![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to tm.texi CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [GCC 2.x] / gcc|
Return to tm.texi CVS log | Up to [GCC 2.x] / gcc |
1.1 root 1: @c Copyright (C) 1988, 1989, 1992 Free Software Foundation, Inc.
2: @c This is part of the GCC manual.
3: @c For copying conditions, see the file gcc.texi.
4:
5: @ifset INTERNALS
1.1.1.2 ! root 6: @node Target Macros, Config, Machine Desc, Top
! 7: @chapter Target Description Macros
1.1 root 8: @cindex machine description macros
1.1.1.2 ! root 9: @cindex target description macros
! 10: @cindex macros, target description
1.1 root 11: @cindex @file{tm.h} macros
12:
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}. This header file defines numerous macros
16: that convey the information about the target machine that does not fit
17: into the scheme of the @file{.md} file. The file @file{tm.h} should be
18: a link to @file{@var{machine}.h}. The header file @file{config.h}
19: includes @file{tm.h} and most compiler source files include
20: @file{config.h}.
21:
22: @menu
23: * Driver:: Controlling how the driver runs the compilation passes.
24: * Run-time Target:: Defining @samp{-m} options like @samp{-m68000} and @samp{-m68020}.
25: * Storage Layout:: Defining sizes and alignments of data.
26: * Type Layout:: Defining sizes and properties of basic user data types.
27: * Registers:: Naming and describing the hardware registers.
28: * Register Classes:: Defining the classes of hardware registers.
29: * Stack and Calling:: Defining which way the stack grows and by how much.
30: * Varargs:: Defining the varargs macros.
31: * Trampolines:: Code set up at run time to enter a nested function.
32: * Library Calls:: Controlling how library routines are implicitly called.
33: * Addressing Modes:: Defining addressing modes valid for memory operands.
34: * Condition Code:: Defining how insns update the condition code.
35: * Costs:: Defining relative costs of different operations.
36: * Sections:: Dividing storage into text, data, and other sections.
37: * PIC:: Macros for position independent code.
38: * Assembler Format:: Defining how to write insns and pseudo-ops to output.
39: * Debugging Info:: Defining the format of debugging output.
40: * Cross-compilation:: Handling floating point for cross-compilers.
41: * Misc:: Everything else.
42: @end menu
43:
1.1.1.2 ! root 44: @node Driver
1.1 root 45: @section Controlling the Compilation Driver, @file{gcc}
46: @cindex driver
47: @cindex controlling the compilation driver
48:
49: @table @code
50: @findex SWITCH_TAKES_ARG
51: @item SWITCH_TAKES_ARG (@var{char})
52: A C expression which determines whether the option @samp{-@var{char}}
53: takes arguments. The value should be the number of arguments that
54: option takes--zero, for many options.
55:
56: By default, this macro is defined to handle the standard options
57: properly. You need not define it unless you wish to add additional
58: options which take arguments.
59:
60: @findex WORD_SWITCH_TAKES_ARG
61: @item WORD_SWITCH_TAKES_ARG (@var{name})
62: A C expression which determines whether the option @samp{-@var{name}}
63: takes arguments. The value should be the number of arguments that
64: option takes--zero, for many options. This macro rather than
65: @code{SWITCH_TAKES_ARG} is used for multi-character option names.
66:
67: By default, this macro is defined to handle the standard options
68: properly. You need not define it unless you wish to add additional
69: options which take arguments.
70:
71: @findex SWITCHES_NEED_SPACES
72: @item SWITCHES_NEED_SPACES
73: A string-valued C expression which is nonempty if the linker needs a
74: space between the @samp{-L} or @samp{-o} option and its argument.
75:
76: If this macro is not defined, the default value is 0.
77:
78: @findex CPP_SPEC
79: @item CPP_SPEC
80: A C string constant that tells the GNU CC driver program options to
81: pass to CPP. It can also specify how to translate options you
82: give to GNU CC into options for GNU CC to pass to the CPP.
83:
84: Do not define this macro if it does not need to do anything.
85:
86: @findex SIGNED_CHAR_SPEC
87: @item SIGNED_CHAR_SPEC
88: A C string constant that tells the GNU CC driver program options to
89: pass to CPP. By default, this macro is defined to pass the option
90: @samp{-D__CHAR_UNSIGNED__} to CPP if @code{char} will be treated as
91: @code{unsigned char} by @code{cc1}.
92:
93: Do not define this macro unless you need to override the default
94: definition.
95:
96: @findex CC1_SPEC
97: @item CC1_SPEC
98: A C string constant that tells the GNU CC driver program options to
99: pass to @code{cc1}. It can also specify how to translate options you
100: give to GNU CC into options for GNU CC to pass to the @code{cc1}.
101:
102: Do not define this macro if it does not need to do anything.
103:
104: @findex CC1PLUS_SPEC
105: @item CC1PLUS_SPEC
106: A C string constant that tells the GNU CC driver program options to
107: pass to @code{cc1plus}. It can also specify how to translate options you
108: give to GNU CC into options for GNU CC to pass to the @code{cc1plus}.
109:
110: Do not define this macro if it does not need to do anything.
111:
112: @findex ASM_SPEC
113: @item ASM_SPEC
114: A C string constant that tells the GNU CC driver program options to
115: pass to the assembler. It can also specify how to translate options
116: you give to GNU CC into options for GNU CC to pass to the assembler.
117: See the file @file{sun3.h} for an example of this.
118:
119: Do not define this macro if it does not need to do anything.
120:
121: @findex ASM_FINAL_SPEC
122: @item ASM_FINAL_SPEC
123: A C string constant that tells the GNU CC driver program how to
124: run any programs which cleanup after the normal assembler.
125: Normally, this is not needed. See the file @file{mips.h} for
126: an example of this.
127:
128: Do not define this macro if it does not need to do anything.
129:
130: @findex LINK_SPEC
131: @item LINK_SPEC
132: A C string constant that tells the GNU CC driver program options to
133: pass to the linker. It can also specify how to translate options you
134: give to GNU CC into options for GNU CC to pass to the linker.
135:
136: Do not define this macro if it does not need to do anything.
137:
138: @findex LIB_SPEC
139: @item LIB_SPEC
140: Another C string constant used much like @code{LINK_SPEC}. The difference
141: between the two is that @code{LIB_SPEC} is used at the end of the
142: command given to the linker.
143:
144: If this macro is not defined, a default is provided that
145: loads the standard C library from the usual place. See @file{gcc.c}.
146:
147: @findex STARTFILE_SPEC
148: @item STARTFILE_SPEC
149: Another C string constant used much like @code{LINK_SPEC}. The
150: difference between the two is that @code{STARTFILE_SPEC} is used at
151: the very beginning of the command given to the linker.
152:
153: If this macro is not defined, a default is provided that loads the
154: standard C startup file from the usual place. See @file{gcc.c}.
155:
156: @findex ENDFILE_SPEC
157: @item ENDFILE_SPEC
158: Another C string constant used much like @code{LINK_SPEC}. The
159: difference between the two is that @code{ENDFILE_SPEC} is used at
160: the very end of the command given to the linker.
161:
162: Do not define this macro if it does not need to do anything.
163:
164: @findex LINK_LIBGCC_SPECIAL
165: @item LINK_LIBGCC_SPECIAL
166: Define this macro meaning that @code{gcc} should find the
167: library @file{libgcc.a} by hand, rather than passing the argument
168: @samp{-lgcc} to tell the linker to do the search.
169:
170: @findex RELATIVE_PREFIX_NOT_LINKDIR
171: @item RELATIVE_PREFIX_NOT_LINKDIR
172: Define this macro to tell @code{gcc} that it should only translate
173: a @samp{-B} prefix into a @samp{-L} linker option if the prefix
174: indicates an absolute file name.
175:
176: @findex STANDARD_EXEC_PREFIX
177: @item STANDARD_EXEC_PREFIX
178: Define this macro as a C string constant if you wish to override the
1.1.1.2 ! root 179: standard choice of @file{/usr/local/lib/gcc-lib/} as the default prefix to
1.1 root 180: try when searching for the executable files of the compiler.
181:
182: @findex MD_EXEC_PREFIX
183: @item MD_EXEC_PREFIX
184: If defined, this macro is an additional prefix to try after
185: @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
186: when the @samp{-b} option is used, or the compiler is built as a cross
187: compiler.
188:
189: @findex STANDARD_STARTFILE_PREFIX
190: @item STANDARD_STARTFILE_PREFIX
191: Define this macro as a C string constant if you wish to override the
192: standard choice of @file{/usr/local/lib/gcc/} as the default prefix to
193: try when searching for startup files such as @file{crt0.o}.
194:
195: @findex MD_STARTFILE_PREFIX
196: @item MD_STARTFILE_PREFIX
197: If defined, this macro supplies an additional prefix to try after
198: the standard prefixes. @code{MD_EXEC_PREFIX} is not searched
199: when the @samp{-b} option is used, or the compiler is built as a cross
200: compiler.
201:
202: @findex LOCAL_INCLUDE_DIR
203: @item LOCAL_INCLUDE_DIR
204: Define this macro as a C string constant if you wish to override the
205: standard choice of @file{/usr/local/include} as the default prefix to
206: try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
207: comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
208:
209: Cross compilers do not use this macro and do not search either
210: @file{/usr/local/include} or its replacement.
211:
212: @findex SYSTEM_INCLUDE_DIR
213: @item SYSTEM_INCLUDE_DIR
214: Define this macro as a C string constant if you wish to specify a
215: system-specific directory to search for header files before the standard
216: directory. @code{SYSTEM_INCLUDE_DIR} comes before
217: @code{STANDARD_INCLUDE_DIR} in the search order.
218:
219: Cross compilers do not use this macro and do not search the directory
220: specified.
221:
222: @findex STANDARD_INCLUDE_DIR
223: @item STANDARD_INCLUDE_DIR
224: Define this macro as a C string constant if you wish to override the
225: standard choice of @file{/usr/include} as the default prefix to
226: try when searching for header files.
227:
228: Cross compilers do not use this macro and do not search either
229: @file{/usr/include} or its replacement.
230:
231: @findex INCLUDE_DEFAULTS
232: @item INCLUDE_DEFAULTS
233: Define this macro if you wish to override the entire default search path
234: for include files. The default search path includes
235: @code{GPLUSPLUS_INCLUDE_DIR}, @code{GCC_INCLUDE_DIR},
236: @code{LOCAL_INCLUDE_DIR}, @code{SYSTEM_INCLUDE_DIR}, and
237: @code{STANDARD_INCLUDE_DIR}. In addition, the macros
238: @code{GPLUSPLUS_INCLUDE_DIR} and @code{GCC_INCLUDE_DIR} are defined
239: automatically by @file{Makefile}, and specify private search areas for
240: GCC. The directory @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++
241: programs.
242:
243: The definition should be an initializer for an array of structures.
244: Each array element should have two elements: the directory name (a
245: string constant) and a flag for C++-only directories. Mark the end of
246: the array with a null element. For example, here is the definition used
247: for VMS:
248:
249: @example
250: #define INCLUDE_DEFAULTS \
251: @{ \
252: @{ "GNU_GXX_INCLUDE:", 1@}, \
253: @{ "GNU_CC_INCLUDE:", 0@}, \
254: @{ "SYS$SYSROOT:[SYSLIB.]", 0@}, \
255: @{ ".", 0@}, \
256: @{ 0, 0@} \
257: @}
258: @end example
259: @end table
260:
261: Here is the order of prefixes tried for exec files:
262:
263: @enumerate
264: @item
265: Any prefixes specified by the user with @samp{-B}.
266:
267: @item
268: The environment variable @code{GCC_EXEC_PREFIX}, if any.
269:
270: @item
271: The directories specified by the environment variable @code{COMPILER_PATH}.
272:
273: @item
274: The macro @code{STANDARD_EXEC_PREFIX}.
275:
276: @item
277: @file{/usr/lib/gcc/}.
278:
279: @item
280: The macro @code{MD_EXEC_PREFIX}, if any.
281: @end enumerate
282:
283: Here is the order of prefixes tried for startfiles:
284:
285: @enumerate
286: @item
287: Any prefixes specified by the user with @samp{-B}.
288:
289: @item
290: The environment variable @code{GCC_EXEC_PREFIX}, if any.
291:
292: @item
293: The directories specified by the environment variable @code{LIBRARY_PATH}.
294:
295: @item
296: The macro @code{STANDARD_EXEC_PREFIX}.
297:
298: @item
299: @file{/usr/lib/gcc/}.
300:
301: @item
302: The macro @code{MD_EXEC_PREFIX}, if any.
303:
304: @item
305: The macro @code{MD_STARTFILE_PREFIX}, if any.
306:
307: @item
308: The macro @code{STANDARD_STARTFILE_PREFIX}.
309:
310: @item
311: @file{/lib/}.
312:
313: @item
314: @file{/usr/lib/}.
315: @end enumerate
316:
1.1.1.2 ! root 317: @node Run-time Target
1.1 root 318: @section Run-time Target Specification
319: @cindex run-time target specification
320: @cindex predefined macros
321: @cindex target specifications
322:
323: @table @code
324: @findex CPP_PREDEFINES
325: @item CPP_PREDEFINES
326: Define this to be a string constant containing @samp{-D} options to
327: define the predefined macros that identify this machine and system.
328: These macros will be predefined unless the @samp{-ansi} option is
329: specified.
330:
331: In addition, a parallel set of macros are predefined, whose names are
332: made by appending @samp{__} at the beginning and at the end. These
333: @samp{__} macros are permitted by the ANSI standard, so they are
334: predefined regardless of whether @samp{-ansi} is specified.
335:
336: For example, on the Sun, one can use the following value:
337:
338: @example
339: "-Dmc68000 -Dsun -Dunix"
340: @end example
341:
342: The result is to define the macros @code{__mc68000__}, @code{__sun__}
343: and @code{__unix__} unconditionally, and the macros @code{mc68000},
344: @code{sun} and @code{unix} provided @samp{-ansi} is not specified.
345:
346: @findex STDC_VALUE
347: @item STDC_VALUE
348: Define the value to be assigned to the built-in macro @code{__STDC__}.
349: The default is the value @samp{1}.
350:
351: @findex extern int target_flags
352: @item extern int target_flags;
353: This declaration should be present.
354:
355: @cindex optional hardware or system features
356: @cindex features, optional, in system conventions
357: @item TARGET_@dots{}
358: This series of macros is to allow compiler command arguments to
359: enable or disable the use of optional features of the target machine.
360: For example, one machine description serves both the 68000 and
361: the 68020; a command argument tells the compiler whether it should
362: use 68020-only instructions or not. This command argument works
363: by means of a macro @code{TARGET_68020} that tests a bit in
364: @code{target_flags}.
365:
366: Define a macro @code{TARGET_@var{featurename}} for each such option.
367: Its definition should test a bit in @code{target_flags}; for example:
368:
369: @example
370: #define TARGET_68020 (target_flags & 1)
371: @end example
372:
373: One place where these macros are used is in the condition-expressions
374: of instruction patterns. Note how @code{TARGET_68020} appears
375: frequently in the 68000 machine description file, @file{m68k.md}.
376: Another place they are used is in the definitions of the other
377: macros in the @file{@var{machine}.h} file.
378:
379: @findex TARGET_SWITCHES
380: @item TARGET_SWITCHES
381: This macro defines names of command options to set and clear
382: bits in @code{target_flags}. Its definition is an initializer
383: with a subgrouping for each command option.
384:
385: Each subgrouping contains a string constant, that defines the option
386: name, and a number, which contains the bits to set in
387: @code{target_flags}. A negative number says to clear bits instead;
388: the negative of the number is which bits to clear. The actual option
389: name is made by appending @samp{-m} to the specified name.
390:
391: One of the subgroupings should have a null string. The number in
392: this grouping is the default value for @code{target_flags}. Any
393: target options act starting with that value.
394:
395: Here is an example which defines @samp{-m68000} and @samp{-m68020}
396: with opposite meanings, and picks the latter as the default:
397:
398: @example
399: #define TARGET_SWITCHES \
400: @{ @{ "68020", 1@}, \
401: @{ "68000", -1@}, \
402: @{ "", 1@}@}
403: @end example
404:
405: @findex TARGET_OPTIONS
406: @item TARGET_OPTIONS
407: This macro is similar to @code{TARGET_SWITCHES} but defines names of command
408: options that have values. Its definition is an initializer with a
409: subgrouping for each command option.
410:
411: Each subgrouping contains a string constant, that defines the fixed part
412: of the option name, and the address of a variable. The variable, type
413: @code{char *}, is set to the variable part of the given option if the fixed
414: part matches. The actual option name is made by appending @samp{-m} to the
415: specified name.
416:
417: Here is an example which defines @samp{-mshort-data-@var{number}}. If the
418: given option is @samp{-mshort-data-512}, the variable @code{m88k_short_data}
419: will be set to the string @code{"512"}.
420:
421: @example
422: extern char *m88k_short_data;
423: #define TARGET_OPTIONS @{ @{ "short-data-", &m88k_short_data @} @}
424: @end example
425:
426: @findex TARGET_VERSION
427: @item TARGET_VERSION
428: This macro is a C statement to print on @code{stderr} a string
429: describing the particular machine description choice. Every machine
430: description should define @code{TARGET_VERSION}. For example:
431:
432: @example
433: #ifdef MOTOROLA
434: #define TARGET_VERSION fprintf (stderr, " (68k, Motorola syntax)");
435: #else
436: #define TARGET_VERSION fprintf (stderr, " (68k, MIT syntax)");
437: #endif
438: @end example
439:
440: @findex OVERRIDE_OPTIONS
441: @item OVERRIDE_OPTIONS
442: Sometimes certain combinations of command options do not make sense on
443: a particular target machine. You can define a macro
444: @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
445: defined, is executed once just after all the command options have been
446: parsed.
447:
448: Don't use this macro to turn on various extra optimizations for
449: @samp{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
450:
451: @findex OPTIMIZATION_OPTIONS
452: @item OPTIMIZATION_OPTIONS (@var{level})
453: Some machines may desire to change what optimizations are performed for
454: various optimization levels. This macro, if defined, is executed once
455: just after the optimization level is determined and before the remainder
456: of the command options have been parsed. Values set in this macro are
457: used as the default values for the other command line options.
458:
459: @var{level} is the optimization level specified; 2 if -O2 is specified,
460: 1 if -O is specified, and 0 if neither is specified.
461:
462: @strong{Do not examine @code{write_symbols} in this macro!}
463: The debugging options are not supposed to alter the generated code.
464: @end table
465:
1.1.1.2 ! root 466: @node Storage Layout
1.1 root 467: @section Storage Layout
468: @cindex storage layout
469:
470: Note that the definitions of the macros in this table which are sizes or
471: alignments measured in bits do not need to be constant. They can be C
472: expressions that refer to static variables, such as the @code{target_flags}.
473: @xref{Run-time Target}.
474:
475: @table @code
476: @findex BITS_BIG_ENDIAN
477: @item BITS_BIG_ENDIAN
478: Define this macro to be the value 1 if the most significant bit in a
479: byte has the lowest number; otherwise define it to be the value zero.
480: This means that bit-field instructions count from the most significant
481: bit. If the machine has no bit-field instructions, this macro is
482: irrelevant.
483:
484: This macro does not affect the way structure fields are packed into
485: bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
486:
487: @findex BYTES_BIG_ENDIAN
488: @item BYTES_BIG_ENDIAN
489: Define this macro to be 1 if the most significant byte in a word has the
490: lowest number.
491:
492: @findex WORDS_BIG_ENDIAN
493: @item WORDS_BIG_ENDIAN
494: Define this macro to be 1 if, in a multiword object, the most
495: significant word has the lowest number.
496:
497: @findex BITS_PER_UNIT
498: @item BITS_PER_UNIT
499: Number of bits in an addressable storage unit (byte); normally 8.
500:
501: @findex BITS_PER_WORD
502: @item BITS_PER_WORD
503: Number of bits in a word; normally 32.
504:
505: @findex MAX_BITS_PER_WORD
506: @item MAX_BITS_PER_WORD
507: Maximum number of bits in a word. If this is undefined, the default is
508: @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
509: largest value that @code{BITS_PER_WORD} can have at run-time.
510:
511: @findex UNITS_PER_WORD
512: @item UNITS_PER_WORD
513: Number of storage units in a word; normally 4.
514:
515: @findex POINTER_SIZE
516: @item POINTER_SIZE
517: Width of a pointer, in bits.
518:
519: @findex PARM_BOUNDARY
520: @item PARM_BOUNDARY
521: Normal alignment required for function parameters on the stack, in
522: bits. All stack parameters receive least this much alignment
523: regardless of data type. On most machines, this is the same as the
524: size of an integer.
525:
526: @findex STACK_BOUNDARY
527: @item STACK_BOUNDARY
528: Define this macro if you wish to preserve a certain alignment for
529: the stack pointer. The definition is a C expression
530: for the desired alignment (measured in bits).
531:
532: @cindex @code{PUSH_ROUNDING}, interaction with @code{STACK_BOUNDARY}
533: If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
534: to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies a
535: less strict alignment than @code{STACK_BOUNDARY}, the stack may be
536: momentarily unaligned while pushing arguments.
537:
538: @findex FUNCTION_BOUNDARY
539: @item FUNCTION_BOUNDARY
540: Alignment required for a function entry point, in bits.
541:
542: @findex BIGGEST_ALIGNMENT
543: @item BIGGEST_ALIGNMENT
544: Biggest alignment that any data type can require on this machine, in bits.
545:
546: @findex BIGGEST_FIELD_ALIGNMENT
547: @item BIGGEST_FIELD_ALIGNMENT
548: Biggest alignment that any structure field can require on this machine,
549: in bits.
550:
551: @findex MAX_OFILE_ALIGNMENT
552: @item MAX_OFILE_ALIGNMENT
553: Biggest alignment supported by the object file format of this machine.
554: Use this macro to limit the alignment which can be specified using the
555: @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
556: the default value is @code{BIGGEST_ALIGNMENT}.
557:
558: @findex DATA_ALIGNMENT
559: @item DATA_ALIGNMENT (@var{type}, @var{basic-align})
560: If defined, a C expression to compute the alignment for a static
561: variable. @var{type} is the data type, and @var{basic-align} is the
562: alignment that the object would ordinarily have. The value of this
563: macro is used instead of that alignment to align the object.
564:
565: If this macro is not defined, then @var{basic-align} is used.
566:
567: @findex strcpy
568: One use of this macro is to increase alignment of medium-size data to
569: make it all fit in fewer cache lines. Another is to cause character
570: arrays to be word-aligned so that @code{strcpy} calls that copy
571: constants to character arrays can be done inline.
572:
573: @findex CONSTANT_ALIGNMENT
574: @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
575: If defined, a C expression to compute the alignment given to a constant
576: that is being placed in memory. @var{constant} is the constant and
577: @var{basic-align} is the alignment that the object would ordinarily
578: have. The value of this macro is used instead of that alignment to
579: align the object.
580:
581: If this macro is not defined, then @var{basic-align} is used.
582:
583: The typical use of this macro is to increase alignment for string
584: constants to be word aligned so that @code{strcpy} calls that copy
585: constants can be done inline.
586:
587: @findex EMPTY_FIELD_BOUNDARY
588: @item EMPTY_FIELD_BOUNDARY
589: Alignment in bits to be given to a structure bit field that follows an
590: empty field such as @code{int : 0;}.
591:
1.1.1.2 ! root 592: Note that @code{PCC_BITFIELD_TYPE_MATTERS} also affects the alignment
! 593: that results from an empty field.
! 594:
1.1 root 595: @findex STRUCTURE_SIZE_BOUNDARY
596: @item STRUCTURE_SIZE_BOUNDARY
597: Number of bits which any structure or union's size must be a multiple of.
598: Each structure or union's size is rounded up to a multiple of this.
599:
600: If you do not define this macro, the default is the same as
601: @code{BITS_PER_UNIT}.
602:
603: @findex STRICT_ALIGNMENT
604: @item STRICT_ALIGNMENT
1.1.1.2 ! root 605: Define this macro to be the value 1 if instructions will fail to work
! 606: if given data not on the nominal alignment. If instructions will merely
! 607: go slower in that case, define this macro as 0.
1.1 root 608:
609: @findex PCC_BITFIELD_TYPE_MATTERS
610: @item PCC_BITFIELD_TYPE_MATTERS
611: Define this if you wish to imitate the way many other C compilers handle
612: alignment of bitfields and the structures that contain them.
613:
614: The behavior is that the type written for a bitfield (@code{int},
615: @code{short}, or other integer type) imposes an alignment for the
616: entire structure, as if the structure really did contain an ordinary
617: field of that type. In addition, the bitfield is placed within the
618: structure so that it would fit within such a field, not crossing a
619: boundary for it.
620:
621: Thus, on most machines, a bitfield whose type is written as @code{int}
622: would not cross a four-byte boundary, and would force four-byte
623: alignment for the whole structure. (The alignment used may not be four
624: bytes; it is controlled by the other alignment parameters.)
625:
626: If the macro is defined, its definition should be a C expression;
627: a nonzero value for the expression enables this behavior.
628:
629: Note that if this macro is not defined, or its value is zero, some
630: bitfields may cross more than one alignment boundary. The compiler can
631: support such references if there are @samp{insv}, @samp{extv}, and
632: @samp{extzv} insns that can directly reference memory.
633:
634: The other known way of making bitfields work is to define
635: @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
636: Then every structure can be accessed with fullwords.
637:
638: Unless the machine has bitfield instructions or you define
639: @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
640: @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
641:
642: @findex BITFIELD_NBYTES_LIMITED
643: @item BITFIELD_NBYTES_LIMITED
644: Like PCC_BITFIELD_TYPE_MATTERS except that its effect is limited to
645: aligning a bitfield within the structure.
646:
647: @findex ROUND_TYPE_SIZE
648: @item ROUND_TYPE_SIZE (@var{struct}, @var{size}, @var{align})
649: Define this macro as an expression for the overall size of a structure
650: (given by @var{struct} as a tree node) when the size computed from the
651: fields is @var{size} and the alignment is @var{align}.
652:
653: The default is to round @var{size} up to a multiple of @var{align}.
654:
655: @findex ROUND_TYPE_ALIGN
656: @item ROUND_TYPE_ALIGN (@var{struct}, @var{computed}, @var{specified})
657: Define this macro as an expression for the alignment of a structure
658: (given by @var{struct} as a tree node) if the alignment computed in the
659: usual way is @var{computed} and the alignment explicitly specified was
660: @var{specified}.
661:
662: The default is to use @var{specified} if it is larger; otherwise, use
663: the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
664:
665: @findex MAX_FIXED_MODE_SIZE
666: @item MAX_FIXED_MODE_SIZE
667: An integer expression for the size in bits of the largest integer
668: machine mode that should actually be used. All integer machine modes of
669: this size or smaller can be used for structures and unions with the
670: appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
671: (DImode)} is assumed.
672:
673: @findex CHECK_FLOAT_VALUE
674: @item CHECK_FLOAT_VALUE (@var{mode}, @var{value})
675: A C statement to validate the value @var{value} (of type
676: @code{double}) for mode @var{mode}. This means that you check whether
677: @var{value} fits within the possible range of values for mode
678: @var{mode} on this target machine. The mode @var{mode} is always
679: @code{SFmode} or @code{DFmode}.
680:
681: @findex error
682: If @var{value} is not valid, you should call @code{error} to print an
683: error message and then assign some valid value to @var{value}.
684: Allowing an invalid value to go through the compiler can produce
685: incorrect assembler code which may even cause Unix assemblers to
686: crash.
687:
688: This macro need not be defined if there is no work for it to do.
