![[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 [Apple XNU] / GNUtools / cc|
Return to tm.texi CVS log | Up to [Apple XNU] / GNUtools / cc |
1.1 root 1: @c Copyright (C) 1988, 1989, 1992, 1993 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: @node Target Macros
6: @chapter Target Description Macros
7: @cindex machine description macros
8: @cindex target description macros
9: @cindex macros, target description
10: @cindex @file{tm.h} macros
11:
12: In addition to the file @file{@var{machine}.md}, a machine description
13: includes a C header file conventionally given the name
14: @file{@var{machine}.h}. This header file defines numerous macros
15: that convey the information about the target machine that does not fit
16: into the scheme of the @file{.md} file. The file @file{tm.h} should be
17: a link to @file{@var{machine}.h}. The header file @file{config.h}
18: includes @file{tm.h} and most compiler source files include
19: @file{config.h}.
20:
21: @menu
22: * Driver:: Controlling how the driver runs the compilation passes.
23: * Run-time Target:: Defining @samp{-m} options like @samp{-m68000} and @samp{-m68020}.
24: * Storage Layout:: Defining sizes and alignments of data.
25: * Type Layout:: Defining sizes and properties of basic user data types.
26: * Registers:: Naming and describing the hardware registers.
27: * Register Classes:: Defining the classes of hardware registers.
28: * Stack and Calling:: Defining which way the stack grows and by how much.
29: * Varargs:: Defining the varargs macros.
30: * Trampolines:: Code set up at run time to enter a nested function.
31: * Library Calls:: Controlling how library routines are implicitly called.
32: * Addressing Modes:: Defining addressing modes valid for memory operands.
33: * Condition Code:: Defining how insns update the condition code.
34: * Costs:: Defining relative costs of different operations.
35: * Sections:: Dividing storage into text, data, and other sections.
36: * PIC:: Macros for position independent code.
37: * Assembler Format:: Defining how to write insns and pseudo-ops to output.
38: * Debugging Info:: Defining the format of debugging output.
39: * Cross-compilation:: Handling floating point for cross-compilers.
40: * Misc:: Everything else.
41: @end menu
42:
43: @node Driver
44: @section Controlling the Compilation Driver, @file{gcc}
45: @cindex driver
46: @cindex controlling the compilation driver
47:
48: @table @code
49: @findex SWITCH_TAKES_ARG
50: @item SWITCH_TAKES_ARG (@var{char})
51: A C expression which determines whether the option @samp{-@var{char}}
52: takes arguments. The value should be the number of arguments that
53: option takes--zero, for many options.
54:
55: By default, this macro is defined to handle the standard options
56: properly. You need not define it unless you wish to add additional
57: options which take arguments.
58:
59: @findex WORD_SWITCH_TAKES_ARG
60: @item WORD_SWITCH_TAKES_ARG (@var{name})
61: A C expression which determines whether the option @samp{-@var{name}}
62: takes arguments. The value should be the number of arguments that
63: option takes--zero, for many options. This macro rather than
64: @code{SWITCH_TAKES_ARG} is used for multi-character option names.
65:
66: By default, this macro is defined as
67: @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
68: properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
69: wish to add additional options which take arguments. Any redefinition
70: should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
71: additional options.
72:
73: @findex SWITCHES_NEED_SPACES
74: @item SWITCHES_NEED_SPACES
75: A string-valued C expression which is nonempty if the linker needs a
76: space between the @samp{-L} or @samp{-o} option and its argument.
77:
78: If this macro is not defined, the default value is 0.
79:
80: @findex CPP_SPEC
81: @item CPP_SPEC
82: A C string constant that tells the GNU CC driver program options to
83: pass to CPP. It can also specify how to translate options you
84: give to GNU CC into options for GNU CC to pass to the CPP.
85:
86: Do not define this macro if it does not need to do anything.
87:
88: @findex NO_BUILTIN_SIZE_TYPE
89: @item NO_BUILTIN_SIZE_TYPE
90: If this macro is defined, the preprocessor will not define the builtin macro
91: @code{__SIZE_TYPE__}. The macro @code{__SIZE_TYPE__} must then be defined
92: by @code{CPP_SPEC} instead.
93:
94: This should be defined if @code{SIZE_TYPE} depends on target dependent flags
95: which are not accessible to the preprocessor. Otherwise, it should not
96: be defined.
97:
98: @findex NO_BUILTIN_PTRDIFF_TYPE
99: @item NO_BUILTIN_PTRDIFF_TYPE
100: If this macro is defined, the preprocessor will not define the builtin macro
101: @code{__PTRDIFF_TYPE__}. The macro @code{__PTRDIFF_TYPE__} must then be
102: defined by @code{CPP_SPEC} instead.
103:
104: This should be defined if @code{PTRDIFF_TYPE} depends on target dependent flags
105: which are not accessible to the preprocessor. Otherwise, it should not
106: be defined.
107:
108: @findex SIGNED_CHAR_SPEC
109: @item SIGNED_CHAR_SPEC
110: A C string constant that tells the GNU CC driver program options to
111: pass to CPP. By default, this macro is defined to pass the option
112: @samp{-D__CHAR_UNSIGNED__} to CPP if @code{char} will be treated as
113: @code{unsigned char} by @code{cc1}.
114:
115: Do not define this macro unless you need to override the default
116: definition.
117:
118: @findex CC1_SPEC
119: @item CC1_SPEC
120: A C string constant that tells the GNU CC driver program options to
121: pass to @code{cc1}. It can also specify how to translate options you
122: give to GNU CC into options for GNU CC to pass to the @code{cc1}.
123:
124: Do not define this macro if it does not need to do anything.
125:
126: @findex CC1PLUS_SPEC
127: @item CC1PLUS_SPEC
128: A C string constant that tells the GNU CC driver program options to
129: pass to @code{cc1plus}. It can also specify how to translate options you
130: give to GNU CC into options for GNU CC to pass to the @code{cc1plus}.
131:
132: Do not define this macro if it does not need to do anything.
133:
134: @findex ASM_SPEC
135: @item ASM_SPEC
136: A C string constant that tells the GNU CC driver program options to
137: pass to the assembler. It can also specify how to translate options
138: you give to GNU CC into options for GNU CC to pass to the assembler.
139: See the file @file{sun3.h} for an example of this.
140:
141: Do not define this macro if it does not need to do anything.
142:
143: @findex ASM_FINAL_SPEC
144: @item ASM_FINAL_SPEC
145: A C string constant that tells the GNU CC driver program how to
146: run any programs which cleanup after the normal assembler.
147: Normally, this is not needed. See the file @file{mips.h} for
148: an example of this.
149:
150: Do not define this macro if it does not need to do anything.
151:
152: @findex LINK_SPEC
153: @item LINK_SPEC
154: A C string constant that tells the GNU CC driver program options to
155: pass to the linker. It can also specify how to translate options you
156: give to GNU CC into options for GNU CC to pass to the linker.
157:
158: Do not define this macro if it does not need to do anything.
159:
160: @findex LIB_SPEC
161: @item LIB_SPEC
162: Another C string constant used much like @code{LINK_SPEC}. The difference
163: between the two is that @code{LIB_SPEC} is used at the end of the
164: command given to the linker.
165:
166: If this macro is not defined, a default is provided that
167: loads the standard C library from the usual place. See @file{gcc.c}.
168:
169: @findex STARTFILE_SPEC
170: @item STARTFILE_SPEC
171: Another C string constant used much like @code{LINK_SPEC}. The
172: difference between the two is that @code{STARTFILE_SPEC} is used at
173: the very beginning of the command given to the linker.
174:
175: If this macro is not defined, a default is provided that loads the
176: standard C startup file from the usual place. See @file{gcc.c}.
177:
178: @findex ENDFILE_SPEC
179: @item ENDFILE_SPEC
180: Another C string constant used much like @code{LINK_SPEC}. The
181: difference between the two is that @code{ENDFILE_SPEC} is used at
182: the very end of the command given to the linker.
183:
184: Do not define this macro if it does not need to do anything.
185:
186: @findex LINK_LIBGCC_SPECIAL
187: @item LINK_LIBGCC_SPECIAL
188: Define this macro meaning that @code{gcc} should find the library
189: @file{libgcc.a} by hand, rather than passing the argument @samp{-lgcc}
190: to tell the linker to do the search; also, @code{gcc} should not
191: generate @samp{-L} options to pass to the linker (as it normally does).
192:
193: @findex LINK_LIBGCC_SPECIAL_1
194: @item LINK_LIBGCC_SPECIAL_1
195: Define this macro meaning that @code{gcc} should find the
196: library @file{libgcc.a} by hand, rather than passing the argument
197: @samp{-lgcc} to tell the linker to do the search.
198:
199: @findex RELATIVE_PREFIX_NOT_LINKDIR
200: @item RELATIVE_PREFIX_NOT_LINKDIR
201: Define this macro to tell @code{gcc} that it should only translate
202: a @samp{-B} prefix into a @samp{-L} linker option if the prefix
203: indicates an absolute file name.
204:
205: @findex STANDARD_EXEC_PREFIX
206: @item STANDARD_EXEC_PREFIX
207: Define this macro as a C string constant if you wish to override the
208: standard choice of @file{/usr/local/lib/gcc-lib/} as the default prefix to
209: try when searching for the executable files of the compiler.
210:
211: @findex MD_EXEC_PREFIX
212: @item MD_EXEC_PREFIX
213: If defined, this macro is an additional prefix to try after
214: @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
215: when the @samp{-b} option is used, or the compiler is built as a cross
216: compiler.
217:
218: @findex STANDARD_STARTFILE_PREFIX
219: @item STANDARD_STARTFILE_PREFIX
220: Define this macro as a C string constant if you wish to override the
221: standard choice of @file{/usr/local/lib/} as the default prefix to
222: try when searching for startup files such as @file{crt0.o}.
223:
224: @findex MD_STARTFILE_PREFIX
225: @item MD_STARTFILE_PREFIX
226: If defined, this macro supplies an additional prefix to try after the
227: standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
228: @samp{-b} option is used, or when the compiler is built as a cross
229: compiler.
230:
231: @findex MD_STARTFILE_PREFIX_1
232: @item MD_STARTFILE_PREFIX_1
233: If defined, this macro supplies yet another prefix to try after the
234: standard prefixes. It is not searched when the @samp{-b} option is
235: used, or when the compiler is built as a cross compiler.
236:
237: @findex LOCAL_INCLUDE_DIR
238: @item LOCAL_INCLUDE_DIR
239: Define this macro as a C string constant if you wish to override the
240: standard choice of @file{/usr/local/include} as the default prefix to
241: try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
242: comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
243:
244: Cross compilers do not use this macro and do not search either
245: @file{/usr/local/include} or its replacement.
246:
247: @findex SYSTEM_INCLUDE_DIR
248: @item SYSTEM_INCLUDE_DIR
249: Define this macro as a C string constant if you wish to specify a
250: system-specific directory to search for header files before the standard
251: directory. @code{SYSTEM_INCLUDE_DIR} comes before
252: @code{STANDARD_INCLUDE_DIR} in the search order.
253:
254: Cross compilers do not use this macro and do not search the directory
255: specified.
256:
257: @findex STANDARD_INCLUDE_DIR
258: @item STANDARD_INCLUDE_DIR
259: Define this macro as a C string constant if you wish to override the
260: standard choice of @file{/usr/include} as the default prefix to
261: try when searching for header files.
262:
263: Cross compilers do not use this macro and do not search either
264: @file{/usr/include} or its replacement.
265:
266: @findex INCLUDE_DEFAULTS
267: @item INCLUDE_DEFAULTS
268: Define this macro if you wish to override the entire default search path
269: for include files. The default search path includes
270: @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
271: @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
272: @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
273: and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
274: and specify private search areas for GCC. The directory
275: @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
276:
277: The definition should be an initializer for an array of structures.
278: Each array element should have two elements: the directory name (a
279: string constant) and a flag for C++-only directories. Mark the end of
280: the array with a null element. For example, here is the definition used
281: for VMS:
282:
283: @example
284: #define INCLUDE_DEFAULTS \
285: @{ \
286: @{ "GNU_GXX_INCLUDE:", 1@}, \
287: @{ "GNU_CC_INCLUDE:", 0@}, \
288: @{ "SYS$SYSROOT:[SYSLIB.]", 0@}, \
289: @{ ".", 0@}, \
290: @{ 0, 0@} \
291: @}
292: @end example
293: @end table
294:
295: Here is the order of prefixes tried for exec files:
296:
297: @enumerate
298: @item
299: Any prefixes specified by the user with @samp{-B}.
300:
301: @item
302: The environment variable @code{GCC_EXEC_PREFIX}, if any.
303:
304: @item
305: The directories specified by the environment variable @code{COMPILER_PATH}.
306:
307: @item
308: The macro @code{STANDARD_EXEC_PREFIX}.
309:
310: @item
311: @file{/usr/lib/gcc/}.
312:
313: @item
314: The macro @code{MD_EXEC_PREFIX}, if any.
315: @end enumerate
316:
317: Here is the order of prefixes tried for startfiles:
318:
319: @enumerate
320: @item
321: Any prefixes specified by the user with @samp{-B}.
322:
323: @item
324: The environment variable @code{GCC_EXEC_PREFIX}, if any.
325:
326: @item
327: The directories specified by the environment variable @code{LIBRARY_PATH}.
328:
329: @item
330: The macro @code{STANDARD_EXEC_PREFIX}.
331:
332: @item
333: @file{/usr/lib/gcc/}.
334:
335: @item
336: The macro @code{MD_EXEC_PREFIX}, if any.
337:
338: @item
339: The macro @code{MD_STARTFILE_PREFIX}, if any.
340:
341: @item
342: The macro @code{STANDARD_STARTFILE_PREFIX}.
343:
344: @item
345: @file{/lib/}.
346:
347: @item
348: @file{/usr/lib/}.
349: @end enumerate
350:
351: @node Run-time Target
352: @section Run-time Target Specification
353: @cindex run-time target specification
354: @cindex predefined macros
355: @cindex target specifications
356:
357: @table @code
358: @findex CPP_PREDEFINES
359: @item CPP_PREDEFINES
360: Define this to be a string constant containing @samp{-D} options to
361: define the predefined macros that identify this machine and system.
362: These macros will be predefined unless the @samp{-ansi} option is
363: specified.
364:
365: In addition, a parallel set of macros are predefined, whose names are
366: made by appending @samp{__} at the beginning and at the end. These
367: @samp{__} macros are permitted by the ANSI standard, so they are
368: predefined regardless of whether @samp{-ansi} is specified.
369:
370: For example, on the Sun, one can use the following value:
371:
372: @smallexample
373: "-Dmc68000 -Dsun -Dunix"
374: @end smallexample
375:
376: The result is to define the macros @code{__mc68000__}, @code{__sun__}
377: and @code{__unix__} unconditionally, and the macros @code{mc68000},
378: @code{sun} and @code{unix} provided @samp{-ansi} is not specified.
379:
380: @findex STDC_VALUE
381: @item STDC_VALUE
382: Define the value to be assigned to the built-in macro @code{__STDC__}.
383: The default is the value @samp{1}.
384:
385: @findex extern int target_flags
386: @item extern int target_flags;
387: This declaration should be present.
388:
389: @cindex optional hardware or system features
390: @cindex features, optional, in system conventions
391: @item TARGET_@dots{}
392: This series of macros is to allow compiler command arguments to
393: enable or disable the use of optional features of the target machine.
394: For example, one machine description serves both the 68000 and
395: the 68020; a command argument tells the compiler whether it should
396: use 68020-only instructions or not. This command argument works
397: by means of a macro @code{TARGET_68020} that tests a bit in
398: @code{target_flags}.
399:
400: Define a macro @code{TARGET_@var{featurename}} for each such option.
401: Its definition should test a bit in @code{target_flags}; for example:
402:
403: @smallexample
404: #define TARGET_68020 (target_flags & 1)
405: @end smallexample
406:
407: One place where these macros are used is in the condition-expressions
408: of instruction patterns. Note how @code{TARGET_68020} appears
409: frequently in the 68000 machine description file, @file{m68k.md}.
410: Another place they are used is in the definitions of the other
411: macros in the @file{@var{machine}.h} file.
412:
413: @findex TARGET_SWITCHES
414: @item TARGET_SWITCHES
415: This macro defines names of command options to set and clear
416: bits in @code{target_flags}. Its definition is an initializer
417: with a subgrouping for each command option.
418:
419: Each subgrouping contains a string constant, that defines the option
420: name, and a number, which contains the bits to set in
421: @code{target_flags}. A negative number says to clear bits instead;
422: the negative of the number is which bits to clear. The actual option
423: name is made by appending @samp{-m} to the specified name.
424:
425: One of the subgroupings should have a null string. The number in
426: this grouping is the default value for @code{target_flags}. Any
427: target options act starting with that value.
428:
429: Here is an example which defines @samp{-m68000} and @samp{-m68020}
430: with opposite meanings, and picks the latter as the default:
431:
432: @smallexample
433: #define TARGET_SWITCHES \
434: @{ @{ "68020", 1@}, \
435: @{ "68000", -1@}, \
436: @{ "", 1@}@}
437: @end smallexample
438:
439: @findex TARGET_OPTIONS
440: @item TARGET_OPTIONS
441: This macro is similar to @code{TARGET_SWITCHES} but defines names of command
442: options that have values. Its definition is an initializer with a
443: subgrouping for each command option.
444:
445: Each subgrouping contains a string constant, that defines the fixed part
446: of the option name, and the address of a variable. The variable, type
447: @code{char *}, is set to the variable part of the given option if the fixed
448: part matches. The actual option name is made by appending @samp{-m} to the
449: specified name.
450:
451: Here is an example which defines @samp{-mshort-data-@var{number}}. If the
452: given option is @samp{-mshort-data-512}, the variable @code{m88k_short_data}
453: will be set to the string @code{"512"}.
454:
455: @smallexample
456: extern char *m88k_short_data;
457: #define TARGET_OPTIONS \
458: @{ @{ "short-data-", &m88k_short_data @} @}
459: @end smallexample
460:
461: @findex TARGET_VERSION
462: @item TARGET_VERSION
463: This macro is a C statement to print on @code{stderr} a string
464: describing the particular machine description choice. Every machine
465: description should define @code{TARGET_VERSION}. For example:
466:
467: @smallexample
468: #ifdef MOTOROLA
469: #define TARGET_VERSION \
470: fprintf (stderr, " (68k, Motorola syntax)");
471: #else
472: #define TARGET_VERSION \
473: fprintf (stderr, " (68k, MIT syntax)");
474: #endif
475: @end smallexample
476:
477: @findex OVERRIDE_OPTIONS
478: @item OVERRIDE_OPTIONS
479: Sometimes certain combinations of command options do not make sense on
480: a particular target machine. You can define a macro
481: @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
482: defined, is executed once just after all the command options have been
483: parsed.
484:
485: Don't use this macro to turn on various extra optimizations for
486: @samp{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
487:
488: @findex OPTIMIZATION_OPTIONS
489: @item OPTIMIZATION_OPTIONS (@var{level})
490: Some machines may desire to change what optimizations are performed for
491: various optimization levels. This macro, if defined, is executed once
492: just after the optimization level is determined and before the remainder
493: of the command options have been parsed. Values set in this macro are
494: used as the default values for the other command line options.
495:
496: @var{level} is the optimization level specified; 2 if -O2 is specified,
497: 1 if -O is specified, and 0 if neither is specified.
498:
499: @strong{Do not examine @code{write_symbols} in this macro!}
500: The debugging options are not supposed to alter the generated code.
501: @end table
502:
503: @node Storage Layout
504: @section Storage Layout
505: @cindex storage layout
506:
507: Note that the definitions of the macros in this table which are sizes or
508: alignments measured in bits do not need to be constant. They can be C
509: expressions that refer to static variables, such as the @code{target_flags}.
510: @xref{Run-time Target}.
511:
512: @table @code
513: @findex BITS_BIG_ENDIAN
514: @item BITS_BIG_ENDIAN
515: Define this macro to be the value 1 if the most significant bit in a
516: byte has the lowest number; otherwise define it to be the value zero.
517: This means that bit-field instructions count from the most significant
518: bit. If the machine has no bit-field instructions, then this must still
519: be defined, but it doesn't matter which value it is defined to.
520:
521: This macro does not affect the way structure fields are packed into
522: bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
523:
524: @findex BYTES_BIG_ENDIAN
525: @item BYTES_BIG_ENDIAN
526: Define this macro to be 1 if the most significant byte in a word has the
527: lowest number.
528:
529: @findex WORDS_BIG_ENDIAN
530: @item WORDS_BIG_ENDIAN
531: Define this macro to be 1 if, in a multiword object, the most
532: significant word has the lowest number. This applies to both memory
533: locations and registers; GNU CC fundamentally assumes that the order of
534: words in memory is the same as the order in registers.
535:
536: @findex FLOAT_WORDS_BIG_ENDIAN
537: @item FLOAT_WORDS_BIG_ENDIAN
538: Define this macro to be 1 if @code{DFmode}, @code{XFmode} or
539: @code{TFmode} floating point numbers are stored in memory with the word
540: containing the sign bit at the lowest address; otherwise define it to be
541: 0.
