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 @} %@{mcall-aix: -D_CALL_AIX @} \
341 %@{!mcall-sysv: %@{!mcall-aix: %(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} defined in @file{c-common.c}. 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 fixed part
765 of the option name, the address of a variable, and a description string.
766 Non-empty description strings should be marked with @code{N_(@dots{})} for
767 @command{xgettext}. Please do not mark empty strings because the empty
768 string is reserved by GNU gettext. @code{gettext("")} returns the header entry
769 of the message catalog with meta information, not the empty string.
771 The variable, type @code{char *}, is set to the variable part of the
772 given option if the fixed part matches. The actual option name is made
773 by appending @samp{-m} to the specified name. Again, each option should
774 also be documented in @file{invoke.texi}.
776 Here is an example which defines @option{-mshort-data-@var{number}}. If the
777 given option is @option{-mshort-data-512}, the variable @code{m88k_short_data}
778 will be set to the string @code{"512"}.
781 extern char *m88k_short_data;
782 #define TARGET_OPTIONS \
783 @{ @{ "short-data-", &m88k_short_data, \
784 N_("Specify the size of the short data section") @} @}
787 @findex TARGET_VERSION
789 This macro is a C statement to print on @code{stderr} a string
790 describing the particular machine description choice. Every machine
791 description should define @code{TARGET_VERSION}. For example:
795 #define TARGET_VERSION \
796 fprintf (stderr, " (68k, Motorola syntax)");
798 #define TARGET_VERSION \
799 fprintf (stderr, " (68k, MIT syntax)");
803 @findex OVERRIDE_OPTIONS
804 @item OVERRIDE_OPTIONS
805 Sometimes certain combinations of command options do not make sense on
806 a particular target machine. You can define a macro
807 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
808 defined, is executed once just after all the command options have been
811 Don't use this macro to turn on various extra optimizations for
812 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
814 @findex OPTIMIZATION_OPTIONS
815 @item OPTIMIZATION_OPTIONS (@var{level}, @var{size})
816 Some machines may desire to change what optimizations are performed for
817 various optimization levels. This macro, if defined, is executed once
818 just after the optimization level is determined and before the remainder
819 of the command options have been parsed. Values set in this macro are
820 used as the default values for the other command line options.
822 @var{level} is the optimization level specified; 2 if @option{-O2} is
823 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
825 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
827 You should not use this macro to change options that are not
828 machine-specific. These should uniformly selected by the same
829 optimization level on all supported machines. Use this macro to enable
830 machine-specific optimizations.
832 @strong{Do not examine @code{write_symbols} in
833 this macro!} The debugging options are not supposed to alter the
836 @findex CAN_DEBUG_WITHOUT_FP
837 @item CAN_DEBUG_WITHOUT_FP
838 Define this macro if debugging can be performed even without a frame
839 pointer. If this macro is defined, GCC will turn on the
840 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
843 @node Per-Function Data
844 @section Defining data structures for per-function information.
845 @cindex per-function data
846 @cindex data structures
848 If the target needs to store information on a per-function basis, GCC
849 provides a macro and a couple of variables to allow this. Note, just
850 using statics to store the information is a bad idea, since GCC supports
851 nested functions, so you can be halfway through encoding one function
852 when another one comes along.
854 GCC defines a data structure called @code{struct function} which
855 contains all of the data specific to an individual function. This
856 structure contains a field called @code{machine} whose type is
857 @code{struct machine_function *}, which can be used by targets to point
858 to their own specific data.
860 If a target needs per-function specific data it should define the type
861 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
862 This macro should be used to initialize the function pointer
863 @code{init_machine_status}. This pointer is explained below.
865 One typical use of per-function, target specific data is to create an
866 RTX to hold the register containing the function's return address. This
867 RTX can then be used to implement the @code{__builtin_return_address}
868 function, for level 0.
870 Note---earlier implementations of GCC used a single data area to hold
871 all of the per-function information. Thus when processing of a nested
872 function began the old per-function data had to be pushed onto a
873 stack, and when the processing was finished, it had to be popped off the
874 stack. GCC used to provide function pointers called
875 @code{save_machine_status} and @code{restore_machine_status} to handle
876 the saving and restoring of the target specific information. Since the
877 single data area approach is no longer used, these pointers are no
880 The macro and function pointers are described below.
883 @findex INIT_EXPANDERS
885 Macro called to initialize any target specific information. This macro
886 is called once per function, before generation of any RTL has begun.
887 The intention of this macro is to allow the initialization of the
888 function pointers below.
890 @findex init_machine_status
891 @item init_machine_status
892 This is a @code{void (*)(struct function *)} function pointer. If this
893 pointer is non-@code{NULL} it will be called once per function, before function
894 compilation starts, in order to allow the target to perform any target
895 specific initialization of the @code{struct function} structure. It is
896 intended that this would be used to initialize the @code{machine} of
899 @code{struct machine_function} structures are expected to be freed by GC.
900 Generally, any memory that they reference must be allocated by using
901 @code{ggc_alloc}, including the structure itself.
906 @section Storage Layout
907 @cindex storage layout
909 Note that the definitions of the macros in this table which are sizes or
910 alignments measured in bits do not need to be constant. They can be C
911 expressions that refer to static variables, such as the @code{target_flags}.
912 @xref{Run-time Target}.
915 @findex BITS_BIG_ENDIAN
916 @item BITS_BIG_ENDIAN
917 Define this macro to have the value 1 if the most significant bit in a
918 byte has the lowest number; otherwise define it to have the value zero.
919 This means that bit-field instructions count from the most significant
920 bit. If the machine has no bit-field instructions, then this must still
921 be defined, but it doesn't matter which value it is defined to. This
922 macro need not be a constant.
924 This macro does not affect the way structure fields are packed into
925 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
927 @findex BYTES_BIG_ENDIAN
928 @item BYTES_BIG_ENDIAN
929 Define this macro to have the value 1 if the most significant byte in a
930 word has the lowest number. This macro need not be a constant.
932 @findex WORDS_BIG_ENDIAN
933 @item WORDS_BIG_ENDIAN
934 Define this macro to have the value 1 if, in a multiword object, the
935 most significant word has the lowest number. This applies to both
936 memory locations and registers; GCC fundamentally assumes that the
937 order of words in memory is the same as the order in registers. This
938 macro need not be a constant.
940 @findex LIBGCC2_WORDS_BIG_ENDIAN
941 @item LIBGCC2_WORDS_BIG_ENDIAN
942 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
943 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
944 used only when compiling @file{libgcc2.c}. Typically the value will be set
945 based on preprocessor defines.
947 @findex FLOAT_WORDS_BIG_ENDIAN
948 @item FLOAT_WORDS_BIG_ENDIAN
949 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
950 @code{TFmode} floating point numbers are stored in memory with the word
951 containing the sign bit at the lowest address; otherwise define it to
952 have the value 0. This macro need not be a constant.
954 You need not define this macro if the ordering is the same as for
957 @findex BITS_PER_UNIT
959 Define this macro to be the number of bits in an addressable storage
960 unit (byte). If you do not define this macro the default is 8.
962 @findex BITS_PER_WORD
964 Number of bits in a word. If you do not define this macro, the default
965 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
967 @findex MAX_BITS_PER_WORD
968 @item MAX_BITS_PER_WORD
969 Maximum number of bits in a word. If this is undefined, the default is
970 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
971 largest value that @code{BITS_PER_WORD} can have at run-time.
973 @findex UNITS_PER_WORD
975 Number of storage units in a word; normally 4.
977 @findex MIN_UNITS_PER_WORD
978 @item MIN_UNITS_PER_WORD
979 Minimum number of units in a word. If this is undefined, the default is
980 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
981 smallest value that @code{UNITS_PER_WORD} can have at run-time.
985 Width of a pointer, in bits. You must specify a value no wider than the
986 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
987 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
988 a value the default is @code{BITS_PER_WORD}.
990 @findex POINTERS_EXTEND_UNSIGNED
991 @item POINTERS_EXTEND_UNSIGNED
992 A C expression whose value is greater than zero if pointers that need to be
993 extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to
994 be zero-extended and zero if they are to be sign-extended. If the value
995 is less then zero then there must be an "ptr_extend" instruction that
996 extends a pointer from @code{POINTER_SIZE} to @code{Pmode}.
998 You need not define this macro if the @code{POINTER_SIZE} is equal
999 to the width of @code{Pmode}.
1001 @findex PROMOTE_MODE
1002 @item PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
1003 A macro to update @var{m} and @var{unsignedp} when an object whose type
1004 is @var{type} and which has the specified mode and signedness is to be
1005 stored in a register. This macro is only called when @var{type} is a
1008 On most RISC machines, which only have operations that operate on a full
1009 register, define this macro to set @var{m} to @code{word_mode} if
1010 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
1011 cases, only integer modes should be widened because wider-precision
1012 floating-point operations are usually more expensive than their narrower
1015 For most machines, the macro definition does not change @var{unsignedp}.
1016 However, some machines, have instructions that preferentially handle
1017 either signed or unsigned quantities of certain modes. For example, on
1018 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
1019 sign-extend the result to 64 bits. On such machines, set
1020 @var{unsignedp} according to which kind of extension is more efficient.
1022 Do not define this macro if it would never modify @var{m}.
1024 @findex PROMOTE_FUNCTION_ARGS
1025 @item PROMOTE_FUNCTION_ARGS
1026 Define this macro if the promotion described by @code{PROMOTE_MODE}
1027 should also be done for outgoing function arguments.
1029 @findex PROMOTE_FUNCTION_RETURN
1030 @item PROMOTE_FUNCTION_RETURN
1031 Define this macro if the promotion described by @code{PROMOTE_MODE}
1032 should also be done for the return value of functions.
1034 If this macro is defined, @code{FUNCTION_VALUE} must perform the same
1035 promotions done by @code{PROMOTE_MODE}.
1037 @findex PROMOTE_FOR_CALL_ONLY
1038 @item PROMOTE_FOR_CALL_ONLY
1039 Define this macro if the promotion described by @code{PROMOTE_MODE}
1040 should @emph{only} be performed for outgoing function arguments or
1041 function return values, as specified by @code{PROMOTE_FUNCTION_ARGS}
1042 and @code{PROMOTE_FUNCTION_RETURN}, respectively.
1044 @findex PARM_BOUNDARY
1046 Normal alignment required for function parameters on the stack, in
1047 bits. All stack parameters receive at least this much alignment
1048 regardless of data type. On most machines, this is the same as the
1051 @findex STACK_BOUNDARY
1052 @item STACK_BOUNDARY
1053 Define this macro to the minimum alignment enforced by hardware for the
1054 stack pointer on this machine. The definition is a C expression for the
1055 desired alignment (measured in bits). This value is used as a default
1056 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1057 this should be the same as @code{PARM_BOUNDARY}.
1059 @findex PREFERRED_STACK_BOUNDARY
1060 @item PREFERRED_STACK_BOUNDARY
1061 Define this macro if you wish to preserve a certain alignment for the
1062 stack pointer, greater than what the hardware enforces. The definition
1063 is a C expression for the desired alignment (measured in bits). This
1064 macro must evaluate to a value equal to or larger than
1065 @code{STACK_BOUNDARY}.
1067 @findex FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1068 @item FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN
1069 A C expression that evaluates true if @code{PREFERRED_STACK_BOUNDARY} is
1070 not guaranteed by the runtime and we should emit code to align the stack
1071 at the beginning of @code{main}.
1073 @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY}
1074 If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
1075 to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies
1076 a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may
1077 be momentarily unaligned while pushing arguments.
1079 @findex FUNCTION_BOUNDARY
1080 @item FUNCTION_BOUNDARY
1081 Alignment required for a function entry point, in bits.
1083 @findex BIGGEST_ALIGNMENT
1084 @item BIGGEST_ALIGNMENT
1085 Biggest alignment that any data type can require on this machine, in bits.
1087 @findex MINIMUM_ATOMIC_ALIGNMENT
1088 @item MINIMUM_ATOMIC_ALIGNMENT
1089 If defined, the smallest alignment, in bits, that can be given to an
1090 object that can be referenced in one operation, without disturbing any
1091 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1092 on machines that don't have byte or half-word store operations.
1094 @findex BIGGEST_FIELD_ALIGNMENT
1095 @item BIGGEST_FIELD_ALIGNMENT
1096 Biggest alignment that any structure or union field can require on this
1097 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1098 structure and union fields only, unless the field alignment has been set
1099 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1101 @findex ADJUST_FIELD_ALIGN
1102 @item ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1103 An expression for the alignment of a structure field @var{field} if the
1104 alignment computed in the usual way (including applying of
1105 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1106 alignment) is @var{computed}. It overrides alignment only if the
1107 field alignment has not been set by the
1108 @code{__attribute__ ((aligned (@var{n})))} construct.
1110 @findex MAX_OFILE_ALIGNMENT
1111 @item MAX_OFILE_ALIGNMENT
1112 Biggest alignment supported by the object file format of this machine.
1113 Use this macro to limit the alignment which can be specified using the
1114 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1115 the default value is @code{BIGGEST_ALIGNMENT}.
1117 @findex DATA_ALIGNMENT
1118 @item DATA_ALIGNMENT (@var{type}, @var{basic-align})
1119 If defined, a C expression to compute the alignment for a variable in
1120 the static store. @var{type} is the data type, and @var{basic-align} is
1121 the alignment that the object would ordinarily have. The value of this
1122 macro is used instead of that alignment to align the object.
1124 If this macro is not defined, then @var{basic-align} is used.
1127 One use of this macro is to increase alignment of medium-size data to
1128 make it all fit in fewer cache lines. Another is to cause character
1129 arrays to be word-aligned so that @code{strcpy} calls that copy
1130 constants to character arrays can be done inline.
1132 @findex CONSTANT_ALIGNMENT
1133 @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1134 If defined, a C expression to compute the alignment given to a constant
1135 that is being placed in memory. @var{constant} is the constant and
1136 @var{basic-align} is the alignment that the object would ordinarily
1137 have. The value of this macro is used instead of that alignment to
1140 If this macro is not defined, then @var{basic-align} is used.
1142 The typical use of this macro is to increase alignment for string
1143 constants to be word aligned so that @code{strcpy} calls that copy
1144 constants can be done inline.
1146 @findex LOCAL_ALIGNMENT
1147 @item LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1148 If defined, a C expression to compute the alignment for a variable in
1149 the local store. @var{type} is the data type, and @var{basic-align} is
1150 the alignment that the object would ordinarily have. The value of this
1151 macro is used instead of that alignment to align the object.
1153 If this macro is not defined, then @var{basic-align} is used.
1155 One use of this macro is to increase alignment of medium-size data to
1156 make it all fit in fewer cache lines.
1158 @findex EMPTY_FIELD_BOUNDARY
1159 @item EMPTY_FIELD_BOUNDARY
1160 Alignment in bits to be given to a structure bit-field that follows an
1161 empty field such as @code{int : 0;}.
1163 Note that @code{PCC_BITFIELD_TYPE_MATTERS} also affects the alignment
1164 that results from an empty field.
1166 @findex STRUCTURE_SIZE_BOUNDARY
1167 @item STRUCTURE_SIZE_BOUNDARY
1168 Number of bits which any structure or union's size must be a multiple of.
1169 Each structure or union's size is rounded up to a multiple of this.
1171 If you do not define this macro, the default is the same as
1172 @code{BITS_PER_UNIT}.
1174 @findex STRICT_ALIGNMENT
1175 @item STRICT_ALIGNMENT
1176 Define this macro to be the value 1 if instructions will fail to work
1177 if given data not on the nominal alignment. If instructions will merely
1178 go slower in that case, define this macro as 0.
1180 @findex PCC_BITFIELD_TYPE_MATTERS
1181 @item PCC_BITFIELD_TYPE_MATTERS
1182 Define this if you wish to imitate the way many other C compilers handle
1183 alignment of bit-fields and the structures that contain them.
1185 The behavior is that the type written for a named bit-field (@code{int},
1186 @code{short}, or other integer type) imposes an alignment for the entire
1187 structure, as if the structure really did contain an ordinary field of
1188 that type. In addition, the bit-field is placed within the structure so
1189 that it would fit within such a field, not crossing a boundary for it.
1191 Thus, on most machines, a named bit-field whose type is written as
1192 @code{int} would not cross a four-byte boundary, and would force
1193 four-byte alignment for the whole structure. (The alignment used may
1194 not be four bytes; it is controlled by the other alignment parameters.)
1196 An unnamed bit-field will not affect the alignment of the containing
1199 If the macro is defined, its definition should be a C expression;
1200 a nonzero value for the expression enables this behavior.
1202 Note that if this macro is not defined, or its value is zero, some
1203 bit-fields may cross more than one alignment boundary. The compiler can
1204 support such references if there are @samp{insv}, @samp{extv}, and
1205 @samp{extzv} insns that can directly reference memory.
1207 The other known way of making bit-fields work is to define
1208 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1209 Then every structure can be accessed with fullwords.
1211 Unless the machine has bit-field instructions or you define
1212 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1213 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1215 If your aim is to make GCC use the same conventions for laying out
1216 bit-fields as are used by another compiler, here is how to investigate
1217 what the other compiler does. Compile and run this program:
1236 printf ("Size of foo1 is %d\n",
1237 sizeof (struct foo1));
1238 printf ("Size of foo2 is %d\n",
1239 sizeof (struct foo2));
1244 If this prints 2 and 5, then the compiler's behavior is what you would
1245 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1247 @findex BITFIELD_NBYTES_LIMITED
1248 @item BITFIELD_NBYTES_LIMITED
1249 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1250 to aligning a bit-field within the structure.
1252 @findex MEMBER_TYPE_FORCES_BLK
1253 @item MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1254 Return 1 if a structure or array containing @var{field} should be accessed using
1257 If @var{field} is the only field in the structure, @var{mode} is its
1258 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1259 case where structures of one field would require the structure's mode to
1260 retain the field's mode.
1262 Normally, this is not needed. See the file @file{c4x.h} for an example
1263 of how to use this macro to prevent a structure having a floating point
1264 field from being accessed in an integer mode.
1266 @findex ROUND_TYPE_SIZE
1267 @item ROUND_TYPE_SIZE (@var{type}, @var{computed}, @var{specified})
1268 Define this macro as an expression for the overall size of a type
1269 (given by @var{type} as a tree node) when the size computed in the
1270 usual way is @var{computed} and the alignment is @var{specified}.
1272 The default is to round @var{computed} up to a multiple of @var{specified}.
1274 @findex ROUND_TYPE_SIZE_UNIT
1275 @item ROUND_TYPE_SIZE_UNIT (@var{type}, @var{computed}, @var{specified})
1276 Similar to @code{ROUND_TYPE_SIZE}, but sizes and alignments are
1277 specified in units (bytes). If you define @code{ROUND_TYPE_SIZE},
1278 you must also define this macro and they must be defined consistently
1281 @findex ROUND_TYPE_ALIGN
1282 @item ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1283 Define this macro as an expression for the alignment of a type (given
1284 by @var{type} as a tree node) if the alignment computed in the usual
1285 way is @var{computed} and the alignment explicitly specified was
1288 The default is to use @var{specified} if it is larger; otherwise, use
1289 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1291 @findex MAX_FIXED_MODE_SIZE
1292 @item MAX_FIXED_MODE_SIZE
1293 An integer expression for the size in bits of the largest integer
1294 machine mode that should actually be used. All integer machine modes of
1295 this size or smaller can be used for structures and unions with the
1296 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1297 (DImode)} is assumed.
1299 @findex VECTOR_MODE_SUPPORTED_P
1300 @item VECTOR_MODE_SUPPORTED_P(@var{mode})
1301 Define this macro to be nonzero if the port is prepared to handle insns
1302 involving vector mode @var{mode}. At the very least, it must have move
1303 patterns for this mode.
1305 @findex STACK_SAVEAREA_MODE
1306 @item STACK_SAVEAREA_MODE (@var{save_level})
1307 If defined, an expression of type @code{enum machine_mode} that
1308 specifies the mode of the save area operand of a
1309 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1310 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1311 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1312 having its mode specified.
1314 You need not define this macro if it always returns @code{Pmode}. You
1315 would most commonly define this macro if the
1316 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1319 @findex STACK_SIZE_MODE
1320 @item STACK_SIZE_MODE
1321 If defined, an expression of type @code{enum machine_mode} that
1322 specifies the mode of the size increment operand of an
1323 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1325 You need not define this macro if it always returns @code{word_mode}.
1326 You would most commonly define this macro if the @code{allocate_stack}
1327 pattern needs to support both a 32- and a 64-bit mode.
1329 @findex TARGET_FLOAT_FORMAT
1330 @item TARGET_FLOAT_FORMAT
1331 A code distinguishing the floating point format of the target machine.
1332 There are five defined values:
1335 @findex IEEE_FLOAT_FORMAT
1336 @item IEEE_FLOAT_FORMAT
1337 This code indicates IEEE floating point. It is the default; there is no
1338 need to define this macro when the format is IEEE@.
1340 @findex VAX_FLOAT_FORMAT
1341 @item VAX_FLOAT_FORMAT
1342 This code indicates the ``F float'' (for @code{float}) and ``D float''
1343 or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@.
1345 @findex IBM_FLOAT_FORMAT
1346 @item IBM_FLOAT_FORMAT
1347 This code indicates the format used on the IBM System/370.
1349 @findex C4X_FLOAT_FORMAT
1350 @item C4X_FLOAT_FORMAT
1351 This code indicates the format used on the TMS320C3x/C4x.
1353 @findex UNKNOWN_FLOAT_FORMAT
1354 @item UNKNOWN_FLOAT_FORMAT
1355 This code indicates any other format.
1359 formats are actually in use on supported machines, new codes should be
1362 The ordering of the component words of floating point values stored in
1363 memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}.
1365 @findex MODE_HAS_NANS
1366 @item MODE_HAS_NANS (@var{mode})
1367 When defined, this macro should be true if @var{mode} has a NaN
1368 representation. The compiler assumes that NaNs are not equal to
1369 anything (including themselves) and that addition, subtraction,
1370 multiplication and division all return NaNs when one operand is
1373 By default, this macro is true if @var{mode} is a floating-point
1374 mode and the target floating-point format is IEEE@.
1376 @findex MODE_HAS_INFINITIES
1377 @item MODE_HAS_INFINITIES (@var{mode})
1378 This macro should be true if @var{mode} can represent infinity. At
1379 present, the compiler uses this macro to decide whether @samp{x - x}
1380 is always defined. By default, the macro is true when @var{mode}
1381 is a floating-point mode and the target format is IEEE@.
1383 @findex MODE_HAS_SIGNED_ZEROS
1384 @item MODE_HAS_SIGNED_ZEROS (@var{mode})
1385 True if @var{mode} distinguishes between positive and negative zero.
1386 The rules are expected to follow the IEEE standard:
1390 @samp{x + x} has the same sign as @samp{x}.
1393 If the sum of two values with opposite sign is zero, the result is
1394 positive for all rounding modes expect towards @minus{}infinity, for
1395 which it is negative.
1398 The sign of a product or quotient is negative when exactly one
1399 of the operands is negative.
1402 The default definition is true if @var{mode} is a floating-point
1403 mode and the target format is IEEE@.
1405 @findex MODE_HAS_SIGN_DEPENDENT_ROUNDING
1406 @item MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode})
1407 If defined, this macro should be true for @var{mode} if it has at
1408 least one rounding mode in which @samp{x} and @samp{-x} can be
1409 rounded to numbers of different magnitude. Two such modes are
1410 towards @minus{}infinity and towards +infinity.
1412 The default definition of this macro is true if @var{mode} is
1413 a floating-point mode and the target format is IEEE@.
1415 @findex ROUND_TOWARDS_ZERO
1416 @item ROUND_TOWARDS_ZERO
1417 If defined, this macro should be true if the prevailing rounding
1418 mode is towards zero. A true value has the following effects:
1422 @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes.
1425 @file{libgcc.a}'s floating-point emulator will round towards zero
1426 rather than towards nearest.
1429 The compiler's floating-point emulator will round towards zero after
1430 doing arithmetic, and when converting from the internal float format to
1434 The macro does not affect the parsing of string literals. When the
1435 primary rounding mode is towards zero, library functions like
1436 @code{strtod} might still round towards nearest, and the compiler's
1437 parser should behave like the target's @code{strtod} where possible.
1439 Not defining this macro is equivalent to returning zero.
1441 @findex LARGEST_EXPONENT_IS_NORMAL
1442 @item LARGEST_EXPONENT_IS_NORMAL (@var{size})
1443 This macro should return true if floats with @var{size}
1444 bits do not have a NaN or infinity representation, but use the largest
1445 exponent for normal numbers instead.
1447 Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS}
1448 and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes.
1449 It also affects the way @file{libgcc.a} and @file{real.c} emulate
1450 floating-point arithmetic.
1452 The default definition of this macro returns false for all sizes.
1455 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1456 This target hook returns @code{true} if bit-fields in the given
1457 @var{record_type} are to be laid out following the rules of Microsoft
1458 Visual C/C++, namely: (i) a bit-field won't share the same storage
1459 unit with the previous bit-field if their underlying types have
1460 different sizes, and the bit-field will be aligned to the highest
1461 alignment of the underlying types of itself and of the previous
1462 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1463 the whole enclosing structure, even if it is unnamed; except that
1464 (iii) a zero-sized bit-field will be disregarded unless it follows
1465 another bit-field of nonzero size. If this hook returns @code{true},
1466 other macros that control bit-field layout are ignored.
1468 When a bit-field is inserted into a packed record, the whole size
1469 of the underlying type is used by one or more same-size adjacent
1470 bit-fields (that is, if its long:3, 32 bits is used in the record,
1471 and any additional adjacent long bit-fields are packed into the same
1472 chunk of 32 bits. However, if the size changes, a new field of that
1473 size is allocated). In an unpacked record, this is the same as using
1474 alignment, but not equivalent when packing.
1476 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1477 the latter will take precedence. If @samp{__attribute__((packed))} is
1478 used on a single field when MS bit-fields are in use, it will take
1479 precedence for that field, but the alignment of the rest of the structure
1480 may affect its placement.
1484 @section Layout of Source Language Data Types
1486 These macros define the sizes and other characteristics of the standard
1487 basic data types used in programs being compiled. Unlike the macros in
1488 the previous section, these apply to specific features of C and related
1489 languages, rather than to fundamental aspects of storage layout.
1492 @findex INT_TYPE_SIZE
1494 A C expression for the size in bits of the type @code{int} on the
1495 target machine. If you don't define this, the default is one word.
1497 @findex SHORT_TYPE_SIZE
1498 @item SHORT_TYPE_SIZE
1499 A C expression for the size in bits of the type @code{short} on the
1500 target machine. If you don't define this, the default is half a word.
1501 (If this would be less than one storage unit, it is rounded up to one
1504 @findex LONG_TYPE_SIZE
1505 @item LONG_TYPE_SIZE
1506 A C expression for the size in bits of the type @code{long} on the
1507 target machine. If you don't define this, the default is one word.
1509 @findex ADA_LONG_TYPE_SIZE
1510 @item ADA_LONG_TYPE_SIZE
1511 On some machines, the size used for the Ada equivalent of the type
1512 @code{long} by a native Ada compiler differs from that used by C. In
1513 that situation, define this macro to be a C expression to be used for
1514 the size of that type. If you don't define this, the default is the
1515 value of @code{LONG_TYPE_SIZE}.
1517 @findex MAX_LONG_TYPE_SIZE
1518 @item MAX_LONG_TYPE_SIZE
1519 Maximum number for the size in bits of the type @code{long} on the
1520 target machine. If this is undefined, the default is
1521 @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the
1522 largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is
1525 @findex LONG_LONG_TYPE_SIZE
1526 @item LONG_LONG_TYPE_SIZE
1527 A C expression for the size in bits of the type @code{long long} on the
1528 target machine. If you don't define this, the default is two
1529 words. If you want to support GNU Ada on your machine, the value of this
1530 macro must be at least 64.
1532 @findex CHAR_TYPE_SIZE
1533 @item CHAR_TYPE_SIZE
1534 A C expression for the size in bits of the type @code{char} on the
1535 target machine. If you don't define this, the default is
1536 @code{BITS_PER_UNIT}.
1538 @findex BOOL_TYPE_SIZE
1539 @item BOOL_TYPE_SIZE
1540 A C expression for the size in bits of the C++ type @code{bool} and
1541 C99 type @code{_Bool} on the target machine. If you don't define
1542 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1544 @findex FLOAT_TYPE_SIZE
1545 @item FLOAT_TYPE_SIZE
1546 A C expression for the size in bits of the type @code{float} on the
1547 target machine. If you don't define this, the default is one word.
1549 @findex DOUBLE_TYPE_SIZE
1550 @item DOUBLE_TYPE_SIZE
1551 A C expression for the size in bits of the type @code{double} on the
1552 target machine. If you don't define this, the default is two
1555 @findex LONG_DOUBLE_TYPE_SIZE
1556 @item LONG_DOUBLE_TYPE_SIZE
1557 A C expression for the size in bits of the type @code{long double} on
1558 the target machine. If you don't define this, the default is two
1561 @findex MAX_LONG_DOUBLE_TYPE_SIZE
1562 Maximum number for the size in bits of the type @code{long double} on the
1563 target machine. If this is undefined, the default is
1564 @code{LONG_DOUBLE_TYPE_SIZE}. Otherwise, it is the constant value that is
1565 the largest value that @code{LONG_DOUBLE_TYPE_SIZE} can have at run-time.
1566 This is used in @code{cpp}.
1568 @findex TARGET_FLT_EVAL_METHOD
1569 @item TARGET_FLT_EVAL_METHOD
1570 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1571 assuming, if applicable, that the floating-point control word is in its
1572 default state. If you do not define this macro the value of
1573 @code{FLT_EVAL_METHOD} will be zero.
1575 @findex WIDEST_HARDWARE_FP_SIZE
1576 @item WIDEST_HARDWARE_FP_SIZE
1577 A C expression for the size in bits of the widest floating-point format
1578 supported by the hardware. If you define this macro, you must specify a
1579 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1580 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1583 @findex DEFAULT_SIGNED_CHAR
1584 @item DEFAULT_SIGNED_CHAR
1585 An expression whose value is 1 or 0, according to whether the type
1586 @code{char} should be signed or unsigned by default. The user can
1587 always override this default with the options @option{-fsigned-char}
1588 and @option{-funsigned-char}.
1590 @findex DEFAULT_SHORT_ENUMS
1591 @item DEFAULT_SHORT_ENUMS
1592 A C expression to determine whether to give an @code{enum} type
1593 only as many bytes as it takes to represent the range of possible values
1594 of that type. A nonzero value means to do that; a zero value means all
1595 @code{enum} types should be allocated like @code{int}.
1597 If you don't define the macro, the default is 0.
1601 A C expression for a string describing the name of the data type to use
1602 for size values. The typedef name @code{size_t} is defined using the
1603 contents of the string.
1605 The string can contain more than one keyword. If so, separate them with
1606 spaces, and write first any length keyword, then @code{unsigned} if
1607 appropriate, and finally @code{int}. The string must exactly match one
1608 of the data type names defined in the function
1609 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1610 omit @code{int} or change the order---that would cause the compiler to
1613 If you don't define this macro, the default is @code{"long unsigned
1616 @findex PTRDIFF_TYPE
1618 A C expression for a string describing the name of the data type to use
1619 for the result of subtracting two pointers. The typedef name
1620 @code{ptrdiff_t} is defined using the contents of the string. See
1621 @code{SIZE_TYPE} above for more information.
1623 If you don't define this macro, the default is @code{"long int"}.
1627 A C expression for a string describing the name of the data type to use
1628 for wide characters. The typedef name @code{wchar_t} is defined using
1629 the contents of the string. See @code{SIZE_TYPE} above for more
1632 If you don't define this macro, the default is @code{"int"}.
1634 @findex WCHAR_TYPE_SIZE
1635 @item WCHAR_TYPE_SIZE
1636 A C expression for the size in bits of the data type for wide
1637 characters. This is used in @code{cpp}, which cannot make use of
1640 @findex MAX_WCHAR_TYPE_SIZE
1641 @item MAX_WCHAR_TYPE_SIZE
1642 Maximum number for the size in bits of the data type for wide
1643 characters. If this is undefined, the default is
1644 @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
1645 largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is
1648 @findex GCOV_TYPE_SIZE
1649 @item GCOV_TYPE_SIZE
1650 A C expression for the size in bits of the type used for gcov counters on the
1651 target machine. If you don't define this, the default is one
1652 @code{LONG_TYPE_SIZE} in case it is greater or equal to 64-bit and
1653 @code{LONG_LONG_TYPE_SIZE} otherwise. You may want to re-define the type to
1654 ensure atomicity for counters in multithreaded programs.
1658 A C expression for a string describing the name of the data type to
1659 use for wide characters passed to @code{printf} and returned from
1660 @code{getwc}. The typedef name @code{wint_t} is defined using the
1661 contents of the string. See @code{SIZE_TYPE} above for more
1664 If you don't define this macro, the default is @code{"unsigned int"}.
1668 A C expression for a string describing the name of the data type that
1669 can represent any value of any standard or extended signed integer type.
1670 The typedef name @code{intmax_t} is defined using the contents of the
1671 string. See @code{SIZE_TYPE} above for more information.
1673 If you don't define this macro, the default is the first of
1674 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1675 much precision as @code{long long int}.
1677 @findex UINTMAX_TYPE
1679 A C expression for a string describing the name of the data type that
1680 can represent any value of any standard or extended unsigned integer
1681 type. The typedef name @code{uintmax_t} is defined using the contents
1682 of the string. See @code{SIZE_TYPE} above for more information.
1684 If you don't define this macro, the default is the first of
1685 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1686 unsigned int"} that has as much precision as @code{long long unsigned
1689 @findex TARGET_PTRMEMFUNC_VBIT_LOCATION
1690 @item TARGET_PTRMEMFUNC_VBIT_LOCATION
1691 The C++ compiler represents a pointer-to-member-function with a struct
1698 ptrdiff_t vtable_index;
1705 The C++ compiler must use one bit to indicate whether the function that
1706 will be called through a pointer-to-member-function is virtual.
1707 Normally, we assume that the low-order bit of a function pointer must
1708 always be zero. Then, by ensuring that the vtable_index is odd, we can
1709 distinguish which variant of the union is in use. But, on some
1710 platforms function pointers can be odd, and so this doesn't work. In
1711 that case, we use the low-order bit of the @code{delta} field, and shift
1712 the remainder of the @code{delta} field to the left.
1714 GCC will automatically make the right selection about where to store
1715 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1716 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1717 set such that functions always start at even addresses, but the lowest
1718 bit of pointers to functions indicate whether the function at that
1719 address is in ARM or Thumb mode. If this is the case of your
1720 architecture, you should define this macro to
1721 @code{ptrmemfunc_vbit_in_delta}.
1723 In general, you should not have to define this macro. On architectures
1724 in which function addresses are always even, according to
1725 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1726 @code{ptrmemfunc_vbit_in_pfn}.
1728 @findex TARGET_VTABLE_USES_DESCRIPTORS
1729 @item TARGET_VTABLE_USES_DESCRIPTORS
1730 Normally, the C++ compiler uses function pointers in vtables. This
1731 macro allows the target to change to use ``function descriptors''
1732 instead. Function descriptors are found on targets for whom a
1733 function pointer is actually a small data structure. Normally the
1734 data structure consists of the actual code address plus a data
1735 pointer to which the function's data is relative.
1737 If vtables are used, the value of this macro should be the number
1738 of words that the function descriptor occupies.
1740 @findex TARGET_VTABLE_ENTRY_ALIGN
1741 @item TARGET_VTABLE_ENTRY_ALIGN
1742 By default, the vtable entries are void pointers, the so the alignment
1743 is the same as pointer alignment. The value of this macro specifies
1744 the alignment of the vtable entry in bits. It should be defined only
1745 when special alignment is necessary. */
1747 @findex TARGET_VTABLE_DATA_ENTRY_DISTANCE
1748 @item TARGET_VTABLE_DATA_ENTRY_DISTANCE
1749 There are a few non-descriptor entries in the vtable at offsets below
1750 zero. If these entries must be padded (say, to preserve the alignment
1751 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1752 of words in each data entry.
1755 @node Escape Sequences
1756 @section Target Character Escape Sequences
1757 @cindex escape sequences
1759 By default, GCC assumes that the C character escape sequences take on
1760 their ASCII values for the target. If this is not correct, you must
1761 explicitly define all of the macros below.
1766 A C constant expression for the integer value for escape sequence
1771 A C constant expression for the integer value of the target escape
1772 character. As an extension, GCC evaluates the escape sequences
1773 @samp{\e} and @samp{\E} to this.
1777 @findex TARGET_NEWLINE
1780 @itemx TARGET_NEWLINE
1781 C constant expressions for the integer values for escape sequences
1782 @samp{\b}, @samp{\t} and @samp{\n}.
1790 C constant expressions for the integer values for escape sequences
1791 @samp{\v}, @samp{\f} and @samp{\r}.
1795 @section Register Usage
1796 @cindex register usage
1798 This section explains how to describe what registers the target machine
1799 has, and how (in general) they can be used.
1801 The description of which registers a specific instruction can use is
1802 done with register classes; see @ref{Register Classes}. For information
1803 on using registers to access a stack frame, see @ref{Frame Registers}.
1804 For passing values in registers, see @ref{Register Arguments}.
1805 For returning values in registers, see @ref{Scalar Return}.
1808 * Register Basics:: Number and kinds of registers.
1809 * Allocation Order:: Order in which registers are allocated.
1810 * Values in Registers:: What kinds of values each reg can hold.
1811 * Leaf Functions:: Renumbering registers for leaf functions.
1812 * Stack Registers:: Handling a register stack such as 80387.
1815 @node Register Basics
1816 @subsection Basic Characteristics of Registers
1818 @c prevent bad page break with this line
1819 Registers have various characteristics.
1822 @findex FIRST_PSEUDO_REGISTER
1823 @item FIRST_PSEUDO_REGISTER
1824 Number of hardware registers known to the compiler. They receive
1825 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1826 pseudo register's number really is assigned the number
1827 @code{FIRST_PSEUDO_REGISTER}.
1829 @item FIXED_REGISTERS
1830 @findex FIXED_REGISTERS
1831 @cindex fixed register
1832 An initializer that says which registers are used for fixed purposes
1833 all throughout the compiled code and are therefore not available for
1834 general allocation. These would include the stack pointer, the frame
1835 pointer (except on machines where that can be used as a general
1836 register when no frame pointer is needed), the program counter on
1837 machines where that is considered one of the addressable registers,
1838 and any other numbered register with a standard use.
1840 This information is expressed as a sequence of numbers, separated by
1841 commas and surrounded by braces. The @var{n}th number is 1 if
1842 register @var{n} is fixed, 0 otherwise.
