INCREMENT (@var{mode})
5500 A C expression used to determine whether a store postincrement is a good
5501 thing to use for a given mode. Defaults to the value of
5502 @code{HAVE_POST_INCREMENT}.
5504 @findex USE_STORE_POST_DECREMENT
5505 @item USE_STORE_POST_DECREMENT (@var{mode})
5506 A C expression used to determine whether a store postdecrement is a good
5507 thing to use for a given mode. Defaults to the value of
5508 @code{HAVE_POST_DECREMENT}.
5510 @findex USE_STORE_PRE_INCREMENT
5511 @item USE_STORE_PRE_INCREMENT (@var{mode})
5512 This macro is used to determine whether a store preincrement is a good
5513 thing to use for a given mode. Defaults to the value of
5514 @code{HAVE_PRE_INCREMENT}.
5516 @findex USE_STORE_PRE_DECREMENT
5517 @item USE_STORE_PRE_DECREMENT (@var{mode})
5518 This macro is used to determine whether a store predecrement is a good
5519 thing to use for a given mode. Defaults to the value of
5520 @code{HAVE_PRE_DECREMENT}.
5522 @findex NO_FUNCTION_CSE
5523 @item NO_FUNCTION_CSE
5524 Define this macro if it is as good or better to call a constant
5525 function address than to call an address kept in a register.
5527 @findex NO_RECURSIVE_FUNCTION_CSE
5528 @item NO_RECURSIVE_FUNCTION_CSE
5529 Define this macro if it is as good or better for a function to call
5530 itself with an explicit address than to call an address kept in a
5533 @findex RANGE_TEST_NON_SHORT_CIRCUIT
5534 @item RANGE_TEST_NON_SHORT_CIRCUIT
5535 Define this macro if a non-short-circuit operation produced by
5536 @samp{fold_range_test ()} is optimal. This macro defaults to true if
5537 @code{BRANCH_COST} is greater than or equal to the value 2.
5540 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total})
5541 This target hook describes the relative costs of RTL expressions.
5543 The cost may depend on the precise form of the expression, which is
5544 available for examination in @var{x}, and the rtx code of the expression
5545 in which it is contained, found in @var{outer_code}. @var{code} is the
5546 expression code---redundant, since it can be obtained with
5547 @code{GET_CODE (@var{x})}.
5549 In implementing this hook, you can use the construct
5550 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
5553 On entry to the hook, @code{*@var{total}} contains a default estimate
5554 for the cost of the expression. The hook should modify this value as
5557 The hook returns true when all subexpressions of @var{x} have been
5558 processed, and false when @code{rtx_cost} should recurse.
5561 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address})
5562 This hook computes the cost of an addressing mode that contains
5563 @var{address}. If not defined, the cost is computed from
5564 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
5566 For most CISC machines, the default cost is a good approximation of the
5567 true cost of the addressing mode. However, on RISC machines, all
5568 instructions normally have the same length and execution time. Hence
5569 all addresses will have equal costs.
5571 In cases where more than one form of an address is known, the form with
5572 the lowest cost will be used. If multiple forms have the same, lowest,
5573 cost, the one that is the most complex will be used.
5575 For example, suppose an address that is equal to the sum of a register
5576 and a constant is used twice in the same basic block. When this macro
5577 is not defined, the address will be computed in a register and memory
5578 references will be indirect through that register. On machines where
5579 the cost of the addressing mode containing the sum is no higher than
5580 that of a simple indirect reference, this will produce an additional
5581 instruction and possibly require an additional register. Proper
5582 specification of this macro eliminates this overhead for such machines.
5584 This hook is never called with an invalid address.
5586 On machines where an address involving more than one register is as
5587 cheap as an address computation involving only one register, defining
5588 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
5589 be live over a region of code where only one would have been if
5590 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
5591 should be considered in the definition of this macro. Equivalent costs
5592 should probably only be given to addresses with different numbers of
5593 registers on machines with lots of registers.
5597 @section Adjusting the Instruction Scheduler
5599 The instruction scheduler may need a fair amount of machine-specific
5600 adjustment in order to produce good code. GCC provides several target
5601 hooks for this purpose. It is usually enough to define just a few of
5602 them: try the first ones in this list first.
5604 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
5605 This hook returns the maximum number of instructions that can ever
5606 issue at the same time on the target machine. The default is one.
5607 Although the insn scheduler can define itself the possibility of issue
5608 an insn on the same cycle, the value can serve as an additional
5609 constraint to issue insns on the same simulated processor cycle (see
5610 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
5611 This value must be constant over the entire compilation. If you need
5612 it to vary depending on what the instructions are, you must use
5613 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
5615 For the automaton based pipeline interface, you could define this hook
5616 to return the value of the macro @code{MAX_DFA_ISSUE_RATE}.
5619 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
5620 This hook is executed by the scheduler after it has scheduled an insn
5621 from the ready list. It should return the number of insns which can
5622 still be issued in the current cycle. The default is
5623 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
5624 @code{USE}, which normally are not counted against the issue rate.
5625 You should define this hook if some insns take more machine resources
5626 than others, so that fewer insns can follow them in the same cycle.
5627 @var{file} is either a null pointer, or a stdio stream to write any
5628 debug output to. @var{verbose} is the verbose level provided by
5629 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
5633 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
5634 This function corrects the value of @var{cost} based on the
5635 relationship between @var{insn} and @var{dep_insn} through the
5636 dependence @var{link}. It should return the new value. The default
5637 is to make no adjustment to @var{cost}. This can be used for example
5638 to specify to the scheduler using the traditional pipeline description
5639 that an output- or anti-dependence does not incur the same cost as a
5640 data-dependence. If the scheduler using the automaton based pipeline
5641 description, the cost of anti-dependence is zero and the cost of
5642 output-dependence is maximum of one and the difference of latency
5643 times of the first and the second insns. If these values are not
5644 acceptable, you could use the hook to modify them too. See also
5645 @pxref{Automaton pipeline description}.
5648 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
5649 This hook adjusts the integer scheduling priority @var{priority} of
5650 @var{insn}. It should return the new priority. Reduce the priority to
5651 execute @var{insn} earlier, increase the priority to execute @var{insn}
5652 later. Do not define this hook if you do not need to adjust the
5653 scheduling priorities of insns.
5656 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
5657 This hook is executed by the scheduler after it has scheduled the ready
5658 list, to allow the machine description to reorder it (for example to
5659 combine two small instructions together on @samp{VLIW} machines).
5660 @var{file} is either a null pointer, or a stdio stream to write any
5661 debug output to. @var{verbose} is the verbose level provided by
5662 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
5663 list of instructions that are ready to be scheduled. @var{n_readyp} is
5664 a pointer to the number of elements in the ready list. The scheduler
5665 reads the ready list in reverse order, starting with
5666 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
5667 is the timer tick of the scheduler. You may modify the ready list and
5668 the number of ready insns. The return value is the number of insns that
5669 can issue this cycle; normally this is just @code{issue_rate}. See also
5670 @samp{TARGET_SCHED_REORDER2}.
5673 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
5674 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
5675 function is called whenever the scheduler starts a new cycle. This one
5676 is called once per iteration over a cycle, immediately after
5677 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
5678 return the number of insns to be scheduled in the same cycle. Defining
5679 this hook can be useful if there are frequent situations where
5680 scheduling one insn causes other insns to become ready in the same
5681 cycle. These other insns can then be taken into account properly.
5684 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
5685 This hook is called after evaluation forward dependencies of insns in
5686 chain given by two parameter values (@var{head} and @var{tail}
5687 correspondingly) but before insns scheduling of the insn chain. For
5688 example, it can be used for better insn classification if it requires
5689 analysis of dependencies. This hook can use backward and forward
5690 dependencies of the insn scheduler because they are already
5694 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
5695 This hook is executed by the scheduler at the beginning of each block of
5696 instructions that are to be scheduled. @var{file} is either a null
5697 pointer, or a stdio stream to write any debug output to. @var{verbose}
5698 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5699 @var{max_ready} is the maximum number of insns in the current scheduling
5700 region that can be live at the same time. This can be used to allocate
5701 scratch space if it is needed, e.g. by @samp{TARGET_SCHED_REORDER}.
5704 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
5705 This hook is executed by the scheduler at the end of each block of
5706 instructions that are to be scheduled. It can be used to perform
5707 cleanup of any actions done by the other scheduling hooks. @var{file}
5708 is either a null pointer, or a stdio stream to write any debug output
5709 to. @var{verbose} is the verbose level provided by
5710 @option{-fsched-verbose-@var{n}}.
5713 @deftypefn {Target Hook} int TARGET_SCHED_USE_DFA_PIPELINE_INTERFACE (void)
5714 This hook is called many times during insn scheduling. If the hook
5715 returns nonzero, the automaton based pipeline description is used for
5716 insn scheduling. Otherwise the traditional pipeline description is
5717 used. The default is usage of the traditional pipeline description.
5719 You should also remember that to simplify the insn scheduler sources
5720 an empty traditional pipeline description interface is generated even
5721 if there is no a traditional pipeline description in the @file{.md}
5722 file. The same is true for the automaton based pipeline description.
5723 That means that you should be accurate in defining the hook.
5726 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
5727 The hook returns an RTL insn. The automaton state used in the
5728 pipeline hazard recognizer is changed as if the insn were scheduled
5729 when the new simulated processor cycle starts. Usage of the hook may
5730 simplify the automaton pipeline description for some @acronym{VLIW}
5731 processors. If the hook is defined, it is used only for the automaton
5732 based pipeline description. The default is not to change the state
5733 when the new simulated processor cycle starts.
5736 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
5737 The hook can be used to initialize data used by the previous hook.
5740 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
5741 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
5742 to changed the state as if the insn were scheduled when the new
5743 simulated processor cycle finishes.
5746 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
5747 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
5748 used to initialize data used by the previous hook.
5751 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
5752 This hook controls better choosing an insn from the ready insn queue
5753 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
5754 chooses the first insn from the queue. If the hook returns a positive
5755 value, an additional scheduler code tries all permutations of
5756 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
5757 subsequent ready insns to choose an insn whose issue will result in
5758 maximal number of issued insns on the same cycle. For the
5759 @acronym{VLIW} processor, the code could actually solve the problem of
5760 packing simple insns into the @acronym{VLIW} insn. Of course, if the
5761 rules of @acronym{VLIW} packing are described in the automaton.
5763 This code also could be used for superscalar @acronym{RISC}
5764 processors. Let us consider a superscalar @acronym{RISC} processor
5765 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
5766 @var{B}, some insns can be executed only in pipelines @var{B} or
5767 @var{C}, and one insn can be executed in pipeline @var{B}. The
5768 processor may issue the 1st insn into @var{A} and the 2nd one into
5769 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
5770 until the next cycle. If the scheduler issues the 3rd insn the first,
5771 the processor could issue all 3 insns per cycle.
5773 Actually this code demonstrates advantages of the automaton based
5774 pipeline hazard recognizer. We try quickly and easy many insn
5775 schedules to choose the best one.
5777 The default is no multipass scheduling.
5780 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
5782 This hook controls what insns from the ready insn queue will be
5783 considered for the multipass insn scheduling. If the hook returns
5784 zero for insn passed as the parameter, the insn will be not chosen to
5787 The default is that any ready insns can be chosen to be issued.
5790 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
5792 This hook is called by the insn scheduler before issuing insn passed
5793 as the third parameter on given cycle. If the hook returns nonzero,
5794 the insn is not issued on given processors cycle. Instead of that,
5795 the processor cycle is advanced. If the value passed through the last
5796 parameter is zero, the insn ready queue is not sorted on the new cycle
5797 start as usually. The first parameter passes file for debugging
5798 output. The second one passes the scheduler verbose level of the
5799 debugging output. The forth and the fifth parameter values are
5800 correspondingly processor cycle on which the previous insn has been
5801 issued and the current processor cycle.
5804 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_BUBBLES (void)
5805 The @acronym{DFA}-based scheduler could take the insertion of nop
5806 operations for better insn scheduling into account. It can be done
5807 only if the multi-pass insn scheduling works (see hook
5808 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD}).
5810 Let us consider a @acronym{VLIW} processor insn with 3 slots. Each
5811 insn can be placed only in one of the three slots. We have 3 ready
5812 insns @var{A}, @var{B}, and @var{C}. @var{A} and @var{C} can be
5813 placed only in the 1st slot, @var{B} can be placed only in the 3rd
5814 slot. We described the automaton which does not permit empty slot
5815 gaps between insns (usually such description is simpler). Without
5816 this code the scheduler would place each insn in 3 separate
5817 @acronym{VLIW} insns. If the scheduler places a nop insn into the 2nd
5818 slot, it could place the 3 insns into 2 @acronym{VLIW} insns. What is
5819 the nop insn is returned by hook @samp{TARGET_SCHED_DFA_BUBBLE}. Hook
5820 @samp{TARGET_SCHED_INIT_DFA_BUBBLES} can be used to initialize or
5821 create the nop insns.
5823 You should remember that the scheduler does not insert the nop insns.
5824 It is not wise because of the following optimizations. The scheduler
5825 only considers such possibility to improve the result schedule. The
5826 nop insns should be inserted lately, e.g. on the final phase.
5829 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_BUBBLE (int @var{index})
5830 This hook @samp{FIRST_CYCLE_MULTIPASS_SCHEDULING} is used to insert
5831 nop operations for better insn scheduling when @acronym{DFA}-based
5832 scheduler makes multipass insn scheduling (see also description of
5833 hook @samp{TARGET_SCHED_INIT_DFA_BUBBLES}). This hook
5834 returns a nop insn with given @var{index}. The indexes start with
5835 zero. The hook should return @code{NULL} if there are no more nop
5836 insns with indexes greater than given index.
5839 Macros in the following table are generated by the program
5840 @file{genattr} and can be useful for writing the hooks.
5843 @findex TRADITIONAL_PIPELINE_INTERFACE
5844 @item TRADITIONAL_PIPELINE_INTERFACE
5845 The macro definition is generated if there is a traditional pipeline
5846 description in @file{.md} file. You should also remember that to
5847 simplify the insn scheduler sources an empty traditional pipeline
5848 description interface is generated even if there is no a traditional
5849 pipeline description in the @file{.md} file. The macro can be used to
5850 distinguish the two types of the traditional interface.
5852 @findex DFA_PIPELINE_INTERFACE
5853 @item DFA_PIPELINE_INTERFACE
5854 The macro definition is generated if there is an automaton pipeline
5855 description in @file{.md} file. You should also remember that to
5856 simplify the insn scheduler sources an empty automaton pipeline
5857 description interface is generated even if there is no an automaton
5858 pipeline description in the @file{.md} file. The macro can be used to
5859 distinguish the two types of the automaton interface.
5861 @findex MAX_DFA_ISSUE_RATE
5862 @item MAX_DFA_ISSUE_RATE
5863 The macro definition is generated in the automaton based pipeline
5864 description interface. Its value is calculated from the automaton
5865 based pipeline description and is equal to maximal number of all insns
5866 described in constructions @samp{define_insn_reservation} which can be
5867 issued on the same processor cycle.
5872 @section Dividing the Output into Sections (Texts, Data, @dots{})
5873 @c the above section title is WAY too long. maybe cut the part between
5874 @c the (...)? --mew 10feb93
5876 An object file is divided into sections containing different types of
5877 data. In the most common case, there are three sections: the @dfn{text
5878 section}, which holds instructions and read-only data; the @dfn{data
5879 section}, which holds initialized writable data; and the @dfn{bss
5880 section}, which holds uninitialized data. Some systems have other kinds
5883 The compiler must tell the assembler when to switch sections. These
5884 macros control what commands to output to tell the assembler this. You
5885 can also define additional sections.
5888 @findex TEXT_SECTION_ASM_OP
5889 @item TEXT_SECTION_ASM_OP
5890 A C expression whose value is a string, including spacing, containing the
5891 assembler operation that should precede instructions and read-only data.
5892 Normally @code{"\t.text"} is right.
5894 @findex TEXT_SECTION
5896 A C statement that switches to the default section containing instructions.
5897 Normally this is not needed, as simply defining @code{TEXT_SECTION_ASM_OP}
5898 is enough. The MIPS port uses this to sort all functions after all data
5901 @findex HOT_TEXT_SECTION_NAME
5902 @item HOT_TEXT_SECTION_NAME
5903 If defined, a C string constant for the name of the section containing most
5904 frequently executed functions of the program. If not defined, GCC will provide
5905 a default definition if the target supports named sections.
5907 @findex UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5908 @item UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5909 If defined, a C string constant for the name of the section containing unlikely
5910 executed functions in the program.
5912 @findex DATA_SECTION_ASM_OP
5913 @item DATA_SECTION_ASM_OP
5914 A C expression whose value is a string, including spacing, containing the
5915 assembler operation to identify the following data as writable initialized
5916 data. Normally @code{"\t.data"} is right.
5918 @findex READONLY_DATA_SECTION_ASM_OP
5919 @item READONLY_DATA_SECTION_ASM_OP
5920 A C expression whose value is a string, including spacing, containing the
5921 assembler operation to identify the following data as read-only initialized
5924 @findex READONLY_DATA_SECTION
5925 @item READONLY_DATA_SECTION
5926 A macro naming a function to call to switch to the proper section for
5927 read-only data. The default is to use @code{READONLY_DATA_SECTION_ASM_OP}
5928 if defined, else fall back to @code{text_section}.
5930 The most common definition will be @code{data_section}, if the target
5931 does not have a special read-only data section, and does not put data
5932 in the text section.