542:
543: You need not define this macro if the ordering is the same as for
544: multi-word integers.
545:
546: @findex BITS_PER_UNIT
547: @item BITS_PER_UNIT
548: Define this macro to be the number of bits in an addressable storage
549: unit (byte); normally 8.
550:
551: @findex BITS_PER_WORD
552: @item BITS_PER_WORD
553: Number of bits in a word; normally 32.
554:
555: @findex MAX_BITS_PER_WORD
556: @item MAX_BITS_PER_WORD
557: Maximum number of bits in a word. If this is undefined, the default is
558: @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
559: largest value that @code{BITS_PER_WORD} can have at run-time.
560:
561: @findex UNITS_PER_WORD
562: @item UNITS_PER_WORD
563: Number of storage units in a word; normally 4.
564:
565: @findex MAX_UNITS_PER_WORD
566: @item MAX_UNITS_PER_WORD
567: Maximum number of units in a word. If this is undefined, the default is
568: @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
569: largest value that @code{UNITS_PER_WORD} can have at run-time.
570:
571: @findex POINTER_SIZE
572: @item POINTER_SIZE
573: Width of a pointer, in bits.
574:
575: @findex PROMOTE_MODE
576: @item PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
577: A macro to update @var{m} and @var{unsignedp} when an object whose type
578: is @var{type} and which has the specified mode and signedness is to be
579: stored in a register. This macro is only called when @var{type} is a
580: scalar type.
581:
582: On most RISC machines, which only have operations that operate on a full
583: register, define this macro to set @var{m} to @code{word_mode} if
584: @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
585: cases, only integer modes should be widened because wider-precision
586: floating-point operations are usually more expensive than their narrower
587: counterparts.
588:
589: For most machines, the macro definition does not change @var{unsignedp}.
590: However, some machines, have instructions that preferentially handle
591: either signed or unsigned quantities of certain modes. For example, on
592: the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
593: sign-extend the result to 64 bits. On such machines, set
594: @var{unsignedp} according to which kind of extension is more efficient.
595:
596: Do not define this macro if it would never modify @var{m}.
597:
598: @findex PROMOTE_FUNCTION_ARGS
599: @item PROMOTE_FUNCTION_ARGS
600: Define this macro if the promotion described by @code{PROMOTE_MODE}
601: should also be done for outgoing function arguments.
602:
603: @findex PROMOTE_FUNCTION_RETURN
604: @item PROMOTE_FUNCTION_RETURN
605: Define this macro if the promotion described by @code{PROMOTE_MODE}
606: should also be done for the return value of functions.
607:
608: If this macro is defined, @code{FUNCTION_VALUE} must perform the same
609: promotions done by @code{PROMOTE_MODE}.
610:
611: @findex PARM_BOUNDARY
612: @item PARM_BOUNDARY
613: Normal alignment required for function parameters on the stack, in
614: bits. All stack parameters receive at least this much alignment
615: regardless of data type. On most machines, this is the same as the
616: size of an integer.
617:
618: @findex STACK_BOUNDARY
619: @item STACK_BOUNDARY
620: Define this macro if you wish to preserve a certain alignment for
621: the stack pointer. The definition is a C expression
622: for the desired alignment (measured in bits).
623:
624: @cindex @code{PUSH_ROUNDING}, interaction with @code{STACK_BOUNDARY}
625: If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned
626: to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies a
627: less strict alignment than @code{STACK_BOUNDARY}, the stack may be
628: momentarily unaligned while pushing arguments.
629:
630: @findex FUNCTION_BOUNDARY
631: @item FUNCTION_BOUNDARY
632: Alignment required for a function entry point, in bits.
633:
634: @findex BIGGEST_ALIGNMENT
635: @item BIGGEST_ALIGNMENT
636: Biggest alignment that any data type can require on this machine, in bits.
637:
638: @findex BIGGEST_FIELD_ALIGNMENT
639: @item BIGGEST_FIELD_ALIGNMENT
640: Biggest alignment that any structure field can require on this machine,
641: in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
642: structure fields only.
643:
644: @findex MAX_OFILE_ALIGNMENT
645: @item MAX_OFILE_ALIGNMENT
646: Biggest alignment supported by the object file format of this machine.
647: Use this macro to limit the alignment which can be specified using the
648: @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
649: the default value is @code{BIGGEST_ALIGNMENT}.
650:
651: @findex DATA_ALIGNMENT
652: @item DATA_ALIGNMENT (@var{type}, @var{basic-align})
653: If defined, a C expression to compute the alignment for a static
654: variable. @var{type} is the data type, and @var{basic-align} is the
655: alignment that the object would ordinarily have. The value of this
656: macro is used instead of that alignment to align the object.
657:
658: If this macro is not defined, then @var{basic-align} is used.
659:
660: @findex strcpy
661: One use of this macro is to increase alignment of medium-size data to
662: make it all fit in fewer cache lines. Another is to cause character
663: arrays to be word-aligned so that @code{strcpy} calls that copy
664: constants to character arrays can be done inline.
665:
666: @findex CONSTANT_ALIGNMENT
667: @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
668: If defined, a C expression to compute the alignment given to a constant
669: that is being placed in memory. @var{constant} is the constant and
670: @var{basic-align} is the alignment that the object would ordinarily
671: have. The value of this macro is used instead of that alignment to
672: align the object.
673:
674: If this macro is not defined, then @var{basic-align} is used.
675:
676: The typical use of this macro is to increase alignment for string
677: constants to be word aligned so that @code{strcpy} calls that copy
678: constants can be done inline.
679:
680: @findex EMPTY_FIELD_BOUNDARY
681: @item EMPTY_FIELD_BOUNDARY
682: Alignment in bits to be given to a structure bit field that follows an
683: empty field such as @code{int : 0;}.
684:
685: Note that @code{PCC_BITFIELD_TYPE_MATTERS} also affects the alignment
686: that results from an empty field.
687:
688: @findex STRUCTURE_SIZE_BOUNDARY
689: @item STRUCTURE_SIZE_BOUNDARY
690: Number of bits which any structure or union's size must be a multiple of.
691: Each structure or union's size is rounded up to a multiple of this.
692:
693: If you do not define this macro, the default is the same as
694: @code{BITS_PER_UNIT}.
695:
696: @findex STRICT_ALIGNMENT
697: @item STRICT_ALIGNMENT
698: Define this macro to be the value 1 if instructions will fail to work
699: if given data not on the nominal alignment. If instructions will merely
700: go slower in that case, define this macro as 0.
701:
702: @findex PCC_BITFIELD_TYPE_MATTERS
703: @item PCC_BITFIELD_TYPE_MATTERS
704: Define this if you wish to imitate the way many other C compilers handle
705: alignment of bitfields and the structures that contain them.
706:
707: The behavior is that the type written for a bitfield (@code{int},
708: @code{short}, or other integer type) imposes an alignment for the
709: entire structure, as if the structure really did contain an ordinary
710: field of that type. In addition, the bitfield is placed within the
711: structure so that it would fit within such a field, not crossing a
712: boundary for it.
713:
714: Thus, on most machines, a bitfield whose type is written as @code{int}
715: would not cross a four-byte boundary, and would force four-byte
716: alignment for the whole structure. (The alignment used may not be four
717: bytes; it is controlled by the other alignment parameters.)
718:
719: If the macro is defined, its definition should be a C expression;
720: a nonzero value for the expression enables this behavior.
721:
722: Note that if this macro is not defined, or its value is zero, some
723: bitfields may cross more than one alignment boundary. The compiler can
724: support such references if there are @samp{insv}, @samp{extv}, and
725: @samp{extzv} insns that can directly reference memory.
726:
727: The other known way of making bitfields work is to define
728: @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
729: Then every structure can be accessed with fullwords.
730:
731: Unless the machine has bitfield instructions or you define
732: @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
733: @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
734:
735: If your aim is to make GNU CC use the same conventions for laying out
736: bitfields as are used by another compiler, here is how to investigate
737: what the other compiler does. Compile and run this program:
738:
739: @example
740: struct foo1
741: @{
742: char x;
743: char :0;
744: char y;
745: @};
746:
747: struct foo2
748: @{
749: char x;
750: int :0;
751: char y;
752: @};
753:
754: main ()
755: @{
756: printf ("Size of foo1 is %d\n",
757: sizeof (struct foo1));
758: printf ("Size of foo2 is %d\n",
759: sizeof (struct foo2));
760: exit (0);
761: @}
762: @end example
763:
764: If this prints 2 and 5, then the compiler's behavior is what you would
765: get from @code{PCC_BITFIELD_TYPE_MATTERS}.
766:
767: @findex BITFIELD_NBYTES_LIMITED
768: @item BITFIELD_NBYTES_LIMITED
769: Like PCC_BITFIELD_TYPE_MATTERS except that its effect is limited to
770: aligning a bitfield within the structure.
771:
772: @findex ROUND_TYPE_SIZE
773: @item ROUND_TYPE_SIZE (@var{struct}, @var{size}, @var{align})
774: Define this macro as an expression for the overall size of a structure
775: (given by @var{struct} as a tree node) when the size computed from the
776: fields is @var{size} and the alignment is @var{align}.
777:
778: The default is to round @var{size} up to a multiple of @var{align}.
779:
780: @findex ROUND_TYPE_ALIGN
781: @item ROUND_TYPE_ALIGN (@var{struct}, @var{computed}, @var{specified})
782: Define this macro as an expression for the alignment of a structure
783: (given by @var{struct} as a tree node) if the alignment computed in the
784: usual way is @var{computed} and the alignment explicitly specified was
785: @var{specified}.
786:
787: The default is to use @var{specified} if it is larger; otherwise, use
788: the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
789:
790: @findex MAX_FIXED_MODE_SIZE
791: @item MAX_FIXED_MODE_SIZE
792: An integer expression for the size in bits of the largest integer
793: machine mode that should actually be used. All integer machine modes of
794: this size or smaller can be used for structures and unions with the
795: appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
796: (DImode)} is assumed.
797:
798: @findex CHECK_FLOAT_VALUE
799: @item CHECK_FLOAT_VALUE (@var{mode}, @var{value})
800: A C statement to validate the value @var{value} (of type
801: @code{double}) for mode @var{mode}. This means that you check whether
802: @var{value} fits within the possible range of values for mode
803: @var{mode} on this target machine. The mode @var{mode} is always
804: @code{SFmode} or @code{DFmode}.
805:
806: @findex error
807: If @var{value} is not valid, you should call @code{error} to print an
808: error message and then assign some valid value to @var{value}.
809: Allowing an invalid value to go through the compiler can produce
810: incorrect assembler code which may even cause Unix assemblers to
811: crash.
812:
813: This macro need not be defined if there is no work for it to do.
814:
815: @findex TARGET_FLOAT_FORMAT
816: @item TARGET_FLOAT_FORMAT
817: A code distinguishing the floating point format of the target machine.
818: There are three defined values:
819:
820: @table @code
821: @findex IEEE_FLOAT_FORMAT
822: @item IEEE_FLOAT_FORMAT
823: This code indicates IEEE floating point. It is the default; there is no
824: need to define this macro when the format is IEEE.
825:
826: @findex VAX_FLOAT_FORMAT
827: @item VAX_FLOAT_FORMAT
828: This code indicates the peculiar format used on the Vax.
829:
830: @findex UNKNOWN_FLOAT_FORMAT
831: @item UNKNOWN_FLOAT_FORMAT
832: This code indicates any other format.
833: @end table
834:
835: The value of this macro is compared with @code{HOST_FLOAT_FORMAT}
836: (@pxref{Config}) to determine whether the target machine has the same
837: format as the host machine. If any other formats are actually in use on
838: supported machines, new codes should be defined for them.
839:
840: The ordering of the component words of floating point values stored in
841: memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN} for the target
842: machine and @code{HOST_FLOAT_WORDS_BIG_ENDIAN} for the host.
843: @end table
844:
845: @node Type Layout
846: @section Layout of Source Language Data Types
847:
848: These macros define the sizes and other characteristics of the standard
849: basic data types used in programs being compiled. Unlike the macros in
850: the previous section, these apply to specific features of C and related
851: languages, rather than to fundamental aspects of storage layout.
852:
853: @table @code
854: @findex INT_TYPE_SIZE
855: @item INT_TYPE_SIZE
856: A C expression for the size in bits of the type @code{int} on the
857: target machine. If you don't define this, the default is one word.
858:
859: @findex MAX_INT_TYPE_SIZE
860: @item MAX_INT_TYPE_SIZE
861: Maximum number for the size in bits of the type @code{int} on the target
862: machine. If this is undefined, the default is @code{INT_TYPE_SIZE}.
863: Otherwise, it is the constant value that is the largest value that
864: @code{INT_TYPE_SIZE} can have at run-time. This is used in @code{cpp}.
865:
866: @findex SHORT_TYPE_SIZE
867: @item SHORT_TYPE_SIZE
868: A C expression for the size in bits of the type @code{short} on the
869: target machine. If you don't define this, the default is half a word.
870: (If this would be less than one storage unit, it is rounded up to one
871: unit.)
872:
873: @findex LONG_TYPE_SIZE
874: @item LONG_TYPE_SIZE
875: A C expression for the size in bits of the type @code{long} on the
876: target machine. If you don't define this, the default is one word.
877:
878: @findex MAX_LONG_TYPE_SIZE
879: @item MAX_LONG_TYPE_SIZE
880: Maximum number for the size in bits of the type @code{long} on the
881: target machine. If this is undefined, the default is
882: @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the
883: largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is
884: used in @code{cpp}.
885:
886: @findex LONG_LONG_TYPE_SIZE
887: @item LONG_LONG_TYPE_SIZE
888: A C expression for the size in bits of the type @code{long long} on the
889: target machine. If you don't define this, the default is two
890: words.
891:
892: @findex CHAR_TYPE_SIZE
893: @item CHAR_TYPE_SIZE
894: A C expression for the size in bits of the type @code{char} on the
895: target machine. If you don't define this, the default is one quarter
896: of a word. (If this would be less than one storage unit, it is rounded up
897: to one unit.)
898:
899: @findex MAX_CHAR_TYPE_SIZE
900: @item MAX_CHAR_TYPE_SIZE
901: Maximum number for the size in bits of the type @code{char} on the
902: target machine. If this is undefined, the default is
903: @code{CHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
904: largest value that @code{CHAR_TYPE_SIZE} can have at run-time. This is
905: used in @code{cpp}.
906:
907: @findex FLOAT_TYPE_SIZE
908: @item FLOAT_TYPE_SIZE
909: A C expression for the size in bits of the type @code{float} on the
910: target machine. If you don't define this, the default is one word.
911:
912: @findex DOUBLE_TYPE_SIZE
913: @item DOUBLE_TYPE_SIZE
914: A C expression for the size in bits of the type @code{double} on the
915: target machine. If you don't define this, the default is two
916: words.
917:
918: @findex LONG_DOUBLE_TYPE_SIZE
919: @item LONG_DOUBLE_TYPE_SIZE
920: A C expression for the size in bits of the type @code{long double} on
921: the target machine. If you don't define this, the default is two
922: words.
923:
924: @findex DEFAULT_SIGNED_CHAR
925: @item DEFAULT_SIGNED_CHAR
926: An expression whose value is 1 or 0, according to whether the type
927: @code{char} should be signed or unsigned by default. The user can
928: always override this default with the options @samp{-fsigned-char}
929: and @samp{-funsigned-char}.
930:
931: @findex DEFAULT_SHORT_ENUMS
932: @item DEFAULT_SHORT_ENUMS
933: A C expression to determine whether to give an @code{enum} type
934: only as many bytes as it takes to represent the range of possible values
935: of that type. A nonzero value means to do that; a zero value means all
936: @code{enum} types should be allocated like @code{int}.
937:
938: If you don't define the macro, the default is 0.
939:
940: @findex SIZE_TYPE
941: @item SIZE_TYPE
942: A C expression for a string describing the name of the data type to use
943: for size values. The typedef name @code{size_t} is defined using the
944: contents of the string.
945:
946: The string can contain more than one keyword. If so, separate them with
947: spaces, and write first any length keyword, then @code{unsigned} if
948: appropriate, and finally @code{int}. The string must exactly match one
949: of the data type names defined in the function
950: @code{init_decl_processing} in the file @file{c-decl.c}. You may not
951: omit @code{int} or change the order---that would cause the compiler to
952: crash on startup.
953:
954: If you don't define this macro, the default is @code{"long unsigned
955: int"}.
956:
957: @findex PTRDIFF_TYPE
958: @item PTRDIFF_TYPE
959: A C expression for a string describing the name of the data type to use
960: for the result of subtracting two pointers. The typedef name
961: @code{ptrdiff_t} is defined using the contents of the string. See
962: @code{SIZE_TYPE} above for more information.
963:
964: If you don't define this macro, the default is @code{"long int"}.
965:
966: @findex WCHAR_TYPE
967: @item WCHAR_TYPE
968: A C expression for a string describing the name of the data type to use
969: for wide characters. The typedef name @code{wchar_t} is defined using
970: the contents of the string. See @code{SIZE_TYPE} above for more
971: information.
972:
973: If you don't define this macro, the default is @code{"int"}.
974:
975: @findex WCHAR_TYPE_SIZE
976: @item WCHAR_TYPE_SIZE
977: A C expression for the size in bits of the data type for wide
978: characters. This is used in @code{cpp}, which cannot make use of
979: @code{WCHAR_TYPE}.
980:
981: @findex MAX_WCHAR_TYPE_SIZE
982: @item MAX_WCHAR_TYPE_SIZE
983: Maximum number for the size in bits of the data type for wide
984: characters. If this is undefined, the default is
985: @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the
986: largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is
987: used in @code{cpp}.
988:
989: @findex OBJC_INT_SELECTORS
990: @item OBJC_INT_SELECTORS
991: Define this macro if the type of Objective C selectors should be
992: @code{int}.
993:
994: If this macro is not defined, then selectors should have the type
995: @code{struct objc_selector *}.
996:
997: @findex OBJC_SELECTORS_WITHOUT_LABELS
998: @item OBJC_SELECTORS_WITHOUT_LABELS
999: Define this macro if the compiler can group all the selectors together
1000: into a vector and use just one label at the beginning of the vector.
1001: Otherwise, the compiler must give each selector its own assembler
1002: label.
1003:
1004: On certain machines, it is important to have a separate label for each
1005: selector because this enables the linker to eliminate duplicate selectors.
1006:
1007: @findex TARGET_BELL
1008: @item TARGET_BELL
1009: A C constant expression for the integer value for escape sequence
1010: @samp{\a}.
1011:
1012: @findex TARGET_TAB
1013: @findex TARGET_BS
1014: @findex TARGET_NEWLINE
1015: @item TARGET_BS
1016: @itemx TARGET_TAB
1017: @itemx TARGET_NEWLINE
1018: C constant expressions for the integer values for escape sequences
1019: @samp{\b}, @samp{\t} and @samp{\n}.
1020:
1021: @findex TARGET_VT
1022: @findex TARGET_FF
1023: @findex TARGET_CR
1024: @item TARGET_VT
1025: @itemx TARGET_FF
1026: @itemx TARGET_CR
1027: C constant expressions for the integer values for escape sequences
1028: @samp{\v}, @samp{\f} and @samp{\r}.
1029: @end table
1030:
1031: @node Registers
1032: @section Register Usage
1033: @cindex register usage
1034:
1035: This section explains how to describe what registers the target machine
1036: has, and how (in general) they can be used.
1037:
1038: The description of which registers a specific instruction can use is
1039: done with register classes; see @ref{Register Classes}. For information
1040: on using registers to access a stack frame, see @ref{Frame Registers}.
1041: For passing values in registers, see @ref{Register Arguments}.
1042: For returning values in registers, see @ref{Scalar Return}.
1043:
1044: @menu
1045: * Register Basics:: Number and kinds of registers.
1046: * Allocation Order:: Order in which registers are allocated.
1047: * Values in Registers:: What kinds of values each reg can hold.
1048: * Leaf Functions:: Renumbering registers for leaf functions.
1049: * Stack Registers:: Handling a register stack such as 80387.
1050: * Obsolete Register Macros:: Macros formerly used for the 80387.
1051: @end menu
1052:
1053: @node Register Basics
1054: @subsection Basic Characteristics of Registers
1055:
1056: @table @code
1057: @findex FIRST_PSEUDO_REGISTER
1058: @item FIRST_PSEUDO_REGISTER
1059: Number of hardware registers known to the compiler. They receive
1060: numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1061: pseudo register's number really is assigned the number
1062: @code{FIRST_PSEUDO_REGISTER}.
1063:
1064: @item FIXED_REGISTERS
1065: @findex FIXED_REGISTERS
1066: @cindex fixed register
1067: An initializer that says which registers are used for fixed purposes
1068: all throughout the compiled code and are therefore not available for
1069: general allocation. These would include the stack pointer, the frame
1070: pointer (except on machines where that can be used as a general
1071: register when no frame pointer is needed), the program counter on
1072: machines where that is considered one of the addressable registers,
1073: and any other numbered register with a standard use.