1844 The table initialized from this macro, and the table initialized by
1845 the following one, may be overridden at run time either automatically,
1846 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1847 the user with the command options @option{-ffixed-@var{reg}},
1848 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1850 @findex CALL_USED_REGISTERS
1851 @item CALL_USED_REGISTERS
1852 @cindex call-used register
1853 @cindex call-clobbered register
1854 @cindex call-saved register
1855 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1856 clobbered (in general) by function calls as well as for fixed
1857 registers. This macro therefore identifies the registers that are not
1858 available for general allocation of values that must live across
1861 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1862 automatically saves it on function entry and restores it on function
1863 exit, if the register is used within the function.
1865 @findex CALL_REALLY_USED_REGISTERS
1866 @item CALL_REALLY_USED_REGISTERS
1867 @cindex call-used register
1868 @cindex call-clobbered register
1869 @cindex call-saved register
1870 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1871 that the entire set of @code{FIXED_REGISTERS} be included.
1872 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1873 This macro is optional. If not specified, it defaults to the value
1874 of @code{CALL_USED_REGISTERS}.
1876 @findex HARD_REGNO_CALL_PART_CLOBBERED
1877 @item HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1878 @cindex call-used register
1879 @cindex call-clobbered register
1880 @cindex call-saved register
1881 A C expression that is nonzero if it is not permissible to store a
1882 value of mode @var{mode} in hard register number @var{regno} across a
1883 call without some part of it being clobbered. For most machines this
1884 macro need not be defined. It is only required for machines that do not
1885 preserve the entire contents of a register across a call.
1887 @findex CONDITIONAL_REGISTER_USAGE
1889 @findex call_used_regs
1890 @item CONDITIONAL_REGISTER_USAGE
1891 Zero or more C statements that may conditionally modify five variables
1892 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1893 @code{reg_names}, and @code{reg_class_contents}, to take into account
1894 any dependence of these register sets on target flags. The first three
1895 of these are of type @code{char []} (interpreted as Boolean vectors).
1896 @code{global_regs} is a @code{const char *[]}, and
1897 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1898 called, @code{fixed_regs}, @code{call_used_regs},
1899 @code{reg_class_contents}, and @code{reg_names} have been initialized
1900 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1901 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1902 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1903 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1904 command options have been applied.
1906 You need not define this macro if it has no work to do.
1908 @cindex disabling certain registers
1909 @cindex controlling register usage
1910 If the usage of an entire class of registers depends on the target
1911 flags, you may indicate this to GCC by using this macro to modify
1912 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1913 registers in the classes which should not be used by GCC@. Also define
1914 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
1915 to return @code{NO_REGS} if it
1916 is called with a letter for a class that shouldn't be used.
1918 (However, if this class is not included in @code{GENERAL_REGS} and all
1919 of the insn patterns whose constraints permit this class are
1920 controlled by target switches, then GCC will automatically avoid using
1921 these registers when the target switches are opposed to them.)
1923 @findex NON_SAVING_SETJMP
1924 @item NON_SAVING_SETJMP
1925 If this macro is defined and has a nonzero value, it means that
1926 @code{setjmp} and related functions fail to save the registers, or that
1927 @code{longjmp} fails to restore them. To compensate, the compiler
1928 avoids putting variables in registers in functions that use
1931 @findex INCOMING_REGNO
1932 @item INCOMING_REGNO (@var{out})
1933 Define this macro if the target machine has register windows. This C
1934 expression returns the register number as seen by the called function
1935 corresponding to the register number @var{out} as seen by the calling
1936 function. Return @var{out} if register number @var{out} is not an
1939 @findex OUTGOING_REGNO
1940 @item OUTGOING_REGNO (@var{in})
1941 Define this macro if the target machine has register windows. This C
1942 expression returns the register number as seen by the calling function
1943 corresponding to the register number @var{in} as seen by the called
1944 function. Return @var{in} if register number @var{in} is not an inbound
1948 @item LOCAL_REGNO (@var{regno})
1949 Define this macro if the target machine has register windows. This C
1950 expression returns true if the register is call-saved but is in the
1951 register window. Unlike most call-saved registers, such registers
1952 need not be explicitly restored on function exit or during non-local
1958 If the program counter has a register number, define this as that
1959 register number. Otherwise, do not define it.
1963 @node Allocation Order
1964 @subsection Order of Allocation of Registers
1965 @cindex order of register allocation
1966 @cindex register allocation order
1968 @c prevent bad page break with this line
1969 Registers are allocated in order.
1972 @findex REG_ALLOC_ORDER
1973 @item REG_ALLOC_ORDER
1974 If defined, an initializer for a vector of integers, containing the
1975 numbers of hard registers in the order in which GCC should prefer
1976 to use them (from most preferred to least).
1978 If this macro is not defined, registers are used lowest numbered first
1979 (all else being equal).
1981 One use of this macro is on machines where the highest numbered
1982 registers must always be saved and the save-multiple-registers
1983 instruction supports only sequences of consecutive registers. On such
1984 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1985 the highest numbered allocable register first.
1987 @findex ORDER_REGS_FOR_LOCAL_ALLOC
1988 @item ORDER_REGS_FOR_LOCAL_ALLOC
1989 A C statement (sans semicolon) to choose the order in which to allocate
1990 hard registers for pseudo-registers local to a basic block.
1992 Store the desired register order in the array @code{reg_alloc_order}.
1993 Element 0 should be the register to allocate first; element 1, the next
1994 register; and so on.
1996 The macro body should not assume anything about the contents of
1997 @code{reg_alloc_order} before execution of the macro.
1999 On most machines, it is not necessary to define this macro.
2002 @node Values in Registers
2003 @subsection How Values Fit in Registers
2005 This section discusses the macros that describe which kinds of values
2006 (specifically, which machine modes) each register can hold, and how many
2007 consecutive registers are needed for a given mode.
2010 @findex HARD_REGNO_NREGS
2011 @item HARD_REGNO_NREGS (@var{regno}, @var{mode})
2012 A C expression for the number of consecutive hard registers, starting
2013 at register number @var{regno}, required to hold a value of mode
2016 On a machine where all registers are exactly one word, a suitable
2017 definition of this macro is
2020 #define HARD_REGNO_NREGS(REGNO, MODE) \
2021 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2025 @findex HARD_REGNO_MODE_OK
2026 @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2027 A C expression that is nonzero if it is permissible to store a value
2028 of mode @var{mode} in hard register number @var{regno} (or in several
2029 registers starting with that one). For a machine where all registers
2030 are equivalent, a suitable definition is
2033 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2036 You need not include code to check for the numbers of fixed registers,
2037 because the allocation mechanism considers them to be always occupied.
2039 @cindex register pairs
2040 On some machines, double-precision values must be kept in even/odd
2041 register pairs. You can implement that by defining this macro to reject
2042 odd register numbers for such modes.
2044 The minimum requirement for a mode to be OK in a register is that the
2045 @samp{mov@var{mode}} instruction pattern support moves between the
2046 register and other hard register in the same class and that moving a
2047 value into the register and back out not alter it.
2049 Since the same instruction used to move @code{word_mode} will work for
2050 all narrower integer modes, it is not necessary on any machine for
2051 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2052 you define patterns @samp{movhi}, etc., to take advantage of this. This
2053 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2054 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2057 Many machines have special registers for floating point arithmetic.
2058 Often people assume that floating point machine modes are allowed only
2059 in floating point registers. This is not true. Any registers that
2060 can hold integers can safely @emph{hold} a floating point machine
2061 mode, whether or not floating arithmetic can be done on it in those
2062 registers. Integer move instructions can be used to move the values.
2064 On some machines, though, the converse is true: fixed-point machine
2065 modes may not go in floating registers. This is true if the floating
2066 registers normalize any value stored in them, because storing a
2067 non-floating value there would garble it. In this case,
2068 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2069 floating registers. But if the floating registers do not automatically
2070 normalize, if you can store any bit pattern in one and retrieve it
2071 unchanged without a trap, then any machine mode may go in a floating
2072 register, so you can define this macro to say so.
2074 The primary significance of special floating registers is rather that
2075 they are the registers acceptable in floating point arithmetic
2076 instructions. However, this is of no concern to
2077 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2078 constraints for those instructions.
2080 On some machines, the floating registers are especially slow to access,
2081 so that it is better to store a value in a stack frame than in such a
2082 register if floating point arithmetic is not being done. As long as the
2083 floating registers are not in class @code{GENERAL_REGS}, they will not
2084 be used unless some pattern's constraint asks for one.
2086 @findex MODES_TIEABLE_P
2087 @item MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2088 A C expression that is nonzero if a value of mode
2089 @var{mode1} is accessible in mode @var{mode2} without copying.
2091 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2092 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2093 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2094 should be nonzero. If they differ for any @var{r}, you should define
2095 this macro to return zero unless some other mechanism ensures the
2096 accessibility of the value in a narrower mode.
2098 You should define this macro to return nonzero in as many cases as
2099 possible since doing so will allow GCC to perform better register
2102 @findex AVOID_CCMODE_COPIES
2103 @item AVOID_CCMODE_COPIES
2104 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2105 registers. You should only define this macro if support for copying to/from
2106 @code{CCmode} is incomplete.
2109 @node Leaf Functions
2110 @subsection Handling Leaf Functions
2112 @cindex leaf functions
2113 @cindex functions, leaf
2114 On some machines, a leaf function (i.e., one which makes no calls) can run
2115 more efficiently if it does not make its own register window. Often this
2116 means it is required to receive its arguments in the registers where they
2117 are passed by the caller, instead of the registers where they would
2120 The special treatment for leaf functions generally applies only when
2121 other conditions are met; for example, often they may use only those
2122 registers for its own variables and temporaries. We use the term ``leaf
2123 function'' to mean a function that is suitable for this special
2124 handling, so that functions with no calls are not necessarily ``leaf
2127 GCC assigns register numbers before it knows whether the function is
2128 suitable for leaf function treatment. So it needs to renumber the
2129 registers in order to output a leaf function. The following macros
2133 @findex LEAF_REGISTERS
2134 @item LEAF_REGISTERS
2135 Name of a char vector, indexed by hard register number, which
2136 contains 1 for a register that is allowable in a candidate for leaf
2139 If leaf function treatment involves renumbering the registers, then the
2140 registers marked here should be the ones before renumbering---those that
2141 GCC would ordinarily allocate. The registers which will actually be
2142 used in the assembler code, after renumbering, should not be marked with 1
2145 Define this macro only if the target machine offers a way to optimize
2146 the treatment of leaf functions.
2148 @findex LEAF_REG_REMAP
2149 @item LEAF_REG_REMAP (@var{regno})
2150 A C expression whose value is the register number to which @var{regno}
2151 should be renumbered, when a function is treated as a leaf function.
2153 If @var{regno} is a register number which should not appear in a leaf
2154 function before renumbering, then the expression should yield @minus{}1, which
2155 will cause the compiler to abort.
2157 Define this macro only if the target machine offers a way to optimize the
2158 treatment of leaf functions, and registers need to be renumbered to do
2162 @findex current_function_is_leaf
2163 @findex current_function_uses_only_leaf_regs
2164 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2165 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2166 specially. They can test the C variable @code{current_function_is_leaf}
2167 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2168 set prior to local register allocation and is valid for the remaining
2169 compiler passes. They can also test the C variable
2170 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2171 functions which only use leaf registers.
2172 @code{current_function_uses_only_leaf_regs} is valid after reload and is
2173 only useful if @code{LEAF_REGISTERS} is defined.
2174 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2175 @c of the next paragraph?! --mew 2feb93
2177 @node Stack Registers
2178 @subsection Registers That Form a Stack
2180 There are special features to handle computers where some of the
2181 ``registers'' form a stack, as in the 80387 coprocessor for the 80386.
2182 Stack registers are normally written by pushing onto the stack, and are
2183 numbered relative to the top of the stack.
2185 Currently, GCC can only handle one group of stack-like registers, and
2186 they must be consecutively numbered.
2191 Define this if the machine has any stack-like registers.
2193 @findex FIRST_STACK_REG
2194 @item FIRST_STACK_REG
2195 The number of the first stack-like register. This one is the top
2198 @findex LAST_STACK_REG
2199 @item LAST_STACK_REG
2200 The number of the last stack-like register. This one is the bottom of
2204 @node Register Classes
2205 @section Register Classes
2206 @cindex register class definitions
2207 @cindex class definitions, register
2209 On many machines, the numbered registers are not all equivalent.
2210 For example, certain registers may not be allowed for indexed addressing;
2211 certain registers may not be allowed in some instructions. These machine
2212 restrictions are described to the compiler using @dfn{register classes}.
2214 You define a number of register classes, giving each one a name and saying
2215 which of the registers belong to it. Then you can specify register classes
2216 that are allowed as operands to particular instruction patterns.
2220 In general, each register will belong to several classes. In fact, one
2221 class must be named @code{ALL_REGS} and contain all the registers. Another
2222 class must be named @code{NO_REGS} and contain no registers. Often the
2223 union of two classes will be another class; however, this is not required.
2225 @findex GENERAL_REGS
2226 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2227 terribly special about the name, but the operand constraint letters
2228 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2229 the same as @code{ALL_REGS}, just define it as a macro which expands
2232 Order the classes so that if class @var{x} is contained in class @var{y}
2233 then @var{x} has a lower class number than @var{y}.
2235 The way classes other than @code{GENERAL_REGS} are specified in operand
2236 constraints is through machine-dependent operand constraint letters.
2237 You can define such letters to correspond to various classes, then use
2238 them in operand constraints.
2240 You should define a class for the union of two classes whenever some
2241 instruction allows both classes. For example, if an instruction allows
2242 either a floating point (coprocessor) register or a general register for a
2243 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2244 which includes both of them. Otherwise you will get suboptimal code.
2246 You must also specify certain redundant information about the register
2247 classes: for each class, which classes contain it and which ones are
2248 contained in it; for each pair of classes, the largest class contained
2251 When a value occupying several consecutive registers is expected in a
2252 certain class, all the registers used must belong to that class.
2253 Therefore, register classes cannot be used to enforce a requirement for
2254 a register pair to start with an even-numbered register. The way to
2255 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2257 Register classes used for input-operands of bitwise-and or shift
2258 instructions have a special requirement: each such class must have, for
2259 each fixed-point machine mode, a subclass whose registers can transfer that
2260 mode to or from memory. For example, on some machines, the operations for
2261 single-byte values (@code{QImode}) are limited to certain registers. When
2262 this is so, each register class that is used in a bitwise-and or shift
2263 instruction must have a subclass consisting of registers from which
2264 single-byte values can be loaded or stored. This is so that
2265 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2268 @findex enum reg_class
2269 @item enum reg_class
2270 An enumeral type that must be defined with all the register class names
2271 as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
2272 must be the last register class, followed by one more enumeral value,
2273 @code{LIM_REG_CLASSES}, which is not a register class but rather
2274 tells how many classes there are.
2276 Each register class has a number, which is the value of casting
2277 the class name to type @code{int}. The number serves as an index
2278 in many of the tables described below.
2280 @findex N_REG_CLASSES
2282 The number of distinct register classes, defined as follows:
2285 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2288 @findex REG_CLASS_NAMES
2289 @item REG_CLASS_NAMES
2290 An initializer containing the names of the register classes as C string
2291 constants. These names are used in writing some of the debugging dumps.
2293 @findex REG_CLASS_CONTENTS
2294 @item REG_CLASS_CONTENTS
2295 An initializer containing the contents of the register classes, as integers
2296 which are bit masks. The @var{n}th integer specifies the contents of class
2297 @var{n}. The way the integer @var{mask} is interpreted is that
2298 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2300 When the machine has more than 32 registers, an integer does not suffice.
2301 Then the integers are replaced by sub-initializers, braced groupings containing
2302 several integers. Each sub-initializer must be suitable as an initializer
2303 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2304 In this situation, the first integer in each sub-initializer corresponds to
2305 registers 0 through 31, the second integer to registers 32 through 63, and
2308 @findex REGNO_REG_CLASS
2309 @item REGNO_REG_CLASS (@var{regno})
2310 A C expression whose value is a register class containing hard register
2311 @var{regno}. In general there is more than one such class; choose a class
2312 which is @dfn{minimal}, meaning that no smaller class also contains the
2315 @findex BASE_REG_CLASS
2316 @item BASE_REG_CLASS
2317 A macro whose definition is the name of the class to which a valid
2318 base register must belong. A base register is one used in an address
2319 which is the register value plus a displacement.
2321 @findex MODE_BASE_REG_CLASS
2322 @item MODE_BASE_REG_CLASS (@var{mode})
2323 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2324 the selection of a base register in a mode dependent manner. If
2325 @var{mode} is VOIDmode then it should return the same value as
2326 @code{BASE_REG_CLASS}.
2328 @findex INDEX_REG_CLASS
2329 @item INDEX_REG_CLASS
2330 A macro whose definition is the name of the class to which a valid
2331 index register must belong. An index register is one used in an
2332 address where its value is either multiplied by a scale factor or
2333 added to another register (as well as added to a displacement).
2335 @findex CONSTRAINT_LEN
2336 @item CONSTRAINT_LEN (@var{char}, @var{str})
2337 For the constraint at the start of @var{str}, which starts with the letter
2338 @var{c}, return the length. This allows you to have register class /
2339 constant / extra constraints that are longer than a single letter;
2340 you don't need to define this macro if you can do with single-letter
2341 constraints only. The definition of this macro should use
2342 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2343 to handle specially.
2344 There are some sanity checks in genoutput.c that check the constraint lengths
2345 for the md file, so you can also use this macro to help you while you are
2346 transitioning from a byzantine single-letter-constraint scheme: when you
2347 return a negative length for a constraint you want to re-use, genoutput
2348 will complain about every instance where it is used in the md file.
2350 @findex REG_CLASS_FROM_LETTER
2351 @item REG_CLASS_FROM_LETTER (@var{char})
2352 A C expression which defines the machine-dependent operand constraint
2353 letters for register classes. If @var{char} is such a letter, the
2354 value should be the register class corresponding to it. Otherwise,
2355 the value should be @code{NO_REGS}. The register letter @samp{r},
2356 corresponding to class @code{GENERAL_REGS}, will not be passed
2357 to this macro; you do not need to handle it.
2359 @findex REG_CLASS_FROM_CONSTRAINT
2360 @item REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2361 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2362 passed in @var{str}, so that you can use suffixes to distinguish between
2365 @findex REGNO_OK_FOR_BASE_P
2366 @item REGNO_OK_FOR_BASE_P (@var{num})
2367 A C expression which is nonzero if register number @var{num} is
2368 suitable for use as a base register in operand addresses. It may be
2369 either a suitable hard register or a pseudo register that has been
2370 allocated such a hard register.
2372 @findex REGNO_MODE_OK_FOR_BASE_P
2373 @item REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2374 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2375 that expression may examine the mode of the memory reference in
2376 @var{mode}. You should define this macro if the mode of the memory
2377 reference affects whether a register may be used as a base register. If
2378 you define this macro, the compiler will use it instead of
2379 @code{REGNO_OK_FOR_BASE_P}.
2381 @findex REGNO_OK_FOR_INDEX_P
2382 @item REGNO_OK_FOR_INDEX_P (@var{num})
2383 A C expression which is nonzero if register number @var{num} is
2384 suitable for use as an index register in operand addresses. It may be
2385 either a suitable hard register or a pseudo register that has been
2386 allocated such a hard register.
2388 The difference between an index register and a base register is that
2389 the index register may be scaled. If an address involves the sum of
2390 two registers, neither one of them scaled, then either one may be
2391 labeled the ``base'' and the other the ``index''; but whichever
2392 labeling is used must fit the machine's constraints of which registers
2393 may serve in each capacity. The compiler will try both labelings,
2394 looking for one that is valid, and will reload one or both registers
2395 only if neither labeling works.
2397 @findex PREFERRED_RELOAD_CLASS
2398 @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2399 A C expression that places additional restrictions on the register class
2400 to use when it is necessary to copy value @var{x} into a register in class
2401 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2402 another, smaller class. On many machines, the following definition is
2406 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2409 Sometimes returning a more restrictive class makes better code. For
2410 example, on the 68000, when @var{x} is an integer constant that is in range
2411 for a @samp{moveq} instruction, the value of this macro is always
2412 @code{DATA_REGS} as long as @var{class} includes the data registers.
2413 Requiring a data register guarantees that a @samp{moveq} will be used.
2415 If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
2416 you can force @var{x} into a memory constant. This is useful on
2417 certain machines where immediate floating values cannot be loaded into
2418 certain kinds of registers.
2420 @findex PREFERRED_OUTPUT_RELOAD_CLASS
2421 @item PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2422 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2423 input reloads. If you don't define this macro, the default is to use
2424 @var{class}, unchanged.
2426 @findex LIMIT_RELOAD_CLASS
2427 @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2428 A C expression that places additional restrictions on the register class
2429 to use when it is necessary to be able to hold a value of mode
2430 @var{mode} in a reload register for which class @var{class} would
2433 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2434 there are certain modes that simply can't go in certain reload classes.
2436 The value is a register class; perhaps @var{class}, or perhaps another,
2439 Don't define this macro unless the target machine has limitations which
2440 require the macro to do something nontrivial.
2442 @findex SECONDARY_RELOAD_CLASS
2443 @findex SECONDARY_INPUT_RELOAD_CLASS
2444 @findex SECONDARY_OUTPUT_RELOAD_CLASS
2445 @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2446 @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2447 @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2448 Many machines have some registers that cannot be copied directly to or
2449 from memory or even from other types of registers. An example is the
2450 @samp{MQ} register, which on most machines, can only be copied to or
2451 from general registers, but not memory. Some machines allow copying all
2452 registers to and from memory, but require a scratch register for stores
2453 to some memory locations (e.g., those with symbolic address on the RT,
2454 and those with certain symbolic address on the SPARC when compiling
2455 PIC)@. In some cases, both an intermediate and a scratch register are
2458 You should define these macros to indicate to the reload phase that it may
2459 need to allocate at least one register for a reload in addition to the
2460 register to contain the data. Specifically, if copying @var{x} to a
2461 register @var{class} in @var{mode} requires an intermediate register,
2462 you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2463 largest register class all of whose registers can be used as
2464 intermediate registers or scratch registers.
2466 If copying a register @var{class} in @var{mode} to @var{x} requires an
2467 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2468 should be defined to return the largest register class required. If the
2469 requirements for input and output reloads are the same, the macro
2470 @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
2473 The values returned by these macros are often @code{GENERAL_REGS}.
2474 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2475 can be directly copied to or from a register of @var{class} in
2476 @var{mode} without requiring a scratch register. Do not define this
2477 macro if it would always return @code{NO_REGS}.
2479 If a scratch register is required (either with or without an
2480 intermediate register), you should define patterns for
2481 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2482 (@pxref{Standard Names}. These patterns, which will normally be
2483 implemented with a @code{define_expand}, should be similar to the
2484 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2487 Define constraints for the reload register and scratch register that
2488 contain a single register class. If the original reload register (whose
2489 class is @var{class}) can meet the constraint given in the pattern, the
2490 value returned by these macros is used for the class of the scratch
2491 register. Otherwise, two additional reload registers are required.
2492 Their classes are obtained from the constraints in the insn pattern.
2494 @var{x} might be a pseudo-register or a @code{subreg} of a
2495 pseudo-register, which could either be in a hard register or in memory.
2496 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2497 in memory and the hard register number if it is in a register.
2499 These macros should not be used in the case where a particular class of
2500 registers can only be copied to memory and not to another class of
2501 registers. In that case, secondary reload registers are not needed and
2502 would not be helpful. Instead, a stack location must be used to perform
2503 the copy and the @code{mov@var{m}} pattern should use memory as an
2504 intermediate storage. This case often occurs between floating-point and
2507 @findex SECONDARY_MEMORY_NEEDED
2508 @item SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2509 Certain machines have the property that some registers cannot be copied
2510 to some other registers without using memory. Define this macro on
2511 those machines to be a C expression that is nonzero if objects of mode
2512 @var{m} in registers of @var{class1} can only be copied to registers of
2513 class @var{class2} by storing a register of @var{class1} into memory
2514 and loading that memory location into a register of @var{class2}.
2516 Do not define this macro if its value would always be zero.
2518 @findex SECONDARY_MEMORY_NEEDED_RTX
2519 @item SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2520 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2521 allocates a stack slot for a memory location needed for register copies.
2522 If this macro is defined, the compiler instead uses the memory location
2523 defined by this macro.
2525 Do not define this macro if you do not define
2526 @code{SECONDARY_MEMORY_NEEDED}.
2528 @findex SECONDARY_MEMORY_NEEDED_MODE
2529 @item SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2530 When the compiler needs a secondary memory location to copy between two
2531 registers of mode @var{mode}, it normally allocates sufficient memory to
2532 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2533 load operations in a mode that many bits wide and whose class is the
2534 same as that of @var{mode}.
2536 This is right thing to do on most machines because it ensures that all
2537 bits of the register are copied and prevents accesses to the registers
2538 in a narrower mode, which some machines prohibit for floating-point
2541 However, this default behavior is not correct on some machines, such as
2542 the DEC Alpha, that store short integers in floating-point registers
2543 differently than in integer registers. On those machines, the default
2544 widening will not work correctly and you must define this macro to
2545 suppress that widening in some cases. See the file @file{alpha.h} for
2548 Do not define this macro if you do not define
2549 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2550 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2552 @findex SMALL_REGISTER_CLASSES
2553 @item SMALL_REGISTER_CLASSES
2554 On some machines, it is risky to let hard registers live across arbitrary
2555 insns. Typically, these machines have instructions that require values
2556 to be in specific registers (like an accumulator), and reload will fail
2557 if the required hard register is used for another purpose across such an
2560 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2561 value on these machines. When this macro has a nonzero value, the
2562 compiler will try to minimize the lifetime of hard registers.
2564 It is always safe to define this macro with a nonzero value, but if you
2565 unnecessarily define it, you will reduce the amount of optimizations
2566 that can be performed in some cases. If you do not define this macro
2567 with a nonzero value when it is required, the compiler will run out of
2568 spill registers and print a fatal error message. For most machines, you
2569 should not define this macro at all.
2571 @findex CLASS_LIKELY_SPILLED_P
2572 @item CLASS_LIKELY_SPILLED_P (@var{class})
2573 A C expression whose value is nonzero if pseudos that have been assigned
2574 to registers of class @var{class} would likely be spilled because
2575 registers of @var{class} are needed for spill registers.
2577 The default value of this macro returns 1 if @var{class} has exactly one
2578 register and zero otherwise. On most machines, this default should be
2579 used. Only define this macro to some other expression if pseudos
2580 allocated by @file{local-alloc.c} end up in memory because their hard
2581 registers were needed for spill registers. If this macro returns nonzero
2582 for those classes, those pseudos will only be allocated by
2583 @file{global.c}, which knows how to reallocate the pseudo to another
2584 register. If there would not be another register available for
2585 reallocation, you should not change the definition of this macro since
2586 the only effect of such a definition would be to slow down register
2589 @findex CLASS_MAX_NREGS
2590 @item CLASS_MAX_NREGS (@var{class}, @var{mode})
2591 A C expression for the maximum number of consecutive registers
2592 of class @var{class} needed to hold a value of mode @var{mode}.
2594 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2595 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2596 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2597 @var{mode})} for all @var{regno} values in the class @var{class}.
2599 This macro helps control the handling of multiple-word values
2602 @item CANNOT_CHANGE_MODE_CLASS(@var{from}, @var{to})
2603 If defined, a C expression that returns a register class for which
2604 a change from mode @var{from} to mode @var{to} is invalid, otherwise the
2605 macro returns @code{NO_REGS}.
2607 For the example, loading 32-bit integer or floating-point objects into
2608 floating-point registers on the Alpha extends them to 64 bits.
2609 Therefore loading a 64-bit object and then storing it as a 32-bit object
2610 does not store the low-order 32 bits, as would be the case for a normal
2611 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2615 #define CANNOT_CHANGE_MODE_CLASS \
2616 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) ? FLOAT_REGS : NO_REGS)
2620 Three other special macros describe which operands fit which constraint
2624 @findex CONST_OK_FOR_LETTER_P
2625 @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2626 A C expression that defines the machine-dependent operand constraint
2627 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2628 particular ranges of integer values. If @var{c} is one of those
2629 letters, the expression should check that @var{value}, an integer, is in
2630 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2631 not one of those letters, the value should be 0 regardless of
2634 @findex CONST_OK_FOR_CONSTRAINT_P
2635 @item CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2636 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2637 string passed in @var{str}, so that you can use suffixes to distinguish
2638 between different variants.
2640 @findex CONST_DOUBLE_OK_FOR_LETTER_P
2641 @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2642 A C expression that defines the machine-dependent operand constraint
2643 letters that specify particular ranges of @code{const_double} values
2644 (@samp{G} or @samp{H}).
2646 If @var{c} is one of those letters, the expression should check that
2647 @var{value}, an RTX of code @code{const_double}, is in the appropriate
2648 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
2649 letters, the value should be 0 regardless of @var{value}.
2651 @code{const_double} is used for all floating-point constants and for
2652 @code{DImode} fixed-point constants. A given letter can accept either
2653 or both kinds of values. It can use @code{GET_MODE} to distinguish
2654 between these kinds.
2656 @findex CONST_DOUBLE_OK_FOR_CONSTRAINT_P
2657 @item CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2658 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
2659 string passed in @var{str}, so that you can use suffixes to distinguish
2660 between different variants.
2662 @findex EXTRA_CONSTRAINT
2663 @item EXTRA_CONSTRAINT (@var{value}, @var{c})
2664 A C expression that defines the optional machine-dependent constraint
2665 letters that can be used to segregate specific types of operands, usually
2666 memory references, for the target machine. Any letter that is not
2667 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
2668 @code{REG_CLASS_FROM_CONSTRAINT}
2669 may be used. Normally this macro will not be defined.
2671 If it is required for a particular target machine, it should return 1
2672 if @var{value} corresponds to the operand type represented by the
2673 constraint letter @var{c}. If @var{c} is not defined as an extra
2674 constraint, the value returned should be 0 regardless of @var{value}.
2676 For example, on the ROMP, load instructions cannot have their output
2677 in r0 if the memory reference contains a symbolic address. Constraint
2678 letter @samp{Q} is defined as representing a memory address that does
2679 @emph{not} contain a symbolic address. An alternative is specified with
2680 a @samp{Q} constraint on the input and @samp{r} on the output. The next
2681 alternative specifies @samp{m} on the input and a register class that
2682 does not include r0 on the output.
2684 @findex EXTRA_CONSTRAINT_STR
2685 @item EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
2686 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
2687 in @var{str}, so that you can use suffixes to distinguish between different
2690 @findex EXTRA_MEMORY_CONSTRAINT
2691 @item EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
2692 A C expression that defines the optional machine-dependent constraint
2693 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
2694 be treated like memory constraints by the reload pass.
2696 It should return 1 if the operand type represented by the constraint
2697 at the start of @var{str}, the first letter of which is the letter @var{c},
2698 comprises a subset of all memory references including
2699 all those whose address is simply a base register. This allows the reload
2700 pass to reload an operand, if it does not directly correspond to the operand
2701 type of @var{c}, by copying its address into a base register.
2703 For example, on the S/390, some instructions do not accept arbitrary
2704 memory references, but only those that do not make use of an index
2705 register. The constraint letter @samp{Q} is defined via
2706 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
2707 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
2708 a @samp{Q} constraint can handle any memory operand, because the
2709 reload pass knows it can be reloaded by copying the memory address
2710 into a base register if required. This is analogous to the way
2711 a @samp{o} constraint can handle any memory operand.
2713 @findex EXTRA_ADDRESS_CONSTRAINT
2714 @item EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
2715 A C expression that defines the optional machine-dependent constraint
2716 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
2717 @code{EXTRA_CONSTRAINT_STR}, that should
2718 be treated like address constraints by the reload pass.
2720 It should return 1 if the operand type represented by the constraint
2721 at the start of @var{str}, which starts with the letter @var{c}, comprises
2722 a subset of all memory addresses including
2723 all those that consist of just a base register. This allows the reload
2724 pass to reload an operand, if it does not directly correspond to the operand
2725 type of @var{str}, by copying it into a base register.
2727 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
2728 be used with the @code{address_operand} predicate. It is treated
2729 analogously to the @samp{p} constraint.
2732 @node Stack and Calling
2733 @section Stack Layout and Calling Conventions
2734 @cindex calling conventions
2736 @c prevent bad page break with this line
2737 This describes the stack layout and calling conventions.
2741 * Exception Handling::
2746 * Register Arguments::
2748 * Aggregate Return::
2756 @subsection Basic Stack Layout
2757 @cindex stack frame layout
2758 @cindex frame layout
2760 @c prevent bad page break with this line
2761 Here is the basic stack layout.
2764 @findex STACK_GROWS_DOWNWARD
2765 @item STACK_GROWS_DOWNWARD
2766 Define this macro if pushing a word onto the stack moves the stack
2767 pointer to a smaller address.
2769 When we say, ``define this macro if @dots{},'' it means that the
2770 compiler checks this macro only with @code{#ifdef} so the precise
2771 definition used does not matter.
2773 @findex STACK_PUSH_CODE
2774 @item STACK_PUSH_CODE
2776 This macro defines the operation used when something is pushed
2777 on the stack. In RTL, a push operation will be
2778 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2780 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2781 and @code{POST_INC}. Which of these is correct depends on
2782 the stack direction and on whether the stack pointer points
2783 to the last item on the stack or whether it points to the
2784 space for the next item on the stack.
2786 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2787 defined, which is almost always right, and @code{PRE_INC} otherwise,
2788 which is often wrong.
2790 @findex FRAME_GROWS_DOWNWARD
2791 @item FRAME_GROWS_DOWNWARD
2792 Define this macro if the addresses of local variable slots are at negative
2793 offsets from the frame pointer.
2795 @findex ARGS_GROW_DOWNWARD
2796 @item ARGS_GROW_DOWNWARD
2797 Define this macro if successive arguments to a function occupy decreasing
2798 addresses on the stack.
2800 @findex STARTING_FRAME_OFFSET
2801 @item STARTING_FRAME_OFFSET
2802 Offset from the frame pointer to the first local variable slot to be allocated.
2804 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2805 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2806 Otherwise, it is found by adding the length of the first slot to the
2807 value @code{STARTING_FRAME_OFFSET}.
2808 @c i'm not sure if the above is still correct.. had to change it to get
2809 @c rid of an overfull. --mew 2feb93
2811 @findex STACK_POINTER_OFFSET
2812 @item STACK_POINTER_OFFSET
2813 Offset from the stack pointer register to the first location at which
2814 outgoing arguments are placed. If not specified, the default value of
2815 zero is used. This is the proper value for most machines.
2817 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2818 the first location at which outgoing arguments are placed.
2820 @findex FIRST_PARM_OFFSET
2821 @item FIRST_PARM_OFFSET (@var{fundecl})
2822 Offset from the argument pointer register to the first argument's
2823 address. On some machines it may depend on the data type of the
2826 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2827 the first argument's address.
2829 @findex STACK_DYNAMIC_OFFSET
2830 @item STACK_DYNAMIC_OFFSET (@var{fundecl})
2831 Offset from the stack pointer register to an item dynamically allocated
2832 on the stack, e.g., by @code{alloca}.
2834 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2835 length of the outgoing arguments. The default is correct for most
2836 machines. See @file{function.c} for details.
2838 @findex DYNAMIC_CHAIN_ADDRESS
2839 @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2840 A C expression whose value is RTL representing the address in a stack
2841 frame where the pointer to the caller's frame is stored. Assume that
2842 @var{frameaddr} is an RTL expression for the address of the stack frame
2845 If you don't define this macro, the default is to return the value
2846 of @var{frameaddr}---that is, the stack frame address is also the
2847 address of the stack word that points to the previous frame.
2849 @findex SETUP_FRAME_ADDRESSES
2850 @item SETUP_FRAME_ADDRESSES
2851 If defined, a C expression that produces the machine-specific code to
2852 setup the stack so that arbitrary frames can be accessed. For example,
2853 on the SPARC, we must flush all of the register windows to the stack
2854 before we can access arbitrary stack frames. You will seldom need to
2857 @findex BUILTIN_SETJMP_FRAME_VALUE
2858 @item BUILTIN_SETJMP_FRAME_VALUE
2859 If defined, a C expression that contains an rtx that is used to store
2860 the address of the current frame into the built in @code{setjmp} buffer.
2861 The default value, @code{virtual_stack_vars_rtx}, is correct for most
2862 machines. One reason you may need to define this macro is if
2863 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
2865 @findex RETURN_ADDR_RTX
2866 @item RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2867 A C expression whose value is RTL representing the value of the return
2868 address for the frame @var{count} steps up from the current frame, after
2869 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2870 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2871 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2873 The value of the expression must always be the correct address when
2874 @var{count} is zero, but may be @code{NULL_RTX} if there is not way to
2875 determine the return address of other frames.
2877 @findex RETURN_ADDR_IN_PREVIOUS_FRAME
2878 @item RETURN_ADDR_IN_PREVIOUS_FRAME
2879 Define this if the return address of a particular stack frame is accessed
2880 from the frame pointer of the previous stack frame.
2882 @findex INCOMING_RETURN_ADDR_RTX
2883 @item INCOMING_RETURN_ADDR_RTX
2884 A C expression whose value is RTL representing the location of the
2885 incoming return address at the beginning of any function, before the
2886 prologue. This RTL is either a @code{REG}, indicating that the return
2887 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2890 You only need to define this macro if you want to support call frame
2891 debugging information like that provided by DWARF 2.
2893 If this RTL is a @code{REG}, you should also define
2894 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2896 @findex INCOMING_FRAME_SP_OFFSET
2897 @item INCOMING_FRAME_SP_OFFSET
2898 A C expression whose value is an integer giving the offset, in bytes,
2899 from the value of the stack pointer register to the top of the stack
2900 frame at the beginning of any function, before the prologue. The top of
2901 the frame is defined to be the value of the stack pointer in the
2902 previous frame, just before the call instruction.
2904 You only need to define this macro if you want to support call frame
2905 debugging information like that provided by DWARF 2.
2907 @findex ARG_POINTER_CFA_OFFSET
2908 @item ARG_POINTER_CFA_OFFSET (@var{fundecl})
2909 A C expression whose value is an integer giving the offset, in bytes,
2910 from the argument pointer to the canonical frame address (cfa). The
2911 final value should coincide with that calculated by
2912 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2913 during virtual register instantiation.
2915 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
2916 which is correct for most machines; in general, the arguments are found
2917 immediately before the stack frame. Note that this is not the case on
2918 some targets that save registers into the caller's frame, such as SPARC
2919 and rs6000, and so such targets need to define this macro.
2921 You only need to define this macro if the default is incorrect, and you
2922 want to support call frame debugging information like that provided by
2927 Define this macro if the stack size for the target is very small. This
2928 has the effect of disabling gcc's built-in @samp{alloca}, though
2929 @samp{__builtin_alloca} is not affected.