5934 @findex SHARED_SECTION_ASM_OP
5935 @item SHARED_SECTION_ASM_OP
5936 If defined, a C expression whose value is a string, including spacing,
5937 containing the assembler operation to identify the following data as
5938 shared data. If not defined, @code{DATA_SECTION_ASM_OP} will be used.
5940 @findex BSS_SECTION_ASM_OP
5941 @item BSS_SECTION_ASM_OP
5942 If defined, a C expression whose value is a string, including spacing,
5943 containing the assembler operation to identify the following data as
5944 uninitialized global data. If not defined, and neither
5945 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5946 uninitialized global data will be output in the data section if
5947 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5950 @findex SHARED_BSS_SECTION_ASM_OP
5951 @item SHARED_BSS_SECTION_ASM_OP
5952 If defined, a C expression whose value is a string, including spacing,
5953 containing the assembler operation to identify the following data as
5954 uninitialized global shared data. If not defined, and
5955 @code{BSS_SECTION_ASM_OP} is, the latter will be used.
5957 @findex INIT_SECTION_ASM_OP
5958 @item INIT_SECTION_ASM_OP
5959 If defined, a C expression whose value is a string, including spacing,
5960 containing the assembler operation to identify the following data as
5961 initialization code. If not defined, GCC will assume such a section does
5964 @findex FINI_SECTION_ASM_OP
5965 @item FINI_SECTION_ASM_OP
5966 If defined, a C expression whose value is a string, including spacing,
5967 containing the assembler operation to identify the following data as
5968 finalization code. If not defined, GCC will assume such a section does
5971 @findex CRT_CALL_STATIC_FUNCTION
5972 @item CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
5973 If defined, an ASM statement that switches to a different section
5974 via @var{section_op}, calls @var{function}, and switches back to
5975 the text section. This is used in @file{crtstuff.c} if
5976 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
5977 to initialization and finalization functions from the init and fini
5978 sections. By default, this macro uses a simple function call. Some
5979 ports need hand-crafted assembly code to avoid dependencies on
5980 registers initialized in the function prologue or to ensure that
5981 constant pools don't end up too far way in the text section.
5983 @findex FORCE_CODE_SECTION_ALIGN
5984 @item FORCE_CODE_SECTION_ALIGN
5985 If defined, an ASM statement that aligns a code section to some
5986 arbitrary boundary. This is used to force all fragments of the
5987 @code{.init} and @code{.fini} sections to have to same alignment
5988 and thus prevent the linker from having to add any padding.
5990 @findex EXTRA_SECTIONS
5993 @item EXTRA_SECTIONS
5994 A list of names for sections other than the standard two, which are
5995 @code{in_text} and @code{in_data}. You need not define this macro
5996 on a system with no other sections (that GCC needs to use).
5998 @findex EXTRA_SECTION_FUNCTIONS
5999 @findex text_section
6000 @findex data_section
6001 @item EXTRA_SECTION_FUNCTIONS
6002 One or more functions to be defined in @file{varasm.c}. These
6003 functions should do jobs analogous to those of @code{text_section} and
6004 @code{data_section}, for your additional sections. Do not define this
6005 macro if you do not define @code{EXTRA_SECTIONS}.
6007 @findex JUMP_TABLES_IN_TEXT_SECTION
6008 @item JUMP_TABLES_IN_TEXT_SECTION
6009 Define this macro to be an expression with a nonzero value if jump
6010 tables (for @code{tablejump} insns) should be output in the text
6011 section, along with the assembler instructions. Otherwise, the
6012 readonly data section is used.
6014 This macro is irrelevant if there is no separate readonly data section.
6017 @deftypefn {Target Hook} void TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
6018 Switches to the appropriate section for output of @var{exp}. You can
6019 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
6020 some sort. @var{reloc} indicates whether the initial value of @var{exp}
6021 requires link-time relocations. Bit 0 is set when variable contains
6022 local relocations only, while bit 1 is set for global relocations.
6023 Select the section by calling @code{data_section} or one of the
6024 alternatives for other sections. @var{align} is the constant alignment
6027 The default version of this function takes care of putting read-only
6028 variables in @code{readonly_data_section}.
6031 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
6032 Build up a unique section name, expressed as a @code{STRING_CST} node,
6033 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
6034 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
6035 the initial value of @var{exp} requires link-time relocations.
6037 The default version of this function appends the symbol name to the
6038 ELF section name that would normally be used for the symbol. For
6039 example, the function @code{foo} would be placed in @code{.text.foo}.
6040 Whatever the actual target object format, this is often good enough.
6043 @deftypefn {Target Hook} void TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
6044 Switches to the appropriate section for output of constant pool entry
6045 @var{x} in @var{mode}. You can assume that @var{x} is some kind of
6046 constant in RTL@. The argument @var{mode} is redundant except in the
6047 case of a @code{const_int} rtx. Select the section by calling
6048 @code{readonly_data_section} or one of the alternatives for other
6049 sections. @var{align} is the constant alignment in bits.
6051 The default version of this function takes care of putting symbolic
6052 constants in @code{flag_pic} mode in @code{data_section} and everything
6053 else in @code{readonly_data_section}.
6056 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
6057 Define this hook if references to a symbol or a constant must be
6058 treated differently depending on something about the variable or
6059 function named by the symbol (such as what section it is in).
6061 The hook is executed immediately after rtl has been created for
6062 @var{decl}, which may be a variable or function declaration or
6063 an entry in the constant pool. In either case, @var{rtl} is the
6064 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
6065 in this hook; that field may not have been initialized yet.
6067 In the case of a constant, it is safe to assume that the rtl is
6068 a @code{mem} whose address is a @code{symbol_ref}. Most decls
6069 will also have this form, but that is not guaranteed. Global
6070 register variables, for instance, will have a @code{reg} for their
6071 rtl. (Normally the right thing to do with such unusual rtl is
6074 The @var{new_decl_p} argument will be true if this is the first time
6075 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
6076 be false for subsequent invocations, which will happen for duplicate
6077 declarations. Whether or not anything must be done for the duplicate
6078 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
6079 @var{new_decl_p} is always true when the hook is called for a constant.
6081 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
6082 The usual thing for this hook to do is to record flags in the
6083 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
6084 Historically, the name string was modified if it was necessary to
6085 encode more than one bit of information, but this practice is now
6086 discouraged; use @code{SYMBOL_REF_FLAGS}.
6088 The default definition of this hook, @code{default_encode_section_info}
6089 in @file{varasm.c}, sets a number of commonly-useful bits in
6090 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
6091 before overriding it.
6094 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
6095 Decode @var{name} and return the real name part, sans
6096 the characters that @code{TARGET_ENCODE_SECTION_INFO}
6100 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
6101 Returns true if @var{exp} should be placed into a ``small data'' section.
6102 The default version of this hook always returns false.
6105 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
6106 Contains the value true if the target places read-only
6107 ``small data'' into a separate section. The default value is false.
6110 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
6111 Returns true if @var{exp} names an object for which name resolution
6112 rules must resolve to the current ``module'' (dynamic shared library
6113 or executable image).
6115 The default version of this hook implements the name resolution rules
6116 for ELF, which has a looser model of global name binding than other
6117 currently supported object file formats.
6120 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
6121 Contains the value true if the target supports thread-local storage.
6122 The default value is false.
6127 @section Position Independent Code
6128 @cindex position independent code
6131 This section describes macros that help implement generation of position
6132 independent code. Simply defining these macros is not enough to
6133 generate valid PIC; you must also add support to the macros
6134 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
6135 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
6136 @samp{movsi} to do something appropriate when the source operand
6137 contains a symbolic address. You may also need to alter the handling of
6138 switch statements so that they use relative addresses.
6139 @c i rearranged the order of the macros above to try to force one of
6140 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
6143 @findex PIC_OFFSET_TABLE_REGNUM
6144 @item PIC_OFFSET_TABLE_REGNUM
6145 The register number of the register used to address a table of static
6146 data addresses in memory. In some cases this register is defined by a
6147 processor's ``application binary interface'' (ABI)@. When this macro
6148 is defined, RTL is generated for this register once, as with the stack
6149 pointer and frame pointer registers. If this macro is not defined, it
6150 is up to the machine-dependent files to allocate such a register (if
6151 necessary). Note that this register must be fixed when in use (e.g.@:
6152 when @code{flag_pic} is true).
6154 @findex PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6155 @item PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6156 Define this macro if the register defined by
6157 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
6158 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
6160 @findex FINALIZE_PIC
6162 By generating position-independent code, when two different programs (A
6163 and B) share a common library (libC.a), the text of the library can be
6164 shared whether or not the library is linked at the same address for both
6165 programs. In some of these environments, position-independent code
6166 requires not only the use of different addressing modes, but also
6167 special code to enable the use of these addressing modes.
6169 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
6170 codes once the function is being compiled into assembly code, but not
6171 before. (It is not done before, because in the case of compiling an
6172 inline function, it would lead to multiple PIC prologues being
6173 included in functions which used inline functions and were compiled to
6176 @findex LEGITIMATE_PIC_OPERAND_P
6177 @item LEGITIMATE_PIC_OPERAND_P (@var{x})
6178 A C expression that is nonzero if @var{x} is a legitimate immediate
6179 operand on the target machine when generating position independent code.
6180 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6181 check this. You can also assume @var{flag_pic} is true, so you need not
6182 check it either. You need not define this macro if all constants
6183 (including @code{SYMBOL_REF}) can be immediate operands when generating
6184 position independent code.
6187 @node Assembler Format
6188 @section Defining the Output Assembler Language
6190 This section describes macros whose principal purpose is to describe how
6191 to write instructions in assembler language---rather than what the
6195 * File Framework:: Structural information for the assembler file.
6196 * Data Output:: Output of constants (numbers, strings, addresses).
6197 * Uninitialized Data:: Output of uninitialized variables.
6198 * Label Output:: Output and generation of labels.
6199 * Initialization:: General principles of initialization
6200 and termination routines.
6201 * Macros for Initialization::
6202 Specific macros that control the handling of
6203 initialization and termination routines.
6204 * Instruction Output:: Output of actual instructions.
6205 * Dispatch Tables:: Output of jump tables.
6206 * Exception Region Output:: Output of exception region code.
6207 * Alignment Output:: Pseudo ops for alignment and skipping data.
6210 @node File Framework
6211 @subsection The Overall Framework of an Assembler File
6212 @cindex assembler format
6213 @cindex output of assembler code
6215 @c prevent bad page break with this line
6216 This describes the overall framework of an assembler file.
6219 @findex ASM_FILE_START
6220 @item ASM_FILE_START (@var{stream})
6221 A C expression which outputs to the stdio stream @var{stream}
6222 some appropriate text to go at the start of an assembler file.
6224 Normally this macro is defined to output a line containing
6225 @samp{#NO_APP}, which is a comment that has no effect on most
6226 assemblers but tells the GNU assembler that it can save time by not
6227 checking for certain assembler constructs.
6229 On systems that use SDB, it is necessary to output certain commands;
6230 see @file{attasm.h}.
6232 @findex ASM_FILE_END
6233 @item ASM_FILE_END (@var{stream})
6234 A C expression which outputs to the stdio stream @var{stream}
6235 some appropriate text to go at the end of an assembler file.
6237 If this macro is not defined, the default is to output nothing
6238 special at the end of the file. Most systems don't require any
6241 On systems that use SDB, it is necessary to output certain commands;
6242 see @file{attasm.h}.
6244 @findex ASM_COMMENT_START
6245 @item ASM_COMMENT_START
6246 A C string constant describing how to begin a comment in the target
6247 assembler language. The compiler assumes that the comment will end at
6248 the end of the line.
6252 A C string constant for text to be output before each @code{asm}
6253 statement or group of consecutive ones. Normally this is
6254 @code{"#APP"}, which is a comment that has no effect on most
6255 assemblers but tells the GNU assembler that it must check the lines
6256 that follow for all valid assembler constructs.
6260 A C string constant for text to be output after each @code{asm}
6261 statement or group of consecutive ones. Normally this is
6262 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6263 time-saving assumptions that are valid for ordinary compiler output.
6265 @findex ASM_OUTPUT_SOURCE_FILENAME
6266 @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6267 A C statement to output COFF information or DWARF debugging information
6268 which indicates that filename @var{name} is the current source file to
6269 the stdio stream @var{stream}.
6271 This macro need not be defined if the standard form of output
6272 for the file format in use is appropriate.
6274 @findex OUTPUT_QUOTED_STRING
6275 @item OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6276 A C statement to output the string @var{string} to the stdio stream
6277 @var{stream}. If you do not call the function @code{output_quoted_string}
6278 in your config files, GCC will only call it to output filenames to
6279 the assembler source. So you can use it to canonicalize the format
6280 of the filename using this macro.
6282 @findex ASM_OUTPUT_SOURCE_LINE
6283 @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
6284 A C statement to output DBX or SDB debugging information before code
6285 for line number @var{line} of the current source file to the
6286 stdio stream @var{stream}.
6288 This macro need not be defined if the standard form of debugging
6289 information for the debugger in use is appropriate.
6291 @findex ASM_OUTPUT_IDENT
6292 @item ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6293 A C statement to output something to the assembler file to handle a
6294 @samp{#ident} directive containing the text @var{string}. If this
6295 macro is not defined, nothing is output for a @samp{#ident} directive.
6297 @findex OBJC_PROLOGUE
6299 A C statement to output any assembler statements which are required to
6300 precede any Objective-C object definitions or message sending. The
6301 statement is executed only when compiling an Objective-C program.
6304 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6305 Output assembly directives to switch to section @var{name}. The section
6306 should have attributes as specified by @var{flags}, which is a bit mask
6307 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6308 is nonzero, it contains an alignment in bytes to be used for the section,
6309 otherwise some target default should be used. Only targets that must
6310 specify an alignment within the section directive need pay attention to
6311 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6314 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6315 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6318 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6319 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6320 based on a variable or function decl, a section name, and whether or not the
6321 declaration's initializer may contain runtime relocations. @var{decl} may be
6322 null, in which case read-write data should be assumed.
6324 The default version if this function handles choosing code vs data,
6325 read-only vs read-write data, and @code{flag_pic}. You should only
6326 need to override this if your target has special flags that might be
6327 set via @code{__attribute__}.
6332 @subsection Output of Data
6335 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6336 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6337 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6338 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6339 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6340 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6341 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6342 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6343 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6344 These hooks specify assembly directives for creating certain kinds
6345 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6346 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6347 aligned two-byte object, and so on. Any of the hooks may be
6348 @code{NULL}, indicating that no suitable directive is available.
6350 The compiler will print these strings at the start of a new line,
6351 followed immediately by the object's initial value. In most cases,
6352 the string should contain a tab, a pseudo-op, and then another tab.
6355 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6356 The @code{assemble_integer} function uses this hook to output an
6357 integer object. @var{x} is the object's value, @var{size} is its size
6358 in bytes and @var{aligned_p} indicates whether it is aligned. The
6359 function should return @code{true} if it was able to output the
6360 object. If it returns false, @code{assemble_integer} will try to
6361 split the object into smaller parts.
6363 The default implementation of this hook will use the
6364 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6365 when the relevant string is @code{NULL}.
6369 @findex OUTPUT_ADDR_CONST_EXTRA
6370 @item OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6371 A C statement to recognize @var{rtx} patterns that
6372 @code{output_addr_const} can't deal with, and output assembly code to
6373 @var{stream} corresponding to the pattern @var{x}. This may be used to
6374 allow machine-dependent @code{UNSPEC}s to appear within constants.
6376 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6377 @code{goto fail}, so that a standard error message is printed. If it
6378 prints an error message itself, by calling, for example,
6379 @code{output_operand_lossage}, it may just complete normally.
6381 @findex ASM_OUTPUT_ASCII
6382 @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6383 A C statement to output to the stdio stream @var{stream} an assembler
6384 instruction to assemble a string constant containing the @var{len}
6385 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6386 @code{char *} and @var{len} a C expression of type @code{int}.
6388 If the assembler has a @code{.ascii} pseudo-op as found in the
6389 Berkeley Unix assembler, do not define the macro
6390 @code{ASM_OUTPUT_ASCII}.
6392 @findex ASM_OUTPUT_FDESC
6393 @item ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6394 A C statement to output word @var{n} of a function descriptor for
6395 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6396 is defined, and is otherwise unused.
6398 @findex CONSTANT_POOL_BEFORE_FUNCTION
6399 @item CONSTANT_POOL_BEFORE_FUNCTION
6400 You may define this macro as a C expression. You should define the
6401 expression to have a nonzero value if GCC should output the constant
6402 pool for a function before the code for the function, or a zero value if
6403 GCC should output the constant pool after the function. If you do
6404 not define this macro, the usual case, GCC will output the constant
6405 pool before the function.
6407 @findex ASM_OUTPUT_POOL_PROLOGUE
6408 @item ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
6409 A C statement to output assembler commands to define the start of the
6410 constant pool for a function. @var{funname} is a string giving
6411 the name of the function. Should the return type of the function
6412 be required, it can be obtained via @var{fundecl}. @var{size}
6413 is the size, in bytes, of the constant pool that will be written
6414 immediately after this call.
6416 If no constant-pool prefix is required, the usual case, this macro need
6419 @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY
6420 @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
6421 A C statement (with or without semicolon) to output a constant in the
6422 constant pool, if it needs special treatment. (This macro need not do
6423 anything for RTL expressions that can be output normally.)