1074:
1075: This information is expressed as a sequence of numbers, separated by
1076: commas and surrounded by braces. The @var{n}th number is 1 if
1077: register @var{n} is fixed, 0 otherwise.
1078:
1079: The table initialized from this macro, and the table initialized by
1080: the following one, may be overridden at run time either automatically,
1081: by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1082: the user with the command options @samp{-ffixed-@var{reg}},
1083: @samp{-fcall-used-@var{reg}} and @samp{-fcall-saved-@var{reg}}.
1084:
1085: @findex CALL_USED_REGISTERS
1086: @item CALL_USED_REGISTERS
1087: @cindex call-used register
1088: @cindex call-clobbered register
1089: @cindex call-saved register
1090: Like @code{FIXED_REGISTERS} but has 1 for each register that is
1091: clobbered (in general) by function calls as well as for fixed
1092: registers. This macro therefore identifies the registers that are not
1093: available for general allocation of values that must live across
1094: function calls.
1095:
1096: If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1097: automatically saves it on function entry and restores it on function
1098: exit, if the register is used within the function.
1099:
1100: @findex CONDITIONAL_REGISTER_USAGE
1101: @findex fixed_regs
1102: @findex call_used_regs
1103: @item CONDITIONAL_REGISTER_USAGE
1104: Zero or more C statements that may conditionally modify two variables
1105: @code{fixed_regs} and @code{call_used_regs} (both of type @code{char
1106: []}) after they have been initialized from the two preceding macros.
1107:
1108: This is necessary in case the fixed or call-clobbered registers depend
1109: on target flags.
1110:
1111: You need not define this macro if it has no work to do.
1112:
1113: @cindex disabling certain registers
1114: @cindex controlling register usage
1115: If the usage of an entire class of registers depends on the target
1116: flags, you may indicate this to GCC by using this macro to modify
1117: @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1118: registers in the classes which should not be used by GCC. Also define
1119: the macro @code{REG_CLASS_FROM_LETTER} to return @code{NO_REGS} if it
1120: is called with a letter for a class that shouldn't be used.
1121:
1122: (However, if this class is not included in @code{GENERAL_REGS} and all
1123: of the insn patterns whose constraints permit this class are
1124: controlled by target switches, then GCC will automatically avoid using
1125: these registers when the target switches are opposed to them.)
1126:
1127: @findex NON_SAVING_SETJMP
1128: @item NON_SAVING_SETJMP
1129: If this macro is defined and has a nonzero value, it means that
1130: @code{setjmp} and related functions fail to save the registers, or that
1131: @code{longjmp} fails to restore them. To compensate, the compiler
1132: avoids putting variables in registers in functions that use
1133: @code{setjmp}.
1134:
1135: @findex INCOMING_REGNO
1136: @item INCOMING_REGNO (@var{out})
1137: Define this macro if the target machine has register windows. This C
1138: expression returns the register number as seen by the called function
1139: corresponding to the register number @var{out} as seen by the calling
1140: function. Return @var{out} if register number @var{out} is not an
1141: outbound register.
1142:
1143: @findex OUTGOING_REGNO
1144: @item OUTGOING_REGNO (@var{in})
1145: Define this macro if the target machine has register windows. This C
1146: expression returns the register number as seen by the calling function
1147: corresponding to the register number @var{in} as seen by the called
1148: function. Return @var{in} if register number @var{in} is not an inbound
1149: register.
1150:
1151: @ignore
1152: @findex PC_REGNUM
1153: @item PC_REGNUM
1154: If the program counter has a register number, define this as that
1155: register number. Otherwise, do not define it.
1156: @end ignore
1157: @end table
1158:
1159: @node Allocation Order
1160: @subsection Order of Allocation of Registers
1161: @cindex order of register allocation
1162: @cindex register allocation order
1163:
1164: @table @code
1165: @findex REG_ALLOC_ORDER
1166: @item REG_ALLOC_ORDER
1167: If defined, an initializer for a vector of integers, containing the
1168: numbers of hard registers in the order in which GNU CC should prefer
1169: to use them (from most preferred to least).
1170:
1171: If this macro is not defined, registers are used lowest numbered first
1172: (all else being equal).
1173:
1174: One use of this macro is on machines where the highest numbered
1175: registers must always be saved and the save-multiple-registers
1176: instruction supports only sequences of consecutive registers. On such
1177: machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1178: the highest numbered allocatable register first.
1179:
1180: @findex ORDER_REGS_FOR_LOCAL_ALLOC
1181: @item ORDER_REGS_FOR_LOCAL_ALLOC
1182: A C statement (sans semicolon) to choose the order in which to allocate
1183: hard registers for pseudo-registers local to a basic block.
1184:
1185: Store the desired register order in the array @code{reg_alloc_order}.
1186: Element 0 should be the register to allocate first; element 1, the next
1187: register; and so on.
1188:
1189: The macro body should not assume anything about the contents of
1190: @code{reg_alloc_order} before execution of the macro.
1191:
1192: On most machines, it is not necessary to define this macro.
1193: @end table
1194:
1195: @node Values in Registers
1196: @subsection How Values Fit in Registers
1197:
1198: This section discusses the macros that describe which kinds of values
1199: (specifically, which machine modes) each register can hold, and how many
1200: consecutive registers are needed for a given mode.
1201:
1202: @table @code
1203: @findex HARD_REGNO_NREGS
1204: @item HARD_REGNO_NREGS (@var{regno}, @var{mode})
1205: A C expression for the number of consecutive hard registers, starting
1206: at register number @var{regno}, required to hold a value of mode
1207: @var{mode}.
1208:
1209: On a machine where all registers are exactly one word, a suitable
1210: definition of this macro is
1211:
1212: @smallexample
1213: #define HARD_REGNO_NREGS(REGNO, MODE) \
1214: ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
1215: / UNITS_PER_WORD))
1216: @end smallexample
1217:
1218: @findex HARD_REGNO_MODE_OK
1219: @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
1220: A C expression that is nonzero if it is permissible to store a value
1221: of mode @var{mode} in hard register number @var{regno} (or in several
1222: registers starting with that one). For a machine where all registers
1223: are equivalent, a suitable definition is
1224:
1225: @smallexample
1226: #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
1227: @end smallexample
1228:
1229: It is not necessary for this macro to check for the numbers of fixed
1230: registers, because the allocation mechanism considers them to be always
1231: occupied.
1232:
1233: @cindex register pairs
1234: On some machines, double-precision values must be kept in even/odd
1235: register pairs. The way to implement that is to define this macro
1236: to reject odd register numbers for such modes.
1237:
1238: @ignore
1239: @c I think this is not true now
1240: GNU CC assumes that it can always move values between registers and
1241: (suitably addressed) memory locations. If it is impossible to move a
1242: value of a certain mode between memory and certain registers, then
1243: @code{HARD_REGNO_MODE_OK} must not allow this mode in those registers.
1244: @end ignore
1245:
1246: The minimum requirement for a mode to be OK in a register is that the
1247: @samp{mov@var{mode}} instruction pattern support moves between the
1248: register and any other hard register for which the mode is OK; and that
1249: moving a value into the register and back out not alter it.
1250:
1251: Since the same instruction used to move @code{SImode} will work for all
1252: narrower integer modes, it is not necessary on any machine for
1253: @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
1254: you define patterns @samp{movhi}, etc., to take advantage of this. This
1255: is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
1256: and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
1257: to be tieable.
1258:
1259: Many machines have special registers for floating point arithmetic.
1260: Often people assume that floating point machine modes are allowed only
1261: in floating point registers. This is not true. Any registers that
1262: can hold integers can safely @emph{hold} a floating point machine
1263: mode, whether or not floating arithmetic can be done on it in those
1264: registers. Integer move instructions can be used to move the values.
1265:
1266: On some machines, though, the converse is true: fixed-point machine
1267: modes may not go in floating registers. This is true if the floating
1268: registers normalize any value stored in them, because storing a
1269: non-floating value there would garble it. In this case,
1270: @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
1271: floating registers. But if the floating registers do not automatically
1272: normalize, if you can store any bit pattern in one and retrieve it
1273: unchanged without a trap, then any machine mode may go in a floating
1274: register, so you can define this macro to say so.
1275:
1276: On some machines, such as the Sparc and the Mips, we get better code
1277: by defining @code{HARD_REGNO_MODE_OK} to forbid integers in floating
1278: registers, even though the hardware is capable of handling them. This
1279: is because transferring values between floating registers and general
1280: registers is so slow that it is better to keep the integer in memory.
1281:
1282: The primary significance of special floating registers is rather that
1283: they are the registers acceptable in floating point arithmetic
1284: instructions. However, this is of no concern to
1285: @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
1286: constraints for those instructions.
1287:
1288: On some machines, the floating registers are especially slow to access,
1289: so that it is better to store a value in a stack frame than in such a
1290: register if floating point arithmetic is not being done. As long as the
1291: floating registers are not in class @code{GENERAL_REGS}, they will not
1292: be used unless some pattern's constraint asks for one.
1293:
1294: @findex MODES_TIEABLE_P
1295: @item MODES_TIEABLE_P (@var{mode1}, @var{mode2})
1296: A C expression that is nonzero if it is desirable to choose register
1297: allocation so as to avoid move instructions between a value of mode
1298: @var{mode1} and a value of mode @var{mode2}.
1299:
1300: If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
1301: @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are ever different
1302: for any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1},
1303: @var{mode2})} must be zero.
1304: @end table
1305:
1306: @node Leaf Functions
1307: @subsection Handling Leaf Functions
1308:
1309: @cindex leaf functions
1310: @cindex functions, leaf
1311: On some machines, a leaf function (i.e., one which makes no calls) can run
1312: more efficiently if it does not make its own register window. Often this
1313: means it is required to receive its arguments in the registers where they
1314: are passed by the caller, instead of the registers where they would
1315: normally arrive.
1316:
1317: The special treatment for leaf functions generally applies only when
1318: other conditions are met; for example, often they may use only those
1319: registers for its own variables and temporaries. We use the term ``leaf
1320: function'' to mean a function that is suitable for this special
1321: handling, so that functions with no calls are not necessarily ``leaf
1322: functions''.
1323:
1324: GNU CC assigns register numbers before it knows whether the function is
1325: suitable for leaf function treatment. So it needs to renumber the
1326: registers in order to output a leaf function. The following macros
1327: accomplish this.
1328:
1329: @table @code
1330: @findex LEAF_REGISTERS
1331: @item LEAF_REGISTERS
1332: A C initializer for a vector, indexed by hard register number, which
1333: contains 1 for a register that is allowable in a candidate for leaf
1334: function treatment.
1335:
1336: If leaf function treatment involves renumbering the registers, then the
1337: registers marked here should be the ones before renumbering---those that
1338: GNU CC would ordinarily allocate. The registers which will actually be
1339: used in the assembler code, after renumbering, should not be marked with 1
1340: in this vector.
1341:
1342: Define this macro only if the target machine offers a way to optimize
1343: the treatment of leaf functions.
1344:
1345: @findex LEAF_REG_REMAP
1346: @item LEAF_REG_REMAP (@var{regno})
1347: A C expression whose value is the register number to which @var{regno}
1348: should be renumbered, when a function is treated as a leaf function.
1349:
1350: If @var{regno} is a register number which should not appear in a leaf
1351: function before renumbering, then the expression should yield -1, which
1352: will cause the compiler to abort.
1353:
1354: Define this macro only if the target machine offers a way to optimize the
1355: treatment of leaf functions, and registers need to be renumbered to do
1356: this.
1357:
1358: @findex REG_LEAF_ALLOC_ORDER
1359: @item REG_LEAF_ALLOC_ORDER
1360: If defined, an initializer for a vector of integers, containing the
1361: numbers of hard registers in the order in which the GNU CC should prefer
1362: to use them (from most preferred to least) in a leaf function. If this
1363: macro is not defined, REG_ALLOC_ORDER is used for both non-leaf and
1364: leaf-functions.
1365: @c *important*.. is the above all-caps phrase supposed tobe in code
1366: @c font? --mew 2feb93
1367: @end table
1368:
1369: @findex leaf_function
1370: Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
1371: treat leaf functions specially. It can test the C variable
1372: @code{leaf_function} which is nonzero for leaf functions. (The variable
1373: @code{leaf_function} is defined only if @code{LEAF_REGISTERS} is
1374: defined.)
1375: @c changed this to fix overfull. ALSO: why the "it" at the beginning
1376: @c of the next paragraph?! --mew 2feb93
1377:
1378: @node Stack Registers
1379: @subsection Registers That Form a Stack
1380:
1381: There are special features to handle computers where some of the
1382: ``registers'' form a stack, as in the 80387 coprocessor for the 80386.
1383: Stack registers are normally written by pushing onto the stack, and are
1384: numbered relative to the top of the stack.
1385:
1386: Currently, GNU CC can only handle one group of stack-like registers, and
1387: they must be consecutively numbered.
1388:
1389: @table @code
1390: @findex STACK_REGS
1391: @item STACK_REGS
1392: Define this if the machine has any stack-like registers.
1393:
1394: @findex FIRST_STACK_REG
1395: @item FIRST_STACK_REG
1396: The number of the first stack-like register. This one is the top
1397: of the stack.
1398:
1399: @findex LAST_STACK_REG
1400: @item LAST_STACK_REG
1401: The number of the last stack-like register. This one is the bottom of
1402: the stack.
1403: @end table
1404:
1405: @node Obsolete Register Macros
1406: @subsection Obsolete Macros for Controlling Register Usage
1407:
1408: These features do not work very well. They exist because they used to
1409: be required to generate correct code for the 80387 coprocessor of the
1410: 80386. They are no longer used by that machine description and may be
1411: removed in a later version of the compiler. Don't use them!
1412:
1413: @table @code
1414: @findex OVERLAPPING_REGNO_P
1415: @item OVERLAPPING_REGNO_P (@var{regno})
1416: If defined, this is a C expression whose value is nonzero if hard
1417: register number @var{regno} is an overlapping register. This means a
1418: hard register which overlaps a hard register with a different number.
1419: (Such overlap is undesirable, but occasionally it allows a machine to
1420: be supported which otherwise could not be.) This macro must return
1421: nonzero for @emph{all} the registers which overlap each other. GNU CC
1422: can use an overlapping register only in certain limited ways. It can
1423: be used for allocation within a basic block, and may be spilled for
1424: reloading; that is all.
1425:
1426: If this macro is not defined, it means that none of the hard registers
1427: overlap each other. This is the usual situation.
1428:
1429: @findex INSN_CLOBBERS_REGNO_P
1430: @item INSN_CLOBBERS_REGNO_P (@var{insn}, @var{regno})
1431: If defined, this is a C expression whose value should be nonzero if
1432: the insn @var{insn} has the effect of mysteriously clobbering the
1433: contents of hard register number @var{regno}. By ``mysterious'' we
1434: mean that the insn's RTL expression doesn't describe such an effect.
1435:
1436: If this macro is not defined, it means that no insn clobbers registers
1437: mysteriously. This is the usual situation; all else being equal,
1438: it is best for the RTL expression to show all the activity.
1439:
1440: @cindex death notes
1441: @findex PRESERVE_DEATH_INFO_REGNO_P
1442: @item PRESERVE_DEATH_INFO_REGNO_P (@var{regno})
1443: If defined, this is a C expression whose value is nonzero if accurate
1444: @code{REG_DEAD} notes are needed for hard register number @var{regno}
1445: at the time of outputting the assembler code. When this is so, a few
1446: optimizations that take place after register allocation and could
1447: invalidate the death notes are not done when this register is
1448: involved.
1449:
1450: You would arrange to preserve death info for a register when some of the
1451: code in the machine description which is executed to write the assembler
1452: code looks at the death notes. This is necessary only when the actual
1453: hardware feature which GNU CC thinks of as a register is not actually a
1454: register of the usual sort. (It might, for example, be a hardware
1455: stack.)
1456:
1457: If this macro is not defined, it means that no death notes need to be
1458: preserved. This is the usual situation.
1459: @end table
1460:
1461: @node Register Classes
1462: @section Register Classes
1463: @cindex register class definitions
1464: @cindex class definitions, register
1465:
1466: On many machines, the numbered registers are not all equivalent.
1467: For example, certain registers may not be allowed for indexed addressing;
1468: certain registers may not be allowed in some instructions. These machine
1469: restrictions are described to the compiler using @dfn{register classes}.
1470:
1471: You define a number of register classes, giving each one a name and saying
1472: which of the registers belong to it. Then you can specify register classes
1473: that are allowed as operands to particular instruction patterns.
1474:
1475: @findex ALL_REGS
1476: @findex NO_REGS
1477: In general, each register will belong to several classes. In fact, one
1478: class must be named @code{ALL_REGS} and contain all the registers. Another
1479: class must be named @code{NO_REGS} and contain no registers. Often the
1480: union of two classes will be another class; however, this is not required.
1481:
1482: @findex GENERAL_REGS
1483: One of the classes must be named @code{GENERAL_REGS}. There is nothing
1484: terribly special about the name, but the operand constraint letters
1485: @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
1486: the same as @code{ALL_REGS}, just define it as a macro which expands
1487: to @code{ALL_REGS}.
1488:
1489: Order the classes so that if class @var{x} is contained in class @var{y}
1490: then @var{x} has a lower class number than @var{y}.
1491:
1492: The way classes other than @code{GENERAL_REGS} are specified in operand
1493: constraints is through machine-dependent operand constraint letters.
1494: You can define such letters to correspond to various classes, then use
1495: them in operand constraints.
1496:
1497: You should define a class for the union of two classes whenever some
1498: instruction allows both classes. For example, if an instruction allows
1499: either a floating point (coprocessor) register or a general register for a
1500: certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
1501: which includes both of them. Otherwise you will get suboptimal code.
1502:
1503: You must also specify certain redundant information about the register
1504: classes: for each class, which classes contain it and which ones are
1505: contained in it; for each pair of classes, the largest class contained
1506: in their union.
1507:
1508: When a value occupying several consecutive registers is expected in a
1509: certain class, all the registers used must belong to that class.
1510: Therefore, register classes cannot be used to enforce a requirement for
1511: a register pair to start with an even-numbered register. The way to
1512: specify this requirement is with @code{HARD_REGNO_MODE_OK}.
1513:
1514: Register classes used for input-operands of bitwise-and or shift
1515: instructions have a special requirement: each such class must have, for
1516: each fixed-point machine mode, a subclass whose registers can transfer that
1517: mode to or from memory. For example, on some machines, the operations for
1518: single-byte values (@code{QImode}) are limited to certain registers. When
1519: this is so, each register class that is used in a bitwise-and or shift
1520: instruction must have a subclass consisting of registers from which
1521: single-byte values can be loaded or stored. This is so that
1522: @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
1523:
1524: @table @code
1525: @findex enum reg_class
1526: @item enum reg_class
1527: An enumeral type that must be defined with all the register class names
1528: as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS}
1529: must be the last register class, followed by one more enumeral value,
1530: @code{LIM_REG_CLASSES}, which is not a register class but rather
1531: tells how many classes there are.
1532:
1533: Each register class has a number, which is the value of casting
1534: the class name to type @code{int}. The number serves as an index
1535: in many of the tables described below.
1536:
1537: @findex N_REG_CLASSES
1538: @item N_REG_CLASSES
1539: The number of distinct register classes, defined as follows:
1540:
1541: @example
1542: #define N_REG_CLASSES (int) LIM_REG_CLASSES
1543: @end example
1544:
1545: @findex REG_CLASS_NAMES
1546: @item REG_CLASS_NAMES
1547: An initializer containing the names of the register classes as C string
1548: constants. These names are used in writing some of the debugging dumps.
1549:
1550: @findex REG_CLASS_CONTENTS
1551: @item REG_CLASS_CONTENTS
1552: An initializer containing the contents of the register classes, as integers
1553: which are bit masks. The @var{n}th integer specifies the contents of class
1554: @var{n}. The way the integer @var{mask} is interpreted is that
1555: register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
1556:
1557: When the machine has more than 32 registers, an integer does not suffice.
1558: Then the integers are replaced by sub-initializers, braced groupings containing
1559: several integers. Each sub-initializer must be suitable as an initializer
1560: for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
1561:
1562: @findex REGNO_REG_CLASS
1563: @item REGNO_REG_CLASS (@var{regno})
1564: A C expression whose value is a register class containing hard register
1565: @var{regno}. In general there is more than one such class; choose a class
1566: which is @dfn{minimal}, meaning that no smaller class also contains the
1567: register.
1568:
1569: @findex BASE_REG_CLASS
1570: @item BASE_REG_CLASS
1571: A macro whose definition is the name of the class to which a valid
1572: base register must belong. A base register is one used in an address
1573: which is the register value plus a displacement.
1574:
1575: @findex INDEX_REG_CLASS
1576: @item INDEX_REG_CLASS
1577: A macro whose definition is the name of the class to which a valid
1578: index register must belong. An index register is one used in an
1579: address where its value is either multiplied by a scale factor or
1580: added to another register (as well as added to a displacement).