2932 @node Exception Handling
2933 @subsection Exception Handling Support
2934 @cindex exception handling
2937 @findex EH_RETURN_DATA_REGNO
2938 @item EH_RETURN_DATA_REGNO (@var{N})
2939 A C expression whose value is the @var{N}th register number used for
2940 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2941 @var{N} registers are usable.
2943 The exception handling library routines communicate with the exception
2944 handlers via a set of agreed upon registers. Ideally these registers
2945 should be call-clobbered; it is possible to use call-saved registers,
2946 but may negatively impact code size. The target must support at least
2947 2 data registers, but should define 4 if there are enough free registers.
2949 You must define this macro if you want to support call frame exception
2950 handling like that provided by DWARF 2.
2952 @findex EH_RETURN_STACKADJ_RTX
2953 @item EH_RETURN_STACKADJ_RTX
2954 A C expression whose value is RTL representing a location in which
2955 to store a stack adjustment to be applied before function return.
2956 This is used to unwind the stack to an exception handler's call frame.
2957 It will be assigned zero on code paths that return normally.
2959 Typically this is a call-clobbered hard register that is otherwise
2960 untouched by the epilogue, but could also be a stack slot.
2962 You must define this macro if you want to support call frame exception
2963 handling like that provided by DWARF 2.
2965 @findex EH_RETURN_HANDLER_RTX
2966 @item EH_RETURN_HANDLER_RTX
2967 A C expression whose value is RTL representing a location in which
2968 to store the address of an exception handler to which we should
2969 return. It will not be assigned on code paths that return normally.
2971 Typically this is the location in the call frame at which the normal
2972 return address is stored. For targets that return by popping an
2973 address off the stack, this might be a memory address just below
2974 the @emph{target} call frame rather than inside the current call
2975 frame. @code{EH_RETURN_STACKADJ_RTX} will have already been assigned,
2976 so it may be used to calculate the location of the target call frame.
2978 Some targets have more complex requirements than storing to an
2979 address calculable during initial code generation. In that case
2980 the @code{eh_return} instruction pattern should be used instead.
2982 If you want to support call frame exception handling, you must
2983 define either this macro or the @code{eh_return} instruction pattern.
2985 @findex ASM_PREFERRED_EH_DATA_FORMAT
2986 @item ASM_PREFERRED_EH_DATA_FORMAT(@var{code}, @var{global})
2987 This macro chooses the encoding of pointers embedded in the exception
2988 handling sections. If at all possible, this should be defined such
2989 that the exception handling section will not require dynamic relocations,
2990 and so may be read-only.
2992 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
2993 @var{global} is true if the symbol may be affected by dynamic relocations.
2994 The macro should return a combination of the @code{DW_EH_PE_*} defines
2995 as found in @file{dwarf2.h}.
2997 If this macro is not defined, pointers will not be encoded but
2998 represented directly.
3000 @findex ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX
3001 @item ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX(@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3002 This macro allows the target to emit whatever special magic is required
3003 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3004 Generic code takes care of pc-relative and indirect encodings; this must
3005 be defined if the target uses text-relative or data-relative encodings.
3007 This is a C statement that branches to @var{done} if the format was
3008 handled. @var{encoding} is the format chosen, @var{size} is the number
3009 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3012 @findex MD_FALLBACK_FRAME_STATE_FOR
3013 @item MD_FALLBACK_FRAME_STATE_FOR(@var{context}, @var{fs}, @var{success})
3014 This macro allows the target to add cpu and operating system specific
3015 code to the call-frame unwinder for use when there is no unwind data
3016 available. The most common reason to implement this macro is to unwind
3017 through signal frames.
3019 This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c}
3020 and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3021 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3022 for the address of the code being executed and @code{context->cfa} for
3023 the stack pointer value. If the frame can be decoded, the register save
3024 addresses should be updated in @var{fs} and the macro should branch to
3025 @var{success}. If the frame cannot be decoded, the macro should do
3029 @node Stack Checking
3030 @subsection Specifying How Stack Checking is Done
3032 GCC will check that stack references are within the boundaries of
3033 the stack, if the @option{-fstack-check} is specified, in one of three ways:
3037 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3038 will assume that you have arranged for stack checking to be done at
3039 appropriate places in the configuration files, e.g., in
3040 @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special
3044 If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern
3045 called @code{check_stack} in your @file{md} file, GCC will call that
3046 pattern with one argument which is the address to compare the stack
3047 value against. You must arrange for this pattern to report an error if
3048 the stack pointer is out of range.
3051 If neither of the above are true, GCC will generate code to periodically
3052 ``probe'' the stack pointer using the values of the macros defined below.
3055 Normally, you will use the default values of these macros, so GCC
3056 will use the third approach.
3059 @findex STACK_CHECK_BUILTIN
3060 @item STACK_CHECK_BUILTIN
3061 A nonzero value if stack checking is done by the configuration files in a
3062 machine-dependent manner. You should define this macro if stack checking
3063 is require by the ABI of your machine or if you would like to have to stack
3064 checking in some more efficient way than GCC's portable approach.
3065 The default value of this macro is zero.
3067 @findex STACK_CHECK_PROBE_INTERVAL
3068 @item STACK_CHECK_PROBE_INTERVAL
3069 An integer representing the interval at which GCC must generate stack
3070 probe instructions. You will normally define this macro to be no larger
3071 than the size of the ``guard pages'' at the end of a stack area. The
3072 default value of 4096 is suitable for most systems.
3074 @findex STACK_CHECK_PROBE_LOAD
3075 @item STACK_CHECK_PROBE_LOAD
3076 A integer which is nonzero if GCC should perform the stack probe
3077 as a load instruction and zero if GCC should use a store instruction.
3078 The default is zero, which is the most efficient choice on most systems.
3080 @findex STACK_CHECK_PROTECT
3081 @item STACK_CHECK_PROTECT
3082 The number of bytes of stack needed to recover from a stack overflow,
3083 for languages where such a recovery is supported. The default value of
3084 75 words should be adequate for most machines.
3086 @findex STACK_CHECK_MAX_FRAME_SIZE
3087 @item STACK_CHECK_MAX_FRAME_SIZE
3088 The maximum size of a stack frame, in bytes. GCC will generate probe
3089 instructions in non-leaf functions to ensure at least this many bytes of
3090 stack are available. If a stack frame is larger than this size, stack
3091 checking will not be reliable and GCC will issue a warning. The
3092 default is chosen so that GCC only generates one instruction on most
3093 systems. You should normally not change the default value of this macro.
3095 @findex STACK_CHECK_FIXED_FRAME_SIZE
3096 @item STACK_CHECK_FIXED_FRAME_SIZE
3097 GCC uses this value to generate the above warning message. It
3098 represents the amount of fixed frame used by a function, not including
3099 space for any callee-saved registers, temporaries and user variables.
3100 You need only specify an upper bound for this amount and will normally
3101 use the default of four words.
3103 @findex STACK_CHECK_MAX_VAR_SIZE
3104 @item STACK_CHECK_MAX_VAR_SIZE
3105 The maximum size, in bytes, of an object that GCC will place in the
3106 fixed area of the stack frame when the user specifies
3107 @option{-fstack-check}.
3108 GCC computed the default from the values of the above macros and you will
3109 normally not need to override that default.
3113 @node Frame Registers
3114 @subsection Registers That Address the Stack Frame
3116 @c prevent bad page break with this line
3117 This discusses registers that address the stack frame.
3120 @findex STACK_POINTER_REGNUM
3121 @item STACK_POINTER_REGNUM
3122 The register number of the stack pointer register, which must also be a
3123 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3124 the hardware determines which register this is.
3126 @findex FRAME_POINTER_REGNUM
3127 @item FRAME_POINTER_REGNUM
3128 The register number of the frame pointer register, which is used to
3129 access automatic variables in the stack frame. On some machines, the
3130 hardware determines which register this is. On other machines, you can
3131 choose any register you wish for this purpose.
3133 @findex HARD_FRAME_POINTER_REGNUM
3134 @item HARD_FRAME_POINTER_REGNUM
3135 On some machines the offset between the frame pointer and starting
3136 offset of the automatic variables is not known until after register
3137 allocation has been done (for example, because the saved registers are
3138 between these two locations). On those machines, define
3139 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3140 be used internally until the offset is known, and define
3141 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3142 used for the frame pointer.
3144 You should define this macro only in the very rare circumstances when it
3145 is not possible to calculate the offset between the frame pointer and
3146 the automatic variables until after register allocation has been
3147 completed. When this macro is defined, you must also indicate in your
3148 definition of @code{ELIMINABLE_REGS} how to eliminate
3149 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3150 or @code{STACK_POINTER_REGNUM}.
3152 Do not define this macro if it would be the same as
3153 @code{FRAME_POINTER_REGNUM}.
3155 @findex ARG_POINTER_REGNUM
3156 @item ARG_POINTER_REGNUM
3157 The register number of the arg pointer register, which is used to access
3158 the function's argument list. On some machines, this is the same as the
3159 frame pointer register. On some machines, the hardware determines which
3160 register this is. On other machines, you can choose any register you
3161 wish for this purpose. If this is not the same register as the frame
3162 pointer register, then you must mark it as a fixed register according to
3163 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3164 (@pxref{Elimination}).
3166 @findex RETURN_ADDRESS_POINTER_REGNUM
3167 @item RETURN_ADDRESS_POINTER_REGNUM
3168 The register number of the return address pointer register, which is used to
3169 access the current function's return address from the stack. On some
3170 machines, the return address is not at a fixed offset from the frame
3171 pointer or stack pointer or argument pointer. This register can be defined
3172 to point to the return address on the stack, and then be converted by
3173 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3175 Do not define this macro unless there is no other way to get the return
3176 address from the stack.
3178 @findex STATIC_CHAIN_REGNUM
3179 @findex STATIC_CHAIN_INCOMING_REGNUM
3180 @item STATIC_CHAIN_REGNUM
3181 @itemx STATIC_CHAIN_INCOMING_REGNUM
3182 Register numbers used for passing a function's static chain pointer. If
3183 register windows are used, the register number as seen by the called
3184 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3185 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3186 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3189 The static chain register need not be a fixed register.
3191 If the static chain is passed in memory, these macros should not be
3192 defined; instead, the next two macros should be defined.
3194 @findex STATIC_CHAIN
3195 @findex STATIC_CHAIN_INCOMING
3197 @itemx STATIC_CHAIN_INCOMING
3198 If the static chain is passed in memory, these macros provide rtx giving
3199 @code{mem} expressions that denote where they are stored.
3200 @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
3201 as seen by the calling and called functions, respectively. Often the former
3202 will be at an offset from the stack pointer and the latter at an offset from
3205 @findex stack_pointer_rtx
3206 @findex frame_pointer_rtx
3207 @findex arg_pointer_rtx
3208 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3209 @code{arg_pointer_rtx} will have been initialized prior to the use of these
3210 macros and should be used to refer to those items.
3212 If the static chain is passed in a register, the two previous macros should
3215 @findex DWARF_FRAME_REGISTERS
3216 @item DWARF_FRAME_REGISTERS
3217 This macro specifies the maximum number of hard registers that can be
3218 saved in a call frame. This is used to size data structures used in
3219 DWARF2 exception handling.
3221 Prior to GCC 3.0, this macro was needed in order to establish a stable
3222 exception handling ABI in the face of adding new hard registers for ISA
3223 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3224 in the number of hard registers. Nevertheless, this macro can still be
3225 used to reduce the runtime memory requirements of the exception handling
3226 routines, which can be substantial if the ISA contains a lot of
3227 registers that are not call-saved.
3229 If this macro is not defined, it defaults to
3230 @code{FIRST_PSEUDO_REGISTER}.
3232 @findex PRE_GCC3_DWARF_FRAME_REGISTERS
3233 @item PRE_GCC3_DWARF_FRAME_REGISTERS
3235 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3236 for backward compatibility in pre GCC 3.0 compiled code.
3238 If this macro is not defined, it defaults to
3239 @code{DWARF_FRAME_REGISTERS}.
3244 @subsection Eliminating Frame Pointer and Arg Pointer
3246 @c prevent bad page break with this line
3247 This is about eliminating the frame pointer and arg pointer.
3250 @findex FRAME_POINTER_REQUIRED
3251 @item FRAME_POINTER_REQUIRED
3252 A C expression which is nonzero if a function must have and use a frame
3253 pointer. This expression is evaluated in the reload pass. If its value is
3254 nonzero the function will have a frame pointer.
3256 The expression can in principle examine the current function and decide
3257 according to the facts, but on most machines the constant 0 or the
3258 constant 1 suffices. Use 0 when the machine allows code to be generated
3259 with no frame pointer, and doing so saves some time or space. Use 1
3260 when there is no possible advantage to avoiding a frame pointer.
3262 In certain cases, the compiler does not know how to produce valid code
3263 without a frame pointer. The compiler recognizes those cases and
3264 automatically gives the function a frame pointer regardless of what
3265 @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
3268 In a function that does not require a frame pointer, the frame pointer
3269 register can be allocated for ordinary usage, unless you mark it as a
3270 fixed register. See @code{FIXED_REGISTERS} for more information.
3272 @findex INITIAL_FRAME_POINTER_OFFSET
3273 @findex get_frame_size
3274 @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3275 A C statement to store in the variable @var{depth-var} the difference
3276 between the frame pointer and the stack pointer values immediately after
3277 the function prologue. The value would be computed from information
3278 such as the result of @code{get_frame_size ()} and the tables of
3279 registers @code{regs_ever_live} and @code{call_used_regs}.
3281 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3282 need not be defined. Otherwise, it must be defined even if
3283 @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
3284 case, you may set @var{depth-var} to anything.
3286 @findex ELIMINABLE_REGS
3287 @item ELIMINABLE_REGS
3288 If defined, this macro specifies a table of register pairs used to
3289 eliminate unneeded registers that point into the stack frame. If it is not
3290 defined, the only elimination attempted by the compiler is to replace
3291 references to the frame pointer with references to the stack pointer.
3293 The definition of this macro is a list of structure initializations, each
3294 of which specifies an original and replacement register.
3296 On some machines, the position of the argument pointer is not known until
3297 the compilation is completed. In such a case, a separate hard register
3298 must be used for the argument pointer. This register can be eliminated by
3299 replacing it with either the frame pointer or the argument pointer,
3300 depending on whether or not the frame pointer has been eliminated.
3302 In this case, you might specify:
3304 #define ELIMINABLE_REGS \
3305 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3306 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3307 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3310 Note that the elimination of the argument pointer with the stack pointer is
3311 specified first since that is the preferred elimination.
3313 @findex CAN_ELIMINATE
3314 @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
3315 A C expression that returns nonzero if the compiler is allowed to try
3316 to replace register number @var{from-reg} with register number
3317 @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
3318 is defined, and will usually be the constant 1, since most of the cases
3319 preventing register elimination are things that the compiler already
3322 @findex INITIAL_ELIMINATION_OFFSET
3323 @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3324 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3325 specifies the initial difference between the specified pair of
3326 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3330 @node Stack Arguments
3331 @subsection Passing Function Arguments on the Stack
3332 @cindex arguments on stack
3333 @cindex stack arguments
3335 The macros in this section control how arguments are passed
3336 on the stack. See the following section for other macros that
3337 control passing certain arguments in registers.
3340 @findex PROMOTE_PROTOTYPES
3341 @item PROMOTE_PROTOTYPES
3342 A C expression whose value is nonzero if an argument declared in
3343 a prototype as an integral type smaller than @code{int} should
3344 actually be passed as an @code{int}. In addition to avoiding
3345 errors in certain cases of mismatch, it also makes for better
3346 code on certain machines. If the macro is not defined in target
3347 header files, it defaults to 0.
3351 A C expression. If nonzero, push insns will be used to pass
3353 If the target machine does not have a push instruction, set it to zero.
3354 That directs GCC to use an alternate strategy: to
3355 allocate the entire argument block and then store the arguments into
3356 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3358 @findex PUSH_ROUNDING
3359 @item PUSH_ROUNDING (@var{npushed})
3360 A C expression that is the number of bytes actually pushed onto the
3361 stack when an instruction attempts to push @var{npushed} bytes.
3363 On some machines, the definition
3366 #define PUSH_ROUNDING(BYTES) (BYTES)
3370 will suffice. But on other machines, instructions that appear
3371 to push one byte actually push two bytes in an attempt to maintain
3372 alignment. Then the definition should be
3375 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3378 @findex ACCUMULATE_OUTGOING_ARGS
3379 @findex current_function_outgoing_args_size
3380 @item ACCUMULATE_OUTGOING_ARGS
3381 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3382 will be computed and placed into the variable
3383 @code{current_function_outgoing_args_size}. No space will be pushed
3384 onto the stack for each call; instead, the function prologue should
3385 increase the stack frame size by this amount.
3387 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3390 @findex REG_PARM_STACK_SPACE
3391 @item REG_PARM_STACK_SPACE (@var{fndecl})
3392 Define this macro if functions should assume that stack space has been
3393 allocated for arguments even when their values are passed in
3396 The value of this macro is the size, in bytes, of the area reserved for
3397 arguments passed in registers for the function represented by @var{fndecl},
3398 which can be zero if GCC is calling a library function.
3400 This space can be allocated by the caller, or be a part of the
3401 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3403 @c above is overfull. not sure what to do. --mew 5feb93 did
3404 @c something, not sure if it looks good. --mew 10feb93
3406 @findex MAYBE_REG_PARM_STACK_SPACE
3407 @findex FINAL_REG_PARM_STACK_SPACE
3408 @item MAYBE_REG_PARM_STACK_SPACE
3409 @itemx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
3410 Define these macros in addition to the one above if functions might
3411 allocate stack space for arguments even when their values are passed
3412 in registers. These should be used when the stack space allocated
3413 for arguments in registers is not a simple constant independent of the
3414 function declaration.
3416 The value of the first macro is the size, in bytes, of the area that
3417 we should initially assume would be reserved for arguments passed in registers.
3419 The value of the second macro is the actual size, in bytes, of the area
3420 that will be reserved for arguments passed in registers. This takes two
3421 arguments: an integer representing the number of bytes of fixed sized
3422 arguments on the stack, and a tree representing the number of bytes of
3423 variable sized arguments on the stack.
3425 When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
3426 called for libcall functions, the current function, or for a function
3427 being called when it is known that such stack space must be allocated.
3428 In each case this value can be easily computed.
3430 When deciding whether a called function needs such stack space, and how
3431 much space to reserve, GCC uses these two macros instead of
3432 @code{REG_PARM_STACK_SPACE}.
3434 @findex OUTGOING_REG_PARM_STACK_SPACE
3435 @item OUTGOING_REG_PARM_STACK_SPACE
3436 Define this if it is the responsibility of the caller to allocate the area
3437 reserved for arguments passed in registers.
3439 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3440 whether the space for these arguments counts in the value of
3441 @code{current_function_outgoing_args_size}.
3443 @findex STACK_PARMS_IN_REG_PARM_AREA
3444 @item STACK_PARMS_IN_REG_PARM_AREA
3445 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3446 stack parameters don't skip the area specified by it.
3447 @c i changed this, makes more sens and it should have taken care of the
3448 @c overfull.. not as specific, tho. --mew 5feb93
3450 Normally, when a parameter is not passed in registers, it is placed on the
3451 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3452 suppresses this behavior and causes the parameter to be passed on the
3453 stack in its natural location.
3455 @findex RETURN_POPS_ARGS
3456 @item RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3457 A C expression that should indicate the number of bytes of its own
3458 arguments that a function pops on returning, or 0 if the
3459 function pops no arguments and the caller must therefore pop them all
3460 after the function returns.
3462 @var{fundecl} is a C variable whose value is a tree node that describes
3463 the function in question. Normally it is a node of type
3464 @code{FUNCTION_DECL} that describes the declaration of the function.
3465 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3467 @var{funtype} is a C variable whose value is a tree node that
3468 describes the function in question. Normally it is a node of type
3469 @code{FUNCTION_TYPE} that describes the data type of the function.
3470 From this it is possible to obtain the data types of the value and
3471 arguments (if known).
3473 When a call to a library function is being considered, @var{fundecl}
3474 will contain an identifier node for the library function. Thus, if
3475 you need to distinguish among various library functions, you can do so
3476 by their names. Note that ``library function'' in this context means
3477 a function used to perform arithmetic, whose name is known specially
3478 in the compiler and was not mentioned in the C code being compiled.
3480 @var{stack-size} is the number of bytes of arguments passed on the
3481 stack. If a variable number of bytes is passed, it is zero, and
3482 argument popping will always be the responsibility of the calling function.
3484 On the VAX, all functions always pop their arguments, so the definition
3485 of this macro is @var{stack-size}. On the 68000, using the standard
3486 calling convention, no functions pop their arguments, so the value of
3487 the macro is always 0 in this case. But an alternative calling
3488 convention is available in which functions that take a fixed number of
3489 arguments pop them but other functions (such as @code{printf}) pop
3490 nothing (the caller pops all). When this convention is in use,
3491 @var{funtype} is examined to determine whether a function takes a fixed
3492 number of arguments.
3494 @findex CALL_POPS_ARGS
3495 @item CALL_POPS_ARGS (@var{cum})
3496 A C expression that should indicate the number of bytes a call sequence
3497 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3498 when compiling a function call.
3500 @var{cum} is the variable in which all arguments to the called function
3501 have been accumulated.
3503 On certain architectures, such as the SH5, a call trampoline is used
3504 that pops certain registers off the stack, depending on the arguments
3505 that have been passed to the function. Since this is a property of the
3506 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3511 @node Register Arguments
3512 @subsection Passing Arguments in Registers
3513 @cindex arguments in registers
3514 @cindex registers arguments
3516 This section describes the macros which let you control how various
3517 types of arguments are passed in registers or how they are arranged in
3521 @findex FUNCTION_ARG
3522 @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3523 A C expression that controls whether a function argument is passed
3524 in a register, and which register.
3526 The arguments are @var{cum}, which summarizes all the previous
3527 arguments; @var{mode}, the machine mode of the argument; @var{type},
3528 the data type of the argument as a tree node or 0 if that is not known
3529 (which happens for C support library functions); and @var{named},
3530 which is 1 for an ordinary argument and 0 for nameless arguments that
3531 correspond to @samp{@dots{}} in the called function's prototype.
3532 @var{type} can be an incomplete type if a syntax error has previously
3535 The value of the expression is usually either a @code{reg} RTX for the
3536 hard register in which to pass the argument, or zero to pass the
3537 argument on the stack.
3539 For machines like the VAX and 68000, where normally all arguments are
3540 pushed, zero suffices as a definition.
3542 The value of the expression can also be a @code{parallel} RTX@. This is
3543 used when an argument is passed in multiple locations. The mode of the
3544 of the @code{parallel} should be the mode of the entire argument. The
3545 @code{parallel} holds any number of @code{expr_list} pairs; each one
3546 describes where part of the argument is passed. In each
3547 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3548 register in which to pass this part of the argument, and the mode of the
3549 register RTX indicates how large this part of the argument is. The
3550 second operand of the @code{expr_list} is a @code{const_int} which gives
3551 the offset in bytes into the entire argument of where this part starts.
3552 As a special exception the first @code{expr_list} in the @code{parallel}
3553 RTX may have a first operand of zero. This indicates that the entire
3554 argument is also stored on the stack.
3556 The last time this macro is called, it is called with @code{MODE ==
3557 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3558 pattern as operands 2 and 3 respectively.
3560 @cindex @file{stdarg.h} and register arguments
3561 The usual way to make the ISO library @file{stdarg.h} work on a machine
3562 where some arguments are usually passed in registers, is to cause
3563 nameless arguments to be passed on the stack instead. This is done
3564 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
3566 @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
3567 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
3568 You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
3569 in the definition of this macro to determine if this argument is of a
3570 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3571 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
3572 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3573 defined, the argument will be computed in the stack and then loaded into
3576 @findex MUST_PASS_IN_STACK
3577 @item MUST_PASS_IN_STACK (@var{mode}, @var{type})
3578 Define as a C expression that evaluates to nonzero if we do not know how
3579 to pass TYPE solely in registers. The file @file{expr.h} defines a
3580 definition that is usually appropriate, refer to @file{expr.h} for additional
3583 @findex FUNCTION_INCOMING_ARG
3584 @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
3585 Define this macro if the target machine has ``register windows'', so
3586 that the register in which a function sees an arguments is not
3587 necessarily the same as the one in which the caller passed the
3590 For such machines, @code{FUNCTION_ARG} computes the register in which
3591 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
3592 be defined in a similar fashion to tell the function being called
3593 where the arguments will arrive.
3595 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
3596 serves both purposes.
3598 @findex FUNCTION_ARG_PARTIAL_NREGS
3599 @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
3600 A C expression for the number of words, at the beginning of an
3601 argument, that must be put in registers. The value must be zero for
3602 arguments that are passed entirely in registers or that are entirely
3603 pushed on the stack.
3605 On some machines, certain arguments must be passed partially in
3606 registers and partially in memory. On these machines, typically the
3607 first @var{n} words of arguments are passed in registers, and the rest
3608 on the stack. If a multi-word argument (a @code{double} or a
3609 structure) crosses that boundary, its first few words must be passed
3610 in registers and the rest must be pushed. This macro tells the
3611 compiler when this occurs, and how many of the words should go in
3614 @code{FUNCTION_ARG} for these arguments should return the first
3615 register to be used by the caller for this argument; likewise
3616 @code{FUNCTION_INCOMING_ARG}, for the called function.
3618 @findex FUNCTION_ARG_PASS_BY_REFERENCE
3619 @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3620 A C expression that indicates when an argument must be passed by reference.
3621 If nonzero for an argument, a copy of that argument is made in memory and a
3622 pointer to the argument is passed instead of the argument itself.
3623 The pointer is passed in whatever way is appropriate for passing a pointer
3626 On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
3627 definition of this macro might be
3629 #define FUNCTION_ARG_PASS_BY_REFERENCE\
3630 (CUM, MODE, TYPE, NAMED) \
3631 MUST_PASS_IN_STACK (MODE, TYPE)
3633 @c this is *still* too long. --mew 5feb93
3635 @findex FUNCTION_ARG_CALLEE_COPIES
3636 @item FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
3637 If defined, a C expression that indicates when it is the called function's
3638 responsibility to make a copy of arguments passed by invisible reference.
3639 Normally, the caller makes a copy and passes the address of the copy to the
3640 routine being called. When @code{FUNCTION_ARG_CALLEE_COPIES} is defined and is
3641 nonzero, the caller does not make a copy. Instead, it passes a pointer to the
3642 ``live'' value. The called function must not modify this value. If it can be
3643 determined that the value won't be modified, it need not make a copy;
3644 otherwise a copy must be made.
3646 @findex CUMULATIVE_ARGS
3647 @item CUMULATIVE_ARGS
3648 A C type for declaring a variable that is used as the first argument of
3649 @code{FUNCTION_ARG} and other related values. For some target machines,
3650 the type @code{int} suffices and can hold the number of bytes of
3653 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3654 arguments that have been passed on the stack. The compiler has other
3655 variables to keep track of that. For target machines on which all
3656 arguments are passed on the stack, there is no need to store anything in
3657 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3658 should not be empty, so use @code{int}.
3660 @findex INIT_CUMULATIVE_ARGS
3661 @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{indirect})
3662 A C statement (sans semicolon) for initializing the variable @var{cum}
3663 for the state at the beginning of the argument list. The variable has
3664 type @code{CUMULATIVE_ARGS}. The value of @var{fntype} is the tree node
3665 for the data type of the function which will receive the args, or 0
3666 if the args are to a compiler support library function. The value of
3667 @var{indirect} is nonzero when processing an indirect call, for example
3668 a call through a function pointer. The value of @var{indirect} is zero
3669 for a call to an explicitly named function, a library function call, or when
3670 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3673 When processing a call to a compiler support library function,
3674 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3675 contains the name of the function, as a string. @var{libname} is 0 when
3676 an ordinary C function call is being processed. Thus, each time this
3677 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3678 never both of them at once.
3680 @findex INIT_CUMULATIVE_LIBCALL_ARGS
3681 @item INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3682 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3683 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3684 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3685 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3686 0)} is used instead.
3688 @findex INIT_CUMULATIVE_INCOMING_ARGS
3689 @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3690 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3691 finding the arguments for the function being compiled. If this macro is
3692 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3694 The value passed for @var{libname} is always 0, since library routines
3695 with special calling conventions are never compiled with GCC@. The
3696 argument @var{libname} exists for symmetry with
3697 @code{INIT_CUMULATIVE_ARGS}.
3698 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3699 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3701 @findex FUNCTION_ARG_ADVANCE
3702 @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
3703 A C statement (sans semicolon) to update the summarizer variable
3704 @var{cum} to advance past an argument in the argument list. The
3705 values @var{mode}, @var{type} and @var{named} describe that argument.
3706 Once this is done, the variable @var{cum} is suitable for analyzing
3707 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
3709 This macro need not do anything if the argument in question was passed
3710 on the stack. The compiler knows how to track the amount of stack space
3711 used for arguments without any special help.
3713 @findex FUNCTION_ARG_PADDING
3714 @item FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3715 If defined, a C expression which determines whether, and in which direction,
3716 to pad out an argument with extra space. The value should be of type
3717 @code{enum direction}: either @code{upward} to pad above the argument,
3718 @code{downward} to pad below, or @code{none} to inhibit padding.
3720 The @emph{amount} of padding is always just enough to reach the next
3721 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
3724 This macro has a default definition which is right for most systems.
3725 For little-endian machines, the default is to pad upward. For
3726 big-endian machines, the default is to pad downward for an argument of
3727 constant size shorter than an @code{int}, and upward otherwise.
3729 @findex PAD_VARARGS_DOWN
3730 @item PAD_VARARGS_DOWN
3731 If defined, a C expression which determines whether the default
3732 implementation of va_arg will attempt to pad down before reading the
3733 next argument, if that argument is smaller than its aligned space as
3734 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3735 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3737 @findex FUNCTION_ARG_BOUNDARY
3738 @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
3739 If defined, a C expression that gives the alignment boundary, in bits,
3740 of an argument with the specified mode and type. If it is not defined,
3741 @code{PARM_BOUNDARY} is used for all arguments.
3743 @findex FUNCTION_ARG_REGNO_P
3744 @item FUNCTION_ARG_REGNO_P (@var{regno})
3745 A C expression that is nonzero if @var{regno} is the number of a hard
3746 register in which function arguments are sometimes passed. This does
3747 @emph{not} include implicit arguments such as the static chain and
3748 the structure-value address. On many machines, no registers can be
3749 used for this purpose since all function arguments are pushed on the
3752 @findex LOAD_ARGS_REVERSED
3753 @item LOAD_ARGS_REVERSED
3754 If defined, the order in which arguments are loaded into their
3755 respective argument registers is reversed so that the last
3756 argument is loaded first. This macro only affects arguments
3757 passed in registers.
3762 @subsection How Scalar Function Values Are Returned
3763 @cindex return values in registers
3764 @cindex values, returned by functions
3765 @cindex scalars, returned as values
3767 This section discusses the macros that control returning scalars as
3768 values---values that can fit in registers.
3771 @findex FUNCTION_VALUE
3772 @item FUNCTION_VALUE (@var{valtype}, @var{func})
3773 A C expression to create an RTX representing the place where a
3774 function returns a value of data type @var{valtype}. @var{valtype} is
3775 a tree node representing a data type. Write @code{TYPE_MODE
3776 (@var{valtype})} to get the machine mode used to represent that type.
3777 On many machines, only the mode is relevant. (Actually, on most
3778 machines, scalar values are returned in the same place regardless of
3781 The value of the expression is usually a @code{reg} RTX for the hard
3782 register where the return value is stored. The value can also be a
3783 @code{parallel} RTX, if the return value is in multiple places. See
3784 @code{FUNCTION_ARG} for an explanation of the @code{parallel} form.
3786 If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
3787 promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
3790 If the precise function being called is known, @var{func} is a tree
3791 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3792 pointer. This makes it possible to use a different value-returning
3793 convention for specific functions when all their calls are
3796 @code{FUNCTION_VALUE} is not used for return vales with aggregate data
3797 types, because these are returned in another way. See
3798 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3800 @findex FUNCTION_OUTGOING_VALUE
3801 @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
3802 Define this macro if the target machine has ``register windows''
3803 so that the register in which a function returns its value is not
3804 the same as the one in which the caller sees the value.
3806 For such machines, @code{FUNCTION_VALUE} computes the register in which
3807 the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
3808 defined in a similar fashion to tell the function where to put the
3811 If @code{FUNCTION_OUTGOING_VALUE} is not defined,
3812 @code{FUNCTION_VALUE} serves both purposes.
3814 @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
3815 aggregate data types, because these are returned in another way. See
3816 @code{STRUCT_VALUE_REGNUM} and related macros, below.
3818 @findex LIBCALL_VALUE
3819 @item LIBCALL_VALUE (@var{mode})
3820 A C expression to create an RTX representing the place where a library
3821 function returns a value of mode @var{mode}. If the precise function
3822 being called is known, @var{func} is a tree node
3823 (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
3824 pointer. This makes it possible to use a different value-returning
3825 convention for specific functions when all their calls are
3828 Note that ``library function'' in this context means a compiler
3829 support routine, used to perform arithmetic, whose name is known
3830 specially by the compiler and was not mentioned in the C code being
3833 The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
3834 data types, because none of the library functions returns such types.
3836 @findex FUNCTION_VALUE_REGNO_P
3837 @item FUNCTION_VALUE_REGNO_P (@var{regno})
3838 A C expression that is nonzero if @var{regno} is the number of a hard
3839 register in which the values of called function may come back.
3841 A register whose use for returning values is limited to serving as the
3842 second of a pair (for a value of type @code{double}, say) need not be
3843 recognized by this macro. So for most machines, this definition
3847 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3850 If the machine has register windows, so that the caller and the called
3851 function use different registers for the return value, this macro
3852 should recognize only the caller's register numbers.
3854 @findex APPLY_RESULT_SIZE
3855 @item APPLY_RESULT_SIZE
3856 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3857 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3858 saving and restoring an arbitrary return value.
3861 @node Aggregate Return
3862 @subsection How Large Values Are Returned
3863 @cindex aggregates as return values
3864 @cindex large return values
3865 @cindex returning aggregate values
3866 @cindex structure value address
3868 When a function value's mode is @code{BLKmode} (and in some other
3869 cases), the value is not returned according to @code{FUNCTION_VALUE}
3870 (@pxref{Scalar Return}). Instead, the caller passes the address of a
3871 block of memory in which the value should be stored. This address
3872 is called the @dfn{structure value address}.
3874 This section describes how to control returning structure values in
3878 @findex RETURN_IN_MEMORY
3879 @item RETURN_IN_MEMORY (@var{type})
3880 A C expression which can inhibit the returning of certain function
3881 values in registers, based on the type of value. A nonzero value says
3882 to return the function value in memory, just as large structures are
3883 always returned. Here @var{type} will be a C expression of type
3884 @code{tree}, representing the data type of the value.
3886 Note that values of mode @code{BLKmode} must be explicitly handled
3887 by this macro. Also, the option @option{-fpcc-struct-return}
3888 takes effect regardless of this macro. On most systems, it is
3889 possible to leave the macro undefined; this causes a default
3890 definition to be used, whose value is the constant 1 for @code{BLKmode}
3891 values, and 0 otherwise.
3893 Do not use this macro to indicate that structures and unions should always
3894 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
3897 @findex DEFAULT_PCC_STRUCT_RETURN
3898 @item DEFAULT_PCC_STRUCT_RETURN
3899 Define this macro to be 1 if all structure and union return values must be
3900 in memory. Since this results in slower code, this should be defined
3901 only if needed for compatibility with other compilers or with an ABI@.
3902 If you define this macro to be 0, then the conventions used for structure
3903 and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
3905 If not defined, this defaults to the value 1.
3907 @findex STRUCT_VALUE_REGNUM
3908 @item STRUCT_VALUE_REGNUM
3909 If the structure value address is passed in a register, then
3910 @code{STRUCT_VALUE_REGNUM} should be the number of that register.
3912 @findex STRUCT_VALUE
3914 If the structure value address is not passed in a register, define
3915 @code{STRUCT_VALUE} as an expression returning an RTX for the place
3916 where the address is passed. If it returns 0, the address is passed as
3917 an ``invisible'' first argument.
3919 @findex STRUCT_VALUE_INCOMING_REGNUM
3920 @item STRUCT_VALUE_INCOMING_REGNUM
3921 On some architectures the place where the structure value address
3922 is found by the called function is not the same place that the
3923 caller put it. This can be due to register windows, or it could
3924 be because the function prologue moves it to a different place.
3926 If the incoming location of the structure value address is in a
3927 register, define this macro as the register number.
3929 @findex STRUCT_VALUE_INCOMING
3930 @item STRUCT_VALUE_INCOMING
3931 If the incoming location is not a register, then you should define
3932 @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the
3933 called function should find the value. If it should find the value on
3934 the stack, define this to create a @code{mem} which refers to the frame
3935 pointer. A definition of 0 means that the address is passed as an
3936 ``invisible'' first argument.
3938 @findex PCC_STATIC_STRUCT_RETURN
3939 @item PCC_STATIC_STRUCT_RETURN
3940 Define this macro if the usual system convention on the target machine
3941 for returning structures and unions is for the called function to return
3942 the address of a static variable containing the value.
3944 Do not define this if the usual system convention is for the caller to
3945 pass an address to the subroutine.
3947 This macro has effect in @option{-fpcc-struct-return} mode, but it does
3948 nothing when you use @option{-freg-struct-return} mode.
3952 @subsection Caller-Saves Register Allocation
3954 If you enable it, GCC can save registers around function calls. This
3955 makes it possible to use call-clobbered registers to hold variables that
3956 must live across calls.
3959 @findex DEFAULT_CALLER_SAVES
3960 @item DEFAULT_CALLER_SAVES
3961 Define this macro if function calls on the target machine do not preserve
3962 any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
3963 for all registers. When defined, this macro enables @option{-fcaller-saves}
3964 by default for all optimization levels. It has no effect for optimization
3965 levels 2 and higher, where @option{-fcaller-saves} is the default.
3967 @findex CALLER_SAVE_PROFITABLE
3968 @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
3969 A C expression to determine whether it is worthwhile to consider placing
3970 a pseudo-register in a call-clobbered hard register and saving and
3971 restoring it around each function call. The expression should be 1 when
3972 this is worth doing, and 0 otherwise.
3974 If you don't define this macro, a default is used which is good on most
3975 machines: @code{4 * @var{calls} < @var{refs}}.
3977 @findex HARD_REGNO_CALLER_SAVE_MODE
3978 @item HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3979 A C expression specifying which mode is required for saving @var{nregs}
3980 of a pseudo-register in call-clobbered hard register @var{regno}. If
3981 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3982 returned. For most machines this macro need not be defined since GCC
3983 will select the smallest suitable mode.
3986 @node Function Entry
3987 @subsection Function Entry and Exit
3988 @cindex function entry and exit
3992 This section describes the macros that output function entry
3993 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3995 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
3996 If defined, a function that outputs the assembler code for entry to a
3997 function. The prologue is responsible for setting up the stack frame,
3998 initializing the frame pointer register, saving registers that must be
3999 saved, and allocating @var{size} additional bytes of storage for the
4000 local variables. @var{size} is an integer. @var{file} is a stdio
4001 stream to which the assembler code should be output.