6425 The argument @var{file} is the standard I/O stream to output the
6426 assembler code on. @var{x} is the RTL expression for the constant to
6427 output, and @var{mode} is the machine mode (in case @var{x} is a
6428 @samp{const_int}). @var{align} is the required alignment for the value
6429 @var{x}; you should output an assembler directive to force this much
6432 The argument @var{labelno} is a number to use in an internal label for
6433 the address of this pool entry. The definition of this macro is
6434 responsible for outputting the label definition at the proper place.
6435 Here is how to do this:
6438 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
6441 When you output a pool entry specially, you should end with a
6442 @code{goto} to the label @var{jumpto}. This will prevent the same pool
6443 entry from being output a second time in the usual manner.
6445 You need not define this macro if it would do nothing.
6447 @findex ASM_OUTPUT_POOL_EPILOGUE
6448 @item ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
6449 A C statement to output assembler commands to at the end of the constant
6450 pool for a function. @var{funname} is a string giving the name of the
6451 function. Should the return type of the function be required, you can
6452 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
6453 constant pool that GCC wrote immediately before this call.
6455 If no constant-pool epilogue is required, the usual case, you need not
6458 @findex IS_ASM_LOGICAL_LINE_SEPARATOR
6459 @item IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
6460 Define this macro as a C expression which is nonzero if @var{C} is
6461 used as a logical line separator by the assembler.
6463 If you do not define this macro, the default is that only
6464 the character @samp{;} is treated as a logical line separator.
6467 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
6468 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
6469 These target hooks are C string constants, describing the syntax in the
6470 assembler for grouping arithmetic expressions. If not overridden, they
6471 default to normal parentheses, which is correct for most assemblers.
6474 These macros are provided by @file{real.h} for writing the definitions
6475 of @code{ASM_OUTPUT_DOUBLE} and the like:
6478 @item REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
6479 @itemx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
6480 @itemx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
6481 @findex REAL_VALUE_TO_TARGET_SINGLE
6482 @findex REAL_VALUE_TO_TARGET_DOUBLE
6483 @findex REAL_VALUE_TO_TARGET_LONG_DOUBLE
6484 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
6485 floating point representation, and store its bit pattern in the variable
6486 @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE}, this variable should
6487 be a simple @code{long int}. For the others, it should be an array of
6488 @code{long int}. The number of elements in this array is determined by
6489 the size of the desired target floating point data type: 32 bits of it
6490 go in each @code{long int} array element. Each array element holds 32
6491 bits of the result, even if @code{long int} is wider than 32 bits on the
6494 The array element values are designed so that you can print them out
6495 using @code{fprintf} in the order they should appear in the target
6499 @node Uninitialized Data
6500 @subsection Output of Uninitialized Variables
6502 Each of the macros in this section is used to do the whole job of
6503 outputting a single uninitialized variable.
6506 @findex ASM_OUTPUT_COMMON
6507 @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6508 A C statement (sans semicolon) to output to the stdio stream
6509 @var{stream} the assembler definition of a common-label named
6510 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6511 is the size rounded up to whatever alignment the caller wants.
6513 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6514 output the name itself; before and after that, output the additional
6515 assembler syntax for defining the name, and a newline.
6517 This macro controls how the assembler definitions of uninitialized
6518 common global variables are output.
6520 @findex ASM_OUTPUT_ALIGNED_COMMON
6521 @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
6522 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
6523 separate, explicit argument. If you define this macro, it is used in
6524 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
6525 handling the required alignment of the variable. The alignment is specified
6526 as the number of bits.
6528 @findex ASM_OUTPUT_ALIGNED_DECL_COMMON
6529 @item ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6530 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
6531 variable to be output, if there is one, or @code{NULL_TREE} if there
6532 is no corresponding variable. If you define this macro, GCC will use it
6533 in place of both @code{ASM_OUTPUT_COMMON} and
6534 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
6535 the variable's decl in order to chose what to output.
6537 @findex ASM_OUTPUT_SHARED_COMMON
6538 @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6539 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
6540 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
6543 @findex ASM_OUTPUT_BSS
6544 @item ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6545 A C statement (sans semicolon) to output to the stdio stream
6546 @var{stream} the assembler definition of uninitialized global @var{decl} named
6547 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6548 is the size rounded up to whatever alignment the caller wants.
6550 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
6551 defining this macro. If unable, use the expression
6552 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
6553 before and after that, output the additional assembler syntax for defining
6554 the name, and a newline.
6556 This macro controls how the assembler definitions of uninitialized global
6557 variables are output. This macro exists to properly support languages like
6558 C++ which do not have @code{common} data. However, this macro currently
6559 is not defined for all targets. If this macro and
6560 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
6561 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
6562 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
6564 @findex ASM_OUTPUT_ALIGNED_BSS
6565 @item ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6566 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
6567 separate, explicit argument. If you define this macro, it is used in
6568 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
6569 handling the required alignment of the variable. The alignment is specified
6570 as the number of bits.
6572 Try to use function @code{asm_output_aligned_bss} defined in file
6573 @file{varasm.c} when defining this macro.
6575 @findex ASM_OUTPUT_SHARED_BSS
6576 @item ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6577 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
6578 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
6581 @findex ASM_OUTPUT_LOCAL
6582 @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6583 A C statement (sans semicolon) to output to the stdio stream
6584 @var{stream} the assembler definition of a local-common-label named
6585 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6586 is the size rounded up to whatever alignment the caller wants.
6588 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6589 output the name itself; before and after that, output the additional
6590 assembler syntax for defining the name, and a newline.
6592 This macro controls how the assembler definitions of uninitialized
6593 static variables are output.
6595 @findex ASM_OUTPUT_ALIGNED_LOCAL
6596 @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
6597 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
6598 separate, explicit argument. If you define this macro, it is used in
6599 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
6600 handling the required alignment of the variable. The alignment is specified
6601 as the number of bits.
6603 @findex ASM_OUTPUT_ALIGNED_DECL_LOCAL
6604 @item ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6605 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
6606 variable to be output, if there is one, or @code{NULL_TREE} if there
6607 is no corresponding variable. If you define this macro, GCC will use it
6608 in place of both @code{ASM_OUTPUT_DECL} and
6609 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
6610 the variable's decl in order to chose what to output.
6612 @findex ASM_OUTPUT_SHARED_LOCAL
6613 @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6614 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
6615 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
6620 @subsection Output and Generation of Labels
6622 @c prevent bad page break with this line
6623 This is about outputting labels.
6626 @findex ASM_OUTPUT_LABEL
6627 @findex assemble_name
6628 @item ASM_OUTPUT_LABEL (@var{stream}, @var{name})
6629 A C statement (sans semicolon) to output to the stdio stream
6630 @var{stream} the assembler definition of a label named @var{name}.
6631 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6632 output the name itself; before and after that, output the additional
6633 assembler syntax for defining the name, and a newline. A default
6634 definition of this macro is provided which is correct for most systems.
6638 A C string containing the appropriate assembler directive to specify the
6639 size of a symbol, without any arguments. On systems that use ELF, the
6640 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
6641 systems, the default is not to define this macro.
6643 Define this macro only if it is correct to use the default definitions
6644 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
6645 for your system. If you need your own custom definitions of those
6646 macros, or if you do not need explicit symbol sizes at all, do not
6649 @findex ASM_OUTPUT_SIZE_DIRECTIVE
6650 @item ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
6651 A C statement (sans semicolon) to output to the stdio stream
6652 @var{stream} a directive telling the assembler that the size of the
6653 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
6654 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6657 @findex ASM_OUTPUT_MEASURED_SIZE
6658 @item ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
6659 A C statement (sans semicolon) to output to the stdio stream
6660 @var{stream} a directive telling the assembler to calculate the size of
6661 the symbol @var{name} by subtracting its address from the current
6664 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6665 provided. The default assumes that the assembler recognizes a special
6666 @samp{.} symbol as referring to the current address, and can calculate
6667 the difference between this and another symbol. If your assembler does
6668 not recognize @samp{.} or cannot do calculations with it, you will need
6669 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
6673 A C string containing the appropriate assembler directive to specify the
6674 type of a symbol, without any arguments. On systems that use ELF, the
6675 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
6676 systems, the default is not to define this macro.
6678 Define this macro only if it is correct to use the default definition of
6679 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6680 custom definition of this macro, or if you do not need explicit symbol
6681 types at all, do not define this macro.
6683 @findex TYPE_OPERAND_FMT
6684 @item TYPE_OPERAND_FMT
6685 A C string which specifies (using @code{printf} syntax) the format of
6686 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
6687 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
6688 the default is not to define this macro.
6690 Define this macro only if it is correct to use the default definition of
6691 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6692 custom definition of this macro, or if you do not need explicit symbol
6693 types at all, do not define this macro.
6695 @findex ASM_OUTPUT_TYPE_DIRECTIVE
6696 @item ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
6697 A C statement (sans semicolon) to output to the stdio stream
6698 @var{stream} a directive telling the assembler that the type of the
6699 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
6700 that string is always either @samp{"function"} or @samp{"object"}, but
6701 you should not count on this.
6703 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
6704 definition of this macro is provided.
6706 @findex ASM_DECLARE_FUNCTION_NAME
6707 @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
6708 A C statement (sans semicolon) to output to the stdio stream
6709 @var{stream} any text necessary for declaring the name @var{name} of a
6710 function which is being defined. This macro is responsible for
6711 outputting the label definition (perhaps using
6712 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
6713 @code{FUNCTION_DECL} tree node representing the function.
6715 If this macro is not defined, then the function name is defined in the
6716 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6718 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6721 @findex ASM_DECLARE_FUNCTION_SIZE
6722 @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
6723 A C statement (sans semicolon) to output to the stdio stream
6724 @var{stream} any text necessary for declaring the size of a function
6725 which is being defined. The argument @var{name} is the name of the
6726 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
6727 representing the function.
6729 If this macro is not defined, then the function size is not defined.
6731 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
6734 @findex ASM_DECLARE_OBJECT_NAME
6735 @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6736 A C statement (sans semicolon) to output to the stdio stream
6737 @var{stream} any text necessary for declaring the name @var{name} of an
6738 initialized variable which is being defined. This macro must output the
6739 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6740 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6742 If this macro is not defined, then the variable name is defined in the
6743 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6745 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
6746 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
6748 @findex ASM_DECLARE_REGISTER_GLOBAL
6749 @item ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6750 A C statement (sans semicolon) to output to the stdio stream
6751 @var{stream} any text necessary for claiming a register @var{regno}
6752 for a global variable @var{decl} with name @var{name}.
6754 If you don't define this macro, that is equivalent to defining it to do
6757 @findex ASM_FINISH_DECLARE_OBJECT
6758 @item ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6759 A C statement (sans semicolon) to finish up declaring a variable name
6760 once the compiler has processed its initializer fully and thus has had a
6761 chance to determine the size of an array when controlled by an
6762 initializer. This is used on systems where it's necessary to declare
6763 something about the size of the object.
6765 If you don't define this macro, that is equivalent to defining it to do
6768 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
6769 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
6772 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
6773 This target hook is a function to output to the stdio stream
6774 @var{stream} some commands that will make the label @var{name} global;
6775 that is, available for reference from other files.
6777 The default implementation relies on a proper definition of
6778 @code{GLOBAL_ASM_OP}.
6782 @findex ASM_WEAKEN_LABEL
6783 @item ASM_WEAKEN_LABEL (@var{stream}, @var{name})
6784 A C statement (sans semicolon) to output to the stdio stream
6785 @var{stream} some commands that will make the label @var{name} weak;
6786 that is, available for reference from other files but only used if
6787 no other definition is available. Use the expression
6788 @code{assemble_name (@var{stream}, @var{name})} to output the name
6789 itself; before and after that, output the additional assembler syntax
6790 for making that name weak, and a newline.
6792 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
6793 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
6796 @findex ASM_WEAKEN_DECL
6797 @item ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
6798 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
6799 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
6800 or variable decl. If @var{value} is not @code{NULL}, this C statement
6801 should output to the stdio stream @var{stream} assembler code which
6802 defines (equates) the weak symbol @var{name} to have the value
6803 @var{value}. If @var{value} is @code{NULL}, it should output commands
6804 to make @var{name} weak.
6806 @findex SUPPORTS_WEAK
6808 A C expression which evaluates to true if the target supports weak symbols.
6810 If you don't define this macro, @file{defaults.h} provides a default
6811 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
6812 is defined, the default definition is @samp{1}; otherwise, it is
6813 @samp{0}. Define this macro if you want to control weak symbol support
6814 with a compiler flag such as @option{-melf}.
6816 @findex MAKE_DECL_ONE_ONLY (@var{decl})
6817 @item MAKE_DECL_ONE_ONLY
6818 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
6819 public symbol such that extra copies in multiple translation units will
6820 be discarded by the linker. Define this macro if your object file
6821 format provides support for this concept, such as the @samp{COMDAT}
6822 section flags in the Microsoft Windows PE/COFF format, and this support
6823 requires changes to @var{decl}, such as putting it in a separate section.
6825 @findex SUPPORTS_ONE_ONLY
6826 @item SUPPORTS_ONE_ONLY
6827 A C expression which evaluates to true if the target supports one-only
6830 If you don't define this macro, @file{varasm.c} provides a default
6831 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6832 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6833 you want to control one-only symbol support with a compiler flag, or if
6834 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6835 be emitted as one-only.
6837 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
6838 This target hook is a function to output to @var{asm_out_file} some
6839 commands that will make the symbol(s) associated with @var{decl} have
6840 hidden, protected or internal visibility as specified by @var{visibility}.
6843 @findex ASM_OUTPUT_EXTERNAL
6844 @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6845 A C statement (sans semicolon) to output to the stdio stream
6846 @var{stream} any text necessary for declaring the name of an external
6847 symbol named @var{name} which is referenced in this compilation but
6848 not defined. The value of @var{decl} is the tree node for the
6851 This macro need not be defined if it does not need to output anything.
6852 The GNU assembler and most Unix assemblers don't require anything.
6854 @findex ASM_OUTPUT_EXTERNAL_LIBCALL
6855 @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref})
6856 A C statement (sans semicolon) to output on @var{stream} an assembler
6857 pseudo-op to declare a library function name external. The name of the
6858 library function is given by @var{symref}, which has type @code{rtx} and
6859 is a @code{symbol_ref}.
6861 This macro need not be defined if it does not need to output anything.
6862 The GNU assembler and most Unix assemblers don't require anything.
6864 @findex ASM_OUTPUT_LABELREF
6865 @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6866 A C statement (sans semicolon) to output to the stdio stream
6867 @var{stream} a reference in assembler syntax to a label named
6868 @var{name}. This should add @samp{_} to the front of the name, if that
6869 is customary on your operating system, as it is in most Berkeley Unix
6870 systems. This macro is used in @code{assemble_name}.
6872 @findex ASM_OUTPUT_SYMBOL_REF
6873 @item ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6874 A C statement (sans semicolon) to output a reference to
6875 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
6876 will be used to output the name of the symbol. This macro may be used
6877 to modify the way a symbol is referenced depending on information
6878 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
6880 @findex ASM_OUTPUT_LABEL_REF
6881 @item ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
6882 A C statement (sans semicolon) to output a reference to @var{buf}, the
6883 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
6884 @code{assemble_name} will be used to output the name of the symbol.
6885 This macro is not used by @code{output_asm_label}, or the @code{%l}
6886 specifier that calls it; the intention is that this macro should be set
6887 when it is necessary to output a label differently when its address is
6891 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
6892 A function to output to the stdio stream @var{stream} a label whose
6893 name is made from the string @var{prefix} and the number @var{labelno}.
6895 It is absolutely essential that these labels be distinct from the labels
6896 used for user-level functions and variables. Otherwise, certain programs
6897 will have name conflicts with internal labels.
6899 It is desirable to exclude internal labels from the symbol table of the
6900 object file. Most assemblers have a naming convention for labels that
6901 should be excluded; on many systems, the letter @samp{L} at the
6902 beginning of a label has this effect. You should find out what
6903 convention your system uses, and follow it.
6905 The default version of this function utilizes ASM_GENERATE_INTERNAL_LABEL.
6910 @findex ASM_OUTPUT_DEBUG_LABEL
6911 @item ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
6912 A C statement to output to the stdio stream @var{stream} a debug info
6913 label whose name is made from the string @var{prefix} and the number
6914 @var{num}. This is useful for VLIW targets, where debug info labels
6915 may need to be treated differently than branch target labels. On some
6916 systems, branch target labels must be at the beginning of instruction
6917 bundles, but debug info labels can occur in the middle of instruction
6920 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
6923 @findex ASM_GENERATE_INTERNAL_LABEL
6924 @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
6925 A C statement to store into the string @var{string} a label whose name
6926 is made from the string @var{prefix} and the number @var{num}.
6928 This string, when output subsequently by @code{assemble_name}, should
6929 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
6930 with the same @var{prefix} and @var{num}.
6932 If the string begins with @samp{*}, then @code{assemble_name} will
6933 output the rest of the string unchanged. It is often convenient for
6934 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
6935 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
6936 to output the string, and may change it. (Of course,
6937 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
6938 you should know what it does on your machine.)
6940 @findex ASM_FORMAT_PRIVATE_NAME
6941 @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
6942 A C expression to assign to @var{outvar} (which is a variable of type
6943 @code{char *}) a newly allocated string made from the string
6944 @var{name} and the number @var{number}, with some suitable punctuation
6945 added. Use @code{alloca} to get space for the string.
6947 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
6948 produce an assembler label for an internal static variable whose name is
6949 @var{name}. Therefore, the string must be such as to result in valid
6950 assembler code. The argument @var{number} is different each time this
6951 macro is executed; it prevents conflicts between similarly-named
6952 internal static variables in different scopes.