1581:
1582: @findex REG_CLASS_FROM_LETTER
1583: @item REG_CLASS_FROM_LETTER (@var{char})
1584: A C expression which defines the machine-dependent operand constraint
1585: letters for register classes. If @var{char} is such a letter, the
1586: value should be the register class corresponding to it. Otherwise,
1587: the value should be @code{NO_REGS}. The register letter @samp{r},
1588: corresponding to class @code{GENERAL_REGS}, will not be passed
1589: to this macro; you do not need to handle it.
1590:
1591: @findex REGNO_OK_FOR_BASE_P
1592: @item REGNO_OK_FOR_BASE_P (@var{num})
1593: A C expression which is nonzero if register number @var{num} is
1594: suitable for use as a base register in operand addresses. It may be
1595: either a suitable hard register or a pseudo register that has been
1596: allocated such a hard register.
1597:
1598: @findex REGNO_OK_FOR_INDEX_P
1599: @item REGNO_OK_FOR_INDEX_P (@var{num})
1600: A C expression which is nonzero if register number @var{num} is
1601: suitable for use as an index register in operand addresses. It may be
1602: either a suitable hard register or a pseudo register that has been
1603: allocated such a hard register.
1604:
1605: The difference between an index register and a base register is that
1606: the index register may be scaled. If an address involves the sum of
1607: two registers, neither one of them scaled, then either one may be
1608: labeled the ``base'' and the other the ``index''; but whichever
1609: labeling is used must fit the machine's constraints of which registers
1610: may serve in each capacity. The compiler will try both labelings,
1611: looking for one that is valid, and will reload one or both registers
1612: only if neither labeling works.
1613:
1614: @findex PREFERRED_RELOAD_CLASS
1615: @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
1616: A C expression that places additional restrictions on the register class
1617: to use when it is necessary to copy value @var{x} into a register in class
1618: @var{class}. The value is a register class; perhaps @var{class}, or perhaps
1619: another, smaller class. On many machines, the following definition is
1620: safe:
1621:
1622: @example
1623: #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
1624: @end example
1625:
1626: Sometimes returning a more restrictive class makes better code. For
1627: example, on the 68000, when @var{x} is an integer constant that is in range
1628: for a @samp{moveq} instruction, the value of this macro is always
1629: @code{DATA_REGS} as long as @var{class} includes the data registers.
1630: Requiring a data register guarantees that a @samp{moveq} will be used.
1631:
1632: If @var{x} is a @code{const_double}, by returning @code{NO_REGS}
1633: you can force @var{x} into a memory constant. This is useful on
1634: certain machines where immediate floating values cannot be loaded into
1635: certain kinds of registers.
1636:
1637: @findex PREFERRED_OUTPUT_RELOAD_CLASS
1638: @item PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
1639: Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
1640: input reloads. If you don't define this macro, the default is to use
1641: @var{class}, unchanged.
1642:
1643: @findex LIMIT_RELOAD_CLASS
1644: @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
1645: A C expression that places additional restrictions on the register class
1646: to use when it is necessary to be able to hold a value of mode
1647: @var{mode} in a reload register for which class @var{class} would
1648: ordinarily be used.
1649:
1650: Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
1651: there are certain modes that simply can't go in certain reload classes.
1652:
1653: The value is a register class; perhaps @var{class}, or perhaps another,
1654: smaller class.
1655:
1656: Don't define this macro unless the target machine has limitations which
1657: require the macro to do something nontrivial.
1658:
1659: @findex SECONDARY_RELOAD_CLASS
1660: @findex SECONDARY_INPUT_RELOAD_CLASS
1661: @findex SECONDARY_OUTPUT_RELOAD_CLASS
1662: @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
1663: @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
1664: @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
1665: Many machines have some registers that cannot be copied directly to or
1666: from memory or even from other types of registers. An example is the
1667: @samp{MQ} register, which on most machines, can only be copied to or
1668: from general registers, but not memory. Some machines allow copying all
1669: registers to and from memory, but require a scratch register for stores
1670: to some memory locations (e.g., those with symbolic address on the RT,
1671: and those with certain symbolic address on the Sparc when compiling
1672: PIC). In some cases, both an intermediate and a scratch register are
1673: required.
1674:
1675: You should define these macros to indicate to the reload phase that it may
1676: need to allocate at least one register for a reload in addition to the
1677: register to contain the data. Specifically, if copying @var{x} to a
1678: register @var{class} in @var{mode} requires an intermediate register,
1679: you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
1680: largest register class all of whose registers can be used as
1681: intermediate registers or scratch registers.
1682:
1683: If copying a register @var{class} in @var{mode} to @var{x} requires an
1684: intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
1685: should be defined to return the largest register class required. If the
1686: requirements for input and output reloads are the same, the macro
1687: @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both
1688: macros identically.
1689:
1690: The values returned by these macros are often @code{GENERAL_REGS}.
1691: Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
1692: can be directly copied to or from a register of @var{class} in
1693: @var{mode} without requiring a scratch register. Do not define this
1694: macro if it would always return @code{NO_REGS}.
1695:
1696: If a scratch register is required (either with or without an
1697: intermediate register), you should define patterns for
1698: @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
1699: (@pxref{Standard Names}. These patterns, which will normally be
1700: implemented with a @code{define_expand}, should be similar to the
1701: @samp{mov@var{m}} patterns, except that operand 2 is the scratch
1702: register.
1703:
1704: Define constraints for the reload register and scratch register that
1705: contain a single register class. If the original reload register (whose
1706: class is @var{class}) can meet the constraint given in the pattern, the
1707: value returned by these macros is used for the class of the scratch
1708: register. Otherwise, two additional reload registers are required.
1709: Their classes are obtained from the constraints in the insn pattern.
1710:
1711: @var{x} might be a pseudo-register or a @code{subreg} of a
1712: pseudo-register, which could either be in a hard register or in memory.
1713: Use @code{true_regnum} to find out; it will return -1 if the pseudo is
1714: in memory and the hard register number if it is in a register.
1715:
1716: These macros should not be used in the case where a particular class of
1717: registers can only be copied to memory and not to another class of
1718: registers. In that case, secondary reload registers are not needed and
1719: would not be helpful. Instead, a stack location must be used to perform
1720: the copy and the @code{mov@var{m}} pattern should use memory as a
1721: intermediate storage. This case often occurs between floating-point and
1722: general registers.
1723:
1724: @findex SECONDARY_MEMORY_NEEDED
1725: @item SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
1726: Certain machines have the property that some registers cannot be copied
1727: to some other registers without using memory. Define this macro on
1728: those machines to be a C expression that is non-zero if objects of mode
1729: @var{m} in registers of @var{class1} can only be copied to registers of
1730: class @var{class2} by storing a register of @var{class1} into memory
1731: and loading that memory location into a register of @var{class2}.
1732:
1733: Do not define this macro if its value would always be zero.
1734:
1735: @findex SECONDARY_MEMORY_NEEDED_RTX
1736: @item SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
1737: Normally, when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
1738: will allocate a stack slot when a memory location for a register copy
1739: is needed. If this macro is defined, the compiler instead uses the
1740: memory location defined by this macro.
1741:
1742: @findex SMALL_REGISTER_CLASSES
1743: @item SMALL_REGISTER_CLASSES
1744: Normally the compiler will avoid choosing spill registers from registers
1745: that have been explicitly mentioned in the rtl (these registers are
1746: normally those used to pass parameters and return values). However,
1747: some machines have so few registers of certain classes that there would
1748: not be enough registers to use as spill registers if this were done.
1749:
1750: You should define @code{SMALL_REGISTER_CLASSES} on those machines. When
1751: it is defined, the compiler allows registers explicitly used in the rtl
1752: to be used as spill registers but prevents the compiler from extending
1753: the lifetime of these registers.
1754:
1755: Defining this macro is always safe, but unnecessarily defining this macro
1756: will reduce the amount of optimizations that can be performed in some
1757: cases. If this macro is not defined but needs to be, the compiler will
1758: run out of reload registers and print a fatal error message.
1759:
1760: For most machines, this macro should not be defined.
1761:
1762: @findex CLASS_LIKELY_SPILLED_P
1763: @item CLASS_LIKELY_SPILLED_P (@var{class})
1764: A C expression whose value is nonzero if pseudos that have been assigned
1765: to registers of class @var{class} would likely be spilled because
1766: registers of @var{class} are needed for spill registers.
1767:
1768: The default value of this macro returns 1 if @var{class} has exactly one
1769: register and zero otherwise. On most machines, this default should be
1770: used. Only define this macro to some other expression if pseudo
1771: allocated by @file{local-alloc.c} end up in memory because their hard
1772: registers were needed for spill regisers. If this macro returns nonzero
1773: for those classes, those pseudos will only be allocated by
1774: @file{global.c}, which knows how to reallocate the pseudo to another
1775: register. If there would not be another register available for
1776: reallocation, you should not change the definition of this macro since
1777: the only effect of such a definition would be to slow down register
1778: allocation.
1779:
1780: @findex CLASS_MAX_NREGS
1781: @item CLASS_MAX_NREGS (@var{class}, @var{mode})
1782: A C expression for the maximum number of consecutive registers
1783: of class @var{class} needed to hold a value of mode @var{mode}.
1784:
1785: This is closely related to the macro @code{HARD_REGNO_NREGS}.
1786: In fact, the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
1787: should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno}, @var{mode})}
1788: for all @var{regno} values in the class @var{class}.
1789:
1790: This macro helps control the handling of multiple-word values
1791: in the reload pass.
1792: @end table
1793:
1794: Three other special macros describe which operands fit which constraint
1795: letters.
1796:
1797: @table @code
1798: @findex CONST_OK_FOR_LETTER_P
1799: @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
1800: A C expression that defines the machine-dependent operand constraint letters
1801: that specify particular ranges of integer values. If @var{c} is one
1802: of those letters, the expression should check that @var{value}, an integer,
1803: is in the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
1804: not one of those letters, the value should be 0 regardless of @var{value}.
1805:
1806: @findex CONST_DOUBLE_OK_FOR_LETTER_P
1807: @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
1808: A C expression that defines the machine-dependent operand constraint
1809: letters that specify particular ranges of @code{const_double} values.
1810:
1811: If @var{c} is one of those letters, the expression should check that
1812: @var{value}, an RTX of code @code{const_double}, is in the appropriate
1813: range and return 1 if so, 0 otherwise. If @var{c} is not one of those
1814: letters, the value should be 0 regardless of @var{value}.
1815:
1816: @code{const_double} is used for all floating-point constants and for
1817: @code{DImode} fixed-point constants. A given letter can accept either
1818: or both kinds of values. It can use @code{GET_MODE} to distinguish
1819: between these kinds.
1820:
1821: @findex EXTRA_CONSTRAINT
1822: @item EXTRA_CONSTRAINT (@var{value}, @var{c})
1823: A C expression that defines the optional machine-dependent constraint
1824: letters that can be used to segregate specific types of operands,
1825: usually memory references, for the target machine. Normally this macro
1826: will not be defined. If it is required for a particular target machine,
1827: it should return 1 if @var{value} corresponds to the operand type
1828: represented by the constraint letter @var{c}. If @var{c} is not defined
1829: as an extra constraint, the value returned should be 0 regardless of
1830: @var{value}.
1831:
1832: For example, on the ROMP, load instructions cannot have their output in r0 if
1833: the memory reference contains a symbolic address. Constraint letter
1834: @samp{Q} is defined as representing a memory address that does
1835: @emph{not} contain a symbolic address. An alternative is specified with
1836: a @samp{Q} constraint on the input and @samp{r} on the output. The next
1837: alternative specifies @samp{m} on the input and a register class that
1838: does not include r0 on the output.
1839: @end table
1840:
1841: @node Stack and Calling
1842: @section Stack Layout and Calling Conventions
1843: @cindex calling conventions
1844:
1845: @menu
1846: * Frame Layout::
1847: * Frame Registers::
1848: * Elimination::
1849: * Stack Arguments::
1850: * Register Arguments::
1851: * Scalar Return::
1852: * Aggregate Return::
1853: * Caller Saves::
1854: * Function Entry::
1855: * Profiling::
1856: @end menu
1857:
1858: @node Frame Layout
1859: @subsection Basic Stack Layout
1860: @cindex stack frame layout
1861: @cindex frame layout
1862:
1863: @table @code
1864: @findex STACK_GROWS_DOWNWARD
1865: @item STACK_GROWS_DOWNWARD
1866: Define this macro if pushing a word onto the stack moves the stack
1867: pointer to a smaller address.
1868:
1869: When we say, ``define this macro if @dots{},'' it means that the
1870: compiler checks this macro only with @code{#ifdef} so the precise
1871: definition used does not matter.
1872:
1873: @findex FRAME_GROWS_DOWNWARD
1874: @item FRAME_GROWS_DOWNWARD
1875: Define this macro if the addresses of local variable slots are at negative
1876: offsets from the frame pointer.
1877:
1878: @findex ARGS_GROW_DOWNWARD
1879: @item ARGS_GROW_DOWNWARD
1880: Define this macro if successive arguments to a function occupy decreasing
1881: addresses on the stack.
1882:
1883: @findex STARTING_FRAME_OFFSET
1884: @item STARTING_FRAME_OFFSET
1885: Offset from the frame pointer to the first local variable slot to be allocated.
1886:
1887: If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
1888: subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
1889: Otherwise, it is found by adding the length of the first slot to the
1890: value @code{STARTING_FRAME_OFFSET}.
1891: @c i'm not sure if the above is still correct.. had to change it to get
1892: @c rid of an overfull. --mew 2feb93
1893:
1894: @findex STACK_POINTER_OFFSET
1895: @item STACK_POINTER_OFFSET
1896: Offset from the stack pointer register to the first location at which
1897: outgoing arguments are placed. If not specified, the default value of
1898: zero is used. This is the proper value for most machines.
1899:
1900: If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
1901: the first location at which outgoing arguments are placed.
1902:
1903: @findex FIRST_PARM_OFFSET
1904: @item FIRST_PARM_OFFSET (@var{fundecl})
1905: Offset from the argument pointer register to the first argument's
1906: address. On some machines it may depend on the data type of the
1907: function.
1908:
1909: If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
1910: the first argument's address.
1911:
1912: @findex STACK_DYNAMIC_OFFSET
1913: @item STACK_DYNAMIC_OFFSET (@var{fundecl})
1914: Offset from the stack pointer register to an item dynamically allocated
1915: on the stack, e.g., by @code{alloca}.
1916:
1917: The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
1918: length of the outgoing arguments. The default is correct for most
1919: machines. See @file{function.c} for details.
1920:
1921: @findex DYNAMIC_CHAIN_ADDRESS
1922: @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
1923: A C expression whose value is RTL representing the address in a stack
1924: frame where the pointer to the caller's frame is stored. Assume that
1925: @var{frameaddr} is an RTL expression for the address of the stack frame
1926: itself.
1927:
1928: If you don't define this macro, the default is to return the value
1929: of @var{frameaddr}---that is, the stack frame address is also the
1930: address of the stack word that points to the previous frame.
1931:
1932: @findex SETUP_FRAME_ADDRESSES
1933: @item SERTUP_FRAME_ADDRESSES ()
1934: If defined, a C expression that produces the machine-specific code to
1935: setup the stack so that arbitrary frames can be accessed. For example,
1936: on the Sparc, we must flush all of the register windows to the stack
1937: before we can access arbitrary stack frames.
1938: This macro will seldom need to be defined.
1939:
1940: @findex RETURN_ADDR_RTX
1941: @item RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
1942: A C expression whose value is RTL representing the value of the return
1943: address for the frame @var{count} steps up from the current frame.
1944: @var{frameaddr} is the frame pointer of the @var{count} frame, or
1945: the frame pointer of the @var{count} @minus{} 1 frame if
1946: @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
1947:
1948: @findex RETURN_ADDR_IN_PREVIOUS_FRAME
1949: @item RETURN_ADDR_IN_PREVIOUS_FRAME
1950: Define this if the return address of a particular stack frame is accessed
1951: from the frame pointer of the previous stack frame.
1952: @end table
1953:
1954: @need 2000
1955: @node Frame Registers
1956: @subsection Registers That Address the Stack Frame
1957:
1958: @table @code
1959: @findex STACK_POINTER_REGNUM
1960: @item STACK_POINTER_REGNUM
1961: The register number of the stack pointer register, which must also be a
1962: fixed register according to @code{FIXED_REGISTERS}. On most machines,
1963: the hardware determines which register this is.
1964:
1965: @findex FRAME_POINTER_REGNUM
1966: @item FRAME_POINTER_REGNUM
1967: The register number of the frame pointer register, which is used to
1968: access automatic variables in the stack frame. On some machines, the
1969: hardware determines which register this is. On other machines, you can
1970: choose any register you wish for this purpose.
1971:
1972: @findex HARD_FRAME_POINTER_REGNUM
1973: @item HARD_FRAME_POINTER_REGNUM
1974: On some machines the offset between the frame pointer and starting
1975: offset of the automatic variables is not known until after register
1976: allocation has been done (for example, because the saved registers are
1977: between these two locations). On those machines,
1978: @code{FRAME_POINTER_REGNUM} as a special, fixed register to be used
1979: internally until the offset is known, and define
1980: @code{HARD_FRAME_POINTER_REGNUM} to be the hard register used for the
1981: frame pointer.
1982:
1983: You should define this macro only in the very rare circumstances when it
1984: is not possible to calculate the offset between the frame pointer and
1985: the automatic variables until after register allocation has been
1986: completed. When this macro is defined, you must also indicate in your
1987: definition of @code{ELIMINABLE_REGS} how to eliminate
1988: @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
1989: or @code{STACK_POINTER_REGNUM}.
1990:
1991: Do not define this macro if it would be the same as
1992: @code{FRAME_POINTER_REGNUM}.
1993:
1994: @findex ARG_POINTER_REGNUM
1995: @item ARG_POINTER_REGNUM
1996: The register number of the arg pointer register, which is used to access
1997: the function's argument list. On some machines, this is the same as the
1998: frame pointer register. On some machines, the hardware determines which
1999: register this is. On other machines, you can choose any register you
2000: wish for this purpose. If this is not the same register as the frame
2001: pointer register, then you must mark it as a fixed register according to
2002: @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
2003: (@pxref{Elimination}).
2004:
2005: @findex STATIC_CHAIN_REGNUM
2006: @findex STATIC_CHAIN_INCOMING_REGNUM
2007: @item STATIC_CHAIN_REGNUM
2008: @itemx STATIC_CHAIN_INCOMING_REGNUM
2009: Register numbers used for passing a function's static chain pointer. If
2010: register windows are used, the register number as seen by the called
2011: function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
2012: number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
2013: these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
2014: not be defined.@refill
2015:
2016: The static chain register need not be a fixed register.
2017:
2018: If the static chain is passed in memory, these macros should not be
2019: defined; instead, the next two macros should be defined.
2020:
2021: @findex STATIC_CHAIN
2022: @findex STATIC_CHAIN_INCOMING
2023: @item STATIC_CHAIN
2024: @itemx STATIC_CHAIN_INCOMING
2025: If the static chain is passed in memory, these macros provide rtx giving
2026: @code{mem} expressions that denote where they are stored.
2027: @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations
2028: as seen by the calling and called functions, respectively. Often the former
2029: will be at an offset from the stack pointer and the latter at an offset from
2030: the frame pointer.@refill
2031:
2032: @findex stack_pointer_rtx
2033: @findex frame_pointer_rtx
2034: @findex arg_pointer_rtx
2035: The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
2036: @code{arg_pointer_rtx} will have been initialized prior to the use of these
2037: macros and should be used to refer to those items.
2038:
2039: If the static chain is passed in a register, the two previous macros should
2040: be defined instead.
2041: @end table
2042:
2043: @node Elimination
2044: @subsection Eliminating Frame Pointer and Arg Pointer
2045:
2046: @table @code
2047: @findex FRAME_POINTER_REQUIRED
2048: @item FRAME_POINTER_REQUIRED
2049: A C expression which is nonzero if a function must have and use a frame
2050: pointer. This expression is evaluated in the reload pass. If its value is
2051: nonzero the function will have a frame pointer.
2052:
2053: The expression can in principle examine the current function and decide
2054: according to the facts, but on most machines the constant 0 or the
2055: constant 1 suffices. Use 0 when the machine allows code to be generated
2056: with no frame pointer, and doing so saves some time or space. Use 1
2057: when there is no possible advantage to avoiding a frame pointer.
2058:
2059: In certain cases, the compiler does not know how to produce valid code
2060: without a frame pointer. The compiler recognizes those cases and
2061: automatically gives the function a frame pointer regardless of what
2062: @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about
2063: them.@refill
2064:
2065: In a function that does not require a frame pointer, the frame pointer
2066: register can be allocated for ordinary usage, unless you mark it as a
2067: fixed register. See @code{FIXED_REGISTERS} for more information.
2068:
2069: This macro is ignored and you do not need to define it if the function
2070: @code{ELIMINABLE_REGS} is defined.