4003 The label for the beginning of the function need not be output by this
4004 macro. That has already been done when the macro is run.
4006 @findex regs_ever_live
4007 To determine which registers to save, the macro can refer to the array
4008 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4009 @var{r} is used anywhere within the function. This implies the function
4010 prologue should save register @var{r}, provided it is not one of the
4011 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4012 @code{regs_ever_live}.)
4014 On machines that have ``register windows'', the function entry code does
4015 not save on the stack the registers that are in the windows, even if
4016 they are supposed to be preserved by function calls; instead it takes
4017 appropriate steps to ``push'' the register stack, if any non-call-used
4018 registers are used in the function.
4020 @findex frame_pointer_needed
4021 On machines where functions may or may not have frame-pointers, the
4022 function entry code must vary accordingly; it must set up the frame
4023 pointer if one is wanted, and not otherwise. To determine whether a
4024 frame pointer is in wanted, the macro can refer to the variable
4025 @code{frame_pointer_needed}. The variable's value will be 1 at run
4026 time in a function that needs a frame pointer. @xref{Elimination}.
4028 The function entry code is responsible for allocating any stack space
4029 required for the function. This stack space consists of the regions
4030 listed below. In most cases, these regions are allocated in the
4031 order listed, with the last listed region closest to the top of the
4032 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4033 the highest address if it is not defined). You can use a different order
4034 for a machine if doing so is more convenient or required for
4035 compatibility reasons. Except in cases where required by standard
4036 or by a debugger, there is no reason why the stack layout used by GCC
4037 need agree with that used by other compilers for a machine.
4040 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4041 If defined, a function that outputs assembler code at the end of a
4042 prologue. This should be used when the function prologue is being
4043 emitted as RTL, and you have some extra assembler that needs to be
4044 emitted. @xref{prologue instruction pattern}.
4047 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4048 If defined, a function that outputs assembler code at the start of an
4049 epilogue. This should be used when the function epilogue is being
4050 emitted as RTL, and you have some extra assembler that needs to be
4051 emitted. @xref{epilogue instruction pattern}.
4054 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4055 If defined, a function that outputs the assembler code for exit from a
4056 function. The epilogue is responsible for restoring the saved
4057 registers and stack pointer to their values when the function was
4058 called, and returning control to the caller. This macro takes the
4059 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4060 registers to restore are determined from @code{regs_ever_live} and
4061 @code{CALL_USED_REGISTERS} in the same way.
4063 On some machines, there is a single instruction that does all the work
4064 of returning from the function. On these machines, give that
4065 instruction the name @samp{return} and do not define the macro
4066 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4068 Do not define a pattern named @samp{return} if you want the
4069 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4070 switches to control whether return instructions or epilogues are used,
4071 define a @samp{return} pattern with a validity condition that tests the
4072 target switches appropriately. If the @samp{return} pattern's validity
4073 condition is false, epilogues will be used.
4075 On machines where functions may or may not have frame-pointers, the
4076 function exit code must vary accordingly. Sometimes the code for these
4077 two cases is completely different. To determine whether a frame pointer
4078 is wanted, the macro can refer to the variable
4079 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4080 a function that needs a frame pointer.
4082 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4083 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4084 The C variable @code{current_function_is_leaf} is nonzero for such a
4085 function. @xref{Leaf Functions}.
4087 On some machines, some functions pop their arguments on exit while
4088 others leave that for the caller to do. For example, the 68020 when
4089 given @option{-mrtd} pops arguments in functions that take a fixed
4090 number of arguments.
4092 @findex current_function_pops_args
4093 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4094 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4095 needs to know what was decided. The variable that is called
4096 @code{current_function_pops_args} is the number of bytes of its
4097 arguments that a function should pop. @xref{Scalar Return}.
4098 @c what is the "its arguments" in the above sentence referring to, pray
4099 @c tell? --mew 5feb93
4106 @findex current_function_pretend_args_size
4107 A region of @code{current_function_pretend_args_size} bytes of
4108 uninitialized space just underneath the first argument arriving on the
4109 stack. (This may not be at the very start of the allocated stack region
4110 if the calling sequence has pushed anything else since pushing the stack
4111 arguments. But usually, on such machines, nothing else has been pushed
4112 yet, because the function prologue itself does all the pushing.) This
4113 region is used on machines where an argument may be passed partly in
4114 registers and partly in memory, and, in some cases to support the
4115 features in @code{<stdarg.h>}.
4118 An area of memory used to save certain registers used by the function.
4119 The size of this area, which may also include space for such things as
4120 the return address and pointers to previous stack frames, is
4121 machine-specific and usually depends on which registers have been used
4122 in the function. Machines with register windows often do not require
4126 A region of at least @var{size} bytes, possibly rounded up to an allocation
4127 boundary, to contain the local variables of the function. On some machines,
4128 this region and the save area may occur in the opposite order, with the
4129 save area closer to the top of the stack.
4132 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4133 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4134 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4135 argument lists of the function. @xref{Stack Arguments}.
4138 Normally, it is necessary for the macros
4139 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4140 @code{TARGET_ASM_FUNCTION_EPILOGUE} to treat leaf functions specially.
4141 The C variable @code{current_function_is_leaf} is nonzero for such a
4144 @findex EXIT_IGNORE_STACK
4145 @item EXIT_IGNORE_STACK
4146 Define this macro as a C expression that is nonzero if the return
4147 instruction or the function epilogue ignores the value of the stack
4148 pointer; in other words, if it is safe to delete an instruction to
4149 adjust the stack pointer before a return from the function.
4151 Note that this macro's value is relevant only for functions for which
4152 frame pointers are maintained. It is never safe to delete a final
4153 stack adjustment in a function that has no frame pointer, and the
4154 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4156 @findex EPILOGUE_USES
4157 @item EPILOGUE_USES (@var{regno})
4158 Define this macro as a C expression that is nonzero for registers that are
4159 used by the epilogue or the @samp{return} pattern. The stack and frame
4160 pointer registers are already be assumed to be used as needed.
4163 @item EH_USES (@var{regno})
4164 Define this macro as a C expression that is nonzero for registers that are
4165 used by the exception handling mechanism, and so should be considered live
4166 on entry to an exception edge.
4168 @findex DELAY_SLOTS_FOR_EPILOGUE
4169 @item DELAY_SLOTS_FOR_EPILOGUE
4170 Define this macro if the function epilogue contains delay slots to which
4171 instructions from the rest of the function can be ``moved''. The
4172 definition should be a C expression whose value is an integer
4173 representing the number of delay slots there.
4175 @findex ELIGIBLE_FOR_EPILOGUE_DELAY
4176 @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4177 A C expression that returns 1 if @var{insn} can be placed in delay
4178 slot number @var{n} of the epilogue.
4180 The argument @var{n} is an integer which identifies the delay slot now
4181 being considered (since different slots may have different rules of
4182 eligibility). It is never negative and is always less than the number
4183 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4184 If you reject a particular insn for a given delay slot, in principle, it
4185 may be reconsidered for a subsequent delay slot. Also, other insns may
4186 (at least in principle) be considered for the so far unfilled delay
4189 @findex current_function_epilogue_delay_list
4190 @findex final_scan_insn
4191 The insns accepted to fill the epilogue delay slots are put in an RTL
4192 list made with @code{insn_list} objects, stored in the variable
4193 @code{current_function_epilogue_delay_list}. The insn for the first
4194 delay slot comes first in the list. Your definition of the macro
4195 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4196 outputting the insns in this list, usually by calling
4197 @code{final_scan_insn}.
4199 You need not define this macro if you did not define
4200 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4204 @findex TARGET_ASM_OUTPUT_MI_THUNK
4205 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, tree @var{function})
4206 A function that outputs the assembler code for a thunk
4207 function, used to implement C++ virtual function calls with multiple
4208 inheritance. The thunk acts as a wrapper around a virtual function,
4209 adjusting the implicit object parameter before handing control off to
4212 First, emit code to add the integer @var{delta} to the location that
4213 contains the incoming first argument. Assume that this argument
4214 contains a pointer, and is the one used to pass the @code{this} pointer
4215 in C++. This is the incoming argument @emph{before} the function prologue,
4216 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4217 all other incoming arguments.
4219 After the addition, emit code to jump to @var{function}, which is a
4220 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4221 not touch the return address. Hence returning from @var{FUNCTION} will
4222 return to whoever called the current @samp{thunk}.
4224 The effect must be as if @var{function} had been called directly with
4225 the adjusted first argument. This macro is responsible for emitting all
4226 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4227 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4229 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4230 have already been extracted from it.) It might possibly be useful on
4231 some targets, but probably not.
4233 If you do not define this macro, the target-independent code in the C++
4234 front end will generate a less efficient heavyweight thunk that calls
4235 @var{function} instead of jumping to it. The generic approach does
4236 not support varargs.
4239 @findex TARGET_ASM_OUTPUT_MI_VCALL_THUNK
4240 @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})
4241 A function like @code{TARGET_ASM_OUTPUT_MI_THUNK}, except that if
4242 @var{vcall_offset} is nonzero, an additional adjustment should be made
4243 after adding @code{delta}. In particular, if @var{p} is the
4244 adjusted pointer, the following adjustment should be made:
4247 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4251 If this function is defined, it will always be used in place of
4252 @code{TARGET_ASM_OUTPUT_MI_THUNK}.
4257 @subsection Generating Code for Profiling
4258 @cindex profiling, code generation
4260 These macros will help you generate code for profiling.
4263 @findex FUNCTION_PROFILER
4264 @item FUNCTION_PROFILER (@var{file}, @var{labelno})
4265 A C statement or compound statement to output to @var{file} some
4266 assembler code to call the profiling subroutine @code{mcount}.
4269 The details of how @code{mcount} expects to be called are determined by
4270 your operating system environment, not by GCC@. To figure them out,
4271 compile a small program for profiling using the system's installed C
4272 compiler and look at the assembler code that results.
4274 Older implementations of @code{mcount} expect the address of a counter
4275 variable to be loaded into some register. The name of this variable is
4276 @samp{LP} followed by the number @var{labelno}, so you would generate
4277 the name using @samp{LP%d} in a @code{fprintf}.
4279 @findex PROFILE_HOOK
4281 A C statement or compound statement to output to @var{file} some assembly
4282 code to call the profiling subroutine @code{mcount} even the target does
4283 not support profiling.
4285 @findex NO_PROFILE_COUNTERS
4286 @item NO_PROFILE_COUNTERS
4287 Define this macro if the @code{mcount} subroutine on your system does
4288 not need a counter variable allocated for each function. This is true
4289 for almost all modern implementations. If you define this macro, you
4290 must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}.
4292 @findex PROFILE_BEFORE_PROLOGUE
4293 @item PROFILE_BEFORE_PROLOGUE
4294 Define this macro if the code for function profiling should come before
4295 the function prologue. Normally, the profiling code comes after.
4299 @subsection Permitting tail calls
4302 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4303 True if it is ok to do sibling call optimization for the specified
4304 call expression @var{exp}. @var{decl} will be the called function,
4305 or @code{NULL} if this is an indirect call.
4307 It is not uncommon for limitations of calling conventions to prevent
4308 tail calls to functions outside the current unit of translation, or
4309 during PIC compilation. The hook is used to enforce these restrictions,
4310 as the @code{sibcall} md pattern can not fail, or fall over to a
4311 ``normal'' call. The criteria for successful sibling call optimization
4312 may vary greatly between different architectures.
4316 @section Implementing the Varargs Macros
4317 @cindex varargs implementation
4319 GCC comes with an implementation of @code{<varargs.h>} and
4320 @code{<stdarg.h>} that work without change on machines that pass arguments
4321 on the stack. Other machines require their own implementations of
4322 varargs, and the two machine independent header files must have
4323 conditionals to include it.
4325 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4326 the calling convention for @code{va_start}. The traditional
4327 implementation takes just one argument, which is the variable in which
4328 to store the argument pointer. The ISO implementation of
4329 @code{va_start} takes an additional second argument. The user is
4330 supposed to write the last named argument of the function here.
4332 However, @code{va_start} should not use this argument. The way to find
4333 the end of the named arguments is with the built-in functions described
4337 @findex __builtin_saveregs
4338 @item __builtin_saveregs ()
4339 Use this built-in function to save the argument registers in memory so
4340 that the varargs mechanism can access them. Both ISO and traditional
4341 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4342 you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
4344 On some machines, @code{__builtin_saveregs} is open-coded under the
4345 control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
4346 it calls a routine written in assembler language, found in
4349 Code generated for the call to @code{__builtin_saveregs} appears at the
4350 beginning of the function, as opposed to where the call to
4351 @code{__builtin_saveregs} is written, regardless of what the code is.
4352 This is because the registers must be saved before the function starts
4353 to use them for its own purposes.
4354 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4357 @findex __builtin_args_info
4358 @item __builtin_args_info (@var{category})
4359 Use this built-in function to find the first anonymous arguments in
4362 In general, a machine may have several categories of registers used for
4363 arguments, each for a particular category of data types. (For example,
4364 on some machines, floating-point registers are used for floating-point
4365 arguments while other arguments are passed in the general registers.)
4366 To make non-varargs functions use the proper calling convention, you
4367 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4368 registers in each category have been used so far
4370 @code{__builtin_args_info} accesses the same data structure of type
4371 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4372 with it, with @var{category} specifying which word to access. Thus, the
4373 value indicates the first unused register in a given category.
4375 Normally, you would use @code{__builtin_args_info} in the implementation
4376 of @code{va_start}, accessing each category just once and storing the
4377 value in the @code{va_list} object. This is because @code{va_list} will
4378 have to update the values, and there is no way to alter the
4379 values accessed by @code{__builtin_args_info}.
4381 @findex __builtin_next_arg
4382 @item __builtin_next_arg (@var{lastarg})
4383 This is the equivalent of @code{__builtin_args_info}, for stack
4384 arguments. It returns the address of the first anonymous stack
4385 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4386 returns the address of the location above the first anonymous stack
4387 argument. Use it in @code{va_start} to initialize the pointer for
4388 fetching arguments from the stack. Also use it in @code{va_start} to
4389 verify that the second parameter @var{lastarg} is the last named argument
4390 of the current function.
4392 @findex __builtin_classify_type
4393 @item __builtin_classify_type (@var{object})
4394 Since each machine has its own conventions for which data types are
4395 passed in which kind of register, your implementation of @code{va_arg}
4396 has to embody these conventions. The easiest way to categorize the
4397 specified data type is to use @code{__builtin_classify_type} together
4398 with @code{sizeof} and @code{__alignof__}.
4400 @code{__builtin_classify_type} ignores the value of @var{object},
4401 considering only its data type. It returns an integer describing what
4402 kind of type that is---integer, floating, pointer, structure, and so on.
4404 The file @file{typeclass.h} defines an enumeration that you can use to
4405 interpret the values of @code{__builtin_classify_type}.
4408 These machine description macros help implement varargs:
4411 @findex EXPAND_BUILTIN_SAVEREGS
4412 @item EXPAND_BUILTIN_SAVEREGS ()
4413 If defined, is a C expression that produces the machine-specific code
4414 for a call to @code{__builtin_saveregs}. This code will be moved to the
4415 very beginning of the function, before any parameter access are made.
4416 The return value of this function should be an RTX that contains the
4417 value to use as the return of @code{__builtin_saveregs}.
4419 @findex SETUP_INCOMING_VARARGS
4420 @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type}, @var{pretend_args_size}, @var{second_time})
4421 This macro offers an alternative to using @code{__builtin_saveregs} and
4422 defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the
4423 anonymous register arguments into the stack so that all the arguments
4424 appear to have been passed consecutively on the stack. Once this is
4425 done, you can use the standard implementation of varargs that works for
4426 machines that pass all their arguments on the stack.
4428 The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data
4429 structure, containing the values that are obtained after processing the
4430 named arguments. The arguments @var{mode} and @var{type} describe the
4431 last named argument---its machine mode and its data type as a tree node.
4433 The macro implementation should do two things: first, push onto the
4434 stack all the argument registers @emph{not} used for the named
4435 arguments, and second, store the size of the data thus pushed into the
4436 @code{int}-valued variable whose name is supplied as the argument
4437 @var{pretend_args_size}. The value that you store here will serve as
4438 additional offset for setting up the stack frame.
4440 Because you must generate code to push the anonymous arguments at
4441 compile time without knowing their data types,
4442 @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just
4443 a single category of argument register and use it uniformly for all data
4446 If the argument @var{second_time} is nonzero, it means that the
4447 arguments of the function are being analyzed for the second time. This
4448 happens for an inline function, which is not actually compiled until the
4449 end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should
4450 not generate any instructions in this case.
4452 @findex STRICT_ARGUMENT_NAMING
4453 @item STRICT_ARGUMENT_NAMING
4454 Define this macro to be a nonzero value if the location where a function
4455 argument is passed depends on whether or not it is a named argument.
4457 This macro controls how the @var{named} argument to @code{FUNCTION_ARG}
4458 is set for varargs and stdarg functions. If this macro returns a
4459 nonzero value, the @var{named} argument is always true for named
4460 arguments, and false for unnamed arguments. If it returns a value of
4461 zero, but @code{SETUP_INCOMING_VARARGS} is defined, then all arguments
4462 are treated as named. Otherwise, all named arguments except the last
4463 are treated as named.
4465 You need not define this macro if it always returns zero.
4467 @findex PRETEND_OUTGOING_VARARGS_NAMED
4468 @item PRETEND_OUTGOING_VARARGS_NAMED
4469 If you need to conditionally change ABIs so that one works with
4470 @code{SETUP_INCOMING_VARARGS}, but the other works like neither
4471 @code{SETUP_INCOMING_VARARGS} nor @code{STRICT_ARGUMENT_NAMING} was
4472 defined, then define this macro to return nonzero if
4473 @code{SETUP_INCOMING_VARARGS} is used, zero otherwise.
4474 Otherwise, you should not define this macro.
4478 @section Trampolines for Nested Functions
4479 @cindex trampolines for nested functions
4480 @cindex nested functions, trampolines for
4482 A @dfn{trampoline} is a small piece of code that is created at run time
4483 when the address of a nested function is taken. It normally resides on
4484 the stack, in the stack frame of the containing function. These macros
4485 tell GCC how to generate code to allocate and initialize a
4488 The instructions in the trampoline must do two things: load a constant
4489 address into the static chain register, and jump to the real address of
4490 the nested function. On CISC machines such as the m68k, this requires
4491 two instructions, a move immediate and a jump. Then the two addresses
4492 exist in the trampoline as word-long immediate operands. On RISC
4493 machines, it is often necessary to load each address into a register in
4494 two parts. Then pieces of each address form separate immediate
4497 The code generated to initialize the trampoline must store the variable
4498 parts---the static chain value and the function address---into the
4499 immediate operands of the instructions. On a CISC machine, this is
4500 simply a matter of copying each address to a memory reference at the
4501 proper offset from the start of the trampoline. On a RISC machine, it
4502 may be necessary to take out pieces of the address and store them
4506 @findex TRAMPOLINE_TEMPLATE
4507 @item TRAMPOLINE_TEMPLATE (@var{file})
4508 A C statement to output, on the stream @var{file}, assembler code for a
4509 block of data that contains the constant parts of a trampoline. This
4510 code should not include a label---the label is taken care of
4513 If you do not define this macro, it means no template is needed
4514 for the target. Do not define this macro on systems where the block move
4515 code to copy the trampoline into place would be larger than the code
4516 to generate it on the spot.
4518 @findex TRAMPOLINE_SECTION
4519 @item TRAMPOLINE_SECTION
4520 The name of a subroutine to switch to the section in which the
4521 trampoline template is to be placed (@pxref{Sections}). The default is
4522 a value of @samp{readonly_data_section}, which places the trampoline in
4523 the section containing read-only data.
4525 @findex TRAMPOLINE_SIZE
4526 @item TRAMPOLINE_SIZE
4527 A C expression for the size in bytes of the trampoline, as an integer.
4529 @findex TRAMPOLINE_ALIGNMENT
4530 @item TRAMPOLINE_ALIGNMENT
4531 Alignment required for trampolines, in bits.
4533 If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
4534 is used for aligning trampolines.
4536 @findex INITIALIZE_TRAMPOLINE
4537 @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
4538 A C statement to initialize the variable parts of a trampoline.
4539 @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
4540 an RTX for the address of the nested function; @var{static_chain} is an
4541 RTX for the static chain value that should be passed to the function
4544 @findex TRAMPOLINE_ADJUST_ADDRESS
4545 @item TRAMPOLINE_ADJUST_ADDRESS (@var{addr})
4546 A C statement that should perform any machine-specific adjustment in
4547 the address of the trampoline. Its argument contains the address that
4548 was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be
4549 used for a function call should be different from the address in which
4550 the template was stored, the different address should be assigned to
4551 @var{addr}. If this macro is not defined, @var{addr} will be used for
4554 @findex ALLOCATE_TRAMPOLINE
4555 @item ALLOCATE_TRAMPOLINE (@var{fp})
4556 A C expression to allocate run-time space for a trampoline. The
4557 expression value should be an RTX representing a memory reference to the
4558 space for the trampoline.
4560 @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines
4561 @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines
4562 If this macro is not defined, by default the trampoline is allocated as
4563 a stack slot. This default is right for most machines. The exceptions
4564 are machines where it is impossible to execute instructions in the stack
4565 area. On such machines, you may have to implement a separate stack,
4566 using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE}
4567 and @code{TARGET_ASM_FUNCTION_EPILOGUE}.
4569 @var{fp} points to a data structure, a @code{struct function}, which
4570 describes the compilation status of the immediate containing function of
4571 the function which the trampoline is for. Normally (when
4572 @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the
4573 trampoline is in the stack frame of this containing function. Other
4574 allocation strategies probably must do something analogous with this
4578 Implementing trampolines is difficult on many machines because they have
4579 separate instruction and data caches. Writing into a stack location
4580 fails to clear the memory in the instruction cache, so when the program
4581 jumps to that location, it executes the old contents.
4583 Here are two possible solutions. One is to clear the relevant parts of
4584 the instruction cache whenever a trampoline is set up. The other is to
4585 make all trampolines identical, by having them jump to a standard
4586 subroutine. The former technique makes trampoline execution faster; the
4587 latter makes initialization faster.
4589 To clear the instruction cache when a trampoline is initialized, define
4590 the following macros which describe the shape of the cache.
4593 @findex INSN_CACHE_SIZE
4594 @item INSN_CACHE_SIZE
4595 The total size in bytes of the cache.
4597 @findex INSN_CACHE_LINE_WIDTH
4598 @item INSN_CACHE_LINE_WIDTH
4599 The length in bytes of each cache line. The cache is divided into cache
4600 lines which are disjoint slots, each holding a contiguous chunk of data
4601 fetched from memory. Each time data is brought into the cache, an
4602 entire line is read at once. The data loaded into a cache line is
4603 always aligned on a boundary equal to the line size.
4605 @findex INSN_CACHE_DEPTH
4606 @item INSN_CACHE_DEPTH
4607 The number of alternative cache lines that can hold any particular memory
4611 Alternatively, if the machine has system calls or instructions to clear
4612 the instruction cache directly, you can define the following macro.
4615 @findex CLEAR_INSN_CACHE
4616 @item CLEAR_INSN_CACHE (@var{beg}, @var{end})
4617 If defined, expands to a C expression clearing the @emph{instruction
4618 cache} in the specified interval. If it is not defined, and the macro
4619 @code{INSN_CACHE_SIZE} is defined, some generic code is generated to clear the
4620 cache. The definition of this macro would typically be a series of
4621 @code{asm} statements. Both @var{beg} and @var{end} are both pointer
4625 To use a standard subroutine, define the following macro. In addition,
4626 you must make sure that the instructions in a trampoline fill an entire
4627 cache line with identical instructions, or else ensure that the
4628 beginning of the trampoline code is always aligned at the same point in
4629 its cache line. Look in @file{m68k.h} as a guide.
4632 @findex TRANSFER_FROM_TRAMPOLINE
4633 @item TRANSFER_FROM_TRAMPOLINE
4634 Define this macro if trampolines need a special subroutine to do their
4635 work. The macro should expand to a series of @code{asm} statements
4636 which will be compiled with GCC@. They go in a library function named
4637 @code{__transfer_from_trampoline}.
4639 If you need to avoid executing the ordinary prologue code of a compiled
4640 C function when you jump to the subroutine, you can do so by placing a
4641 special label of your own in the assembler code. Use one @code{asm}
4642 statement to generate an assembler label, and another to make the label
4643 global. Then trampolines can use that label to jump directly to your
4644 special assembler code.
4648 @section Implicit Calls to Library Routines
4649 @cindex library subroutine names
4650 @cindex @file{libgcc.a}
4652 @c prevent bad page break with this line
4653 Here is an explanation of implicit calls to library routines.
4656 @findex MULSI3_LIBCALL
4657 @item MULSI3_LIBCALL
4658 A C string constant giving the name of the function to call for
4659 multiplication of one signed full-word by another. If you do not
4660 define this macro, the default name is used, which is @code{__mulsi3},
4661 a function defined in @file{libgcc.a}.
4663 @findex DIVSI3_LIBCALL
4664 @item DIVSI3_LIBCALL
4665 A C string constant giving the name of the function to call for
4666 division of one signed full-word by another. If you do not define
4667 this macro, the default name is used, which is @code{__divsi3}, a
4668 function defined in @file{libgcc.a}.
4670 @findex UDIVSI3_LIBCALL
4671 @item UDIVSI3_LIBCALL
4672 A C string constant giving the name of the function to call for
4673 division of one unsigned full-word by another. If you do not define
4674 this macro, the default name is used, which is @code{__udivsi3}, a
4675 function defined in @file{libgcc.a}.
4677 @findex MODSI3_LIBCALL
4678 @item MODSI3_LIBCALL
4679 A C string constant giving the name of the function to call for the
4680 remainder in division of one signed full-word by another. If you do
4681 not define this macro, the default name is used, which is
4682 @code{__modsi3}, a function defined in @file{libgcc.a}.
4684 @findex UMODSI3_LIBCALL
4685 @item UMODSI3_LIBCALL
4686 A C string constant giving the name of the function to call for the
4687 remainder in division of one unsigned full-word by another. If you do
4688 not define this macro, the default name is used, which is
4689 @code{__umodsi3}, a function defined in @file{libgcc.a}.
4691 @findex MULDI3_LIBCALL
4692 @item MULDI3_LIBCALL
4693 A C string constant giving the name of the function to call for
4694 multiplication of one signed double-word by another. If you do not
4695 define this macro, the default name is used, which is @code{__muldi3},
4696 a function defined in @file{libgcc.a}.
4698 @findex DIVDI3_LIBCALL
4699 @item DIVDI3_LIBCALL
4700 A C string constant giving the name of the function to call for
4701 division of one signed double-word by another. If you do not define
4702 this macro, the default name is used, which is @code{__divdi3}, a
4703 function defined in @file{libgcc.a}.
4705 @findex UDIVDI3_LIBCALL
4706 @item UDIVDI3_LIBCALL
4707 A C string constant giving the name of the function to call for
4708 division of one unsigned full-word by another. If you do not define
4709 this macro, the default name is used, which is @code{__udivdi3}, a
4710 function defined in @file{libgcc.a}.
4712 @findex MODDI3_LIBCALL
4713 @item MODDI3_LIBCALL
4714 A C string constant giving the name of the function to call for the
4715 remainder in division of one signed double-word by another. If you do
4716 not define this macro, the default name is used, which is
4717 @code{__moddi3}, a function defined in @file{libgcc.a}.
4719 @findex UMODDI3_LIBCALL
4720 @item UMODDI3_LIBCALL
4721 A C string constant giving the name of the function to call for the
4722 remainder in division of one unsigned full-word by another. If you do
4723 not define this macro, the default name is used, which is
4724 @code{__umoddi3}, a function defined in @file{libgcc.a}.
4726 @findex DECLARE_LIBRARY_RENAMES
4727 @item DECLARE_LIBRARY_RENAMES
4728 This macro, if defined, should expand to a piece of C code that will get
4729 expanded when compiling functions for libgcc.a. It can be used to
4730 provide alternate names for gcc's internal library functions if there
4731 are ABI-mandated names that the compiler should provide.
4733 @findex INIT_TARGET_OPTABS
4734 @item INIT_TARGET_OPTABS
4735 Define this macro as a C statement that declares additional library
4736 routines renames existing ones. @code{init_optabs} calls this macro after
4737 initializing all the normal library routines.
4739 @findex FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
4740 @item FLOAT_LIB_COMPARE_RETURNS_BOOL
4741 Define this macro as a C statement that returns nonzero if a call to
4742 the floating point comparison library function will return a boolean
4743 value that indicates the result of the comparison. It should return
4744 zero if one of gcc's own libgcc functions is called.
4746 Most ports don't need to define this macro.
4749 @cindex @code{EDOM}, implicit usage
4751 The value of @code{EDOM} on the target machine, as a C integer constant
4752 expression. If you don't define this macro, GCC does not attempt to
4753 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4754 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4757 If you do not define @code{TARGET_EDOM}, then compiled code reports
4758 domain errors by calling the library function and letting it report the
4759 error. If mathematical functions on your system use @code{matherr} when
4760 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4761 that @code{matherr} is used normally.
4763 @findex GEN_ERRNO_RTX
4764 @cindex @code{errno}, implicit usage
4766 Define this macro as a C expression to create an rtl expression that
4767 refers to the global ``variable'' @code{errno}. (On certain systems,
4768 @code{errno} may not actually be a variable.) If you don't define this
4769 macro, a reasonable default is used.
4771 @findex TARGET_MEM_FUNCTIONS
4772 @cindex @code{bcopy}, implicit usage
4773 @cindex @code{memcpy}, implicit usage
4774 @cindex @code{memmove}, implicit usage
4775 @cindex @code{bzero}, implicit usage
4776 @cindex @code{memset}, implicit usage
4777 @item TARGET_MEM_FUNCTIONS
4778 Define this macro if GCC should generate calls to the ISO C
4779 (and System V) library functions @code{memcpy}, @code{memmove} and
4780 @code{memset} rather than the BSD functions @code{bcopy} and @code{bzero}.
4782 @findex TARGET_C99_FUNCTIONS
4783 @cindex C99 math functions, implicit usage
4784 @item TARGET_C99_FUNCTIONS
4785 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
4786 @code{sinf} and similary for other functions defined by C99 standard. The
4787 default is nonzero that should be proper value for most modern systems, however
4788 number of existing systems lacks support for these functions in the runtime so
4789 they needs this macro to be redefined to 0.
4791 @findex LIBGCC_NEEDS_DOUBLE
4792 @item LIBGCC_NEEDS_DOUBLE
4793 Define this macro if @code{float} arguments cannot be passed to library
4794 routines (so they must be converted to @code{double}). This macro
4795 affects both how library calls are generated and how the library
4796 routines in @file{libgcc.a} accept their arguments. It is useful on
4797 machines where floating and fixed point arguments are passed
4798 differently, such as the i860.
4800 @findex NEXT_OBJC_RUNTIME
4801 @item NEXT_OBJC_RUNTIME
4802 Define this macro to generate code for Objective-C message sending using
4803 the calling convention of the NeXT system. This calling convention
4804 involves passing the object, the selector and the method arguments all
4805 at once to the method-lookup library function.
4807 The default calling convention passes just the object and the selector
4808 to the lookup function, which returns a pointer to the method.
4811 @node Addressing Modes
4812 @section Addressing Modes
4813 @cindex addressing modes
4815 @c prevent bad page break with this line
4816 This is about addressing modes.
4819 @findex HAVE_PRE_INCREMENT
4820 @findex HAVE_PRE_DECREMENT
4821 @findex HAVE_POST_INCREMENT
4822 @findex HAVE_POST_DECREMENT
4823 @item HAVE_PRE_INCREMENT
4824 @itemx HAVE_PRE_DECREMENT
4825 @itemx HAVE_POST_INCREMENT
4826 @itemx HAVE_POST_DECREMENT
4827 A C expression that is nonzero if the machine supports pre-increment,
4828 pre-decrement, post-increment, or post-decrement addressing respectively.
4830 @findex HAVE_POST_MODIFY_DISP
4831 @findex HAVE_PRE_MODIFY_DISP
4832 @item HAVE_PRE_MODIFY_DISP
4833 @itemx HAVE_POST_MODIFY_DISP
4834 A C expression that is nonzero if the machine supports pre- or
4835 post-address side-effect generation involving constants other than
4836 the size of the memory operand.
4838 @findex HAVE_POST_MODIFY_REG
4839 @findex HAVE_PRE_MODIFY_REG
4840 @item HAVE_PRE_MODIFY_REG
4841 @itemx HAVE_POST_MODIFY_REG
4842 A C expression that is nonzero if the machine supports pre- or
4843 post-address side-effect generation involving a register displacement.
4845 @findex CONSTANT_ADDRESS_P
4846 @item CONSTANT_ADDRESS_P (@var{x})
4847 A C expression that is 1 if the RTX @var{x} is a constant which
4848 is a valid address. On most machines, this can be defined as
4849 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
4850 in which constant addresses are supported.
4853 @code{CONSTANT_P} accepts integer-values expressions whose values are
4854 not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and
4855 @code{high} expressions and @code{const} arithmetic expressions, in
4856 addition to @code{const_int} and @code{const_double} expressions.
4858 @findex MAX_REGS_PER_ADDRESS
4859 @item MAX_REGS_PER_ADDRESS
4860 A number, the maximum number of registers that can appear in a valid
4861 memory address. Note that it is up to you to specify a value equal to
4862 the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
4865 @findex GO_IF_LEGITIMATE_ADDRESS
4866 @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
4867 A C compound statement with a conditional @code{goto @var{label};}
4868 executed if @var{x} (an RTX) is a legitimate memory address on the
4869 target machine for a memory operand of mode @var{mode}.
4871 It usually pays to define several simpler macros to serve as
4872 subroutines for this one. Otherwise it may be too complicated to
4875 This macro must exist in two variants: a strict variant and a
4876 non-strict one. The strict variant is used in the reload pass. It
4877 must be defined so that any pseudo-register that has not been
4878 allocated a hard register is considered a memory reference. In
4879 contexts where some kind of register is required, a pseudo-register
4880 with no hard register must be rejected.
4882 The non-strict variant is used in other passes. It must be defined to
4883 accept all pseudo-registers in every context where some kind of
4884 register is required.
4886 @findex REG_OK_STRICT
4887 Compiler source files that want to use the strict variant of this
4888 macro define the macro @code{REG_OK_STRICT}. You should use an
4889 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
4890 in that case and the non-strict variant otherwise.
4892 Subroutines to check for acceptable registers for various purposes (one
4893 for base registers, one for index registers, and so on) are typically
4894 among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
4895 Then only these subroutine macros need have two variants; the higher
4896 levels of macros may be the same whether strict or not.
4898 Normally, constant addresses which are the sum of a @code{symbol_ref}
4899 and an integer are stored inside a @code{const} RTX to mark them as
4900 constant. Therefore, there is no need to recognize such sums
4901 specifically as legitimate addresses. Normally you would simply
4902 recognize any @code{const} as legitimate.
4904 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
4905 sums that are not marked with @code{const}. It assumes that a naked
4906 @code{plus} indicates indexing. If so, then you @emph{must} reject such
4907 naked constant sums as illegitimate addresses, so that none of them will
4908 be given to @code{PRINT_OPERAND_ADDRESS}.
4910 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
4911 On some machines, whether a symbolic address is legitimate depends on
4912 the section that the address refers to. On these machines, define the
4913 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
4914 into the @code{symbol_ref}, and then check for it here. When you see a
4915 @code{const}, you will have to look inside it to find the
4916 @code{symbol_ref} in order to determine the section. @xref{Assembler
4919 @findex saveable_obstack
4920 The best way to modify the name string is by adding text to the
4921 beginning, with suitable punctuation to prevent any ambiguity. Allocate
4922 the new name in @code{saveable_obstack}. You will have to modify
4923 @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and
4924 output the name accordingly, and define @code{TARGET_STRIP_NAME_ENCODING}
4925 to access the original name string.
4927 You can check the information stored here into the @code{symbol_ref} in
4928 the definitions of the macros @code{GO_IF_LEGITIMATE_ADDRESS} and
4929 @code{PRINT_OPERAND_ADDRESS}.
4931 @findex REG_OK_FOR_BASE_P
4932 @item REG_OK_FOR_BASE_P (@var{x})
4933 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4934 RTX) is valid for use as a base register. For hard registers, it
4935 should always accept those which the hardware permits and reject the
4936 others. Whether the macro accepts or rejects pseudo registers must be
4937 controlled by @code{REG_OK_STRICT} as described above. This usually
4938 requires two variant definitions, of which @code{REG_OK_STRICT}
4939 controls the one actually used.
4941 @findex REG_MODE_OK_FOR_BASE_P
4942 @item REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode})
4943 A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that
4944 that expression may examine the mode of the memory reference in
4945 @var{mode}. You should define this macro if the mode of the memory
4946 reference affects whether a register may be used as a base register. If
4947 you define this macro, the compiler will use it instead of
4948 @code{REG_OK_FOR_BASE_P}.
4950 @findex REG_OK_FOR_INDEX_P
4951 @item REG_OK_FOR_INDEX_P (@var{x})
4952 A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
4953 RTX) is valid for use as an index register.
4955 The difference between an index register and a base register is that
4956 the index register may be scaled. If an address involves the sum of
4957 two registers, neither one of them scaled, then either one may be
4958 labeled the ``base'' and the other the ``index''; but whichever
4959 labeling is used must fit the machine's constraints of which registers
4960 may serve in each capacity. The compiler will try both labelings,
4961 looking for one that is valid, and will reload one or both registers
4962 only if neither labeling works.
4964 @findex FIND_BASE_TERM
4965 @item FIND_BASE_TERM (@var{x})
4966 A C expression to determine the base term of address @var{x}.
4967 This macro is used in only one place: `find_base_term' in alias.c.
4969 It is always safe for this macro to not be defined. It exists so
4970 that alias analysis can understand machine-dependent addresses.
4972 The typical use of this macro is to handle addresses containing
4973 a label_ref or symbol_ref within an UNSPEC@.
4975 @findex LEGITIMIZE_ADDRESS
4976 @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
4977 A C compound statement that attempts to replace @var{x} with a valid
4978 memory address for an operand of mode @var{mode}. @var{win} will be a
4979 C statement label elsewhere in the code; the macro definition may use
4982 GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
4986 to avoid further processing if the address has become legitimate.
4988 @findex break_out_memory_refs
4989 @var{x} will always be the result of a call to @code{break_out_memory_refs},
4990 and @var{oldx} will be the operand that was given to that function to produce
4993 The code generated by this macro should not alter the substructure of
4994 @var{x}. If it transforms @var{x} into a more legitimate form, it
4995 should assign @var{x} (which will always be a C variable) a new value.