689:
690: @findex TARGET_FLOAT_FORMAT
691: @item TARGET_FLOAT_FORMAT
692: A code distinguishing the floating point format of the target machine.
693: There are three defined values:
694:
695: @table @code
696: @findex IEEE_FLOAT_FORMAT
697: @item IEEE_FLOAT_FORMAT
698: This code indicates IEEE floating point. It is the default; there is no
699: need to define this macro when the format is IEEE.
700:
701: @findex VAX_FLOAT_FORMAT
702: @item VAX_FLOAT_FORMAT
703: This code indicates the peculiar format used on the Vax.
704:
705: @findex UNKNOWN_FLOAT_FORMAT
706: @item UNKNOWN_FLOAT_FORMAT
707: This code indicates any other format.
708: @end table
709:
710: The value of this macro is compared with @code{HOST_FLOAT_FORMAT}
711: (@pxref{Config}) to determine whether the target machine has the same
712: format as the host machine. If any other formats are actually in use on
713: supported machines, new codes should be defined for them.
714: @end table
715:
1.1.1.2 ! root 716: @node Type Layout
1.1 root 717: @section Layout of Source Language Data Types
718:
719: These macros define the sizes and other characteristics of the standard
720: basic data types used in programs being compiled. Unlike the macros in
721: the previous section, these apply to specific features of C and related
722: languages, rather than to fundamental aspects of storage layout.
723:
724: @table @code
725: @findex INT_TYPE_SIZE
726: @item INT_TYPE_SIZE
727: A C expression for the size in bits of the type @code{int} on the
728: target machine. If you don't define this, the default is one word.
729:
730: @findex SHORT_TYPE_SIZE
731: @item SHORT_TYPE_SIZE
732: A C expression for the size in bits of the type @code{short} on the
733: target machine. If you don't define this, the default is half a word.
734: (If this would be less than one storage unit, it is rounded up to one
735: unit.)
736:
737: @findex LONG_TYPE_SIZE
738: @item LONG_TYPE_SIZE
739: A C expression for the size in bits of the type @code{long} on the
740: target machine. If you don't define this, the default is one word.
741:
742: @findex LONG_LONG_TYPE_SIZE
743: @item LONG_LONG_TYPE_SIZE
744: A C expression for the size in bits of the type @code{long long} on the
745: target machine. If you don't define this, the default is two
746: words.
747:
748: @findex CHAR_TYPE_SIZE
749: @item CHAR_TYPE_SIZE
750: A C expression for the size in bits of the type @code{char} on the
751: target machine. If you don't define this, the default is one quarter
752: of a word. (If this would be less than one storage unit, it is rounded up
753: to one unit.)
754:
755: @findex FLOAT_TYPE_SIZE
756: @item FLOAT_TYPE_SIZE
757: A C expression for the size in bits of the type @code{float} on the
758: target machine. If you don't define this, the default is one word.
759:
760: @findex DOUBLE_TYPE_SIZE
761: @item DOUBLE_TYPE_SIZE
762: A C expression for the size in bits of the type @code{double} on the
763: target machine. If you don't define this, the default is two
764: words.
765:
766: @findex LONG_DOUBLE_TYPE_SIZE
767: @item LONG_DOUBLE_TYPE_SIZE
768: A C expression for the size in bits of the type @code{long double} on
769: the target machine. If you don't define this, the default is two
770: words.
771:
772: @findex DEFAULT_SIGNED_CHAR
773: @item DEFAULT_SIGNED_CHAR
774: An expression whose value is 1 or 0, according to whether the type
775: @code{char} should be signed or unsigned by default. The user can
776: always override this default with the options @samp{-fsigned-char}
777: and @samp{-funsigned-char}.
778:
779: @findex DEFAULT_SHORT_ENUMS
780: @item DEFAULT_SHORT_ENUMS
781: A C expression to determine whether to give an @code{enum} type
782: only as many bytes as it takes to represent the range of possible values
783: of that type. A nonzero value means to do that; a zero value means all
784: @code{enum} types should be allocated like @code{int}.
785:
786: If you don't define the macro, the default is 0.
787:
788: @findex SIZE_TYPE
789: @item SIZE_TYPE
790: A C expression for a string describing the name of the data type to use
791: for size values. The typedef name @code{size_t} is defined using the
792: contents of the string.
793:
794: The string can contain more than one keyword. If so, separate them with
795: spaces, and write first any length keyword, then @code{unsigned} if
796: appropriate, and finally @code{int}. The string must exactly match one
797: of the data type names defined in the function
798: @code{init_decl_processing} in the file @file{c-decl.c}. You may not
799: omit @code{int} or change the order---that would cause the compiler to
800: crash on startup.
801:
802: If you don't define this macro, the default is @code{"long unsigned
803: int"}.
804:
805: @findex PTRDIFF_TYPE
806: @item PTRDIFF_TYPE
807: A C expression for a string describing the name of the data type to use
808: for the result of subtracting two pointers. The typedef name
809: @code{ptrdiff_t} is defined using the contents of the string. See
810: @code{SIZE_TYPE} above for more information.
811:
812: If you don't define this macro, the default is @code{"long int"}.
813:
814: @findex WCHAR_TYPE
815: @item WCHAR_TYPE
816: A C expression for a string describing the name of the data type to use
817: for wide characters. The typedef name @code{wchar_t} is defined using
818: the contents of the string. See @code{SIZE_TYPE} above for more
819: information.
820:
821: If you don't define this macro, the default is @code{"int"}.
822:
823: @findex WCHAR_TYPE_SIZE
824: @item WCHAR_TYPE_SIZE
825: A C expression for the size in bits of the data type for wide
826: characters. This is used in @code{cpp}, which cannot make use of
827: @code{WCHAR_TYPE}.
828:
829: @findex OBJC_INT_SELECTORS
830: @item OBJC_INT_SELECTORS
831: Define this macro if the type of Objective C selectors should be
832: @code{int}.
833:
834: If this macro is not defined, then selectors should have the type
835: @code{struct objc_selector *}.
836:
837: @findex OBJC_NONUNIQUE_SELECTORS
838: @item OBJC_NONUNIQUE_SELECTORS
839: Define this macro if Objective C selector-references will be made unique
840: by the linker (this is the default). In this case, each
841: selector-reference will be given a separate assembler label. Otherwise,
842: the selector-references will be gathered into an array with a single
843: assembler label.
844:
845: @findex MULTIBYTE_CHARS
846: @item MULTIBYTE_CHARS
847: Define this macro to enable support for multibyte characters in the input
848: to GNU CC. This requires that the host system support the ANSI C library
849: functions for converting multibyte characters to wide characters.
850:
851: @findex TARGET_BELL
852: @item TARGET_BELL
853: A C constant expression for the integer value for escape sequence
854: @samp{\a}.
855:
856: @findex TARGET_TAB
857: @findex TARGET_BS
858: @findex TARGET_NEWLINE
859: @item TARGET_BS
860: @itemx TARGET_TAB
861: @itemx TARGET_NEWLINE
862: C constant expressions for the integer values for escape sequences
863: @samp{\b}, @samp{\t} and @samp{\n}.
864:
865: @findex TARGET_VT
866: @findex TARGET_FF
867: @findex TARGET_CR
868: @item TARGET_VT
869: @itemx TARGET_FF
870: @itemx TARGET_CR
871: C constant expressions for the integer values for escape sequences
872: @samp{\v}, @samp{\f} and @samp{\r}.
873: @end table
874:
1.1.1.2 ! root 875: @node Registers
1.1 root 876: @section Register Usage
877: @cindex register usage
878:
879: This section explains how to describe what registers the target machine
880: has, and how (in general) they can be used.
881:
882: The description of which registers a specific instruction can use is
883: done with register classes; see @ref{Register Classes}. For information
884: on using registers to access a stack frame, see @ref{Frame Registers}.
885: For passing values in registers, see @ref{Register Arguments}.
886: For returning values in registers, see @ref{Scalar Return}.
887:
888: @menu
889: * Register Basics:: Number and kinds of registers.
890: * Allocation Order:: Order in which registers are allocated.
891: * Values in Registers:: What kinds of values each reg can hold.
892: * Leaf Functions:: Renumbering registers for leaf functions.
893: * Stack Registers:: Handling a register stack such as 80387.
894: * Obsolete Register Macros:: Macros formerly used for the 80387.
895: @end menu
896:
897: @node Register Basics
898: @subsection Basic Characteristics of Registers
899:
900: @table @code
901: @findex FIRST_PSEUDO_REGISTER
902: @item FIRST_PSEUDO_REGISTER
903: Number of hardware registers known to the compiler. They receive
904: numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
905: pseudo register's number really is assigned the number
906: @code{FIRST_PSEUDO_REGISTER}.
907:
908: @item FIXED_REGISTERS
909: @findex FIXED_REGISTERS
910: @cindex fixed register
911: An initializer that says which registers are used for fixed purposes
912: all throughout the compiled code and are therefore not available for
913: general allocation. These would include the stack pointer, the frame
914: pointer (except on machines where that can be used as a general
915: register when no frame pointer is needed), the program counter on
916: machines where that is considered one of the addressable registers,
917: and any other numbered register with a standard use.
918:
919: This information is expressed as a sequence of numbers, separated by
920: commas and surrounded by braces. The @var{n}th number is 1 if
921: register @var{n} is fixed, 0 otherwise.
922:
923: The table initialized from this macro, and the table initialized by
924: the following one, may be overridden at run time either automatically,
925: by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
926: the user with the command options @samp{-ffixed-@var{reg}},
927: @samp{-fcall-used-@var{reg}} and @samp{-fcall-saved-@var{reg}}.
928:
929: @findex CALL_USED_REGISTERS
930: @item CALL_USED_REGISTERS
931: @cindex call-used register
932: @cindex call-clobbered register
933: @cindex call-saved register
934: Like @code{FIXED_REGISTERS} but has 1 for each register that is
935: clobbered (in general) by function calls as well as for fixed
936: registers. This macro therefore identifies the registers that are not
937: available for general allocation of values that must live across
938: function calls.
939:
940: If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
941: automatically saves it on function entry and restores it on function
942: exit, if the register is used within the function.
943:
944: @findex CONDITIONAL_REGISTER_USAGE
945: @findex fixed_regs
946: @findex call_used_regs
947: @item CONDITIONAL_REGISTER_USAGE
948: Zero or more C statements that may conditionally modify two variables
949: @code{fixed_regs} and @code{call_used_regs} (both of type @code{char
950: []}) after they have been initialized from the two preceding macros.
951:
952: This is necessary in case the fixed or call-clobbered registers depend
953: on target flags.
954:
955: You need not define this macro if it has no work to do.
956:
957: @cindex disabling certain registers
958: @cindex controlling register usage
959: If the usage of an entire class of registers depends on the target
960: flags, you may indicate this to GCC by using this macro to modify
961: @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
962: registers in the classes which should not be used by GCC. Also define
963: the macro @code{REG_CLASS_FROM_LETTER} to return @code{NO_REGS} if it
964: is called with a letter for a class that shouldn't be used.
965:
966: (However, if this class is not included in @code{GENERAL_REGS} and all
967: of the insn patterns whose constraints permit this class are
968: controlled by target switches, then GCC will automatically avoid using
969: these registers when the target switches are opposed to them.)
970:
971: @findex NON_SAVING_SETJMP
972: @item NON_SAVING_SETJMP
973: If this macro is defined and has a nonzero value, it means that
974: @code{setjmp} and related functions fail to save the registers, or that
975: @code{longjmp} fails to restore them. To compensate, the compiler
976: avoids putting variables in registers in functions that use
977: @code{setjmp}.
978:
979: @ignore
980: @findex PC_REGNUM
981: @item PC_REGNUM
982: If the program counter has a register number, define this as that
983: register number. Otherwise, do not define it.
984: @end ignore
985: @end table
986:
987: @node Allocation Order
988: @subsection Order of Allocation of Registers
989: @cindex order of register allocation
990: @cindex register allocation order
991:
992: @table @code
993: @findex REG_ALLOC_ORDER
994: @item REG_ALLOC_ORDER
995: If defined, an initializer for a vector of integers, containing the
996: numbers of hard registers in the order in which GNU CC should prefer
997: to use them (from most preferred to least).
998:
999: If this macro is not defined, registers are used lowest numbered first
1000: (all else being equal).
1001:
1002: One use of this macro is on machines where the highest numbered
1003: registers must always be saved and the save-multiple-registers
1004: instruction supports only sequences of consecutive registers. On such
1005: machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1006: the highest numbered allocatable register first.
1007:
1008: @findex ORDER_REGS_FOR_LOCAL_ALLOC
1009: @item ORDER_REGS_FOR_LOCAL_ALLOC
1010: A C statement (sans semicolon) to choose the order in which to allocate
1011: hard registers for pseudo-registers local to a basic block.
1012:
1013: Store the desired order of registers in the array
1014: @code{reg_alloc_order}. Element 0 should be the register to allocate
1015: first; element 1, the next register; and so on.
1016:
1017: The macro body should not assume anything about the contents of
1018: @code{reg_alloc_order} before execution of the macro.
1019:
1020: On most machines, it is not necessary to define this macro.
1021: @end table
1022:
1023: @node Values in Registers
1024: @subsection How Values Fit in Registers
1025:
1026: This section discusses the macros that describe which kinds of values
1027: (specifically, which machine modes) each register can hold, and how many
1028: consecutive registers are needed for a given mode.
1029:
1030: @table @code
1031: @findex HARD_REGNO_NREGS
1032: @item HARD_REGNO_NREGS (@var{regno}, @var{mode})
1033: A C expression for the number of consecutive hard registers, starting
1034: at register number @var{regno}, required to hold a value of mode
1035: @var{mode}.
1036:
1037: On a machine where all registers are exactly one word, a suitable
1038: definition of this macro is
1039:
1040: @example
1041: #define HARD_REGNO_NREGS(REGNO, MODE) \
1042: ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
1043: / UNITS_PER_WORD))
1044: @end example
1045:
1046: @findex HARD_REGNO_MODE_OK
1047: @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
1048: A C expression that is nonzero if it is permissible to store a value
1049: of mode @var{mode} in hard register number @var{regno} (or in several
1050: registers starting with that one). For a machine where all registers
1051: are equivalent, a suitable definition is
1052:
1053: @example
1054: #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
1055: @end example
1056:
1057: It is not necessary for this macro to check for the numbers of fixed
1058: registers, because the allocation mechanism considers them to be always
1059: occupied.
1060:
1061: @cindex register pairs
1062: On some machines, double-precision values must be kept in even/odd
1063: register pairs. The way to implement that is to define this macro
1064: to reject odd register numbers for such modes.
1065:
1066: @ignore
1067: @c I think this is not true now
1068: GNU CC assumes that it can always move values between registers and
1069: (suitably addressed) memory locations. If it is impossible to move a
1070: value of a certain mode between memory and certain registers, then
1071: @code{HARD_REGNO_MODE_OK} must not allow this mode in those registers.
1072: @end ignore
1073:
1074: The minimum requirement for a mode to be OK in a register is that the
1075: @samp{mov@var{mode}} instruction pattern support moves between the
1076: register and any other hard register for which the mode is OK; and that
1077: moving a value into the register and back out not alter it.
1078:
1079: Since the same instruction used to move @code{SImode} will work for all
1080: narrower integer modes, it is not necessary on any machine for
1081: @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
1082: you define patterns @samp{movhi}, etc., to take advantage of this. This
1083: is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
1084: and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
1085: to be tieable.
1086:
1087: Many machines have special registers for floating point arithmetic.
1088: Often people assume that floating point machine modes are allowed only
1089: in floating point registers. This is not true. Any registers that
1090: can hold integers can safely @emph{hold} a floating point machine
1091: mode, whether or not floating arithmetic can be done on it in those
1092: registers. Integer move instructions can be used to move the values.
1093:
1094: On some machines, though, the converse is true: fixed-point machine
1095: modes may not go in floating registers. This is true if the floating
1096: registers normalize any value stored in them, because storing a
1097: non-floating value there would garble it. In this case,
1098: @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
1099: floating registers. But if the floating registers do not automatically
1100: normalize, if you can store any bit pattern in one and retrieve it
1101: unchanged without a trap, then any machine mode may go in a floating
1102: register and this macro should say so.
1103:
1104: The primary significance of special floating registers is rather that
1105: they are the registers acceptable in floating point arithmetic
1106: instructions. However, this is of no concern to
1107: @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
1108: constraints for those instructions.
1109:
1110: On some machines, the floating registers are especially slow to access,
1111: so that it is better to store a value in a stack frame than in such a
1112: register if floating point arithmetic is not being done. As long as the
1113: floating registers are not in class @code{GENERAL_REGS}, they will not
1114: be used unless some pattern's constraint asks for one.
1115:
1116: @findex MODES_TIEABLE_P
1117: @item MODES_TIEABLE_P (@var{mode1}, @var{mode2})
1118: A C expression that is nonzero if it is desirable to choose register
1119: allocation so as to avoid move instructions between a value of mode
1120: @var{mode1} and a value of mode @var{mode2}.
1121:
1122: If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
1123: @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are ever different
1124: for any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1},
1125: @var{mode2})} must be zero.
1126: @end table
1127:
1128: @node Leaf Functions
1129: @subsection Handling Leaf Functions
1130:
1131: @cindex leaf functions
1132: @cindex functions, leaf
1133: On some machines, a leaf function (i.e., one which make no calls) can run
1134: more efficiently if it does not make its own register window. Often this
1135: means it is required to receive its arguments in the registers where they
1136: are passed by the caller, instead of the registers where they would
1137: normally arrive. Also, the leaf function may use only those registers for
1138: its own variables and temporaries.
1139:
1140: GNU CC assigns register numbers before it knows whether the function is
1141: suitable for leaf function treatment. So it needs to renumber the
1142: registers in order to output a leaf function. The following macros
1143: accomplish this.
1144:
1145: @table @code
1146: @findex LEAF_REGISTERS
1147: @item LEAF_REGISTERS
1148: A C initializer for a vector, indexed by hard register number, which
1149: contains 1 for a register that is allowable in a candidate for leaf
1150: function treatment.
1151:
1152: If leaf function treatment involves renumbering the registers, then the
1153: registers marked here should be the ones before renumbering---those that
1154: GNU CC would ordinarily allocate. The registers which will actually be
1155: used in the assembler code, after renumbering, should not be marked with 1
1156: in this vector.
1157:
1158: Define this macro only if the target machine offers a way to optimize
1159: the treatment of leaf functions.
1160:
1161: @findex LEAF_REG_REMAP
1162: @item LEAF_REG_REMAP (@var{regno})
1163: A C expression whose value is the register number to which @var{regno}
1164: should be renumbered, when a function is treated as a leaf function.
1165:
1166: If @var{regno} is a register number which should not appear in a leaf
1167: function before renumbering, then the expression should yield -1, which
1168: will cause the compiler to abort.
1169:
1170: Define this macro only if the target machine offers a way to optimize the
1171: treatment of leaf functions, and registers need to be renumbered to do
1172: this.
1173:
1174: @findex REG_LEAF_ALLOC_ORDER
1175: @item REG_LEAF_ALLOC_ORDER
1176: If defined, an initializer for a vector of integers, containing the
1177: numbers of hard registers in the order in which the GNU CC should prefer
1178: to use them (from most preferred to least) in a leaf function. If this
1179: macro is not defined, REG_ALLOC_ORDER is used for both non-leaf and
1180: leaf-functions.
1181: @end table
1182:
1183: @findex leaf_function
1184: Normally, it is necessary for @code{FUNCTION_PROLOGUE} and
1185: @code{FUNCTION_EPILOGUE} to treat leaf functions specially. The C variable
1186: @code{leaf_function} is nonzero for such a function.
1187:
1188: @node Stack Registers
1189: @subsection Registers That Form a Stack
1190:
1191: There are special features to handle computers where some of the
1192: ``registers'' form a stack, as in the 80387 coprocessor for the 80386.
1193: Stack registers are normally written by pushing onto the stack, and are
1194: numbered relative to the top of the stack.
1195:
1196: Currently, GNU CC can only handle one group of stack-like registers, and
1197: they must be consecutively numbered.
1198:
1199: @table @code
1200: @findex STACK_REGS
1201: @item STACK_REGS
1202: Define this if the machine has any stack-like registers.
1203:
1204: @findex FIRST_STACK_REG
1205: @item FIRST_STACK_REG
1206: The number of the first stack-like register. This one is the top
1207: of the stack.
1208:
1209: @findex LAST_STACK_REG
1210: @item LAST_STACK_REG
1211: The number of the last stack-like register. This one is the bottom of
1212: the stack.
1213: @end table
1214:
1215: @node Obsolete Register Macros
1216: @subsection Obsolete Macros for Controlling Register Usage
1217:
1218: These features do not work very well. They exist because they used to
1219: be required to generate correct code for the 80387 coprocessor of the
1220: 80386. They are no longer used by that machine description and may be
1221: removed in a later version of the compiler. Don't use them!
1222:
1223: @table @code
1224: @findex OVERLAPPING_REGNO_P
1225: @item OVERLAPPING_REGNO_P (@var{regno})
1226: If defined, this is a C expression whose value is nonzero if hard
1227: register number @var{regno} is an overlapping register. This means a
1228: hard register which overlaps a hard register with a different number.
1229: (Such overlap is undesirable, but occasionally it allows a machine to
1230: be supported which otherwise could not be.) This macro must return
1231: nonzero for @emph{all} the registers which overlap each other. GNU CC
1232: can use an overlapping register only in certain limited ways. It can
1233: be used for allocation within a basic block, and may be spilled for
1234: reloading; that is all.
1235:
1236: If this macro is not defined, it means that none of the hard registers
1237: overlap each other. This is the usual situation.
1238:
1239: @findex INSN_CLOBBERS_REGNO_P
1240: @item INSN_CLOBBERS_REGNO_P (@var{insn}, @var{regno})
1241: If defined, this is a C expression whose value should be nonzero if
1242: the insn @var{insn} has the effect of mysteriously clobbering the
1243: contents of hard register number @var{regno}. By ``mysterious'' we
1244: mean that the insn's RTL expression doesn't describe such an effect.
1245:
1246: If this macro is not defined, it means that no insn clobbers registers
1247: mysteriously. This is the usual situation; all else being equal,
1248: it is best for the RTL expression to show all the activity.
1249:
1250: @cindex death notes
1251: @findex PRESERVE_DEATH_INFO_REGNO_P
1252: @item PRESERVE_DEATH_INFO_REGNO_P (@var{regno})
1253: If defined, this is a C expression whose value is nonzero if accurate
1254: @code{REG_DEAD} notes are needed for hard register number @var{regno}
1255: at the time of outputting the assembler code. When this is so, a few
1256: optimizations that take place after register allocation and could
1257: invalidate the death notes are not done when this register is
1258: involved.
1259:
1260: You would arrange to preserve death info for a register when some of the
1261: code in the machine description which is executed to write the assembler
1262: code looks at the death notes. This is necessary only when the actual
1263: hardware feature which GNU CC thinks of as a register is not actually a
1264: register of the usual sort. (It might, for example, be a hardware
1265: stack.)
1266:
1267: If this macro is not defined, it means that no death notes need to be
1268: preserved. This is the usual situation.
1269: @end table
1270:
1.1.1.2 ! root 1271: @node Register Classes
1.1 root 1272: @section Register Classes
1273: @cindex register class definitions
1274: @cindex class definitions, register
1275:
1276: On many machines, the numbered registers are not all equivalent.
1277: For example, certain registers may not be allowed for indexed addressing;
1278: certain registers may not be allowed in some instructions. These machine
1279: restrictions are described to the compiler using @dfn{register classes}.
1280:
1281: You define a number of register classes, giving each one a name and saying
1282: which of the registers belong to it. Then you can specify register classes
1283: that are allowed as operands to particular instruction patterns.
1284:
1285: @findex ALL_REGS
1286: @findex NO_REGS
1287: In general, each register will belong to several classes. In fact, one
1288: class must be named @code{ALL_REGS} and contain all the registers. Another
1289: class must be named @code{NO_REGS} and contain no registers. Often the
1290: union of two classes will be another class; however, this is not required.
1291:
1292: @findex GENERAL_REGS
1293: One of the classes must be named @code{GENERAL_REGS}. There is nothing
1294: terribly special about the name, but the operand constraint letters
1295: @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
1296: the same as @code{ALL_REGS}, just define it as a macro which expands
1297: to @code{ALL_REGS}.
1298:
1299: Order the classes so that if class @var{x} is contained in class @var{y}
1300: then @var{x} has a lower class number than @var{y}.
1301:
1302: The way classes other than @code{GENERAL_REGS} are specified in operand
1303: constraints is through machine-dependent operand constraint letters.
1304: You can define such letters to correspond to various classes, then use
1305: them in operand constraints.
1306:
1307: You should define a class for the union of two classes whenever some
1308: instruction allows both classes. For example, if an instruction allows
1309: either a floating point (coprocessor) register or a general register for a
1310: certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
1311: which includes both of them. Otherwise you will get suboptimal code.
1312:
1313: You must also specify certain redundant information about the register
1314: classes: for each class, which classes contain it and which ones are
1315: contained in it; for each pair of classes, the largest class contained
1316: in their union.
1317:
1318: When a value occupying several consecutive registers is expected in a
1319: certain class, all the registers used must belong to that class.
1320: Therefore, register classes cannot be used to enforce a requirement for
1321: a register pair to start with an even-numbered register. The way to
1322: specify this requirement is with @code{HARD_REGNO_MODE_OK}.
1323:
1324: Register classes used for input-operands of bitwise-and or shift
1325: instructions have a special requirement: each such class must have, for
1326: each fixed-point machine mode, a subclass whose registers can transfer that
1327: mode to or from memory. For example, on some machines, the operations for
1328: single-byte values (@code{QImode}) are limited to certain registers. When
1329: this is so, each register class that is used in a bitwise-and or shift
1330: instruction must have a subclass consisting of registers from which
1331: single-byte values can be loaded or stored. This is so that
1332: @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
1333:
1334: @table @code
1335: @findex enum reg_class
1336: @item enum reg_class
1337: An enumeral type that must be defined with all the register class names
1338: as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
1339: must be the last register class, followed by one more enumeral value,
1340: @code{LIM_REG_CLASSES}, which is not a register class but rather
1341: tells how many classes there are.
1342:
1343: Each register class has a number, which is the value of casting
1344: the class name to type @code{int}. The number serves as an index
1345: in many of the tables described below.
1346:
1347: @findex N_REG_CLASSES
1348: @item N_REG_CLASSES
1349: The number of distinct register classes, defined as follows:
1350:
1351: @example
1352: #define N_REG_CLASSES (int) LIM_REG_CLASSES
1353: @end example
1354:
1355: @findex REG_CLASS_NAMES
1356: @item REG_CLASS_NAMES
1357: An initializer containing the names of the register classes as C string
1358: constants. These names are used in writing some of the debugging dumps.
1359:
1360: @findex REG_CLASS_CONTENTS
1361: @item REG_CLASS_CONTENTS
1362: An initializer containing the contents of the register classes, as integers
1363: which are bit masks. The @var{n}th integer specifies the contents of class
1364: @var{n}. The way the integer @var{mask} is interpreted is that
1365: register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
1366:
1367: When the machine has more than 32 registers, an integer does not suffice.
1368: Then the integers are replaced by sub-initializers, braced groupings containing
1369: several integers. Each sub-initializer must be suitable as an initializer
1370: for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
1371:
1372: @findex REGNO_REG_CLASS
1373: @item REGNO_REG_CLASS (@var{regno})
1374: A C expression whose value is a register class containing hard register
1375: @var{regno}. In general there is more that one such class; choose a class
1376: which is @dfn{minimal}, meaning that no smaller class also contains the
1377: register.