2071:
2072: @findex INITIAL_FRAME_POINTER_OFFSET
2073: @findex get_frame_size
2074: @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
2075: A C statement to store in the variable @var{depth-var} the difference
2076: between the frame pointer and the stack pointer values immediately after
2077: the function prologue. The value would be computed from information
2078: such as the result of @code{get_frame_size ()} and the tables of
2079: registers @code{regs_ever_live} and @code{call_used_regs}.
2080:
2081: If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
2082: need not be defined. Otherwise, it must be defined even if
2083: @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that
2084: case, you may set @var{depth-var} to anything.
2085:
2086: @findex ELIMINABLE_REGS
2087: @item ELIMINABLE_REGS
2088: If defined, this macro specifies a table of register pairs used to
2089: eliminate unneeded registers that point into the stack frame. If it is not
2090: defined, the only elimination attempted by the compiler is to replace
2091: references to the frame pointer with references to the stack pointer.
2092:
2093: The definition of this macro is a list of structure initializations, each
2094: of which specifies an original and replacement register.
2095:
2096: On some machines, the position of the argument pointer is not known until
2097: the compilation is completed. In such a case, a separate hard register
2098: must be used for the argument pointer. This register can be eliminated by
2099: replacing it with either the frame pointer or the argument pointer,
2100: depending on whether or not the frame pointer has been eliminated.
2101:
2102: In this case, you might specify:
2103: @example
2104: #define ELIMINABLE_REGS \
2105: @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
2106: @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
2107: @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
2108: @end example
2109:
2110: Note that the elimination of the argument pointer with the stack pointer is
2111: specified first since that is the preferred elimination.
2112:
2113: @findex CAN_ELIMINATE
2114: @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg})
2115: A C expression that returns non-zero if the compiler is allowed to try
2116: to replace register number @var{from-reg} with register number
2117: @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS}
2118: is defined, and will usually be the constant 1, since most of the cases
2119: preventing register elimination are things that the compiler already
2120: knows about.
2121:
2122: @findex INITIAL_ELIMINATION_OFFSET
2123: @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
2124: This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
2125: specifies the initial difference between the specified pair of
2126: registers. This macro must be defined if @code{ELIMINABLE_REGS} is
2127: defined.
2128:
2129: @findex LONGJMP_RESTORE_FROM_STACK
2130: @item LONGJMP_RESTORE_FROM_STACK
2131: Define this macro if the @code{longjmp} function restores registers from
2132: the stack frames, rather than from those saved specifically by
2133: @code{setjmp}. Certain quantities must not be kept in registers across
2134: a call to @code{setjmp} on such machines.
2135: @end table
2136:
2137: @node Stack Arguments
2138: @subsection Passing Function Arguments on the Stack
2139: @cindex arguments on stack
2140: @cindex stack arguments
2141:
2142: The macros in this section control how arguments are passed
2143: on the stack. See the following section for other macros that
2144: control passing certain arguments in registers.
2145:
2146: @table @code
2147: @findex PROMOTE_PROTOTYPES
2148: @item PROMOTE_PROTOTYPES
2149: Define this macro if an argument declared in a prototype as an
2150: integral type smaller than @code{int} should actually be passed as an
2151: @code{int}. In addition to avoiding errors in certain cases of
2152: mismatch, it also makes for better code on certain machines.
2153:
2154: @findex PUSH_ROUNDING
2155: @item PUSH_ROUNDING (@var{npushed})
2156: A C expression that is the number of bytes actually pushed onto the
2157: stack when an instruction attempts to push @var{npushed} bytes.
2158:
2159: If the target machine does not have a push instruction, do not define
2160: this macro. That directs GNU CC to use an alternate strategy: to
2161: allocate the entire argument block and then store the arguments into
2162: it.
2163:
2164: On some machines, the definition
2165:
2166: @example
2167: #define PUSH_ROUNDING(BYTES) (BYTES)
2168: @end example
2169:
2170: @noindent
2171: will suffice. But on other machines, instructions that appear
2172: to push one byte actually push two bytes in an attempt to maintain
2173: alignment. Then the definition should be
2174:
2175: @example
2176: #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
2177: @end example
2178:
2179: @findex ACCUMULATE_OUTGOING_ARGS
2180: @findex current_function_outgoing_args_size
2181: @item ACCUMULATE_OUTGOING_ARGS
2182: If defined, the maximum amount of space required for outgoing arguments
2183: will be computed and placed into the variable
2184: @code{current_function_outgoing_args_size}. No space will be pushed
2185: onto the stack for each call; instead, the function prologue should
2186: increase the stack frame size by this amount.
2187:
2188: Defining both @code{PUSH_ROUNDING} and @code{ACCUMULATE_OUTGOING_ARGS}
2189: is not proper.
2190:
2191: @findex REG_PARM_STACK_SPACE
2192: @item REG_PARM_STACK_SPACE (@var{fndecl})
2193: Define this macro if functions should assume that stack space has been
2194: allocated for arguments even when their values are passed in
2195: registers.
2196:
2197: The value of this macro is the size, in bytes, of the area reserved for
2198: arguments passed in registers for the function represented by @var{fndecl}.
2199:
2200: This space can be allocated by the caller, or be a part of the
2201: machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
2202: which.
2203: @c above is overfull. not sure what to do. --mew 5feb93 did
2204: @c something, not sure if it looks good. --mew 10feb93
2205:
2206: @findex MAYBE_REG_PARM_STACK_SPACE
2207: @findex FINAL_REG_PARM_STACK_SPACE
2208: @item MAYBE_REG_PARM_STACK_SPACE
2209: @itemx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size})
2210: Define these macros in addition to the one above if functions might
2211: allocate stack space for arguments even when their values are passed
2212: in registers. These should be used when the stack space allocated
2213: for arguments in registers is not a simple constant independent of the
2214: function declaration.
2215:
2216: The value of the first macro is the size, in bytes, of the area that
2217: we should initially assume would be reserved for arguments passed in registers.
2218:
2219: The value of the second macro is the actual size, in bytes, of the area
2220: that will be reserved for arguments passed in registers. This takes two
2221: arguments: an integer representing the number of bytes of fixed sized
2222: arguments on the stack, and a tree representing the number of bytes of
2223: variable sized arguments on the stack.
2224:
2225: When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be
2226: called for libcall functions, the current function, or for a function
2227: being called when it is known that such stack space must be allocated.
2228: In each case this value can be easily computed.
2229:
2230: When deciding whether a called function needs such stack space, and how
2231: much space to reserve, GNU CC uses these two macros instead of
2232: @code{REG_PARM_STACK_SPACE}.
2233:
2234: @findex OUTGOING_REG_PARM_STACK_SPACE
2235: @item OUTGOING_REG_PARM_STACK_SPACE
2236: Define this if it is the responsibility of the caller to allocate the area
2237: reserved for arguments passed in registers.
2238:
2239: If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
2240: whether the space for these arguments counts in the value of
2241: @code{current_function_outgoing_args_size}.
2242:
2243: @findex STACK_PARMS_IN_REG_PARM_AREA
2244: @item STACK_PARMS_IN_REG_PARM_AREA
2245: Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
2246: stack parameters don't skip the area specified by it.
2247: @c i changed this, makes more sens and it should have taken care of the
2248: @c overfull.. not as specific, tho. --mew 5feb93
2249:
2250: Normally, when a parameter is not passed in registers, it is placed on the
2251: stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
2252: suppresses this behavior and causes the parameter to be passed on the
2253: stack in its natural location.
2254:
2255: @findex RETURN_POPS_ARGS
2256: @item RETURN_POPS_ARGS (@var{funtype}, @var{stack-size})
2257: A C expression that should indicate the number of bytes of its own
2258: arguments that a function pops on returning, or 0 if the
2259: function pops no arguments and the caller must therefore pop them all
2260: after the function returns.
2261:
2262: @var{funtype} is a C variable whose value is a tree node that
2263: describes the function in question. Normally it is a node of type
2264: @code{FUNCTION_TYPE} that describes the data type of the function.
2265: From this it is possible to obtain the data types of the value and
2266: arguments (if known).
2267:
2268: When a call to a library function is being considered, @var{funtype}
2269: will contain an identifier node for the library function. Thus, if
2270: you need to distinguish among various library functions, you can do so
2271: by their names. Note that ``library function'' in this context means
2272: a function used to perform arithmetic, whose name is known specially
2273: in the compiler and was not mentioned in the C code being compiled.
2274:
2275: @var{stack-size} is the number of bytes of arguments passed on the
2276: stack. If a variable number of bytes is passed, it is zero, and
2277: argument popping will always be the responsibility of the calling function.
2278:
2279: On the Vax, all functions always pop their arguments, so the definition
2280: of this macro is @var{stack-size}. On the 68000, using the standard
2281: calling convention, no functions pop their arguments, so the value of
2282: the macro is always 0 in this case. But an alternative calling
2283: convention is available in which functions that take a fixed number of
2284: arguments pop them but other functions (such as @code{printf}) pop
2285: nothing (the caller pops all). When this convention is in use,
2286: @var{funtype} is examined to determine whether a function takes a fixed
2287: number of arguments.
2288: @end table
2289:
2290: @node Register Arguments
2291: @subsection Passing Arguments in Registers
2292: @cindex arguments in registers
2293: @cindex registers arguments
2294:
2295: This section describes the macros which let you control how various
2296: types of arguments are passed in registers or how they are arranged in
2297: the stack.
2298:
2299: @table @code
2300: @findex FUNCTION_ARG
2301: @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
2302: A C expression that controls whether a function argument is passed
2303: in a register, and which register.
2304:
2305: The arguments are @var{cum}, which summarizes all the previous
2306: arguments; @var{mode}, the machine mode of the argument; @var{type},
2307: the data type of the argument as a tree node or 0 if that is not known
2308: (which happens for C support library functions); and @var{named},
2309: which is 1 for an ordinary argument and 0 for nameless arguments that
2310: correspond to @samp{@dots{}} in the called function's prototype.
2311:
2312: The value of the expression should either be a @code{reg} RTX for the
2313: hard register in which to pass the argument, or zero to pass the
2314: argument on the stack.
2315:
2316: For machines like the Vax and 68000, where normally all arguments are
2317: pushed, zero suffices as a definition.
2318:
2319: @cindex @file{stdarg.h} and register arguments
2320: The usual way to make the ANSI library @file{stdarg.h} work on a machine
2321: where some arguments are usually passed in registers, is to cause
2322: nameless arguments to be passed on the stack instead. This is done
2323: by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
2324:
2325: @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
2326: @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
2327: You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})}
2328: in the definition of this macro to determine if this argument is of a
2329: type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
2330: is not defined and @code{FUNCTION_ARG} returns non-zero for such an
2331: argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
2332: defined, the argument will be computed in the stack and then loaded into
2333: a register.
2334:
2335: @findex FUNCTION_INCOMING_ARG
2336: @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
2337: Define this macro if the target machine has ``register windows'', so
2338: that the register in which a function sees an arguments is not
2339: necessarily the same as the one in which the caller passed the
2340: argument.
2341:
2342: For such machines, @code{FUNCTION_ARG} computes the register in which
2343: the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
2344: be defined in a similar fashion to tell the function being called
2345: where the arguments will arrive.
2346:
2347: If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
2348: serves both purposes.@refill
2349:
2350: @findex FUNCTION_ARG_PARTIAL_NREGS
2351: @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named})
2352: A C expression for the number of words, at the beginning of an
2353: argument, must be put in registers. The value must be zero for
2354: arguments that are passed entirely in registers or that are entirely
2355: pushed on the stack.
2356:
2357: On some machines, certain arguments must be passed partially in
2358: registers and partially in memory. On these machines, typically the
2359: first @var{n} words of arguments are passed in registers, and the rest
2360: on the stack. If a multi-word argument (a @code{double} or a
2361: structure) crosses that boundary, its first few words must be passed
2362: in registers and the rest must be pushed. This macro tells the
2363: compiler when this occurs, and how many of the words should go in
2364: registers.
2365:
2366: @code{FUNCTION_ARG} for these arguments should return the first
2367: register to be used by the caller for this argument; likewise
2368: @code{FUNCTION_INCOMING_ARG}, for the called function.
2369:
2370: @findex FUNCTION_ARG_PASS_BY_REFERENCE
2371: @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named})
2372: A C expression that indicates when an argument must be passed by reference.
2373: If nonzero for an argument, a copy of that argument is made in memory and a
2374: pointer to the argument is passed instead of the argument itself.
2375: The pointer is passed in whatever way is appropriate for passing a pointer
2376: to that type.
2377:
2378: On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable
2379: definition of this macro might be
2380: @smallexample
2381: #define FUNCTION_ARG_PASS_BY_REFERENCE\
2382: (CUM, MODE, TYPE, NAMED) \
2383: MUST_PASS_IN_STACK (MODE, TYPE)
2384: @end smallexample
2385: @c this is *still* too long. --mew 5feb93
2386:
2387: @findex FUNCTION_ARG_CALLEE_COPIES
2388: @item FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named})
2389: If defined, a C expression that indicates when it is the called function's
2390: responsibility to make a copy of arguments passed by invisible reference.
2391: Normally, the caller makes a copy and passes the address of the copy to the
2392: routine being called. When FUNCTION_ARG_CALLEE_COPIES is defined and is
2393: nonzero, the caller does not make a copy. Instead, it passes a pointer to the
2394: ``live'' value. The called function must not modify this value. If it can be
2395: determined that the value won't be modified, it need not make a copy;
2396: otherwise a copy must be made.
2397:
2398: @findex CUMULATIVE_ARGS
2399: @item CUMULATIVE_ARGS
2400: A C type for declaring a variable that is used as the first argument of
2401: @code{FUNCTION_ARG} and other related values. For some target machines,
2402: the type @code{int} suffices and can hold the number of bytes of
2403: argument so far.
2404:
2405: There is no need to record in @code{CUMULATIVE_ARGS} anything about the
2406: arguments that have been passed on the stack. The compiler has other
2407: variables to keep track of that. For target machines on which all
2408: arguments are passed on the stack, there is no need to store anything in
2409: @code{CUMULATIVE_ARGS}; however, the data structure must exist and
2410: should not be empty, so use @code{int}.
2411:
2412: @findex INIT_CUMULATIVE_ARGS
2413: @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname})
2414: A C statement (sans semicolon) for initializing the variable @var{cum}
2415: for the state at the beginning of the argument list. The variable has
2416: type @code{CUMULATIVE_ARGS}. The value of @var{fntype} is the tree node
2417: for the data type of the function which will receive the args, or 0
2418: if the args are to a compiler support library function.
2419:
2420: When processing a call to a compiler support library function,
2421: @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
2422: contains the name of the function, as a string. @var{libname} is 0 when
2423: an ordinary C function call is being processed. Thus, each time this
2424: macro is called, either @var{libname} or @var{fntype} is nonzero, but
2425: never both of them at once.
2426:
2427: @findex INIT_CUMULATIVE_INCOMING_ARGS
2428: @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
2429: Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
2430: finding the arguments for the function being compiled. If this macro is
2431: undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
2432:
2433: The value passed for @var{libname} is always 0, since library routines
2434: with special calling conventions are never compiled with GNU CC. The
2435: argument @var{libname} exists for symmetry with
2436: @code{INIT_CUMULATIVE_ARGS}.
2437: @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
2438: @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
2439:
2440: @findex FUNCTION_ARG_ADVANCE
2441: @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
2442: A C statement (sans semicolon) to update the summarizer variable
2443: @var{cum} to advance past an argument in the argument list. The
2444: values @var{mode}, @var{type} and @var{named} describe that argument.
2445: Once this is done, the variable @var{cum} is suitable for analyzing
2446: the @emph{following} argument with @code{FUNCTION_ARG}, etc.@refill
2447:
2448: This macro need not do anything if the argument in question was passed
2449: on the stack. The compiler knows how to track the amount of stack space
2450: used for arguments without any special help.
2451:
2452: @findex FUNCTION_ARG_PADDING
2453: @item FUNCTION_ARG_PADDING (@var{mode}, @var{type})
2454: If defined, a C expression which determines whether, and in which direction,
2455: to pad out an argument with extra space. The value should be of type
2456: @code{enum direction}: either @code{upward} to pad above the argument,
2457: @code{downward} to pad below, or @code{none} to inhibit padding.
2458:
2459: The @emph{amount} of padding is always just enough to reach the next
2460: multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
2461: it.
2462:
2463: This macro has a default definition which is right for most systems.
2464: For little-endian machines, the default is to pad upward. For
2465: big-endian machines, the default is to pad downward for an argument of
2466: constant size shorter than an @code{int}, and upward otherwise.
2467:
2468: @findex FUNCTION_ARG_BOUNDARY
2469: @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
2470: If defined, a C expression that gives the alignment boundary, in bits,
2471: of an argument with the specified mode and type. If it is not defined,
2472: @code{PARM_BOUNDARY} is used for all arguments.
2473:
2474: @findex FUNCTION_ARG_REGNO_P
2475: @item FUNCTION_ARG_REGNO_P (@var{regno})
2476: A C expression that is nonzero if @var{regno} is the number of a hard
2477: register in which function arguments are sometimes passed. This does
2478: @emph{not} include implicit arguments such as the static chain and
2479: the structure-value address. On many machines, no registers can be
2480: used for this purpose since all function arguments are pushed on the
2481: stack.
2482: @end table
2483:
2484: @node Scalar Return
2485: @subsection How Scalar Function Values Are Returned
2486: @cindex return values in registers
2487: @cindex values, returned by functions
2488: @cindex scalars, returned as values
2489:
2490: This section discusses the macros that control returning scalars as
2491: values---values that can fit in registers.
2492:
2493: @table @code
2494: @findex TRADITIONAL_RETURN_FLOAT
2495: @item TRADITIONAL_RETURN_FLOAT
2496: Define this macro if @samp{-traditional} should not cause functions
2497: declared to return @code{float} to convert the value to @code{double}.
2498:
2499: @findex FUNCTION_VALUE
2500: @item FUNCTION_VALUE (@var{valtype}, @var{func})
2501: A C expression to create an RTX representing the place where a
2502: function returns a value of data type @var{valtype}. @var{valtype} is
2503: a tree node representing a data type. Write @code{TYPE_MODE
2504: (@var{valtype})} to get the machine mode used to represent that type.
2505: On many machines, only the mode is relevant. (Actually, on most
2506: machines, scalar values are returned in the same place regardless of
2507: mode).@refill
2508:
2509: If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same
2510: promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a
2511: scalar type.
2512:
2513: If the precise function being called is known, @var{func} is a tree
2514: node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
2515: pointer. This makes it possible to use a different value-returning
2516: convention for specific functions when all their calls are
2517: known.@refill
2518:
2519: @code{FUNCTION_VALUE} is not used for return vales with aggregate data
2520: types, because these are returned in another way. See
2521: @code{STRUCT_VALUE_REGNUM} and related macros, below.
2522:
2523: @findex FUNCTION_OUTGOING_VALUE
2524: @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
2525: Define this macro if the target machine has ``register windows''
2526: so that the register in which a function returns its value is not
2527: the same as the one in which the caller sees the value.
2528:
2529: For such machines, @code{FUNCTION_VALUE} computes the register in which
2530: the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be
2531: defined in a similar fashion to tell the function where to put the
2532: value.@refill
2533:
2534: If @code{FUNCTION_OUTGOING_VALUE} is not defined,
2535: @code{FUNCTION_VALUE} serves both purposes.@refill
2536:
2537: @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with
2538: aggregate data types, because these are returned in another way. See
2539: @code{STRUCT_VALUE_REGNUM} and related macros, below.
2540:
2541: @findex LIBCALL_VALUE
2542: @item LIBCALL_VALUE (@var{mode})
2543: A C expression to create an RTX representing the place where a library
2544: function returns a value of mode @var{mode}. If the precise function
2545: being called is known, @var{func} is a tree node
2546: (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
2547: pointer. This makes it possible to use a different value-returning
2548: convention for specific functions when all their calls are
2549: known.@refill
2550:
2551: Note that ``library function'' in this context means a compiler
2552: support routine, used to perform arithmetic, whose name is known
2553: specially by the compiler and was not mentioned in the C code being
2554: compiled.
2555:
2556: The definition of @code{LIBRARY_VALUE} need not be concerned aggregate
2557: data types, because none of the library functions returns such types.
2558:
2559: @findex FUNCTION_VALUE_REGNO_P
2560: @item FUNCTION_VALUE_REGNO_P (@var{regno})
2561: A C expression that is nonzero if @var{regno} is the number of a hard
2562: register in which the values of called function may come back.
2563:
2564: A register whose use for returning values is limited to serving as the
2565: second of a pair (for a value of type @code{double}, say) need not be
2566: recognized by this macro. So for most machines, this definition
2567: suffices:
2568:
2569: @example
2570: #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
2571: @end example
2572:
2573: If the machine has register windows, so that the caller and the called
2574: function use different registers for the return value, this macro
2575: should recognize only the caller's register numbers.