6954 Ideally this string should not be a valid C identifier, to prevent any
6955 conflict with the user's own symbols. Most assemblers allow periods
6956 or percent signs in assembler symbols; putting at least one of these
6957 between the name and the number will suffice.
6959 If this macro is not defined, a default definition will be provided
6960 which is correct for most systems.
6962 @findex ASM_OUTPUT_DEF
6963 @item ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
6964 A C statement to output to the stdio stream @var{stream} assembler code
6965 which defines (equates) the symbol @var{name} to have the value @var{value}.
6968 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6969 correct for most systems.
6971 @findex ASM_OUTPUT_DEF_FROM_DECLS
6972 @item ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
6973 A C statement to output to the stdio stream @var{stream} assembler code
6974 which defines (equates) the symbol whose tree node is @var{decl_of_name}
6975 to have the value of the tree node @var{decl_of_value}. This macro will
6976 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
6977 the tree nodes are available.
6980 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6981 correct for most systems.
6983 @findex ASM_OUTPUT_WEAK_ALIAS
6984 @item ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
6985 A C statement to output to the stdio stream @var{stream} assembler code
6986 which defines (equates) the weak symbol @var{name} to have the value
6987 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
6988 an undefined weak symbol.
6990 Define this macro if the target only supports weak aliases; define
6991 @code{ASM_OUTPUT_DEF} instead if possible.
6993 @findex OBJC_GEN_METHOD_LABEL
6994 @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
6995 Define this macro to override the default assembler names used for
6996 Objective-C methods.
6998 The default name is a unique method number followed by the name of the
6999 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
7000 the category is also included in the assembler name (e.g.@:
7003 These names are safe on most systems, but make debugging difficult since
7004 the method's selector is not present in the name. Therefore, particular
7005 systems define other ways of computing names.
7007 @var{buf} is an expression of type @code{char *} which gives you a
7008 buffer in which to store the name; its length is as long as
7009 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
7010 50 characters extra.
7012 The argument @var{is_inst} specifies whether the method is an instance
7013 method or a class method; @var{class_name} is the name of the class;
7014 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
7015 in a category); and @var{sel_name} is the name of the selector.
7017 On systems where the assembler can handle quoted names, you can use this
7018 macro to provide more human-readable names.
7020 @findex ASM_DECLARE_CLASS_REFERENCE
7021 @item ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
7022 A C statement (sans semicolon) to output to the stdio stream
7023 @var{stream} commands to declare that the label @var{name} is an
7024 Objective-C class reference. This is only needed for targets whose
7025 linkers have special support for NeXT-style runtimes.
7027 @findex ASM_DECLARE_UNRESOLVED_REFERENCE
7028 @item ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
7029 A C statement (sans semicolon) to output to the stdio stream
7030 @var{stream} commands to declare that the label @var{name} is an
7031 unresolved Objective-C class reference. This is only needed for targets
7032 whose linkers have special support for NeXT-style runtimes.
7035 @node Initialization
7036 @subsection How Initialization Functions Are Handled
7037 @cindex initialization routines
7038 @cindex termination routines
7039 @cindex constructors, output of
7040 @cindex destructors, output of
7042 The compiled code for certain languages includes @dfn{constructors}
7043 (also called @dfn{initialization routines})---functions to initialize
7044 data in the program when the program is started. These functions need
7045 to be called before the program is ``started''---that is to say, before
7046 @code{main} is called.
7048 Compiling some languages generates @dfn{destructors} (also called
7049 @dfn{termination routines}) that should be called when the program
7052 To make the initialization and termination functions work, the compiler
7053 must output something in the assembler code to cause those functions to
7054 be called at the appropriate time. When you port the compiler to a new
7055 system, you need to specify how to do this.
7057 There are two major ways that GCC currently supports the execution of
7058 initialization and termination functions. Each way has two variants.
7059 Much of the structure is common to all four variations.
7061 @findex __CTOR_LIST__
7062 @findex __DTOR_LIST__
7063 The linker must build two lists of these functions---a list of
7064 initialization functions, called @code{__CTOR_LIST__}, and a list of
7065 termination functions, called @code{__DTOR_LIST__}.
7067 Each list always begins with an ignored function pointer (which may hold
7068 0, @minus{}1, or a count of the function pointers after it, depending on
7069 the environment). This is followed by a series of zero or more function
7070 pointers to constructors (or destructors), followed by a function
7071 pointer containing zero.
7073 Depending on the operating system and its executable file format, either
7074 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
7075 time and exit time. Constructors are called in reverse order of the
7076 list; destructors in forward order.
7078 The best way to handle static constructors works only for object file
7079 formats which provide arbitrarily-named sections. A section is set
7080 aside for a list of constructors, and another for a list of destructors.
7081 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
7082 object file that defines an initialization function also puts a word in
7083 the constructor section to point to that function. The linker
7084 accumulates all these words into one contiguous @samp{.ctors} section.
7085 Termination functions are handled similarly.
7087 This method will be chosen as the default by @file{target-def.h} if
7088 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
7089 support arbitrary sections, but does support special designated
7090 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
7091 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
7093 When arbitrary sections are available, there are two variants, depending
7094 upon how the code in @file{crtstuff.c} is called. On systems that
7095 support a @dfn{.init} section which is executed at program startup,
7096 parts of @file{crtstuff.c} are compiled into that section. The
7097 program is linked by the @command{gcc} driver like this:
7100 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
7103 The prologue of a function (@code{__init}) appears in the @code{.init}
7104 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
7105 for the function @code{__fini} in the @dfn{.fini} section. Normally these
7106 files are provided by the operating system or by the GNU C library, but
7107 are provided by GCC for a few targets.
7109 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
7110 compiled from @file{crtstuff.c}. They contain, among other things, code
7111 fragments within the @code{.init} and @code{.fini} sections that branch
7112 to routines in the @code{.text} section. The linker will pull all parts
7113 of a section together, which results in a complete @code{__init} function
7114 that invokes the routines we need at startup.
7116 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
7119 If no init section is available, when GCC compiles any function called
7120 @code{main} (or more accurately, any function designated as a program
7121 entry point by the language front end calling @code{expand_main_function}),
7122 it inserts a procedure call to @code{__main} as the first executable code
7123 after the function prologue. The @code{__main} function is defined
7124 in @file{libgcc2.c} and runs the global constructors.
7126 In file formats that don't support arbitrary sections, there are again
7127 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
7128 and an `a.out' format must be used. In this case,
7129 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
7130 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
7131 and with the address of the void function containing the initialization
7132 code as its value. The GNU linker recognizes this as a request to add
7133 the value to a @dfn{set}; the values are accumulated, and are eventually
7134 placed in the executable as a vector in the format described above, with
7135 a leading (ignored) count and a trailing zero element.
7136 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
7137 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
7138 the compilation of @code{main} to call @code{__main} as above, starting
7139 the initialization process.
7141 The last variant uses neither arbitrary sections nor the GNU linker.
7142 This is preferable when you want to do dynamic linking and when using
7143 file formats which the GNU linker does not support, such as `ECOFF'@. In
7144 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
7145 termination functions are recognized simply by their names. This requires
7146 an extra program in the linkage step, called @command{collect2}. This program
7147 pretends to be the linker, for use with GCC; it does its job by running
7148 the ordinary linker, but also arranges to include the vectors of
7149 initialization and termination functions. These functions are called
7150 via @code{__main} as described above. In order to use this method,
7151 @code{use_collect2} must be defined in the target in @file{config.gcc}.
7154 The following section describes the specific macros that control and
7155 customize the handling of initialization and termination functions.
7158 @node Macros for Initialization
7159 @subsection Macros Controlling Initialization Routines
7161 Here are the macros that control how the compiler handles initialization
7162 and termination functions:
7165 @findex INIT_SECTION_ASM_OP
7166 @item INIT_SECTION_ASM_OP
7167 If defined, a C string constant, including spacing, for the assembler
7168 operation to identify the following data as initialization code. If not
7169 defined, GCC will assume such a section does not exist. When you are
7170 using special sections for initialization and termination functions, this
7171 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7172 run the initialization functions.
7174 @item HAS_INIT_SECTION
7175 @findex HAS_INIT_SECTION
7176 If defined, @code{main} will not call @code{__main} as described above.
7177 This macro should be defined for systems that control start-up code
7178 on a symbol-by-symbol basis, such as OSF/1, and should not
7179 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7181 @item LD_INIT_SWITCH
7182 @findex LD_INIT_SWITCH
7183 If defined, a C string constant for a switch that tells the linker that
7184 the following symbol is an initialization routine.
7186 @item LD_FINI_SWITCH
7187 @findex LD_FINI_SWITCH
7188 If defined, a C string constant for a switch that tells the linker that
7189 the following symbol is a finalization routine.
7191 @item COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7192 If defined, a C statement that will write a function that can be
7193 automatically called when a shared library is loaded. The function
7194 should call @var{func}, which takes no arguments. If not defined, and
7195 the object format requires an explicit initialization function, then a
7196 function called @code{_GLOBAL__DI} will be generated.
7198 This function and the following one are used by collect2 when linking a
7199 shared library that needs constructors or destructors, or has DWARF2
7200 exception tables embedded in the code.
7202 @item COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7203 If defined, a C statement that will write a function that can be
7204 automatically called when a shared library is unloaded. The function
7205 should call @var{func}, which takes no arguments. If not defined, and
7206 the object format requires an explicit finalization function, then a
7207 function called @code{_GLOBAL__DD} will be generated.
7210 @findex INVOKE__main
7211 If defined, @code{main} will call @code{__main} despite the presence of
7212 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7213 where the init section is not actually run automatically, but is still
7214 useful for collecting the lists of constructors and destructors.
7216 @item SUPPORTS_INIT_PRIORITY
7217 @findex SUPPORTS_INIT_PRIORITY
7218 If nonzero, the C++ @code{init_priority} attribute is supported and the
7219 compiler should emit instructions to control the order of initialization
7220 of objects. If zero, the compiler will issue an error message upon
7221 encountering an @code{init_priority} attribute.
7224 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7225 This value is true if the target supports some ``native'' method of
7226 collecting constructors and destructors to be run at startup and exit.
7227 It is false if we must use @command{collect2}.
7230 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7231 If defined, a function that outputs assembler code to arrange to call
7232 the function referenced by @var{symbol} at initialization time.
7234 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7235 no arguments and with no return value. If the target supports initialization
7236 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7237 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7239 If this macro is not defined by the target, a suitable default will
7240 be chosen if (1) the target supports arbitrary section names, (2) the
7241 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7245 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7246 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7247 functions rather than initialization functions.
7250 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7251 generated for the generated object file will have static linkage.
7253 If your system uses @command{collect2} as the means of processing
7254 constructors, then that program normally uses @command{nm} to scan
7255 an object file for constructor functions to be called.
7257 On certain kinds of systems, you can define these macros to make
7258 @command{collect2} work faster (and, in some cases, make it work at all):
7261 @findex OBJECT_FORMAT_COFF
7262 @item OBJECT_FORMAT_COFF
7263 Define this macro if the system uses COFF (Common Object File Format)
7264 object files, so that @command{collect2} can assume this format and scan
7265 object files directly for dynamic constructor/destructor functions.
7267 @findex OBJECT_FORMAT_ROSE
7268 @item OBJECT_FORMAT_ROSE
7269 Define this macro if the system uses ROSE format object files, so that
7270 @command{collect2} can assume this format and scan object files directly
7271 for dynamic constructor/destructor functions.
7273 These macros are effective only in a native compiler; @command{collect2} as
7274 part of a cross compiler always uses @command{nm} for the target machine.
7276 @findex REAL_NM_FILE_NAME
7277 @item REAL_NM_FILE_NAME
7278 Define this macro as a C string constant containing the file name to use
7279 to execute @command{nm}. The default is to search the path normally for
7282 If your system supports shared libraries and has a program to list the
7283 dynamic dependencies of a given library or executable, you can define
7284 these macros to enable support for running initialization and
7285 termination functions in shared libraries:
7289 Define this macro to a C string constant containing the name of the program
7290 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7292 @findex PARSE_LDD_OUTPUT
7293 @item PARSE_LDD_OUTPUT (@var{ptr})
7294 Define this macro to be C code that extracts filenames from the output
7295 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7296 of type @code{char *} that points to the beginning of a line of output
7297 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7298 code must advance @var{ptr} to the beginning of the filename on that
7299 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7302 @node Instruction Output
7303 @subsection Output of Assembler Instructions
7305 @c prevent bad page break with this line
7306 This describes assembler instruction output.
7309 @findex REGISTER_NAMES
7310 @item REGISTER_NAMES
7311 A C initializer containing the assembler's names for the machine
7312 registers, each one as a C string constant. This is what translates
7313 register numbers in the compiler into assembler language.
7315 @findex ADDITIONAL_REGISTER_NAMES
7316 @item ADDITIONAL_REGISTER_NAMES
7317 If defined, a C initializer for an array of structures containing a name
7318 and a register number. This macro defines additional names for hard
7319 registers, thus allowing the @code{asm} option in declarations to refer
7320 to registers using alternate names.
7322 @findex ASM_OUTPUT_OPCODE
7323 @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7324 Define this macro if you are using an unusual assembler that
7325 requires different names for the machine instructions.
7327 The definition is a C statement or statements which output an
7328 assembler instruction opcode to the stdio stream @var{stream}. The
7329 macro-operand @var{ptr} is a variable of type @code{char *} which
7330 points to the opcode name in its ``internal'' form---the form that is
7331 written in the machine description. The definition should output the
7332 opcode name to @var{stream}, performing any translation you desire, and
7333 increment the variable @var{ptr} to point at the end of the opcode
7334 so that it will not be output twice.
7336 In fact, your macro definition may process less than the entire opcode
7337 name, or more than the opcode name; but if you want to process text
7338 that includes @samp{%}-sequences to substitute operands, you must take
7339 care of the substitution yourself. Just be sure to increment
7340 @var{ptr} over whatever text should not be output normally.
7342 @findex recog_data.operand
7343 If you need to look at the operand values, they can be found as the
7344 elements of @code{recog_data.operand}.
7346 If the macro definition does nothing, the instruction is output
7349 @findex FINAL_PRESCAN_INSN
7350 @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7351 If defined, a C statement to be executed just prior to the output of
7352 assembler code for @var{insn}, to modify the extracted operands so
7353 they will be output differently.
7355 Here the argument @var{opvec} is the vector containing the operands
7356 extracted from @var{insn}, and @var{noperands} is the number of
7357 elements of the vector which contain meaningful data for this insn.
7358 The contents of this vector are what will be used to convert the insn
7359 template into assembler code, so you can change the assembler output
7360 by changing the contents of the vector.
7362 This macro is useful when various assembler syntaxes share a single
7363 file of instruction patterns; by defining this macro differently, you
7364 can cause a large class of instructions to be output differently (such
7365 as with rearranged operands). Naturally, variations in assembler
7366 syntax affecting individual insn patterns ought to be handled by
7367 writing conditional output routines in those patterns.
7369 If this macro is not defined, it is equivalent to a null statement.
7371 @findex FINAL_PRESCAN_LABEL
7372 @item FINAL_PRESCAN_LABEL
7373 If defined, @code{FINAL_PRESCAN_INSN} will be called on each
7374 @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and
7375 @var{noperands} will be zero.
7377 @findex PRINT_OPERAND
7378 @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
7379 A C compound statement to output to stdio stream @var{stream} the
7380 assembler syntax for an instruction operand @var{x}. @var{x} is an
7383 @var{code} is a value that can be used to specify one of several ways
7384 of printing the operand. It is used when identical operands must be
7385 printed differently depending on the context. @var{code} comes from
7386 the @samp{%} specification that was used to request printing of the
7387 operand. If the specification was just @samp{%@var{digit}} then
7388 @var{code} is 0; if the specification was @samp{%@var{ltr}
7389 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
7392 If @var{x} is a register, this macro should print the register's name.
7393 The names can be found in an array @code{reg_names} whose type is
7394 @code{char *[]}. @code{reg_names} is initialized from
7395 @code{REGISTER_NAMES}.
7397 When the machine description has a specification @samp{%@var{punct}}
7398 (a @samp{%} followed by a punctuation character), this macro is called
7399 with a null pointer for @var{x} and the punctuation character for
7402 @findex PRINT_OPERAND_PUNCT_VALID_P
7403 @item PRINT_OPERAND_PUNCT_VALID_P (@var{code})
7404 A C expression which evaluates to true if @var{code} is a valid
7405 punctuation character for use in the @code{PRINT_OPERAND} macro. If
7406 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
7407 punctuation characters (except for the standard one, @samp{%}) are used
7410 @findex PRINT_OPERAND_ADDRESS
7411 @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
7412 A C compound statement to output to stdio stream @var{stream} the
7413 assembler syntax for an instruction operand that is a memory reference
7414 whose address is @var{x}. @var{x} is an RTL expression.
7416 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
7417 On some machines, the syntax for a symbolic address depends on the
7418 section that the address refers to. On these machines, define the hook
7419 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
7420 @code{symbol_ref}, and then check for it here. @xref{Assembler Format}.
7422 @findex DBR_OUTPUT_SEQEND
7423 @findex dbr_sequence_length
7424 @item DBR_OUTPUT_SEQEND(@var{file})
7425 A C statement, to be executed after all slot-filler instructions have
7426 been output. If necessary, call @code{dbr_sequence_length} to
7427 determine the number of slots filled in a sequence (zero if not
7428 currently outputting a sequence), to decide how many no-ops to output,
7431 Don't define this macro if it has nothing to do, but it is helpful in
7432 reading assembly output if the extent of the delay sequence is made
7433 explicit (e.g.@: with white space).