1378:
1379: @findex BASE_REG_CLASS
1380: @item BASE_REG_CLASS
1381: A macro whose definition is the name of the class to which a valid
1382: base register must belong. A base register is one used in an address
1383: which is the register value plus a displacement.
1384:
1385: @findex INDEX_REG_CLASS
1386: @item INDEX_REG_CLASS
1387: A macro whose definition is the name of the class to which a valid
1388: index register must belong. An index register is one used in an
1389: address where its value is either multiplied by a scale factor or
1390: added to another register (as well as added to a displacement).
1391:
1392: @findex REG_CLASS_FROM_LETTER
1393: @item REG_CLASS_FROM_LETTER (@var{char})
1394: A C expression which defines the machine-dependent operand constraint
1395: letters for register classes. If @var{char} is such a letter, the
1396: value should be the register class corresponding to it. Otherwise,
1.1.1.2 ! root 1397: the value should be @code{NO_REGS}. The register letter @samp{r},
! 1398: corresponding to class @code{GENERAL_REGS}, will not be passed
! 1399: to this macro; you do not need to handle it.
1.1 root 1400:
1401: @findex REGNO_OK_FOR_BASE_P
1402: @item REGNO_OK_FOR_BASE_P (@var{num})
1403: A C expression which is nonzero if register number @var{num} is
1404: suitable for use as a base register in operand addresses. It may be
1405: either a suitable hard register or a pseudo register that has been
1406: allocated such a hard register.
1407:
1408: @findex REGNO_OK_FOR_INDEX_P
1409: @item REGNO_OK_FOR_INDEX_P (@var{num})
1410: A C expression which is nonzero if register number @var{num} is
1411: suitable for use as an index register in operand addresses. It may be
1412: either a suitable hard register or a pseudo register that has been
1413: allocated such a hard register.
1414:
1415: The difference between an index register and a base register is that
1416: the index register may be scaled. If an address involves the sum of
1417: two registers, neither one of them scaled, then either one may be
1418: labeled the ``base'' and the other the ``index''; but whichever
1419: labeling is used must fit the machine's constraints of which registers
1420: may serve in each capacity. The compiler will try both labelings,
1421: looking for one that is valid, and will reload one or both registers
1422: only if neither labeling works.
1423:
1424: @findex PREFERRED_RELOAD_CLASS
1425: @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
1426: A C expression that places additional restrictions on the register class
1427: to use when it is necessary to copy value @var{x} into a register in class
1428: @var{class}. The value is a register class; perhaps @var{class}, or perhaps
1429: another, smaller class. On many machines, the definition
1430:
1431: @example
1432: #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
1433: @end example
1434:
1435: @noindent
1436: is safe.
1437:
1438: Sometimes returning a more restrictive class makes better code. For
1439: example, on the 68000, when @var{x} is an integer constant that is in range
1440: for a @samp{moveq} instruction, the value of this macro is always
1441: @code{DATA_REGS} as long as @var{class} includes the data registers.
1442: Requiring a data register guarantees that a @samp{moveq} will be used.
1443:
1444: If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
1445: you can force @var{x} into a memory constant. This is useful on
1446: certain machines where immediate floating values cannot be loaded into
1447: certain kinds of registers.
1448:
1449: @findex LIMIT_RELOAD_CLASS
1450: @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
1451: A C expression that places additional restrictions on the register class
1452: to use when it is necessary to be able to hold a value of mode
1453: @var{mode} in a reload register for which class @var{class} would
1454: ordinarily be used.
1455:
1456: Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
1457: there are certain modes that simply can't go in certain reload classes.
1458:
1459: The value is a register class; perhaps @var{class}, or perhaps another,
1460: smaller class.
1461:
1462: Don't define this macro unless the target machine has limitations which
1463: require the macro to do something nontrivial.
1464:
1465: @findex SECONDARY_RELOAD_CLASS
1466: @findex SECONDARY_INPUT_RELOAD_CLASS
1467: @findex SECONDARY_OUTPUT_RELOAD_CLASS
1468: @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
1469: @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
1470: @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
1471: Many machines have some registers that cannot be copied directly to or
1472: from memory or even from other types of registers. An example is the
1473: @samp{MQ} register, which on most machines, can only be copied to or
1474: from general registers, but not memory. Some machines allow copying all
1475: registers to and from memory, but require a scratch register for stores
1476: to some memory locations (e.g., those with symbolic address on the RT,
1477: and those with certain symbolic address on the Sparc when compiling
1478: PIC). In some cases, both an intermediate and a scratch register are
1479: required.
1480:
1481: You should define these macros to indicate to the reload phase that it may
1482: need to allocate at least one register for a reload in addition to the
1483: register to contain the data. Specifically, if copying @var{x} to a
1484: register @var{class} in @var{mode} requires an intermediate register,
1485: you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
1486: largest register class all of whose registers can be used as
1487: intermediate registers or scratch registers.
1488:
1489: If copying a register @var{class} in @var{mode} to @var{x} requires an
1490: intermediate or scratch register, you should define
1491: @code{SECONDARY_OUTPUT_RELOAD_CLASS} to return the largest register
1492: class required. If the requirements for input and output reloads are
1493: the same, the macro @code{SECONDARY_RELOAD_CLASS} should be used instead
1494: of defining both macros identically.
1495:
1496: The values returned by these macros are often @code{GENERAL_REGS}.
1497: Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
1498: can be directly copied to or from a register of @var{class} in
1499: @var{mode} without requiring a scratch register. Do not define this
1500: macro if it would always return @code{NO_REGS}.
1501:
1502: If a scratch register is required (either with or without an
1503: intermediate register), you should define patterns for
1504: @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
1505: (@pxref{Standard Names}. These patterns, which will normally be
1506: implemented with a @code{define_expand}, should be similar to the
1507: @samp{mov@var{m}} patterns, except that operand 2 is the scratch
1508: register.
1509:
1510: Define constraints for the reload register and scratch register that
1511: contain a single register class. If the original reload register (whose
1512: class is @var{class}) can meet the constraint given in the pattern, the
1513: value returned by these macros is used for the class of the scratch
1514: register. Otherwise, two additional reload registers are required.
1515: Their classes are obtained from the constraints in the insn pattern.
1516:
1517: @var{x} might be a pseudo-register or a @code{subreg} of a
1518: pseudo-register, which could either be in a hard register or in memory.
1519: Use @code{true_regnum} to find out; it will return -1 if the pseudo is
1520: in memory and the hard register number if it is in a register.
1521:
1522: These macros should not be used in the case where a particular class of
1523: registers can only be copied to memory and not to another class of
1524: registers. In that case, secondary reload registers are not needed and
1525: would not be helpful. Instead, a stack location must be used to perform
1526: the copy and the @code{mov@var{m}} pattern should use memory as a
1527: intermediate storage. This case often occurs between floating-point and
1528: general registers.
1529:
1530: @findex SMALL_REGISTER_CLASSES
1531: @item SMALL_REGISTER_CLASSES
1532: Normally the compiler will avoid choosing spill registers from registers
1533: that have been explicitly mentioned in the rtl (these registers are
1534: normally those used to pass parameters and return values). However,
1535: some machines have so few registers of certain classes that there would
1536: not be enough registers to use as spill registers if this were done.
1537:
1538: On those machines, you should define @code{SMALL_REGISTER_CLASSES}.
1539: When it is defined, the compiler allows registers explicitly used in the
1540: rtl to be used as spill registers but prevents the compiler from
1541: extending the lifetime of these registers.
1542:
1543: Defining this macro is always safe, but unnecessarily defining this macro
1544: will reduce the amount of optimizations that can be performed in some
1545: cases. If this macro is not defined but needs to be, the compiler will
1546: run out of reload registers and print a fatal error message.
1547:
1548: For most machines, this macro should not be defined.
1549:
1550: @findex CLASS_MAX_NREGS
1551: @item CLASS_MAX_NREGS (@var{class}, @var{mode})
1552: A C expression for the maximum number of consecutive registers
1553: of class @var{class} needed to hold a value of mode @var{mode}.
1554:
1555: This is closely related to the macro @code{HARD_REGNO_NREGS}.
1556: In fact, the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
1557: should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno}, @var{mode})}
1558: for all @var{regno} values in the class @var{class}.
1559:
1560: This macro helps control the handling of multiple-word values
1561: in the reload pass.
1562: @end table
1563:
1564: Three other special macros describe which operands fit which constraint
1565: letters.
1566:
1567: @table @code
1568: @findex CONST_OK_FOR_LETTER_P
1569: @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
1570: A C expression that defines the machine-dependent operand constraint letters
1571: that specify particular ranges of integer values. If @var{c} is one
1572: of those letters, the expression should check that @var{value}, an integer,
1573: is in the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
1574: not one of those letters, the value should be 0 regardless of @var{value}.
1575:
1576: @findex CONST_DOUBLE_OK_FOR_LETTER_P
1577: @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
1578: A C expression that defines the machine-dependent operand constraint
1579: letters that specify particular ranges of @code{const_double} values.
1580:
1581: If @var{c} is one of those letters, the expression should check that
1582: @var{value}, an RTX of code @code{const_double}, is in the appropriate
1583: range and return 1 if so, 0 otherwise. If @var{c} is not one of those
1584: letters, the value should be 0 regardless of @var{value}.
1585:
1586: @code{const_double} is used for all floating-point constants and for
1587: @code{DImode} fixed-point constants. A given letter can accept either
1588: or both kinds of values. It can use @code{GET_MODE} to distinguish
1589: between these kinds.
1590:
1591: @findex EXTRA_CONSTRAINT
1592: @item EXTRA_CONSTRAINT (@var{value}, @var{c})
1593: A C expression that defines the optional machine-dependent constraint
1594: letters that can be used to segregate specific types of operands,
1595: usually memory references, for the target machine. Normally this macro
1596: will not be defined. If it is required for a particular target machine,
1597: it should return 1 if @var{value} corresponds to the operand type
1598: represented by the constraint letter @var{c}. If @var{c} is not defined
1599: as an extra constraint, the value returned should be 0 regardless of
1600: @var{value}.
1601:
1602: For example, on the ROMP, load instructions cannot have their output in r0 if
1603: the memory reference contains a symbolic address. Constraint letter
1604: @samp{Q} is defined as representing a memory address that does
1605: @emph{not} contain a symbolic address. An alternative is specified with
1606: a @samp{Q} constraint on the input and @samp{r} on the output. The next
1607: alternative specifies @samp{m} on the input and a register class that
1608: does not include r0 on the output.
1609: @end table
1610:
1.1.1.2 ! root 1611: @node Stack and Calling
1.1 root 1612: @section Describing Stack Layout and Calling Conventions
1613: @cindex calling conventions
1614:
1615: @menu
1616: * Frame Layout::
1617: * Frame Registers::
1618: * Elimination::
1619: * Stack Arguments::
1620: * Register Arguments::
1621: * Scalar Return::
1622: * Aggregate Return::
1623: * Caller Saves::
1624: * Function Entry::
1625: * Profiling::
1626: @end menu
1627:
1628: @node Frame Layout
1629: @subsection Basic Stack Layout
1630: @cindex stack frame layout
1631: @cindex frame layout
1632:
1633: @table @code
1634: @findex STACK_GROWS_DOWNWARD
1635: @item STACK_GROWS_DOWNWARD
1636: Define this macro if pushing a word onto the stack moves the stack
1637: pointer to a smaller address.
1638:
1639: When we say, ``define this macro if @dots{},'' it means that the
1640: compiler checks this macro only with @code{#ifdef} so the precise
1641: definition used does not matter.
1642:
1643: @findex FRAME_GROWS_DOWNWARD
1644: @item FRAME_GROWS_DOWNWARD
1645: Define this macro if the addresses of local variable slots are at negative
1646: offsets from the frame pointer.
1647:
1648: @findex ARGS_GROW_DOWNWARD
1649: @item ARGS_GROW_DOWNWARD
1650: Define this macro if successive arguments to a function occupy decreasing
1651: addresses on the stack.
1652:
1653: @findex STARTING_FRAME_OFFSET
1654: @item STARTING_FRAME_OFFSET
1655: Offset from the frame pointer to the first local variable slot to be allocated.
1656:
1657: If @code{FRAME_GROWS_DOWNWARD}, the next slot's offset is found by
1658: subtracting the length of the first slot from @code{STARTING_FRAME_OFFSET}.
1659: Otherwise, it is found by adding the length of the first slot to
1660: the value @code{STARTING_FRAME_OFFSET}.
1661:
1662: @findex STACK_POINTER_OFFSET
1663: @item STACK_POINTER_OFFSET
1664: Offset from the stack pointer register to the first location at which
1665: outgoing arguments are placed. If not specified, the default value of
1666: zero is used. This is the proper value for most machines.
1667:
1668: If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
1669: the first location at which outgoing arguments are placed.
1670:
1671: @findex FIRST_PARM_OFFSET
1672: @item FIRST_PARM_OFFSET (@var{fundecl})
1673: Offset from the argument pointer register to the first argument's
1674: address. On some machines it may depend on the data type of the
1675: function.
1676:
1677: If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
1678: the first argument's address.
1679:
1680: @findex STACK_DYNAMIC_OFFSET
1681: @item STACK_DYNAMIC_OFFSET (@var{fundecl})
1682: Offset from the stack pointer register to an item dynamically allocated
1683: on the stack, e.g., by @code{alloca}.
1684:
1685: The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
1686: length of the outgoing arguments. The default is correct for most
1687: machines. See @file{function.c} for details.
1688:
1689: @findex DYNAMIC_CHAIN_ADDRESS
1690: @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
1691: A C expression whose value is RTL representing the address in a stack
1692: frame where the pointer to the caller's frame is stored. Assume that
1693: @var{frameaddr} is an RTL expression for the address of the stack frame
1694: itself.
1695:
1696: If you don't define this macro, the default is to return the value
1697: of @var{frameaddr}---that is, the stack frame address is also the
1698: address of the stack word that points to the previous frame.
1699: @end table
1700:
1701: @node Frame Registers
1702: @subsection Registers That Address the Stack Frame
1703:
1704: @table @code
1705: @findex STACK_POINTER_REGNUM
1706: @item STACK_POINTER_REGNUM
1707: The register number of the stack pointer register, which must also be a
1708: fixed register according to @code{FIXED_REGISTERS}. On most machines,
1709: the hardware determines which register this is.
1710:
1711: @findex FRAME_POINTER_REGNUM
1712: @item FRAME_POINTER_REGNUM
1713: The register number of the frame pointer register, which is used to
1714: access automatic variables in the stack frame. On some machines, the
1715: hardware determines which register this is. On other machines, you can
1716: choose any register you wish for this purpose.
1717:
1718: @findex ARG_POINTER_REGNUM
1719: @item ARG_POINTER_REGNUM
1720: The register number of the arg pointer register, which is used to access
1721: the function's argument list. On some machines, this is the same as the
1722: frame pointer register. On some machines, the hardware determines which
1723: register this is. On other machines, you can choose any register you
1724: wish for this purpose. If this is not the same register as the frame
1725: pointer register, then you must mark it as a fixed register according to
1726: @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
1727: (@pxref{Elimination}).
1728:
1729: @findex STATIC_CHAIN_REGNUM
1730: @findex STATIC_CHAIN_INCOMING_REGNUM
1731: @item STATIC_CHAIN_REGNUM
1732: @itemx STATIC_CHAIN_INCOMING_REGNUM
1733: Register numbers used for passing a function's static chain
1734: pointer. If register windows are used, @code{STATIC_CHAIN_INCOMING_REGNUM}
1735: is the register number as seen by the called function, while
1736: @code{STATIC_CHAIN_REGNUM} is the register number as seen by the calling
1737: function. If these registers are the same,
1738: @code{STATIC_CHAIN_INCOMING_REGNUM} need not be defined.@refill
1739:
1740: The static chain register need not be a fixed register.
1741:
1742: If the static chain is passed in memory, these macros should not be
1743: defined; instead, the next two macros should be defined.
1744:
1745: @findex STATIC_CHAIN
1746: @findex STATIC_CHAIN_INCOMING
1747: @item STATIC_CHAIN
1748: @itemx STATIC_CHAIN_INCOMING
1749: If the static chain is passed in memory, these macros provide rtx giving
1750: @code{mem} expressions that denote where they are stored.
1751: @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
1752: as seen by the calling and called functions, respectively. Often the former
1753: will be at an offset from the stack pointer and the latter at an offset from
1754: the frame pointer.@refill
1755:
1756: @findex stack_pointer_rtx
1757: @findex frame_pointer_rtx
1758: @findex arg_pointer_rtx
1759: The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
1760: @code{arg_pointer_rtx} will have been initialized prior to the use of these
1761: macros and should be used to refer to those items.
1762:
1763: If the static chain is passed in a register, the two previous macros should
1764: be defined instead.
1765: @end table
1766:
1767: @node Elimination
1768: @subsection Eliminating Frame Pointer and Arg Pointer
1769:
1770: @table @code
1771: @findex FRAME_POINTER_REQUIRED
1772: @item FRAME_POINTER_REQUIRED
1773: A C expression which is nonzero if a function must have and use a frame
1774: pointer. This expression is evaluated in the reload pass. If its value is
1775: nonzero the function will have a frame pointer.
1776:
1777: The expression can in principle examine the current function and decide
1778: according to the facts, but on most machines the constant 0 or the
1779: constant 1 suffices. Use 0 when the machine allows code to be generated
1780: with no frame pointer, and doing so saves some time or space. Use 1
1781: when there is no possible advantage to avoiding a frame pointer.
1782:
1783: In certain cases, the compiler does not know how to produce valid code
1784: without a frame pointer. The compiler recognizes those cases and
1785: automatically gives the function a frame pointer regardless of what
1786: @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
1787: them.@refill
1788:
1789: In a function that does not require a frame pointer, the frame pointer
1790: register can be allocated for ordinary usage, unless you mark it as a
1791: fixed register. See @code{FIXED_REGISTERS} for more information.
1792:
1793: This macro is ignored and need not be defined if @code{ELIMINABLE_REGS}
1794: is defined.
1795:
1796: @findex INITIAL_FRAME_POINTER_OFFSET
1797: @findex get_frame_size
1798: @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
1799: A C statement to store in the variable @var{depth-var} the difference
1800: between the frame pointer and the stack pointer values immediately after
1801: the function prologue. The value would be computed from information
1802: such as the result of @code{get_frame_size ()} and the tables of
1803: registers @code{regs_ever_live} and @code{call_used_regs}.
1804:
1805: If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
1806: need not be defined. Otherwise, it must be defined even if
1807: @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
1808: case, you may set @var{depth-var} to anything.
1809:
1810: @findex ELIMINABLE_REGS
1811: @item ELIMINABLE_REGS
1812: If defined, this macro specifies a table of register pairs used to
1813: eliminate unneeded registers that point into the stack frame. If it is not
1814: defined, the only elimination attempted by the compiler is to replace
1815: references to the frame pointer with references to the stack pointer.
1816:
1817: The definition of this macro is a list of structure initializations, each
1818: of which specifies an original and replacement register.
1819:
1820: On some machines, the position of the argument pointer is not known until
1821: the compilation is completed. In such a case, a separate hard register
1822: must be used for the argument pointer. This register can be eliminated by
1823: replacing it with either the frame pointer or the argument pointer,
1824: depending on whether or not the frame pointer has been eliminated.
1825:
1826: In this case, you might specify:
1827: @example
1828: #define ELIMINABLE_REGS \
1829: @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
1830: @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
1831: @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
1832: @end example
1833:
1834: Note that the elimination of the argument pointer with the stack pointer is
1835: specified first since that is the preferred elimination.
1836:
1837: @findex CAN_ELIMINATE
1838: @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
1839: A C expression that returns non-zero if the compiler is allowed to try
1840: to replace register number @var{from-reg} with register number
1841: @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
1842: is defined, and will usually be the constant 1, since most of the cases
1843: preventing register elimination are things that the compiler already
1844: knows about.
1845:
1846: @findex INITIAL_ELIMINATION_OFFSET
1847: @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
1848: This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
1849: specifies the initial difference between the specified pair of
1850: registers. This macro must be defined if @code{ELIMINABLE_REGS} is
1851: defined.
1852:
1853: @findex LONGJMP_RESTORE_FROM_STACK
1854: @item LONGJMP_RESTORE_FROM_STACK
1855: Define this macro if the @code{longjmp} function restores registers from
1856: the stack frames, rather than from those saved specifically by
1857: @code{setjmp}. Certain quantities must not be kept in registers across
1858: a call to @code{setjmp} on such machines.
1859: @end table
1860:
1861: @node Stack Arguments
1862: @subsection Passing Function Arguments on the Stack
1863: @cindex arguments on stack
1864: @cindex stack arguments
1865:
1866: The macros in this section control how arguments are passed
1867: on the stack. See the following section for other macros that
1868: control passing certain arguments in registers.
1869:
1870: @table @code
1871: @findex PROMOTE_PROTOTYPES
1872: @item PROMOTE_PROTOTYPES
1873: Define this macro if an argument declared as @code{char} or
1874: @code{short} in a prototype should actually be passed as an
1875: @code{int}. In addition to avoiding errors in certain cases of
1876: mismatch, it also makes for better code on certain machines.
1877:
1878: @findex PUSH_ROUNDING
1879: @item PUSH_ROUNDING (@var{npushed})
1880: A C expression that is the number of bytes actually pushed onto the
1881: stack when an instruction attempts to push @var{npushed} bytes.
1882:
1883: If the target machine does not have a push instruction, do not define
1884: this macro. That directs GNU CC to use an alternate strategy: to
1885: allocate the entire argument block and then store the arguments into
1886: it.
1887:
1888: On some machines, the definition
1889:
1890: @example
1891: #define PUSH_ROUNDING(BYTES) (BYTES)
1892: @end example
1893:
1894: @noindent
1895: will suffice. But on other machines, instructions that appear
1896: to push one byte actually push two bytes in an attempt to maintain
1897: alignment. Then the definition should be
1898:
1899: @example
1900: #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
1901: @end example
1902:
1903: @findex ACCUMULATE_OUTGOING_ARGS
1904: @findex current_function_outgoing_args_size
1905: @item ACCUMULATE_OUTGOING_ARGS
1906: If defined, the maximum amount of space required for outgoing arguments
1907: will be computed and placed into the variable
1908: @code{current_function_outgoing_args_size}. No space will be pushed
1909: onto the stack for each call; instead, the function prologue should
1910: increase the stack frame size by this amount.
1911:
1912: It is not proper to define both @code{PUSH_ROUNDING} and
1913: @code{ACCUMULATE_OUTGOING_ARGS}.
1914:
1915: @findex REG_PARM_STACK_SPACE
1916: @item REG_PARM_STACK_SPACE
1917: Define this macro if functions should assume that stack space has been
1918: allocated for arguments even when their values are passed in
1919: registers.
1920:
1921: The value of this macro is the size, in bytes, of the area reserved for
1922: arguments passed in registers.
1923:
1924: This space can either be allocated by the caller or be a part of the
1925: machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE}
1926: says which.
1927:
1928: @findex OUTGOING_REG_PARM_STACK_SPACE
1929: @item OUTGOING_REG_PARM_STACK_SPACE
1930: Define this if it is the responsibility of the caller to allocate the area
1931: reserved for arguments passed in registers.
1932:
1933: If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
1934: whether the space for these arguments counts in the value of
1935: @code{current_function_outgoing_args_size}.
1936:
1937: @findex STACK_PARMS_IN_REG_PARM_AREA
1938: @item STACK_PARMS_IN_REG_PARM_AREA
1939: Define this macro if @code{REG_PARM_STACK_SPACE} is defined but stack
1940: parameters don't skip the area specified by @code{REG_PARM_STACK_SPACE}.
1941:
1942: Normally, when a parameter is not passed in registers, it is placed on the
1943: stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
1944: suppresses this behavior and causes the parameter to be passed on the
1945: stack in its natural location.
1946:
1947: @findex RETURN_POPS_ARGS
1948: @item RETURN_POPS_ARGS (@var{funtype}, @var{stack-size})
1949: A C expression that should indicate the number of bytes of its own
1950: arguments that a function pops on returning, or 0 if the
1951: function pops no arguments and the caller must therefore pop them all
1952: after the function returns.
1953:
1954: @var{funtype} is a C variable whose value is a tree node that
1955: describes the function in question. Normally it is a node of type
1956: @code{FUNCTION_TYPE} that describes the data type of the function.
1957: From this it is possible to obtain the data types of the value and
1958: arguments (if known).
1959:
1960: When a call to a library function is being considered, @var{funtype}
1961: will contain an identifier node for the library function. Thus, if
1962: you need to distinguish among various library functions, you can do so
1963: by their names. Note that ``library function'' in this context means
1964: a function used to perform arithmetic, whose name is known specially
1965: in the compiler and was not mentioned in the C code being compiled.
1966:
1967: @var{stack-size} is the number of bytes of arguments passed on the
1968: stack. If a variable number of bytes is passed, it is zero, and
1969: argument popping will always be the responsibility of the calling function.
1970:
1971: On the Vax, all functions always pop their arguments, so the definition
1972: of this macro is @var{stack-size}. On the 68000, using the standard
1973: calling convention, no functions pop their arguments, so the value of
1974: the macro is always 0 in this case. But an alternative calling
1975: convention is available in which functions that take a fixed number of
1976: arguments pop them but other functions (such as @code{printf}) pop
1977: nothing (the caller pops all). When this convention is in use,
1978: @var{funtype} is examined to determine whether a function takes a fixed
1979: number of arguments.
1980: @end table
1981:
1982: @node Register Arguments
1983: @subsection Passing Arguments in Registers
1984: @cindex arguments in registers
1985: @cindex registers arguments
1986:
1987: This section describes the macros which let you control how various
1988: types of arguments are passed in registers or how they are arranged in
1989: the stack.
1990:
1991: @table @code
1992: @findex FUNCTION_ARG
1993: @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
1994: A C expression that controls whether a function argument is passed
1995: in a register, and which register.
1996:
1997: The arguments are @var{cum}, which summarizes all the previous
1998: arguments; @var{mode}, the machine mode of the argument; @var{type},
1999: the data type of the argument as a tree node or 0 if that is not known
2000: (which happens for C support library functions); and @var{named},
2001: which is 1 for an ordinary argument and 0 for nameless arguments that
2002: correspond to @samp{@dots{}} in the called function's prototype.
2003:
2004: The value of the expression should either be a @code{reg} RTX for the
2005: hard register in which to pass the argument, or zero to pass the
2006: argument on the stack.
2007:
2008: For machines like the Vax and 68000, where normally all arguments are
2009: pushed, zero suffices as a definition.
2010:
2011: @cindex @file{stdarg.h} and register arguments
2012: The usual way to make the ANSI library @file{stdarg.h} work on a machine
2013: where some arguments are usually passed in registers, is to cause
2014: nameless arguments to be passed on the stack instead. This is done
2015: by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
2016:
2017: @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
2018: @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
2019: You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
2020: in the definition of this macro to determine if this argument is of a
2021: type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
2022: is not defined and @code{FUNCTION_ARG} returns non-zero for such an
2023: argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
2024: defined, the argument will be computed in the stack and then loaded into
2025: a register.
2026:
2027: @findex FUNCTION_INCOMING_ARG
2028: @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
2029: Define this macro if the target machine has ``register windows'', so
2030: that the register in which a function sees an arguments is not
2031: necessarily the same as the one in which the caller passed the
2032: argument.
2033:
2034: For such machines, @code{FUNCTION_ARG} computes the register in which
2035: the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
2036: be defined in a similar fashion to tell the function being called
2037: where the arguments will arrive.