2576:
2577: @findex APPLY_RESULT_SIZE
2578: @item APPLY_RESULT_SIZE
2579: Define this macro if @samp{untyped_call} and @samp{untyped_return}
2580: need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
2581: saving and restoring an arbitrary return value.
2582: @end table
2583:
2584: @node Aggregate Return
2585: @subsection How Large Values Are Returned
2586: @cindex aggregates as return values
2587: @cindex large return values
2588: @cindex returning aggregate values
2589: @cindex structure value address
2590:
2591: When a function value's mode is @code{BLKmode} (and in some other
2592: cases), the value is not returned according to @code{FUNCTION_VALUE}
2593: (@pxref{Scalar Return}). Instead, the caller passes the address of a
2594: block of memory in which the value should be stored. This address
2595: is called the @dfn{structure value address}.
2596:
2597: This section describes how to control returning structure values in
2598: memory.
2599:
2600: @table @code
2601: @findex RETURN_IN_MEMORY
2602: @item RETURN_IN_MEMORY (@var{type})
2603: A C expression which can inhibit the returning of certain function
2604: values in registers, based on the type of value. A nonzero value says
2605: to return the function value in memory, just as large structures are
2606: always returned. Here @var{type} will be a C expression of type
2607: @code{tree}, representing the data type of the value.
2608:
2609: Note that values of mode @code{BLKmode} must be explicitly handled
2610: by this macro. Also, the option @samp{-fpcc-struct-return}
2611: takes effect regardless of this macro. On most systems, it is
2612: possible to leave the macro undefined; this causes a default
2613: definition to be used, whose value is the constant 1 for @code{BLKmode}
2614: values, and 0 otherwise.
2615:
2616: Do not use this macro to indicate that structures and unions should always
2617: be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
2618: to indicate this.
2619:
2620: @findex DEFAULT_PCC_STRUCT_RETURN
2621: @item DEFAULT_PCC_STRUCT_RETURN
2622: Define this macro to be 1 if all structure and union return values must be
2623: in memory. Since this results in slower code, this should be defined
2624: only if needed for compatibility with other compilers or with an ABI.
2625: If you define this macro to be 0, then the conventions used for structure
2626: and union return values are decided by the @code{RETURN_IN_MEMORY} macro.
2627:
2628: If not defined, this defaults to the value 1.
2629:
2630: @findex STRUCT_VALUE_REGNUM
2631: @item STRUCT_VALUE_REGNUM
2632: If the structure value address is passed in a register, then
2633: @code{STRUCT_VALUE_REGNUM} should be the number of that register.
2634:
2635: @findex STRUCT_VALUE
2636: @item STRUCT_VALUE
2637: If the structure value address is not passed in a register, define
2638: @code{STRUCT_VALUE} as an expression returning an RTX for the place
2639: where the address is passed. If it returns 0, the address is passed as
2640: an ``invisible'' first argument.
2641:
2642: @findex STRUCT_VALUE_INCOMING_REGNUM
2643: @item STRUCT_VALUE_INCOMING_REGNUM
2644: On some architectures the place where the structure value address
2645: is found by the called function is not the same place that the
2646: caller put it. This can be due to register windows, or it could
2647: be because the function prologue moves it to a different place.
2648:
2649: If the incoming location of the structure value address is in a
2650: register, define this macro as the register number.
2651:
2652: @findex STRUCT_VALUE_INCOMING
2653: @item STRUCT_VALUE_INCOMING
2654: If the incoming location is not a register, then you should define
2655: @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the
2656: called function should find the value. If it should find the value on
2657: the stack, define this to create a @code{mem} which refers to the frame
2658: pointer. A definition of 0 means that the address is passed as an
2659: ``invisible'' first argument.
2660:
2661: @findex PCC_STATIC_STRUCT_RETURN
2662: @item PCC_STATIC_STRUCT_RETURN
2663: Define this macro if the usual system convention on the target machine
2664: for returning structures and unions is for the called function to return
2665: the address of a static variable containing the value.
2666:
2667: Do not define this if the usual system convention is for the caller to
2668: pass an address to the subroutine.
2669:
2670: This macro has effect in @samp{-fpcc-struct-return} mode, but it does
2671: nothing when you use @samp{-freg-struct-return} mode.
2672: @end table
2673:
2674: @node Caller Saves
2675: @subsection Caller-Saves Register Allocation
2676:
2677: If you enable it, GNU CC can save registers around function calls. This
2678: makes it possible to use call-clobbered registers to hold variables that
2679: must live across calls.
2680:
2681: @table @code
2682: @findex DEFAULT_CALLER_SAVES
2683: @item DEFAULT_CALLER_SAVES
2684: Define this macro if function calls on the target machine do not preserve
2685: any registers; in other words, if @code{CALL_USED_REGISTERS} has 1
2686: for all registers. This macro enables @samp{-fcaller-saves} by default.
2687: Eventually that option will be enabled by default on all machines and both
2688: the option and this macro will be eliminated.
2689:
2690: @findex CALLER_SAVE_PROFITABLE
2691: @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
2692: A C expression to determine whether it is worthwhile to consider placing
2693: a pseudo-register in a call-clobbered hard register and saving and
2694: restoring it around each function call. The expression should be 1 when
2695: this is worth doing, and 0 otherwise.
2696:
2697: If you don't define this macro, a default is used which is good on most
2698: machines: @code{4 * @var{calls} < @var{refs}}.
2699: @end table
2700:
2701: @node Function Entry
2702: @subsection Function Entry and Exit
2703: @cindex function entry and exit
2704: @cindex prologue
2705: @cindex epilogue
2706:
2707: This section describes the macros that output function entry
2708: (@dfn{prologue}) and exit (@dfn{epilogue}) code.
2709:
2710: @table @code
2711: @findex FUNCTION_PROLOGUE
2712: @item FUNCTION_PROLOGUE (@var{file}, @var{size})
2713: A C compound statement that outputs the assembler code for entry to a
2714: function. The prologue is responsible for setting up the stack frame,
2715: initializing the frame pointer register, saving registers that must be
2716: saved, and allocating @var{size} additional bytes of storage for the
2717: local variables. @var{size} is an integer. @var{file} is a stdio
2718: stream to which the assembler code should be output.
2719:
2720: The label for the beginning of the function need not be output by this
2721: macro. That has already been done when the macro is run.
2722:
2723: @findex regs_ever_live
2724: To determine which registers to save, the macro can refer to the array
2725: @code{regs_ever_live}: element @var{r} is nonzero if hard register
2726: @var{r} is used anywhere within the function. This implies the function
2727: prologue should save register @var{r}, provided it is not one of the
2728: call-used registers. (@code{FUNCTION_EPILOGUE} must likewise use
2729: @code{regs_ever_live}.)
2730:
2731: On machines that have ``register windows'', the function entry code does
2732: not save on the stack the registers that are in the windows, even if
2733: they are supposed to be preserved by function calls; instead it takes
2734: appropriate steps to ``push'' the register stack, if any non-call-used
2735: registers are used in the function.
2736:
2737: @findex frame_pointer_needed
2738: On machines where functions may or may not have frame-pointers, the
2739: function entry code must vary accordingly; it must set up the frame
2740: pointer if one is wanted, and not otherwise. To determine whether a
2741: frame pointer is in wanted, the macro can refer to the variable
2742: @code{frame_pointer_needed}. The variable's value will be 1 at run
2743: time in a function that needs a frame pointer. @xref{Elimination}.
2744:
2745: The function entry code is responsible for allocating any stack space
2746: required for the function. This stack space consists of the regions
2747: listed below. In most cases, these regions are allocated in the
2748: order listed, with the last listed region closest to the top of the
2749: stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
2750: the highest address if it is not defined). You can use a different order
2751: for a machine if doing so is more convenient or required for
2752: compatibility reasons. Except in cases where required by standard
2753: or by a debugger, there is no reason why the stack layout used by GCC
2754: need agree with that used by other compilers for a machine.
2755:
2756: @itemize @bullet
2757: @item
2758: @findex current_function_pretend_args_size
2759: A region of @code{current_function_pretend_args_size} bytes of
2760: uninitialized space just underneath the first argument arriving on the
2761: stack. (This may not be at the very start of the allocated stack region
2762: if the calling sequence has pushed anything else since pushing the stack
2763: arguments. But usually, on such machines, nothing else has been pushed
2764: yet, because the function prologue itself does all the pushing.) This
2765: region is used on machines where an argument may be passed partly in
2766: registers and partly in memory, and, in some cases to support the
2767: features in @file{varargs.h} and @file{stdargs.h}.
2768:
2769: @item
2770: An area of memory used to save certain registers used by the function.
2771: The size of this area, which may also include space for such things as
2772: the return address and pointers to previous stack frames, is
2773: machine-specific and usually depends on which registers have been used
2774: in the function. Machines with register windows often do not require
2775: a save area.
2776:
2777: @item
2778: A region of at least @var{size} bytes, possibly rounded up to an allocation
2779: boundary, to contain the local variables of the function. On some machines,
2780: this region and the save area may occur in the opposite order, with the
2781: save area closer to the top of the stack.
2782:
2783: @item
2784: @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
2785: Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
2786: @code{current_function_outgoing_args_size} bytes to be used for outgoing
2787: argument lists of the function. @xref{Stack Arguments}.
2788: @end itemize
2789:
2790: Normally, it is necessary for the macros @code{FUNCTION_PROLOGUE} and
2791: @code{FUNCTION_EPILOGUE} to treat leaf functions specially. The C
2792: variable @code{leaf_function} is nonzero for such a function.
2793:
2794: @findex EXIT_IGNORE_STACK
2795: @item EXIT_IGNORE_STACK
2796: Define this macro as a C expression that is nonzero if the return
2797: instruction or the function epilogue ignores the value of the stack
2798: pointer; in other words, if it is safe to delete an instruction to
2799: adjust the stack pointer before a return from the function.
2800:
2801: Note that this macro's value is relevant only for functions for which
2802: frame pointers are maintained. It is never safe to delete a final
2803: stack adjustment in a function that has no frame pointer, and the
2804: compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
2805:
2806: @findex FUNCTION_EPILOGUE
2807: @item FUNCTION_EPILOGUE (@var{file}, @var{size})
2808: A C compound statement that outputs the assembler code for exit from a
2809: function. The epilogue is responsible for restoring the saved
2810: registers and stack pointer to their values when the function was
2811: called, and returning control to the caller. This macro takes the
2812: same arguments as the macro @code{FUNCTION_PROLOGUE}, and the
2813: registers to restore are determined from @code{regs_ever_live} and
2814: @code{CALL_USED_REGISTERS} in the same way.
2815:
2816: On some machines, there is a single instruction that does all the work
2817: of returning from the function. On these machines, give that
2818: instruction the name @samp{return} and do not define the macro
2819: @code{FUNCTION_EPILOGUE} at all.
2820:
2821: Do not define a pattern named @samp{return} if you want the
2822: @code{FUNCTION_EPILOGUE} to be used. If you want the target switches
2823: to control whether return instructions or epilogues are used, define a
2824: @samp{return} pattern with a validity condition that tests the target
2825: switches appropriately. If the @samp{return} pattern's validity
2826: condition is false, epilogues will be used.
2827:
2828: On machines where functions may or may not have frame-pointers, the
2829: function exit code must vary accordingly. Sometimes the code for these
2830: two cases is completely different. To determine whether a frame pointer
2831: is wanted, the macro can refer to the variable
2832: @code{frame_pointer_needed}. The variable's value will be 1 when compiling
2833: a function that needs a frame pointer.
2834:
2835: Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must
2836: treat leaf functions specially. The C variable @code{leaf_function} is
2837: nonzero for such a function. @xref{Leaf Functions}.
2838:
2839: On some machines, some functions pop their arguments on exit while
2840: others leave that for the caller to do. For example, the 68020 when
2841: given @samp{-mrtd} pops arguments in functions that take a fixed
2842: number of arguments.
2843:
2844: @findex current_function_pops_args
2845: Your definition of the macro @code{RETURN_POPS_ARGS} decides which
2846: functions pop their own arguments. @code{FUNCTION_EPILOGUE} needs to
2847: know what was decided. The variable that is called
2848: @code{current_function_pops_args} is the number of bytes of its
2849: arguments that a function should pop. @xref{Scalar Return}.
2850: @c what is the "its arguments" in the above sentence referring to, pray
2851: @c tell? --mew 5feb93
2852:
2853: @findex DELAY_SLOTS_FOR_EPILOGUE
2854: @item DELAY_SLOTS_FOR_EPILOGUE
2855: Define this macro if the function epilogue contains delay slots to which
2856: instructions from the rest of the function can be ``moved''. The
2857: definition should be a C expression whose value is an integer
2858: representing the number of delay slots there.
2859:
2860: @findex ELIGIBLE_FOR_EPILOGUE_DELAY
2861: @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
2862: A C expression that returns 1 if @var{insn} can be placed in delay
2863: slot number @var{n} of the epilogue.
2864:
2865: The argument @var{n} is an integer which identifies the delay slot now
2866: being considered (since different slots may have different rules of
2867: eligibility). It is never negative and is always less than the number
2868: of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
2869: If you reject a particular insn for a given delay slot, in principle, it
2870: may be reconsidered for a subsequent delay slot. Also, other insns may
2871: (at least in principle) be considered for the so far unfilled delay
2872: slot.
2873:
2874: @findex current_function_epilogue_delay_list
2875: @findex final_scan_insn
2876: The insns accepted to fill the epilogue delay slots are put in an RTL
2877: list made with @code{insn_list} objects, stored in the variable
2878: @code{current_function_epilogue_delay_list}. The insn for the first
2879: delay slot comes first in the list. Your definition of the macro
2880: @code{FUNCTION_EPILOGUE} should fill the delay slots by outputting the
2881: insns in this list, usually by calling @code{final_scan_insn}.
2882:
2883: You need not define this macro if you did not define
2884: @code{DELAY_SLOTS_FOR_EPILOGUE}.
2885: @end table
2886:
2887: @node Profiling
2888: @subsection Generating Code for Profiling
2889: @cindex profiling, code generation
2890:
2891: These macros will help you generate code for profiling.
2892:
2893: @table @code
2894: @findex FUNCTION_PROFILER
2895: @item FUNCTION_PROFILER (@var{file}, @var{labelno})
2896: A C statement or compound statement to output to @var{file} some
2897: assembler code to call the profiling subroutine @code{mcount}.
2898: Before calling, the assembler code must load the address of a
2899: counter variable into a register where @code{mcount} expects to
2900: find the address. The name of this variable is @samp{LP} followed
2901: by the number @var{labelno}, so you would generate the name using
2902: @samp{LP%d} in a @code{fprintf}.
2903:
2904: @findex mcount
2905: The details of how the address should be passed to @code{mcount} are
2906: determined by your operating system environment, not by GNU CC. To
2907: figure them out, compile a small program for profiling using the
2908: system's installed C compiler and look at the assembler code that
2909: results.
2910:
2911: @findex PROFILE_BEFORE_PROLOGUE
2912: @item PROFILE_BEFORE_PROLOGUE
2913: Define this macro if the code for function profiling should come before
2914: the function prologue. Normally, the profiling code comes after.
2915:
2916: @findex FUNCTION_BLOCK_PROFILER
2917: @findex __bb_init_func
2918: @item FUNCTION_BLOCK_PROFILER (@var{file}, @var{labelno})
2919: A C statement or compound statement to output to @var{file} some
2920: assembler code to initialize basic-block profiling for the current
2921: object module. This code should call the subroutine
2922: @code{__bb_init_func} once per object module, passing it as its sole
2923: argument the address of a block allocated in the object module.
2924:
2925: The name of the block is a local symbol made with this statement:
2926:
2927: @example
2928: ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0);
2929: @end example
2930:
2931: Of course, since you are writing the definition of
2932: @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
2933: can take a short cut in the definition of this macro and use the name
2934: that you know will result.
2935:
2936: The first word of this block is a flag which will be nonzero if the
2937: object module has already been initialized. So test this word first,
2938: and do not call @code{__bb_init_func} if the flag is nonzero.
2939:
2940: @findex BLOCK_PROFILER
2941: @item BLOCK_PROFILER (@var{file}, @var{blockno})
2942: A C statement or compound statement to increment the count associated
2943: with the basic block number @var{blockno}. Basic blocks are numbered
2944: separately from zero within each compilation. The count associated
2945: with block number @var{blockno} is at index @var{blockno} in a vector
2946: of words; the name of this array is a local symbol made with this
2947: statement:
2948:
2949: @example
2950: ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 2);
2951: @end example
2952:
2953: @c This paragraph is the same as one a few paragraphs up.
2954: @c That is not an error.
2955: Of course, since you are writing the definition of
2956: @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you
2957: can take a short cut in the definition of this macro and use the name
2958: that you know will result.
2959:
2960: @findex BLOCK_PROFILER_CODE
2961: @item BLOCK_PROFILER_CODE
2962: A C function or functions which are needed in the library to
2963: support block profiling.
2964: @end table
2965:
2966: @node Varargs
2967: @section Implementing the Varargs Macros
2968: @cindex varargs implementation
2969:
2970: GNU CC comes with an implementation of @file{varargs.h} and
2971: @file{stdarg.h} that work without change on machines that pass arguments
2972: on the stack. Other machines require their own implementations of
2973: varargs, and the two machine independent header files must have
2974: conditionals to include it.
2975:
2976: ANSI @file{stdarg.h} differs from traditional @file{varargs.h} mainly in
2977: the calling convention for @code{va_start}. The traditional
2978: implementation takes just one argument, which is the variable in which
2979: to store the argument pointer. The ANSI implementation of
2980: @code{va_start} takes an additional second argument. The user is
2981: supposed to write the last named argument of the function here.
2982:
2983: However, @code{va_start} should not use this argument. The way to find
2984: the end of the named arguments is with the built-in functions described
2985: below.
2986:
2987: @table @code
2988: @findex __builtin_saveregs
2989: @item __builtin_saveregs ()
2990: Use this built-in function to save the argument registers in memory so
2991: that the varargs mechanism can access them. Both ANSI and traditional
2992: versions of @code{va_start} must use @code{__builtin_saveregs}, unless
2993: you use @code{SETUP_INCOMING_VARARGS} (see below) instead.
2994:
2995: On some machines, @code{__builtin_saveregs} is open-coded under the
2996: control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines,
2997: it calls a routine written in assembler language, found in
2998: @file{libgcc2.c}.
2999:
3000: Code generated for the call to @code{__builtin_saveregs} appears at the
3001: beginning of the function, as opposed to where the call to
3002: @code{__builtin_saveregs} is written, regardless of what the code is.
3003: This is because the registers must be saved before the function starts
3004: to use them for its own purposes.
3005: @c i rewrote the first sentence above to fix an overfull hbox. --mew
3006: @c 10feb93
3007:
3008: @findex __builtin_args_info
3009: @item __builtin_args_info (@var{category})
3010: Use this built-in function to find the first anonymous arguments in
3011: registers.
3012:
3013: In general, a machine may have several categories of registers used for
3014: arguments, each for a particular category of data types. (For example,
3015: on some machines, floating-point registers are used for floating-point
3016: arguments while other arguments are passed in the general registers.)
3017: To make non-varargs functions use the proper calling convention, you
3018: have defined the @code{CUMULATIVE_ARGS} data type to record how many
3019: registers in each category have been used so far
3020:
3021: @code{__builtin_args_info} accesses the same data structure of type
3022: @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
3023: with it, with @var{category} specifying which word to access. Thus, the
3024: value indicates the first unused register in a given category.
3025:
3026: Normally, you would use @code{__builtin_args_info} in the implementation
3027: of @code{va_start}, accessing each category just once and storing the
3028: value in the @code{va_list} object. This is because @code{va_list} will
3029: have to update the values, and there is no way to alter the
3030: values accessed by @code{__builtin_args_info}.
3031:
3032: @findex __builtin_next_arg
3033: @item __builtin_next_arg ()
3034: This is the equivalent of @code{__builtin_args_info}, for stack
3035: arguments. It returns the address of the first anonymous stack
3036: argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
3037: returns the address of the location above the first anonymous stack
3038: argument. Use it in @code{va_start} to initialize the pointer for
3039: fetching arguments from the stack.
3040:
3041: @findex __builtin_classify_type
3042: @item __builtin_classify_type (@var{object})
3043: Since each machine has its own conventions for which data types are
3044: passed in which kind of register, your implementation of @code{va_arg}
3045: has to embody these conventions. The easiest way to categorize the
3046: specified data type is to use @code{__builtin_classify_type} together
3047: with @code{sizeof} and @code{__alignof__}.
3048:
3049: @code{__builtin_classify_type} ignores the value of @var{object},
3050: considering only its data type. It returns an integer describing what
3051: kind of type that is---integer, floating, pointer, structure, and so on.
3052:
3053: The file @file{typeclass.h} defines an enumeration that you can use to
3054: interpret the values of @code{__builtin_classify_type}.
3055: @end table
3056:
3057: These machine description macros help implement varargs:
3058:
3059: @table @code
3060: @findex EXPAND_BUILTIN_SAVEREGS
3061: @item EXPAND_BUILTIN_SAVEREGS (@var{args})
3062: If defined, is a C expression that produces the machine-specific code
3063: for a call to @code{__builtin_saveregs}. This code will be moved to the
3064: very beginning of the function, before any parameter access are made.