4997 It is not necessary for this macro to come up with a legitimate
4998 address. The compiler has standard ways of doing so in all cases. In
4999 fact, it is safe for this macro to do nothing. But often a
5000 machine-dependent strategy can generate better code.
5002 @findex LEGITIMIZE_RELOAD_ADDRESS
5003 @item LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5004 A C compound statement that attempts to replace @var{x}, which is an address
5005 that needs reloading, with a valid memory address for an operand of mode
5006 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5007 It is not necessary to define this macro, but it might be useful for
5008 performance reasons.
5010 For example, on the i386, it is sometimes possible to use a single
5011 reload register instead of two by reloading a sum of two pseudo
5012 registers into a register. On the other hand, for number of RISC
5013 processors offsets are limited so that often an intermediate address
5014 needs to be generated in order to address a stack slot. By defining
5015 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5016 generated for adjacent some stack slots can be made identical, and thus
5019 @emph{Note}: This macro should be used with caution. It is necessary
5020 to know something of how reload works in order to effectively use this,
5021 and it is quite easy to produce macros that build in too much knowledge
5022 of reload internals.
5024 @emph{Note}: This macro must be able to reload an address created by a
5025 previous invocation of this macro. If it fails to handle such addresses
5026 then the compiler may generate incorrect code or abort.
5029 The macro definition should use @code{push_reload} to indicate parts that
5030 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5031 suitable to be passed unaltered to @code{push_reload}.
5033 The code generated by this macro must not alter the substructure of
5034 @var{x}. If it transforms @var{x} into a more legitimate form, it
5035 should assign @var{x} (which will always be a C variable) a new value.
5036 This also applies to parts that you change indirectly by calling
5039 @findex strict_memory_address_p
5040 The macro definition may use @code{strict_memory_address_p} to test if
5041 the address has become legitimate.
5044 If you want to change only a part of @var{x}, one standard way of doing
5045 this is to use @code{copy_rtx}. Note, however, that is unshares only a
5046 single level of rtl. Thus, if the part to be changed is not at the
5047 top level, you'll need to replace first the top level.
5048 It is not necessary for this macro to come up with a legitimate
5049 address; but often a machine-dependent strategy can generate better code.
5051 @findex GO_IF_MODE_DEPENDENT_ADDRESS
5052 @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5053 A C statement or compound statement with a conditional @code{goto
5054 @var{label};} executed if memory address @var{x} (an RTX) can have
5055 different meanings depending on the machine mode of the memory
5056 reference it is used for or if the address is valid for some modes
5059 Autoincrement and autodecrement addresses typically have mode-dependent
5060 effects because the amount of the increment or decrement is the size
5061 of the operand being addressed. Some machines have other mode-dependent
5062 addresses. Many RISC machines have no mode-dependent addresses.
5064 You may assume that @var{addr} is a valid address for the machine.
5066 @findex LEGITIMATE_CONSTANT_P
5067 @item LEGITIMATE_CONSTANT_P (@var{x})
5068 A C expression that is nonzero if @var{x} is a legitimate constant for
5069 an immediate operand on the target machine. You can assume that
5070 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5071 @samp{1} is a suitable definition for this macro on machines where
5072 anything @code{CONSTANT_P} is valid.
5075 @node Condition Code
5076 @section Condition Code Status
5077 @cindex condition code status
5079 @c prevent bad page break with this line
5080 This describes the condition code status.
5083 The file @file{conditions.h} defines a variable @code{cc_status} to
5084 describe how the condition code was computed (in case the interpretation of
5085 the condition code depends on the instruction that it was set by). This
5086 variable contains the RTL expressions on which the condition code is
5087 currently based, and several standard flags.
5089 Sometimes additional machine-specific flags must be defined in the machine
5090 description header file. It can also add additional machine-specific
5091 information by defining @code{CC_STATUS_MDEP}.
5094 @findex CC_STATUS_MDEP
5095 @item CC_STATUS_MDEP
5096 C code for a data type which is used for declaring the @code{mdep}
5097 component of @code{cc_status}. It defaults to @code{int}.
5099 This macro is not used on machines that do not use @code{cc0}.
5101 @findex CC_STATUS_MDEP_INIT
5102 @item CC_STATUS_MDEP_INIT
5103 A C expression to initialize the @code{mdep} field to ``empty''.
5104 The default definition does nothing, since most machines don't use
5105 the field anyway. If you want to use the field, you should probably
5106 define this macro to initialize it.
5108 This macro is not used on machines that do not use @code{cc0}.
5110 @findex NOTICE_UPDATE_CC
5111 @item NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5112 A C compound statement to set the components of @code{cc_status}
5113 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5114 this macro's responsibility to recognize insns that set the condition
5115 code as a byproduct of other activity as well as those that explicitly
5118 This macro is not used on machines that do not use @code{cc0}.
5120 If there are insns that do not set the condition code but do alter
5121 other machine registers, this macro must check to see whether they
5122 invalidate the expressions that the condition code is recorded as
5123 reflecting. For example, on the 68000, insns that store in address
5124 registers do not set the condition code, which means that usually
5125 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5126 insns. But suppose that the previous insn set the condition code
5127 based on location @samp{a4@@(102)} and the current insn stores a new
5128 value in @samp{a4}. Although the condition code is not changed by
5129 this, it will no longer be true that it reflects the contents of
5130 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5131 @code{cc_status} in this case to say that nothing is known about the
5132 condition code value.
5134 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5135 with the results of peephole optimization: insns whose patterns are
5136 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5137 constants which are just the operands. The RTL structure of these
5138 insns is not sufficient to indicate what the insns actually do. What
5139 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5140 @code{CC_STATUS_INIT}.
5142 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5143 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5144 @samp{cc}. This avoids having detailed information about patterns in
5145 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5147 @findex EXTRA_CC_MODES
5148 @item EXTRA_CC_MODES
5149 Condition codes are represented in registers by machine modes of class
5150 @code{MODE_CC}. By default, there is just one mode, @code{CCmode}, with
5151 this class. If you need more such modes, create a file named
5152 @file{@var{machine}-modes.def} in your @file{config/@var{machine}}
5153 directory (@pxref{Back End, , Anatomy of a Target Back End}), containing
5154 a list of these modes. Each entry in the list should be a call to the
5155 macro @code{CC}. This macro takes one argument, which is the name of
5156 the mode: it should begin with @samp{CC}. Do not put quotation marks
5157 around the name, or include the trailing @samp{mode}; these are
5158 automatically added. There should not be anything else in the file
5161 A sample @file{@var{machine}-modes.def} file might look like this:
5164 CC (CC_NOOV) /* @r{Comparison only valid if there was no overflow.} */
5165 CC (CCFP) /* @r{Floating point comparison that cannot trap.} */
5166 CC (CCFPE) /* @r{Floating point comparison that may trap.} */
5169 When you create this file, the macro @code{EXTRA_CC_MODES} is
5170 automatically defined by @command{configure}, with value @samp{1}.
5172 @findex SELECT_CC_MODE
5173 @item SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5174 Returns a mode from class @code{MODE_CC} to be used when comparison
5175 operation code @var{op} is applied to rtx @var{x} and @var{y}. For
5176 example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see
5177 @pxref{Jump Patterns} for a description of the reason for this
5181 #define SELECT_CC_MODE(OP,X,Y) \
5182 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5183 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5184 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5185 || GET_CODE (X) == NEG) \
5186 ? CC_NOOVmode : CCmode))
5189 You need not define this macro if @code{EXTRA_CC_MODES} is not defined.
5191 @findex CANONICALIZE_COMPARISON
5192 @item CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5193 On some machines not all possible comparisons are defined, but you can
5194 convert an invalid comparison into a valid one. For example, the Alpha
5195 does not have a @code{GT} comparison, but you can use an @code{LT}
5196 comparison instead and swap the order of the operands.
5198 On such machines, define this macro to be a C statement to do any
5199 required conversions. @var{code} is the initial comparison code
5200 and @var{op0} and @var{op1} are the left and right operands of the
5201 comparison, respectively. You should modify @var{code}, @var{op0}, and
5202 @var{op1} as required.
5204 GCC will not assume that the comparison resulting from this macro is
5205 valid but will see if the resulting insn matches a pattern in the
5208 You need not define this macro if it would never change the comparison
5211 @findex REVERSIBLE_CC_MODE
5212 @item REVERSIBLE_CC_MODE (@var{mode})
5213 A C expression whose value is one if it is always safe to reverse a
5214 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5215 can ever return @var{mode} for a floating-point inequality comparison,
5216 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5218 You need not define this macro if it would always returns zero or if the
5219 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5220 For example, here is the definition used on the SPARC, where floating-point
5221 inequality comparisons are always given @code{CCFPEmode}:
5224 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5227 @findex REVERSE_CONDITION (@var{code}, @var{mode})
5228 A C expression whose value is reversed condition code of the @var{code} for
5229 comparison done in CC_MODE @var{mode}. The macro is used only in case
5230 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5231 machine has some non-standard way how to reverse certain conditionals. For
5232 instance in case all floating point conditions are non-trapping, compiler may
5233 freely convert unordered compares to ordered one. Then definition may look
5237 #define REVERSE_CONDITION(CODE, MODE) \
5238 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5239 : reverse_condition_maybe_unordered (CODE))
5242 @findex REVERSE_CONDEXEC_PREDICATES_P
5243 @item REVERSE_CONDEXEC_PREDICATES_P (@var{code1}, @var{code2})
5244 A C expression that returns true if the conditional execution predicate
5245 @var{code1} is the inverse of @var{code2} and vice versa. Define this to
5246 return 0 if the target has conditional execution predicates that cannot be
5247 reversed safely. If no expansion is specified, this macro is defined as
5251 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
5252 ((x) == reverse_condition (y))
5258 @section Describing Relative Costs of Operations
5259 @cindex costs of instructions
5260 @cindex relative costs
5261 @cindex speed of instructions
5263 These macros let you describe the relative speed of various operations
5264 on the target machine.
5268 @item CONST_COSTS (@var{x}, @var{code}, @var{outer_code})
5269 A part of a C @code{switch} statement that describes the relative costs
5270 of constant RTL expressions. It must contain @code{case} labels for
5271 expression codes @code{const_int}, @code{const}, @code{symbol_ref},
5272 @code{label_ref} and @code{const_double}. Each case must ultimately
5273 reach a @code{return} statement to return the relative cost of the use
5274 of that kind of constant value in an expression. The cost may depend on
5275 the precise value of the constant, which is available for examination in
5276 @var{x}, and the rtx code of the expression in which it is contained,
5277 found in @var{outer_code}.
5279 @var{code} is the expression code---redundant, since it can be
5280 obtained with @code{GET_CODE (@var{x})}.
5283 @findex COSTS_N_INSNS
5284 @item RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
5285 Like @code{CONST_COSTS} but applies to nonconstant RTL expressions.
5286 This can be used, for example, to indicate how costly a multiply
5287 instruction is. In writing this macro, you can use the construct
5288 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
5289 instructions. @var{outer_code} is the code of the expression in which
5290 @var{x} is contained.
5292 This macro is optional; do not define it if the default cost assumptions
5293 are adequate for the target machine.
5295 @findex DEFAULT_RTX_COSTS
5296 @item DEFAULT_RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
5297 This macro, if defined, is called for any case not handled by the
5298 @code{RTX_COSTS} or @code{CONST_COSTS} macros. This eliminates the need
5299 to put case labels into the macro, but the code, or any functions it
5300 calls, must assume that the RTL in @var{x} could be of any type that has
5301 not already been handled. The arguments are the same as for
5302 @code{RTX_COSTS}, and the macro should execute a return statement giving
5303 the cost of any RTL expressions that it can handle. The default cost
5304 calculation is used for any RTL for which this macro does not return a
5307 This macro is optional; do not define it if the default cost assumptions
5308 are adequate for the target machine.
5310 @findex ADDRESS_COST
5311 @item ADDRESS_COST (@var{address})
5312 An expression giving the cost of an addressing mode that contains
5313 @var{address}. If not defined, the cost is computed from
5314 the @var{address} expression and the @code{CONST_COSTS} values.
5316 For most CISC machines, the default cost is a good approximation of the
5317 true cost of the addressing mode. However, on RISC machines, all
5318 instructions normally have the same length and execution time. Hence
5319 all addresses will have equal costs.
5321 In cases where more than one form of an address is known, the form with
5322 the lowest cost will be used. If multiple forms have the same, lowest,
5323 cost, the one that is the most complex will be used.
5325 For example, suppose an address that is equal to the sum of a register
5326 and a constant is used twice in the same basic block. When this macro
5327 is not defined, the address will be computed in a register and memory
5328 references will be indirect through that register. On machines where
5329 the cost of the addressing mode containing the sum is no higher than
5330 that of a simple indirect reference, this will produce an additional
5331 instruction and possibly require an additional register. Proper
5332 specification of this macro eliminates this overhead for such machines.
5334 Similar use of this macro is made in strength reduction of loops.
5336 @var{address} need not be valid as an address. In such a case, the cost
5337 is not relevant and can be any value; invalid addresses need not be
5338 assigned a different cost.
5340 On machines where an address involving more than one register is as
5341 cheap as an address computation involving only one register, defining
5342 @code{ADDRESS_COST} to reflect this can cause two registers to be live
5343 over a region of code where only one would have been if
5344 @code{ADDRESS_COST} were not defined in that manner. This effect should
5345 be considered in the definition of this macro. Equivalent costs should
5346 probably only be given to addresses with different numbers of registers
5347 on machines with lots of registers.
5349 This macro will normally either not be defined or be defined as a
5352 @findex REGISTER_MOVE_COST
5353 @item REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
5354 A C expression for the cost of moving data of mode @var{mode} from a
5355 register in class @var{from} to one in class @var{to}. The classes are
5356 expressed using the enumeration values such as @code{GENERAL_REGS}. A
5357 value of 2 is the default; other values are interpreted relative to
5360 It is not required that the cost always equal 2 when @var{from} is the
5361 same as @var{to}; on some machines it is expensive to move between
5362 registers if they are not general registers.
5364 If reload sees an insn consisting of a single @code{set} between two
5365 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
5366 classes returns a value of 2, reload does not check to ensure that the
5367 constraints of the insn are met. Setting a cost of other than 2 will
5368 allow reload to verify that the constraints are met. You should do this
5369 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
5371 @findex MEMORY_MOVE_COST
5372 @item MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
5373 A C expression for the cost of moving data of mode @var{mode} between a
5374 register of class @var{class} and memory; @var{in} is zero if the value
5375 is to be written to memory, nonzero if it is to be read in. This cost
5376 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
5377 registers and memory is more expensive than between two registers, you
5378 should define this macro to express the relative cost.
5380 If you do not define this macro, GCC uses a default cost of 4 plus
5381 the cost of copying via a secondary reload register, if one is
5382 needed. If your machine requires a secondary reload register to copy
5383 between memory and a register of @var{class} but the reload mechanism is
5384 more complex than copying via an intermediate, define this macro to
5385 reflect the actual cost of the move.
5387 GCC defines the function @code{memory_move_secondary_cost} if
5388 secondary reloads are needed. It computes the costs due to copying via
5389 a secondary register. If your machine copies from memory using a
5390 secondary register in the conventional way but the default base value of
5391 4 is not correct for your machine, define this macro to add some other
5392 value to the result of that function. The arguments to that function
5393 are the same as to this macro.
5397 A C expression for the cost of a branch instruction. A value of 1 is
5398 the default; other values are interpreted relative to that.
5401 Here are additional macros which do not specify precise relative costs,
5402 but only that certain actions are more expensive than GCC would
5406 @findex SLOW_BYTE_ACCESS
5407 @item SLOW_BYTE_ACCESS
5408 Define this macro as a C expression which is nonzero if accessing less
5409 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
5410 faster than accessing a word of memory, i.e., if such access
5411 require more than one instruction or if there is no difference in cost
5412 between byte and (aligned) word loads.
5414 When this macro is not defined, the compiler will access a field by
5415 finding the smallest containing object; when it is defined, a fullword
5416 load will be used if alignment permits. Unless bytes accesses are
5417 faster than word accesses, using word accesses is preferable since it
5418 may eliminate subsequent memory access if subsequent accesses occur to
5419 other fields in the same word of the structure, but to different bytes.
5421 @findex SLOW_UNALIGNED_ACCESS
5422 @item SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
5423 Define this macro to be the value 1 if memory accesses described by the
5424 @var{mode} and @var{alignment} parameters have a cost many times greater
5425 than aligned accesses, for example if they are emulated in a trap
5428 When this macro is nonzero, the compiler will act as if
5429 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
5430 moves. This can cause significantly more instructions to be produced.
5431 Therefore, do not set this macro nonzero if unaligned accesses only add a
5432 cycle or two to the time for a memory access.
5434 If the value of this macro is always zero, it need not be defined. If
5435 this macro is defined, it should produce a nonzero value when
5436 @code{STRICT_ALIGNMENT} is nonzero.
5438 @findex DONT_REDUCE_ADDR
5439 @item DONT_REDUCE_ADDR
5440 Define this macro to inhibit strength reduction of memory addresses.
5441 (On some machines, such strength reduction seems to do harm rather
5446 The threshold of number of scalar memory-to-memory move insns, @emph{below}
5447 which a sequence of insns should be generated instead of a
5448 string move insn or a library call. Increasing the value will always
5449 make code faster, but eventually incurs high cost in increased code size.
5451 Note that on machines where the corresponding move insn is a
5452 @code{define_expand} that emits a sequence of insns, this macro counts
5453 the number of such sequences.
5455 If you don't define this, a reasonable default is used.
5457 @findex MOVE_BY_PIECES_P
5458 @item MOVE_BY_PIECES_P (@var{size}, @var{alignment})
5459 A C expression used to determine whether @code{move_by_pieces} will be used to
5460 copy a chunk of memory, or whether some other block move mechanism
5461 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5462 than @code{MOVE_RATIO}.
5464 @findex MOVE_MAX_PIECES
5465 @item MOVE_MAX_PIECES
5466 A C expression used by @code{move_by_pieces} to determine the largest unit
5467 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
5471 The threshold of number of scalar move insns, @emph{below} which a sequence
5472 of insns should be generated to clear memory instead of a string clear insn
5473 or a library call. Increasing the value will always make code faster, but
5474 eventually incurs high cost in increased code size.
5476 If you don't define this, a reasonable default is used.
5478 @findex CLEAR_BY_PIECES_P
5479 @item CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
5480 A C expression used to determine whether @code{clear_by_pieces} will be used
5481 to clear a chunk of memory, or whether some other block clear mechanism
5482 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
5483 than @code{CLEAR_RATIO}.
5485 @findex STORE_BY_PIECES_P
5486 @item STORE_BY_PIECES_P (@var{size}, @var{alignment})
5487 A C expression used to determine whether @code{store_by_pieces} will be
5488 used to set a chunk of memory to a constant value, or whether some other
5489 mechanism will be used. Used by @code{__builtin_memset} when storing
5490 values other than constant zero and by @code{__builtin_strcpy} when
5491 when called with a constant source string.
5492 Defaults to @code{MOVE_BY_PIECES_P}.
5494 @findex USE_LOAD_POST_INCREMENT
5495 @item USE_LOAD_POST_INCREMENT (@var{mode})
5496 A C expression used to determine whether a load postincrement is a good
5497 thing to use for a given mode. Defaults to the value of
5498 @code{HAVE_POST_INCREMENT}.
5500 @findex USE_LOAD_POST_DECREMENT
5501 @item USE_LOAD_POST_DECREMENT (@var{mode})
5502 A C expression used to determine whether a load postdecrement is a good
5503 thing to use for a given mode. Defaults to the value of
5504 @code{HAVE_POST_DECREMENT}.
5506 @findex USE_LOAD_PRE_INCREMENT
5507 @item USE_LOAD_PRE_INCREMENT (@var{mode})
5508 A C expression used to determine whether a load preincrement is a good
5509 thing to use for a given mode. Defaults to the value of
5510 @code{HAVE_PRE_INCREMENT}.
5512 @findex USE_LOAD_PRE_DECREMENT
5513 @item USE_LOAD_PRE_DECREMENT (@var{mode})
5514 A C expression used to determine whether a load predecrement is a good
5515 thing to use for a given mode. Defaults to the value of
5516 @code{HAVE_PRE_DECREMENT}.
5518 @findex USE_STORE_POST_INCREMENT
5519 @item USE_STORE_POST_INCREMENT (@var{mode})
5520 A C expression used to determine whether a store postincrement is a good
5521 thing to use for a given mode. Defaults to the value of
5522 @code{HAVE_POST_INCREMENT}.
5524 @findex USE_STORE_POST_DECREMENT
5525 @item USE_STORE_POST_DECREMENT (@var{mode})
5526 A C expression used to determine whether a store postdecrement is a good
5527 thing to use for a given mode. Defaults to the value of
5528 @code{HAVE_POST_DECREMENT}.
5530 @findex USE_STORE_PRE_INCREMENT
5531 @item USE_STORE_PRE_INCREMENT (@var{mode})
5532 This macro is used to determine whether a store preincrement is a good
5533 thing to use for a given mode. Defaults to the value of
5534 @code{HAVE_PRE_INCREMENT}.
5536 @findex USE_STORE_PRE_DECREMENT
5537 @item USE_STORE_PRE_DECREMENT (@var{mode})
5538 This macro is used to determine whether a store predecrement is a good
5539 thing to use for a given mode. Defaults to the value of
5540 @code{HAVE_PRE_DECREMENT}.
5542 @findex NO_FUNCTION_CSE
5543 @item NO_FUNCTION_CSE
5544 Define this macro if it is as good or better to call a constant
5545 function address than to call an address kept in a register.
5547 @findex NO_RECURSIVE_FUNCTION_CSE
5548 @item NO_RECURSIVE_FUNCTION_CSE
5549 Define this macro if it is as good or better for a function to call
5550 itself with an explicit address than to call an address kept in a
5555 @section Adjusting the Instruction Scheduler
5557 The instruction scheduler may need a fair amount of machine-specific
5558 adjustment in order to produce good code. GCC provides several target
5559 hooks for this purpose. It is usually enough to define just a few of
5560 them: try the first ones in this list first.
5562 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
5563 This hook returns the maximum number of instructions that can ever
5564 issue at the same time on the target machine. The default is one.
5565 Although the insn scheduler can define itself the possibility of issue
5566 an insn on the same cycle, the value can serve as an additional
5567 constraint to issue insns on the same simulated processor cycle (see
5568 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
5569 This value must be constant over the entire compilation. If you need
5570 it to vary depending on what the instructions are, you must use
5571 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
5573 For the automaton based pipeline interface, you could define this hook
5574 to return the value of the macro @code{MAX_DFA_ISSUE_RATE}.
5577 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
5578 This hook is executed by the scheduler after it has scheduled an insn
5579 from the ready list. It should return the number of insns which can
5580 still be issued in the current cycle. The default is
5581 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
5582 @code{USE}, which normally are not counted against the issue rate.
5583 You should define this hook if some insns take more machine resources
5584 than others, so that fewer insns can follow them in the same cycle.
5585 @var{file} is either a null pointer, or a stdio stream to write any
5586 debug output to. @var{verbose} is the verbose level provided by
5587 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
5591 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
5592 This function corrects the value of @var{cost} based on the
5593 relationship between @var{insn} and @var{dep_insn} through the
5594 dependence @var{link}. It should return the new value. The default
5595 is to make no adjustment to @var{cost}. This can be used for example
5596 to specify to the scheduler using the traditional pipeline description
5597 that an output- or anti-dependence does not incur the same cost as a
5598 data-dependence. If the scheduler using the automaton based pipeline
5599 description, the cost of anti-dependence is zero and the cost of
5600 output-dependence is maximum of one and the difference of latency
5601 times of the first and the second insns. If these values are not
5602 acceptable, you could use the hook to modify them too. See also
5603 @pxref{Automaton pipeline description}.
5606 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
5607 This hook adjusts the integer scheduling priority @var{priority} of
5608 @var{insn}. It should return the new priority. Reduce the priority to
5609 execute @var{insn} earlier, increase the priority to execute @var{insn}
5610 later. Do not define this hook if you do not need to adjust the
5611 scheduling priorities of insns.
5614 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
5615 This hook is executed by the scheduler after it has scheduled the ready
5616 list, to allow the machine description to reorder it (for example to
5617 combine two small instructions together on @samp{VLIW} machines).
5618 @var{file} is either a null pointer, or a stdio stream to write any
5619 debug output to. @var{verbose} is the verbose level provided by
5620 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
5621 list of instructions that are ready to be scheduled. @var{n_readyp} is
5622 a pointer to the number of elements in the ready list. The scheduler
5623 reads the ready list in reverse order, starting with
5624 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
5625 is the timer tick of the scheduler. You may modify the ready list and
5626 the number of ready insns. The return value is the number of insns that
5627 can issue this cycle; normally this is just @code{issue_rate}. See also
5628 @samp{TARGET_SCHED_REORDER2}.
5631 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
5632 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
5633 function is called whenever the scheduler starts a new cycle. This one
5634 is called once per iteration over a cycle, immediately after
5635 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
5636 return the number of insns to be scheduled in the same cycle. Defining
5637 this hook can be useful if there are frequent situations where
5638 scheduling one insn causes other insns to become ready in the same
5639 cycle. These other insns can then be taken into account properly.
5642 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
5643 This hook is called after evaluation forward dependencies of insns in
5644 chain given by two parameter values (@var{head} and @var{tail}
5645 correspondingly) but before insns scheduling of the insn chain. For
5646 example, it can be used for better insn classification if it requires
5647 analysis of dependencies. This hook can use backward and forward
5648 dependencies of the insn scheduler because they are already
5652 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
5653 This hook is executed by the scheduler at the beginning of each block of
5654 instructions that are to be scheduled. @var{file} is either a null
5655 pointer, or a stdio stream to write any debug output to. @var{verbose}
5656 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
5657 @var{max_ready} is the maximum number of insns in the current scheduling
5658 region that can be live at the same time. This can be used to allocate
5659 scratch space if it is needed, e.g. by @samp{TARGET_SCHED_REORDER}.
5662 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
5663 This hook is executed by the scheduler at the end of each block of
5664 instructions that are to be scheduled. It can be used to perform
5665 cleanup of any actions done by the other scheduling hooks. @var{file}
5666 is either a null pointer, or a stdio stream to write any debug output
5667 to. @var{verbose} is the verbose level provided by
5668 @option{-fsched-verbose-@var{n}}.
5671 @deftypefn {Target Hook} int TARGET_SCHED_USE_DFA_PIPELINE_INTERFACE (void)
5672 This hook is called many times during insn scheduling. If the hook
5673 returns nonzero, the automaton based pipeline description is used for
5674 insn scheduling. Otherwise the traditional pipeline description is
5675 used. The default is usage of the traditional pipeline description.
5677 You should also remember that to simplify the insn scheduler sources
5678 an empty traditional pipeline description interface is generated even
5679 if there is no a traditional pipeline description in the @file{.md}
5680 file. The same is true for the automaton based pipeline description.
5681 That means that you should be accurate in defining the hook.
5684 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
5685 The hook returns an RTL insn. The automaton state used in the
5686 pipeline hazard recognizer is changed as if the insn were scheduled
5687 when the new simulated processor cycle starts. Usage of the hook may
5688 simplify the automaton pipeline description for some @acronym{VLIW}
5689 processors. If the hook is defined, it is used only for the automaton
5690 based pipeline description. The default is not to change the state
5691 when the new simulated processor cycle starts.
5694 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
5695 The hook can be used to initialize data used by the previous hook.
5698 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
5699 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
5700 to changed the state as if the insn were scheduled when the new
5701 simulated processor cycle finishes.
5704 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
5705 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
5706 used to initialize data used by the previous hook.
5709 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
5710 This hook controls better choosing an insn from the ready insn queue
5711 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
5712 chooses the first insn from the queue. If the hook returns a positive
5713 value, an additional scheduler code tries all permutations of
5714 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
5715 subsequent ready insns to choose an insn whose issue will result in
5716 maximal number of issued insns on the same cycle. For the
5717 @acronym{VLIW} processor, the code could actually solve the problem of
5718 packing simple insns into the @acronym{VLIW} insn. Of course, if the
5719 rules of @acronym{VLIW} packing are described in the automaton.
5721 This code also could be used for superscalar @acronym{RISC}
5722 processors. Let us consider a superscalar @acronym{RISC} processor
5723 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
5724 @var{B}, some insns can be executed only in pipelines @var{B} or
5725 @var{C}, and one insn can be executed in pipeline @var{B}. The
5726 processor may issue the 1st insn into @var{A} and the 2nd one into
5727 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
5728 until the next cycle. If the scheduler issues the 3rd insn the first,
5729 the processor could issue all 3 insns per cycle.
5731 Actually this code demonstrates advantages of the automaton based
5732 pipeline hazard recognizer. We try quickly and easy many insn
5733 schedules to choose the best one.
5735 The default is no multipass scheduling.
5738 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
5740 This hook controls what insns from the ready insn queue will be
5741 considered for the multipass insn scheduling. If the hook returns
5742 zero for insn passed as the parameter, the insn will be not chosen to
5745 The default is that any ready insns can be choosen to be issued.
5748 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
5750 This hook is called by the insn scheduler before issuing insn passed
5751 as the third parameter on given cycle. If the hook returns nonzero,
5752 the insn is not issued on given processors cycle. Instead of that,
5753 the processor cycle is advanced. If the value passed through the last
5754 parameter is zero, the insn ready queue is not sorted on the new cycle
5755 start as usually. The first parameter passes file for debugging
5756 output. The second one passes the scheduler verbose level of the
5757 debugging output. The forth and the fifth parameter values are
5758 correspondingly processor cycle on which the previous insn has been
5759 issued and the current processor cycle.
5762 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_BUBBLES (void)
5763 The @acronym{DFA}-based scheduler could take the insertion of nop
5764 operations for better insn scheduling into account. It can be done
5765 only if the multi-pass insn scheduling works (see hook
5766 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD}).
5768 Let us consider a @acronym{VLIW} processor insn with 3 slots. Each
5769 insn can be placed only in one of the three slots. We have 3 ready
5770 insns @var{A}, @var{B}, and @var{C}. @var{A} and @var{C} can be
5771 placed only in the 1st slot, @var{B} can be placed only in the 3rd
5772 slot. We described the automaton which does not permit empty slot
5773 gaps between insns (usually such description is simpler). Without
5774 this code the scheduler would place each insn in 3 separate
5775 @acronym{VLIW} insns. If the scheduler places a nop insn into the 2nd
5776 slot, it could place the 3 insns into 2 @acronym{VLIW} insns. What is
5777 the nop insn is returned by hook @samp{TARGET_SCHED_DFA_BUBBLE}. Hook
5778 @samp{TARGET_SCHED_INIT_DFA_BUBBLES} can be used to initialize or
5779 create the nop insns.
5781 You should remember that the scheduler does not insert the nop insns.
5782 It is not wise because of the following optimizations. The scheduler
5783 only considers such possibility to improve the result schedule. The
5784 nop insns should be inserted lately, e.g. on the final phase.
5787 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_BUBBLE (int @var{index})
5788 This hook @samp{FIRST_CYCLE_MULTIPASS_SCHEDULING} is used to insert
5789 nop operations for better insn scheduling when @acronym{DFA}-based
5790 scheduler makes multipass insn scheduling (see also description of
5791 hook @samp{TARGET_SCHED_INIT_DFA_BUBBLES}). This hook
5792 returns a nop insn with given @var{index}. The indexes start with
5793 zero. The hook should return @code{NULL} if there are no more nop
5794 insns with indexes greater than given index.
5797 Macros in the following table are generated by the program
5798 @file{genattr} and can be useful for writing the hooks.
5801 @findex TRADITIONAL_PIPELINE_INTERFACE
5802 @item TRADITIONAL_PIPELINE_INTERFACE
5803 The macro definition is generated if there is a traditional pipeline
5804 description in @file{.md} file. You should also remember that to
5805 simplify the insn scheduler sources an empty traditional pipeline
5806 description interface is generated even if there is no a traditional
5807 pipeline description in the @file{.md} file. The macro can be used to
5808 distinguish the two types of the traditional interface.
5810 @findex DFA_PIPELINE_INTERFACE
5811 @item DFA_PIPELINE_INTERFACE
5812 The macro definition is generated if there is an automaton pipeline
5813 description in @file{.md} file. You should also remember that to
5814 simplify the insn scheduler sources an empty automaton pipeline
5815 description interface is generated even if there is no an automaton
5816 pipeline description in the @file{.md} file. The macro can be used to
5817 distinguish the two types of the automaton interface.
5819 @findex MAX_DFA_ISSUE_RATE
5820 @item MAX_DFA_ISSUE_RATE
5821 The macro definition is generated in the automaton based pipeline
5822 description interface. Its value is calculated from the automaton
5823 based pipeline description and is equal to maximal number of all insns
5824 described in constructions @samp{define_insn_reservation} which can be
5825 issued on the same processor cycle.
5830 @section Dividing the Output into Sections (Texts, Data, @dots{})
5831 @c the above section title is WAY too long. maybe cut the part between
5832 @c the (...)? --mew 10feb93
5834 An object file is divided into sections containing different types of
5835 data. In the most common case, there are three sections: the @dfn{text
5836 section}, which holds instructions and read-only data; the @dfn{data
5837 section}, which holds initialized writable data; and the @dfn{bss
5838 section}, which holds uninitialized data. Some systems have other kinds
5841 The compiler must tell the assembler when to switch sections. These
5842 macros control what commands to output to tell the assembler this. You
5843 can also define additional sections.
5846 @findex TEXT_SECTION_ASM_OP
5847 @item TEXT_SECTION_ASM_OP
5848 A C expression whose value is a string, including spacing, containing the
5849 assembler operation that should precede instructions and read-only data.
5850 Normally @code{"\t.text"} is right.
5852 @findex TEXT_SECTION
5854 A C statement that switches to the default section containing instructions.
5855 Normally this is not needed, as simply defining @code{TEXT_SECTION_ASM_OP}
5856 is enough. The MIPS port uses this to sort all functions after all data
5859 @findex HOT_TEXT_SECTION_NAME
5860 @item HOT_TEXT_SECTION_NAME
5861 If defined, a C string constant for the name of the section containing most
5862 frequently executed functions of the program. If not defined, GCC will provide
5863 a default definition if the target supports named sections.
5865 @findex UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5866 @item UNLIKELY_EXECUTED_TEXT_SECTION_NAME
5867 If defined, a C string constant for the name of the section containing unlikely
5868 executed functions in the program.
5870 @findex DATA_SECTION_ASM_OP
5871 @item DATA_SECTION_ASM_OP
5872 A C expression whose value is a string, including spacing, containing the
5873 assembler operation to identify the following data as writable initialized
5874 data. Normally @code{"\t.data"} is right.
5876 @findex READONLY_DATA_SECTION_ASM_OP
5877 @item READONLY_DATA_SECTION_ASM_OP
5878 A C expression whose value is a string, including spacing, containing the
5879 assembler operation to identify the following data as read-only initialized
5882 @findex READONLY_DATA_SECTION
5883 @item READONLY_DATA_SECTION
5884 A macro naming a function to call to switch to the proper section for
5885 read-only data. The default is to use @code{READONLY_DATA_SECTION_ASM_OP}
5886 if defined, else fall back to @code{text_section}.
5888 The most common definition will be @code{data_section}, if the target
5889 does not have a special read-only data section, and does not put data
5890 in the text section.
5892 @findex SHARED_SECTION_ASM_OP
5893 @item SHARED_SECTION_ASM_OP
5894 If defined, a C expression whose value is a string, including spacing,
5895 containing the assembler operation to identify the following data as
5896 shared data. If not defined, @code{DATA_SECTION_ASM_OP} will be used.
5898 @findex BSS_SECTION_ASM_OP
5899 @item BSS_SECTION_ASM_OP
5900 If defined, a C expression whose value is a string, including spacing,
5901 containing the assembler operation to identify the following data as
5902 uninitialized global data. If not defined, and neither
5903 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
5904 uninitialized global data will be output in the data section if
5905 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
5908 @findex SHARED_BSS_SECTION_ASM_OP
5909 @item SHARED_BSS_SECTION_ASM_OP
5910 If defined, a C expression whose value is a string, including spacing,
5911 containing the assembler operation to identify the following data as
5912 uninitialized global shared data. If not defined, and
5913 @code{BSS_SECTION_ASM_OP} is, the latter will be used.
5915 @findex INIT_SECTION_ASM_OP
5916 @item INIT_SECTION_ASM_OP
5917 If defined, a C expression whose value is a string, including spacing,
5918 containing the assembler operation to identify the following data as
5919 initialization code. If not defined, GCC will assume such a section does
5922 @findex FINI_SECTION_ASM_OP
5923 @item FINI_SECTION_ASM_OP
5924 If defined, a C expression whose value is a string, including spacing,
5925 containing the assembler operation to identify the following data as
5926 finalization code. If not defined, GCC will assume such a section does
5929 @findex CRT_CALL_STATIC_FUNCTION
5930 @item CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
5931 If defined, an ASM statement that switches to a different section
5932 via @var{section_op}, calls @var{function}, and switches back to
5933 the text section. This is used in @file{crtstuff.c} if
5934 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
5935 to initialization and finalization functions from the init and fini
5936 sections. By default, this macro uses a simple function call. Some
5937 ports need hand-crafted assembly code to avoid dependencies on
5938 registers initialized in the function prologue or to ensure that
5939 constant pools don't end up too far way in the text section.
5941 @findex FORCE_CODE_SECTION_ALIGN
5942 @item FORCE_CODE_SECTION_ALIGN
5943 If defined, an ASM statement that aligns a code section to some
5944 arbitrary boundary. This is used to force all fragments of the
5945 @code{.init} and @code{.fini} sections to have to same alignment
5946 and thus prevent the linker from having to add any padding.
5948 @findex EXTRA_SECTIONS
5951 @item EXTRA_SECTIONS
5952 A list of names for sections other than the standard two, which are
5953 @code{in_text} and @code{in_data}. You need not define this macro
5954 on a system with no other sections (that GCC needs to use).
5956 @findex EXTRA_SECTION_FUNCTIONS
5957 @findex text_section
5958 @findex data_section
5959 @item EXTRA_SECTION_FUNCTIONS
5960 One or more functions to be defined in @file{varasm.c}. These
5961 functions should do jobs analogous to those of @code{text_section} and
5962 @code{data_section}, for your additional sections. Do not define this
5963 macro if you do not define @code{EXTRA_SECTIONS}.
5965 @findex JUMP_TABLES_IN_TEXT_SECTION
5966 @item JUMP_TABLES_IN_TEXT_SECTION
5967 Define this macro to be an expression with a nonzero value if jump
5968 tables (for @code{tablejump} insns) should be output in the text
5969 section, along with the assembler instructions. Otherwise, the
5970 readonly data section is used.