2038:
2039: If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
2040: serves both purposes.@refill
2041:
2042: @findex FUNCTION_ARG_PARTIAL_NREGS
2043: @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
2044: A C expression for the number of words, at the beginning of an
2045: argument, must be put in registers. The value must be zero for
2046: arguments that are passed entirely in registers or that are entirely
2047: pushed on the stack.
2048:
2049: On some machines, certain arguments must be passed partially in
2050: registers and partially in memory. On these machines, typically the
2051: first @var{n} words of arguments are passed in registers, and the rest
2052: on the stack. If a multi-word argument (a @code{double} or a
2053: structure) crosses that boundary, its first few words must be passed
2054: in registers and the rest must be pushed. This macro tells the
2055: compiler when this occurs, and how many of the words should go in
2056: registers.
2057:
2058: @code{FUNCTION_ARG} for these arguments should return the first
2059: register to be used by the caller for this argument; likewise
2060: @code{FUNCTION_INCOMING_ARG}, for the called function.
2061:
2062: @findex FUNCTION_ARG_PASS_BY_REFERENCE
2063: @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
2064: A C expression that indicates when an argument must be passed by reference.
2065: If nonzero for an argument, a copy of that argument is made in memory and a
2066: pointer to the argument is passed instead of the argument itself.
2067: The pointer is passed in whatever way is appropriate for passing a pointer
2068: to that type.
2069:
2070: On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
2071: definition of this macro might be
2072: @example
2073: #define FUNCTION_ARG_PASS_BY_REFERENCE(CUM, MODE, TYPE, NAMED) \
2074: MUST_PASS_IN_STACK (MODE, TYPE)
2075: @end example
2076:
2077: @findex CUMULATIVE_ARGS
2078: @item CUMULATIVE_ARGS
2079: A C type for declaring a variable that is used as the first argument of
2080: @code{FUNCTION_ARG} and other related values. For some target machines,
2081: the type @code{int} suffices and can hold the number of bytes of
2082: argument so far.
2083:
2084: There is no need to record in @code{CUMULATIVE_ARGS} anything about the
2085: arguments that have been passed on the stack. The compiler has other
2086: variables to keep track of that. For target machines on which all
2087: arguments are passed on the stack, there is no need to store anything in
2088: @code{CUMULATIVE_ARGS}; however, the data structure must exist and
2089: should not be empty, so use @code{int}.
2090:
2091: @findex INIT_CUMULATIVE_ARGS
2092: @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname})
2093: A C statement (sans semicolon) for initializing the variable @var{cum}
2094: for the state at the beginning of the argument list. The variable has
2095: type @code{CUMULATIVE_ARGS}. The value of @var{fntype} is the tree node
2096: for the data type of the function which will receive the args, or 0
2097: if the args are to a compiler support library function.
2098:
2099: When processing a call to a compiler support library function,
2100: @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
2101: contains the name of the function, as a string. @var{libname} is 0 when
2102: an ordinary C function call is being processed. Thus, each time this
2103: macro is called, either @var{libname} or @var{fntype} is nonzero, but
2104: never both of them at once.
2105:
2106: @findex INIT_CUMULATIVE_INCOMING_ARGS
2107: @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
2108: Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
2109: finding the arguments for the function being compiled. If this macro is
2110: undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
2111:
2112: The argument @var{libname} exists for symmetry with
2113: @code{INIT_CUMULATIVE_ARGS}. The value passed for @var{libname} is
2114: always 0, since library routines with special calling conventions are
2115: never compiled with GNU CC.
2116:
2117: @findex FUNCTION_ARG_ADVANCE
2118: @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
2119: A C statement (sans semicolon) to update the summarizer variable
2120: @var{cum} to advance past an argument in the argument list. The
2121: values @var{mode}, @var{type} and @var{named} describe that argument.
2122: Once this is done, the variable @var{cum} is suitable for analyzing
2123: the @emph{following} argument with @code{FUNCTION_ARG}, etc.@refill
2124:
2125: This macro need not do anything if the argument in question was passed
2126: on the stack. The compiler knows how to track the amount of stack space
2127: used for arguments without any special help.
2128:
2129: @findex FUNCTION_ARG_PADDING
2130: @item FUNCTION_ARG_PADDING (@var{mode}, @var{type})
2131: If defined, a C expression which determines whether, and in which direction,
2132: to pad out an argument with extra space. The value should be of type
2133: @code{enum direction}: either @code{upward} to pad above the argument,
2134: @code{downward} to pad below, or @code{none} to inhibit padding.
2135:
2136: This macro does not control the @emph{amount} of padding; that is
2137: always just enough to reach the next multiple of @code{FUNCTION_ARG_BOUNDARY}.
2138:
2139: This macro has a default definition which is right for most systems.
2140: For little-endian machines, the default is to pad upward. For
2141: big-endian machines, the default is to pad downward for an argument of
2142: constant size shorter than an @code{int}, and upward otherwise.
2143:
2144: @findex FUNCTION_ARG_BOUNDARY
2145: @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
2146: If defined, a C expression that gives the alignment boundary, in bits,
2147: of an argument with the specified mode and type. If it is not defined,
2148: @code{PARM_BOUNDARY} is used for all arguments.
2149:
2150: @findex FUNCTION_ARG_REGNO_P
2151: @item FUNCTION_ARG_REGNO_P (@var{regno})
2152: A C expression that is nonzero if @var{regno} is the number of a hard
2153: register in which function arguments are sometimes passed. This does
2154: @emph{not} include implicit arguments such as the static chain and
2155: the structure-value address. On many machines, no registers can be
2156: used for this purpose since all function arguments are pushed on the
2157: stack.
2158: @end table
2159:
2160: @node Scalar Return
2161: @subsection How Scalar Function Values Are Returned
2162: @cindex return values in registers
2163: @cindex values, returned by functions
2164: @cindex scalars, returned as values
2165:
2166: This section discusses the macros that control returning scalars as
2167: values---values that can fit in registers.
2168:
2169: @table @code
2170: @findex TRADITIONAL_RETURN_FLOAT
2171: @item TRADITIONAL_RETURN_FLOAT
2172: Define this macro if @samp{-traditional} should not cause functions
2173: declared to return @code{float} to convert the value to @code{double}.
2174:
2175: @findex FUNCTION_VALUE
2176: @item FUNCTION_VALUE (@var{valtype}, @var{func})
2177: A C expression to create an RTX representing the place where a
2178: function returns a value of data type @var{valtype}. @var{valtype} is
2179: a tree node representing a data type. Write @code{TYPE_MODE
2180: (@var{valtype})} to get the machine mode used to represent that type.
2181: On many machines, only the mode is relevant. (Actually, on most
2182: machines, scalar values are returned in the same place regardless of
2183: mode).@refill
2184:
2185: If the precise function being called is known, @var{func} is a tree
2186: node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
2187: pointer. This makes it possible to use a different value-returning
2188: convention for specific functions when all their calls are
2189: known.@refill
2190:
2191: @code{FUNCTION_VALUE} is not used for return vales with aggregate data
2192: types, because these are returned in another way. See
2193: @code{STRUCT_VALUE_REGNUM} and related macros, below.
2194:
2195: @findex FUNCTION_OUTGOING_VALUE
2196: @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
2197: Define this macro if the target machine has ``register windows''
2198: so that the register in which a function returns its value is not
2199: the same as the one in which the caller sees the value.
2200:
2201: For such machines, @code{FUNCTION_VALUE} computes the register in
2202: which the caller will see the value, and
2203: @code{FUNCTION_OUTGOING_VALUE} should be defined in a similar fashion
2204: to tell the function where to put the value.@refill
2205:
2206: If @code{FUNCTION_OUTGOING_VALUE} is not defined,
2207: @code{FUNCTION_VALUE} serves both purposes.@refill
2208:
2209: @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
2210: aggregate data types, because these are returned in another way. See
2211: @code{STRUCT_VALUE_REGNUM} and related macros, below.
2212:
2213: @findex LIBCALL_VALUE
2214: @item LIBCALL_VALUE (@var{mode})
2215: A C expression to create an RTX representing the place where a library
2216: function returns a value of mode @var{mode}. If the precise function
2217: being called is known, @var{func} is a tree node
2218: (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
2219: pointer. This makes it possible to use a different value-returning
2220: convention for specific functions when all their calls are
2221: known.@refill
2222:
2223: Note that ``library function'' in this context means a compiler
2224: support routine, used to perform arithmetic, whose name is known
2225: specially by the compiler and was not mentioned in the C code being
2226: compiled.
2227:
2228: The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
2229: data types, because none of the library functions returns such types.
2230:
2231: @findex FUNCTION_VALUE_REGNO_P
2232: @item FUNCTION_VALUE_REGNO_P (@var{regno})
2233: A C expression that is nonzero if @var{regno} is the number of a hard
2234: register in which the values of called function may come back.
2235:
2236: A register whose use for returning values is limited to serving as the
2237: second of a pair (for a value of type @code{double}, say) need not be
2238: recognized by this macro. So for most machines, this definition
2239: suffices:
2240:
2241: @example
2242: #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
2243: @end example
2244:
2245: If the machine has register windows, so that the caller and the called
2246: function use different registers for the return value, this macro
2247: should recognize only the caller's register numbers.
2248: @end table
2249:
2250: @node Aggregate Return
1.1.1.2 ! root 2251: @subsection How Large Values Are Returned
1.1 root 2252: @cindex aggregates as return values
2253: @cindex large return values
2254: @cindex returning aggregate values
2255: @cindex structure value address
2256:
2257: When a function value's mode is @code{BLKmode} (and in some other
2258: cases), the value is not returned according to @code{FUNCTION_VALUE}
2259: (@pxref{Scalar Return}). Instead, the caller passes the address of a
2260: block of memory in which the value should be stored. This address
2261: is called the @dfn{structure value address}.
2262:
2263: This section describes how to control returning structure values in
2264: memory.
2265:
2266: @table @code
2267: @findex RETURN_IN_MEMORY
2268: @item RETURN_IN_MEMORY (@var{type})
2269: A C expression which can inhibit the returning of certain function
2270: values in registers, based on the type of value. A nonzero value says
2271: to return the function value in memory, just as large structures are
2272: always returned. Here @var{type} will be a C expression of type
2273: @code{tree}, representing the data type of the value.
2274:
2275: Note that values of mode @code{BLKmode} are returned in memory
2276: regardless of this macro. Also, the option @samp{-fpcc-struct-return}
2277: takes effect regardless of this macro. On most systems, it is
2278: possible to leave the macro undefined; this causes a default
2279: definition to be used, whose value is the constant 0.
2280:
2281: @findex STRUCT_VALUE_REGNUM
2282: @item STRUCT_VALUE_REGNUM
2283: If the structure value address is passed in a register, then
2284: @code{STRUCT_VALUE_REGNUM} should be the number of that register.
2285:
2286: @findex STRUCT_VALUE
2287: @item STRUCT_VALUE
2288: If the structure value address is not passed in a register, define
2289: @code{STRUCT_VALUE} as an expression returning an RTX for the place
2290: where the address is passed. If it returns 0, the address is passed as
2291: an ``invisible'' first argument.
2292:
2293: @findex STRUCT_VALUE_INCOMING_REGNUM
2294: @item STRUCT_VALUE_INCOMING_REGNUM
2295: On some architectures the place where the structure value address
2296: is found by the called function is not the same place that the
2297: caller put it. This can be due to register windows, or it could
2298: be because the function prologue moves it to a different place.
2299:
2300: If the incoming location of the structure value address is in a
2301: register, define this macro as the register number.
2302:
2303: @findex STRUCT_VALUE_INCOMING
2304: @item STRUCT_VALUE_INCOMING
2305: If the incoming location is not a register, define
2306: @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the
2307: called function should find the value. If it should find the value on
2308: the stack, define this to create a @code{mem} which refers to the frame
2309: pointer. A definition of 0 means that the address is passed as an
2310: ``invisible'' first argument.
2311:
2312: @findex PCC_STATIC_STRUCT_RETURN
2313: @item PCC_STATIC_STRUCT_RETURN
2314: Define this macro if the usual system convention on the target machine
2315: for returning structures and unions is for the called function to return
2316: the address of a static variable containing the value. GNU CC does not
2317: normally use this convention, even if it is the usual one, but does use
2318: it if @samp{-fpcc-struct-value} is specified.
2319:
2320: Do not define this if the usual system convention is for the caller to
2321: pass an address to the subroutine.
2322: @end table
2323:
2324: @node Caller Saves
2325: @subsection Caller-Saves Register Allocation
2326:
2327: If you enable it, GNU CC can save registers around function calls. This
2328: makes it possible to use call-clobbered registers to hold variables that
2329: must live across calls.
2330:
2331: @table @code
2332: @findex DEFAULT_CALLER_SAVES
2333: @item DEFAULT_CALLER_SAVES
2334: Define this macro if function calls on the target machine do not preserve
2335: any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
2336: for all registers. This macro enables @samp{-fcaller-saves} by default.
2337: Eventually that option will be enabled by default on all machines and both
2338: the option and this macro will be eliminated.
2339:
2340: @findex CALLER_SAVE_PROFITABLE
2341: @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
2342: A C expression to determine whether it is worthwhile to consider placing
2343: a pseudo-register in a call-clobbered hard register and saving and
2344: restoring it around each function call. The expression should be 1 when
2345: this is worth doing, and 0 otherwise.
2346:
2347: If you don't define this macro, a default is used which is good on most
2348: machines: @code{4 * @var{calls} < @var{refs}}.
2349: @end table
2350:
2351: @node Function Entry
2352: @subsection Function Entry and Exit
2353: @cindex function entry and exit
2354: @cindex prologue
2355: @cindex epilogue
2356:
2357: This section describes the macros that output function entry
2358: (@dfn{prologue}) and exit (@dfn{epilogue}) code.
2359:
2360: @table @code
2361: @findex FUNCTION_PROLOGUE
2362: @item FUNCTION_PROLOGUE (@var{file}, @var{size})
2363: A C compound statement that outputs the assembler code for entry to a
2364: function. The prologue is responsible for setting up the stack frame,
2365: initializing the frame pointer register, saving registers that must be
2366: saved, and allocating @var{size} additional bytes of storage for the
2367: local variables. @var{size} is an integer. @var{file} is a stdio
2368: stream to which the assembler code should be output.
2369:
2370: The label for the beginning of the function need not be output by this
2371: macro. That has already been done when the macro is run.
2372:
2373: @findex regs_ever_live
2374: To determine which registers to save, the macro can refer to the array
2375: @code{regs_ever_live}: element @var{r} is nonzero if hard register
2376: @var{r} is used anywhere within the function. This implies the function
2377: prologue should save register @var{r}, provided it is not one of the
2378: call-used registers. (@code{FUNCTION_EPILOGUE} must likewise use
2379: @code{regs_ever_live}.)
2380:
2381: On machines that have ``register windows'', the function entry code does
2382: not save on the stack the registers that are in the windows, even if
2383: they are supposed to be preserved by function calls; instead it takes
2384: appropriate steps to ``push'' the register stack, if any non-call-used
2385: registers are used in the function.
2386:
2387: @findex frame_pointer_needed
2388: On machines where functions may or may not have frame-pointers, the
2389: function entry code must vary accordingly; it must set up the frame
2390: pointer if one is wanted, and not otherwise. To determine whether a
2391: frame pointer is in wanted, the macro can refer to the variable
2392: @code{frame_pointer_needed}. The variable's value will be 1 at run
2393: time in a function that needs a frame pointer. @xref{Elimination}.
2394:
2395: The function entry code is responsible for allocating any stack space
2396: required for the function. This stack space consists of the regions
2397: listed below. In most cases, these regions are allocated in the
2398: order listed, with the last listed region closest to the top of the
2399: stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
2400: the highest address if it is not defined). You can use a different order
2401: for a machine if doing so is more convenient or required for
2402: compatibility reasons. Except in cases where required by standard
2403: or by a debugger, there is no reason why the stack layout used by GCC
2404: need agree with that used by other compilers for a machine.
2405:
2406: @itemize @bullet
2407: @item
2408: @findex current_function_pretend_args_size
2409: A region of @code{current_function_pretend_args_size} bytes of
2410: uninitialized space just underneath the first argument arriving on the
2411: stack. (This may not be at the very start of the allocated stack region
2412: if the calling sequence has pushed anything else since pushing the stack
2413: arguments. But usually, on such machines, nothing else has been pushed
2414: yet, because the function prologue itself does all the pushing.) This
2415: region is used on machines where an argument may be passed partly in
2416: registers and partly in memory, and, in some cases to support the
2417: features in @file{varargs.h} and @file{stdargs.h}.
2418:
2419: @item
2420: An area of memory used to save certain registers used by the function.
2421: The size of this area, which may also include space for such things as
2422: the return address and pointers to previous stack frames, is
2423: machine-specific and usually depends on which registers have been used
2424: in the function. Machines with register windows often do not require
2425: a save area.
2426:
2427: @item
2428: A region of at least @var{size} bytes, possibly rounded up to an allocation
2429: boundary, to contain the local variables of the function. On some machines,
2430: this region and the save area may occur in the opposite order, with the
2431: save area closer to the top of the stack.
2432:
2433: @item
2434: @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
2435: Optionally, in the case that @code{ACCUMULATE_OUTGOING_ARGS} is defined,
2436: a region of @code{current_function_outgoing_args_size} bytes to be used
2437: for outgoing argument lists of the function. @xref{Stack Arguments}.
2438: @end itemize
2439:
2440: Normally, it is necessary for @code{FUNCTION_PROLOGUE} and
2441: @code{FUNCTION_EPILOGUE} to treat leaf functions specially. The C
2442: variable @code{leaf_function} is nonzero for such a function.
2443:
2444: @findex EXIT_IGNORE_STACK
2445: @item EXIT_IGNORE_STACK
2446: Define this macro as a C expression that is nonzero if the return
2447: instruction or the function epilogue ignores the value of the stack
2448: pointer; in other words, if it is safe to delete an instruction to
2449: adjust the stack pointer before a return from the function.
2450:
2451: Note that this macro's value is relevant only for functions for which
2452: frame pointers are maintained. It is never safe to delete a final
2453: stack adjustment in a function that has no frame pointer, and the
2454: compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
2455:
2456: @findex FUNCTION_EPILOGUE
2457: @item FUNCTION_EPILOGUE (@var{file}, @var{size})
2458: A C compound statement that outputs the assembler code for exit from a
2459: function. The epilogue is responsible for restoring the saved
2460: registers and stack pointer to their values when the function was
2461: called, and returning control to the caller. This macro takes the
2462: same arguments as the macro @code{FUNCTION_PROLOGUE}, and the
2463: registers to restore are determined from @code{regs_ever_live} and
2464: @code{CALL_USED_REGISTERS} in the same way.
2465:
2466: On some machines, there is a single instruction that does all the work
2467: of returning from the function. On these machines, give that
2468: instruction the name @samp{return} and do not define the macro
2469: @code{FUNCTION_EPILOGUE} at all.
2470:
2471: Do not define a pattern named @samp{return} if you want the
2472: @code{FUNCTION_EPILOGUE} to be used. If you want the target switches
2473: to control whether return instructions or epilogues are used, define a
2474: @samp{return} pattern with a validity condition that tests the target
2475: switches appropriately. If the @samp{return} pattern's validity
2476: condition is false, epilogues will be used.
2477:
2478: On machines where functions may or may not have frame-pointers, the
2479: function exit code must vary accordingly. Sometimes the code for
2480: these two cases is completely different. To determine whether a frame
2481: pointer is in wanted, the macro can refer to the variable
2482: @code{frame_pointer_needed}. The variable's value will be 1 at run
2483: time in a function that needs a frame pointer.
2484:
2485: Normally, it is necessary for @code{FUNCTION_PROLOGUE} and
2486: @code{FUNCTION_EPILOGUE} to treat leaf functions specially. The C
2487: variable @code{leaf_function} is nonzero for such a function.
2488: @xref{Leaf Functions}.
2489:
2490: On some machines, some functions pop their arguments on exit while
2491: others leave that for the caller to do. For example, the 68020 when
2492: given @samp{-mrtd} pops arguments in functions that take a fixed
2493: number of arguments.
2494:
2495: @findex current_function_pops_args
2496: Your definition of the macro @code{RETURN_POPS_ARGS} decides which
2497: functions pop their own arguments. @code{FUNCTION_EPILOGUE} needs to
2498: know what was decided. The variable @code{current_function_pops_args}
2499: is the number of bytes of its arguments that a function should pop.
2500: @xref{Scalar Return}.
2501:
2502: @findex DELAY_SLOTS_FOR_EPILOGUE
2503: @item DELAY_SLOTS_FOR_EPILOGUE
2504: Define this macro if the function epilogue contains delay slots to which
2505: instructions from the rest of the function can be ``moved''. The
2506: definition should be a C expression whose value is an integer
2507: representing the number of delay slots there.
2508:
2509: @findex ELIGIBLE_FOR_EPILOGUE_DELAY
2510: @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
2511: A C expression that returns 1 if @var{insn} can be placed in delay
2512: slot number @var{n} of the epilogue.
2513:
2514: The argument @var{n} is an integer which identifies the delay slot now
2515: being considered (since different slots may have different rules of
2516: eligibility). It is never negative and is always less than the number
2517: of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
2518: If you reject a particular insn for a given delay slot, in principle, it
2519: may be reconsidered for a subsequent delay slot. Also, other insns may
2520: (at least in principle) be considered for the so far unfilled delay
2521: slot.
2522:
2523: @findex current_function_epilogue_delay_list
2524: @findex final_scan_insn
2525: The insns accepted to fill the epilogue delay slots are put in an RTL
2526: list made with @code{insn_list} objects, stored in the variable
2527: @code{current_function_epilogue_delay_list}. The insn for the first
2528: delay slot comes first in the list. Your definition of the macro
2529: @code{FUNCTION_EPILOGUE} should fill the delay slots by outputting the
2530: insns in this list, usually by calling @code{final_scan_insn}.
2531:
2532: You need not define this macro if you did not define
2533: @code{DELAY_SLOTS_FOR_EPILOGUE}.
2534: @end table
2535:
2536: @node Profiling
2537: @subsection Generating Code for Profiling
2538: @cindex profiling, code generation
2539:
2540: @table @code
2541: @findex FUNCTION_PROFILER
2542: @item FUNCTION_PROFILER (@var{file}, @var{labelno})
2543: A C statement or compound statement to output to @var{file} some
2544: assembler code to call the profiling subroutine @code{mcount}.
2545: Before calling, the assembler code must load the address of a
2546: counter variable into a register where @code{mcount} expects to
2547: find the address. The name of this variable is @samp{LP} followed
2548: by the number @var{labelno}, so you would generate the name using
2549: @samp{LP%d} in a @code{fprintf}.
2550:
2551: @findex mcount
2552: The details of how the address should be passed to @code{mcount} are
2553: determined by your operating system environment, not by GNU CC. To
2554: figure them out, compile a small program for profiling using the
2555: system's installed C compiler and look at the assembler code that
2556: results.
2557:
2558: @findex PROFILE_BEFORE_PROLOGUE
2559: @item PROFILE_BEFORE_PROLOGUE
2560: Define this macro if the code for function profiling should come before
2561: the function prologue. Normally, the profiling code comes after.
2562:
2563: @findex FUNCTION_BLOCK_PROFILER
2564: @findex __bb_init_func
2565: @item FUNCTION_BLOCK_PROFILER (@var{file}, @var{labelno})
2566: A C statement or compound statement to output to @var{file} some
2567: assembler code to initialize basic-block profiling for the current
2568: object module. This code should call the subroutine
2569: @code{__bb_init_func} once per object module, passing it as its sole
2570: argument the address of a block allocated in the object module.
2571:
2572: The name of the block is a local symbol made with this statement:
2573:
2574: @example
2575: ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0);
2576: @end example
2577:
2578: Of course, since you are writing the definition of
2579: @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
2580: can take a short cut in the definition of this macro and use the name
2581: that you know will result.
2582:
2583: The first word of this block is a flag which will be nonzero if the
2584: object module has already been initialized. So test this word first,
2585: and do not call @code{__bb_init_func} if the flag is nonzero.
2586:
2587: @findex BLOCK_PROFILER
2588: @item BLOCK_PROFILER (@var{file}, @var{blockno})
2589: A C statement or compound statement to increment the count associated
2590: with the basic block number @var{blockno}. Basic blocks are numbered
2591: separately from zero within each compilation. The count associated
2592: with block number @var{blockno} is at index @var{blockno} in a vector
2593: of words; the name of this array is a local symbol made with this
2594: statement:
2595:
2596: @example
2597: ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 2);
2598: @end example
2599:
2600: Of course, since you are writing the definition of
2601: @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
2602: can take a short cut in the definition of this macro and use the name
2603: that you know will result.
2604: @end table
2605:
1.1.1.2 ! root 2606: @node Varargs
1.1 root 2607: @section Implementing the Varargs Macros
2608: @cindex varargs implementation
2609:
2610: GNU CC comes with an implementation of @file{varargs.h} and
2611: @file{stdarg.h} that work without change on machines that pass arguments
2612: on the stack. Other machines require their own implementations of
2613: varargs, and the two machine independent header files must have
2614: conditionals to include it.
2615:
2616: ANSI @file{stdarg.h} differs from traditional @file{varargs.h} mainly in
2617: the calling convention for @code{va_start}. The traditional
2618: implementation takes just one argument, which is the variable in which
2619: to store the argument pointer. The ANSI implementation takes an
2620: additional first argument, which is the last named argument of the
2621: function. However, it should not use this argument. The way to find
2622: the end of the named arguments is with the built-in functions described
2623: below.
2624:
2625: @table @code
2626: @findex __builtin_saveregs
2627: @item __builtin_saveregs ()
2628: Use this built-in function to save the argument registers in memory so
2629: that the varargs mechanism can access them. Both ANSI and traditional
2630: versions of @code{va_start} must use @code{__builtin_saveregs}, unless
2631: you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
2632:
2633: On some machines, @code{__builtin_saveregs} is open-coded under the
2634: control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
2635: it calls a routine written in assembler language, found in
2636: @file{libgcc2.c}.
2637:
2638: Regardless of what code is generated for the call to
2639: @code{__builtin_saveregs}, it appears at the beginning of the function,
2640: not where the call to @code{__builtin_saveregs} is written. This is
2641: because the registers must be saved before the function starts to use
2642: them for its own purposes.
2643:
2644: @findex __builtin_args_info
2645: @item __builtin_args_info (@var{category})
2646: Use this built-in function to find the first anonymous arguments in
2647: registers.
2648:
2649: In general, a machine may have several categories of registers used for
2650: arguments, each for a particular category of data types. (For example,
2651: on some machines, floating-point registers are used for floating-point
2652: arguments while other arguments are passed in the general registers.)
2653: To make non-varargs functions use the proper calling convention, you
2654: have defined the @code{CUMULATIVE_ARGS} data type to record how many
2655: registers in each category have been used so far
2656:
2657: @code{__builtin_args_info} accesses the same data structure of type
2658: @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
2659: with it, with @var{category} specifying which word to access. Thus, the
2660: value indicates the first unused register in a given category.
2661:
2662: Normally, you would use @code{__builtin_args_info} in the implementation
2663: of @code{va_start}, accessing each category just once and storing the
2664: value in the @code{va_list} object. This is because @code{va_list} will
2665: have to update the values, and there is no way to alter the
2666: values accessed by @code{__builtin_args_info}.
2667:
2668: @findex __builtin_next_arg
2669: @item __builtin_next_arg ()
2670: This is the equivalent of @code{__builtin_args_info}, for stack
2671: arguments. It returns the address of the first anonymous stack
2672: argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
2673: returns the address of the location above the first anonymous stack
2674: argument. Use it in @code{va_start} to initialize the pointer for
2675: fetching arguments from the stack.