7435 @findex final_sequence
7436 Note that output routines for instructions with delay slots must be
7437 prepared to deal with not being output as part of a sequence
7438 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
7439 found.) The variable @code{final_sequence} is null when not
7440 processing a sequence, otherwise it contains the @code{sequence} rtx
7443 @findex REGISTER_PREFIX
7444 @findex LOCAL_LABEL_PREFIX
7445 @findex USER_LABEL_PREFIX
7446 @findex IMMEDIATE_PREFIX
7448 @item REGISTER_PREFIX
7449 @itemx LOCAL_LABEL_PREFIX
7450 @itemx USER_LABEL_PREFIX
7451 @itemx IMMEDIATE_PREFIX
7452 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
7453 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
7454 @file{final.c}). These are useful when a single @file{md} file must
7455 support multiple assembler formats. In that case, the various @file{tm.h}
7456 files can define these macros differently.
7458 @item ASM_FPRINTF_EXTENSIONS(@var{file}, @var{argptr}, @var{format})
7459 @findex ASM_FPRINTF_EXTENSIONS
7460 If defined this macro should expand to a series of @code{case}
7461 statements which will be parsed inside the @code{switch} statement of
7462 the @code{asm_fprintf} function. This allows targets to define extra
7463 printf formats which may useful when generating their assembler
7464 statements. Note that upper case letters are reserved for future
7465 generic extensions to asm_fprintf, and so are not available to target
7466 specific code. The output file is given by the parameter @var{file}.
7467 The varargs input pointer is @var{argptr} and the rest of the format
7468 string, starting the character after the one that is being switched
7469 upon, is pointed to by @var{format}.
7471 @findex ASSEMBLER_DIALECT
7472 @item ASSEMBLER_DIALECT
7473 If your target supports multiple dialects of assembler language (such as
7474 different opcodes), define this macro as a C expression that gives the
7475 numeric index of the assembler language dialect to use, with zero as the
7478 If this macro is defined, you may use constructs of the form
7480 @samp{@{option0|option1|option2@dots{}@}}
7483 in the output templates of patterns (@pxref{Output Template}) or in the
7484 first argument of @code{asm_fprintf}. This construct outputs
7485 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
7486 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
7487 within these strings retain their usual meaning. If there are fewer
7488 alternatives within the braces than the value of
7489 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
7491 If you do not define this macro, the characters @samp{@{}, @samp{|} and
7492 @samp{@}} do not have any special meaning when used in templates or
7493 operands to @code{asm_fprintf}.
7495 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
7496 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
7497 the variations in assembler language syntax with that mechanism. Define
7498 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
7499 if the syntax variant are larger and involve such things as different
7500 opcodes or operand order.
7502 @findex ASM_OUTPUT_REG_PUSH
7503 @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
7504 A C expression to output to @var{stream} some assembler code
7505 which will push hard register number @var{regno} onto the stack.
7506 The code need not be optimal, since this macro is used only when
7509 @findex ASM_OUTPUT_REG_POP
7510 @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
7511 A C expression to output to @var{stream} some assembler code
7512 which will pop hard register number @var{regno} off of the stack.
7513 The code need not be optimal, since this macro is used only when
7517 @node Dispatch Tables
7518 @subsection Output of Dispatch Tables
7520 @c prevent bad page break with this line
7521 This concerns dispatch tables.
7524 @cindex dispatch table
7525 @findex ASM_OUTPUT_ADDR_DIFF_ELT
7526 @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
7527 A C statement to output to the stdio stream @var{stream} an assembler
7528 pseudo-instruction to generate a difference between two labels.
7529 @var{value} and @var{rel} are the numbers of two internal labels. The
7530 definitions of these labels are output using
7531 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
7532 way here. For example,
7535 fprintf (@var{stream}, "\t.word L%d-L%d\n",
7536 @var{value}, @var{rel})
7539 You must provide this macro on machines where the addresses in a
7540 dispatch table are relative to the table's own address. If defined, GCC
7541 will also use this macro on all machines when producing PIC@.
7542 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
7543 mode and flags can be read.
7545 @findex ASM_OUTPUT_ADDR_VEC_ELT
7546 @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
7547 This macro should be provided on machines where the addresses
7548 in a dispatch table are absolute.
7550 The definition should be a C statement to output to the stdio stream
7551 @var{stream} an assembler pseudo-instruction to generate a reference to
7552 a label. @var{value} is the number of an internal label whose
7553 definition is output using @code{(*targetm.asm_out.internal_label)}.
7557 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
7560 @findex ASM_OUTPUT_CASE_LABEL
7561 @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
7562 Define this if the label before a jump-table needs to be output
7563 specially. The first three arguments are the same as for
7564 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
7565 jump-table which follows (a @code{jump_insn} containing an
7566 @code{addr_vec} or @code{addr_diff_vec}).
7568 This feature is used on system V to output a @code{swbeg} statement
7571 If this macro is not defined, these labels are output with
7572 @code{(*targetm.asm_out.internal_label)}.
7574 @findex ASM_OUTPUT_CASE_END
7575 @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
7576 Define this if something special must be output at the end of a
7577 jump-table. The definition should be a C statement to be executed
7578 after the assembler code for the table is written. It should write
7579 the appropriate code to stdio stream @var{stream}. The argument
7580 @var{table} is the jump-table insn, and @var{num} is the label-number
7581 of the preceding label.
7583 If this macro is not defined, nothing special is output at the end of
7587 @node Exception Region Output
7588 @subsection Assembler Commands for Exception Regions
7590 @c prevent bad page break with this line
7592 This describes commands marking the start and the end of an exception
7596 @findex EH_FRAME_SECTION_NAME
7597 @item EH_FRAME_SECTION_NAME
7598 If defined, a C string constant for the name of the section containing
7599 exception handling frame unwind information. If not defined, GCC will
7600 provide a default definition if the target supports named sections.
7601 @file{crtstuff.c} uses this macro to switch to the appropriate section.
7603 You should define this symbol if your target supports DWARF 2 frame
7604 unwind information and the default definition does not work.
7606 @findex EH_FRAME_IN_DATA_SECTION
7607 @item EH_FRAME_IN_DATA_SECTION
7608 If defined, DWARF 2 frame unwind information will be placed in the
7609 data section even though the target supports named sections. This
7610 might be necessary, for instance, if the system linker does garbage
7611 collection and sections cannot be marked as not to be collected.
7613 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
7616 @findex MASK_RETURN_ADDR
7617 @item MASK_RETURN_ADDR
7618 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
7619 that it does not contain any extraneous set bits in it.
7621 @findex DWARF2_UNWIND_INFO
7622 @item DWARF2_UNWIND_INFO
7623 Define this macro to 0 if your target supports DWARF 2 frame unwind
7624 information, but it does not yet work with exception handling.
7625 Otherwise, if your target supports this information (if it defines
7626 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
7627 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
7630 If this macro is defined to 1, the DWARF 2 unwinder will be the default
7631 exception handling mechanism; otherwise, @code{setjmp}/@code{longjmp} will be used by
7634 If this macro is defined to anything, the DWARF 2 unwinder will be used
7635 instead of inline unwinders and @code{__unwind_function} in the non-@code{setjmp} case.
7637 @findex DWARF_CIE_DATA_ALIGNMENT
7638 @item DWARF_CIE_DATA_ALIGNMENT
7639 This macro need only be defined if the target might save registers in the
7640 function prologue at an offset to the stack pointer that is not aligned to
7641 @code{UNITS_PER_WORD}. The definition should be the negative minimum
7642 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
7643 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
7644 the target supports DWARF 2 frame unwind information.
7648 @deftypefn {Target Hook} void TARGET_ASM_EXCEPTION_SECTION ()
7649 If defined, a function that switches to the section in which the main
7650 exception table is to be placed (@pxref{Sections}). The default is a
7651 function that switches to a section named @code{.gcc_except_table} on
7652 machines that support named sections via
7653 @code{TARGET_ASM_NAMED_SECTION}, otherwise if @option{-fpic} or
7654 @option{-fPIC} is in effect, the @code{data_section}, otherwise the
7655 @code{readonly_data_section}.
7658 @deftypefn {Target Hook} void TARGET_ASM_EH_FRAME_SECTION ()
7659 If defined, a function that switches to the section in which the DWARF 2
7660 frame unwind information to be placed (@pxref{Sections}). The default
7661 is a function that outputs a standard GAS section directive, if
7662 @code{EH_FRAME_SECTION_NAME} is defined, or else a data section
7663 directive followed by a synthetic label.
7666 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
7667 Contains the value true if the target should add a zero word onto the
7668 end of a Dwarf-2 frame info section when used for exception handling.
7669 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
7673 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
7674 Given a register, this hook should return a parallel of registers to
7675 represent where to find the register pieces. Define this hook if the
7676 register and its mode are represented in Dwarf in non-contiguous
7677 locations, or if the register should be represented in more than one
7678 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
7679 If not defined, the default is to return @code{NULL_RTX}.
7682 @node Alignment Output
7683 @subsection Assembler Commands for Alignment
7685 @c prevent bad page break with this line
7686 This describes commands for alignment.
7690 @item JUMP_ALIGN (@var{label})
7691 The alignment (log base 2) to put in front of @var{label}, which is
7692 a common destination of jumps and has no fallthru incoming edge.
7694 This macro need not be defined if you don't want any special alignment
7695 to be done at such a time. Most machine descriptions do not currently
7698 Unless it's necessary to inspect the @var{label} parameter, it is better
7699 to set the variable @var{align_jumps} in the target's
7700 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7701 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
7703 @findex LABEL_ALIGN_AFTER_BARRIER
7704 @item LABEL_ALIGN_AFTER_BARRIER (@var{label})
7705 The alignment (log base 2) to put in front of @var{label}, which follows
7708 This macro need not be defined if you don't want any special alignment
7709 to be done at such a time. Most machine descriptions do not currently
7712 @findex LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7713 @item LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7714 The maximum number of bytes to skip when applying
7715 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
7716 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7719 @item LOOP_ALIGN (@var{label})
7720 The alignment (log base 2) to put in front of @var{label}, which follows
7721 a @code{NOTE_INSN_LOOP_BEG} note.
7723 This macro need not be defined if you don't want any special alignment
7724 to be done at such a time. Most machine descriptions do not currently
7727 Unless it's necessary to inspect the @var{label} parameter, it is better
7728 to set the variable @code{align_loops} in the target's
7729 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7730 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
7732 @findex LOOP_ALIGN_MAX_SKIP
7733 @item LOOP_ALIGN_MAX_SKIP
7734 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
7735 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7738 @item LABEL_ALIGN (@var{label})
7739 The alignment (log base 2) to put in front of @var{label}.
7740 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
7741 the maximum of the specified values is used.
7743 Unless it's necessary to inspect the @var{label} parameter, it is better
7744 to set the variable @code{align_labels} in the target's
7745 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7746 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
7748 @findex LABEL_ALIGN_MAX_SKIP
7749 @item LABEL_ALIGN_MAX_SKIP
7750 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
7751 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7753 @findex ASM_OUTPUT_SKIP
7754 @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
7755 A C statement to output to the stdio stream @var{stream} an assembler
7756 instruction to advance the location counter by @var{nbytes} bytes.
7757 Those bytes should be zero when loaded. @var{nbytes} will be a C
7758 expression of type @code{int}.
7760 @findex ASM_NO_SKIP_IN_TEXT
7761 @item ASM_NO_SKIP_IN_TEXT
7762 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
7763 text section because it fails to put zeros in the bytes that are skipped.
7764 This is true on many Unix systems, where the pseudo--op to skip bytes
7765 produces no-op instructions rather than zeros when used in the text
7768 @findex ASM_OUTPUT_ALIGN
7769 @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
7770 A C statement to output to the stdio stream @var{stream} an assembler
7771 command to advance the location counter to a multiple of 2 to the
7772 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
7774 @findex ASM_OUTPUT_ALIGN_WITH_NOP
7775 @item ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
7776 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
7777 for padding, if necessary.
7779 @findex ASM_OUTPUT_MAX_SKIP_ALIGN
7780 @item ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
7781 A C statement to output to the stdio stream @var{stream} an assembler
7782 command to advance the location counter to a multiple of 2 to the
7783 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
7784 satisfy the alignment request. @var{power} and @var{max_skip} will be
7785 a C expression of type @code{int}.
7789 @node Debugging Info
7790 @section Controlling Debugging Information Format
7792 @c prevent bad page break with this line
7793 This describes how to specify debugging information.
7796 * All Debuggers:: Macros that affect all debugging formats uniformly.
7797 * DBX Options:: Macros enabling specific options in DBX format.
7798 * DBX Hooks:: Hook macros for varying DBX format.
7799 * File Names and DBX:: Macros controlling output of file names in DBX format.
7800 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
7801 * VMS Debug:: Macros for VMS debug format.
7805 @subsection Macros Affecting All Debugging Formats
7807 @c prevent bad page break with this line
7808 These macros affect all debugging formats.
7811 @findex DBX_REGISTER_NUMBER
7812 @item DBX_REGISTER_NUMBER (@var{regno})
7813 A C expression that returns the DBX register number for the compiler
7814 register number @var{regno}. In the default macro provided, the value
7815 of this expression will be @var{regno} itself. But sometimes there are
7816 some registers that the compiler knows about and DBX does not, or vice
7817 versa. In such cases, some register may need to have one number in the
7818 compiler and another for DBX@.
7820 If two registers have consecutive numbers inside GCC, and they can be
7821 used as a pair to hold a multiword value, then they @emph{must} have
7822 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
7823 Otherwise, debuggers will be unable to access such a pair, because they
7824 expect register pairs to be consecutive in their own numbering scheme.
7826 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
7827 does not preserve register pairs, then what you must do instead is
7828 redefine the actual register numbering scheme.
7830 @findex DEBUGGER_AUTO_OFFSET
7831 @item DEBUGGER_AUTO_OFFSET (@var{x})
7832 A C expression that returns the integer offset value for an automatic
7833 variable having address @var{x} (an RTL expression). The default
7834 computation assumes that @var{x} is based on the frame-pointer and
7835 gives the offset from the frame-pointer. This is required for targets
7836 that produce debugging output for DBX or COFF-style debugging output
7837 for SDB and allow the frame-pointer to be eliminated when the
7838 @option{-g} options is used.
7840 @findex DEBUGGER_ARG_OFFSET
7841 @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
7842 A C expression that returns the integer offset value for an argument
7843 having address @var{x} (an RTL expression). The nominal offset is
7846 @findex PREFERRED_DEBUGGING_TYPE
7847 @item PREFERRED_DEBUGGING_TYPE
7848 A C expression that returns the type of debugging output GCC should
7849 produce when the user specifies just @option{-g}. Define
7850 this if you have arranged for GCC to support more than one format of
7851 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7852 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
7853 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
7855 When the user specifies @option{-ggdb}, GCC normally also uses the
7856 value of this macro to select the debugging output format, but with two
7857 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined and
7858 @code{LINKER_DOES_NOT_WORK_WITH_DWARF2} is not defined, GCC uses the
7859 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7860 defined, GCC uses @code{DBX_DEBUG}.
7862 The value of this macro only affects the default debugging output; the
7863 user can always get a specific type of output by using @option{-gstabs},
7864 @option{-gcoff}, @option{-gdwarf-1}, @option{-gdwarf-2}, @option{-gxcoff},
7869 @subsection Specific Options for DBX Output
7871 @c prevent bad page break with this line
7872 These are specific options for DBX output.
7875 @findex DBX_DEBUGGING_INFO
7876 @item DBX_DEBUGGING_INFO
7877 Define this macro if GCC should produce debugging output for DBX
7878 in response to the @option{-g} option.
7880 @findex XCOFF_DEBUGGING_INFO
7881 @item XCOFF_DEBUGGING_INFO
7882 Define this macro if GCC should produce XCOFF format debugging output
7883 in response to the @option{-g} option. This is a variant of DBX format.
7885 @findex DEFAULT_GDB_EXTENSIONS
7886 @item DEFAULT_GDB_EXTENSIONS
7887 Define this macro to control whether GCC should by default generate
7888 GDB's extended version of DBX debugging information (assuming DBX-format
7889 debugging information is enabled at all). If you don't define the
7890 macro, the default is 1: always generate the extended information
7891 if there is any occasion to.
7893 @findex DEBUG_SYMS_TEXT
7894 @item DEBUG_SYMS_TEXT
7895 Define this macro if all @code{.stabs} commands should be output while
7896 in the text section.
7898 @findex ASM_STABS_OP
7900 A C string constant, including spacing, naming the assembler pseudo op to
7901 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
7902 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
7903 applies only to DBX debugging information format.
7905 @findex ASM_STABD_OP
7907 A C string constant, including spacing, naming the assembler pseudo op to
7908 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
7909 value is the current location. If you don't define this macro,
7910 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
7913 @findex ASM_STABN_OP
7915 A C string constant, including spacing, naming the assembler pseudo op to
7916 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
7917 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
7918 macro applies only to DBX debugging information format.
7920 @findex DBX_NO_XREFS
7922 Define this macro if DBX on your system does not support the construct
7923 @samp{xs@var{tagname}}. On some systems, this construct is used to
7924 describe a forward reference to a structure named @var{tagname}.