3065: The return value of this function should be an RTX that contains the
3066: value to use as the return of @code{__builtin_saveregs}.
3067:
3068: The argument @var{args} is a @code{tree_list} containing the arguments
3069: that were passed to @code{__builtin_saveregs}.
3070:
3071: If this macro is not defined, the compiler will output an ordinary
3072: call to the library function @samp{__builtin_saveregs}.
3073:
3074: @c !!! a bug in texinfo; how to make the entry on the @item line allow
3075: @c more than one line of text... help... --mew 10feb93
3076: @findex SETUP_INCOMING_VARARGS
3077: @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type},
3078: @var{pretend_args_size}, @var{second_time})
3079: This macro offers an alternative to using @code{__builtin_saveregs} and
3080: defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the
3081: anonymous register arguments into the stack so that all the arguments
3082: appear to have been passed consecutively on the stack. Once this is
3083: done, you can use the standard implementation of varargs that works for
3084: machines that pass all their arguments on the stack.
3085:
3086: The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data
3087: structure, containing the values that obtain after processing of the
3088: named arguments. The arguments @var{mode} and @var{type} describe the
3089: last named argument---its machine mode and its data type as a tree node.
3090:
3091: The macro implementation should do two things: first, push onto the
3092: stack all the argument registers @emph{not} used for the named
3093: arguments, and second, store the size of the data thus pushed into the
3094: @code{int}-valued variable whose name is supplied as the argument
3095: @var{pretend_args_size}. The value that you store here will serve as
3096: additional offset for setting up the stack frame.
3097:
3098: Because you must generate code to push the anonymous arguments at
3099: compile time without knowing their data types,
3100: @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just
3101: a single category of argument register and use it uniformly for all data
3102: types.
3103:
3104: If the argument @var{second_time} is nonzero, it means that the
3105: arguments of the function are being analyzed for the second time. This
3106: happens for an inline function, which is not actually compiled until the
3107: end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should
3108: not generate any instructions in this case.
3109: @end table
3110:
3111: @node Trampolines
3112: @section Trampolines for Nested Functions
3113: @cindex trampolines for nested functions
3114: @cindex nested functions, trampolines for
3115:
3116: A @dfn{trampoline} is a small piece of code that is created at run time
3117: when the address of a nested function is taken. It normally resides on
3118: the stack, in the stack frame of the containing function. These macros
3119: tell GNU CC how to generate code to allocate and initialize a
3120: trampoline.
3121:
3122: The instructions in the trampoline must do two things: load a constant
3123: address into the static chain register, and jump to the real address of
3124: the nested function. On CISC machines such as the m68k, this requires
3125: two instructions, a move immediate and a jump. Then the two addresses
3126: exist in the trampoline as word-long immediate operands. On RISC
3127: machines, it is often necessary to load each address into a register in
3128: two parts. Then pieces of each address form separate immediate
3129: operands.
3130:
3131: The code generated to initialize the trampoline must store the variable
3132: parts---the static chain value and the function address---into the
3133: immediate operands of the instructions. On a CISC machine, this is
3134: simply a matter of copying each address to a memory reference at the
3135: proper offset from the start of the trampoline. On a RISC machine, it
3136: may be necessary to take out pieces of the address and store them
3137: separately.
3138:
3139: @table @code
3140: @findex TRAMPOLINE_TEMPLATE
3141: @item TRAMPOLINE_TEMPLATE (@var{file})
3142: A C statement to output, on the stream @var{file}, assembler code for a
3143: block of data that contains the constant parts of a trampoline. This
3144: code should not include a label---the label is taken care of
3145: automatically.
3146:
3147: @findex TRAMPOLINE_SECTION
3148: @item TRAMPOLINE_SECTION
3149: The name of a subroutine to switch to the section in which the
3150: trampoline template is to be placed (@pxref{Sections}). The default is
3151: a value of @samp{readonly_data_section}, which places the trampoline in
3152: the section containing read-only data.
3153:
3154: @findex TRAMPOLINE_SIZE
3155: @item TRAMPOLINE_SIZE
3156: A C expression for the size in bytes of the trampoline, as an integer.
3157:
3158: @findex TRAMPOLINE_ALIGNMENT
3159: @item TRAMPOLINE_ALIGNMENT
3160: Alignment required for trampolines, in bits.
3161:
3162: If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT}
3163: is used for aligning trampolines.
3164:
3165: @findex INITIALIZE_TRAMPOLINE
3166: @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain})
3167: A C statement to initialize the variable parts of a trampoline.
3168: @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is
3169: an RTX for the address of the nested function; @var{static_chain} is an
3170: RTX for the static chain value that should be passed to the function
3171: when it is called.
3172:
3173: @findex ALLOCATE_TRAMPOLINE
3174: @item ALLOCATE_TRAMPOLINE (@var{fp})
3175: A C expression to allocate run-time space for a trampoline. The
3176: expression value should be an RTX representing a memory reference to the
3177: space for the trampoline.
3178:
3179: @cindex @code{FUNCTION_EPILOGUE} and trampolines
3180: @cindex @code{FUNCTION_PROLOGUE} and trampolines
3181: If this macro is not defined, by default the trampoline is allocated as
3182: a stack slot. This default is right for most machines. The exceptions
3183: are machines where it is impossible to execute instructions in the stack
3184: area. On such machines, you may have to implement a separate stack,
3185: using this macro in conjunction with @code{FUNCTION_PROLOGUE} and
3186: @code{FUNCTION_EPILOGUE}.
3187:
3188: @var{fp} points to a data structure, a @code{struct function}, which
3189: describes the compilation status of the immediate containing function of
3190: the function which the trampoline is for. Normally (when
3191: @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the
3192: trampoline is in the stack frame of this containing function. Other
3193: allocation strategies probably must do something analogous with this
3194: information.
3195: @end table
3196:
3197: Implementing trampolines is difficult on many machines because they have
3198: separate instruction and data caches. Writing into a stack location
3199: fails to clear the memory in the instruction cache, so when the program
3200: jumps to that location, it executes the old contents.
3201:
3202: Here are two possible solutions. One is to clear the relevant parts of
3203: the instruction cache whenever a trampoline is set up. The other is to
3204: make all trampolines identical, by having them jump to a standard
3205: subroutine. The former technique makes trampoline execution faster; the
3206: latter makes initialization faster.
3207:
3208: To clear the instruction cache when a trampoline is initialized, define
3209: the following macros which describe the shape of the cache.
3210:
3211: @table @code
3212: @findex INSN_CACHE_SIZE
3213: @item INSN_CACHE_SIZE
3214: The total size in bytes of the cache.
3215:
3216: @findex INSN_CACHE_LINE_WIDTH
3217: @item INSN_CACHE_LINE_WIDTH
3218: The length in bytes of each cache line. The cache is divided into cache
3219: lines which are disjoint slots, each holding a contiguous chunk of data
3220: fetched from memory. Each time data is brought into the cache, an
3221: entire line is read at once. The data loaded into a cache line is
3222: always aligned on a boundary equal to the line size.
3223:
3224: @findex INSN_CACHE_DEPTH
3225: @item INSN_CACHE_DEPTH
3226: The number of alternative cache lines that can hold any particular memory
3227: location.
3228: @end table
3229:
3230: Alternatively, if the machine has system calls or instructions to clear
3231: the instruction cache directly, you can define the following macro.
3232:
3233: @table @code
3234: @findex CLEAR_INSN_CACHE (BEG, END)
3235: @item
3236: If defined, expands to a C expression clearing the @emph{instruction
3237: cache} in the specified interval. If it is not defined, and the macro
3238: INSN_CACHE_SIZE is defined, some generic code is generated to clear the
3239: cache. The definition of this macro would typically be a series of
3240: @code{asm} statements. Both @var{BEG} and @var{END} are both pointer
3241: expressions.
3242: @end table
3243:
3244: To use a standard subroutine, define the following macro. In addition,
3245: you must make sure that the instructions in a trampoline fill an entire
3246: cache line with identical instructions, or else ensure that the
3247: beginning of the trampoline code is always aligned at the same point in
3248: its cache line. Look in @file{m68k.h} as a guide.
3249:
3250: @table @code
3251: @findex TRANSFER_FROM_TRAMPOLINE
3252: @item TRANSFER_FROM_TRAMPOLINE
3253: Define this macro if trampolines need a special subroutine to do their
3254: work. The macro should expand to a series of @code{asm} statements
3255: which will be compiled with GNU CC. They go in a library function named
3256: @code{__transfer_from_trampoline}.
3257:
3258: If you need to avoid executing the ordinary prologue code of a compiled
3259: C function when you jump to the subroutine, you can do so by placing a
3260: special label of your own in the assembler code. Use one @code{asm}
3261: statement to generate an assembler label, and another to make the label
3262: global. Then trampolines can use that label to jump directly to your
3263: special assembler code.
3264: @end table
3265:
3266: @node Library Calls
3267: @section Implicit Calls to Library Routines
3268: @cindex library subroutine names
3269: @cindex @file{libgcc.a}
3270:
3271: @table @code
3272: @findex MULSI3_LIBCALL
3273: @item MULSI3_LIBCALL
3274: A C string constant giving the name of the function to call for
3275: multiplication of one signed full-word by another. If you do not
3276: define this macro, the default name is used, which is @code{__mulsi3},
3277: a function defined in @file{libgcc.a}.
3278:
3279: @findex DIVSI3_LIBCALL
3280: @item DIVSI3_LIBCALL
3281: A C string constant giving the name of the function to call for
3282: division of one signed full-word by another. If you do not define
3283: this macro, the default name is used, which is @code{__divsi3}, a
3284: function defined in @file{libgcc.a}.
3285:
3286: @findex UDIVSI3_LIBCALL
3287: @item UDIVSI3_LIBCALL
3288: A C string constant giving the name of the function to call for
3289: division of one unsigned full-word by another. If you do not define
3290: this macro, the default name is used, which is @code{__udivsi3}, a
3291: function defined in @file{libgcc.a}.
3292:
3293: @findex MODSI3_LIBCALL
3294: @item MODSI3_LIBCALL
3295: A C string constant giving the name of the function to call for the
3296: remainder in division of one signed full-word by another. If you do
3297: not define this macro, the default name is used, which is
3298: @code{__modsi3}, a function defined in @file{libgcc.a}.
3299:
3300: @findex UMODSI3_LIBCALL
3301: @item UMODSI3_LIBCALL
3302: A C string constant giving the name of the function to call for the
3303: remainder in division of one unsigned full-word by another. If you do
3304: not define this macro, the default name is used, which is
3305: @code{__umodsi3}, a function defined in @file{libgcc.a}.
3306:
3307: @findex MULDI3_LIBCALL
3308: @item MULDI3_LIBCALL
3309: A C string constant giving the name of the function to call for
3310: multiplication of one signed double-word by another. If you do not
3311: define this macro, the default name is used, which is @code{__muldi3},
3312: a function defined in @file{libgcc.a}.
3313:
3314: @findex DIVDI3_LIBCALL
3315: @item DIVDI3_LIBCALL
3316: A C string constant giving the name of the function to call for
3317: division of one signed double-word by another. If you do not define
3318: this macro, the default name is used, which is @code{__divdi3}, a
3319: function defined in @file{libgcc.a}.
3320:
3321: @findex UDIVDI3_LIBCALL
3322: @item UDIVDI3_LIBCALL
3323: A C string constant giving the name of the function to call for
3324: division of one unsigned full-word by another. If you do not define
3325: this macro, the default name is used, which is @code{__udivdi3}, a
3326: function defined in @file{libgcc.a}.
3327:
3328: @findex MODDI3_LIBCALL
3329: @item MODDI3_LIBCALL
3330: A C string constant giving the name of the function to call for the
3331: remainder in division of one signed double-word by another. If you do
3332: not define this macro, the default name is used, which is
3333: @code{__moddi3}, a function defined in @file{libgcc.a}.
3334:
3335: @findex UMODDI3_LIBCALL
3336: @item UMODDI3_LIBCALL
3337: A C string constant giving the name of the function to call for the
3338: remainder in division of one unsigned full-word by another. If you do
3339: not define this macro, the default name is used, which is
3340: @code{__umoddi3}, a function defined in @file{libgcc.a}.
3341:
3342: @findex TARGET_EDOM
3343: @cindex @code{EDOM}, implicit usage
3344: @item TARGET_EDOM
3345: The value of @code{EDOM} on the target machine, as a C integer constant
3346: expression. If you don't define this macro, GNU CC does not attempt to
3347: deposit the value of @code{EDOM} into @code{errno} directly. Look in
3348: @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
3349: system.
3350:
3351: If you do not define @code{TARGET_EDOM}, then compiled code reports
3352: domain errors by calling the library function and letting it report the
3353: error. If mathematical functions on your system use @code{matherr} when
3354: there is an error, then you should leave @code{TARGET_EDOM} undefined so
3355: that @code{matherr} is used normally.
3356:
3357: @findex GEN_ERRNO_RTX
3358: @cindex @code{errno}, implicit usage
3359: @item GEN_ERRNO_RTX
3360: Define this macro as a C expression to create an rtl expression that
3361: refers to the global ``variable'' @code{errno}. (On certain systems,
3362: @code{errno} may not actually be a variable.) If you don't define this
3363: macro, a reasonable default is used.
3364:
3365: @findex TARGET_MEM_FUNCTIONS
3366: @cindex @code{bcopy}, implicit usage
3367: @cindex @code{memcpy}, implicit usage
3368: @cindex @code{bzero}, implicit usage
3369: @cindex @code{memset}, implicit usage
3370: @item TARGET_MEM_FUNCTIONS
3371: Define this macro if GNU CC should generate calls to the System V
3372: (and ANSI C) library functions @code{memcpy} and @code{memset}
3373: rather than the BSD functions @code{bcopy} and @code{bzero}.
3374:
3375: @findex LIBGCC_NEEDS_DOUBLE
3376: @item LIBGCC_NEEDS_DOUBLE
3377: Define this macro if only @code{float} arguments cannot be passed to
3378: library routines (so they must be converted to @code{double}). This
3379: macro affects both how library calls are generated and how the library
3380: routines in @file{libgcc1.c} accept their arguments. It is useful on
3381: machines where floating and fixed point arguments are passed
3382: differently, such as the i860.
3383:
3384: @findex FLOAT_ARG_TYPE
3385: @item FLOAT_ARG_TYPE
3386: Define this macro to override the type used by the library routines to
3387: pick up arguments of type @code{float}. (By default, they use a union
3388: of @code{float} and @code{int}.)
3389:
3390: The obvious choice would be @code{float}---but that won't work with
3391: traditional C compilers that expect all arguments declared as @code{float}
3392: to arrive as @code{double}. To avoid this conversion, the library routines
3393: ask for the value as some other type and then treat it as a @code{float}.
3394:
3395: On some systems, no other type will work for this. For these systems,
3396: you must use @code{LIBGCC_NEEDS_DOUBLE} instead, to force conversion of
3397: the values @code{double} before they are passed.
3398:
3399: @findex FLOATIFY
3400: @item FLOATIFY (@var{passed-value})
3401: Define this macro to override the way library routines redesignate a
3402: @code{float} argument as a @code{float} instead of the type it was
3403: passed as. The default is an expression which takes the @code{float}
3404: field of the union.
3405:
3406: @findex FLOAT_VALUE_TYPE
3407: @item FLOAT_VALUE_TYPE
3408: Define this macro to override the type used by the library routines to
3409: return values that ought to have type @code{float}. (By default, they
3410: use @code{int}.)
3411:
3412: The obvious choice would be @code{float}---but that won't work with
3413: traditional C compilers gratuitously convert values declared as
3414: @code{float} into @code{double}.
3415:
3416: @findex INTIFY
3417: @item INTIFY (@var{float-value})
3418: Define this macro to override the way the value of a
3419: @code{float}-returning library routine should be packaged in order to
3420: return it. These functions are actually declared to return type
3421: @code{FLOAT_VALUE_TYPE} (normally @code{int}).
3422:
3423: These values can't be returned as type @code{float} because traditional
3424: C compilers would gratuitously convert the value to a @code{double}.
3425:
3426: A local variable named @code{intify} is always available when the macro
3427: @code{INTIFY} is used. It is a union of a @code{float} field named
3428: @code{f} and a field named @code{i} whose type is
3429: @code{FLOAT_VALUE_TYPE} or @code{int}.
3430:
3431: If you don't define this macro, the default definition works by copying
3432: the value through that union.
3433:
3434: @findex nongcc_SI_type
3435: @item nongcc_SI_type
3436: Define this macro as the name of the data type corresponding to
3437: @code{SImode} in the system's own C compiler.
3438:
3439: You need not define this macro if that type is @code{long int}, as it usually
3440: is.
3441:
3442: @findex nongcc_word_type
3443: @item nongcc_word_type
3444: Define this macro as the name of the data type corresponding to the
3445: word_mode in the system's own C compiler.
3446:
3447: You need not define this macro if that type is @code{long int}, as it usually
3448: is.
3449:
3450: @findex perform_@dots{}
3451: @item perform_@dots{}
3452: Define these macros to supply explicit C statements to carry out various
3453: arithmetic operations on types @code{float} and @code{double} in the
3454: library routines in @file{libgcc1.c}. See that file for a full list
3455: of these macros and their arguments.
3456:
3457: On most machines, you don't need to define any of these macros, because
3458: the C compiler that comes with the system takes care of doing them.
3459:
3460: @findex NEXT_OBJC_RUNTIME
3461: @item NEXT_OBJC_RUNTIME
3462: Define this macro to generate code for Objective C message sending using
3463: the calling convention of the NeXT system. This calling convention
3464: involves passing the object, the selector and the method arguments all
3465: at once to the method-lookup library function.
3466:
3467: The default calling convention passes just the object and the selector
3468: to the lookup function, which returns a pointer to the method.
3469: @end table
3470:
3471: @node Addressing Modes
3472: @section Addressing Modes
3473: @cindex addressing modes
3474:
3475: @table @code
3476: @findex HAVE_POST_INCREMENT
3477: @item HAVE_POST_INCREMENT
3478: Define this macro if the machine supports post-increment addressing.
3479:
3480: @findex HAVE_PRE_INCREMENT
3481: @findex HAVE_POST_DECREMENT
3482: @findex HAVE_PRE_DECREMENT
3483: @item HAVE_PRE_INCREMENT
3484: @itemx HAVE_POST_DECREMENT
3485: @itemx HAVE_PRE_DECREMENT
3486: Similar for other kinds of addressing.
3487:
3488: @findex CONSTANT_ADDRESS_P
3489: @item CONSTANT_ADDRESS_P (@var{x})
3490: A C expression that is 1 if the RTX @var{x} is a constant which
3491: is a valid address. On most machines, this can be defined as
3492: @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
3493: in which constant addresses are supported.
3494:
3495: @findex CONSTANT_P
3496: @code{CONSTANT_P} accepts integer-values expressions whose values are
3497: not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and
3498: @code{high} expressions and @code{const} arithmetic expressions, in
3499: addition to @code{const_int} and @code{const_double} expressions.
3500:
3501: @findex MAX_REGS_PER_ADDRESS
3502: @item MAX_REGS_PER_ADDRESS
3503: A number, the maximum number of registers that can appear in a valid
3504: memory address. Note that it is up to you to specify a value equal to
3505: the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever
3506: accept.
3507:
3508: @findex GO_IF_LEGITIMATE_ADDRESS
3509: @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
3510: A C compound statement with a conditional @code{goto @var{label};}
3511: executed if @var{x} (an RTX) is a legitimate memory address on the
3512: target machine for a memory operand of mode @var{mode}.
3513:
3514: It usually pays to define several simpler macros to serve as
3515: subroutines for this one. Otherwise it may be too complicated to
3516: understand.
3517:
3518: This macro must exist in two variants: a strict variant and a
3519: non-strict one. The strict variant is used in the reload pass. It
3520: must be defined so that any pseudo-register that has not been
3521: allocated a hard register is considered a memory reference. In
3522: contexts where some kind of register is required, a pseudo-register
3523: with no hard register must be rejected.
3524:
3525: The non-strict variant is used in other passes. It must be defined to
3526: accept all pseudo-registers in every context where some kind of
3527: register is required.
3528:
3529: @findex REG_OK_STRICT
3530: Compiler source files that want to use the strict variant of this
3531: macro define the macro @code{REG_OK_STRICT}. You should use an
3532: @code{#ifdef REG_OK_STRICT} conditional to define the strict variant
3533: in that case and the non-strict variant otherwise.
3534:
3535: Subroutines to check for acceptable registers for various purposes (one
3536: for base registers, one for index registers, and so on) are typically
3537: among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}.
3538: Then only these subroutine macros need have two variants; the higher
3539: levels of macros may be the same whether strict or not.@refill
3540:
3541: Normally, constant addresses which are the sum of a @code{symbol_ref}
3542: and an integer are stored inside a @code{const} RTX to mark them as
3543: constant. Therefore, there is no need to recognize such sums
3544: specifically as legitimate addresses. Normally you would simply
3545: recognize any @code{const} as legitimate.