5972 This macro is irrelevant if there is no separate readonly data section.
5975 @deftypefn {Target Hook} void TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
5976 Switches to the appropriate section for output of @var{exp}. You can
5977 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
5978 some sort. @var{reloc} indicates whether the initial value of @var{exp}
5979 requires link-time relocations. Bit 0 is set when variable contains
5980 local relocations only, while bit 1 is set for global relocations.
5981 Select the section by calling @code{data_section} or one of the
5982 alternatives for other sections. @var{align} is the constant alignment
5985 The default version of this function takes care of putting read-only
5986 variables in @code{readonly_data_section}.
5989 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
5990 Build up a unique section name, expressed as a @code{STRING_CST} node,
5991 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
5992 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
5993 the initial value of @var{exp} requires link-time relocations.
5995 The default version of this function appends the symbol name to the
5996 ELF section name that would normally be used for the symbol. For
5997 example, the function @code{foo} would be placed in @code{.text.foo}.
5998 Whatever the actual target object format, this is often good enough.
6001 @deftypefn {Target Hook} void TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
6002 Switches to the appropriate section for output of constant pool entry
6003 @var{x} in @var{mode}. You can assume that @var{x} is some kind of
6004 constant in RTL@. The argument @var{mode} is redundant except in the
6005 case of a @code{const_int} rtx. Select the section by calling
6006 @code{readonly_data_section} or one of the alternatives for other
6007 sections. @var{align} is the constant alignment in bits.
6009 The default version of this function takes care of putting symbolic
6010 constants in @code{flag_pic} mode in @code{data_section} and everything
6011 else in @code{readonly_data_section}.
6014 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, int @var{new_decl_p})
6015 Define this hook if references to a symbol or a constant must be
6016 treated differently depending on something about the variable or
6017 function named by the symbol (such as what section it is in).
6019 The hook is executed under two circumstances. One is immediately after
6020 the rtl for @var{decl} that represents a variable or a function has been
6021 created and stored in @code{DECL_RTL(@var{decl})}. The value of the rtl
6022 will be a @code{mem} whose address is a @code{symbol_ref}. The other is
6023 immediately after the rtl for @var{decl} that represents a constant has
6024 been created and stored in @code{TREE_CST_RTL (@var{decl})}. The macro
6025 is called once for each distinct constant in a source file.
6027 The @var{new_decl_p} argument will be true if this is the first time
6028 that @code{ENCODE_SECTION_INFO} has been invoked on this decl. It will
6029 be false for subsequent invocations, which will happen for duplicate
6030 declarations. Whether or not anything must be done for the duplicate
6031 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
6033 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
6034 The usual thing for this hook to do is to record a flag in the
6035 @code{symbol_ref} (such as @code{SYMBOL_REF_FLAG}) or to store a
6036 modified name string in the @code{symbol_ref} (if one bit is not
6037 enough information).
6040 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
6041 Decode @var{name} and return the real name part, sans
6042 the characters that @code{TARGET_ENCODE_SECTION_INFO}
6046 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
6047 Returns true if @var{exp} should be placed into a ``small data'' section.
6048 The default version of this hook always returns false.
6051 @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
6052 Contains the value true if the target places read-only
6053 ``small data'' into a separate section. The default value is false.
6056 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
6057 Returns true if @var{exp} names an object for which name resolution
6058 rules must resolve to the current ``module'' (dynamic shared library
6059 or executable image).
6061 The default version of this hook implements the name resolution rules
6062 for ELF, which has a looser model of global name binding than other
6063 currently supported object file formats.
6066 @deftypevar {Target Hook} bool TARGET_HAVE_TLS
6067 Contains the value true if the target supports thread-local storage.
6068 The default value is false.
6073 @section Position Independent Code
6074 @cindex position independent code
6077 This section describes macros that help implement generation of position
6078 independent code. Simply defining these macros is not enough to
6079 generate valid PIC; you must also add support to the macros
6080 @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
6081 well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
6082 @samp{movsi} to do something appropriate when the source operand
6083 contains a symbolic address. You may also need to alter the handling of
6084 switch statements so that they use relative addresses.
6085 @c i rearranged the order of the macros above to try to force one of
6086 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
6089 @findex PIC_OFFSET_TABLE_REGNUM
6090 @item PIC_OFFSET_TABLE_REGNUM
6091 The register number of the register used to address a table of static
6092 data addresses in memory. In some cases this register is defined by a
6093 processor's ``application binary interface'' (ABI)@. When this macro
6094 is defined, RTL is generated for this register once, as with the stack
6095 pointer and frame pointer registers. If this macro is not defined, it
6096 is up to the machine-dependent files to allocate such a register (if
6097 necessary). Note that this register must be fixed when in use (e.g.@:
6098 when @code{flag_pic} is true).
6100 @findex PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6101 @item PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
6102 Define this macro if the register defined by
6103 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
6104 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
6106 @findex FINALIZE_PIC
6108 By generating position-independent code, when two different programs (A
6109 and B) share a common library (libC.a), the text of the library can be
6110 shared whether or not the library is linked at the same address for both
6111 programs. In some of these environments, position-independent code
6112 requires not only the use of different addressing modes, but also
6113 special code to enable the use of these addressing modes.
6115 The @code{FINALIZE_PIC} macro serves as a hook to emit these special
6116 codes once the function is being compiled into assembly code, but not
6117 before. (It is not done before, because in the case of compiling an
6118 inline function, it would lead to multiple PIC prologues being
6119 included in functions which used inline functions and were compiled to
6122 @findex LEGITIMATE_PIC_OPERAND_P
6123 @item LEGITIMATE_PIC_OPERAND_P (@var{x})
6124 A C expression that is nonzero if @var{x} is a legitimate immediate
6125 operand on the target machine when generating position independent code.
6126 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
6127 check this. You can also assume @var{flag_pic} is true, so you need not
6128 check it either. You need not define this macro if all constants
6129 (including @code{SYMBOL_REF}) can be immediate operands when generating
6130 position independent code.
6133 @node Assembler Format
6134 @section Defining the Output Assembler Language
6136 This section describes macros whose principal purpose is to describe how
6137 to write instructions in assembler language---rather than what the
6141 * File Framework:: Structural information for the assembler file.
6142 * Data Output:: Output of constants (numbers, strings, addresses).
6143 * Uninitialized Data:: Output of uninitialized variables.
6144 * Label Output:: Output and generation of labels.
6145 * Initialization:: General principles of initialization
6146 and termination routines.
6147 * Macros for Initialization::
6148 Specific macros that control the handling of
6149 initialization and termination routines.
6150 * Instruction Output:: Output of actual instructions.
6151 * Dispatch Tables:: Output of jump tables.
6152 * Exception Region Output:: Output of exception region code.
6153 * Alignment Output:: Pseudo ops for alignment and skipping data.
6156 @node File Framework
6157 @subsection The Overall Framework of an Assembler File
6158 @cindex assembler format
6159 @cindex output of assembler code
6161 @c prevent bad page break with this line
6162 This describes the overall framework of an assembler file.
6165 @findex ASM_FILE_START
6166 @item ASM_FILE_START (@var{stream})
6167 A C expression which outputs to the stdio stream @var{stream}
6168 some appropriate text to go at the start of an assembler file.
6170 Normally this macro is defined to output a line containing
6171 @samp{#NO_APP}, which is a comment that has no effect on most
6172 assemblers but tells the GNU assembler that it can save time by not
6173 checking for certain assembler constructs.
6175 On systems that use SDB, it is necessary to output certain commands;
6176 see @file{attasm.h}.
6178 @findex ASM_FILE_END
6179 @item ASM_FILE_END (@var{stream})
6180 A C expression which outputs to the stdio stream @var{stream}
6181 some appropriate text to go at the end of an assembler file.
6183 If this macro is not defined, the default is to output nothing
6184 special at the end of the file. Most systems don't require any
6187 On systems that use SDB, it is necessary to output certain commands;
6188 see @file{attasm.h}.
6190 @findex ASM_COMMENT_START
6191 @item ASM_COMMENT_START
6192 A C string constant describing how to begin a comment in the target
6193 assembler language. The compiler assumes that the comment will end at
6194 the end of the line.
6198 A C string constant for text to be output before each @code{asm}
6199 statement or group of consecutive ones. Normally this is
6200 @code{"#APP"}, which is a comment that has no effect on most
6201 assemblers but tells the GNU assembler that it must check the lines
6202 that follow for all valid assembler constructs.
6206 A C string constant for text to be output after each @code{asm}
6207 statement or group of consecutive ones. Normally this is
6208 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
6209 time-saving assumptions that are valid for ordinary compiler output.
6211 @findex ASM_OUTPUT_SOURCE_FILENAME
6212 @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
6213 A C statement to output COFF information or DWARF debugging information
6214 which indicates that filename @var{name} is the current source file to
6215 the stdio stream @var{stream}.
6217 This macro need not be defined if the standard form of output
6218 for the file format in use is appropriate.
6220 @findex OUTPUT_QUOTED_STRING
6221 @item OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
6222 A C statement to output the string @var{string} to the stdio stream
6223 @var{stream}. If you do not call the function @code{output_quoted_string}
6224 in your config files, GCC will only call it to output filenames to
6225 the assembler source. So you can use it to canonicalize the format
6226 of the filename using this macro.
6228 @findex ASM_OUTPUT_SOURCE_LINE
6229 @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
6230 A C statement to output DBX or SDB debugging information before code
6231 for line number @var{line} of the current source file to the
6232 stdio stream @var{stream}.
6234 This macro need not be defined if the standard form of debugging
6235 information for the debugger in use is appropriate.
6237 @findex ASM_OUTPUT_IDENT
6238 @item ASM_OUTPUT_IDENT (@var{stream}, @var{string})
6239 A C statement to output something to the assembler file to handle a
6240 @samp{#ident} directive containing the text @var{string}. If this
6241 macro is not defined, nothing is output for a @samp{#ident} directive.
6243 @findex OBJC_PROLOGUE
6245 A C statement to output any assembler statements which are required to
6246 precede any Objective-C object definitions or message sending. The
6247 statement is executed only when compiling an Objective-C program.
6250 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
6251 Output assembly directives to switch to section @var{name}. The section
6252 should have attributes as specified by @var{flags}, which is a bit mask
6253 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
6254 is nonzero, it contains an alignment in bytes to be used for the section,
6255 otherwise some target default should be used. Only targets that must
6256 specify an alignment within the section directive need pay attention to
6257 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
6260 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
6261 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
6264 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
6265 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
6266 based on a variable or function decl, a section name, and whether or not the
6267 declaration's initializer may contain runtime relocations. @var{decl} may be
6268 null, in which case read-write data should be assumed.
6270 The default version if this function handles choosing code vs data,
6271 read-only vs read-write data, and @code{flag_pic}. You should only
6272 need to override this if your target has special flags that might be
6273 set via @code{__attribute__}.
6278 @subsection Output of Data
6281 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
6282 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
6283 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
6284 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
6285 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
6286 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
6287 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
6288 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
6289 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
6290 These hooks specify assembly directives for creating certain kinds
6291 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
6292 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
6293 aligned two-byte object, and so on. Any of the hooks may be
6294 @code{NULL}, indicating that no suitable directive is available.
6296 The compiler will print these strings at the start of a new line,
6297 followed immediately by the object's initial value. In most cases,
6298 the string should contain a tab, a pseudo-op, and then another tab.
6301 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
6302 The @code{assemble_integer} function uses this hook to output an
6303 integer object. @var{x} is the object's value, @var{size} is its size
6304 in bytes and @var{aligned_p} indicates whether it is aligned. The
6305 function should return @code{true} if it was able to output the
6306 object. If it returns false, @code{assemble_integer} will try to
6307 split the object into smaller parts.
6309 The default implementation of this hook will use the
6310 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
6311 when the relevant string is @code{NULL}.
6315 @findex OUTPUT_ADDR_CONST_EXTRA
6316 @item OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
6317 A C statement to recognize @var{rtx} patterns that
6318 @code{output_addr_const} can't deal with, and output assembly code to
6319 @var{stream} corresponding to the pattern @var{x}. This may be used to
6320 allow machine-dependent @code{UNSPEC}s to appear within constants.
6322 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
6323 @code{goto fail}, so that a standard error message is printed. If it
6324 prints an error message itself, by calling, for example,
6325 @code{output_operand_lossage}, it may just complete normally.
6327 @findex ASM_OUTPUT_ASCII
6328 @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
6329 A C statement to output to the stdio stream @var{stream} an assembler
6330 instruction to assemble a string constant containing the @var{len}
6331 bytes at @var{ptr}. @var{ptr} will be a C expression of type
6332 @code{char *} and @var{len} a C expression of type @code{int}.
6334 If the assembler has a @code{.ascii} pseudo-op as found in the
6335 Berkeley Unix assembler, do not define the macro
6336 @code{ASM_OUTPUT_ASCII}.
6338 @findex ASM_OUTPUT_FDESC
6339 @item ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
6340 A C statement to output word @var{n} of a function descriptor for
6341 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
6342 is defined, and is otherwise unused.
6344 @findex CONSTANT_POOL_BEFORE_FUNCTION
6345 @item CONSTANT_POOL_BEFORE_FUNCTION
6346 You may define this macro as a C expression. You should define the
6347 expression to have a nonzero value if GCC should output the constant
6348 pool for a function before the code for the function, or a zero value if
6349 GCC should output the constant pool after the function. If you do
6350 not define this macro, the usual case, GCC will output the constant
6351 pool before the function.
6353 @findex ASM_OUTPUT_POOL_PROLOGUE
6354 @item ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
6355 A C statement to output assembler commands to define the start of the
6356 constant pool for a function. @var{funname} is a string giving
6357 the name of the function. Should the return type of the function
6358 be required, it can be obtained via @var{fundecl}. @var{size}
6359 is the size, in bytes, of the constant pool that will be written
6360 immediately after this call.
6362 If no constant-pool prefix is required, the usual case, this macro need
6365 @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY
6366 @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
6367 A C statement (with or without semicolon) to output a constant in the
6368 constant pool, if it needs special treatment. (This macro need not do
6369 anything for RTL expressions that can be output normally.)
6371 The argument @var{file} is the standard I/O stream to output the
6372 assembler code on. @var{x} is the RTL expression for the constant to
6373 output, and @var{mode} is the machine mode (in case @var{x} is a
6374 @samp{const_int}). @var{align} is the required alignment for the value
6375 @var{x}; you should output an assembler directive to force this much
6378 The argument @var{labelno} is a number to use in an internal label for
6379 the address of this pool entry. The definition of this macro is
6380 responsible for outputting the label definition at the proper place.
6381 Here is how to do this:
6384 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
6387 When you output a pool entry specially, you should end with a
6388 @code{goto} to the label @var{jumpto}. This will prevent the same pool
6389 entry from being output a second time in the usual manner.
6391 You need not define this macro if it would do nothing.
6393 @findex CONSTANT_AFTER_FUNCTION_P
6394 @item CONSTANT_AFTER_FUNCTION_P (@var{exp})
6395 Define this macro as a C expression which is nonzero if the constant
6396 @var{exp}, of type @code{tree}, should be output after the code for a
6397 function. The compiler will normally output all constants before the
6398 function; you need not define this macro if this is OK@.
6400 @findex ASM_OUTPUT_POOL_EPILOGUE
6401 @item ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
6402 A C statement to output assembler commands to at the end of the constant
6403 pool for a function. @var{funname} is a string giving the name of the
6404 function. Should the return type of the function be required, you can
6405 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
6406 constant pool that GCC wrote immediately before this call.
6408 If no constant-pool epilogue is required, the usual case, you need not
6411 @findex IS_ASM_LOGICAL_LINE_SEPARATOR
6412 @item IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C})
6413 Define this macro as a C expression which is nonzero if @var{C} is
6414 used as a logical line separator by the assembler.
6416 If you do not define this macro, the default is that only
6417 the character @samp{;} is treated as a logical line separator.
6420 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
6421 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
6422 These target hooks are C string constants, describing the syntax in the
6423 assembler for grouping arithmetic expressions. If not overridden, they
6424 default to normal parentheses, which is correct for most assemblers.
6427 These macros are provided by @file{real.h} for writing the definitions
6428 of @code{ASM_OUTPUT_DOUBLE} and the like:
6431 @item REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
6432 @itemx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
6433 @itemx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
6434 @findex REAL_VALUE_TO_TARGET_SINGLE
6435 @findex REAL_VALUE_TO_TARGET_DOUBLE
6436 @findex REAL_VALUE_TO_TARGET_LONG_DOUBLE
6437 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
6438 floating point representation, and store its bit pattern in the variable
6439 @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE}, this variable should
6440 be a simple @code{long int}. For the others, it should be an array of
6441 @code{long int}. The number of elements in this array is determined by
6442 the size of the desired target floating point data type: 32 bits of it
6443 go in each @code{long int} array element. Each array element holds 32
6444 bits of the result, even if @code{long int} is wider than 32 bits on the
6447 The array element values are designed so that you can print them out
6448 using @code{fprintf} in the order they should appear in the target
6452 @node Uninitialized Data
6453 @subsection Output of Uninitialized Variables
6455 Each of the macros in this section is used to do the whole job of
6456 outputting a single uninitialized variable.
6459 @findex ASM_OUTPUT_COMMON
6460 @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6461 A C statement (sans semicolon) to output to the stdio stream
6462 @var{stream} the assembler definition of a common-label named
6463 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6464 is the size rounded up to whatever alignment the caller wants.
6466 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6467 output the name itself; before and after that, output the additional
6468 assembler syntax for defining the name, and a newline.
6470 This macro controls how the assembler definitions of uninitialized
6471 common global variables are output.
6473 @findex ASM_OUTPUT_ALIGNED_COMMON
6474 @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
6475 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
6476 separate, explicit argument. If you define this macro, it is used in
6477 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
6478 handling the required alignment of the variable. The alignment is specified
6479 as the number of bits.
6481 @findex ASM_OUTPUT_ALIGNED_DECL_COMMON
6482 @item ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6483 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
6484 variable to be output, if there is one, or @code{NULL_TREE} if there
6485 is no corresponding variable. If you define this macro, GCC will use it
6486 in place of both @code{ASM_OUTPUT_COMMON} and
6487 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
6488 the variable's decl in order to chose what to output.
6490 @findex ASM_OUTPUT_SHARED_COMMON
6491 @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
6492 If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
6493 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
6496 @findex ASM_OUTPUT_BSS
6497 @item ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6498 A C statement (sans semicolon) to output to the stdio stream
6499 @var{stream} the assembler definition of uninitialized global @var{decl} named
6500 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6501 is the size rounded up to whatever alignment the caller wants.
6503 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
6504 defining this macro. If unable, use the expression
6505 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
6506 before and after that, output the additional assembler syntax for defining
6507 the name, and a newline.
6509 This macro controls how the assembler definitions of uninitialized global
6510 variables are output. This macro exists to properly support languages like
6511 C++ which do not have @code{common} data. However, this macro currently
6512 is not defined for all targets. If this macro and
6513 @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON}
6514 or @code{ASM_OUTPUT_ALIGNED_COMMON} or
6515 @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used.
6517 @findex ASM_OUTPUT_ALIGNED_BSS
6518 @item ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6519 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
6520 separate, explicit argument. If you define this macro, it is used in
6521 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
6522 handling the required alignment of the variable. The alignment is specified
6523 as the number of bits.
6525 Try to use function @code{asm_output_aligned_bss} defined in file
6526 @file{varasm.c} when defining this macro.
6528 @findex ASM_OUTPUT_SHARED_BSS
6529 @item ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
6530 If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it
6531 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS}
6534 @findex ASM_OUTPUT_LOCAL
6535 @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6536 A C statement (sans semicolon) to output to the stdio stream
6537 @var{stream} the assembler definition of a local-common-label named
6538 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
6539 is the size rounded up to whatever alignment the caller wants.
6541 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6542 output the name itself; before and after that, output the additional
6543 assembler syntax for defining the name, and a newline.
6545 This macro controls how the assembler definitions of uninitialized
6546 static variables are output.
6548 @findex ASM_OUTPUT_ALIGNED_LOCAL
6549 @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
6550 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
6551 separate, explicit argument. If you define this macro, it is used in
6552 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
6553 handling the required alignment of the variable. The alignment is specified
6554 as the number of bits.
6556 @findex ASM_OUTPUT_ALIGNED_DECL_LOCAL
6557 @item ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
6558 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
6559 variable to be output, if there is one, or @code{NULL_TREE} if there
6560 is no corresponding variable. If you define this macro, GCC will use it
6561 in place of both @code{ASM_OUTPUT_DECL} and
6562 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
6563 the variable's decl in order to chose what to output.
6565 @findex ASM_OUTPUT_SHARED_LOCAL
6566 @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
6567 If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
6568 is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
6573 @subsection Output and Generation of Labels
6575 @c prevent bad page break with this line
6576 This is about outputting labels.
6579 @findex ASM_OUTPUT_LABEL
6580 @findex assemble_name
6581 @item ASM_OUTPUT_LABEL (@var{stream}, @var{name})
6582 A C statement (sans semicolon) to output to the stdio stream
6583 @var{stream} the assembler definition of a label named @var{name}.
6584 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
6585 output the name itself; before and after that, output the additional
6586 assembler syntax for defining the name, and a newline. A default
6587 definition of this macro is provided which is correct for most systems.
6591 A C string containing the appropriate assembler directive to specify the
6592 size of a symbol, without any arguments. On systems that use ELF, the
6593 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
6594 systems, the default is not to define this macro.
6596 Define this macro only if it is correct to use the default definitions
6597 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
6598 for your system. If you need your own custom definitions of those
6599 macros, or if you do not need explicit symbol sizes at all, do not
6602 @findex ASM_OUTPUT_SIZE_DIRECTIVE
6603 @item ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
6604 A C statement (sans semicolon) to output to the stdio stream
6605 @var{stream} a directive telling the assembler that the size of the
6606 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
6607 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6610 @findex ASM_OUTPUT_MEASURED_SIZE
6611 @item ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
6612 A C statement (sans semicolon) to output to the stdio stream
6613 @var{stream} a directive telling the assembler to calculate the size of
6614 the symbol @var{name} by subtracting its address from the current
6617 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
6618 provided. The default assumes that the assembler recognizes a special
6619 @samp{.} symbol as referring to the current address, and can calculate
6620 the difference between this and another symbol. If your assembler does
6621 not recognize @samp{.} or cannot do calculations with it, you will need
6622 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
6626 A C string containing the appropriate assembler directive to specify the
6627 type of a symbol, without any arguments. On systems that use ELF, the
6628 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
6629 systems, the default is not to define this macro.
6631 Define this macro only if it is correct to use the default definition of
6632 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6633 custom definition of this macro, or if you do not need explicit symbol
6634 types at all, do not define this macro.
6636 @findex TYPE_OPERAND_FMT
6637 @item TYPE_OPERAND_FMT
6638 A C string which specifies (using @code{printf} syntax) the format of
6639 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
6640 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
6641 the default is not to define this macro.
6643 Define this macro only if it is correct to use the default definition of
6644 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
6645 custom definition of this macro, or if you do not need explicit symbol
6646 types at all, do not define this macro.
6648 @findex ASM_OUTPUT_TYPE_DIRECTIVE
6649 @item ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
6650 A C statement (sans semicolon) to output to the stdio stream
6651 @var{stream} a directive telling the assembler that the type of the
6652 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
6653 that string is always either @samp{"function"} or @samp{"object"}, but
6654 you should not count on this.
6656 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
6657 definition of this macro is provided.
6659 @findex ASM_DECLARE_FUNCTION_NAME
6660 @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
6661 A C statement (sans semicolon) to output to the stdio stream
6662 @var{stream} any text necessary for declaring the name @var{name} of a
6663 function which is being defined. This macro is responsible for
6664 outputting the label definition (perhaps using
6665 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
6666 @code{FUNCTION_DECL} tree node representing the function.
6668 If this macro is not defined, then the function name is defined in the
6669 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6671 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
6674 @findex ASM_DECLARE_FUNCTION_SIZE
6675 @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
6676 A C statement (sans semicolon) to output to the stdio stream
6677 @var{stream} any text necessary for declaring the size of a function
6678 which is being defined. The argument @var{name} is the name of the
6679 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
6680 representing the function.
6682 If this macro is not defined, then the function size is not defined.
6684 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
6687 @findex ASM_DECLARE_OBJECT_NAME
6688 @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
6689 A C statement (sans semicolon) to output to the stdio stream
6690 @var{stream} any text necessary for declaring the name @var{name} of an
6691 initialized variable which is being defined. This macro must output the
6692 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
6693 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
6695 If this macro is not defined, then the variable name is defined in the
6696 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
6698 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
6699 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
6701 @findex ASM_DECLARE_REGISTER_GLOBAL
6702 @item ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
6703 A C statement (sans semicolon) to output to the stdio stream
6704 @var{stream} any text necessary for claiming a register @var{regno}
6705 for a global variable @var{decl} with name @var{name}.
6707 If you don't define this macro, that is equivalent to defining it to do
6710 @findex ASM_FINISH_DECLARE_OBJECT
6711 @item ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
6712 A C statement (sans semicolon) to finish up declaring a variable name
6713 once the compiler has processed its initializer fully and thus has had a
6714 chance to determine the size of an array when controlled by an
6715 initializer. This is used on systems where it's necessary to declare
6716 something about the size of the object.
6718 If you don't define this macro, that is equivalent to defining it to do
6721 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
6722 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
6725 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
6726 This target hook is a function to output to the stdio stream
6727 @var{stream} some commands that will make the label @var{name} global;
6728 that is, available for reference from other files.
6730 The default implementation relies on a proper definition of
6731 @code{GLOBAL_ASM_OP}.
6735 @findex ASM_WEAKEN_LABEL
6736 @item ASM_WEAKEN_LABEL (@var{stream}, @var{name})
6737 A C statement (sans semicolon) to output to the stdio stream
6738 @var{stream} some commands that will make the label @var{name} weak;
6739 that is, available for reference from other files but only used if
6740 no other definition is available. Use the expression
6741 @code{assemble_name (@var{stream}, @var{name})} to output the name
6742 itself; before and after that, output the additional assembler syntax
6743 for making that name weak, and a newline.
6745 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
6746 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
6749 @findex ASM_WEAKEN_DECL
6750 @item ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
6751 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
6752 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
6753 or variable decl. If @var{value} is not @code{NULL}, this C statement
6754 should output to the stdio stream @var{stream} assembler code which
6755 defines (equates) the weak symbol @var{name} to have the value
6756 @var{value}. If @var{value} is @code{NULL}, it should output commands
6757 to make @var{name} weak.
6759 @findex SUPPORTS_WEAK
6761 A C expression which evaluates to true if the target supports weak symbols.
6763 If you don't define this macro, @file{defaults.h} provides a default
6764 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
6765 is defined, the default definition is @samp{1}; otherwise, it is
6766 @samp{0}. Define this macro if you want to control weak symbol support
6767 with a compiler flag such as @option{-melf}.
6769 @findex MAKE_DECL_ONE_ONLY (@var{decl})
6770 @item MAKE_DECL_ONE_ONLY
6771 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
6772 public symbol such that extra copies in multiple translation units will
6773 be discarded by the linker. Define this macro if your object file
6774 format provides support for this concept, such as the @samp{COMDAT}
6775 section flags in the Microsoft Windows PE/COFF format, and this support
6776 requires changes to @var{decl}, such as putting it in a separate section.
6778 @findex SUPPORTS_ONE_ONLY
6779 @item SUPPORTS_ONE_ONLY
6780 A C expression which evaluates to true if the target supports one-only
6783 If you don't define this macro, @file{varasm.c} provides a default
6784 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
6785 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
6786 you want to control one-only symbol support with a compiler flag, or if
6787 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
6788 be emitted as one-only.
6790 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
6791 This target hook is a function to output to @var{asm_out_file} some
6792 commands that will make the symbol(s) associated with @var{decl} have
6793 hidden, protected or internal visibility as specified by @var{visibility}.
6796 @findex ASM_OUTPUT_EXTERNAL
6797 @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
6798 A C statement (sans semicolon) to output to the stdio stream
6799 @var{stream} any text necessary for declaring the name of an external
6800 symbol named @var{name} which is referenced in this compilation but
6801 not defined. The value of @var{decl} is the tree node for the
6804 This macro need not be defined if it does not need to output anything.
6805 The GNU assembler and most Unix assemblers don't require anything.
6807 @findex ASM_OUTPUT_EXTERNAL_LIBCALL
6808 @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref})
6809 A C statement (sans semicolon) to output on @var{stream} an assembler
6810 pseudo-op to declare a library function name external. The name of the
6811 library function is given by @var{symref}, which has type @code{rtx} and
6812 is a @code{symbol_ref}.
6814 This macro need not be defined if it does not need to output anything.
6815 The GNU assembler and most Unix assemblers don't require anything.
6817 @findex ASM_OUTPUT_LABELREF
6818 @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
6819 A C statement (sans semicolon) to output to the stdio stream
6820 @var{stream} a reference in assembler syntax to a label named
6821 @var{name}. This should add @samp{_} to the front of the name, if that
6822 is customary on your operating system, as it is in most Berkeley Unix
6823 systems. This macro is used in @code{assemble_name}.
6825 @findex ASM_OUTPUT_SYMBOL_REF
6826 @item ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
6827 A C statement (sans semicolon) to output a reference to
6828 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
6829 will be used to output the name of the symbol. This macro may be used
6830 to modify the way a symbol is referenced depending on information
6831 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
6833 @findex ASM_OUTPUT_LABEL_REF
6834 @item ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
6835 A C statement (sans semicolon) to output a reference to @var{buf}, the
6836 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
6837 @code{assemble_name} will be used to output the name of the symbol.
6838 This macro is not used by @code{output_asm_label}, or the @code{%l}
6839 specifier that calls it; the intention is that this macro should be set
6840 when it is necessary to output a label differently when its address is
6844 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
6845 A function to output to the stdio stream @var{stream} a label whose
6846 name is made from the string @var{prefix} and the number @var{labelno}.
6848 It is absolutely essential that these labels be distinct from the labels
6849 used for user-level functions and variables. Otherwise, certain programs
6850 will have name conflicts with internal labels.
6852 It is desirable to exclude internal labels from the symbol table of the
6853 object file. Most assemblers have a naming convention for labels that
6854 should be excluded; on many systems, the letter @samp{L} at the
6855 beginning of a label has this effect. You should find out what
6856 convention your system uses, and follow it.
6858 The default version of this function utilizes ASM_GENERATE_INTERNAL_LABEL.
6863 @findex ASM_OUTPUT_DEBUG_LABEL
6864 @item ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
6865 A C statement to output to the stdio stream @var{stream} a debug info
6866 label whose name is made from the string @var{prefix} and the number
6867 @var{num}. This is useful for VLIW targets, where debug info labels
6868 may need to be treated differently than branch target labels. On some
6869 systems, branch target labels must be at the beginning of instruction
6870 bundles, but debug info labels can occur in the middle of instruction
6873 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
6876 @findex ASM_GENERATE_INTERNAL_LABEL
6877 @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
6878 A C statement to store into the string @var{string} a label whose name
6879 is made from the string @var{prefix} and the number @var{num}.
6881 This string, when output subsequently by @code{assemble_name}, should
6882 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
6883 with the same @var{prefix} and @var{num}.
6885 If the string begins with @samp{*}, then @code{assemble_name} will
6886 output the rest of the string unchanged. It is often convenient for
6887 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
6888 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
6889 to output the string, and may change it. (Of course,
6890 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
6891 you should know what it does on your machine.)
6893 @findex ASM_FORMAT_PRIVATE_NAME
6894 @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
6895 A C expression to assign to @var{outvar} (which is a variable of type
6896 @code{char *}) a newly allocated string made from the string
6897 @var{name} and the number @var{number}, with some suitable punctuation
6898 added. Use @code{alloca} to get space for the string.
6900 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
6901 produce an assembler label for an internal static variable whose name is
6902 @var{name}. Therefore, the string must be such as to result in valid
6903 assembler code. The argument @var{number} is different each time this
6904 macro is executed; it prevents conflicts between similarly-named
6905 internal static variables in different scopes.
6907 Ideally this string should not be a valid C identifier, to prevent any
6908 conflict with the user's own symbols. Most assemblers allow periods
6909 or percent signs in assembler symbols; putting at least one of these
6910 between the name and the number will suffice.
6912 If this macro is not defined, a default definition will be provided
6913 which is correct for most systems.
6915 @findex ASM_OUTPUT_DEF
6916 @item ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
6917 A C statement to output to the stdio stream @var{stream} assembler code
6918 which defines (equates) the symbol @var{name} to have the value @var{value}.
6921 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6922 correct for most systems.
6924 @findex ASM_OUTPUT_DEF_FROM_DECLS
6925 @item ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
6926 A C statement to output to the stdio stream @var{stream} assembler code
6927 which defines (equates) the symbol whose tree node is @var{decl_of_name}
6928 to have the value of the tree node @var{decl_of_value}. This macro will
6929 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
6930 the tree nodes are available.
6933 If @code{SET_ASM_OP} is defined, a default definition is provided which is
6934 correct for most systems.
6936 @findex ASM_OUTPUT_WEAK_ALIAS
6937 @item ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
6938 A C statement to output to the stdio stream @var{stream} assembler code
6939 which defines (equates) the weak symbol @var{name} to have the value
6940 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
6941 an undefined weak symbol.
6943 Define this macro if the target only supports weak aliases; define
6944 @code{ASM_OUTPUT_DEF} instead if possible.
6946 @findex OBJC_GEN_METHOD_LABEL
6947 @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
6948 Define this macro to override the default assembler names used for
6949 Objective-C methods.
6951 The default name is a unique method number followed by the name of the
6952 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
6953 the category is also included in the assembler name (e.g.@:
6956 These names are safe on most systems, but make debugging difficult since
6957 the method's selector is not present in the name. Therefore, particular
6958 systems define other ways of computing names.
6960 @var{buf} is an expression of type @code{char *} which gives you a
6961 buffer in which to store the name; its length is as long as
6962 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
6963 50 characters extra.
6965 The argument @var{is_inst} specifies whether the method is an instance
6966 method or a class method; @var{class_name} is the name of the class;
6967 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
6968 in a category); and @var{sel_name} is the name of the selector.
6970 On systems where the assembler can handle quoted names, you can use this
6971 macro to provide more human-readable names.
6973 @findex ASM_DECLARE_CLASS_REFERENCE
6974 @item ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
6975 A C statement (sans semicolon) to output to the stdio stream
6976 @var{stream} commands to declare that the label @var{name} is an
6977 Objective-C class reference. This is only needed for targets whose
6978 linkers have special support for NeXT-style runtimes.
6980 @findex ASM_DECLARE_UNRESOLVED_REFERENCE
6981 @item ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
6982 A C statement (sans semicolon) to output to the stdio stream
6983 @var{stream} commands to declare that the label @var{name} is an
6984 unresolved Objective-C class reference. This is only needed for targets
6985 whose linkers have special support for NeXT-style runtimes.
6988 @node Initialization
6989 @subsection How Initialization Functions Are Handled
6990 @cindex initialization routines
6991 @cindex termination routines
6992 @cindex constructors, output of
6993 @cindex destructors, output of
6995 The compiled code for certain languages includes @dfn{constructors}
6996 (also called @dfn{initialization routines})---functions to initialize
6997 data in the program when the program is started. These functions need
6998 to be called before the program is ``started''---that is to say, before
6999 @code{main} is called.
7001 Compiling some languages generates @dfn{destructors} (also called
7002 @dfn{termination routines}) that should be called when the program
7005 To make the initialization and termination functions work, the compiler
7006 must output something in the assembler code to cause those functions to
7007 be called at the appropriate time. When you port the compiler to a new
7008 system, you need to specify how to do this.
7010 There are two major ways that GCC currently supports the execution of
7011 initialization and termination functions. Each way has two variants.
7012 Much of the structure is common to all four variations.
7014 @findex __CTOR_LIST__
7015 @findex __DTOR_LIST__
7016 The linker must build two lists of these functions---a list of
7017 initialization functions, called @code{__CTOR_LIST__}, and a list of
7018 termination functions, called @code{__DTOR_LIST__}.
7020 Each list always begins with an ignored function pointer (which may hold
7021 0, @minus{}1, or a count of the function pointers after it, depending on
7022 the environment). This is followed by a series of zero or more function
7023 pointers to constructors (or destructors), followed by a function
7024 pointer containing zero.
7026 Depending on the operating system and its executable file format, either
7027 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
7028 time and exit time. Constructors are called in reverse order of the
7029 list; destructors in forward order.
7031 The best way to handle static constructors works only for object file
7032 formats which provide arbitrarily-named sections. A section is set
7033 aside for a list of constructors, and another for a list of destructors.
7034 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
7035 object file that defines an initialization function also puts a word in
7036 the constructor section to point to that function. The linker
7037 accumulates all these words into one contiguous @samp{.ctors} section.
7038 Termination functions are handled similarly.
7040 This method will be chosen as the default by @file{target-def.h} if
7041 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
7042 support arbitrary sections, but does support special designated
7043 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
7044 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
7046 When arbitrary sections are available, there are two variants, depending
7047 upon how the code in @file{crtstuff.c} is called. On systems that
7048 support a @dfn{.init} section which is executed at program startup,
7049 parts of @file{crtstuff.c} are compiled into that section. The
7050 program is linked by the @command{gcc} driver like this:
7053 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
7056 The prologue of a function (@code{__init}) appears in the @code{.init}
7057 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
7058 for the function @code{__fini} in the @dfn{.fini} section. Normally these
7059 files are provided by the operating system or by the GNU C library, but
7060 are provided by GCC for a few targets.
7062 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
7063 compiled from @file{crtstuff.c}. They contain, among other things, code
7064 fragments within the @code{.init} and @code{.fini} sections that branch
7065 to routines in the @code{.text} section. The linker will pull all parts
7066 of a section together, which results in a complete @code{__init} function
7067 that invokes the routines we need at startup.
7069 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
7072 If no init section is available, when GCC compiles any function called
7073 @code{main} (or more accurately, any function designated as a program
7074 entry point by the language front end calling @code{expand_main_function}),
7075 it inserts a procedure call to @code{__main} as the first executable code
7076 after the function prologue. The @code{__main} function is defined
7077 in @file{libgcc2.c} and runs the global constructors.
7079 In file formats that don't support arbitrary sections, there are again
7080 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
7081 and an `a.out' format must be used. In this case,
7082 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
7083 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
7084 and with the address of the void function containing the initialization
7085 code as its value. The GNU linker recognizes this as a request to add
7086 the value to a @dfn{set}; the values are accumulated, and are eventually
7087 placed in the executable as a vector in the format described above, with
7088 a leading (ignored) count and a trailing zero element.
7089 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
7090 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
7091 the compilation of @code{main} to call @code{__main} as above, starting
7092 the initialization process.