2676:
2677: @findex __builtin_classify_type
2678: @item __builtin_classify_type (@var{object})
2679: Since each machine has its own conventions for which data types are
2680: passed in which kind of register, your implementation of @code{va_arg}
2681: has to embody these conventions. The easiest way to categorize the
2682: specified data type is to use @code{__builtin_classify_type} together
2683: with @code{sizeof} and @code{__alignof__}.
2684:
2685: @code{__builtin_classify_type} ignores the value of @var{object},
2686: considering only its data type. It returns an integer describing what
2687: kind of type that is---integer, floating, pointer, structure, and so on.
2688:
2689: The file @file{typeclass.h} defines an enumeration that you can use to
2690: interpret the values of @code{__builtin_classify_type}.
2691: @end table
2692:
2693: These machine description macros help implement varargs:
2694:
2695: @table @code
2696: @findex EXPAND_BUILTIN_SAVEREGS
2697: @item EXPAND_BUILTIN_SAVEREGS (@var{args})
2698: If defined, is a C expression that produces the machine-specific code
2699: for a call to @code{__builtin_saveregs}. This code will be moved to the
2700: very beginning of the function, before any parameter access are made.
2701: The return value of this function should be an RTX that contains the
2702: value to use as the return of @code{__builtin_saveregs}.
2703:
2704: The argument @var{args} is a @code{tree_list} containing the arguments
2705: that were passed to @code{__builtin_saveregs}.
2706:
2707: If this macro is not defined, the compiler will output an ordinary
2708: call to the library function @samp{__builtin_saveregs}.
2709:
2710: @findex SETUP_INCOMING_VARARGS
2711: @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type}, @var{pretend_args_size}, @var{second_time})
2712: This macro offers an alternative to using @code{__builtin_saveregs} and
2713: defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the
2714: anonymous register arguments into the stack so that all the arguments
2715: appear to have been passed consecutively on the stack. Once this is
2716: done, you can use the standard implementation of varargs that works for
2717: machines that pass all their arguments on the stack.
2718:
2719: The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data
2720: structure, containing the values that obtain after processing of the
2721: named arguments. The arguments @var{mode} and @var{type} describe the
2722: last named argument---its machine mode and its data type as a tree node.
2723:
2724: The macro implementation should do two things: first, push onto the
2725: stack all the argument registers @emph{not} used for the named
2726: arguments, and second, store the size of the data thus pushed into the
2727: @code{int}-valued variable whose name is supplied as the argument
2728: @var{pretend_args_size}. The value that you store here will serve as
2729: additional offset for setting up the stack frame.
2730:
2731: Because you must generate code to push the anonymous arguments at
2732: compile time without knowing their data types,
2733: @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just
2734: a single category of argument register and use it uniformly for all data
2735: types.
2736:
2737: If the argument @var{second_time} is nonzero, it means that the
2738: arguments of the function are being analyzed for the second time. This
2739: happens for an inline function, which is not actually compiled until the
2740: end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should
2741: not generate any instructions in this case.
2742: @end table
2743:
1.1.1.2 ! root 2744: @node Trampolines
1.1 root 2745: @section Trampolines for Nested Functions
2746: @cindex trampolines for nested functions
2747: @cindex nested functions, trampolines for
2748:
2749: A @dfn{trampoline} is a small piece of code that is created at run time
2750: when the address of a nested function is taken. It normally resides on
2751: the stack, in the stack frame of the containing function. These macros
2752: tell GNU CC how to generate code to allocate and initialize a
2753: trampoline.
2754:
2755: The instructions in the trampoline must do two things: load a constant
2756: address into the static chain register, and jump to the real address of
2757: the nested function. On CISC machines such as the m68k, this requires
2758: two instructions, a move immediate and a jump. Then the two addresses
2759: exist in the trampoline as word-long immediate operands. On RISC
2760: machines, it is often necessary to load each address into a register in
2761: two parts. Then pieces of each address form separate immediate
2762: operands.
2763:
2764: The code generated to initialize the trampoline must store the variable
2765: parts---the static chain value and the function address---into the
2766: immediate operands of the instructions. On a CISC machine, this is
2767: simply a matter of copying each address to a memory reference at the
2768: proper offset from the start of the trampoline. On a RISC machine, it
2769: may be necessary to take out pieces of the address and store them
2770: separately.
2771:
2772: @table @code
2773: @findex TRAMPOLINE_TEMPLATE
2774: @item TRAMPOLINE_TEMPLATE (@var{file})
2775: A C statement to output, on the stream @var{file}, assembler code for a
2776: block of data that contains the constant parts of a trampoline. This
2777: code should not include a label---the label is taken care of
2778: automatically.
2779:
2780: @findex TRAMPOLINE_SIZE
2781: @item TRAMPOLINE_SIZE
2782: A C expression for the size in bytes of the trampoline, as an integer.
2783:
2784: @findex TRAMPOLINE_ALIGNMENT
2785: @item TRAMPOLINE_ALIGNMENT
2786: Alignment required for trampolines, in bits.
2787:
2788: If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
2789: is used for aligning trampolines.
2790:
2791: @findex INITIALIZE_TRAMPOLINE
2792: @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
2793: A C statement to initialize the variable parts of a trampoline.
2794: @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
2795: an RTX for the address of the nested function; @var{static_chain} is an
2796: RTX for the static chain value that should be passed to the function
2797: when it is called.
2798:
2799: @findex ALLOCATE_TRAMPOLINE
2800: @item ALLOCATE_TRAMPOLINE (@var{fp})
2801: A C expression to allocate run-time space for a trampoline. The
2802: expression value should be an RTX representing a memory reference to the
2803: space for the trampoline.
2804:
2805: @cindex @code{FUNCTION_EPILOGUE} and trampolines
2806: @cindex @code{FUNCTION_PROLOGUE} and trampolines
2807: If this macro is not defined, by default the trampoline is allocated as
2808: a stack slot. This default is right for most machines. The exceptions
2809: are machines where it is impossible to execute instructions in the stack
2810: area. On such machines, you may have to implement a separate stack,
2811: using this macro in conjunction with @code{FUNCTION_PROLOGUE} and
2812: @code{FUNCTION_EPILOGUE}.
2813:
2814: @var{fp} points to a data structure, a @code{struct function}, which
2815: describes the compilation status of the immediate containing function of
2816: the function which the trampoline is for. Normally (when
2817: @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the
2818: trampoline is in the stack frame of this containing function. Other
2819: allocation strategies probably must do something analogous with this
2820: information.
2821: @end table
2822:
2823: Implementing trampolines is difficult on many machines because they have
2824: separate instruction and data caches. Writing into a stack location
2825: fails to clear the memory in the instruction cache, so when the program
2826: jumps to that location, it executes the old contents.
2827:
2828: Here are two possible solutions. One is to clear the relevant parts of
2829: the instruction cache whenever a trampoline is set up. The other is to
2830: make all trampolines identical, by having them jump to a standard
2831: subroutine. The former technique makes trampoline execution faster; the
2832: latter makes initialization faster.
2833:
2834: To clear the instruction cache when a trampoline is initialized, define
2835: the following macros which describe the shape of the cache.
2836:
2837: @table @code
2838: @findex INSN_CACHE_SIZE
2839: @item INSN_CACHE_SIZE
2840: The total size in bytes of the cache.
2841:
2842: @findex INSN_CACHE_LINE_WIDTH
2843: @item INSN_CACHE_LINE_WIDTH
2844: The length in bytes of each cache line. The cache is divided into cache
2845: lines which are disjoint slots, each holding a contiguous chunk of data
2846: fetched from memory. Each time data is brought into the cache, an
2847: entire line is read at once. The data loaded into a cache line is
2848: always aligned on a boundary equal to the line size.
2849:
2850: @findex INSN_CACHE_DEPTH
2851: @item INSN_CACHE_DEPTH
2852: The number of alternative cache lines that can hold any particular memory
2853: location.
2854: @end table
2855:
2856: To use a standard subroutine, define the following macro. In addition,
2857: you must make sure that the instructions in a trampoline fill an entire
2858: cache line with identical instructions, or else ensure that the
2859: beginning of the trampoline code is always aligned at the same point in
2860: its cache line. Look in @file{m68k.h} as a guide.
2861:
2862: @table @code
2863: @findex TRANSFER_FROM_TRAMPOLINE
2864: @item TRANSFER_FROM_TRAMPOLINE
2865: Define this macro if trampolines need a special subroutine to do their
2866: work. The macro should expand to a series of @code{asm} statements
2867: which will be compiled with GNU CC. They go in a library function named
2868: @code{__transfer_from_trampoline}.
2869:
2870: If you need to avoid executing the ordinary prologue code of a compiled
2871: C function when you jump to the subroutine, you can do so by placing a
2872: special label of your own in the assembler code. Use one @code{asm}
2873: statement to generate an assembler label, and another to make the label
2874: global. Then trampolines can use that label to jump directly to your
2875: special assembler code.
2876: @end table
2877:
1.1.1.2 ! root 2878: @node Library Calls
1.1 root 2879: @section Implicit Calls to Library Routines
2880: @cindex library subroutine names
2881: @cindex @file{libgcc.a}
2882:
2883: @table @code
2884: @findex MULSI3_LIBCALL
2885: @item MULSI3_LIBCALL
2886: A C string constant giving the name of the function to call for
2887: multiplication of one signed full-word by another. If you do not
2888: define this macro, the default name is used, which is @code{__mulsi3},
2889: a function defined in @file{libgcc.a}.
2890:
2891: @findex DIVSI3_LIBCALL
2892: @item DIVSI3_LIBCALL
2893: A C string constant giving the name of the function to call for
2894: division of one signed full-word by another. If you do not define
2895: this macro, the default name is used, which is @code{__divsi3}, a
2896: function defined in @file{libgcc.a}.
2897:
2898: @findex UDIVSI3_LIBCALL
2899: @item UDIVSI3_LIBCALL
2900: A C string constant giving the name of the function to call for
2901: division of one unsigned full-word by another. If you do not define
2902: this macro, the default name is used, which is @code{__udivsi3}, a
2903: function defined in @file{libgcc.a}.
2904:
2905: @findex MODSI3_LIBCALL
2906: @item MODSI3_LIBCALL
2907: A C string constant giving the name of the function to call for the
2908: remainder in division of one signed full-word by another. If you do
2909: not define this macro, the default name is used, which is
2910: @code{__modsi3}, a function defined in @file{libgcc.a}.
2911:
2912: @findex UMODSI3_LIBCALL
2913: @item UMODSI3_LIBCALL
2914: A C string constant giving the name of the function to call for the
2915: remainder in division of one unsigned full-word by another. If you do
2916: not define this macro, the default name is used, which is
2917: @code{__umodsi3}, a function defined in @file{libgcc.a}.
2918:
2919: @findex MULDI3_LIBCALL
2920: @item MULDI3_LIBCALL
2921: A C string constant giving the name of the function to call for
2922: multiplication of one signed double-word by another. If you do not
2923: define this macro, the default name is used, which is @code{__muldi3},
2924: a function defined in @file{libgcc.a}.
2925:
2926: @findex DIVDI3_LIBCALL
2927: @item DIVDI3_LIBCALL
2928: A C string constant giving the name of the function to call for
2929: division of one signed double-word by another. If you do not define
2930: this macro, the default name is used, which is @code{__divdi3}, a
2931: function defined in @file{libgcc.a}.
2932:
2933: @findex UDIVDI3_LIBCALL
2934: @item UDIVDI3_LIBCALL
2935: A C string constant giving the name of the function to call for
2936: division of one unsigned full-word by another. If you do not define
2937: this macro, the default name is used, which is @code{__udivdi3}, a
2938: function defined in @file{libgcc.a}.
2939:
2940: @findex MODDI3_LIBCALL
2941: @item MODDI3_LIBCALL
2942: A C string constant giving the name of the function to call for the
2943: remainder in division of one signed double-word by another. If you do
2944: not define this macro, the default name is used, which is
2945: @code{__moddi3}, a function defined in @file{libgcc.a}.
2946:
2947: @findex UMODDI3_LIBCALL
2948: @item UMODDI3_LIBCALL
2949: A C string constant giving the name of the function to call for the
2950: remainder in division of one unsigned full-word by another. If you do
2951: not define this macro, the default name is used, which is
2952: @code{__umoddi3}, a function defined in @file{libgcc.a}.
2953:
2954: @findex TARGET_MEM_FUNCTIONS
2955: @cindex @code{bcopy}, implicit usage
2956: @cindex @code{memcpy}, implicit usage
2957: @cindex @code{bzero}, implicit usage
2958: @cindex @code{memset}, implicit usage
2959: @item TARGET_MEM_FUNCTIONS
2960: Define this macro if GNU CC should generate calls to the System V
2961: (and ANSI C) library functions @code{memcpy} and @code{memset}
2962: rather than the BSD functions @code{bcopy} and @code{bzero}.
2963:
2964: @findex LIBGCC_NEEDS_DOUBLE
2965: @item LIBGCC_NEEDS_DOUBLE
2966: Define this macro if only @code{float} arguments cannot be passed to
2967: library routines (so they must be converted to @code{double}). This
2968: macro affects both how library calls are generated and how the library
2969: routines in @file{libgcc1.c} accept their arguments. It is useful on
2970: machines where floating and fixed point arguments are passed
2971: differently, such as the i860.
2972:
2973: @findex FLOAT_ARG_TYPE
2974: @item FLOAT_ARG_TYPE
2975: Define this macro to override the type used by the library routines to
2976: pick up arguments of type @code{float}. (By default, they use a union
2977: of @code{float} and @code{int}.)
2978:
2979: The obvious choice would be @code{float}---but that won't work with
2980: traditional C compilers that expect all arguments declared as @code{float}
2981: to arrive as @code{double}. To avoid this conversion, the library routines
2982: ask for the value as some other type and then treat it as a @code{float}.
2983:
2984: On some systems, no other type will work for this. For these systems,
2985: you must use @code{LIBGCC_NEEDS_DOUBLE} instead, to force conversion of
2986: the values @code{double} before they are passed.
2987:
2988: @findex FLOATIFY
2989: @item FLOATIFY (@var{passed-value})
2990: Define this macro to override the way library routines redesignate a
2991: @code{float} argument as a @code{float} instead of the type it was
2992: passed as. The default is an expression which takes the @code{float}
2993: field of the union.
2994:
2995: @findex FLOAT_VALUE_TYPE
2996: @item FLOAT_VALUE_TYPE
2997: Define this macro to override the type used by the library routines to
2998: return values that ought to have type @code{float}. (By default, they
2999: use @code{int}.)
3000:
3001: The obvious choice would be @code{float}---but that won't work with
3002: traditional C compilers gratuitously convert values declared as
3003: @code{float} into @code{double}.
3004:
3005: @findex INTIFY
3006: @item INTIFY (@var{float-value})
3007: Define this macro to override the way the value of a
3008: @code{float}-returning library routine should be packaged in order to
3009: return it. These functions are actually declared to return type
3010: @code{FLOAT_VALUE_TYPE} (normally @code{int}).
3011:
3012: These values can't be returned as type @code{float} because traditional
3013: C compilers would gratuitously convert the value to a @code{double}.
3014:
3015: A local variable named @code{intify} is always available when the macro
3016: @code{INTIFY} is used. It is a union of a @code{float} field named
3017: @code{f} and a field named @code{i} whose type is
3018: @code{FLOAT_VALUE_TYPE} or @code{int}.
3019:
3020: If you don't define this macro, the default definition works by copying
3021: the value through that union.
3022:
3023: @findex SItype
3024: @item SItype
3025: Define this macro as the name of the data type corresponding to
3026: @code{SImode} in the system's own C compiler.
3027:
3028: You need not define this macro if that type is @code{int}, as it usually
3029: is.
3030:
3031: @findex perform_@dots{}
3032: @item perform_@dots{}
3033: Define these macros to supply explicit C statements to carry out various
3034: arithmetic operations on types @code{float} and @code{double} in the
3035: library routines in @file{libgcc1.c}. See that file for a full list
3036: of these macros and their arguments.
3037:
3038: On most machines, you don't need to define any of these macros, because
3039: the C compiler that comes with the system takes care of doing them.
3040:
3041: @findex NEXT_OBJC_RUNTIME
3042: @item NEXT_OBJC_RUNTIME
3043: Define this macro to generate code for Objective C message sending using
3044: the calling convention of the NeXT system. This calling convention
3045: involves passing the object, the selector and the method arguments all
3046: at once to the method-lookup library function.
3047:
3048: The default calling convention passes just the object and the selector
3049: to the lookup function, which returns a pointer to the method.
3050: @end table
3051:
1.1.1.2 ! root 3052: @node Addressing Modes
1.1 root 3053: @section Addressing Modes
3054: @cindex addressing modes
3055:
3056: @table @code
3057: @findex HAVE_POST_INCREMENT
3058: @item HAVE_POST_INCREMENT
3059: Define this macro if the machine supports post-increment addressing.
3060:
3061: @findex HAVE_PRE_INCREMENT
3062: @findex HAVE_POST_DECREMENT
3063: @findex HAVE_PRE_DECREMENT
3064: @item HAVE_PRE_INCREMENT
3065: @itemx HAVE_POST_DECREMENT
3066: @itemx HAVE_PRE_DECREMENT
3067: Similar for other kinds of addressing.
3068:
3069: @findex CONSTANT_ADDRESS_P
3070: @item CONSTANT_ADDRESS_P (@var{x})
3071: A C expression that is 1 if the RTX @var{x} is a constant which
3072: is a valid address. On most machines, this can be defined as
3073: @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
3074: in which constant addresses are supported.
3075:
3076: @findex CONSTANT_P
3077: @code{CONSTANT_P} accepts integer-values expressions whose values are
3078: not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and
3079: @code{high} expressions and @code{const} arithmetic expressions, in
3080: addition to @code{const_int} and @code{const_double} expressions.
3081:
3082: @findex MAX_REGS_PER_ADDRESS
3083: @item MAX_REGS_PER_ADDRESS
3084: A number, the maximum number of registers that can appear in a valid
3085: memory address. Note that it is up to you to specify a value equal to
3086: the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
3087: accept.
3088:
3089: @findex GO_IF_LEGITIMATE_ADDRESS
3090: @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
3091: A C compound statement with a conditional @code{goto @var{label};}
3092: executed if @var{x} (an RTX) is a legitimate memory address on the
3093: target machine for a memory operand of mode @var{mode}.
3094:
3095: It usually pays to define several simpler macros to serve as
3096: subroutines for this one. Otherwise it may be too complicated to
3097: understand.
3098:
3099: This macro must exist in two variants: a strict variant and a
3100: non-strict one. The strict variant is used in the reload pass. It
3101: must be defined so that any pseudo-register that has not been
3102: allocated a hard register is considered a memory reference. In
3103: contexts where some kind of register is required, a pseudo-register
3104: with no hard register must be rejected.
3105:
3106: The non-strict variant is used in other passes. It must be defined to
3107: accept all pseudo-registers in every context where some kind of
3108: register is required.
3109:
3110: @findex REG_OK_STRICT
3111: Compiler source files that want to use the strict variant of this
3112: macro define the macro @code{REG_OK_STRICT}. You should use an
3113: @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
3114: in that case and the non-strict variant otherwise.
3115:
3116: Typically among the subroutines used to define
3117: @code{GO_IF_LEGITIMATE_ADDRESS} are subroutines to check for
3118: acceptable registers for various purposes (one for base registers, one
3119: for index registers, and so on). Then only these subroutine macros
3120: need have two variants; the higher levels of macros may be the same
3121: whether strict or not.@refill
3122:
3123: Normally, constant addresses which are the sum of a @code{symbol_ref}
3124: and an integer are stored inside a @code{const} RTX to mark them as
3125: constant. Therefore, there is no need to recognize such sums
3126: specifically as legitimate addresses. Normally you would simply
3127: recognize any @code{const} as legitimate.
3128:
3129: Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
3130: sums that are not marked with @code{const}. It assumes that a naked
3131: @code{plus} indicates indexing. If so, then you @emph{must} reject such
3132: naked constant sums as illegitimate addresses, so that none of them will
3133: be given to @code{PRINT_OPERAND_ADDRESS}.
3134:
3135: @cindex @code{ENCODE_SECTION_INFO} and address validation
3136: On some machines, whether a symbolic address is legitimate depends on
3137: the section that the address refers to. On these machines, define the
3138: macro @code{ENCODE_SECTION_INFO} to store the information into the
3139: @code{symbol_ref}, and then check for it here. When you see a
3140: @code{const}, you will have to look inside it to find the
3141: @code{symbol_ref} in order to determine the section. @xref{Assembler
3142: Format}.
3143:
3144: @findex saveable_obstack
3145: The best way to modify the name string is by adding text to the
3146: beginning, with suitable punctuation to prevent any ambiguity. Allocate
3147: the new name in @code{saveable_obstack}. You will have to modify
3148: @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and
3149: output the name accordingly.
3150:
3151: You can check the information stored here into the @code{symbol_ref} in
3152: the definitions of @code{GO_IF_LEGITIMATE_ADDRESS} and
3153: @code{PRINT_OPERAND_ADDRESS}.
3154:
3155: @findex REG_OK_FOR_BASE_P
3156: @item REG_OK_FOR_BASE_P (@var{x})
3157: A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
3158: RTX) is valid for use as a base register. For hard registers, it
3159: should always accept those which the hardware permits and reject the
3160: others. Whether the macro accepts or rejects pseudo registers must be
3161: controlled by @code{REG_OK_STRICT} as described above. This usually
3162: requires two variant definitions, of which @code{REG_OK_STRICT}
3163: controls the one actually used.
3164:
3165: @findex REG_OK_FOR_INDEX_P
3166: @item REG_OK_FOR_INDEX_P (@var{x})
3167: A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
3168: RTX) is valid for use as an index register.
3169:
3170: The difference between an index register and a base register is that
3171: the index register may be scaled. If an address involves the sum of
3172: two registers, neither one of them scaled, then either one may be
3173: labeled the ``base'' and the other the ``index''; but whichever
3174: labeling is used must fit the machine's constraints of which registers
3175: may serve in each capacity. The compiler will try both labelings,
3176: looking for one that is valid, and will reload one or both registers
3177: only if neither labeling works.
3178:
3179: @findex LEGITIMIZE_ADDRESS
3180: @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
3181: A C compound statement that attempts to replace @var{x} with a valid
3182: memory address for an operand of mode @var{mode}. @var{win} will be a
3183: C statement label elsewhere in the code; the macro definition may use
3184:
3185: @example
3186: GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
3187: @end example
3188:
3189: @noindent
3190: to avoid further processing if the address has become legitimate.
3191:
3192: @findex break_out_memory_refs
3193: @var{x} will always be the result of a call to @code{break_out_memory_refs},
3194: and @var{oldx} will be the operand that was given to that function to produce
3195: @var{x}.
3196:
3197: The code generated by this macro should not alter the substructure of
3198: @var{x}. If it transforms @var{x} into a more legitimate form, it
3199: should assign @var{x} (which will always be a C variable) a new value.
3200:
3201: It is not necessary for this macro to come up with a legitimate
3202: address. The compiler has standard ways of doing so in all cases. In
3203: fact, it is safe for this macro to do nothing. But often a
3204: machine-dependent strategy can generate better code.
3205:
3206: @findex GO_IF_MODE_DEPENDENT_ADDRESS
3207: @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
3208: A C statement or compound statement with a conditional @code{goto
3209: @var{label};} executed if memory address @var{x} (an RTX) can have
3210: different meanings depending on the machine mode of the memory
3211: reference it is used for.
3212:
3213: Autoincrement and autodecrement addresses typically have mode-dependent
3214: effects because the amount of the increment or decrement is the size
3215: of the operand being addressed. Some machines have other mode-dependent
3216: addresses. Many RISC machines have no mode-dependent addresses.
3217:
3218: You may assume that @var{addr} is a valid address for the machine.
3219:
3220: @findex LEGITIMATE_CONSTANT_P
3221: @item LEGITIMATE_CONSTANT_P (@var{x})
3222: A C expression that is nonzero if @var{x} is a legitimate constant for
3223: an immediate operand on the target machine. You can assume that
3224: @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
3225: @samp{1} is a suitable definition for this macro on machines where
3226: anything @code{CONSTANT_P} is valid.@refill
3227:
3228: @findex LEGITIMATE_PIC_OPERAND_P
3229: @item LEGITIMATE_PIC_OPERAND_P (@var{x})
3230: A C expression that is nonzero if @var{x} is a legitimate immediate
3231: operand on the target machine when generating position independent code.
3232: You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
3233: check this. You can also assume @var{flag_pic} is true, so you need not
3234: check it either. You need not define this macro if all constants
3235: (including @code{SYMBOL_REF}) can be immediate operands when generating
3236: position independent code.
3237: @end table
3238:
1.1.1.2 ! root 3239: @node Condition Code
1.1 root 3240: @section Condition Code Status
3241: @cindex condition code status
3242:
3243: @findex cc_status
3244: The file @file{conditions.h} defines a variable @code{cc_status} to
3245: describe how the condition code was computed (in case the interpretation of
3246: the condition code depends on the instruction that it was set by). This
3247: variable contains the RTL expressions on which the condition code is
3248: currently based, and several standard flags.
3249:
3250: Sometimes additional machine-specific flags must be defined in the machine
3251: description header file. It can also add additional machine-specific
3252: information by defining @code{CC_STATUS_MDEP}.
3253:
3254: @table @code
3255: @findex CC_STATUS_MDEP
3256: @item CC_STATUS_MDEP
3257: C code for a data type which is used for declaring the @code{mdep}
3258: component of @code{cc_status}. It defaults to @code{int}.
3259:
3260: This macro is not used on machines that do not use @code{cc0}.
3261:
3262: @findex CC_STATUS_MDEP_INIT
3263: @item CC_STATUS_MDEP_INIT
3264: A C expression to initialize the @code{mdep} field to ``empty''.
3265: The default definition does nothing, since most machines don't use
3266: the field anyway. If you want to use the field, you should probably
3267: define this macro to initialize it.
3268:
3269: This macro is not used on machines that do not use @code{cc0}.
3270:
3271: @findex NOTICE_UPDATE_CC
3272: @item NOTICE_UPDATE_CC (@var{exp}, @var{insn})
3273: A C compound statement to set the components of @code{cc_status}
3274: appropriately for an insn @var{insn} whose body is @var{exp}. It is
3275: this macro's responsibility to recognize insns that set the condition
3276: code as a byproduct of other activity as well as those that explicitly
3277: set @code{(cc0)}.
3278:
3279: This macro is not used on machines that do not use @code{cc0}.
3280:
3281: If there are insns that do not set the condition code but do alter
3282: other machine registers, this macro must check to see whether they
3283: invalidate the expressions that the condition code is recorded as
3284: reflecting. For example, on the 68000, insns that store in address
3285: registers do not set the condition code, which means that usually
3286: @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
3287: insns. But suppose that the previous insn set the condition code
3288: based on location @samp{a4@@(102)} and the current insn stores a new
3289: value in @samp{a4}. Although the condition code is not changed by
3290: this, it will no longer be true that it reflects the contents of
3291: @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
3292: @code{cc_status} in this case to say that nothing is known about the
3293: condition code value.
3294:
3295: The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
3296: with the results of peephole optimization: insns whose patterns are
3297: @code{parallel} RTXs containing various @code{reg}, @code{mem} or
3298: constants which are just the operands. The RTL structure of these
3299: insns is not sufficient to indicate what the insns actually do. What
3300: @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
3301: @code{CC_STATUS_INIT}.