7925 On other systems, this construct is not supported at all.
7927 @findex DBX_CONTIN_LENGTH
7928 @item DBX_CONTIN_LENGTH
7929 A symbol name in DBX-format debugging information is normally
7930 continued (split into two separate @code{.stabs} directives) when it
7931 exceeds a certain length (by default, 80 characters). On some
7932 operating systems, DBX requires this splitting; on others, splitting
7933 must not be done. You can inhibit splitting by defining this macro
7934 with the value zero. You can override the default splitting-length by
7935 defining this macro as an expression for the length you desire.
7937 @findex DBX_CONTIN_CHAR
7938 @item DBX_CONTIN_CHAR
7939 Normally continuation is indicated by adding a @samp{\} character to
7940 the end of a @code{.stabs} string when a continuation follows. To use
7941 a different character instead, define this macro as a character
7942 constant for the character you want to use. Do not define this macro
7943 if backslash is correct for your system.
7945 @findex DBX_STATIC_STAB_DATA_SECTION
7946 @item DBX_STATIC_STAB_DATA_SECTION
7947 Define this macro if it is necessary to go to the data section before
7948 outputting the @samp{.stabs} pseudo-op for a non-global static
7951 @findex DBX_TYPE_DECL_STABS_CODE
7952 @item DBX_TYPE_DECL_STABS_CODE
7953 The value to use in the ``code'' field of the @code{.stabs} directive
7954 for a typedef. The default is @code{N_LSYM}.
7956 @findex DBX_STATIC_CONST_VAR_CODE
7957 @item DBX_STATIC_CONST_VAR_CODE
7958 The value to use in the ``code'' field of the @code{.stabs} directive
7959 for a static variable located in the text section. DBX format does not
7960 provide any ``right'' way to do this. The default is @code{N_FUN}.
7962 @findex DBX_REGPARM_STABS_CODE
7963 @item DBX_REGPARM_STABS_CODE
7964 The value to use in the ``code'' field of the @code{.stabs} directive
7965 for a parameter passed in registers. DBX format does not provide any
7966 ``right'' way to do this. The default is @code{N_RSYM}.
7968 @findex DBX_REGPARM_STABS_LETTER
7969 @item DBX_REGPARM_STABS_LETTER
7970 The letter to use in DBX symbol data to identify a symbol as a parameter
7971 passed in registers. DBX format does not customarily provide any way to
7972 do this. The default is @code{'P'}.
7974 @findex DBX_MEMPARM_STABS_LETTER
7975 @item DBX_MEMPARM_STABS_LETTER
7976 The letter to use in DBX symbol data to identify a symbol as a stack
7977 parameter. The default is @code{'p'}.
7979 @findex DBX_FUNCTION_FIRST
7980 @item DBX_FUNCTION_FIRST
7981 Define this macro if the DBX information for a function and its
7982 arguments should precede the assembler code for the function. Normally,
7983 in DBX format, the debugging information entirely follows the assembler
7986 @findex DBX_LBRAC_FIRST
7987 @item DBX_LBRAC_FIRST
7988 Define this macro if the @code{N_LBRAC} symbol for a block should
7989 precede the debugging information for variables and functions defined in
7990 that block. Normally, in DBX format, the @code{N_LBRAC} symbol comes
7993 @findex DBX_BLOCKS_FUNCTION_RELATIVE
7994 @item DBX_BLOCKS_FUNCTION_RELATIVE
7995 Define this macro if the value of a symbol describing the scope of a
7996 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
7997 of the enclosing function. Normally, GCC uses an absolute address.
7999 @findex DBX_USE_BINCL
8001 Define this macro if GCC should generate @code{N_BINCL} and
8002 @code{N_EINCL} stabs for included header files, as on Sun systems. This
8003 macro also directs GCC to output a type number as a pair of a file
8004 number and a type number within the file. Normally, GCC does not
8005 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
8006 number for a type number.
8010 @subsection Open-Ended Hooks for DBX Format
8012 @c prevent bad page break with this line
8013 These are hooks for DBX format.
8016 @findex DBX_OUTPUT_LBRAC
8017 @item DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
8018 Define this macro to say how to output to @var{stream} the debugging
8019 information for the start of a scope level for variable names. The
8020 argument @var{name} is the name of an assembler symbol (for use with
8021 @code{assemble_name}) whose value is the address where the scope begins.
8023 @findex DBX_OUTPUT_RBRAC
8024 @item DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
8025 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
8027 @findex DBX_OUTPUT_NFUN
8028 @item DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
8029 Define this macro if the target machine requires special handling to
8030 output an @code{N_FUN} entry for the function @var{decl}.
8032 @findex DBX_OUTPUT_ENUM
8033 @item DBX_OUTPUT_ENUM (@var{stream}, @var{type})
8034 Define this macro if the target machine requires special handling to
8035 output an enumeration type. The definition should be a C statement
8036 (sans semicolon) to output the appropriate information to @var{stream}
8037 for the type @var{type}.
8039 @findex DBX_OUTPUT_FUNCTION_END
8040 @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
8041 Define this macro if the target machine requires special output at the
8042 end of the debugging information for a function. The definition should
8043 be a C statement (sans semicolon) to output the appropriate information
8044 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
8047 @findex DBX_OUTPUT_STANDARD_TYPES
8048 @item DBX_OUTPUT_STANDARD_TYPES (@var{syms})
8049 Define this macro if you need to control the order of output of the
8050 standard data types at the beginning of compilation. The argument
8051 @var{syms} is a @code{tree} which is a chain of all the predefined
8052 global symbols, including names of data types.
8054 Normally, DBX output starts with definitions of the types for integers
8055 and characters, followed by all the other predefined types of the
8056 particular language in no particular order.
8058 On some machines, it is necessary to output different particular types
8059 first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
8060 those symbols in the necessary order. Any predefined types that you
8061 don't explicitly output will be output afterward in no particular order.
8063 Be careful not to define this macro so that it works only for C@. There
8064 are no global variables to access most of the built-in types, because
8065 another language may have another set of types. The way to output a
8066 particular type is to look through @var{syms} to see if you can find it.
8072 for (decl = syms; decl; decl = TREE_CHAIN (decl))
8073 if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
8075 dbxout_symbol (decl);
8081 This does nothing if the expected type does not exist.
8083 See the function @code{init_decl_processing} in @file{c-decl.c} to find
8084 the names to use for all the built-in C types.
8086 Here is another way of finding a particular type:
8088 @c this is still overfull. --mew 10feb93
8092 for (decl = syms; decl; decl = TREE_CHAIN (decl))
8093 if (TREE_CODE (decl) == TYPE_DECL
8094 && (TREE_CODE (TREE_TYPE (decl))
8096 && TYPE_PRECISION (TREE_TYPE (decl)) == 16
8097 && TYPE_UNSIGNED (TREE_TYPE (decl)))
8099 /* @r{This must be @code{unsigned short}.} */
8100 dbxout_symbol (decl);
8106 @findex NO_DBX_FUNCTION_END
8107 @item NO_DBX_FUNCTION_END
8108 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
8109 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
8110 On those machines, define this macro to turn this feature off without
8111 disturbing the rest of the gdb extensions.
8115 @node File Names and DBX
8116 @subsection File Names in DBX Format
8118 @c prevent bad page break with this line
8119 This describes file names in DBX format.
8122 @findex DBX_WORKING_DIRECTORY
8123 @item DBX_WORKING_DIRECTORY
8124 Define this if DBX wants to have the current directory recorded in each
8127 Note that the working directory is always recorded if GDB extensions are
8130 @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME
8131 @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
8132 A C statement to output DBX debugging information to the stdio stream
8133 @var{stream} which indicates that file @var{name} is the main source
8134 file---the file specified as the input file for compilation.
8135 This macro is called only once, at the beginning of compilation.
8137 This macro need not be defined if the standard form of output
8138 for DBX debugging information is appropriate.
8140 @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY
8141 @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
8142 A C statement to output DBX debugging information to the stdio stream
8143 @var{stream} which indicates that the current directory during
8144 compilation is named @var{name}.
8146 This macro need not be defined if the standard form of output
8147 for DBX debugging information is appropriate.
8149 @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END
8150 @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
8151 A C statement to output DBX debugging information at the end of
8152 compilation of the main source file @var{name}.
8154 If you don't define this macro, nothing special is output at the end
8155 of compilation, which is correct for most machines.
8157 @findex DBX_OUTPUT_SOURCE_FILENAME
8158 @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
8159 A C statement to output DBX debugging information to the stdio stream
8160 @var{stream} which indicates that file @var{name} is the current source
8161 file. This output is generated each time input shifts to a different
8162 source file as a result of @samp{#include}, the end of an included file,
8163 or a @samp{#line} command.
8165 This macro need not be defined if the standard form of output
8166 for DBX debugging information is appropriate.
8171 @subsection Macros for SDB and DWARF Output
8173 @c prevent bad page break with this line
8174 Here are macros for SDB and DWARF output.
8177 @findex SDB_DEBUGGING_INFO
8178 @item SDB_DEBUGGING_INFO
8179 Define this macro if GCC should produce COFF-style debugging output
8180 for SDB in response to the @option{-g} option.
8182 @findex DWARF_DEBUGGING_INFO
8183 @item DWARF_DEBUGGING_INFO
8184 Define this macro if GCC should produce dwarf format debugging output
8185 in response to the @option{-g} option.
8187 @findex DWARF2_DEBUGGING_INFO
8188 @item DWARF2_DEBUGGING_INFO
8189 Define this macro if GCC should produce dwarf version 2 format
8190 debugging output in response to the @option{-g} option.
8192 To support optional call frame debugging information, you must also
8193 define @code{INCOMING_RETURN_ADDR_RTX} and either set
8194 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
8195 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
8196 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
8198 @findex DWARF2_FRAME_INFO
8199 @item DWARF2_FRAME_INFO
8200 Define this macro to a nonzero value if GCC should always output
8201 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
8202 (@pxref{Exception Region Output} is nonzero, GCC will output this
8203 information not matter how you define @code{DWARF2_FRAME_INFO}.
8205 @findex LINKER_DOES_NOT_WORK_WITH_DWARF2
8206 @item LINKER_DOES_NOT_WORK_WITH_DWARF2
8207 Define this macro if the linker does not work with Dwarf version 2.
8208 Normally, if the user specifies only @option{-ggdb} GCC will use Dwarf
8209 version 2 if available; this macro disables this. See the description
8210 of the @code{PREFERRED_DEBUGGING_TYPE} macro for more details.
8212 @findex DWARF2_GENERATE_TEXT_SECTION_LABEL
8213 @item DWARF2_GENERATE_TEXT_SECTION_LABEL
8214 By default, the Dwarf 2 debugging information generator will generate a
8215 label to mark the beginning of the text section. If it is better simply
8216 to use the name of the text section itself, rather than an explicit label,
8217 to indicate the beginning of the text section, define this macro to zero.
8219 @findex DWARF2_ASM_LINE_DEBUG_INFO
8220 @item DWARF2_ASM_LINE_DEBUG_INFO
8221 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8222 line debug info sections. This will result in much more compact line number
8223 tables, and hence is desirable if it works.
8225 @findex ASM_OUTPUT_DWARF_DELTA
8226 @item ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8227 A C statement to issue assembly directives that create a difference
8228 between the two given labels, using an integer of the given size.
8230 @findex ASM_OUTPUT_DWARF_OFFSET
8231 @item ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label})
8232 A C statement to issue assembly directives that create a
8233 section-relative reference to the given label, using an integer of the
8236 @findex ASM_OUTPUT_DWARF_PCREL
8237 @item ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8238 A C statement to issue assembly directives that create a self-relative
8239 reference to the given label, using an integer of the given size.
8241 @findex PUT_SDB_@dots{}
8242 @item PUT_SDB_@dots{}
8243 Define these macros to override the assembler syntax for the special
8244 SDB assembler directives. See @file{sdbout.c} for a list of these
8245 macros and their arguments. If the standard syntax is used, you need
8246 not define them yourself.
8250 Some assemblers do not support a semicolon as a delimiter, even between
8251 SDB assembler directives. In that case, define this macro to be the
8252 delimiter to use (usually @samp{\n}). It is not necessary to define
8253 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8256 @findex SDB_GENERATE_FAKE
8257 @item SDB_GENERATE_FAKE
8258 Define this macro to override the usual method of constructing a dummy
8259 name for anonymous structure and union types. See @file{sdbout.c} for
8262 @findex SDB_ALLOW_UNKNOWN_REFERENCES
8263 @item SDB_ALLOW_UNKNOWN_REFERENCES
8264 Define this macro to allow references to unknown structure,
8265 union, or enumeration tags to be emitted. Standard COFF does not
8266 allow handling of unknown references, MIPS ECOFF has support for
8269 @findex SDB_ALLOW_FORWARD_REFERENCES
8270 @item SDB_ALLOW_FORWARD_REFERENCES
8271 Define this macro to allow references to structure, union, or
8272 enumeration tags that have not yet been seen to be handled. Some
8273 assemblers choke if forward tags are used, while some require it.
8278 @subsection Macros for VMS Debug Format
8280 @c prevent bad page break with this line
8281 Here are macros for VMS debug format.
8284 @findex VMS_DEBUGGING_INFO
8285 @item VMS_DEBUGGING_INFO
8286 Define this macro if GCC should produce debugging output for VMS
8287 in response to the @option{-g} option. The default behavior for VMS
8288 is to generate minimal debug info for a traceback in the absence of
8289 @option{-g} unless explicitly overridden with @option{-g0}. This
8290 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8291 @code{OVERRIDE_OPTIONS}.
8294 @node Floating Point
8295 @section Cross Compilation and Floating Point
8296 @cindex cross compilation and floating point
8297 @cindex floating point and cross compilation
8299 While all modern machines use twos-complement representation for integers,
8300 there are a variety of representations for floating point numbers. This
8301 means that in a cross-compiler the representation of floating point numbers
8302 in the compiled program may be different from that used in the machine
8303 doing the compilation.
8305 Because different representation systems may offer different amounts of
8306 range and precision, all floating point constants must be represented in
8307 the target machine's format. Therefore, the cross compiler cannot
8308 safely use the host machine's floating point arithmetic; it must emulate
8309 the target's arithmetic. To ensure consistency, GCC always uses
8310 emulation to work with floating point values, even when the host and
8311 target floating point formats are identical.
8313 The following macros are provided by @file{real.h} for the compiler to
8314 use. All parts of the compiler which generate or optimize
8315 floating-point calculations must use these macros. They may evaluate
8316 their operands more than once, so operands must not have side effects.
8318 @defmac REAL_VALUE_TYPE
8319 The C data type to be used to hold a floating point value in the target
8320 machine's format. Typically this is a @code{struct} containing an
8321 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8325 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8326 Compares for equality the two values, @var{x} and @var{y}. If the target
8327 floating point format supports negative zeroes and/or NaNs,
8328 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8329 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8332 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8333 Tests whether @var{x} is less than @var{y}.
8336 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8337 Truncates @var{x} to a signed integer, rounding toward zero.
8340 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8341 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8342 @var{x} is negative, returns zero.
8345 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8346 Converts @var{string} into a floating point number in the target machine's
8347 representation for mode @var{mode}. This routine can handle both
8348 decimal and hexadecimal floating point constants, using the syntax
8349 defined by the C language for both.
8352 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8353 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8356 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8357 Determines whether @var{x} represents infinity (positive or negative).
8360 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8361 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8364 @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})
8365 Calculates an arithmetic operation on the two floating point values
8366 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8369 The operation to be performed is specified by @var{code}. Only the
8370 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8371 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8373 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8374 target's floating point format cannot represent infinity, it will call
8375 @code{abort}. Callers should check for this situation first, using
8376 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8379 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8380 Returns the negative of the floating point value @var{x}.
8383 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8384 Returns the absolute value of @var{x}.
8387 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8388 Truncates the floating point value @var{x} to fit in @var{mode}. The
8389 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8390 appropriate bit pattern to be output asa floating constant whose
8391 precision accords with mode @var{mode}.
8394 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8395 Converts a floating point value @var{x} into a double-precision integer
8396 which is then stored into @var{low} and @var{high}. If the value is not
8397 integral, it is truncated.
8400 @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})
8401 @findex REAL_VALUE_FROM_INT
8402 Converts a double-precision integer found in @var{low} and @var{high},
8403 into a floating point value which is then stored into @var{x}. The
8404 value is truncated to fit in mode @var{mode}.
8407 @node Mode Switching
8408 @section Mode Switching Instructions
8409 @cindex mode switching
8410 The following macros control mode switching optimizations:
8413 @findex OPTIMIZE_MODE_SWITCHING
8414 @item OPTIMIZE_MODE_SWITCHING (@var{entity})
8415 Define this macro if the port needs extra instructions inserted for mode
8416 switching in an optimizing compilation.
8418 For an example, the SH4 can perform both single and double precision
8419 floating point operations, but to perform a single precision operation,
8420 the FPSCR PR bit has to be cleared, while for a double precision
8421 operation, this bit has to be set. Changing the PR bit requires a general
8422 purpose register as a scratch register, hence these FPSCR sets have to
8423 be inserted before reload, i.e.@: you can't put this into instruction emitting
8424 or @code{MACHINE_DEPENDENT_REORG}.
8426 You can have multiple entities that are mode-switched, and select at run time
8427 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
8428 return nonzero for any @var{entity} that needs mode-switching.
8429 If you define this macro, you also have to define
8430 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
8431 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
8432 @code{NORMAL_MODE} is optional.