7094 The last variant uses neither arbitrary sections nor the GNU linker.
7095 This is preferable when you want to do dynamic linking and when using
7096 file formats which the GNU linker does not support, such as `ECOFF'@. In
7097 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
7098 termination functions are recognized simply by their names. This requires
7099 an extra program in the linkage step, called @command{collect2}. This program
7100 pretends to be the linker, for use with GCC; it does its job by running
7101 the ordinary linker, but also arranges to include the vectors of
7102 initialization and termination functions. These functions are called
7103 via @code{__main} as described above. In order to use this method,
7104 @code{use_collect2} must be defined in the target in @file{config.gcc}.
7107 The following section describes the specific macros that control and
7108 customize the handling of initialization and termination functions.
7111 @node Macros for Initialization
7112 @subsection Macros Controlling Initialization Routines
7114 Here are the macros that control how the compiler handles initialization
7115 and termination functions:
7118 @findex INIT_SECTION_ASM_OP
7119 @item INIT_SECTION_ASM_OP
7120 If defined, a C string constant, including spacing, for the assembler
7121 operation to identify the following data as initialization code. If not
7122 defined, GCC will assume such a section does not exist. When you are
7123 using special sections for initialization and termination functions, this
7124 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
7125 run the initialization functions.
7127 @item HAS_INIT_SECTION
7128 @findex HAS_INIT_SECTION
7129 If defined, @code{main} will not call @code{__main} as described above.
7130 This macro should be defined for systems that control start-up code
7131 on a symbol-by-symbol basis, such as OSF/1, and should not
7132 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
7134 @item LD_INIT_SWITCH
7135 @findex LD_INIT_SWITCH
7136 If defined, a C string constant for a switch that tells the linker that
7137 the following symbol is an initialization routine.
7139 @item LD_FINI_SWITCH
7140 @findex LD_FINI_SWITCH
7141 If defined, a C string constant for a switch that tells the linker that
7142 the following symbol is a finalization routine.
7144 @item COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
7145 If defined, a C statement that will write a function that can be
7146 automatically called when a shared library is loaded. The function
7147 should call @var{func}, which takes no arguments. If not defined, and
7148 the object format requires an explicit initialization function, then a
7149 function called @code{_GLOBAL__DI} will be generated.
7151 This function and the following one are used by collect2 when linking a
7152 shared library that needs constructors or destructors, or has DWARF2
7153 exception tables embedded in the code.
7155 @item COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
7156 If defined, a C statement that will write a function that can be
7157 automatically called when a shared library is unloaded. The function
7158 should call @var{func}, which takes no arguments. If not defined, and
7159 the object format requires an explicit finalization function, then a
7160 function called @code{_GLOBAL__DD} will be generated.
7163 @findex INVOKE__main
7164 If defined, @code{main} will call @code{__main} despite the presence of
7165 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
7166 where the init section is not actually run automatically, but is still
7167 useful for collecting the lists of constructors and destructors.
7169 @item SUPPORTS_INIT_PRIORITY
7170 @findex SUPPORTS_INIT_PRIORITY
7171 If nonzero, the C++ @code{init_priority} attribute is supported and the
7172 compiler should emit instructions to control the order of initialization
7173 of objects. If zero, the compiler will issue an error message upon
7174 encountering an @code{init_priority} attribute.
7177 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
7178 This value is true if the target supports some ``native'' method of
7179 collecting constructors and destructors to be run at startup and exit.
7180 It is false if we must use @command{collect2}.
7183 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
7184 If defined, a function that outputs assembler code to arrange to call
7185 the function referenced by @var{symbol} at initialization time.
7187 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
7188 no arguments and with no return value. If the target supports initialization
7189 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
7190 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
7192 If this macro is not defined by the target, a suitable default will
7193 be chosen if (1) the target supports arbitrary section names, (2) the
7194 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
7198 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
7199 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
7200 functions rather than initialization functions.
7203 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
7204 generated for the generated object file will have static linkage.
7206 If your system uses @command{collect2} as the means of processing
7207 constructors, then that program normally uses @command{nm} to scan
7208 an object file for constructor functions to be called.
7210 On certain kinds of systems, you can define these macros to make
7211 @command{collect2} work faster (and, in some cases, make it work at all):
7214 @findex OBJECT_FORMAT_COFF
7215 @item OBJECT_FORMAT_COFF
7216 Define this macro if the system uses COFF (Common Object File Format)
7217 object files, so that @command{collect2} can assume this format and scan
7218 object files directly for dynamic constructor/destructor functions.
7220 @findex OBJECT_FORMAT_ROSE
7221 @item OBJECT_FORMAT_ROSE
7222 Define this macro if the system uses ROSE format object files, so that
7223 @command{collect2} can assume this format and scan object files directly
7224 for dynamic constructor/destructor functions.
7226 These macros are effective only in a native compiler; @command{collect2} as
7227 part of a cross compiler always uses @command{nm} for the target machine.
7229 @findex REAL_NM_FILE_NAME
7230 @item REAL_NM_FILE_NAME
7231 Define this macro as a C string constant containing the file name to use
7232 to execute @command{nm}. The default is to search the path normally for
7235 If your system supports shared libraries and has a program to list the
7236 dynamic dependencies of a given library or executable, you can define
7237 these macros to enable support for running initialization and
7238 termination functions in shared libraries:
7242 Define this macro to a C string constant containing the name of the program
7243 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
7245 @findex PARSE_LDD_OUTPUT
7246 @item PARSE_LDD_OUTPUT (@var{ptr})
7247 Define this macro to be C code that extracts filenames from the output
7248 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
7249 of type @code{char *} that points to the beginning of a line of output
7250 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
7251 code must advance @var{ptr} to the beginning of the filename on that
7252 line. Otherwise, it must set @var{ptr} to @code{NULL}.
7255 @node Instruction Output
7256 @subsection Output of Assembler Instructions
7258 @c prevent bad page break with this line
7259 This describes assembler instruction output.
7262 @findex REGISTER_NAMES
7263 @item REGISTER_NAMES
7264 A C initializer containing the assembler's names for the machine
7265 registers, each one as a C string constant. This is what translates
7266 register numbers in the compiler into assembler language.
7268 @findex ADDITIONAL_REGISTER_NAMES
7269 @item ADDITIONAL_REGISTER_NAMES
7270 If defined, a C initializer for an array of structures containing a name
7271 and a register number. This macro defines additional names for hard
7272 registers, thus allowing the @code{asm} option in declarations to refer
7273 to registers using alternate names.
7275 @findex ASM_OUTPUT_OPCODE
7276 @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
7277 Define this macro if you are using an unusual assembler that
7278 requires different names for the machine instructions.
7280 The definition is a C statement or statements which output an
7281 assembler instruction opcode to the stdio stream @var{stream}. The
7282 macro-operand @var{ptr} is a variable of type @code{char *} which
7283 points to the opcode name in its ``internal'' form---the form that is
7284 written in the machine description. The definition should output the
7285 opcode name to @var{stream}, performing any translation you desire, and
7286 increment the variable @var{ptr} to point at the end of the opcode
7287 so that it will not be output twice.
7289 In fact, your macro definition may process less than the entire opcode
7290 name, or more than the opcode name; but if you want to process text
7291 that includes @samp{%}-sequences to substitute operands, you must take
7292 care of the substitution yourself. Just be sure to increment
7293 @var{ptr} over whatever text should not be output normally.
7295 @findex recog_data.operand
7296 If you need to look at the operand values, they can be found as the
7297 elements of @code{recog_data.operand}.
7299 If the macro definition does nothing, the instruction is output
7302 @findex FINAL_PRESCAN_INSN
7303 @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
7304 If defined, a C statement to be executed just prior to the output of
7305 assembler code for @var{insn}, to modify the extracted operands so
7306 they will be output differently.
7308 Here the argument @var{opvec} is the vector containing the operands
7309 extracted from @var{insn}, and @var{noperands} is the number of
7310 elements of the vector which contain meaningful data for this insn.
7311 The contents of this vector are what will be used to convert the insn
7312 template into assembler code, so you can change the assembler output
7313 by changing the contents of the vector.
7315 This macro is useful when various assembler syntaxes share a single
7316 file of instruction patterns; by defining this macro differently, you
7317 can cause a large class of instructions to be output differently (such
7318 as with rearranged operands). Naturally, variations in assembler
7319 syntax affecting individual insn patterns ought to be handled by
7320 writing conditional output routines in those patterns.
7322 If this macro is not defined, it is equivalent to a null statement.
7324 @findex FINAL_PRESCAN_LABEL
7325 @item FINAL_PRESCAN_LABEL
7326 If defined, @code{FINAL_PRESCAN_INSN} will be called on each
7327 @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and
7328 @var{noperands} will be zero.
7330 @findex PRINT_OPERAND
7331 @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
7332 A C compound statement to output to stdio stream @var{stream} the
7333 assembler syntax for an instruction operand @var{x}. @var{x} is an
7336 @var{code} is a value that can be used to specify one of several ways
7337 of printing the operand. It is used when identical operands must be
7338 printed differently depending on the context. @var{code} comes from
7339 the @samp{%} specification that was used to request printing of the
7340 operand. If the specification was just @samp{%@var{digit}} then
7341 @var{code} is 0; if the specification was @samp{%@var{ltr}
7342 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
7345 If @var{x} is a register, this macro should print the register's name.
7346 The names can be found in an array @code{reg_names} whose type is
7347 @code{char *[]}. @code{reg_names} is initialized from
7348 @code{REGISTER_NAMES}.
7350 When the machine description has a specification @samp{%@var{punct}}
7351 (a @samp{%} followed by a punctuation character), this macro is called
7352 with a null pointer for @var{x} and the punctuation character for
7355 @findex PRINT_OPERAND_PUNCT_VALID_P
7356 @item PRINT_OPERAND_PUNCT_VALID_P (@var{code})
7357 A C expression which evaluates to true if @var{code} is a valid
7358 punctuation character for use in the @code{PRINT_OPERAND} macro. If
7359 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
7360 punctuation characters (except for the standard one, @samp{%}) are used
7363 @findex PRINT_OPERAND_ADDRESS
7364 @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
7365 A C compound statement to output to stdio stream @var{stream} the
7366 assembler syntax for an instruction operand that is a memory reference
7367 whose address is @var{x}. @var{x} is an RTL expression.
7369 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
7370 On some machines, the syntax for a symbolic address depends on the
7371 section that the address refers to. On these machines, define the hook
7372 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
7373 @code{symbol_ref}, and then check for it here. @xref{Assembler Format}.
7375 @findex DBR_OUTPUT_SEQEND
7376 @findex dbr_sequence_length
7377 @item DBR_OUTPUT_SEQEND(@var{file})
7378 A C statement, to be executed after all slot-filler instructions have
7379 been output. If necessary, call @code{dbr_sequence_length} to
7380 determine the number of slots filled in a sequence (zero if not
7381 currently outputting a sequence), to decide how many no-ops to output,
7384 Don't define this macro if it has nothing to do, but it is helpful in
7385 reading assembly output if the extent of the delay sequence is made
7386 explicit (e.g.@: with white space).
7388 @findex final_sequence
7389 Note that output routines for instructions with delay slots must be
7390 prepared to deal with not being output as part of a sequence
7391 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
7392 found.) The variable @code{final_sequence} is null when not
7393 processing a sequence, otherwise it contains the @code{sequence} rtx
7396 @findex REGISTER_PREFIX
7397 @findex LOCAL_LABEL_PREFIX
7398 @findex USER_LABEL_PREFIX
7399 @findex IMMEDIATE_PREFIX
7401 @item REGISTER_PREFIX
7402 @itemx LOCAL_LABEL_PREFIX
7403 @itemx USER_LABEL_PREFIX
7404 @itemx IMMEDIATE_PREFIX
7405 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
7406 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
7407 @file{final.c}). These are useful when a single @file{md} file must
7408 support multiple assembler formats. In that case, the various @file{tm.h}
7409 files can define these macros differently.
7411 @item ASM_FPRINTF_EXTENSIONS(@var{file}, @var{argptr}, @var{format})
7412 @findex ASM_FPRINTF_EXTENSIONS
7413 If defined this macro should expand to a series of @code{case}
7414 statements which will be parsed inside the @code{switch} statement of
7415 the @code{asm_fprintf} function. This allows targets to define extra
7416 printf formats which may useful when generating their assembler
7417 statements. Note that upper case letters are reserved for future
7418 generic extensions to asm_fprintf, and so are not available to target
7419 specific code. The output file is given by the parameter @var{file}.
7420 The varargs input pointer is @var{argptr} and the rest of the format
7421 string, starting the character after the one that is being switched
7422 upon, is pointed to by @var{format}.
7424 @findex ASSEMBLER_DIALECT
7425 @item ASSEMBLER_DIALECT
7426 If your target supports multiple dialects of assembler language (such as
7427 different opcodes), define this macro as a C expression that gives the
7428 numeric index of the assembler language dialect to use, with zero as the
7431 If this macro is defined, you may use constructs of the form
7433 @samp{@{option0|option1|option2@dots{}@}}
7436 in the output templates of patterns (@pxref{Output Template}) or in the
7437 first argument of @code{asm_fprintf}. This construct outputs
7438 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
7439 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
7440 within these strings retain their usual meaning. If there are fewer
7441 alternatives within the braces than the value of
7442 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
7444 If you do not define this macro, the characters @samp{@{}, @samp{|} and
7445 @samp{@}} do not have any special meaning when used in templates or
7446 operands to @code{asm_fprintf}.
7448 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
7449 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
7450 the variations in assembler language syntax with that mechanism. Define
7451 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
7452 if the syntax variant are larger and involve such things as different
7453 opcodes or operand order.
7455 @findex ASM_OUTPUT_REG_PUSH
7456 @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
7457 A C expression to output to @var{stream} some assembler code
7458 which will push hard register number @var{regno} onto the stack.
7459 The code need not be optimal, since this macro is used only when
7462 @findex ASM_OUTPUT_REG_POP
7463 @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
7464 A C expression to output to @var{stream} some assembler code
7465 which will pop hard register number @var{regno} off of the stack.
7466 The code need not be optimal, since this macro is used only when
7470 @node Dispatch Tables
7471 @subsection Output of Dispatch Tables
7473 @c prevent bad page break with this line
7474 This concerns dispatch tables.
7477 @cindex dispatch table
7478 @findex ASM_OUTPUT_ADDR_DIFF_ELT
7479 @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
7480 A C statement to output to the stdio stream @var{stream} an assembler
7481 pseudo-instruction to generate a difference between two labels.
7482 @var{value} and @var{rel} are the numbers of two internal labels. The
7483 definitions of these labels are output using
7484 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
7485 way here. For example,
7488 fprintf (@var{stream}, "\t.word L%d-L%d\n",
7489 @var{value}, @var{rel})
7492 You must provide this macro on machines where the addresses in a
7493 dispatch table are relative to the table's own address. If defined, GCC
7494 will also use this macro on all machines when producing PIC@.
7495 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
7496 mode and flags can be read.
7498 @findex ASM_OUTPUT_ADDR_VEC_ELT
7499 @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
7500 This macro should be provided on machines where the addresses
7501 in a dispatch table are absolute.
7503 The definition should be a C statement to output to the stdio stream
7504 @var{stream} an assembler pseudo-instruction to generate a reference to
7505 a label. @var{value} is the number of an internal label whose
7506 definition is output using @code{(*targetm.asm_out.internal_label)}.
7510 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
7513 @findex ASM_OUTPUT_CASE_LABEL
7514 @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
7515 Define this if the label before a jump-table needs to be output
7516 specially. The first three arguments are the same as for
7517 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
7518 jump-table which follows (a @code{jump_insn} containing an
7519 @code{addr_vec} or @code{addr_diff_vec}).
7521 This feature is used on system V to output a @code{swbeg} statement
7524 If this macro is not defined, these labels are output with
7525 @code{(*targetm.asm_out.internal_label)}.
7527 @findex ASM_OUTPUT_CASE_END
7528 @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
7529 Define this if something special must be output at the end of a
7530 jump-table. The definition should be a C statement to be executed
7531 after the assembler code for the table is written. It should write
7532 the appropriate code to stdio stream @var{stream}. The argument
7533 @var{table} is the jump-table insn, and @var{num} is the label-number
7534 of the preceding label.
7536 If this macro is not defined, nothing special is output at the end of
7540 @node Exception Region Output
7541 @subsection Assembler Commands for Exception Regions
7543 @c prevent bad page break with this line
7545 This describes commands marking the start and the end of an exception
7549 @findex EH_FRAME_SECTION_NAME
7550 @item EH_FRAME_SECTION_NAME
7551 If defined, a C string constant for the name of the section containing
7552 exception handling frame unwind information. If not defined, GCC will
7553 provide a default definition if the target supports named sections.
7554 @file{crtstuff.c} uses this macro to switch to the appropriate section.
7556 You should define this symbol if your target supports DWARF 2 frame
7557 unwind information and the default definition does not work.
7559 @findex EH_FRAME_IN_DATA_SECTION
7560 @item EH_FRAME_IN_DATA_SECTION
7561 If defined, DWARF 2 frame unwind information will be placed in the
7562 data section even though the target supports named sections. This
7563 might be necessary, for instance, if the system linker does garbage
7564 collection and sections cannot be marked as not to be collected.
7566 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
7569 @findex MASK_RETURN_ADDR
7570 @item MASK_RETURN_ADDR
7571 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
7572 that it does not contain any extraneous set bits in it.
7574 @findex DWARF2_UNWIND_INFO
7575 @item DWARF2_UNWIND_INFO
7576 Define this macro to 0 if your target supports DWARF 2 frame unwind
7577 information, but it does not yet work with exception handling.
7578 Otherwise, if your target supports this information (if it defines
7579 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
7580 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of
7583 If this macro is defined to 1, the DWARF 2 unwinder will be the default
7584 exception handling mechanism; otherwise, @code{setjmp}/@code{longjmp} will be used by
7587 If this macro is defined to anything, the DWARF 2 unwinder will be used
7588 instead of inline unwinders and @code{__unwind_function} in the non-@code{setjmp} case.
7590 @findex DWARF_CIE_DATA_ALIGNMENT
7591 @item DWARF_CIE_DATA_ALIGNMENT
7592 This macro need only be defined if the target might save registers in the
7593 function prologue at an offset to the stack pointer that is not aligned to
7594 @code{UNITS_PER_WORD}. The definition should be the negative minimum
7595 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
7596 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
7597 the target supports DWARF 2 frame unwind information.
7601 @deftypefn {Target Hook} void TARGET_ASM_EXCEPTION_SECTION ()
7602 If defined, a function that switches to the section in which the main
7603 exception table is to be placed (@pxref{Sections}). The default is a
7604 function that switches to a section named @code{.gcc_except_table} on
7605 machines that support named sections via
7606 @code{TARGET_ASM_NAMED_SECTION}, otherwise if @option{-fpic} or
7607 @option{-fPIC} is in effect, the @code{data_section}, otherwise the
7608 @code{readonly_data_section}.
7611 @deftypefn {Target Hook} void TARGET_ASM_EH_FRAME_SECTION ()
7612 If defined, a function that switches to the section in which the DWARF 2
7613 frame unwind information to be placed (@pxref{Sections}). The default
7614 is a function that outputs a standard GAS section directive, if
7615 @code{EH_FRAME_SECTION_NAME} is defined, or else a data section
7616 directive followed by a synthetic label.
7619 @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
7620 Contains the value true if the target should add a zero word onto the
7621 end of a Dwarf-2 frame info section when used for exception handling.
7622 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
7626 @node Alignment Output
7627 @subsection Assembler Commands for Alignment
7629 @c prevent bad page break with this line
7630 This describes commands for alignment.
7634 @item JUMP_ALIGN (@var{label})
7635 The alignment (log base 2) to put in front of @var{label}, which is
7636 a common destination of jumps and has no fallthru incoming edge.
7638 This macro need not be defined if you don't want any special alignment
7639 to be done at such a time. Most machine descriptions do not currently
7642 Unless it's necessary to inspect the @var{label} parameter, it is better
7643 to set the variable @var{align_jumps} in the target's
7644 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7645 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
7647 @findex LABEL_ALIGN_AFTER_BARRIER
7648 @item LABEL_ALIGN_AFTER_BARRIER (@var{label})
7649 The alignment (log base 2) to put in front of @var{label}, which follows
7652 This macro need not be defined if you don't want any special alignment
7653 to be done at such a time. Most machine descriptions do not currently
7656 @findex LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7657 @item LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
7658 The maximum number of bytes to skip when applying
7659 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
7660 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7663 @item LOOP_ALIGN (@var{label})
7664 The alignment (log base 2) to put in front of @var{label}, which follows
7665 a @code{NOTE_INSN_LOOP_BEG} note.
7667 This macro need not be defined if you don't want any special alignment
7668 to be done at such a time. Most machine descriptions do not currently
7671 Unless it's necessary to inspect the @var{label} parameter, it is better
7672 to set the variable @code{align_loops} in the target's
7673 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7674 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
7676 @findex LOOP_ALIGN_MAX_SKIP
7677 @item LOOP_ALIGN_MAX_SKIP
7678 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
7679 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7682 @item LABEL_ALIGN (@var{label})
7683 The alignment (log base 2) to put in front of @var{label}.
7684 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
7685 the maximum of the specified values is used.
7687 Unless it's necessary to inspect the @var{label} parameter, it is better
7688 to set the variable @code{align_labels} in the target's
7689 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
7690 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
7692 @findex LABEL_ALIGN_MAX_SKIP
7693 @item LABEL_ALIGN_MAX_SKIP
7694 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
7695 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
7697 @findex ASM_OUTPUT_SKIP
7698 @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
7699 A C statement to output to the stdio stream @var{stream} an assembler
7700 instruction to advance the location counter by @var{nbytes} bytes.
7701 Those bytes should be zero when loaded. @var{nbytes} will be a C
7702 expression of type @code{int}.
7704 @findex ASM_NO_SKIP_IN_TEXT
7705 @item ASM_NO_SKIP_IN_TEXT
7706 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
7707 text section because it fails to put zeros in the bytes that are skipped.
7708 This is true on many Unix systems, where the pseudo--op to skip bytes
7709 produces no-op instructions rather than zeros when used in the text
7712 @findex ASM_OUTPUT_ALIGN
7713 @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
7714 A C statement to output to the stdio stream @var{stream} an assembler
7715 command to advance the location counter to a multiple of 2 to the
7716 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
7718 @findex ASM_OUTPUT_ALIGN_WITH_NOP
7719 @item ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
7720 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
7721 for padding, if necessary.
7723 @findex ASM_OUTPUT_MAX_SKIP_ALIGN
7724 @item ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
7725 A C statement to output to the stdio stream @var{stream} an assembler
7726 command to advance the location counter to a multiple of 2 to the
7727 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
7728 satisfy the alignment request. @var{power} and @var{max_skip} will be
7729 a C expression of type @code{int}.
7733 @node Debugging Info
7734 @section Controlling Debugging Information Format
7736 @c prevent bad page break with this line
7737 This describes how to specify debugging information.
7740 * All Debuggers:: Macros that affect all debugging formats uniformly.
7741 * DBX Options:: Macros enabling specific options in DBX format.
7742 * DBX Hooks:: Hook macros for varying DBX format.
7743 * File Names and DBX:: Macros controlling output of file names in DBX format.
7744 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
7745 * VMS Debug:: Macros for VMS debug format.
7749 @subsection Macros Affecting All Debugging Formats
7751 @c prevent bad page break with this line
7752 These macros affect all debugging formats.
7755 @findex DBX_REGISTER_NUMBER
7756 @item DBX_REGISTER_NUMBER (@var{regno})
7757 A C expression that returns the DBX register number for the compiler
7758 register number @var{regno}. In the default macro provided, the value
7759 of this expression will be @var{regno} itself. But sometimes there are
7760 some registers that the compiler knows about and DBX does not, or vice
7761 versa. In such cases, some register may need to have one number in the
7762 compiler and another for DBX@.
7764 If two registers have consecutive numbers inside GCC, and they can be
7765 used as a pair to hold a multiword value, then they @emph{must} have
7766 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
7767 Otherwise, debuggers will be unable to access such a pair, because they
7768 expect register pairs to be consecutive in their own numbering scheme.
7770 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
7771 does not preserve register pairs, then what you must do instead is
7772 redefine the actual register numbering scheme.
7774 @findex DEBUGGER_AUTO_OFFSET
7775 @item DEBUGGER_AUTO_OFFSET (@var{x})
7776 A C expression that returns the integer offset value for an automatic
7777 variable having address @var{x} (an RTL expression). The default
7778 computation assumes that @var{x} is based on the frame-pointer and
7779 gives the offset from the frame-pointer. This is required for targets
7780 that produce debugging output for DBX or COFF-style debugging output
7781 for SDB and allow the frame-pointer to be eliminated when the
7782 @option{-g} options is used.
7784 @findex DEBUGGER_ARG_OFFSET
7785 @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
7786 A C expression that returns the integer offset value for an argument
7787 having address @var{x} (an RTL expression). The nominal offset is
7790 @findex PREFERRED_DEBUGGING_TYPE
7791 @item PREFERRED_DEBUGGING_TYPE
7792 A C expression that returns the type of debugging output GCC should
7793 produce when the user specifies just @option{-g}. Define
7794 this if you have arranged for GCC to support more than one format of
7795 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
7796 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
7797 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
7799 When the user specifies @option{-ggdb}, GCC normally also uses the
7800 value of this macro to select the debugging output format, but with two
7801 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined and
7802 @code{LINKER_DOES_NOT_WORK_WITH_DWARF2} is not defined, GCC uses the
7803 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
7804 defined, GCC uses @code{DBX_DEBUG}.
7806 The value of this macro only affects the default debugging output; the
7807 user can always get a specific type of output by using @option{-gstabs},
7808 @option{-gcoff}, @option{-gdwarf-1}, @option{-gdwarf-2}, @option{-gxcoff},
7813 @subsection Specific Options for DBX Output
7815 @c prevent bad page break with this line
7816 These are specific options for DBX output.
7819 @findex DBX_DEBUGGING_INFO
7820 @item DBX_DEBUGGING_INFO
7821 Define this macro if GCC should produce debugging output for DBX
7822 in response to the @option{-g} option.
7824 @findex XCOFF_DEBUGGING_INFO
7825 @item XCOFF_DEBUGGING_INFO
7826 Define this macro if GCC should produce XCOFF format debugging output
7827 in response to the @option{-g} option. This is a variant of DBX format.
7829 @findex DEFAULT_GDB_EXTENSIONS
7830 @item DEFAULT_GDB_EXTENSIONS
7831 Define this macro to control whether GCC should by default generate
7832 GDB's extended version of DBX debugging information (assuming DBX-format
7833 debugging information is enabled at all). If you don't define the
7834 macro, the default is 1: always generate the extended information
7835 if there is any occasion to.
7837 @findex DEBUG_SYMS_TEXT
7838 @item DEBUG_SYMS_TEXT
7839 Define this macro if all @code{.stabs} commands should be output while
7840 in the text section.
7842 @findex ASM_STABS_OP
7844 A C string constant, including spacing, naming the assembler pseudo op to
7845 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
7846 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
7847 applies only to DBX debugging information format.
7849 @findex ASM_STABD_OP
7851 A C string constant, including spacing, naming the assembler pseudo op to
7852 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
7853 value is the current location. If you don't define this macro,
7854 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
7857 @findex ASM_STABN_OP
7859 A C string constant, including spacing, naming the assembler pseudo op to
7860 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
7861 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
7862 macro applies only to DBX debugging information format.
7864 @findex DBX_NO_XREFS
7866 Define this macro if DBX on your system does not support the construct
7867 @samp{xs@var{tagname}}. On some systems, this construct is used to
7868 describe a forward reference to a structure named @var{tagname}.
7869 On other systems, this construct is not supported at all.
7871 @findex DBX_CONTIN_LENGTH
7872 @item DBX_CONTIN_LENGTH
7873 A symbol name in DBX-format debugging information is normally
7874 continued (split into two separate @code{.stabs} directives) when it
7875 exceeds a certain length (by default, 80 characters). On some
7876 operating systems, DBX requires this splitting; on others, splitting
7877 must not be done. You can inhibit splitting by defining this macro
7878 with the value zero. You can override the default splitting-length by
7879 defining this macro as an expression for the length you desire.
7881 @findex DBX_CONTIN_CHAR
7882 @item DBX_CONTIN_CHAR
7883 Normally continuation is indicated by adding a @samp{\} character to
7884 the end of a @code{.stabs} string when a continuation follows. To use
7885 a different character instead, define this macro as a character
7886 constant for the character you want to use. Do not define this macro
7887 if backslash is correct for your system.
7889 @findex DBX_STATIC_STAB_DATA_SECTION
7890 @item DBX_STATIC_STAB_DATA_SECTION
7891 Define this macro if it is necessary to go to the data section before
7892 outputting the @samp{.stabs} pseudo-op for a non-global static
7895 @findex DBX_TYPE_DECL_STABS_CODE
7896 @item DBX_TYPE_DECL_STABS_CODE
7897 The value to use in the ``code'' field of the @code{.stabs} directive
7898 for a typedef. The default is @code{N_LSYM}.
7900 @findex DBX_STATIC_CONST_VAR_CODE
7901 @item DBX_STATIC_CONST_VAR_CODE
7902 The value to use in the ``code'' field of the @code{.stabs} directive
7903 for a static variable located in the text section. DBX format does not
7904 provide any ``right'' way to do this. The default is @code{N_FUN}.
7906 @findex DBX_REGPARM_STABS_CODE
7907 @item DBX_REGPARM_STABS_CODE
7908 The value to use in the ``code'' field of the @code{.stabs} directive
7909 for a parameter passed in registers. DBX format does not provide any
7910 ``right'' way to do this. The default is @code{N_RSYM}.
7912 @findex DBX_REGPARM_STABS_LETTER
7913 @item DBX_REGPARM_STABS_LETTER
7914 The letter to use in DBX symbol data to identify a symbol as a parameter
7915 passed in registers. DBX format does not customarily provide any way to
7916 do this. The default is @code{'P'}.
7918 @findex DBX_MEMPARM_STABS_LETTER
7919 @item DBX_MEMPARM_STABS_LETTER
7920 The letter to use in DBX symbol data to identify a symbol as a stack
7921 parameter. The default is @code{'p'}.
7923 @findex DBX_FUNCTION_FIRST
7924 @item DBX_FUNCTION_FIRST
7925 Define this macro if the DBX information for a function and its
7926 arguments should precede the assembler code for the function. Normally,
7927 in DBX format, the debugging information entirely follows the assembler
7930 @findex DBX_LBRAC_FIRST
7931 @item DBX_LBRAC_FIRST
7932 Define this macro if the @code{N_LBRAC} symbol for a block should
7933 precede the debugging information for variables and functions defined in
7934 that block. Normally, in DBX format, the @code{N_LBRAC} symbol comes
7937 @findex DBX_BLOCKS_FUNCTION_RELATIVE
7938 @item DBX_BLOCKS_FUNCTION_RELATIVE
7939 Define this macro if the value of a symbol describing the scope of a
7940 block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start
7941 of the enclosing function. Normally, GCC uses an absolute address.
7943 @findex DBX_USE_BINCL
7945 Define this macro if GCC should generate @code{N_BINCL} and
7946 @code{N_EINCL} stabs for included header files, as on Sun systems. This
7947 macro also directs GCC to output a type number as a pair of a file
7948 number and a type number within the file. Normally, GCC does not
7949 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
7950 number for a type number.
7954 @subsection Open-Ended Hooks for DBX Format
7956 @c prevent bad page break with this line
7957 These are hooks for DBX format.
7960 @findex DBX_OUTPUT_LBRAC
7961 @item DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
7962 Define this macro to say how to output to @var{stream} the debugging
7963 information for the start of a scope level for variable names. The
7964 argument @var{name} is the name of an assembler symbol (for use with
7965 @code{assemble_name}) whose value is the address where the scope begins.
7967 @findex DBX_OUTPUT_RBRAC
7968 @item DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
7969 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
7971 @findex DBX_OUTPUT_NFUN
7972 @item DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
7973 Define this macro if the target machine requires special handling to
7974 output an @code{N_FUN} entry for the function @var{decl}.
7976 @findex DBX_OUTPUT_ENUM
7977 @item DBX_OUTPUT_ENUM (@var{stream}, @var{type})
7978 Define this macro if the target machine requires special handling to
7979 output an enumeration type. The definition should be a C statement
7980 (sans semicolon) to output the appropriate information to @var{stream}
7981 for the type @var{type}.
7983 @findex DBX_OUTPUT_FUNCTION_END
7984 @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
7985 Define this macro if the target machine requires special output at the
7986 end of the debugging information for a function. The definition should
7987 be a C statement (sans semicolon) to output the appropriate information
7988 to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
7991 @findex DBX_OUTPUT_STANDARD_TYPES
7992 @item DBX_OUTPUT_STANDARD_TYPES (@var{syms})
7993 Define this macro if you need to control the order of output of the
7994 standard data types at the beginning of compilation. The argument
7995 @var{syms} is a @code{tree} which is a chain of all the predefined
7996 global symbols, including names of data types.
7998 Normally, DBX output starts with definitions of the types for integers
7999 and characters, followed by all the other predefined types of the
8000 particular language in no particular order.
8002 On some machines, it is necessary to output different particular types
8003 first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
8004 those symbols in the necessary order. Any predefined types that you
8005 don't explicitly output will be output afterward in no particular order.
8007 Be careful not to define this macro so that it works only for C@. There
8008 are no global variables to access most of the built-in types, because
8009 another language may have another set of types. The way to output a
8010 particular type is to look through @var{syms} to see if you can find it.
8016 for (decl = syms; decl; decl = TREE_CHAIN (decl))
8017 if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
8019 dbxout_symbol (decl);
8025 This does nothing if the expected type does not exist.
8027 See the function @code{init_decl_processing} in @file{c-decl.c} to find
8028 the names to use for all the built-in C types.
8030 Here is another way of finding a particular type:
8032 @c this is still overfull. --mew 10feb93
8036 for (decl = syms; decl; decl = TREE_CHAIN (decl))
8037 if (TREE_CODE (decl) == TYPE_DECL
8038 && (TREE_CODE (TREE_TYPE (decl))
8040 && TYPE_PRECISION (TREE_TYPE (decl)) == 16
8041 && TYPE_UNSIGNED (TREE_TYPE (decl)))
8043 /* @r{This must be @code{unsigned short}.} */
8044 dbxout_symbol (decl);
8050 @findex NO_DBX_FUNCTION_END
8051 @item NO_DBX_FUNCTION_END
8052 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
8053 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
8054 On those machines, define this macro to turn this feature off without
8055 disturbing the rest of the gdb extensions.
8059 @node File Names and DBX
8060 @subsection File Names in DBX Format
8062 @c prevent bad page break with this line
8063 This describes file names in DBX format.
8066 @findex DBX_WORKING_DIRECTORY
8067 @item DBX_WORKING_DIRECTORY
8068 Define this if DBX wants to have the current directory recorded in each
8071 Note that the working directory is always recorded if GDB extensions are
8074 @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME
8075 @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
8076 A C statement to output DBX debugging information to the stdio stream
8077 @var{stream} which indicates that file @var{name} is the main source
8078 file---the file specified as the input file for compilation.
8079 This macro is called only once, at the beginning of compilation.
8081 This macro need not be defined if the standard form of output
8082 for DBX debugging information is appropriate.
8084 @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY
8085 @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
8086 A C statement to output DBX debugging information to the stdio stream
8087 @var{stream} which indicates that the current directory during
8088 compilation is named @var{name}.
8090 This macro need not be defined if the standard form of output
8091 for DBX debugging information is appropriate.
8093 @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END
8094 @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
8095 A C statement to output DBX debugging information at the end of
8096 compilation of the main source file @var{name}.
8098 If you don't define this macro, nothing special is output at the end
8099 of compilation, which is correct for most machines.
8101 @findex DBX_OUTPUT_SOURCE_FILENAME
8102 @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
8103 A C statement to output DBX debugging information to the stdio stream
8104 @var{stream} which indicates that file @var{name} is the current source
8105 file. This output is generated each time input shifts to a different
8106 source file as a result of @samp{#include}, the end of an included file,
8107 or a @samp{#line} command.
8109 This macro need not be defined if the standard form of output
8110 for DBX debugging information is appropriate.
8115 @subsection Macros for SDB and DWARF Output
8117 @c prevent bad page break with this line
8118 Here are macros for SDB and DWARF output.
8121 @findex SDB_DEBUGGING_INFO
8122 @item SDB_DEBUGGING_INFO
8123 Define this macro if GCC should produce COFF-style debugging output
8124 for SDB in response to the @option{-g} option.
8126 @findex DWARF_DEBUGGING_INFO
8127 @item DWARF_DEBUGGING_INFO
8128 Define this macro if GCC should produce dwarf format debugging output
8129 in response to the @option{-g} option.
8131 @findex DWARF2_DEBUGGING_INFO
8132 @item DWARF2_DEBUGGING_INFO
8133 Define this macro if GCC should produce dwarf version 2 format
8134 debugging output in response to the @option{-g} option.
8136 To support optional call frame debugging information, you must also
8137 define @code{INCOMING_RETURN_ADDR_RTX} and either set
8138 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
8139 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
8140 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
8142 @findex DWARF2_FRAME_INFO
8143 @item DWARF2_FRAME_INFO
8144 Define this macro to a nonzero value if GCC should always output
8145 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
8146 (@pxref{Exception Region Output} is nonzero, GCC will output this
8147 information not matter how you define @code{DWARF2_FRAME_INFO}.
8149 @findex LINKER_DOES_NOT_WORK_WITH_DWARF2
8150 @item LINKER_DOES_NOT_WORK_WITH_DWARF2
8151 Define this macro if the linker does not work with Dwarf version 2.
8152 Normally, if the user specifies only @option{-ggdb} GCC will use Dwarf
8153 version 2 if available; this macro disables this. See the description
8154 of the @code{PREFERRED_DEBUGGING_TYPE} macro for more details.
8156 @findex DWARF2_GENERATE_TEXT_SECTION_LABEL
8157 @item DWARF2_GENERATE_TEXT_SECTION_LABEL
8158 By default, the Dwarf 2 debugging information generator will generate a
8159 label to mark the beginning of the text section. If it is better simply
8160 to use the name of the text section itself, rather than an explicit label,
8161 to indicate the beginning of the text section, define this macro to zero.
8163 @findex DWARF2_ASM_LINE_DEBUG_INFO
8164 @item DWARF2_ASM_LINE_DEBUG_INFO
8165 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
8166 line debug info sections. This will result in much more compact line number
8167 tables, and hence is desirable if it works.
8169 @findex ASM_OUTPUT_DWARF_DELTA
8170 @item ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
8171 A C statement to issue assembly directives that create a difference
8172 between the two given labels, using an integer of the given size.
8174 @findex ASM_OUTPUT_DWARF_OFFSET
8175 @item ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label})
8176 A C statement to issue assembly directives that create a
8177 section-relative reference to the given label, using an integer of the
8180 @findex ASM_OUTPUT_DWARF_PCREL
8181 @item ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
8182 A C statement to issue assembly directives that create a self-relative
8183 reference to the given label, using an integer of the given size.