3302:
3303: A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
3304: that looks at an attribute (@pxref{Insn Attributes}) named, for example,
3305: @samp{cc}. This avoids having detailed information about patterns in
3306: two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
3307:
3308: @findex EXTRA_CC_MODES
3309: @item EXTRA_CC_MODES
3310: A list of names to be used for additional modes for condition code
3311: values in registers (@pxref{Jump Patterns}). These names are added
3312: to @code{enum machine_mode} and all have class @code{MODE_CC}. By
3313: convention, they should start with @samp{CC} and end with @samp{mode}.
3314:
3315: You should only define this macro if your machine does not use @code{cc0}
3316: and only if additional modes are required.
3317:
3318: @findex EXTRA_CC_NAMES
3319: @item EXTRA_CC_NAMES
3320: A list of C strings giving the names for the modes listed in
3321: @code{EXTRA_CC_MODES}. For example, the Sparc defines this macro and
3322: @code{EXTRA_CC_MODES} as
3323:
3324: @example
3325: #define EXTRA_CC_MODES CC_NOOVmode, CCFPmode
3326: #define EXTRA_CC_NAMES "CC_NOOV", "CCFP"
3327: @end example
3328:
3329: This macro is not required if @code{EXTRA_CC_MODES} is not defined.
3330:
3331: @findex SELECT_CC_MODE
3332: @item SELECT_CC_MODE (@var{op}, @var{x})
3333: Returns a mode from class @code{MODE_CC} to be used when comparison operation
3334: code @var{op} is applied to rtx @var{x}. For example, on the Sparc,
3335: @code{SELECT_CC_MODE} is defined as (see @pxref{Jump Patterns} for a
3336: description of the reason for this definition)
3337:
3338: @example
3339: #define SELECT_CC_MODE(OP,X) \
3340: (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT ? CCFPmode \
3341: : (GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
3342: || GET_CODE (X) == NEG) \
3343: ? CC_NOOVmode : CCmode)
3344: @end example
3345:
3346: This macro is not required if @code{EXTRA_CC_MODES} is not defined.
3347: @end table
3348:
1.1.1.2 ! root 3349: @node Costs
1.1 root 3350: @section Describing Relative Costs of Operations
3351: @cindex costs of instructions
3352: @cindex relative costs
3353: @cindex speed of instructions
3354:
3355: These macros let you describe the relative speed of various operations
3356: on the target machine.
3357:
3358: @table @code
3359: @findex CONST_COSTS
3360: @item CONST_COSTS (@var{x}, @var{code})
3361: A part of a C @code{switch} statement that describes the relative costs
3362: of constant RTL expressions. It must contain @code{case} labels for
3363: expression codes @code{const_int}, @code{const}, @code{symbol_ref},
3364: @code{label_ref} and @code{const_double}. Each case must ultimately
3365: reach a @code{return} statement to return the relative cost of the use
3366: of that kind of constant value in an expression. The cost may depend on
3367: the precise value of the constant, which is available for examination in
3368: @var{x}.
3369:
3370: @var{code} is the expression code---redundant, since it can be
3371: obtained with @code{GET_CODE (@var{x})}.
3372:
3373: @findex RTX_COSTS
3374: @findex COSTS_N_INSNS
3375: @item RTX_COSTS (@var{x}, @var{code})
3376: Like @code{CONST_COSTS} but applies to nonconstant RTL expressions.
3377: This can be used, for example, to indicate how costly a multiply
3378: instruction is. In writing this macro, you can use the construct
3379: @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
3380: instructions.
3381:
3382: This macro is optional; do not define it if the default cost assumptions
3383: are adequate for the target machine.
3384:
3385: @findex ADDRESS_COST
3386: @item ADDRESS_COST (@var{address})
3387: An expression giving the cost of an addressing mode that contains
3388: @var{address}. If not defined, the cost is computed from
3389: the @var{address} expression and the @code{CONST_COSTS} values.
3390:
3391: For most CISC machines, the default cost is a good approximation of the
3392: true cost of the addressing mode. However, on RISC machines, all
3393: instructions normally have the same length and execution time. Hence
3394: all addresses will have equal costs.
3395:
3396: In cases where more than one form of an address is known, the form with
3397: the lowest cost will be used. If multiple forms have the same, lowest,
3398: cost, the one that is the most complex will be used.
3399:
3400: For example, suppose an address that is equal to the sum of a register
3401: and a constant is used twice in the same basic block. When this macro
3402: is not defined, the address will be computed in a register and memory
3403: references will be indirect through that register. On machines where
3404: the cost of the addressing mode containing the sum is no higher than
3405: that of a simple indirect reference, this will produce an additional
3406: instruction and possibly require an additional register. Proper
3407: specification of this macro eliminates this overhead for such machines.
3408:
3409: Similar use of this macro is made in strength reduction of loops.
3410:
3411: @var{address} need not be valid as an address. In such a case, the cost
3412: is not relevant and can be any value; invalid addresses need not be
3413: assigned a different cost.
3414:
3415: On machines where an address involving more than one register is as
3416: cheap as an address computation involving only one register, defining
3417: @code{ADDRESS_COST} to reflect this can cause two registers to be live
3418: over a region of code where only one would have been if
3419: @code{ADDRESS_COST} were not defined in that manner. This effect should
3420: be considered in the definition of this macro. Equivalent costs should
3421: probably only be given to addresses with different numbers of registers
3422: on machines with lots of registers.
3423:
3424: This macro will normally either not be defined or be defined as a
3425: constant.
3426:
3427: @findex REGISTER_MOVE_COST
3428: @item REGISTER_MOVE_COST (@var{from}, @var{to})
3429: A C expression for the cost of moving data from a register in class
3430: @var{from} to one in class @var{to}. The classes are expressed using
3431: the enumeration values such as @code{GENERAL_REGS}. A value of 2 is the
3432: default; other values are interpreted relative to that.
3433:
3434: It is not required that the cost always equal 2 when @var{from} is the
3435: same as @var{to}; on some machines it is expensive to move between
3436: registers if they are not general registers.
3437:
3438: If reload sees an insn consisting of a single @code{set} between two
3439: hard registers, and if @code{REGISTER_MOVE_COST} applied to their
3440: classes returns a value of 2, reload does not check to ensure that the
3441: constraints of the insn are met. Setting a cost of other than 2 will
3442: allow reload to verify that the constraints are met. You should do this
3443: if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
3444:
3445: @findex MEMORY_MOVE_COST
3446: @item MEMORY_MOVE_COST (@var{m})
3447: A C expression for the cost of moving data of mode @var{m} between a
3448: register and memory. A value of 2 is the default; this cost is relative
3449: to those in @code{REGISTER_MOVE_COST}.
3450:
3451: If moving between registers and memory is more expensive than between
3452: two registers, you should define this macro to express the relative cost.
3453:
3454: @findex BRANCH_COST
3455: @item BRANCH_COST
3456: A C expression for the cost of a branch instruction. A value of 1 is
3457: the default; other values are interpreted relative to that.
3458: @end table
3459:
3460: Here are additional macros which do not specify precise relative costs,
3461: but only that certain actions are more expensive than GNU CC would
3462: ordinarily expect.
3463:
3464: @table @code
3465: @findex SLOW_BYTE_ACCESS
3466: @item SLOW_BYTE_ACCESS
3467: Define this macro as a C expression which is nonzero if accessing less
3468: than a word of memory (i.e. a @code{char} or a @code{short}) is no
3469: faster than accessing a word of memory, i.e., if such access
3470: require more than one instruction or if there is no difference in cost
3471: between byte and (aligned) word loads.
3472:
3473: When this macro is not defined, the compiler will access a field by
3474: finding the smallest containing object; when it is defined, a fullword
3475: load will be used if alignment permits. Unless bytes accesses are
3476: faster than word accesses, using word accesses is preferable since it
3477: may eliminate subsequent memory access if subsequent accesses occur to
3478: other fields in the same word of the structure, but to different bytes.
3479:
3480: @findex SLOW_ZERO_EXTEND
3481: @item SLOW_ZERO_EXTEND
3482: Define this macro if zero-extension (of a @code{char} or @code{short}
3483: to an @code{int}) can be done faster if the destination is a register
3484: that is known to be zero.
3485:
3486: If you define this macro, you must have instruction patterns that
3487: recognize RTL structures like this:
3488:
3489: @example
3490: (set (strict_low_part (subreg:QI (reg:SI @dots{}) 0)) @dots{})
3491: @end example
3492:
3493: @noindent
3494: and likewise for @code{HImode}.
3495:
3496: @findex SLOW_UNALIGNED_ACCESS
3497: @item SLOW_UNALIGNED_ACCESS
1.1.1.2 ! root 3498: Define this macro to be the value 1 if unaligned accesses have a cost
! 3499: many times greater than aligned accesses, for example if they are
! 3500: emulated in a trap handler.
1.1 root 3501:
1.1.1.2 ! root 3502: When this macro is non-zero, the compiler will act as if
! 3503: @code{STRICT_ALIGNMENT} were non-zero when generating code for block
1.1 root 3504: moves. This can cause significantly more instructions to be produced.
1.1.1.2 ! root 3505: Therefore, do not set this macro non-zero if unaligned accesses only add a
1.1 root 3506: cycle or two to the time for a memory access.
3507:
1.1.1.2 ! root 3508: If the value of this macro is always zero, it need not be defined.
! 3509:
1.1 root 3510: @findex DONT_REDUCE_ADDR
3511: @item DONT_REDUCE_ADDR
3512: Define this macro to inhibit strength reduction of memory addresses.
3513: (On some machines, such strength reduction seems to do harm rather
3514: than good.)
3515:
3516: @findex MOVE_RATIO
3517: @item MOVE_RATIO
3518: The number of scalar move insns which should be generated instead of a
3519: string move insn or a library call. Increasing the value will always
3520: make code faster, but eventually incurs high cost in increased code size.
3521:
3522: If you don't define this, a reasonable default is used.
3523:
3524: @findex NO_FUNCTION_CSE
3525: @item NO_FUNCTION_CSE
3526: Define this macro if it is as good or better to call a constant
3527: function address than to call an address kept in a register.
3528:
3529: @findex NO_RECURSIVE_FUNCTION_CSE
3530: @item NO_RECURSIVE_FUNCTION_CSE
3531: Define this macro if it is as good or better for a function to call
3532: itself with an explicit address than to call an address kept in a
3533: register.
3534: @end table
3535:
1.1.1.2 ! root 3536: @node Sections
1.1 root 3537: @section Dividing the Output into Sections (Texts, Data, @dots{})
3538:
3539: An object file is divided into sections containing different types of
3540: data. In the most common case, there are three sections: the @dfn{text
3541: section}, which holds instructions and read-only data; the @dfn{data
3542: section}, which holds initialized writable data; and the @dfn{bss
3543: section}, which holds uninitialized data. Some systems have other kinds
3544: of sections.
3545:
3546: The compiler must tell the assembler when to switch sections. These
3547: macros control what commands to output to tell the assembler this. You
3548: can also define additional sections.
3549:
3550: @table @code
3551: @findex TEXT_SECTION_ASM_OP
3552: @item TEXT_SECTION_ASM_OP
3553: A C string constant for the assembler operation that should precede
3554: instructions and read-only data. Normally @code{".text"} is right.
3555:
3556: @findex DATA_SECTION_ASM_OP
3557: @item DATA_SECTION_ASM_OP
3558: A C string constant for the assembler operation to identify the
3559: following data as writable initialized data. Normally @code{".data"}
3560: is right.
3561:
3562: @findex SHARED_SECTION_ASM_OP
3563: @item SHARED_SECTION_ASM_OP
3564: If defined, a C string constant for the assembler operation to identify the
3565: following data as shared data. If not defined, @code{DATA_SECTION_ASM_OP}
3566: will be used.
3567:
3568: @findex INIT_SECTION_ASM_OP
3569: @item INIT_SECTION_ASM_OP
3570: If defined, a C string constant for the assembler operation to identify the
3571: following data as initialization code. If not defined, GNU CC will
3572: assume such a section does not exist.
3573:
3574: @findex EXTRA_SECTIONS
3575: @findex in_text
3576: @findex in_data
3577: @item EXTRA_SECTIONS
3578: A list of names for sections other than the standard two, which are
3579: @code{in_text} and @code{in_data}. You need not define this macro
3580: on a system with no other sections (that GCC needs to use).
3581:
3582: @findex EXTRA_SECTION_FUNCTIONS
3583: @findex text_section
3584: @findex data_section
3585: @item EXTRA_SECTION_FUNCTIONS
3586: One or more functions to be defined in @file{varasm.c}. These
3587: functions should do jobs analogous to those of @code{text_section} and
3588: @code{data_section}, for your additional sections. Do not define this
3589: macro if you do not define @code{EXTRA_SECTIONS}.
3590:
3591: @findex READONLY_DATA_SECTION
3592: @item READONLY_DATA_SECTION
3593: On most machines, read-only variables, constants, and jump tables are
3594: placed in the text section. If this is not the case on your machine,
3595: this macro should be defined to be the name of a function (either
3596: @code{data_section} or a function defined in @code{EXTRA_SECTIONS}) that
3597: switches to the section to be used for read-only items.
3598:
3599: If these items should be placed in the text section, this macro should
3600: not be defined.
3601:
3602: @findex SELECT_SECTION
3603: @item SELECT_SECTION (@var{exp}, @var{reloc})
3604: A C statement or statements to switch to the appropriate section for
3605: output of @var{exp}. You can assume that @var{exp} is either a
3606: @code{VAR_DECL} node or a constant of some sort. @var{reloc}
3607: indicates whether the initial value of @var{exp} requires link-time
3608: relocations. Select the section by calling @code{text_section} or one
3609: of the alternatives for other sections.
3610:
3611: Do not define this macro if you put all read-only variables and
3612: constants in the read-only data section (usually the text section).
3613:
3614: @findex SELECT_RTX_SECTION
3615: @item SELECT_RTX_SECTION (@var{mode}, @var{rtx})
3616: A C statement or statements to switch to the appropriate section for
3617: output of @var{rtx} in mode @var{mode}. You can assume that @var{rtx}
3618: is some kind of constant in RTL. The argument @var{mode} is redundant
3619: except in the case of a @code{const_int} rtx. Select the section by
3620: calling @code{text_section} or one of the alternatives for other
3621: sections.
3622:
3623: Do not define this macro if you put all constants in the read-only
3624: data section.
3625:
3626: @findex JUMP_TABLES_IN_TEXT_SECTION
3627: @item JUMP_TABLES_IN_TEXT_SECTION
3628: Define this macro if jump tables (for @code{tablejump} insns) should be
3629: output in the text section, along with the assembler instructions.
3630: Otherwise, the readonly data section is used.
3631:
3632: This macro is irrelevant if there is no separate readonly data section.
3633:
3634: @findex ENCODE_SECTION_INFO
3635: @item ENCODE_SECTION_INFO (@var{decl})
3636: Define this macro if references to a symbol must be treated differently
3637: depending on something about the variable or function named by the
3638: symbol (such as what section it is in).
3639:
3640: The macro definition, if any, is executed immediately after the rtl for
3641: @var{decl} has been created and stored in @code{DECL_RTL (@var{decl})}.
3642: The value of the rtl will be a @code{mem} whose address is a
3643: @code{symbol_ref}.
3644:
3645: @cindex @code{SYMBOL_REF_FLAG}, in @code{ENCODE_SECTION_INFO}
3646: The usual thing for this macro to do is to record a flag in the
3647: @code{symbol_ref} (such as @code{SYMBOL_REF_FLAG}) or to store a
3648: modified name string in the @code{symbol_ref} (if one bit is not enough
3649: information).
3650: @end table
3651:
1.1.1.2 ! root 3652: @node PIC
1.1 root 3653: @section Position Independent Code
3654: @cindex position independent code
3655: @cindex PIC
3656:
3657: This section describes macros that help implement generation of position
3658: independent code. Simply defining these macros is not enough to
3659: generate valid PIC; you must also add support to the macros
3660: @code{GO_IF_LEGITIMATE_ADDRESS} and @code{LEGITIMIZE_ADDRESS}, and
3661: @code{PRINT_OPERAND_ADDRESS} as well. You must modify the definition of
3662: @samp{movsi} to do something appropriate when the source operand
3663: contains a symbolic address. You may also need to alter the handling of
3664: switch statements so that they use relative addresses.
3665:
3666: @table @code
3667: @findex PIC_OFFSET_TABLE_REGNUM
3668: @item PIC_OFFSET_TABLE_REGNUM
3669: The register number of the register used to address a table of static
3670: data addresses in memory. In some cases this register is defined by a
3671: processor's ``application binary interface'' (ABI). When this macro
3672: is defined, RTL is generated for this register once, as with the stack
3673: pointer and frame pointer registers. If this macro is not defined, it
3674: is up to the machine-dependent files to allocate such a register (if
3675: necessary).
3676:
3677: @findex FINALIZE_PIC
3678: @item FINALIZE_PIC
3679: By generating position-independent code, when two different programs (A
3680: and B) share a common library (libC.a), the text of the library can be
3681: shared whether or not the library is linked at the same address for both
3682: programs. In some of these environments, position-independent code
3683: requires not only the use of different addressing modes, but also
3684: special code to enable the use of these addressing modes.
3685:
3686: The @code{FINALIZE_PIC} macro serves as a hook to emit these special
3687: codes once the function is being compiled into assembly code, but not
3688: before. (It is not done before, because in the case of compiling an
3689: inline function, it would lead to multiple PIC prologues being
3690: included in functions which used inline functions and were compiled to
3691: assembly language.)
3692:
3693: @end table
3694:
1.1.1.2 ! root 3695: @node Assembler Format
1.1 root 3696: @section Defining the Output Assembler Language
3697:
3698: This section describes macros whose principal purpose is to describe how
3699: to write instructions in assembler language--rather than what the
3700: instructions do.
3701:
3702: @menu
3703: * File Framework:: Structural information for the assembler file.
3704: * Data Output:: Output of constants (numbers, strings, addresses).
3705: * Uninitialized Data:: Output of uninitialized variables.
3706: * Label Output:: Output and generation of labels.
3707: * Constructor Output:: Output of initialization and termination routines.
3708: * Instruction Output:: Output of actual instructions.
3709: * Dispatch Tables:: Output of jump tables.
3710: * Alignment Output:: Pseudo ops for alignment and skipping data.
3711: @end menu
3712:
1.1.1.2 ! root 3713: @node File Framework
1.1 root 3714: @subsection The Overall Framework of an Assembler File
3715: @cindex assembler format
3716: @cindex output of assembler code
3717:
3718: @table @code
3719: @findex ASM_FILE_START
3720: @item ASM_FILE_START (@var{stream})
3721: A C expression which outputs to the stdio stream @var{stream}
3722: some appropriate text to go at the start of an assembler file.
3723:
3724: Normally this macro is defined to output a line containing
3725: @samp{#NO_APP}, which is a comment that has no effect on most
3726: assemblers but tells the GNU assembler that it can save time by not
3727: checking for certain assembler constructs.
3728:
3729: On systems that use SDB, it is necessary to output certain commands;
3730: see @file{attasm.h}.
3731:
3732: @findex ASM_FILE_END
3733: @item ASM_FILE_END (@var{stream})
3734: A C expression which outputs to the stdio stream @var{stream}
3735: some appropriate text to go at the end of an assembler file.
3736:
3737: If this macro is not defined, the default is to output nothing
3738: special at the end of the file. Most systems don't require any
3739: definition.
3740:
3741: On systems that use SDB, it is necessary to output certain commands;
3742: see @file{attasm.h}.
3743:
3744: @findex ASM_IDENTIFY_GCC
3745: @item ASM_IDENTIFY_GCC (@var{file})
3746: A C statement to output assembler commands which will identify
3747: the object file as having been compiled with GNU CC (or another
3748: GNU compiler).
3749:
3750: If you don't define this macro, the string @samp{gcc_compiled.:}
3751: is output. This string is calculated to define a symbol which,
3752: on BSD systems, will never be defined for any other reason.
3753: GDB checks for the presence of this symbol when reading the
3754: symbol table of an executable.
3755:
3756: On non-BSD systems, you must arrange communication with GDB in
3757: some other fashion. If GDB is not used on your system, you can
3758: define this macro with an empty body.
3759:
3760: @findex ASM_COMMENT_START
3761: @item ASM_COMMENT_START
3762: A C string constant describing how to begin a comment in the target
3763: assembler language. The compiler assumes that the comment will end at
3764: the end of the line.
3765:
3766: @findex ASM_APP_ON
3767: @item ASM_APP_ON
3768: A C string constant for text to be output before each @code{asm}
3769: statement or group of consecutive ones. Normally this is
3770: @code{"#APP"}, which is a comment that has no effect on most
3771: assemblers but tells the GNU assembler that it must check the lines
3772: that follow for all valid assembler constructs.
3773:
3774: @findex ASM_APP_OFF
3775: @item ASM_APP_OFF
3776: A C string constant for text to be output after each @code{asm}
3777: statement or group of consecutive ones. Normally this is
3778: @code{"#NO_APP"}, which tells the GNU assembler to resume making the
3779: time-saving assumptions that are valid for ordinary compiler output.
3780:
3781: @findex ASM_OUTPUT_SOURCE_FILENAME
3782: @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
3783: A C statement to output COFF information or DWARF debugging information
3784: which indicates that filename @var{name} is the current source file to
3785: the stdio stream @var{stream}.
3786:
3787: This macro need not be defined if the standard form of output
3788: for the file format in use is appropriate.
3789:
3790: @findex ASM_OUTPUT_SOURCE_LINE
3791: @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
3792: A C statement to output DBX or SDB debugging information before code
3793: for line number @var{line} of the current source file to the
3794: stdio stream @var{stream}.
3795:
3796: This macro need not be defined if the standard form of debugging
3797: information for the debugger in use is appropriate.
3798:
3799: @findex ASM_OUTPUT_IDENT
3800: @item ASM_OUTPUT_IDENT (@var{stream}, @var{string})
3801: A C statement to output something to the assembler file to handle a
3802: @samp{#ident} directive containing the text @var{string}. If this
3803: macro is not defined, nothing is output for a @samp{#ident} directive.
3804:
3805: @findex OBJC_PROLOGUE
3806: @item OBJC_PROLOGUE
3807: A C statement to output any assembler statements which are required to
3808: precede any Objective C object definitions or message sending. The
3809: statement is executed only when compiling an Objective C program.
3810: @end table
3811:
1.1.1.2 ! root 3812: @node Data Output
1.1 root 3813: @subsection Output of Data
3814:
3815: @table @code
3816: @findex ASM_OUTPUT_LONG_DOUBLE
3817: @findex ASM_OUTPUT_DOUBLE
3818: @findex ASM_OUTPUT_FLOAT
3819: @item ASM_OUTPUT_LONG_DOUBLE (@var{stream}, @var{value})
3820: @item ASM_OUTPUT_DOUBLE (@var{stream}, @var{value})
3821: @item ASM_OUTPUT_FLOAT (@var{stream}, @var{value})
3822: A C statement to output to the stdio stream @var{stream} an assembler
3823: instruction to assemble a floating-point constant of @code{TFmode},
3824: @code{DFmode} or @code{SFmode}, respectively, whose value is
3825: @var{value}. @var{value} will be a C expression of type
3826: @code{REAL_VALUE__TYPE}, usually @code{double}.@refill
3827:
3828: @findex ASM_OUTPUT_QUADRUPLE_INT
3829: @findex ASM_OUTPUT_DOUBLE_INT
3830: @findex ASM_OUTPUT_INT
3831: @findex ASM_OUTPUT_SHORT
3832: @findex ASM_OUTPUT_CHAR
3833: @findex output_addr_const
3834: @item ASM_OUTPUT_QUADRUPLE_INT (@var{stream}, @var{exp})
3835: @item ASM_OUTPUT_DOUBLE_INT (@var{stream}, @var{exp})
3836: @item ASM_OUTPUT_INT (@var{stream}, @var{exp})
3837: @itemx ASM_OUTPUT_SHORT (@var{stream}, @var{exp})
3838: @itemx ASM_OUTPUT_CHAR (@var{stream}, @var{exp})
3839: A C statement to output to the stdio stream @var{stream} an assembler
3840: instruction to assemble an integer of 16, 8, 4, 2 or 1 bytes,
3841: respectively, whose value is @var{value}. The argument @var{exp} will
3842: be an RTL expression which represents a constant value. Use
3843: @samp{output_addr_const (@var{stream}, @var{exp})} to output this value
3844: as an assembler expression.@refill
3845:
3846: For sizes larger than @code{UNITS_PER_WORD}, if the action of a macro
3847: would be identical to repeatedly calling the macro corresponding to
3848: a size of @code{UNITS_PER_WORD}, once for each word, you need not define
3849: the macro.
3850:
3851: @findex ASM_OUTPUT_BYTE
3852: @item ASM_OUTPUT_BYTE (@var{stream}, @var{value})
3853: A C statement to output to the stdio stream @var{stream} an assembler
3854: instruction to assemble a single byte containing the number @var{value}.
3855:
3856: @findex ASM_BYTE_OP
3857: @item ASM_BYTE_OP
3858: A C string constant giving the pseudo-op to use for a sequence of
3859: single-byte constants. If this macro is not defined, the default is
3860: @code{"byte"}.
3861:
3862: @findex ASM_OUTPUT_ASCII
3863: @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
3864: A C statement to output to the stdio stream @var{stream} an assembler
3865: instruction to assemble a string constant containing the @var{len}
3866: bytes at @var{ptr}. @var{ptr} will be a C expression of type
3867: @code{char *} and @var{len} a C expression of type @code{int}.
3868:
3869: If the assembler has a @code{.ascii} pseudo-op as found in the
3870: Berkeley Unix assembler, do not define the macro
3871: @code{ASM_OUTPUT_ASCII}.
3872:
3873: @findex ASM_OUTPUT_POOL_PROLOGUE
3874: @item ASM_OUTPUT_POOL_PROLOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
3875: A C statement to output assembler commands to define the start of the
3876: constant pool for a function. @var{funname} is a string giving
3877: the name of the function. Should the return type of the function
3878: be required, it can be obtained via @var{fundecl}. @var{size}
3879: is the size, in bytes, of the constant pool that will be written
3880: immediately after this call.
3881:
3882: If no constant-pool prefix is required, the usual case, this macro need
3883: not be defined.
3884:
3885: @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY
3886: @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
3887: A C statement (with or without semicolon) to output a constant in the
3888: constant pool, if it needs special treatment. (This macro need not do
3889: anything for RTL expressions that can be output normally.)
3890:
3891: The argument @var{file} is the standard I/O stream to output the
3892: assembler code on. @var{x} is the RTL expression for the constant to
3893: output, and @var{mode} is the machine mode (in case @var{x} is a
3894: @samp{const_int}). @var{align} is the required alignment for the value
3895: @var{x}; you should output an assembler directive to force this much
3896: alignment.
3897:
3898: The argument @var{labelno} is a number to use in an internal label for
3899: the address of this pool entry. The definition of this macro is
3900: responsible for outputting the label definition at the proper place.
3901: Here is how to do this:
3902:
3903: @example
3904: ASM_OUTPUT_INTERNAL_LABEL (@var{file}, "LC", @var{labelno});
3905: @end example
3906:
3907: When you output a pool entry specially, you should end with a
3908: @code{goto} to the label @var{jumpto}. This will prevent the same pool
3909: entry from being output a second time in the usual manner.
3910:
3911: You need not define this macro if it would do nothing.