8434 @findex NUM_MODES_FOR_MODE_SWITCHING
8435 @item NUM_MODES_FOR_MODE_SWITCHING
8436 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
8437 initializer for an array of integers. Each initializer element
8438 N refers to an entity that needs mode switching, and specifies the number
8439 of different modes that might need to be set for this entity.
8440 The position of the initializer in the initializer - starting counting at
8441 zero - determines the integer that is used to refer to the mode-switched
8443 In macros that take mode arguments / yield a mode result, modes are
8444 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
8445 switch is needed / supplied.
8448 @item MODE_NEEDED (@var{entity}, @var{insn})
8449 @var{entity} is an integer specifying a mode-switched entity. If
8450 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
8451 return an integer value not larger than the corresponding element in
8452 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
8453 be switched into prior to the execution of @var{insn}.
8456 @item NORMAL_MODE (@var{entity})
8457 If this macro is defined, it is evaluated for every @var{entity} that needs
8458 mode switching. It should evaluate to an integer, which is a mode that
8459 @var{entity} is assumed to be switched to at function entry and exit.
8461 @findex MODE_PRIORITY_TO_MODE
8462 @item MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
8463 This macro specifies the order in which modes for @var{entity} are processed.
8464 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
8465 lowest. The value of the macro should be an integer designating a mode
8466 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
8467 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
8468 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
8470 @findex EMIT_MODE_SET
8471 @item EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
8472 Generate one or more insns to set @var{entity} to @var{mode}.
8473 @var{hard_reg_live} is the set of hard registers live at the point where
8474 the insn(s) are to be inserted.
8477 @node Target Attributes
8478 @section Defining target-specific uses of @code{__attribute__}
8479 @cindex target attributes
8480 @cindex machine attributes
8481 @cindex attributes, target-specific
8483 Target-specific attributes may be defined for functions, data and types.
8484 These are described using the following target hooks; they also need to
8485 be documented in @file{extend.texi}.
8487 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
8488 If defined, this target hook points to an array of @samp{struct
8489 attribute_spec} (defined in @file{tree.h}) specifying the machine
8490 specific attributes for this target and some of the restrictions on the
8491 entities to which these attributes are applied and the arguments they
8495 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8496 If defined, this target hook is a function which returns zero if the attributes on
8497 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8498 and two if they are nearly compatible (which causes a warning to be
8499 generated). If this is not defined, machine-specific attributes are
8500 supposed always to be compatible.
8503 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
8504 If defined, this target hook is a function which assigns default attributes to
8505 newly defined @var{type}.
8508 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8509 Define this target hook if the merging of type attributes needs special
8510 handling. If defined, the result is a list of the combined
8511 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
8512 that @code{comptypes} has already been called and returned 1. This
8513 function may call @code{merge_attributes} to handle machine-independent
8517 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
8518 Define this target hook if the merging of decl attributes needs special
8519 handling. If defined, the result is a list of the combined
8520 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
8521 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
8522 when this is needed are when one attribute overrides another, or when an
8523 attribute is nullified by a subsequent definition. This function may
8524 call @code{merge_attributes} to handle machine-independent merging.
8526 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
8527 If the only target-specific handling you require is @samp{dllimport} for
8528 Windows targets, you should define the macro
8529 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. This links in a function
8530 called @code{merge_dllimport_decl_attributes} which can then be defined
8531 as the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. This is done
8532 in @file{i386/cygwin.h} and @file{i386/i386.c}, for example.
8535 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
8536 Define this target hook if you want to be able to add attributes to a decl
8537 when it is being created. This is normally useful for back ends which
8538 wish to implement a pragma by using the attributes which correspond to
8539 the pragma's effect. The @var{node} argument is the decl which is being
8540 created. The @var{attr_ptr} argument is a pointer to the attribute list
8541 for this decl. The list itself should not be modified, since it may be
8542 shared with other decls, but attributes may be chained on the head of
8543 the list and @code{*@var{attr_ptr}} modified to point to the new
8544 attributes, or a copy of the list may be made if further changes are
8548 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
8550 This target hook returns @code{true} if it is ok to inline @var{fndecl}
8551 into the current function, despite its having target-specific
8552 attributes, @code{false} otherwise. By default, if a function has a
8553 target specific attribute attached to it, it will not be inlined.
8556 @node MIPS Coprocessors
8557 @section Defining coprocessor specifics for MIPS targets.
8558 @cindex MIPS coprocessor-definition macros
8560 The MIPS specification allows MIPS implementations to have as many as 4
8561 coprocessors, each with as many as 32 private registers. gcc supports
8562 accessing these registers and transferring values between the registers
8563 and memory using asm-ized variables. For example:
8566 register unsigned int cp0count asm ("c0r1");
8572 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
8573 names may be added as described below, or the default names may be
8574 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
8576 Coprocessor registers are assumed to be epilogue-used; sets to them will
8577 be preserved even if it does not appear that the register is used again
8578 later in the function.
8580 Another note: according to the MIPS spec, coprocessor 1 (if present) is
8581 the FPU. One accesses COP1 registers through standard mips
8582 floating-point support; they are not included in this mechanism.
8584 There is one macro used in defining the MIPS coprocessor interface which
8585 you may want to override in subtargets; it is described below.
8589 @item ALL_COP_ADDITIONAL_REGISTER_NAMES
8590 @findex ALL_COP_ADDITIONAL_REGISTER_NAMES
8591 A comma-separated list (with leading comma) of pairs describing the
8592 alternate names of coprocessor registers. The format of each entry should be
8594 @{ @var{alternatename}, @var{register_number}@}
8601 @section Miscellaneous Parameters
8602 @cindex parameters, miscellaneous
8604 @c prevent bad page break with this line
8605 Here are several miscellaneous parameters.
8608 @item PREDICATE_CODES
8609 @findex PREDICATE_CODES
8610 Define this if you have defined special-purpose predicates in the file
8611 @file{@var{machine}.c}. This macro is called within an initializer of an
8612 array of structures. The first field in the structure is the name of a
8613 predicate and the second field is an array of rtl codes. For each
8614 predicate, list all rtl codes that can be in expressions matched by the
8615 predicate. The list should have a trailing comma. Here is an example
8616 of two entries in the list for a typical RISC machine:
8619 #define PREDICATE_CODES \
8620 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
8621 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
8624 Defining this macro does not affect the generated code (however,
8625 incorrect definitions that omit an rtl code that may be matched by the
8626 predicate can cause the compiler to malfunction). Instead, it allows
8627 the table built by @file{genrecog} to be more compact and efficient,
8628 thus speeding up the compiler. The most important predicates to include
8629 in the list specified by this macro are those used in the most insn
8632 For each predicate function named in @code{PREDICATE_CODES}, a
8633 declaration will be generated in @file{insn-codes.h}.
8635 @item SPECIAL_MODE_PREDICATES
8636 @findex SPECIAL_MODE_PREDICATES
8637 Define this if you have special predicates that know special things
8638 about modes. Genrecog will warn about certain forms of
8639 @code{match_operand} without a mode; if the operand predicate is
8640 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
8643 Here is an example from the IA-32 port (@code{ext_register_operand}
8644 specially checks for @code{HImode} or @code{SImode} in preparation
8645 for a byte extraction from @code{%ah} etc.).
8648 #define SPECIAL_MODE_PREDICATES \
8649 "ext_register_operand",
8652 @findex CASE_VECTOR_MODE
8653 @item CASE_VECTOR_MODE
8654 An alias for a machine mode name. This is the machine mode that
8655 elements of a jump-table should have.
8657 @findex CASE_VECTOR_SHORTEN_MODE
8658 @item CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
8659 Optional: return the preferred mode for an @code{addr_diff_vec}
8660 when the minimum and maximum offset are known. If you define this,
8661 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
8662 To make this work, you also have to define @code{INSN_ALIGN} and
8663 make the alignment for @code{addr_diff_vec} explicit.
8664 The @var{body} argument is provided so that the offset_unsigned and scale
8665 flags can be updated.
8667 @findex CASE_VECTOR_PC_RELATIVE
8668 @item CASE_VECTOR_PC_RELATIVE
8669 Define this macro to be a C expression to indicate when jump-tables
8670 should contain relative addresses. If jump-tables never contain
8671 relative addresses, then you need not define this macro.
8673 @findex CASE_DROPS_THROUGH
8674 @item CASE_DROPS_THROUGH
8675 Define this if control falls through a @code{case} insn when the index
8676 value is out of range. This means the specified default-label is
8677 actually ignored by the @code{case} insn proper.
8679 @findex CASE_VALUES_THRESHOLD
8680 @item CASE_VALUES_THRESHOLD
8681 Define this to be the smallest number of different values for which it
8682 is best to use a jump-table instead of a tree of conditional branches.
8683 The default is four for machines with a @code{casesi} instruction and
8684 five otherwise. This is best for most machines.
8686 @findex CASE_USE_BIT_TESTS
8687 @item CASE_USE_BIT_TESTS
8688 Define this macro to be a C expression to indicate whether C switch
8689 statements may be implemented by a sequence of bit tests. This is
8690 advantageous on processors that can efficiently implement left shift
8691 of 1 by the number of bits held in a register, but inappropriate on
8692 targets that would require a loop. By default, this macro returns
8693 @code{true} if the target defines an @code{ashlsi3} pattern, and
8694 @code{false} otherwise.
8696 @findex WORD_REGISTER_OPERATIONS
8697 @item WORD_REGISTER_OPERATIONS
8698 Define this macro if operations between registers with integral mode
8699 smaller than a word are always performed on the entire register.
8700 Most RISC machines have this property and most CISC machines do not.
8702 @findex LOAD_EXTEND_OP
8703 @item LOAD_EXTEND_OP (@var{mode})
8704 Define this macro to be a C expression indicating when insns that read
8705 memory in @var{mode}, an integral mode narrower than a word, set the
8706 bits outside of @var{mode} to be either the sign-extension or the
8707 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
8708 of @var{mode} for which the
8709 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
8710 @code{NIL} for other modes.
8712 This macro is not called with @var{mode} non-integral or with a width
8713 greater than or equal to @code{BITS_PER_WORD}, so you may return any
8714 value in this case. Do not define this macro if it would always return
8715 @code{NIL}. On machines where this macro is defined, you will normally
8716 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
8718 @findex SHORT_IMMEDIATES_SIGN_EXTEND
8719 @item SHORT_IMMEDIATES_SIGN_EXTEND
8720 Define this macro if loading short immediate values into registers sign
8723 @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC
8724 @item FIXUNS_TRUNC_LIKE_FIX_TRUNC
8725 Define this macro if the same instructions that convert a floating
8726 point number to a signed fixed point number also convert validly to an
8731 The maximum number of bytes that a single instruction can move quickly
8732 between memory and registers or between two memory locations.
8734 @findex MAX_MOVE_MAX
8736 The maximum number of bytes that a single instruction can move quickly
8737 between memory and registers or between two memory locations. If this
8738 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
8739 constant value that is the largest value that @code{MOVE_MAX} can have
8742 @findex SHIFT_COUNT_TRUNCATED
8743 @item SHIFT_COUNT_TRUNCATED
8744 A C expression that is nonzero if on this machine the number of bits
8745 actually used for the count of a shift operation is equal to the number
8746 of bits needed to represent the size of the object being shifted. When
8747 this macro is nonzero, the compiler will assume that it is safe to omit
8748 a sign-extend, zero-extend, and certain bitwise `and' instructions that
8749 truncates the count of a shift operation. On machines that have
8750 instructions that act on bit-fields at variable positions, which may
8751 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
8752 also enables deletion of truncations of the values that serve as
8753 arguments to bit-field instructions.
8755 If both types of instructions truncate the count (for shifts) and
8756 position (for bit-field operations), or if no variable-position bit-field
8757 instructions exist, you should define this macro.
8759 However, on some machines, such as the 80386 and the 680x0, truncation
8760 only applies to shift operations and not the (real or pretended)
8761 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
8762 such machines. Instead, add patterns to the @file{md} file that include
8763 the implied truncation of the shift instructions.
8765 You need not define this macro if it would always have the value of zero.
8767 @findex TRULY_NOOP_TRUNCATION
8768 @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
8769 A C expression which is nonzero if on this machine it is safe to
8770 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
8771 bits (where @var{outprec} is smaller than @var{inprec}) by merely
8772 operating on it as if it had only @var{outprec} bits.
8774 On many machines, this expression can be 1.
8776 @c rearranged this, removed the phrase "it is reported that". this was
8777 @c to fix an overfull hbox. --mew 10feb93
8778 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
8779 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
8780 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
8781 such cases may improve things.
8783 @findex STORE_FLAG_VALUE
8784 @item STORE_FLAG_VALUE
8785 A C expression describing the value returned by a comparison operator
8786 with an integral mode and stored by a store-flag instruction
8787 (@samp{s@var{cond}}) when the condition is true. This description must
8788 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
8789 comparison operators whose results have a @code{MODE_INT} mode.
8791 A value of 1 or @minus{}1 means that the instruction implementing the
8792 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
8793 and 0 when the comparison is false. Otherwise, the value indicates
8794 which bits of the result are guaranteed to be 1 when the comparison is
8795 true. This value is interpreted in the mode of the comparison
8796 operation, which is given by the mode of the first operand in the
8797 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
8798 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
8801 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
8802 generate code that depends only on the specified bits. It can also
8803 replace comparison operators with equivalent operations if they cause
8804 the required bits to be set, even if the remaining bits are undefined.
8805 For example, on a machine whose comparison operators return an
8806 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
8807 @samp{0x80000000}, saying that just the sign bit is relevant, the
8811 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
8818 (ashift:SI @var{x} (const_int @var{n}))
8822 where @var{n} is the appropriate shift count to move the bit being
8823 tested into the sign bit.
8825 There is no way to describe a machine that always sets the low-order bit
8826 for a true value, but does not guarantee the value of any other bits,
8827 but we do not know of any machine that has such an instruction. If you
8828 are trying to port GCC to such a machine, include an instruction to
8829 perform a logical-and of the result with 1 in the pattern for the
8830 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
8832 Often, a machine will have multiple instructions that obtain a value
8833 from a comparison (or the condition codes). Here are rules to guide the
8834 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
8839 Use the shortest sequence that yields a valid definition for
8840 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
8841 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
8842 comparison operators to do so because there may be opportunities to
8843 combine the normalization with other operations.
8846 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
8847 slightly preferred on machines with expensive jumps and 1 preferred on
8851 As a second choice, choose a value of @samp{0x80000001} if instructions
8852 exist that set both the sign and low-order bits but do not define the
8856 Otherwise, use a value of @samp{0x80000000}.
8859 Many machines can produce both the value chosen for
8860 @code{STORE_FLAG_VALUE} and its negation in the same number of
8861 instructions. On those machines, you should also define a pattern for
8862 those cases, e.g., one matching
8865 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
8868 Some machines can also perform @code{and} or @code{plus} operations on
8869 condition code values with less instructions than the corresponding
8870 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
8871 machines, define the appropriate patterns. Use the names @code{incscc}
8872 and @code{decscc}, respectively, for the patterns which perform
8873 @code{plus} or @code{minus} operations on condition code values. See
8874 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
8875 find such instruction sequences on other machines.
8877 You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
8880 @findex FLOAT_STORE_FLAG_VALUE
8881 @item FLOAT_STORE_FLAG_VALUE (@var{mode})
8882 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
8883 returned when comparison operators with floating-point results are true.
8884 Define this macro on machine that have comparison operations that return
8885 floating-point values. If there are no such operations, do not define
8888 @findex CLZ_DEFINED_VALUE_AT_ZERO
8889 @findex CTZ_DEFINED_VALUE_AT_ZERO
8890 @item CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
8891 @itemx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
8892 A C expression that evaluates to true if the architecture defines a value
8893 for @code{clz} or @code{ctz} with a zero operand. If so, @var{value}
8894 should be set to this value. If this macro is not defined, the value of
8895 @code{clz} or @code{ctz} is assumed to be undefined.
8897 This macro must be defined if the target's expansion for @code{ffs}
8898 relies on a particular value to get correct results. Otherwise it
8899 is not necessary, though it may be used to optimize some corner cases.
8901 Note that regardless of this macro the ``definedness'' of @code{clz}
8902 and @code{ctz} at zero do @emph{not} extend to the builtin functions
8903 visible to the user. Thus one may be free to adjust the value at will
8904 to match the target expansion of these operations without fear of
8909 An alias for the machine mode for pointers. On most machines, define
8910 this to be the integer mode corresponding to the width of a hardware
8911 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
8912 On some machines you must define this to be one of the partial integer
8913 modes, such as @code{PSImode}.
8915 The width of @code{Pmode} must be at least as large as the value of
8916 @code{POINTER_SIZE}. If it is not equal, you must define the macro
8917 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
8920 @findex FUNCTION_MODE
8922 An alias for the machine mode used for memory references to functions
8923 being called, in @code{call} RTL expressions. On most machines this
8924 should be @code{QImode}.
8926 @findex INTEGRATE_THRESHOLD
8927 @item INTEGRATE_THRESHOLD (@var{decl})
8928 A C expression for the maximum number of instructions above which the
8929 function @var{decl} should not be inlined. @var{decl} is a
8930 @code{FUNCTION_DECL} node.
8932 The default definition of this macro is 64 plus 8 times the number of
8933 arguments that the function accepts. Some people think a larger
8934 threshold should be used on RISC machines.