8185 @findex PUT_SDB_@dots{}
8186 @item PUT_SDB_@dots{}
8187 Define these macros to override the assembler syntax for the special
8188 SDB assembler directives. See @file{sdbout.c} for a list of these
8189 macros and their arguments. If the standard syntax is used, you need
8190 not define them yourself.
8194 Some assemblers do not support a semicolon as a delimiter, even between
8195 SDB assembler directives. In that case, define this macro to be the
8196 delimiter to use (usually @samp{\n}). It is not necessary to define
8197 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
8200 @findex SDB_GENERATE_FAKE
8201 @item SDB_GENERATE_FAKE
8202 Define this macro to override the usual method of constructing a dummy
8203 name for anonymous structure and union types. See @file{sdbout.c} for
8206 @findex SDB_ALLOW_UNKNOWN_REFERENCES
8207 @item SDB_ALLOW_UNKNOWN_REFERENCES
8208 Define this macro to allow references to unknown structure,
8209 union, or enumeration tags to be emitted. Standard COFF does not
8210 allow handling of unknown references, MIPS ECOFF has support for
8213 @findex SDB_ALLOW_FORWARD_REFERENCES
8214 @item SDB_ALLOW_FORWARD_REFERENCES
8215 Define this macro to allow references to structure, union, or
8216 enumeration tags that have not yet been seen to be handled. Some
8217 assemblers choke if forward tags are used, while some require it.
8222 @subsection Macros for VMS Debug Format
8224 @c prevent bad page break with this line
8225 Here are macros for VMS debug format.
8228 @findex VMS_DEBUGGING_INFO
8229 @item VMS_DEBUGGING_INFO
8230 Define this macro if GCC should produce debugging output for VMS
8231 in response to the @option{-g} option. The default behavior for VMS
8232 is to generate minimal debug info for a traceback in the absence of
8233 @option{-g} unless explicitly overridden with @option{-g0}. This
8234 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
8235 @code{OVERRIDE_OPTIONS}.
8238 @node Floating Point
8239 @section Cross Compilation and Floating Point
8240 @cindex cross compilation and floating point
8241 @cindex floating point and cross compilation
8243 While all modern machines use twos-complement representation for integers,
8244 there are a variety of representations for floating point numbers. This
8245 means that in a cross-compiler the representation of floating point numbers
8246 in the compiled program may be different from that used in the machine
8247 doing the compilation.
8249 Because different representation systems may offer different amounts of
8250 range and precision, all floating point constants must be represented in
8251 the target machine's format. Therefore, the cross compiler cannot
8252 safely use the host machine's floating point arithmetic; it must emulate
8253 the target's arithmetic. To ensure consistency, GCC always uses
8254 emulation to work with floating point values, even when the host and
8255 target floating point formats are identical.
8257 The following macros are provided by @file{real.h} for the compiler to
8258 use. All parts of the compiler which generate or optimize
8259 floating-point calculations must use these macros. They may evaluate
8260 their operands more than once, so operands must not have side effects.
8262 @defmac REAL_VALUE_TYPE
8263 The C data type to be used to hold a floating point value in the target
8264 machine's format. Typically this is a @code{struct} containing an
8265 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
8269 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8270 Compares for equality the two values, @var{x} and @var{y}. If the target
8271 floating point format supports negative zeroes and/or NaNs,
8272 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
8273 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
8276 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
8277 Tests whether @var{x} is less than @var{y}.
8280 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
8281 Truncates @var{x} to a signed integer, rounding toward zero.
8284 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
8285 Truncates @var{x} to an unsigned integer, rounding toward zero. If
8286 @var{x} is negative, returns zero.
8289 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
8290 Converts @var{string} into a floating point number in the target machine's
8291 representation for mode @var{mode}. This routine can handle both
8292 decimal and hexadecimal floating point constants, using the syntax
8293 defined by the C language for both.
8296 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
8297 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
8300 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
8301 Determines whether @var{x} represents infinity (positive or negative).
8304 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
8305 Determines whether @var{x} represents a ``NaN'' (not-a-number).
8308 @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})
8309 Calculates an arithmetic operation on the two floating point values
8310 @var{x} and @var{y}, storing the result in @var{output} (which must be a
8313 The operation to be performed is specified by @var{code}. Only the
8314 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
8315 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
8317 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
8318 target's floating point format cannot represent infinity, it will call
8319 @code{abort}. Callers should check for this situation first, using
8320 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
8323 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
8324 Returns the negative of the floating point value @var{x}.
8327 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
8328 Returns the absolute value of @var{x}.
8331 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
8332 Truncates the floating point value @var{x} to fit in @var{mode}. The
8333 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
8334 appropriate bit pattern to be output asa floating constant whose
8335 precision accords with mode @var{mode}.
8338 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
8339 Converts a floating point value @var{x} into a double-precision integer
8340 which is then stored into @var{low} and @var{high}. If the value is not
8341 integral, it is truncated.
8344 @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})
8345 @findex REAL_VALUE_FROM_INT
8346 Converts a double-precision integer found in @var{low} and @var{high},
8347 into a floating point value which is then stored into @var{x}. The
8348 value is truncated to fit in mode @var{mode}.
8351 @node Mode Switching
8352 @section Mode Switching Instructions
8353 @cindex mode switching
8354 The following macros control mode switching optimizations:
8357 @findex OPTIMIZE_MODE_SWITCHING
8358 @item OPTIMIZE_MODE_SWITCHING (@var{entity})
8359 Define this macro if the port needs extra instructions inserted for mode
8360 switching in an optimizing compilation.
8362 For an example, the SH4 can perform both single and double precision
8363 floating point operations, but to perform a single precision operation,
8364 the FPSCR PR bit has to be cleared, while for a double precision
8365 operation, this bit has to be set. Changing the PR bit requires a general
8366 purpose register as a scratch register, hence these FPSCR sets have to
8367 be inserted before reload, i.e.@: you can't put this into instruction emitting
8368 or @code{MACHINE_DEPENDENT_REORG}.
8370 You can have multiple entities that are mode-switched, and select at run time
8371 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
8372 return nonzero for any @var{entity} that needs mode-switching.
8373 If you define this macro, you also have to define
8374 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
8375 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
8376 @code{NORMAL_MODE} is optional.
8378 @findex NUM_MODES_FOR_MODE_SWITCHING
8379 @item NUM_MODES_FOR_MODE_SWITCHING
8380 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
8381 initializer for an array of integers. Each initializer element
8382 N refers to an entity that needs mode switching, and specifies the number
8383 of different modes that might need to be set for this entity.
8384 The position of the initializer in the initializer - starting counting at
8385 zero - determines the integer that is used to refer to the mode-switched
8387 In macros that take mode arguments / yield a mode result, modes are
8388 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
8389 switch is needed / supplied.
8392 @item MODE_NEEDED (@var{entity}, @var{insn})
8393 @var{entity} is an integer specifying a mode-switched entity. If
8394 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
8395 return an integer value not larger than the corresponding element in
8396 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
8397 be switched into prior to the execution of @var{insn}.
8400 @item NORMAL_MODE (@var{entity})
8401 If this macro is defined, it is evaluated for every @var{entity} that needs
8402 mode switching. It should evaluate to an integer, which is a mode that
8403 @var{entity} is assumed to be switched to at function entry and exit.
8405 @findex MODE_PRIORITY_TO_MODE
8406 @item MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
8407 This macro specifies the order in which modes for @var{entity} are processed.
8408 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
8409 lowest. The value of the macro should be an integer designating a mode
8410 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
8411 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
8412 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
8414 @findex EMIT_MODE_SET
8415 @item EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
8416 Generate one or more insns to set @var{entity} to @var{mode}.
8417 @var{hard_reg_live} is the set of hard registers live at the point where
8418 the insn(s) are to be inserted.
8421 @node Target Attributes
8422 @section Defining target-specific uses of @code{__attribute__}
8423 @cindex target attributes
8424 @cindex machine attributes
8425 @cindex attributes, target-specific
8427 Target-specific attributes may be defined for functions, data and types.
8428 These are described using the following target hooks; they also need to
8429 be documented in @file{extend.texi}.
8431 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
8432 If defined, this target hook points to an array of @samp{struct
8433 attribute_spec} (defined in @file{tree.h}) specifying the machine
8434 specific attributes for this target and some of the restrictions on the
8435 entities to which these attributes are applied and the arguments they
8439 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8440 If defined, this target hook is a function which returns zero if the attributes on
8441 @var{type1} and @var{type2} are incompatible, one if they are compatible,
8442 and two if they are nearly compatible (which causes a warning to be
8443 generated). If this is not defined, machine-specific attributes are
8444 supposed always to be compatible.
8447 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
8448 If defined, this target hook is a function which assigns default attributes to
8449 newly defined @var{type}.
8452 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
8453 Define this target hook if the merging of type attributes needs special
8454 handling. If defined, the result is a list of the combined
8455 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
8456 that @code{comptypes} has already been called and returned 1. This
8457 function may call @code{merge_attributes} to handle machine-independent
8461 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
8462 Define this target hook if the merging of decl attributes needs special
8463 handling. If defined, the result is a list of the combined
8464 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
8465 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
8466 when this is needed are when one attribute overrides another, or when an
8467 attribute is nullified by a subsequent definition. This function may
8468 call @code{merge_attributes} to handle machine-independent merging.
8470 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
8471 If the only target-specific handling you require is @samp{dllimport} for
8472 Windows targets, you should define the macro
8473 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. This links in a function
8474 called @code{merge_dllimport_decl_attributes} which can then be defined
8475 as the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. This is done
8476 in @file{i386/cygwin.h} and @file{i386/i386.c}, for example.
8479 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
8480 Define this target hook if you want to be able to add attributes to a decl
8481 when it is being created. This is normally useful for back ends which
8482 wish to implement a pragma by using the attributes which correspond to
8483 the pragma's effect. The @var{node} argument is the decl which is being
8484 created. The @var{attr_ptr} argument is a pointer to the attribute list
8485 for this decl. The list itself should not be modified, since it may be
8486 shared with other decls, but attributes may be chained on the head of
8487 the list and @code{*@var{attr_ptr}} modified to point to the new
8488 attributes, or a copy of the list may be made if further changes are
8492 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
8494 This target hook returns @code{true} if it is ok to inline @var{fndecl}
8495 into the current function, despite its having target-specific
8496 attributes, @code{false} otherwise. By default, if a function has a
8497 target specific attribute attached to it, it will not be inlined.
8500 @node MIPS Coprocessors
8501 @section Defining coprocessor specifics for MIPS targets.
8502 @cindex MIPS coprocessor-definition macros
8504 The MIPS specification allows MIPS implementations to have as many as 4
8505 coprocessors, each with as many as 32 private registers. gcc supports
8506 accessing these registers and transferring values between the registers
8507 and memory using asm-ized variables. For example:
8510 register unsigned int cp0count asm ("c0r1");
8516 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
8517 names may be added as described below, or the default names may be
8518 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
8520 Coprocessor registers are assumed to be epilogue-used; sets to them will
8521 be preserved even if it does not appear that the register is used again
8522 later in the function.
8524 Another note: according to the MIPS spec, coprocessor 1 (if present) is
8525 the FPU. One accesses COP1 registers through standard mips
8526 floating-point support; they are not included in this mechanism.
8528 There is one macro used in defining the MIPS coprocessor interface which
8529 you may want to override in subtargets; it is described below.
8533 @item ALL_COP_ADDITIONAL_REGISTER_NAMES
8534 @findex ALL_COP_ADDITIONAL_REGISTER_NAMES
8535 A comma-separated list (with leading comma) of pairs describing the
8536 alternate names of coprocessor registers. The format of each entry should be
8538 @{ @var{alternatename}, @var{register_number}@}
8545 @section Miscellaneous Parameters
8546 @cindex parameters, miscellaneous
8548 @c prevent bad page break with this line
8549 Here are several miscellaneous parameters.
8552 @item PREDICATE_CODES
8553 @findex PREDICATE_CODES
8554 Define this if you have defined special-purpose predicates in the file
8555 @file{@var{machine}.c}. This macro is called within an initializer of an
8556 array of structures. The first field in the structure is the name of a
8557 predicate and the second field is an array of rtl codes. For each
8558 predicate, list all rtl codes that can be in expressions matched by the
8559 predicate. The list should have a trailing comma. Here is an example
8560 of two entries in the list for a typical RISC machine:
8563 #define PREDICATE_CODES \
8564 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
8565 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
8568 Defining this macro does not affect the generated code (however,
8569 incorrect definitions that omit an rtl code that may be matched by the
8570 predicate can cause the compiler to malfunction). Instead, it allows
8571 the table built by @file{genrecog} to be more compact and efficient,
8572 thus speeding up the compiler. The most important predicates to include
8573 in the list specified by this macro are those used in the most insn
8576 For each predicate function named in @code{PREDICATE_CODES}, a
8577 declaration will be generated in @file{insn-codes.h}.
8579 @item SPECIAL_MODE_PREDICATES
8580 @findex SPECIAL_MODE_PREDICATES
8581 Define this if you have special predicates that know special things
8582 about modes. Genrecog will warn about certain forms of
8583 @code{match_operand} without a mode; if the operand predicate is
8584 listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be
8587 Here is an example from the IA-32 port (@code{ext_register_operand}
8588 specially checks for @code{HImode} or @code{SImode} in preparation
8589 for a byte extraction from @code{%ah} etc.).
8592 #define SPECIAL_MODE_PREDICATES \
8593 "ext_register_operand",
8596 @findex CASE_VECTOR_MODE
8597 @item CASE_VECTOR_MODE
8598 An alias for a machine mode name. This is the machine mode that
8599 elements of a jump-table should have.
8601 @findex CASE_VECTOR_SHORTEN_MODE
8602 @item CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
8603 Optional: return the preferred mode for an @code{addr_diff_vec}
8604 when the minimum and maximum offset are known. If you define this,
8605 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
8606 To make this work, you also have to define @code{INSN_ALIGN} and
8607 make the alignment for @code{addr_diff_vec} explicit.
8608 The @var{body} argument is provided so that the offset_unsigned and scale
8609 flags can be updated.
8611 @findex CASE_VECTOR_PC_RELATIVE
8612 @item CASE_VECTOR_PC_RELATIVE
8613 Define this macro to be a C expression to indicate when jump-tables
8614 should contain relative addresses. If jump-tables never contain
8615 relative addresses, then you need not define this macro.
8617 @findex CASE_DROPS_THROUGH
8618 @item CASE_DROPS_THROUGH
8619 Define this if control falls through a @code{case} insn when the index
8620 value is out of range. This means the specified default-label is
8621 actually ignored by the @code{case} insn proper.
8623 @findex CASE_VALUES_THRESHOLD
8624 @item CASE_VALUES_THRESHOLD
8625 Define this to be the smallest number of different values for which it
8626 is best to use a jump-table instead of a tree of conditional branches.
8627 The default is four for machines with a @code{casesi} instruction and
8628 five otherwise. This is best for most machines.
8630 @findex WORD_REGISTER_OPERATIONS
8631 @item WORD_REGISTER_OPERATIONS
8632 Define this macro if operations between registers with integral mode
8633 smaller than a word are always performed on the entire register.
8634 Most RISC machines have this property and most CISC machines do not.
8636 @findex LOAD_EXTEND_OP
8637 @item LOAD_EXTEND_OP (@var{mode})
8638 Define this macro to be a C expression indicating when insns that read
8639 memory in @var{mode}, an integral mode narrower than a word, set the
8640 bits outside of @var{mode} to be either the sign-extension or the
8641 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
8642 of @var{mode} for which the
8643 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
8644 @code{NIL} for other modes.
8646 This macro is not called with @var{mode} non-integral or with a width
8647 greater than or equal to @code{BITS_PER_WORD}, so you may return any
8648 value in this case. Do not define this macro if it would always return
8649 @code{NIL}. On machines where this macro is defined, you will normally
8650 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
8652 @findex SHORT_IMMEDIATES_SIGN_EXTEND
8653 @item SHORT_IMMEDIATES_SIGN_EXTEND
8654 Define this macro if loading short immediate values into registers sign
8657 @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC
8658 @item FIXUNS_TRUNC_LIKE_FIX_TRUNC
8659 Define this macro if the same instructions that convert a floating
8660 point number to a signed fixed point number also convert validly to an
8665 The maximum number of bytes that a single instruction can move quickly
8666 between memory and registers or between two memory locations.
8668 @findex MAX_MOVE_MAX
8670 The maximum number of bytes that a single instruction can move quickly
8671 between memory and registers or between two memory locations. If this
8672 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
8673 constant value that is the largest value that @code{MOVE_MAX} can have
8676 @findex SHIFT_COUNT_TRUNCATED
8677 @item SHIFT_COUNT_TRUNCATED
8678 A C expression that is nonzero if on this machine the number of bits
8679 actually used for the count of a shift operation is equal to the number
8680 of bits needed to represent the size of the object being shifted. When
8681 this macro is nonzero, the compiler will assume that it is safe to omit
8682 a sign-extend, zero-extend, and certain bitwise `and' instructions that
8683 truncates the count of a shift operation. On machines that have
8684 instructions that act on bit-fields at variable positions, which may
8685 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
8686 also enables deletion of truncations of the values that serve as
8687 arguments to bit-field instructions.
8689 If both types of instructions truncate the count (for shifts) and
8690 position (for bit-field operations), or if no variable-position bit-field
8691 instructions exist, you should define this macro.
8693 However, on some machines, such as the 80386 and the 680x0, truncation
8694 only applies to shift operations and not the (real or pretended)
8695 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
8696 such machines. Instead, add patterns to the @file{md} file that include
8697 the implied truncation of the shift instructions.
8699 You need not define this macro if it would always have the value of zero.
8701 @findex TRULY_NOOP_TRUNCATION
8702 @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
8703 A C expression which is nonzero if on this machine it is safe to
8704 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
8705 bits (where @var{outprec} is smaller than @var{inprec}) by merely
8706 operating on it as if it had only @var{outprec} bits.
8708 On many machines, this expression can be 1.
8710 @c rearranged this, removed the phrase "it is reported that". this was
8711 @c to fix an overfull hbox. --mew 10feb93
8712 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
8713 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
8714 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
8715 such cases may improve things.
8717 @findex STORE_FLAG_VALUE
8718 @item STORE_FLAG_VALUE
8719 A C expression describing the value returned by a comparison operator
8720 with an integral mode and stored by a store-flag instruction
8721 (@samp{s@var{cond}}) when the condition is true. This description must
8722 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
8723 comparison operators whose results have a @code{MODE_INT} mode.
8725 A value of 1 or @minus{}1 means that the instruction implementing the
8726 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
8727 and 0 when the comparison is false. Otherwise, the value indicates
8728 which bits of the result are guaranteed to be 1 when the comparison is
8729 true. This value is interpreted in the mode of the comparison
8730 operation, which is given by the mode of the first operand in the
8731 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
8732 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
8735 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
8736 generate code that depends only on the specified bits. It can also
8737 replace comparison operators with equivalent operations if they cause
8738 the required bits to be set, even if the remaining bits are undefined.
8739 For example, on a machine whose comparison operators return an
8740 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
8741 @samp{0x80000000}, saying that just the sign bit is relevant, the
8745 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
8752 (ashift:SI @var{x} (const_int @var{n}))
8756 where @var{n} is the appropriate shift count to move the bit being
8757 tested into the sign bit.
8759 There is no way to describe a machine that always sets the low-order bit
8760 for a true value, but does not guarantee the value of any other bits,
8761 but we do not know of any machine that has such an instruction. If you
8762 are trying to port GCC to such a machine, include an instruction to
8763 perform a logical-and of the result with 1 in the pattern for the
8764 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
8766 Often, a machine will have multiple instructions that obtain a value
8767 from a comparison (or the condition codes). Here are rules to guide the
8768 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
8773 Use the shortest sequence that yields a valid definition for
8774 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
8775 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
8776 comparison operators to do so because there may be opportunities to
8777 combine the normalization with other operations.
8780 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
8781 slightly preferred on machines with expensive jumps and 1 preferred on
8785 As a second choice, choose a value of @samp{0x80000001} if instructions
8786 exist that set both the sign and low-order bits but do not define the
8790 Otherwise, use a value of @samp{0x80000000}.
8793 Many machines can produce both the value chosen for
8794 @code{STORE_FLAG_VALUE} and its negation in the same number of
8795 instructions. On those machines, you should also define a pattern for
8796 those cases, e.g., one matching
8799 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
8802 Some machines can also perform @code{and} or @code{plus} operations on
8803 condition code values with less instructions than the corresponding
8804 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
8805 machines, define the appropriate patterns. Use the names @code{incscc}
8806 and @code{decscc}, respectively, for the patterns which perform
8807 @code{plus} or @code{minus} operations on condition code values. See
8808 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
8809 find such instruction sequences on other machines.
8811 You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
8814 @findex FLOAT_STORE_FLAG_VALUE
8815 @item FLOAT_STORE_FLAG_VALUE (@var{mode})
8816 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
8817 returned when comparison operators with floating-point results are true.
8818 Define this macro on machine that have comparison operations that return
8819 floating-point values. If there are no such operations, do not define
8824 An alias for the machine mode for pointers. On most machines, define
8825 this to be the integer mode corresponding to the width of a hardware
8826 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
8827 On some machines you must define this to be one of the partial integer
8828 modes, such as @code{PSImode}.
8830 The width of @code{Pmode} must be at least as large as the value of
8831 @code{POINTER_SIZE}. If it is not equal, you must define the macro
8832 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
8835 @findex FUNCTION_MODE
8837 An alias for the machine mode used for memory references to functions
8838 being called, in @code{call} RTL expressions. On most machines this
8839 should be @code{QImode}.
8841 @findex INTEGRATE_THRESHOLD
8842 @item INTEGRATE_THRESHOLD (@var{decl})
8843 A C expression for the maximum number of instructions above which the
8844 function @var{decl} should not be inlined. @var{decl} is a
8845 @code{FUNCTION_DECL} node.
8847 The default definition of this macro is 64 plus 8 times the number of
8848 arguments that the function accepts. Some people think a larger
8849 threshold should be used on RISC machines.
8851 @findex STDC_0_IN_SYSTEM_HEADERS
8852 @item STDC_0_IN_SYSTEM_HEADERS
8853 In normal operation, the preprocessor expands @code{__STDC__} to the
8854 constant 1, to signify that GCC conforms to ISO Standard C@. On some
8855 hosts, like Solaris, the system compiler uses a different convention,
8856 where @code{__STDC__} is normally 0, but is 1 if the user specifies
8857 strict conformance to the C Standard.
8859 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
8860 convention when processing system header files, but when processing user
8861 files @code{__STDC__} will always expand to 1.
8863 @findex NO_IMPLICIT_EXTERN_C
8864 @item NO_IMPLICIT_EXTERN_C
8865 Define this macro if the system header files support C++ as well as C@.
8866 This macro inhibits the usual method of using system header files in
8867 C++, which is to pretend that the file's contents are enclosed in
8868 @samp{extern "C" @{@dots{}@}}.
8870 @findex HANDLE_PRAGMA
8871 @item HANDLE_PRAGMA (@var{getc}, @var{ungetc}, @var{name})
8872 This macro is no longer supported. You must use
8873 @code{REGISTER_TARGET_PRAGMAS} instead.
8875 @findex REGISTER_TARGET_PRAGMAS
8878 @item REGISTER_TARGET_PRAGMAS ()
8879 Define this macro if you want to implement any target-specific pragmas.
8880 If defined, it is a C expression which makes a series of calls to
8881 @code{c_register_pragma} for each pragma. The macro may also do any
8882 setup required for the pragmas.
8884 The primary reason to define this macro is to provide compatibility with
8885 other compilers for the same target. In general, we discourage
8886 definition of target-specific pragmas for GCC@.
8888 If the pragma can be implemented by attributes then you should consider
8889 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
8891 Preprocessor macros that appear on pragma lines are not expanded. All
8892 @samp{#pragma} directives that do not match any registered pragma are
8893 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
8895 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
8897 Each call to @code{c_register_pragma} establishes one pragma. The
8898 @var{callback} routine will be called when the preprocessor encounters a
8902 #pragma [@var{space}] @var{name} @dots{}
8905 @var{space} is the case-sensitive namespace of the pragma, or
8906 @code{NULL} to put the pragma in the global namespace. The callback
8907 routine receives @var{pfile} as its first argument, which can be passed
8908 on to cpplib's functions if necessary. You can lex tokens after the
8909 @var{name} by calling @code{c_lex}. Tokens that are not read by the
8910 callback will be silently ignored. The end of the line is indicated by
8911 a token of type @code{CPP_EOF}.
8913 For an example use of this routine, see @file{c4x.h} and the callback
8914 routines defined in @file{c4x-c.c}.
8916 Note that the use of @code{c_lex} is specific to the C and C++
8917 compilers. It will not work in the Java or Fortran compilers, or any
8918 other language compilers for that matter. Thus if @code{c_lex} is going
8919 to be called from target-specific code, it must only be done so when
8920 building the C and C++ compilers. This can be done by defining the
8921 variables @code{c_target_objs} and @code{cxx_target_objs} in the
8922 target entry in the @file{config.gcc} file. These variables should name
8923 the target-specific, language-specific object file which contains the
8924 code that uses @code{c_lex}. Note it will also be necessary to add a
8925 rule to the makefile fragment pointed to by @code{tmake_file} that shows
8926 how to build this object file.
8929 @findex HANDLE_SYSV_PRAGMA
8932 @item HANDLE_SYSV_PRAGMA
8933 Define this macro (to a value of 1) if you want the System V style
8934 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
8935 [=<value>]} to be supported by gcc.
8937 The pack pragma specifies the maximum alignment (in bytes) of fields
8938 within a structure, in much the same way as the @samp{__aligned__} and
8939 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
8940 the behavior to the default.
8942 A subtlety for Microsoft Visual C/C++ style bit-field packing
8943 (e.g. -mms-bitfields) for targets that support it:
8944 When a bit-field is inserted into a packed record, the whole size
8945 of the underlying type is used by one or more same-size adjacent
8946 bit-fields (that is, if its long:3, 32 bits is used in the record,
8947 and any additional adjacent long bit-fields are packed into the same
8948 chunk of 32 bits. However, if the size changes, a new field of that
8951 If both MS bit-fields and @samp{__attribute__((packed))} are used,
8952 the latter will take precedence. If @samp{__attribute__((packed))} is
8953 used on a single field when MS bit-fields are in use, it will take
8954 precedence for that field, but the alignment of the rest of the structure
8955 may affect its placement.
8957 The weak pragma only works if @code{SUPPORTS_WEAK} and
8958 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
8959 of specifically named weak labels, optionally with a value.
8961 @findex HANDLE_PRAGMA_PACK_PUSH_POP
8964 @item HANDLE_PRAGMA_PACK_PUSH_POP
8965 Define this macro (to a value of 1) if you want to support the Win32
8966 style pragmas @samp{#pragma pack(push,@var{n})} and @samp{#pragma
8967 pack(pop)}. The @samp{pack(push,@var{n})} pragma specifies the maximum alignment
8968 (in bytes) of fields within a structure, in much the same way as the
8969 @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
8970 pack value of zero resets the behavior to the default. Successive
8971 invocations of this pragma cause the previous values to be stacked, so
8972 that invocations of @samp{#pragma pack(pop)} will return to the previous
8975 @findex DOLLARS_IN_IDENTIFIERS
8976 @item DOLLARS_IN_IDENTIFIERS
8977 Define this macro to control use of the character @samp{$} in identifier
8978 names. 0 means @samp{$} is not allowed by default; 1 means it is allowed.
8979 1 is the default; there is no need to define this macro in that case.
8980 This macro controls the compiler proper; it does not affect the preprocessor.
8982 @findex NO_DOLLAR_IN_LABEL
8983 @item NO_DOLLAR_IN_LABEL
8984 Define this macro if the assembler does not accept the character
8985 @samp{$} in label names. By default constructors and destructors in
8986 G++ have @samp{$} in the identifiers. If this macro is defined,
8987 @samp{.} is used instead.
8989 @findex NO_DOT_IN_LABEL
8990 @item NO_DOT_IN_LABEL
8991 Define this macro if the assembler does not accept the character
8992 @samp{.} in label names. By default constructors and destructors in G++
8993 have names that use @samp{.}. If this macro is defined, these names
8994 are rewritten to avoid @samp{.}.
8996 @findex DEFAULT_MAIN_RETURN
8997 @item DEFAULT_MAIN_RETURN
8998 Define this macro if the target system expects every program's @code{main}
8999 function to return a standard ``success'' value by default (if no other
9000 value is explicitly returned).
9002 The definition should be a C statement (sans semicolon) to generate the
9003 appropriate rtl instructions. It is used only when compiling the end of
9008 Define this if the target system lacks the function @code{atexit}
9009 from the ISO C standard. If this macro is defined, a default definition
9010 will be provided to support C++. If @code{ON_EXIT} is not defined,
9011 a default @code{exit} function will also be provided.
9015 Define this macro if the target has another way to implement atexit
9016 functionality without replacing @code{exit}. For instance, SunOS 4 has
9017 a similar @code{on_exit} library function.
9019 The definition should be a functional macro which can be used just like
9020 the @code{atexit} function.
9024 Define this if your @code{exit} function needs to do something
9025 besides calling an external function @code{_cleanup} before
9026 terminating with @code{_exit}. The @code{EXIT_BODY} macro is
9027 only needed if @code{NEED_ATEXIT} is defined and @code{ON_EXIT} is not
9030 @findex INSN_SETS_ARE_DELAYED
9031 @item INSN_SETS_ARE_DELAYED (@var{insn})
9032 Define this macro as a C expression that is nonzero if it is safe for the
9033 delay slot scheduler to place instructions in the delay slot of @var{insn},
9034 even if they appear to use a resource set or clobbered in @var{insn}.
9035 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
9036 every @code{call_insn} has this behavior. On machines where some @code{insn}
9037 or @code{jump_insn} is really a function call and hence has this behavior,
9038 you should define this macro.
9040 You need not define this macro if it would always return zero.
9042 @findex INSN_REFERENCES_ARE_DELAYED
9043 @item INSN_REFERENCES_ARE_DELAYED (@var{insn})
9044 Define this macro as a C expression that is nonzero if it is safe for the
9045 delay slot scheduler to place instructions in the delay slot of @var{insn},
9046 even if they appear to set or clobber a resource referenced in @var{insn}.
9047 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
9048 some @code{insn} or @code{jump_insn} is really a function call and its operands
9049 are registers whose use is actually in the subroutine it calls, you should
9050 define this macro. Doing so allows the delay slot scheduler to move
9051 instructions which copy arguments into the argument registers into the delay
9054 You need not define this macro if it would always return zero.
9056 @findex MACHINE_DEPENDENT_REORG
9057 @item MACHINE_DEPENDENT_REORG (@var{insn})
9058 In rare cases, correct code generation requires extra machine
9059 dependent processing between the second jump optimization pass and
9060 delayed branch scheduling. On those machines, define this macro as a C
9061 statement to act on the code starting at @var{insn}.
9063 @findex MULTIPLE_SYMBOL_SPACES
9064 @item MULTIPLE_SYMBOL_SPACES
9065 Define this macro if in some cases global symbols from one translation
9066 unit may not be bound to undefined symbols in another translation unit
9067 without user intervention. For instance, under Microsoft Windows
9068 symbols must be explicitly imported from shared libraries (DLLs).
9070 @findex MD_ASM_CLOBBERS
9071 @item MD_ASM_CLOBBERS (@var{clobbers})
9072 A C statement that adds to @var{clobbers} @code{STRING_CST} trees for
9073 any hard regs the port wishes to automatically clobber for all asms.
9075 @findex MAX_INTEGER_COMPUTATION_MODE
9076 @item MAX_INTEGER_COMPUTATION_MODE
9077 Define this to the largest integer machine mode which can be used for
9078 operations other than load, store and copy operations.
9080 You need only define this macro if the target holds values larger than
9081 @code{word_mode} in general purpose registers. Most targets should not define
9084 @findex MATH_LIBRARY
9086 Define this macro as a C string constant for the linker argument to link
9087 in the system math library, or @samp{""} if the target does not have a
9088 separate math library.
9090 You need only define this macro if the default of @samp{"-lm"} is wrong.
9092 @findex LIBRARY_PATH_ENV
9093 @item LIBRARY_PATH_ENV
9094 Define this macro as a C string constant for the environment variable that
9095 specifies where the linker should look for libraries.
9097 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
9100 @findex TARGET_HAS_F_SETLKW
9101 @item TARGET_HAS_F_SETLKW
9102 Define this macro if the target supports file locking with fcntl / F_SETLKW@.
9103 Note that this functionality is part of POSIX@.
9104 Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code
9105 to use file locking when exiting a program, which avoids race conditions
9106 if the program has forked.
9108 @findex MAX_CONDITIONAL_EXECUTE
9109 @item MAX_CONDITIONAL_EXECUTE
9111 A C expression for the maximum number of instructions to execute via
9112 conditional execution instructions instead of a branch. A value of
9113 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
9114 1 if it does use cc0.
9116 @findex IFCVT_MODIFY_TESTS
9117 @item IFCVT_MODIFY_TESTS(@var{ce_info}, @var{true_expr}, @var{false_expr})
9118 Used if the target needs to perform machine-dependent modifications on the
9119 conditionals used for turning basic blocks into conditionally executed code.
9120 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
9121 contains information about the currently processed blocks. @var{true_expr}
9122 and @var{false_expr} are the tests that are used for converting the
9123 then-block and the else-block, respectively. Set either @var{true_expr} or
9124 @var{false_expr} to a null pointer if the tests cannot be converted.
9126 @findex IFCVT_MODIFY_MULTIPLE_TESTS
9127 @item IFCVT_MODIFY_MULTIPLE_TESTS(@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
9128 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
9129 if-statements into conditions combined by @code{and} and @code{or} operations.
9130 @var{bb} contains the basic block that contains the test that is currently
9131 being processed and about to be turned into a condition.
9133 @findex IFCVT_MODIFY_INSN
9134 @item IFCVT_MODIFY_INSN(@var{ce_info}, @var{pattern}, @var{insn})
9135 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
9136 be converted to conditional execution format. @var{ce_info} points to
9137 a data structure, @code{struct ce_if_block}, which contains information
9138 about the currently processed blocks.
9140 @findex IFCVT_MODIFY_FINAL
9141 @item IFCVT_MODIFY_FINAL(@var{ce_info})
9142 A C expression to perform any final machine dependent modifications in
9143 converting code to conditional execution. The involved basic blocks
9144 can be found in the @code{struct ce_if_block} structure that is pointed
9145 to by @var{ce_info}.
9147 @findex IFCVT_MODIFY_CANCEL
9148 @item IFCVT_MODIFY_CANCEL(@var{ce_info})
9149 A C expression to cancel any machine dependent modifications in
9150 converting code to conditional execution. The involved basic blocks
9151 can be found in the @code{struct ce_if_block} structure that is pointed
9152 to by @var{ce_info}.
9154 @findex IFCVT_INIT_EXTRA_FIELDS
9155 @item IFCVT_INIT_EXTRA_FIELDS(@var{ce_info})
9156 A C expression to initialize any extra fields in a @code{struct ce_if_block}
9157 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
9159 @findex IFCVT_EXTRA_FIELDS
9160 @item IFCVT_EXTRA_FIELDS
9161 If defined, it should expand to a set of field declarations that will be
9162 added to the @code{struct ce_if_block} structure. These should be initialized
9163 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
9167 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
9168 Define this hook if you have any machine-specific built-in functions
9169 that need to be defined. It should be a function that performs the
9172 Machine specific built-in functions can be useful to expand special machine
9173 instructions that would otherwise not normally be generated because
9174 they have no equivalent in the source language (for example, SIMD vector
9175 instructions or prefetch instructions).
9177 To create a built-in function, call the function @code{builtin_function}
9178 which is defined by the language front end. You can use any type nodes set
9179 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
9180 only language front ends that use those two functions will call
9181 @samp{TARGET_INIT_BUILTINS}.
9184 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
9186 Expand a call to a machine specific built-in function that was set up by
9187 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
9188 function call; the result should go to @var{target} if that is
9189 convenient, and have mode @var{mode} if that is convenient.
9190 @var{subtarget} may be used as the target for computing one of
9191 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
9192 ignored. This function should return the result of the call to the
9197 @findex MD_CAN_REDIRECT_BRANCH
9198 @item MD_CAN_REDIRECT_BRANCH(@var{branch1}, @var{branch2})
9200 Take a branch insn in @var{branch1} and another in @var{branch2}.
9201 Return true if redirecting @var{branch1} to the destination of
9202 @var{branch2} is possible.
9204 On some targets, branches may have a limited range. Optimizing the
9205 filling of delay slots can result in branches being redirected, and this
9206 may in turn cause a branch offset to overflow.
9208 @findex ALLOCATE_INITIAL_VALUE
9209 @item ALLOCATE_INITIAL_VALUE(@var{hard_reg})
9211 When the initial value of a hard register has been copied in a pseudo
9212 register, it is often not necessary to actually allocate another register
9213 to this pseudo register, because the original hard register or a stack slot
9214 it has been saved into can be used. @code{ALLOCATE_INITIAL_VALUE}, if
9215 defined, is called at the start of register allocation once for each
9216 hard register that had its initial value copied by using
9217 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
9218 Possible values are @code{NULL_RTX}, if you don't want
9219 to do any special allocation, a @code{REG} rtx---that would typically be
9220 the hard register itself, if it is known not to be clobbered---or a
9222 If you are returning a @code{MEM}, this is only a hint for the allocator;
9223 it might decide to use another register anyways.
9224 You may use @code{current_function_leaf_function} in the definition of the
9225 macro, functions that use @code{REG_N_SETS}, to determine if the hard
9226 register in question will not be clobbered.
9228 @findex TARGET_OBJECT_SUFFIX
9229 @item TARGET_OBJECT_SUFFIX
9230 Define this macro to be a C string representing the suffix for object
9231 files on your target machine. If you do not define this macro, GCC will
9232 use @samp{.o} as the suffix for object files.
9234 @findex TARGET_EXECUTABLE_SUFFIX
9235 @item TARGET_EXECUTABLE_SUFFIX
9236 Define this macro to be a C string representing the suffix to be
9237 automatically added to executable files on your target machine. If you
9238 do not define this macro, GCC will use the null string as the suffix for
9241 @findex COLLECT_EXPORT_LIST
9242 @item COLLECT_EXPORT_LIST
9243 If defined, @code{collect2} will scan the individual object files
9244 specified on its command line and create an export list for the linker.
9245 Define this macro for systems like AIX, where the linker discards
9246 object files that are not referenced from @code{main} and uses export
9251 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
9252 This target hook returns @code{true} past the point in which new jump
9253 instructions could be created. On machines that require a register for
9254 every jump such as the SHmedia ISA of SH5, this point would typically be
9255 reload, so this target hook should be defined to a function such as:
9259 cannot_modify_jumps_past_reload_p ()
9261 return (reload_completed || reload_in_progress);