3912:
3913: @findex ASM_OPEN_PAREN
3914: @findex ASM_CLOSE_PAREN
3915: @item ASM_OPEN_PAREN
3916: @itemx ASM_CLOSE_PAREN
3917: These macros are defined as C string constant, describing the syntax
3918: in the assembler for grouping arithmetic expressions. The following
3919: definitions are correct for most assemblers:
3920:
3921: @example
3922: #define ASM_OPEN_PAREN "("
3923: #define ASM_CLOSE_PAREN ")"
3924: @end example
3925: @end table
3926:
1.1.1.2 ! root 3927: @node Uninitialized Data
1.1 root 3928: @subsection Output of Uninitialized Variables
3929:
3930: Each of the macros in this section is used to do the whole job of
3931: outputting a single uninitialized variable.
3932:
3933: @table @code
3934: @findex ASM_OUTPUT_COMMON
3935: @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
3936: A C statement (sans semicolon) to output to the stdio stream
3937: @var{stream} the assembler definition of a common-label named
3938: @var{name} whose size is @var{size} bytes. The variable @var{rounded}
3939: is the size rounded up to whatever alignment the caller wants.
3940:
3941: Use the expression @code{assemble_name (@var{stream}, @var{name})} to
3942: output the name itself; before and after that, output the additional
3943: assembler syntax for defining the name, and a newline.
3944:
3945: This macro controls how the assembler definitions of uninitialized
3946: global variables are output.
3947:
3948: @findex ASM_OUTPUT_ALIGNED_COMMON
3949: @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
3950: Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
3951: separate, explicit argument. If you define this macro, it is used in
3952: place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
3953: handling the required alignment of the variable.
3954:
3955: @findex ASM_OUTPUT_SHARED_COMMON
3956: @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
3957: If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
3958: is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
3959: will be used.
3960:
3961: @findex ASM_OUTPUT_LOCAL
3962: @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
3963: A C statement (sans semicolon) to output to the stdio stream
3964: @var{stream} the assembler definition of a local-common-label named
3965: @var{name} whose size is @var{size} bytes. The variable @var{rounded}
3966: is the size rounded up to whatever alignment the caller wants.
3967:
3968: Use the expression @code{assemble_name (@var{stream}, @var{name})} to
3969: output the name itself; before and after that, output the additional
3970: assembler syntax for defining the name, and a newline.
3971:
3972: This macro controls how the assembler definitions of uninitialized
3973: static variables are output.
3974:
3975: @findex ASM_OUTPUT_ALIGNED_LOCAL
3976: @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
3977: Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
3978: separate, explicit argument. If you define this macro, it is used in
3979: place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
3980: handling the required alignment of the variable.
3981:
3982: @findex ASM_OUTPUT_SHARED_LOCAL
3983: @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
3984: If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
3985: is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
3986: will be used.
3987: @end table
3988:
1.1.1.2 ! root 3989: @node Label Output
1.1 root 3990: @subsection Output and Generation of Labels
3991:
3992: @table @code
3993: @findex ASM_OUTPUT_LABEL
3994: @findex assemble_name
3995: @item ASM_OUTPUT_LABEL (@var{stream}, @var{name})
3996: A C statement (sans semicolon) to output to the stdio stream
3997: @var{stream} the assembler definition of a label named @var{name}.
3998: Use the expression @code{assemble_name (@var{stream}, @var{name})} to
3999: output the name itself; before and after that, output the additional
4000: assembler syntax for defining the name, and a newline.
4001:
4002: @findex ASM_DECLARE_FUNCTION_NAME
4003: @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
4004: A C statement (sans semicolon) to output to the stdio stream
4005: @var{stream} any text necessary for declaring the name @var{name} of a
4006: function which is being defined. This macro is responsible for
4007: outputting the label definition (perhaps using
4008: @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
4009: @code{FUNCTION_DECL} tree node representing the function.
4010:
4011: If this macro is not defined, then the function name is defined in the
4012: usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
4013:
4014: @findex ASM_DECLARE_FUNCTION_SIZE
4015: @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
4016: A C statement (sans semicolon) to output to the stdio stream
4017: @var{stream} any text necessary for declaring the size of a function
4018: which is being defined. The argument @var{name} is the name of the
4019: function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
4020: representing the function.
4021:
4022: If this macro is not defined, then the function size is not defined.
4023:
4024: @findex ASM_DECLARE_OBJECT_NAME
4025: @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
4026: A C statement (sans semicolon) to output to the stdio stream
4027: @var{stream} any text necessary for declaring the name @var{name} of an
4028: initialized variable which is being defined. This macro must output the
4029: label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
4030: @var{decl} is the @code{VAR_DECL} tree node representing the variable.
4031:
4032: If this macro is not defined, then the variable name is defined in the
4033: usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
4034:
4035: @findex ASM_GLOBALIZE_LABEL
4036: @item ASM_GLOBALIZE_LABEL (@var{stream}, @var{name})
4037: A C statement (sans semicolon) to output to the stdio stream
4038: @var{stream} some commands that will make the label @var{name} global;
4039: that is, available for reference from other files. Use the expression
4040: @code{assemble_name (@var{stream}, @var{name})} to output the name
4041: itself; before and after that, output the additional assembler syntax
4042: for making that name global, and a newline.
4043:
4044: @findex ASM_OUTPUT_EXTERNAL
4045: @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
4046: A C statement (sans semicolon) to output to the stdio stream
4047: @var{stream} any text necessary for declaring the name of an external
4048: symbol named @var{name} which is referenced in this compilation but
4049: not defined. The value of @var{decl} is the tree node for the
4050: declaration.
4051:
4052: This macro need not be defined if it does not need to output anything.
4053: The GNU assembler and most Unix assemblers don't require anything.
4054:
4055: @findex ASM_OUTPUT_EXTERNAL_LIBCALL
4056: @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref})
4057: A C statement (sans semicolon) to output on @var{stream} an assembler
4058: pseudo-op to declare a library function name external. The name of the
4059: library function is given by @var{symref}, which has type @code{rtx} and
4060: is a @code{symbol_ref}.
4061:
4062: This macro need not be defined if it does not need to output anything.
4063: The GNU assembler and most Unix assemblers don't require anything.
4064:
4065: @findex ASM_OUTPUT_LABELREF
4066: @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
4067: A C statement (sans semicolon) to output to the stdio stream
4068: @var{stream} a reference in assembler syntax to a label named
4069: @var{name}. This should add @samp{_} to the front of the name, if that
4070: is customary on your operating system, as it is in most Berkeley Unix
4071: systems. This macro is used in @code{assemble_name}.
4072:
4073: @findex ASM_OUTPUT_LABELREF_AS_INT
4074: @item ASM_OUTPUT_LABELREF_AS_INT (@var{file}, @var{label})
4075: Define this macro for systems that use the program @code{collect2}.
4076: The definition should be a C statement to output a word containing
4077: a reference to the label @var{label}.
4078:
4079: @findex ASM_OUTPUT_INTERNAL_LABEL
4080: @item ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{prefix}, @var{num})
4081: A C statement to output to the stdio stream @var{stream} a label whose
4082: name is made from the string @var{prefix} and the number @var{num}.
1.1.1.2 ! root 4083:
! 4084: It is absolutely essential that these labels be distinct from the labels
! 4085: used for user-level functions and variables. Otherwise, certain programs
! 4086: will have name conflicts with internal labels.
! 4087:
! 4088: It is desirable to exclude internal labels from the symbol table of the
! 4089: object file. Most assemblers have a naming convention for labels that
! 4090: should be excluded; on many systems, the letter @samp{L} at the
! 4091: beginning of a label has this effect. You should find out what
! 4092: convention your system uses, and follow it.
! 4093:
! 4094: The usual definition of this macro is as follows:
1.1 root 4095:
4096: @example
4097: fprintf (@var{stream}, "L%s%d:\n", @var{prefix}, @var{num})
4098: @end example
4099:
1.1.1.2 ! root 4100: @findex ASM_GENERATE_INTERNAL_LABEL
! 4101: @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
! 4102: A C statement to store into the string @var{string} a label whose name
! 4103: is made from the string @var{prefix} and the number @var{num}.
! 4104:
! 4105: This string, when output subsequently by @code{assemble_name},
! 4106: should produce the same output that @code{ASM_OUTPUT_INTERNAL_LABEL}
! 4107: would produce with the same @var{prefix} and @var{num}.
! 4108:
! 4109: If the string begins with @samp{*}, then @code{assemble_name} will
! 4110: output the rest of the string unchanged. It is often convenient for
! 4111: @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
! 4112: string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
! 4113: to output the string, and may change it. (Of course,
! 4114: @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
! 4115: you should know what it does on your machine.)
! 4116:
1.1 root 4117: @findex ASM_FORMAT_PRIVATE_NAME
4118: @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
4119: A C expression to assign to @var{outvar} (which is a variable of type
4120: @code{char *}) a newly allocated string made from the string
4121: @var{name} and the number @var{number}, with some suitable punctuation
4122: added. Use @code{alloca} to get space for the string.
4123:
4124: This string will be used as the argument to @code{ASM_OUTPUT_LABELREF}
4125: to produce an assembler label for an internal static variable whose
4126: name is @var{name}. Therefore, the string must be such as to result
4127: in valid assembler code. The argument @var{number} is different each
4128: time this macro is executed; it prevents conflicts between
4129: similarly-named internal static variables in different scopes.
4130:
4131: Ideally this string should not be a valid C identifier, to prevent any
4132: conflict with the user's own symbols. Most assemblers allow periods
4133: or percent signs in assembler symbols; putting at least one of these
4134: between the name and the number will suffice.
4135:
4136: @findex OBJC_GEN_METHOD_LABEL
4137: @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
4138: Define this macro to override the default assembler names used for
4139: Objective C methods.
4140:
4141: The default name is a unique method number followed by the name of the
4142: class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
4143: the category is also included in the assembler name (e.g.@:
4144: @samp{_1_Foo_Bar}).
4145:
4146: These names are safe on most systems, but make debugging difficult since
4147: the method's selector is not present in the name. Therefore, particular
4148: systems define other ways of computing names.
4149:
1.1.1.2 ! root 4150: @var{buf} is an expression of type @code{char *} which gives you a
! 4151: buffer in which to store the name; its length is as long as
! 4152: @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
! 4153: 50 characters extra.
! 4154:
! 4155: The argument @var{is_inst} specifies whether the method is an instance
! 4156: method or a class method; @var{class_name} is the name of the class;
! 4157: @var{cat_name} is the name of the category (or NULL if the method is not
! 4158: in a category); and @var{sel_name} is the name of the selector.
1.1 root 4159:
4160: On systems where the assembler can handle quoted names, you can use this
4161: macro to provide more human-readable names.
4162: @end table
4163:
1.1.1.2 ! root 4164: @node Constructor Output
1.1 root 4165: @subsection Output of Initialization Routines
4166: @cindex initialization routines
4167: @cindex termination routines
4168: @cindex constructors, output of
4169: @cindex destructors, output of
4170:
4171: The compiled code for certain languages includes @dfn{constructors}
4172: (also called @dfn{initialization routines})---functions to initialize
4173: data in the program when the program is started. These functions need
4174: to be called before the program is ``started''---that is to say, before
4175: @code{main} is called.
4176:
4177: Compiling some languages generates @dfn{destructors} (also called
4178: @dfn{termination routines}) that should be called when the program
4179: terminates.
4180:
4181: To make the initialization and termination functions work, the compiler
4182: must output something in the assembler code to cause those functions to
4183: be called at the appropriate time. When you port the compiler to a new
4184: system, you need to specify what assembler code is needed to do this.
4185:
4186: Here are the two macros you should define if necessary:
4187:
4188: @table @code
4189: @item ASM_OUTPUT_CONSTRUCTOR (@var{stream}, @var{name})
4190: @findex ASM_OUTPUT_CONSTRUCTOR
4191: Define this macro as a C statement to output on the stream @var{stream}
4192: the assembler code to arrange to call the function named @var{name} at
4193: initialization time.
4194:
4195: Assume that @var{name} is the name of a C function generated
4196: automatically by the compiler. This function takes no arguments. Use
4197: the function @code{assemble_name} to output the name @var{name}; this
4198: performs any system-specific syntactic transformations such as adding an
4199: underscore.
4200:
4201: If you don't define this macro, nothing special is output to arrange to
4202: call the function. This is correct when the function will be called in
4203: some other manner---for example, by means of the @code{collect} program,
4204: which looks through the symbol table to find these functions by their
4205: names. If you want to use @code{collect}, then you need to arrange for
4206: it to be built and installed and used on your system.
4207:
4208: @item ASM_OUTPUT_DESTRUCTOR (@var{stream}, @var{name})
4209: @findex ASM_OUTPUT_DESTRUCTOR
4210: This is like @code{ASM_OUTPUT_CONSTRUCTOR} but used for termination
4211: functions rather than initialization functions.
4212: @end table
4213:
1.1.1.2 ! root 4214: @node Instruction Output
1.1 root 4215: @subsection Output of Assembler Instructions
4216:
4217: @table @code
4218: @findex REGISTER_NAMES
4219: @item REGISTER_NAMES
4220: A C initializer containing the assembler's names for the machine
4221: registers, each one as a C string constant. This is what translates
4222: register numbers in the compiler into assembler language.
4223:
4224: @findex ADDITIONAL_REGISTER_NAMES
4225: @item ADDITIONAL_REGISTER_NAMES
4226: If defined, a C initializer for an array of structures containing a name
4227: and a register number. This macro defines additional names for hard
4228: registers, thus allowing the @code{asm} option in declarations to refer
4229: to registers using alternate names.
4230:
4231: @findex ASM_OUTPUT_OPCODE
4232: @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
4233: Define this macro if you are using an unusual assembler that
4234: requires different names for the machine instructions.
4235:
4236: The definition is a C statement or statements which output an
4237: assembler instruction opcode to the stdio stream @var{stream}. The
4238: macro-operand @var{ptr} is a variable of type @code{char *} which
4239: points to the opcode name in its ``internal'' form---the form that is
4240: written in the machine description. The definition should output the
4241: opcode name to @var{stream}, performing any translation you desire, and
4242: increment the variable @var{ptr} to point at the end of the opcode
4243: so that it will not be output twice.
4244:
4245: In fact, your macro definition may process less than the entire opcode
4246: name, or more than the opcode name; but if you want to process text
4247: that includes @samp{%}-sequences to substitute operands, you must take
4248: care of the substitution yourself. Just be sure to increment
4249: @var{ptr} over whatever text should not be output normally.
4250:
4251: @findex recog_operand
4252: If you need to look at the operand values, they can be found as the
4253: elements of @code{recog_operand}.
4254:
4255: If the macro definition does nothing, the instruction is output
4256: in the usual way.
4257:
4258: @findex FINAL_PRESCAN_INSN
4259: @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
4260: If defined, a C statement to be executed just prior to the output of
4261: assembler code for @var{insn}, to modify the extracted operands so
4262: they will be output differently.
4263:
4264: Here the argument @var{opvec} is the vector containing the operands
4265: extracted from @var{insn}, and @var{noperands} is the number of
4266: elements of the vector which contain meaningful data for this insn.
4267: The contents of this vector are what will be used to convert the insn
4268: template into assembler code, so you can change the assembler output
4269: by changing the contents of the vector.
4270:
4271: This macro is useful when various assembler syntaxes share a single
4272: file of instruction patterns; by defining this macro differently, you
4273: can cause a large class of instructions to be output differently (such
4274: as with rearranged operands). Naturally, variations in assembler
4275: syntax affecting individual insn patterns ought to be handled by
4276: writing conditional output routines in those patterns.
4277:
4278: If this macro is not defined, it is equivalent to a null statement.
4279:
4280: @findex PRINT_OPERAND
4281: @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
4282: A C compound statement to output to stdio stream @var{stream} the
4283: assembler syntax for an instruction operand @var{x}. @var{x} is an
4284: RTL expression.
4285:
4286: @var{code} is a value that can be used to specify one of several ways
4287: of printing the operand. It is used when identical operands must be
4288: printed differently depending on the context. @var{code} comes from
4289: the @samp{%} specification that was used to request printing of the
4290: operand. If the specification was just @samp{%@var{digit}} then
4291: @var{code} is 0; if the specification was @samp{%@var{ltr}
4292: @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
4293:
4294: @findex reg_names
4295: If @var{x} is a register, this macro should print the register's name.
4296: The names can be found in an array @code{reg_names} whose type is
4297: @code{char *[]}. @code{reg_names} is initialized from
4298: @code{REGISTER_NAMES}.
4299:
4300: When the machine description has a specification @samp{%@var{punct}}
4301: (a @samp{%} followed by a punctuation character), this macro is called
4302: with a null pointer for @var{x} and the punctuation character for
4303: @var{code}.
4304:
4305: @findex PRINT_OPERAND_PUNCT_VALID_P
4306: @item PRINT_OPERAND_PUNCT_VALID_P (@var{code})
4307: A C expression which evaluates to true if @var{code} is a valid
4308: punctuation character for use in the @code{PRINT_OPERAND} macro. If
4309: @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
4310: punctuation characters (except for the standard one, @samp{%}) are used
4311: in this way.
4312:
4313: @findex PRINT_OPERAND_ADDRESS
4314: @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
4315: A C compound statement to output to stdio stream @var{stream} the
4316: assembler syntax for an instruction operand that is a memory reference
4317: whose address is @var{x}. @var{x} is an RTL expression.
4318:
4319: @cindex @code{ENCODE_SECTION_INFO} usage
4320: On some machines, the syntax for a symbolic address depends on the
4321: section that the address refers to. On these machines, define the macro
4322: @code{ENCODE_SECTION_INFO} to store the information into the
4323: @code{symbol_ref}, and then check for it here. @xref{Assembler Format}.
4324:
4325: @findex DBR_OUTPUT_SEQEND
4326: @findex dbr_sequence_length
4327: @item DBR_OUTPUT_SEQEND(@var{file})
4328: A C statement, to be executed after all slot-filler instructions have
4329: been output. If necessary, call @code{dbr_sequence_length} to
4330: determine the number of slots filled in a sequence (zero if not
4331: currently outputting a sequence), to decide how many no-ops to output,
4332: or whatever.
4333:
4334: Don't define this macro if it has nothing to do, but it is helpful in
4335: reading assembly output if the extent of the delay sequence is made
4336: explicit (e.g. with white space).
4337:
4338: @findex final_sequence
4339: Note that output routines for instructions with delay slots must be
4340: prepared to deal with not being output as part of a sequence (i.e.
4341: when the scheduling pass is not run, or when no slot fillers could be
4342: found.) The variable @code{final_sequence} is null when not
4343: processing a sequence, otherwise it contains the @code{sequence} rtx
4344: being output.
4345:
4346: @findex REGISTER_PREFIX
4347: @findex LOCAL_LABEL_PREFIX
4348: @findex USER_LABEL_PREFIX
4349: @findex IMMEDIATE_PREFIX
4350: @findex asm_fprintf
4351: @item REGISTER_PREFIX
4352: @itemx LOCAL_LABEL_PREFIX
4353: @itemx USER_LABEL_PREFIX
4354: @itemx IMMEDIATE_PREFIX
4355: If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
4356: @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
4357: @file{final.c}). These are useful when a single @file{md} file must
4358: support multiple assembler formats. In that case, the various @file{tm.h}
4359: files can define these macros differently.
4360:
4361: @findex ASM_OUTPUT_REG_PUSH
4362: @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
4363: A C expression to output to @var{stream} some assembler code
4364: which will push hard register number @var{regno} onto the stack.
4365: The code need not be optimal, since this macro is used only when
4366: profiling.
4367:
4368: @findex ASM_OUTPUT_REG_POP
4369: @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
4370: A C expression to output to @var{stream} some assembler code
4371: which will pop hard register number @var{regno} off of the stack.
4372: The code need not be optimal, since this macro is used only when
4373: profiling.
4374: @end table
4375:
1.1.1.2 ! root 4376: @node Dispatch Tables
1.1 root 4377: @subsection Output of Dispatch Tables
4378:
4379: @table @code
4380: @cindex dispatch table
4381: @findex ASM_OUTPUT_ADDR_DIFF_ELT
4382: @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{value}, @var{rel})
4383: This macro should be provided on machines where the addresses
4384: in a dispatch table are relative to the table's own address.
4385:
4386: The definition should be a C statement to output to the stdio stream
4387: @var{stream} an assembler pseudo-instruction to generate a difference
4388: between two labels. @var{value} and @var{rel} are the numbers of two
4389: internal labels. The definitions of these labels are output using
4390: @code{ASM_OUTPUT_INTERNAL_LABEL}, and they must be printed in the same
4391: way here. For example,
4392:
4393: @example
4394: fprintf (@var{stream}, "\t.word L%d-L%d\n",
4395: @var{value}, @var{rel})
4396: @end example
4397:
4398: @findex ASM_OUTPUT_ADDR_VEC_ELT
4399: @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
4400: This macro should be provided on machines where the addresses
4401: in a dispatch table are absolute.
4402:
4403: The definition should be a C statement to output to the stdio stream
4404: @var{stream} an assembler pseudo-instruction to generate a reference to
4405: a label. @var{value} is the number of an internal label whose
4406: definition is output using @code{ASM_OUTPUT_INTERNAL_LABEL}.
4407: For example,
4408:
4409: @example
4410: fprintf (@var{stream}, "\t.word L%d\n", @var{value})
4411: @end example
4412:
4413: @findex ASM_OUTPUT_CASE_LABEL
4414: @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
4415: Define this if the label before a jump-table needs to be output
4416: specially. The first three arguments are the same as for
4417: @code{ASM_OUTPUT_INTERNAL_LABEL}; the fourth argument is the
4418: jump-table which follows (a @code{jump_insn} containing an
4419: @code{addr_vec} or @code{addr_diff_vec}).
4420:
4421: This feature is used on system V to output a @code{swbeg} statement
4422: for the table.
4423:
4424: If this macro is not defined, these labels are output with
4425: @code{ASM_OUTPUT_INTERNAL_LABEL}.
4426:
4427: @findex ASM_OUTPUT_CASE_END
4428: @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
4429: Define this if something special must be output at the end of a
4430: jump-table. The definition should be a C statement to be executed
4431: after the assembler code for the table is written. It should write
4432: the appropriate code to stdio stream @var{stream}. The argument
4433: @var{table} is the jump-table insn, and @var{num} is the label-number
4434: of the preceding label.
4435:
4436: If this macro is not defined, nothing special is output at the end of
4437: the jump-table.
4438: @end table
4439:
1.1.1.2 ! root 4440: @node Alignment Output
1.1 root 4441: @subsection Assembler Commands for Alignment
4442:
4443: @table @code
4444: @findex ASM_OUTPUT_ALIGN_CODE
4445: @item ASM_OUTPUT_ALIGN_CODE (@var{file})
4446: A C expression to output text to align the location counter in the way
4447: that is desirable at a point in the code that is reached only by
4448: jumping.
4449:
4450: This macro need not be defined if you don't want any special alignment
4451: to be done at such a time. Most machine descriptions do not currently
4452: define the macro.
4453:
4454: @findex ASM_OUTPUT_LOOP_ALIGN
4455: @item ASM_OUTPUT_LOOP_ALIGN (@var{file})
4456: A C expression to output text to align the location counter in the way
4457: that is desirable at the beginning of a loop.
4458:
4459: This macro need not be defined if you don't want any special alignment
4460: to be done at such a time. Most machine descriptions do not currently
4461: define the macro.
4462:
4463: @findex ASM_OUTPUT_SKIP
4464: @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
4465: A C statement to output to the stdio stream @var{stream} an assembler
4466: instruction to advance the location counter by @var{nbytes} bytes.
4467: Those bytes should be zero when loaded. @var{nbytes} will be a C
4468: expression of type @code{int}.
4469:
4470: @findex ASM_NO_SKIP_IN_TEXT
4471: @item ASM_NO_SKIP_IN_TEXT
4472: Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
4473: text section because it fails put zeros in the bytes that are skipped.
4474: This is true on many Unix systems, where the pseudo--op to skip bytes
4475: produces no-op instructions rather than zeros when used in the text
4476: section.
4477:
4478: @findex ASM_OUTPUT_ALIGN
4479: @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
4480: A C statement to output to the stdio stream @var{stream} an assembler
4481: command to advance the location counter to a multiple of 2 to the
4482: @var{power} bytes. @var{power} will be a C expression of type @code{int}.
4483: @end table
4484:
1.1.1.2 ! root 4485: @node Debugging Info
1.1 root 4486: @section Controlling Debugging Information Format
4487:
4488: @table @code
4489: @findex DBX_REGISTER_NUMBER
4490: @item DBX_REGISTER_NUMBER (@var{regno})
4491: A C expression that returns the DBX register number for the compiler
4492: register number @var{regno}. In simple cases, the value of this
4493: expression may be @var{regno} itself. But sometimes there are some
4494: registers that the compiler knows about and DBX does not, or vice
4495: versa. In such cases, some register may need to have one number in
4496: the compiler and another for DBX.
4497:
4498: If two registers have consecutive numbers inside GNU CC, and they can be
4499: used as a pair to hold a multiword value, then they @emph{must} have
4500: consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
4501: Otherwise, debuggers will be unable to access such a pair, because they
4502: expect register pairs to be consecutive in their own numbering scheme.
4503:
4504: If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
4505: does not preserve register pairs, then what you must do instead is
4506: redefine the actual register numbering scheme.
4507:
4508: @findex DBX_DEBUGGING_INFO
4509: @item DBX_DEBUGGING_INFO
4510: Define this macro if GNU CC should produce debugging output for DBX
4511: in response to the @samp{-g} option.
4512:
4513: @findex SDB_DEBUGGING_INFO
4514: @item SDB_DEBUGGING_INFO
4515: Define this macro if GNU CC should produce COFF-style debugging output
4516: for SDB in response to the @samp{-g} option.
4517:
4518: @findex DWARF_DEBUGGING_INFO
4519: @item DWARF_DEBUGGING_INFO
4520: Define this macro if GNU CC should produce dwarf format debugging output
4521: in response to the @samp{-g} option.
4522:
1.1.1.2 ! root 4523: @findex XCOFF_DEBUGGING_INFO
! 4524: @item XCOFF_DEBUGGING_INFO
! 4525: Define this macro if GNU CC should produce XCOFF format debugging output
! 4526: in response to the @samp{-g} option.
! 4527:
1.1 root 4528: @findex DEFAULT_GDB_EXTENSIONS
4529: @item DEFAULT_GDB_EXTENSIONS
4530: Define this macro to control whether GNU CC should by default generate
4531: GDB's extended version of DBX debugging information (assuming DBX-format
4532: debugging information is enabled at all). If you don't define the
4533: macro, the default is 1: always generate the extended information.
4534:
4535: @findex DEBUG_SYMS_TEXT
4536: @item DEBUG_SYMS_TEXT
4537: Define this macro if all @code{.stabs} commands should be output while
4538: in the text section.
4539:
4540: @findex DEBUGGER_AUTO_OFFSET
4541: @item DEBUGGER_AUTO_OFFSET (@var{x})
4542: A C expression that returns the integer offset value for an automatic
4543: variable having address @var{x} (an RTL expression). The default
4544: computation assumes that @var{x} is based on the frame-pointer and
4545: gives the offset from the frame-pointer. This is required for targets
4546: that produce debugging output for DBX or COFF-style debugging output
4547: for SDB and allow the frame-pointer to be eliminated when the
4548: @samp{-g} options is used.
4549:
4550: @findex DEBUGGER_ARG_OFFSET
4551: @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
4552: A C expression that returns the integer offset value for an argument
4553: having address @var{x} (an RTL expression). The nominal offset is
4554: @var{offset}.