8936 @findex STDC_0_IN_SYSTEM_HEADERS
8937 @item STDC_0_IN_SYSTEM_HEADERS
8938 In normal operation, the preprocessor expands @code{__STDC__} to the
8939 constant 1, to signify that GCC conforms to ISO Standard C@. On some
8940 hosts, like Solaris, the system compiler uses a different convention,
8941 where @code{__STDC__} is normally 0, but is 1 if the user specifies
8942 strict conformance to the C Standard.
8944 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
8945 convention when processing system header files, but when processing user
8946 files @code{__STDC__} will always expand to 1.
8948 @findex NO_IMPLICIT_EXTERN_C
8949 @item NO_IMPLICIT_EXTERN_C
8950 Define this macro if the system header files support C++ as well as C@.
8951 This macro inhibits the usual method of using system header files in
8952 C++, which is to pretend that the file's contents are enclosed in
8953 @samp{extern "C" @{@dots{}@}}.
8955 @findex HANDLE_PRAGMA
8956 @item HANDLE_PRAGMA (@var{getc}, @var{ungetc}, @var{name})
8957 This macro is no longer supported. You must use
8958 @code{REGISTER_TARGET_PRAGMAS} instead.
8960 @findex REGISTER_TARGET_PRAGMAS
8963 @item REGISTER_TARGET_PRAGMAS ()
8964 Define this macro if you want to implement any target-specific pragmas.
8965 If defined, it is a C expression which makes a series of calls to
8966 @code{c_register_pragma} for each pragma. The macro may also do any
8967 setup required for the pragmas.
8969 The primary reason to define this macro is to provide compatibility with
8970 other compilers for the same target. In general, we discourage
8971 definition of target-specific pragmas for GCC@.
8973 If the pragma can be implemented by attributes then you should consider
8974 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
8976 Preprocessor macros that appear on pragma lines are not expanded. All
8977 @samp{#pragma} directives that do not match any registered pragma are
8978 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
8980 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
8982 Each call to @code{c_register_pragma} establishes one pragma. The
8983 @var{callback} routine will be called when the preprocessor encounters a
8987 #pragma [@var{space}] @var{name} @dots{}
8990 @var{space} is the case-sensitive namespace of the pragma, or
8991 @code{NULL} to put the pragma in the global namespace. The callback
8992 routine receives @var{pfile} as its first argument, which can be passed
8993 on to cpplib's functions if necessary. You can lex tokens after the
8994 @var{name} by calling @code{c_lex}. Tokens that are not read by the
8995 callback will be silently ignored. The end of the line is indicated by
8996 a token of type @code{CPP_EOF}.
8998 For an example use of this routine, see @file{c4x.h} and the callback
8999 routines defined in @file{c4x-c.c}.
9001 Note that the use of @code{c_lex} is specific to the C and C++
9002 compilers. It will not work in the Java or Fortran compilers, or any
9003 other language compilers for that matter. Thus if @code{c_lex} is going
9004 to be called from target-specific code, it must only be done so when
9005 building the C and C++ compilers. This can be done by defining the
9006 variables @code{c_target_objs} and @code{cxx_target_objs} in the
9007 target entry in the @file{config.gcc} file. These variables should name
9008 the target-specific, language-specific object file which contains the
9009 code that uses @code{c_lex}. Note it will also be necessary to add a
9010 rule to the makefile fragment pointed to by @code{tmake_file} that shows
9011 how to build this object file.
9014 @findex HANDLE_SYSV_PRAGMA
9017 @item HANDLE_SYSV_PRAGMA
9018 Define this macro (to a value of 1) if you want the System V style
9019 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
9020 [=<value>]} to be supported by gcc.
9022 The pack pragma specifies the maximum alignment (in bytes) of fields
9023 within a structure, in much the same way as the @samp{__aligned__} and
9024 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
9025 the behavior to the default.
9027 A subtlety for Microsoft Visual C/C++ style bit-field packing
9028 (e.g. -mms-bitfields) for targets that support it:
9029 When a bit-field is inserted into a packed record, the whole size
9030 of the underlying type is used by one or more same-size adjacent
9031 bit-fields (that is, if its long:3, 32 bits is used in the record,
9032 and any additional adjacent long bit-fields are packed into the same
9033 chunk of 32 bits. However, if the size changes, a new field of that
9036 If both MS bit-fields and @samp{__attribute__((packed))} are used,
9037 the latter will take precedence. If @samp{__attribute__((packed))} is
9038 used on a single field when MS bit-fields are in use, it will take
9039 precedence for that field, but the alignment of the rest of the structure
9040 may affect its placement.
9042 The weak pragma only works if @code{SUPPORTS_WEAK} and
9043 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
9044 of specifically named weak labels, optionally with a value.
9046 @findex HANDLE_PRAGMA_PACK_PUSH_POP
9049 @item HANDLE_PRAGMA_PACK_PUSH_POP
9050 Define this macro (to a value of 1) if you want to support the Win32
9051 style pragmas @samp{#pragma pack(push,@var{n})} and @samp{#pragma
9052 pack(pop)}. The @samp{pack(push,@var{n})} pragma specifies the maximum alignment
9053 (in bytes) of fields within a structure, in much the same way as the
9054 @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
9055 pack value of zero resets the behavior to the default. Successive
9056 invocations of this pragma cause the previous values to be stacked, so
9057 that invocations of @samp{#pragma pack(pop)} will return to the previous
9060 @findex DOLLARS_IN_IDENTIFIERS
9061 @item DOLLARS_IN_IDENTIFIERS
9062 Define this macro to control use of the character @samp{$} in identifier
9063 names. 0 means @samp{$} is not allowed by default; 1 means it is allowed.
9064 1 is the default; there is no need to define this macro in that case.
9065 This macro controls the compiler proper; it does not affect the preprocessor.
9067 @findex NO_DOLLAR_IN_LABEL
9068 @item NO_DOLLAR_IN_LABEL
9069 Define this macro if the assembler does not accept the character
9070 @samp{$} in label names. By default constructors and destructors in
9071 G++ have @samp{$} in the identifiers. If this macro is defined,
9072 @samp{.} is used instead.
9074 @findex NO_DOT_IN_LABEL
9075 @item NO_DOT_IN_LABEL
9076 Define this macro if the assembler does not accept the character
9077 @samp{.} in label names. By default constructors and destructors in G++
9078 have names that use @samp{.}. If this macro is defined, these names
9079 are rewritten to avoid @samp{.}.
9081 @findex DEFAULT_MAIN_RETURN
9082 @item DEFAULT_MAIN_RETURN
9083 Define this macro if the target system expects every program's @code{main}
9084 function to return a standard ``success'' value by default (if no other
9085 value is explicitly returned).
9087 The definition should be a C statement (sans semicolon) to generate the
9088 appropriate rtl instructions. It is used only when compiling the end of
9093 Define this if the target system lacks the function @code{atexit}
9094 from the ISO C standard. If this macro is defined, a default definition
9095 will be provided to support C++. If @code{ON_EXIT} is not defined,
9096 a default @code{exit} function will also be provided.
9100 Define this macro if the target has another way to implement atexit
9101 functionality without replacing @code{exit}. For instance, SunOS 4 has
9102 a similar @code{on_exit} library function.
9104 The definition should be a functional macro which can be used just like
9105 the @code{atexit} function.
9109 Define this if your @code{exit} function needs to do something
9110 besides calling an external function @code{_cleanup} before
9111 terminating with @code{_exit}. The @code{EXIT_BODY} macro is
9112 only needed if @code{NEED_ATEXIT} is defined and @code{ON_EXIT} is not
9115 @findex INSN_SETS_ARE_DELAYED
9116 @item INSN_SETS_ARE_DELAYED (@var{insn})
9117 Define this macro as a C expression that is nonzero if it is safe for the
9118 delay slot scheduler to place instructions in the delay slot of @var{insn},
9119 even if they appear to use a resource set or clobbered in @var{insn}.
9120 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
9121 every @code{call_insn} has this behavior. On machines where some @code{insn}
9122 or @code{jump_insn} is really a function call and hence has this behavior,
9123 you should define this macro.
9125 You need not define this macro if it would always return zero.
9127 @findex INSN_REFERENCES_ARE_DELAYED
9128 @item INSN_REFERENCES_ARE_DELAYED (@var{insn})
9129 Define this macro as a C expression that is nonzero if it is safe for the
9130 delay slot scheduler to place instructions in the delay slot of @var{insn},
9131 even if they appear to set or clobber a resource referenced in @var{insn}.
9132 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
9133 some @code{insn} or @code{jump_insn} is really a function call and its operands
9134 are registers whose use is actually in the subroutine it calls, you should
9135 define this macro. Doing so allows the delay slot scheduler to move
9136 instructions which copy arguments into the argument registers into the delay
9139 You need not define this macro if it would always return zero.
9141 @findex MACHINE_DEPENDENT_REORG
9142 @item MACHINE_DEPENDENT_REORG (@var{insn})
9143 In rare cases, correct code generation requires extra machine
9144 dependent processing between the second jump optimization pass and
9145 delayed branch scheduling. On those machines, define this macro as a C
9146 statement to act on the code starting at @var{insn}.
9148 @findex MULTIPLE_SYMBOL_SPACES
9149 @item MULTIPLE_SYMBOL_SPACES
9150 Define this macro if in some cases global symbols from one translation
9151 unit may not be bound to undefined symbols in another translation unit
9152 without user intervention. For instance, under Microsoft Windows
9153 symbols must be explicitly imported from shared libraries (DLLs).
9155 @findex MD_ASM_CLOBBERS
9156 @item MD_ASM_CLOBBERS (@var{clobbers})
9157 A C statement that adds to @var{clobbers} @code{STRING_CST} trees for
9158 any hard regs the port wishes to automatically clobber for all asms.
9160 @findex MAX_INTEGER_COMPUTATION_MODE
9161 @item MAX_INTEGER_COMPUTATION_MODE
9162 Define this to the largest integer machine mode which can be used for
9163 operations other than load, store and copy operations.
9165 You need only define this macro if the target holds values larger than
9166 @code{word_mode} in general purpose registers. Most targets should not define
9169 @findex MATH_LIBRARY
9171 Define this macro as a C string constant for the linker argument to link
9172 in the system math library, or @samp{""} if the target does not have a
9173 separate math library.
9175 You need only define this macro if the default of @samp{"-lm"} is wrong.
9177 @findex LIBRARY_PATH_ENV
9178 @item LIBRARY_PATH_ENV
9179 Define this macro as a C string constant for the environment variable that
9180 specifies where the linker should look for libraries.
9182 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
9185 @findex TARGET_HAS_F_SETLKW
9186 @item TARGET_HAS_F_SETLKW
9187 Define this macro if the target supports file locking with fcntl / F_SETLKW@.
9188 Note that this functionality is part of POSIX@.
9189 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
9190 to use file locking when exiting a program, which avoids race conditions
9191 if the program has forked.
9193 @findex MAX_CONDITIONAL_EXECUTE
9194 @item MAX_CONDITIONAL_EXECUTE
9196 A C expression for the maximum number of instructions to execute via
9197 conditional execution instructions instead of a branch. A value of
9198 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
9199 1 if it does use cc0.
9201 @findex IFCVT_MODIFY_TESTS
9202 @item IFCVT_MODIFY_TESTS(@var{ce_info}, @var{true_expr}, @var{false_expr})
9203 Used if the target needs to perform machine-dependent modifications on the
9204 conditionals used for turning basic blocks into conditionally executed code.
9205 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
9206 contains information about the currently processed blocks. @var{true_expr}
9207 and @var{false_expr} are the tests that are used for converting the
9208 then-block and the else-block, respectively. Set either @var{true_expr} or
9209 @var{false_expr} to a null pointer if the tests cannot be converted.
9211 @findex IFCVT_MODIFY_MULTIPLE_TESTS
9212 @item IFCVT_MODIFY_MULTIPLE_TESTS(@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
9213 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
9214 if-statements into conditions combined by @code{and} and @code{or} operations.
9215 @var{bb} contains the basic block that contains the test that is currently
9216 being processed and about to be turned into a condition.
9218 @findex IFCVT_MODIFY_INSN
9219 @item IFCVT_MODIFY_INSN(@var{ce_info}, @var{pattern}, @var{insn})
9220 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
9221 be converted to conditional execution format. @var{ce_info} points to
9222 a data structure, @code{struct ce_if_block}, which contains information
9223 about the currently processed blocks.
9225 @findex IFCVT_MODIFY_FINAL
9226 @item IFCVT_MODIFY_FINAL(@var{ce_info})
9227 A C expression to perform any final machine dependent modifications in
9228 converting code to conditional execution. The involved basic blocks
9229 can be found in the @code{struct ce_if_block} structure that is pointed
9230 to by @var{ce_info}.
9232 @findex IFCVT_MODIFY_CANCEL
9233 @item IFCVT_MODIFY_CANCEL(@var{ce_info})
9234 A C expression to cancel any machine dependent modifications in
9235 converting code to conditional execution. The involved basic blocks
9236 can be found in the @code{struct ce_if_block} structure that is pointed
9237 to by @var{ce_info}.
9239 @findex IFCVT_INIT_EXTRA_FIELDS
9240 @item IFCVT_INIT_EXTRA_FIELDS(@var{ce_info})
9241 A C expression to initialize any extra fields in a @code{struct ce_if_block}
9242 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
9244 @findex IFCVT_EXTRA_FIELDS
9245 @item IFCVT_EXTRA_FIELDS
9246 If defined, it should expand to a set of field declarations that will be
9247 added to the @code{struct ce_if_block} structure. These should be initialized
9248 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
9252 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
9253 Define this hook if you have any machine-specific built-in functions
9254 that need to be defined. It should be a function that performs the
9257 Machine specific built-in functions can be useful to expand special machine
9258 instructions that would otherwise not normally be generated because
9259 they have no equivalent in the source language (for example, SIMD vector
9260 instructions or prefetch instructions).
9262 To create a built-in function, call the function @code{builtin_function}
9263 which is defined by the language front end. You can use any type nodes set
9264 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
9265 only language front ends that use those two functions will call
9266 @samp{TARGET_INIT_BUILTINS}.
9269 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
9271 Expand a call to a machine specific built-in function that was set up by
9272 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
9273 function call; the result should go to @var{target} if that is
9274 convenient, and have mode @var{mode} if that is convenient.
9275 @var{subtarget} may be used as the target for computing one of
9276 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
9277 ignored. This function should return the result of the call to the
9282 @findex MD_CAN_REDIRECT_BRANCH
9283 @item MD_CAN_REDIRECT_BRANCH(@var{branch1}, @var{branch2})
9285 Take a branch insn in @var{branch1} and another in @var{branch2}.
9286 Return true if redirecting @var{branch1} to the destination of
9287 @var{branch2} is possible.
9289 On some targets, branches may have a limited range. Optimizing the
9290 filling of delay slots can result in branches being redirected, and this
9291 may in turn cause a branch offset to overflow.
9293 @findex ALLOCATE_INITIAL_VALUE
9294 @item ALLOCATE_INITIAL_VALUE(@var{hard_reg})
9296 When the initial value of a hard register has been copied in a pseudo
9297 register, it is often not necessary to actually allocate another register
9298 to this pseudo register, because the original hard register or a stack slot
9299 it has been saved into can be used. @code{ALLOCATE_INITIAL_VALUE}, if
9300 defined, is called at the start of register allocation once for each
9301 hard register that had its initial value copied by using
9302 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
9303 Possible values are @code{NULL_RTX}, if you don't want
9304 to do any special allocation, a @code{REG} rtx---that would typically be
9305 the hard register itself, if it is known not to be clobbered---or a
9307 If you are returning a @code{MEM}, this is only a hint for the allocator;
9308 it might decide to use another register anyways.
9309 You may use @code{current_function_leaf_function} in the definition of the
9310 macro, functions that use @code{REG_N_SETS}, to determine if the hard
9311 register in question will not be clobbered.
9313 @findex TARGET_OBJECT_SUFFIX
9314 @item TARGET_OBJECT_SUFFIX
9315 Define this macro to be a C string representing the suffix for object
9316 files on your target machine. If you do not define this macro, GCC will
9317 use @samp{.o} as the suffix for object files.
9319 @findex TARGET_EXECUTABLE_SUFFIX
9320 @item TARGET_EXECUTABLE_SUFFIX
9321 Define this macro to be a C string representing the suffix to be
9322 automatically added to executable files on your target machine. If you
9323 do not define this macro, GCC will use the null string as the suffix for
9326 @findex COLLECT_EXPORT_LIST
9327 @item COLLECT_EXPORT_LIST
9328 If defined, @code{collect2} will scan the individual object files
9329 specified on its command line and create an export list for the linker.
9330 Define this macro for systems like AIX, where the linker discards
9331 object files that are not referenced from @code{main} and uses export
9334 @findex MODIFY_JNI_METHOD_CALL
9335 @item MODIFY_JNI_METHOD_CALL (@var{mdecl})
9336 Define this macro to a C expression representing a variant of the
9337 method call @var{mdecl}, if Java Native Interface (JNI) methods
9338 must be invoked differently from other methods on your target.
9339 For example, on 32-bit Windows, JNI methods must be invoked using
9340 the @code{stdcall} calling convention and this macro is then
9341 defined as this expression:
9344 build_type_attribute_variant (@var{mdecl},
9346 (get_identifier ("stdcall"),
9352 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
9353 This target hook returns @code{true} past the point in which new jump
9354 instructions could be created. On machines that require a register for
9355 every jump such as the SHmedia ISA of SH5, this point would typically be
9356 reload, so this target hook should be defined to a function such as:
9360 cannot_modify_jumps_past_reload_p ()
9362 return (reload_completed || reload_in_progress);