4555:
4556: @findex ASM_STABS_OP
4557: @item ASM_STABS_OP
4558: A C string constant naming the assembler pseudo op to use instead of
4559: @code{.stabs} to define an ordinary debugging symbol. If you don't
4560: define this macro, @code{.stabs} is used. This macro applies only to
4561: DBX debugging information format.
4562:
4563: @findex ASM_STABD_OP
4564: @item ASM_STABD_OP
4565: A C string constant naming the assembler pseudo op to use instead of
4566: @code{.stabd} to define a debugging symbol whose value is the current
4567: location. If you don't define this macro, @code{.stabd} is used.
4568: This macro applies only to DBX debugging information format.
4569:
4570: @findex ASM_STABN_OP
4571: @item ASM_STABN_OP
4572: A C string constant naming the assembler pseudo op to use instead of
4573: @code{.stabn} to define a debugging symbol with no name. If you don't
4574: define this macro, @code{.stabn} is used. This macro applies only to
4575: DBX debugging information format.
4576:
4577: @findex PUT_SDB_@dots{}
4578: @item PUT_SDB_@dots{}
4579: Define these macros to override the assembler syntax for the special
4580: SDB assembler directives. See @file{sdbout.c} for a list of these
4581: macros and their arguments. If the standard syntax is used, you need
4582: not define them yourself.
4583:
4584: @findex SDB_DELIM
4585: @item SDB_DELIM
4586: Some assemblers do not support a semicolon as a delimiter, even between
4587: SDB assembler directives. In that case, define this macro to be the
4588: delimiter to use (usually @samp{\n}). It is not necessary to define
4589: a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
4590: required.
4591:
4592: @findex SDB_GENERATE_FAKE
4593: @item SDB_GENERATE_FAKE
4594: Define this macro to override the usual method of constructing a dummy
4595: name for anonymous structure and union types. See @file{sdbout.c} for
4596: more information.
4597:
4598: @findex SDB_ALLOW_UNKNOWN_REFERENCES
4599: @item SDB_ALLOW_UNKNOWN_REFERENCES
4600: Define this macro to allow references to unknown structure,
4601: union, or enumeration tags to be emitted. Standard COFF does not
4602: allow handling of unknown references, MIPS ECOFF has support for
4603: it.
4604:
4605: @findex SDB_ALLOW_FORWARD_REFERENCES
4606: @item SDB_ALLOW_FORWARD_REFERENCES
4607: Define this macro to allow references to structure, union, or
4608: enumeration tags that have not yet been seen to be handled. Some
4609: assemblers choke if forward tags are used, while some require it.
4610:
4611: @findex DBX_NO_XREFS
4612: @item DBX_NO_XREFS
4613: Define this macro if DBX on your system does not support the construct
4614: @samp{xs@var{tagname}}. On some systems, this construct is used to
4615: describe a forward reference to a structure named @var{tagname}.
4616: On other systems, this construct is not supported at all.
4617:
4618: @findex DBX_CONTIN_LENGTH
4619: @item DBX_CONTIN_LENGTH
4620: A symbol name in DBX-format debugging information is normally
4621: continued (split into two separate @code{.stabs} directives) when it
4622: exceeds a certain length (by default, 80 characters). On some
4623: operating systems, DBX requires this splitting; on others, splitting
4624: must not be done. You can inhibit splitting by defining this macro
4625: with the value zero. You can override the default splitting-length by
4626: defining this macro as an expression for the length you desire.
4627:
4628: @findex DBX_CONTIN_CHAR
4629: @item DBX_CONTIN_CHAR
4630: Normally continuation is indicated by adding a @samp{\} character to
4631: the end of a @code{.stabs} string when a continuation follows. To use
4632: a different character instead, define this macro as a character
4633: constant for the character you want to use. Do not define this macro
4634: if backslash is correct for your system.
4635:
1.1.1.2 ! root 4636: @findex DBX_WORKING_DIRECTORY
! 4637: @item DBX_WORKING_DIRECTORY
! 4638: Define this if DBX wants to have the current directory recorded in each
! 4639: object file.
! 4640:
! 4641: Note that the working directory is always recorded if GDB extensions are
! 4642: enabled.
! 4643:
1.1 root 4644: @findex DBX_STATIC_STAB_DATA_SECTION
4645: @item DBX_STATIC_STAB_DATA_SECTION
4646: Define this macro if it is necessary to go to the data section before
4647: outputting the @samp{.stabs} pseudo-op for a non-global static
4648: variable.
4649:
4650: @findex DBX_LBRAC_FIRST
4651: @item DBX_LBRAC_FIRST
4652: Define this macro if the @code{N_LBRAC} symbol for a block should
4653: precede the debugging information for variables and functions defined in
4654: that block. Normally, in DBX format, the @code{N_LBRAC} symbol comes
4655: first.
4656:
4657: @findex DBX_FUNCTION_FIRST
4658: @item DBX_FUNCTION_FIRST
4659: Define this macro if the DBX information for a function and its
4660: arguments should precede the assembler code for the function. Normally,
4661: in DBX format, the debugging information entirely follows the assembler
4662: code.
4663:
4664: @findex DBX_OUTPUT_FUNCTION_END
4665: @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
4666: Define this macro if the target machine requires special output at the
4667: end of the debugging information for a function. The definition should
4668: be a C statement (sans semicolon) to output the appropriate information
4669: to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
4670: the function.
4671:
4672: @findex DBX_OUTPUT_STANDARD_TYPES
4673: @item DBX_OUTPUT_STANDARD_TYPES (@var{syms})
4674: Define this macro if you need to control the order of output of the
4675: standard data types at the beginning of compilation. The argument
4676: @var{syms} is a @code{tree} which is a chain of all the predefined
4677: global symbols, including names of data types.
4678:
4679: Normally, DBX output starts with definitions of the types for integers
4680: and characters, followed by all the other predefined types of the
4681: particular language in no particular order.
4682:
4683: On some machines, it is necessary to output different particular types
4684: first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
4685: those symbols in the necessary order. Any predefined types that you
4686: don't explicitly output will be output afterward in no particular order.
4687:
4688: Be careful not to define this macro so that it works only for C. There
4689: are no global variables to access most of the built-in types, because
4690: another language may have another set of types. The way to output a
4691: particular type is to look through @var{syms} to see if you can find it.
4692: Here is an example:
4693:
4694: @example
4695: @{
4696: tree decl;
4697: for (decl = syms; decl; decl = TREE_CHAIN (decl))
4698: if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)), "long int"))
4699: dbxout_symbol (decl);
4700: @dots{}
4701: @}
4702: @end example
4703:
4704: @noindent
4705: This does nothing if the expected type does not exist.
4706:
4707: See the function @code{init_decl_processing} in source file
4708: @file{c-decl.c} to find the names to use for all the built-in C types.
4709:
4710: @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME
4711: @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
4712: A C statement to output DBX debugging information to the stdio stream
4713: @var{stream} which indicates that file @var{name} is the main source
4714: file---the file specified as the input file for compilation.
4715: This macro is called only once, at the beginning of compilation.
4716:
4717: This macro need not be defined if the standard form of output
4718: for DBX debugging information is appropriate.
4719:
4720: @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY
4721: @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
4722: A C statement to output DBX debugging information to the stdio stream
4723: @var{stream} which indicates that the current directory during
4724: compilation is named @var{name}.
4725:
4726: This macro need not be defined if the standard form of output
4727: for DBX debugging information is appropriate.
4728:
4729: @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END
4730: @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
4731: A C statement to output DBX debugging information at the end of
4732: compilation of the main source file @var{name}.
4733:
4734: If you don't define this macro, nothing special is output at the end
4735: of compilation, which is correct for most machines.
4736:
4737: @findex DBX_OUTPUT_SOURCE_FILENAME
4738: @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
4739: A C statement to output DBX debugging information to the stdio stream
4740: @var{stream} which indicates that file @var{name} is the current source
4741: file. This output is generated each time input shifts to a different
4742: source file as a result of @samp{#include}, the end of an included file,
4743: or a @samp{#line} command.
4744:
4745: This macro need not be defined if the standard form of output
4746: for DBX debugging information is appropriate.
4747: @end table
4748:
1.1.1.2 ! root 4749: @node Cross-compilation
1.1 root 4750: @section Cross Compilation and Floating Point Format
4751: @cindex cross compilation and floating point
4752: @cindex floating point format and cross compilation
4753:
4754: While all modern machines use 2's complement representation for integers,
4755: there are a variety of representations for floating point numbers. This
4756: means that in a cross-compiler the representation of floating point numbers
4757: in the compiled program may be different from that used in the machine
4758: doing the compilation.
4759:
4760: @findex atof
4761: Because different representation systems may offer different amounts of
4762: range and precision, the cross compiler cannot safely use the host
4763: machine's floating point arithmetic. Therefore, floating point constants
4764: must be represented in the target machine's format. This means that the
4765: cross compiler cannot use @code{atof} to parse a floating point constant;
4766: it must have its own special routine to use instead. Also, constant
4767: folding must emulate the target machine's arithmetic (or must not be done
4768: at all).
4769:
4770: The macros in the following table should be defined only if you are cross
4771: compiling between different floating point formats.
4772:
4773: Otherwise, don't define them. Then default definitions will be set up which
4774: use @code{double} as the data type, @code{==} to test for equality, etc.
4775:
4776: You don't need to worry about how many times you use an operand of any
4777: of these macros. The compiler never uses operands which have side effects.
4778:
4779: @table @code
4780: @findex REAL_VALUE_TYPE
4781: @item REAL_VALUE_TYPE
4782: A macro for the C data type to be used to hold a floating point value
4783: in the target machine's format. Typically this would be a
4784: @code{struct} containing an array of @code{int}.
4785:
4786: @findex REAL_VALUES_EQUAL
4787: @item REAL_VALUES_EQUAL (@var{x}, @var{y})
4788: A macro for a C expression which compares for equality the two values,
4789: @var{x} and @var{y}, both of type @code{REAL_VALUE_TYPE}.
4790:
4791: @findex REAL_VALUES_LESS
4792: @item REAL_VALUES_LESS (@var{x}, @var{y})
4793: A macro for a C expression which tests whether @var{x} is less than
4794: @var{y}, both values being of type @code{REAL_VALUE_TYPE} and
4795: interpreted as floating point numbers in the target machine's
4796: representation.
4797:
4798: @findex REAL_VALUE_LDEXP
4799: @findex ldexp
4800: @item REAL_VALUE_LDEXP (@var{x}, @var{scale})
4801: A macro for a C expression which performs the standard library
4802: function @code{ldexp}, but using the target machine's floating point
4803: representation. Both @var{x} and the value of the expression have
4804: type @code{REAL_VALUE_TYPE}. The second argument, @var{scale}, is an
4805: integer.
4806:
4807: @findex REAL_VALUE_FIX
4808: @item REAL_VALUE_FIX (@var{x})
4809: A macro whose definition is a C expression to convert the target-machine
4810: floating point value @var{x} to a signed integer. @var{x} has type
4811: @code{REAL_VALUE_TYPE}.
4812:
4813: @findex REAL_VALUE_UNSIGNED_FIX
4814: @item REAL_VALUE_UNSIGNED_FIX (@var{x})
4815: A macro whose definition is a C expression to convert the target-machine
4816: floating point value @var{x} to an unsigned integer. @var{x} has type
4817: @code{REAL_VALUE_TYPE}.
4818:
4819: @findex REAL_VALUE_FIX_TRUNCATE
4820: @item REAL_VALUE_FIX_TRUNCATE (@var{x})
4821: A macro whose definition is a C expression to convert the target-machine
4822: floating point value @var{x} to a signed integer, rounding toward 0.
4823: @var{x} has type @code{REAL_VALUE_TYPE}.
4824:
4825: @findex REAL_VALUE_UNSIGNED_FIX_TRUNCATE
4826: @item REAL_VALUE_UNSIGNED_FIX_TRUNCATE (@var{x})
4827: A macro whose definition is a C expression to convert the target-machine
4828: floating point value @var{x} to an unsigned integer, rounding toward 0.
4829: @var{x} has type @code{REAL_VALUE_TYPE}.
4830:
4831: @findex REAL_VALUE_ATOF
4832: @item REAL_VALUE_ATOF (@var{string})
4833: A macro for a C expression which converts @var{string}, an expression
4834: of type @code{char *}, into a floating point number in the target
4835: machine's representation. The value has type @code{REAL_VALUE_TYPE}.
4836:
4837: @findex REAL_INFINITY
4838: @item REAL_INFINITY
4839: Define this macro if infinity is a possible floating point value, and
4840: therefore division by 0 is legitimate.
4841:
4842: @findex REAL_VALUE_ISINF
4843: @findex isinf
4844: @item REAL_VALUE_ISINF (@var{x})
4845: A macro for a C expression which determines whether @var{x}, a floating
4846: point value, is infinity. The value has type @code{int}.
4847: By default, this is defined to call @code{isinf}.
4848:
4849: @findex REAL_VALUE_ISNAN
4850: @findex isnan
4851: @item REAL_VALUE_ISNAN (@var{x})
4852: A macro for a C expression which determines whether @var{x}, a floating
4853: point value, is a ``nan'' (not-a-number). The value has type
4854: @code{int}. By default, this is defined to call @code{isnan}.
4855: @end table
4856:
4857: @cindex constant folding and floating point
4858: Define the following additional macros if you want to make floating
4859: point constant folding work while cross compiling. If you don't
4860: define them, cross compilation is still possible, but constant folding
4861: will not happen for floating point values.
4862:
4863: @table @code
4864: @findex REAL_ARITHMETIC
4865: @item REAL_ARITHMETIC (@var{output}, @var{code}, @var{x}, @var{y})
4866: A macro for a C statement which calculates an arithmetic operation of
4867: the two floating point values @var{x} and @var{y}, both of type
4868: @code{REAL_VALUE_TYPE} in the target machine's representation, to
4869: produce a result of the same type and representation which is stored
4870: in @var{output} (which will be a variable).
4871:
4872: The operation to be performed is specified by @var{code}, a tree code
4873: which will always be one of the following: @code{PLUS_EXPR},
4874: @code{MINUS_EXPR}, @code{MULT_EXPR}, @code{RDIV_EXPR},
4875: @code{MAX_EXPR}, @code{MIN_EXPR}.@refill
4876:
4877: @cindex overflow while constant folding
4878: The expansion of this macro is responsible for checking for overflow.
4879: If overflow happens, the macro expansion should execute the statement
4880: @code{return 0;}, which indicates the inability to perform the
4881: arithmetic operation requested.
4882:
4883: @findex REAL_VALUE_NEGATE
4884: @item REAL_VALUE_NEGATE (@var{x})
4885: A macro for a C expression which returns the negative of the floating
4886: point value @var{x}. Both @var{x} and the value of the expression
4887: have type @code{REAL_VALUE_TYPE} and are in the target machine's
4888: floating point representation.
4889:
4890: There is no way for this macro to report overflow, since overflow
4891: can't happen in the negation operation.
4892:
4893: @findex REAL_VALUE_TRUNCATE
4894: @item REAL_VALUE_TRUNCATE (@var{x})
4895: A macro for a C expression which converts the double-precision floating
4896: point value @var{x} to single-precision.
4897:
4898: Both @var{x} and the value of the expression have type
4899: @code{REAL_VALUE_TYPE} and are in the target machine's floating point
4900: representation. However, the value should have an appropriate bit
4901: pattern to be output properly as a single-precision floating constant.
4902:
4903: There is no way for this macro to report overflow.
4904:
4905: @findex REAL_VALUE_TO_INT
4906: @item REAL_VALUE_TO_INT (@var{low}, @var{high}, @var{x})
4907: A macro for a C expression which converts a floating point value
4908: @var{x} into a double-precision integer which is then stored into
4909: @var{low} and @var{high}, two variables of type @var{int}.
4910:
4911: @item REAL_VALUE_FROM_INT (@var{x}, @var{low}, @var{high})
4912: @findex REAL_VALUE_FROM_INT
4913: A macro for a C expression which converts a double-precision integer
4914: found in @var{low} and @var{high}, two variables of type @var{int},
4915: into a floating point value which is then stored into @var{x}.
4916: @end table
4917:
1.1.1.2 ! root 4918: @node Misc
1.1 root 4919: @section Miscellaneous Parameters
4920: @cindex parameters, miscellaneous
4921:
4922: @table @code
4923: @item PREDICATE_CODES
4924: @findex PREDICATE_CODES
4925: Optionally define this if you have added predicates to
4926: @file{@var{machine}.c}. This macro is called within an initializer of an
4927: array of structures. The first field in the structure is the name of a
1.1.1.2 ! root 4928: predicate and the second field is an array of rtl codes. For each
1.1 root 4929: predicate, list all rtl codes that can be in expressions matched by the
4930: predicate. The list should have a trailing comma. Here is an example
4931: of two entries in the list for a typical RISC machine:
4932:
4933: @example
4934: #define PREDICATE_CODES \
4935: @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
4936: @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
4937: @end example
4938:
4939: Defining this macro does not affect the generated code (however,
4940: incorrect definitions that omit an rtl code that may be matched by the
4941: predicate can cause the compiler to malfunction). Instead, it allows
4942: the table built by @file{genrecog} to be more compact and efficient,
4943: thus speeding up the compiler. The most important predicates to include
4944: in the list specified by this macro are thoses used in the most insn
4945: patterns.
4946:
4947: @findex CASE_VECTOR_MODE
4948: @item CASE_VECTOR_MODE
4949: An alias for a machine mode name. This is the machine mode that
4950: elements of a jump-table should have.
4951:
4952: @findex CASE_VECTOR_PC_RELATIVE
4953: @item CASE_VECTOR_PC_RELATIVE
4954: Define this macro if jump-tables should contain relative addresses.
4955:
4956: @findex CASE_DROPS_THROUGH
4957: @item CASE_DROPS_THROUGH
4958: Define this if control falls through a @code{case} insn when the index
4959: value is out of range. This means the specified default-label is
4960: actually ignored by the @code{case} insn proper.
4961:
4962: @findex BYTE_LOADS_ZERO_EXTEND
4963: @item BYTE_LOADS_ZERO_EXTEND
4964: Define this macro if an instruction to load a value narrower than a
4965: word from memory into a register also zero-extends the value to the whole
4966: register.
4967:
4968: @findex IMPLICIT_FIX_EXPR
4969: @item IMPLICIT_FIX_EXPR
4970: An alias for a tree code that should be used by default for conversion
4971: of floating point values to fixed point. Normally,
4972: @code{FIX_ROUND_EXPR} is used.@refill
4973:
4974: @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC
4975: @item FIXUNS_TRUNC_LIKE_FIX_TRUNC
4976: Define this macro if the same instructions that convert a floating
4977: point number to a signed fixed point number also convert validly to an
4978: unsigned one.
4979:
4980: @findex EASY_DIV_EXPR
4981: @item EASY_DIV_EXPR
4982: An alias for a tree code that is the easiest kind of division to
4983: compile code for in the general case. It may be
4984: @code{TRUNC_DIV_EXPR}, @code{FLOOR_DIV_EXPR}, @code{CEIL_DIV_EXPR} or
4985: @code{ROUND_DIV_EXPR}. These four division operators differ in how
4986: they round the result to an integer. @code{EASY_DIV_EXPR} is used
4987: when it is permissible to use any of those kinds of division and the
4988: choice should be made on the basis of efficiency.@refill
4989:
4990: @findex MOVE_MAX
4991: @item MOVE_MAX
4992: The maximum number of bytes that a single instruction can move quickly
4993: from memory to memory.
4994:
4995: @findex SHIFT_COUNT_TRUNCATED
4996: @item SHIFT_COUNT_TRUNCATED
4997: Defining this macro causes the compiler to omit a sign-extend,
4998: zero-extend, or bitwise `and' instruction that truncates the count of a
4999: shift operation to a width equal to the number of bits needed to
5000: represent the size of the object being shifted. On machines that have
5001: instructions that act on bitfields at variable positions, including `bit
5002: test' instructions, defining @code{SHIFT_COUNT_TRUNCATED} also causes
5003: truncation not to be applied to these instructions.
5004:
5005: If both types of instructions truncate the count (for shifts) and
5006: position (for bitfield operations), or if no variable-position bitfield
5007: instructions exist, you should define this macro.
5008:
1.1.1.2 ! root 5009: However, on some machines, such as the 80386 and the 680x0, truncation
! 5010: only applies to shift operations and not the (real or pretended)
! 5011: bitfield operations. Do not define @code{SHIFT_COUNT_TRUNCATED} on such
! 5012: machines. Instead, add patterns to the @file{md} file that include the
! 5013: implied truncation of the shift instructions.
1.1 root 5014:
5015: @findex TRULY_NOOP_TRUNCATION
5016: @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
5017: A C expression which is nonzero if on this machine it is safe to
5018: ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
5019: bits (where @var{outprec} is smaller than @var{inprec}) by merely
5020: operating on it as if it had only @var{outprec} bits.
5021:
5022: On many machines, this expression can be 1.
5023:
5024: It is reported that suboptimal code can result when
5025: @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for modes for
5026: which @code{MODES_TIEABLE_P} is 0. If this is the case, making
5027: @code{TRULY_NOOP_TRUNCATION} return 0 in such cases may improve things.
5028:
5029: @findex STORE_FLAG_VALUE
5030: @item STORE_FLAG_VALUE
5031: A C expression describing the value returned by a comparison operator
5032: and stored by a store-flag instruction (@samp{s@var{cond}}) when the
5033: condition is true. This description must apply to @emph{all} the
5034: @samp{s@var{cond}} patterns and all the comparison operators.
5035:
5036: A value of 1 or -1 means that the instruction implementing the
5037: comparison operator returns exactly 1 or -1 when the comparison is true
5038: and 0 when the comparison is false. Otherwise, the value indicates
5039: which bits of the result are guaranteed to be 1 when the comparison is
5040: true. This value is interpreted in the mode of the comparison
5041: operation, which is given by the mode of the first operand in the
5042: @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
5043: @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
5044: the compiler.
5045:
5046: If @code{STORE_FLAG_VALUE} is neither 1 or -1, the compiler will
5047: generate code that depends only on the specified bits. It can also
5048: replace comparison operators with equivalent operations if they cause
5049: the required bits to be set, even if the remaining bits are undefined.
5050: For example, on a machine whose comparison operators return an
5051: @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
5052: @samp{0x80000000}, saying that just the sign bit is relevant, the
5053: expression
5054:
5055: @example
5056: (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
5057: @end example
5058:
5059: @noindent
5060: can be converted to
5061:
5062: @example
5063: (ashift:SI @var{x} (const_int @var{n}))
5064: @end example
5065:
5066: @noindent
5067: where @var{n} is the appropriate shift count to move the bit being
5068: tested into the sign bit.
5069:
5070: There is no way to describe a machine that always sets the low-order bit
5071: for a true value, but does not guarantee the value of any other bits,
5072: but we do not know of any machine that has such an instruction. If you
5073: are trying to port GNU CC to such a machine, include an instruction to
5074: perform a logical-and of the result with 1 in the pattern for the
5075: comparison operators and let us know (@pxref{Bug Reporting}).
5076:
5077: Often, a machine will have multiple instructions that obtain a value
5078: from a comparison (or the condition codes). Here are rules to guide the
5079: choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
5080: to be used:
5081:
5082: @itemize @bullet
5083: @item
5084: Use the shortest sequence that yields a valid definition for
5085: @code{STORE_FLAG_VALUE}. It is more efficent for the compiler to
5086: ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
5087: comparison operators to do so because there may be opportunities to
5088: combine the normalization with other operations.
5089:
5090: @item
5091: For equal-length sequences, use a value of 1 or -1, with -1 being
5092: slightly preferred on machines with expensive jumps and 1 preferred on
5093: other machines.
5094:
5095: @item
5096: As a second choice, choose a value of @samp{0x80000001} if instructions
5097: exist that set both the sign and low-order bits but do not define the
5098: others.
5099:
5100: @item
5101: Otherwise, use a value of @samp{0x80000000}.
5102: @end itemize
5103:
5104: You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
5105: instructions.
5106:
5107: @findex Pmode
5108: @item Pmode
5109: An alias for the machine mode for pointers. Normally the definition
5110: can be
5111:
5112: @example
5113: #define Pmode SImode
5114: @end example
5115:
5116: @findex FUNCTION_MODE
5117: @item FUNCTION_MODE
5118: An alias for the machine mode used for memory references to functions
5119: being called, in @code{call} RTL expressions. On most machines this
5120: should be @code{QImode}.
5121:
5122: @findex INTEGRATE_THRESHOLD
5123: @item INTEGRATE_THRESHOLD (@var{decl})
5124: A C expression for the maximum number of instructions above which the
5125: function @var{decl} should not be inlined. @var{decl} is a
5126: @code{FUNCTION_DECL} node.
5127:
5128: The default definition of this macro is 64 plus 8 times the number of
5129: arguments that the function accepts. Some people think a larger
5130: threshold should be used on RISC machines.
5131:
5132: @findex SCCS_DIRECTIVE
5133: @item SCCS_DIRECTIVE
5134: Define this if the preprocessor should ignore @code{#sccs} directives
5135: and print no error message.
5136:
5137: @findex HANDLE_PRAGMA
5138: @findex #pragma
5139: @findex pragma
5140: @item HANDLE_PRAGMA (@var{stream})
5141: Define this macro if you want to implement any pragmas. If defined, it
5142: should be a C statement to be executed when @code{#pragma} is seen. The
5143: argument @var{stream} is the stdio input stream from which the source
5144: text can be read.
5145:
5146: It is generally a bad idea to implement new uses of @code{#pragma}. The
5147: only reason to define this macro is for compatibility with other
5148: compilers that do support @code{#pragma} for the sake of any user
5149: programs which already use it.
5150:
5151: @findex DOLLARS_IN_IDENTIFIERS
5152: @item DOLLARS_IN_IDENTIFIERS
5153: Define this macro to control use of the character @samp{$} in identifier
5154: names. The value should be 0, 1, or 2. 0 means @samp{$} is not allowed
5155: by default; 1 means it is allowed by default if @samp{-traditional} is
5156: used; 2 means it is allowed by default provided @samp{-ansi} is not used.
5157: 1 is the default; there is no need to define this macro in that case.
5158:
1.1.1.2 ! root 5159: @findex NO_DOLLAR_IN_LABEL
! 5160: @item NO_DOLLAR_IN_LABEL
! 5161: Define this macro if the assembler does not accept the character
! 5162: @samp{$} in label names. By default constructors and destructors in
! 5163: G++ have @samp{$} in the identifiers. If this macro is defined,
! 5164: @samp{.} is used instead.
! 5165:
1.1 root 5166: @findex DEFAULT_MAIN_RETURN
5167: @item DEFAULT_MAIN_RETURN
5168: Define this macro if the target system expects every program's @code{main}
5169: function to return a standard ``success'' value by default (if no other
5170: value is explicitly returned).
5171:
5172: The definition should be a C statement (sans semicolon) to generate the
5173: appropriate rtl instructions. It is used only when compiling the end of
5174: @code{main}.
5175:
5176: @item HAVE_ATEXIT
5177: @findex HAVE_ATEXIT
5178: Define this if the target system supports the function
5179: @code{atexit} from the ANSI C standard. If this is not defined,
5180: and @code{INIT_SECTION_ASM_OP} is not defined, a default
5181: @code{exit} function will be provided to support C++.
5182:
5183: @item EXIT_BODY
5184: @findex EXIT_BODY
5185: Define this if your @code{exit} function needs to do something
5186: besides calling an external function @code{_cleanup} before
5187: terminating with @code{_exit}. The @code{EXIT_BODY} macro is
5188: only needed if netiher @code{HAVE_ATEXIT} nor
5189: @code{INIT_SECTION_ASM_OP} are defined.
5190: @end table
5191: @end ifset
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.