3546:
3547: Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
3548: sums that are not marked with @code{const}. It assumes that a naked
3549: @code{plus} indicates indexing. If so, then you @emph{must} reject such
3550: naked constant sums as illegitimate addresses, so that none of them will
3551: be given to @code{PRINT_OPERAND_ADDRESS}.
3552:
3553: @cindex @code{ENCODE_SECTION_INFO} and address validation
3554: On some machines, whether a symbolic address is legitimate depends on
3555: the section that the address refers to. On these machines, define the
3556: macro @code{ENCODE_SECTION_INFO} to store the information into the
3557: @code{symbol_ref}, and then check for it here. When you see a
3558: @code{const}, you will have to look inside it to find the
3559: @code{symbol_ref} in order to determine the section. @xref{Assembler
3560: Format}.
3561:
3562: @findex saveable_obstack
3563: The best way to modify the name string is by adding text to the
3564: beginning, with suitable punctuation to prevent any ambiguity. Allocate
3565: the new name in @code{saveable_obstack}. You will have to modify
3566: @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and
3567: output the name accordingly, and define @code{STRIP_NAME_ENCODING} to
3568: access the original name string.
3569:
3570: You can check the information stored here into the @code{symbol_ref} in
3571: the definitions of the macros @code{GO_IF_LEGITIMATE_ADDRESS} and
3572: @code{PRINT_OPERAND_ADDRESS}.
3573:
3574: @findex REG_OK_FOR_BASE_P
3575: @item REG_OK_FOR_BASE_P (@var{x})
3576: A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
3577: RTX) is valid for use as a base register. For hard registers, it
3578: should always accept those which the hardware permits and reject the
3579: others. Whether the macro accepts or rejects pseudo registers must be
3580: controlled by @code{REG_OK_STRICT} as described above. This usually
3581: requires two variant definitions, of which @code{REG_OK_STRICT}
3582: controls the one actually used.
3583:
3584: @findex REG_OK_FOR_INDEX_P
3585: @item REG_OK_FOR_INDEX_P (@var{x})
3586: A C expression that is nonzero if @var{x} (assumed to be a @code{reg}
3587: RTX) is valid for use as an index register.
3588:
3589: The difference between an index register and a base register is that
3590: the index register may be scaled. If an address involves the sum of
3591: two registers, neither one of them scaled, then either one may be
3592: labeled the ``base'' and the other the ``index''; but whichever
3593: labeling is used must fit the machine's constraints of which registers
3594: may serve in each capacity. The compiler will try both labelings,
3595: looking for one that is valid, and will reload one or both registers
3596: only if neither labeling works.
3597:
3598: @findex LEGITIMIZE_ADDRESS
3599: @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win})
3600: A C compound statement that attempts to replace @var{x} with a valid
3601: memory address for an operand of mode @var{mode}. @var{win} will be a
3602: C statement label elsewhere in the code; the macro definition may use
3603:
3604: @example
3605: GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win});
3606: @end example
3607:
3608: @noindent
3609: to avoid further processing if the address has become legitimate.
3610:
3611: @findex break_out_memory_refs
3612: @var{x} will always be the result of a call to @code{break_out_memory_refs},
3613: and @var{oldx} will be the operand that was given to that function to produce
3614: @var{x}.
3615:
3616: The code generated by this macro should not alter the substructure of
3617: @var{x}. If it transforms @var{x} into a more legitimate form, it
3618: should assign @var{x} (which will always be a C variable) a new value.
3619:
3620: It is not necessary for this macro to come up with a legitimate
3621: address. The compiler has standard ways of doing so in all cases. In
3622: fact, it is safe for this macro to do nothing. But often a
3623: machine-dependent strategy can generate better code.
3624:
3625: @findex GO_IF_MODE_DEPENDENT_ADDRESS
3626: @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
3627: A C statement or compound statement with a conditional @code{goto
3628: @var{label};} executed if memory address @var{x} (an RTX) can have
3629: different meanings depending on the machine mode of the memory
3630: reference it is used for or if the address is valid for some modes
3631: but not others.
3632:
3633: Autoincrement and autodecrement addresses typically have mode-dependent
3634: effects because the amount of the increment or decrement is the size
3635: of the operand being addressed. Some machines have other mode-dependent
3636: addresses. Many RISC machines have no mode-dependent addresses.
3637:
3638: You may assume that @var{addr} is a valid address for the machine.
3639:
3640: @findex LEGITIMATE_CONSTANT_P
3641: @item LEGITIMATE_CONSTANT_P (@var{x})
3642: A C expression that is nonzero if @var{x} is a legitimate constant for
3643: an immediate operand on the target machine. You can assume that
3644: @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
3645: @samp{1} is a suitable definition for this macro on machines where
3646: anything @code{CONSTANT_P} is valid.@refill
3647: @end table
3648:
3649: @node Condition Code
3650: @section Condition Code Status
3651: @cindex condition code status
3652:
3653: @findex cc_status
3654: The file @file{conditions.h} defines a variable @code{cc_status} to
3655: describe how the condition code was computed (in case the interpretation of
3656: the condition code depends on the instruction that it was set by). This
3657: variable contains the RTL expressions on which the condition code is
3658: currently based, and several standard flags.
3659:
3660: Sometimes additional machine-specific flags must be defined in the machine
3661: description header file. It can also add additional machine-specific
3662: information by defining @code{CC_STATUS_MDEP}.
3663:
3664: @table @code
3665: @findex CC_STATUS_MDEP
3666: @item CC_STATUS_MDEP
3667: C code for a data type which is used for declaring the @code{mdep}
3668: component of @code{cc_status}. It defaults to @code{int}.
3669:
3670: This macro is not used on machines that do not use @code{cc0}.
3671:
3672: @findex CC_STATUS_MDEP_INIT
3673: @item CC_STATUS_MDEP_INIT
3674: A C expression to initialize the @code{mdep} field to ``empty''.
3675: The default definition does nothing, since most machines don't use
3676: the field anyway. If you want to use the field, you should probably
3677: define this macro to initialize it.
3678:
3679: This macro is not used on machines that do not use @code{cc0}.
3680:
3681: @findex NOTICE_UPDATE_CC
3682: @item NOTICE_UPDATE_CC (@var{exp}, @var{insn})
3683: A C compound statement to set the components of @code{cc_status}
3684: appropriately for an insn @var{insn} whose body is @var{exp}. It is
3685: this macro's responsibility to recognize insns that set the condition
3686: code as a byproduct of other activity as well as those that explicitly
3687: set @code{(cc0)}.
3688:
3689: This macro is not used on machines that do not use @code{cc0}.
3690:
3691: If there are insns that do not set the condition code but do alter
3692: other machine registers, this macro must check to see whether they
3693: invalidate the expressions that the condition code is recorded as
3694: reflecting. For example, on the 68000, insns that store in address
3695: registers do not set the condition code, which means that usually
3696: @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
3697: insns. But suppose that the previous insn set the condition code
3698: based on location @samp{a4@@(102)} and the current insn stores a new
3699: value in @samp{a4}. Although the condition code is not changed by
3700: this, it will no longer be true that it reflects the contents of
3701: @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
3702: @code{cc_status} in this case to say that nothing is known about the
3703: condition code value.
3704:
3705: The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
3706: with the results of peephole optimization: insns whose patterns are
3707: @code{parallel} RTXs containing various @code{reg}, @code{mem} or
3708: constants which are just the operands. The RTL structure of these
3709: insns is not sufficient to indicate what the insns actually do. What
3710: @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
3711: @code{CC_STATUS_INIT}.
3712:
3713: A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
3714: that looks at an attribute (@pxref{Insn Attributes}) named, for example,
3715: @samp{cc}. This avoids having detailed information about patterns in
3716: two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
3717:
3718: @findex EXTRA_CC_MODES
3719: @item EXTRA_CC_MODES
3720: A list of names to be used for additional modes for condition code
3721: values in registers (@pxref{Jump Patterns}). These names are added
3722: to @code{enum machine_mode} and all have class @code{MODE_CC}. By
3723: convention, they should start with @samp{CC} and end with @samp{mode}.
3724:
3725: You should only define this macro if your machine does not use @code{cc0}
3726: and only if additional modes are required.
3727:
3728: @findex EXTRA_CC_NAMES
3729: @item EXTRA_CC_NAMES
3730: A list of C strings giving the names for the modes listed in
3731: @code{EXTRA_CC_MODES}. For example, the Sparc defines this macro and
3732: @code{EXTRA_CC_MODES} as
3733:
3734: @smallexample
3735: #define EXTRA_CC_MODES CC_NOOVmode, CCFPmode
3736: #define EXTRA_CC_NAMES "CC_NOOV", "CCFP"
3737: @end smallexample
3738:
3739: This macro is not required if @code{EXTRA_CC_MODES} is not defined.
3740:
3741: @findex SELECT_CC_MODE
3742: @item SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
3743: Returns a mode from class @code{MODE_CC} to be used when comparison
3744: operation code @var{op} is applied to rtx @var{x} and @var{y}. For
3745: example, on the Sparc, @code{SELECT_CC_MODE} is defined as (see
3746: @pxref{Jump Patterns} for a description of the reason for this
3747: definition)
3748:
3749: @smallexample
3750: #define SELECT_CC_MODE(OP,X,Y) \
3751: (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
3752: ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
3753: : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
3754: || GET_CODE (X) == NEG) \
3755: ? CC_NOOVmode : CCmode))
3756: @end smallexample
3757:
3758: This macro is not required if @code{EXTRA_CC_MODES} is not defined.
3759: @end table
3760:
3761: @node Costs
3762: @section Describing Relative Costs of Operations
3763: @cindex costs of instructions
3764: @cindex relative costs
3765: @cindex speed of instructions
3766:
3767: These macros let you describe the relative speed of various operations
3768: on the target machine.
3769:
3770: @table @code
3771: @findex CONST_COSTS
3772: @item CONST_COSTS (@var{x}, @var{code}, @var{outer_code})
3773: A part of a C @code{switch} statement that describes the relative costs
3774: of constant RTL expressions. It must contain @code{case} labels for
3775: expression codes @code{const_int}, @code{const}, @code{symbol_ref},
3776: @code{label_ref} and @code{const_double}. Each case must ultimately
3777: reach a @code{return} statement to return the relative cost of the use
3778: of that kind of constant value in an expression. The cost may depend on
3779: the precise value of the constant, which is available for examination in
3780: @var{x}, and the rtx code of the expression in which it is contained,
3781: found in @var{outer_code}.
3782:
3783: @var{code} is the expression code---redundant, since it can be
3784: obtained with @code{GET_CODE (@var{x})}.
3785:
3786: @findex RTX_COSTS
3787: @findex COSTS_N_INSNS
3788: @item RTX_COSTS (@var{x}, @var{code}, @var{outer_code})
3789: Like @code{CONST_COSTS} but applies to nonconstant RTL expressions.
3790: This can be used, for example, to indicate how costly a multiply
3791: instruction is. In writing this macro, you can use the construct
3792: @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
3793: instructions. @var{outer_code} is the code of the expression in which
3794: @var{x} is contained.
3795:
3796: This macro is optional; do not define it if the default cost assumptions
3797: are adequate for the target machine.
3798:
3799: @findex ADDRESS_COST
3800: @item ADDRESS_COST (@var{address})
3801: An expression giving the cost of an addressing mode that contains
3802: @var{address}. If not defined, the cost is computed from
3803: the @var{address} expression and the @code{CONST_COSTS} values.
3804:
3805: For most CISC machines, the default cost is a good approximation of the
3806: true cost of the addressing mode. However, on RISC machines, all
3807: instructions normally have the same length and execution time. Hence
3808: all addresses will have equal costs.
3809:
3810: In cases where more than one form of an address is known, the form with
3811: the lowest cost will be used. If multiple forms have the same, lowest,
3812: cost, the one that is the most complex will be used.
3813:
3814: For example, suppose an address that is equal to the sum of a register
3815: and a constant is used twice in the same basic block. When this macro
3816: is not defined, the address will be computed in a register and memory
3817: references will be indirect through that register. On machines where
3818: the cost of the addressing mode containing the sum is no higher than
3819: that of a simple indirect reference, this will produce an additional
3820: instruction and possibly require an additional register. Proper
3821: specification of this macro eliminates this overhead for such machines.
3822:
3823: Similar use of this macro is made in strength reduction of loops.
3824:
3825: @var{address} need not be valid as an address. In such a case, the cost
3826: is not relevant and can be any value; invalid addresses need not be
3827: assigned a different cost.
3828:
3829: On machines where an address involving more than one register is as
3830: cheap as an address computation involving only one register, defining
3831: @code{ADDRESS_COST} to reflect this can cause two registers to be live
3832: over a region of code where only one would have been if
3833: @code{ADDRESS_COST} were not defined in that manner. This effect should
3834: be considered in the definition of this macro. Equivalent costs should
3835: probably only be given to addresses with different numbers of registers
3836: on machines with lots of registers.
3837:
3838: This macro will normally either not be defined or be defined as a
3839: constant.
3840:
3841: @findex REGISTER_MOVE_COST
3842: @item REGISTER_MOVE_COST (@var{from}, @var{to})
3843: A C expression for the cost of moving data from a register in class
3844: @var{from} to one in class @var{to}. The classes are expressed using
3845: the enumeration values such as @code{GENERAL_REGS}. A value of 4 is the
3846: default; other values are interpreted relative to that.
3847:
3848: It is not required that the cost always equal 2 when @var{from} is the
3849: same as @var{to}; on some machines it is expensive to move between
3850: registers if they are not general registers.
3851:
3852: If reload sees an insn consisting of a single @code{set} between two
3853: hard registers, and if @code{REGISTER_MOVE_COST} applied to their
3854: classes returns a value of 2, reload does not check to ensure that the
3855: constraints of the insn are met. Setting a cost of other than 2 will
3856: allow reload to verify that the constraints are met. You should do this
3857: if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
3858:
3859: @findex MEMORY_MOVE_COST
3860: @item MEMORY_MOVE_COST (@var{m})
3861: A C expression for the cost of moving data of mode @var{m} between a
3862: register and memory. A value of 2 is the default; this cost is relative
3863: to those in @code{REGISTER_MOVE_COST}.
3864:
3865: If moving between registers and memory is more expensive than between
3866: two registers, you should define this macro to express the relative cost.
3867:
3868: @findex BRANCH_COST
3869: @item BRANCH_COST
3870: A C expression for the cost of a branch instruction. A value of 1 is
3871: the default; other values are interpreted relative to that.
3872: @end table
3873:
3874: Here are additional macros which do not specify precise relative costs,
3875: but only that certain actions are more expensive than GNU CC would
3876: ordinarily expect.
3877:
3878: @table @code
3879: @findex SLOW_BYTE_ACCESS
3880: @item SLOW_BYTE_ACCESS
3881: Define this macro as a C expression which is nonzero if accessing less
3882: than a word of memory (i.e. a @code{char} or a @code{short}) is no
3883: faster than accessing a word of memory, i.e., if such access
3884: require more than one instruction or if there is no difference in cost
3885: between byte and (aligned) word loads.
3886:
3887: When this macro is not defined, the compiler will access a field by
3888: finding the smallest containing object; when it is defined, a fullword
3889: load will be used if alignment permits. Unless bytes accesses are
3890: faster than word accesses, using word accesses is preferable since it
3891: may eliminate subsequent memory access if subsequent accesses occur to
3892: other fields in the same word of the structure, but to different bytes.
3893:
3894: @findex SLOW_ZERO_EXTEND
3895: @item SLOW_ZERO_EXTEND
3896: Define this macro if zero-extension (of a @code{char} or @code{short}
3897: to an @code{int}) can be done faster if the destination is a register
3898: that is known to be zero.
3899:
3900: If you define this macro, you must have instruction patterns that
3901: recognize RTL structures like this:
3902:
3903: @smallexample
3904: (set (strict_low_part (subreg:QI (reg:SI @dots{}) 0)) @dots{})
3905: @end smallexample
3906:
3907: @noindent
3908: and likewise for @code{HImode}.
3909:
3910: @findex SLOW_UNALIGNED_ACCESS
3911: @item SLOW_UNALIGNED_ACCESS
3912: Define this macro to be the value 1 if unaligned accesses have a cost
3913: many times greater than aligned accesses, for example if they are
3914: emulated in a trap handler.
3915:
3916: When this macro is non-zero, the compiler will act as if
3917: @code{STRICT_ALIGNMENT} were non-zero when generating code for block
3918: moves. This can cause significantly more instructions to be produced.
3919: Therefore, do not set this macro non-zero if unaligned accesses only add a
3920: cycle or two to the time for a memory access.
3921:
3922: If the value of this macro is always zero, it need not be defined.
3923:
3924: @findex DONT_REDUCE_ADDR
3925: @item DONT_REDUCE_ADDR
3926: Define this macro to inhibit strength reduction of memory addresses.
3927: (On some machines, such strength reduction seems to do harm rather
3928: than good.)
3929:
3930: @findex MOVE_RATIO
3931: @item MOVE_RATIO
3932: The number of scalar move insns which should be generated instead of a
3933: string move insn or a library call. Increasing the value will always
3934: make code faster, but eventually incurs high cost in increased code size.
3935:
3936: If you don't define this, a reasonable default is used.
3937:
3938: @findex NO_FUNCTION_CSE
3939: @item NO_FUNCTION_CSE
3940: Define this macro if it is as good or better to call a constant
3941: function address than to call an address kept in a register.
3942:
3943: @findex NO_RECURSIVE_FUNCTION_CSE
3944: @item NO_RECURSIVE_FUNCTION_CSE
3945: Define this macro if it is as good or better for a function to call
3946: itself with an explicit address than to call an address kept in a
3947: register.
3948:
3949: @findex ADJUST_COST
3950: @item ADJUST_COST (@var{insn}, @var{link}, @var{dep_insn}, @var{cost})
3951: A C statement (sans semicolon) to update the integer variable @var{cost}
3952: based on the relationship between @var{insn} that is dependent on
3953: @var{dep_insn} through the dependence @var{link}. The default is to
3954: make no adjustment to @var{cost}. This can be used for example to
3955: specify to the scheduler that an output- or anti-dependence does not
3956: incur the same cost as a data-dependence.
3957: @end table
3958:
3959: @node Sections
3960: @section Dividing the Output into Sections (Texts, Data, @dots{})
3961: @c the above section title is WAY too long. maybe cut the part between
3962: @c the (...)? --mew 10feb93
3963:
3964: An object file is divided into sections containing different types of
3965: data. In the most common case, there are three sections: the @dfn{text
3966: section}, which holds instructions and read-only data; the @dfn{data
3967: section}, which holds initialized writable data; and the @dfn{bss
3968: section}, which holds uninitialized data. Some systems have other kinds
3969: of sections.
3970:
3971: The compiler must tell the assembler when to switch sections. These
3972: macros control what commands to output to tell the assembler this. You
3973: can also define additional sections.
3974:
3975: @table @code
3976: @findex TEXT_SECTION_ASM_OP
3977: @item TEXT_SECTION_ASM_OP
3978: A C expression whose value is a string containing the assembler
3979: operation that should precede instructions and read-only data. Normally
3980: @code{".text"} is right.
3981:
3982: @findex DATA_SECTION_ASM_OP
3983: @item DATA_SECTION_ASM_OP
3984: A C expression whose value is a string containing the assembler
3985: operation to identify the following data as writable initialized data.
3986: Normally @code{".data"} is right.
3987:
3988: @findex SHARED_SECTION_ASM_OP
3989: @item SHARED_SECTION_ASM_OP
3990: if defined, a C expression whose value is a string containing the
3991: assembler operation to identify the following data as shared data. If
3992: not defined, @code{DATA_SECTION_ASM_OP} will be used.
3993:
3994: @findex INIT_SECTION_ASM_OP
3995: @item INIT_SECTION_ASM_OP
3996: if defined, a C expression whose value is a string containing the
3997: assembler operation to identify the following data as initialization
3998: code. If not defined, GNU CC will assume such a section does not
3999: exist.
4000:
4001: @findex EXTRA_SECTIONS
4002: @findex in_text
4003: @findex in_data
4004: @item EXTRA_SECTIONS
4005: A list of names for sections other than the standard two, which are
4006: @code{in_text} and @code{in_data}. You need not define this macro
4007: on a system with no other sections (that GCC needs to use).
4008:
4009: @findex EXTRA_SECTION_FUNCTIONS
4010: @findex text_section
4011: @findex data_section
4012: @item EXTRA_SECTION_FUNCTIONS
4013: One or more functions to be defined in @file{varasm.c}. These
4014: functions should do jobs analogous to those of @code{text_section} and
4015: @code{data_section}, for your additional sections. Do not define this
4016: macro if you do not define @code{EXTRA_SECTIONS}.
4017:
4018: @findex READONLY_DATA_SECTION
4019: @item READONLY_DATA_SECTION
4020: On most machines, read-only variables, constants, and jump tables are
4021: placed in the text section. If this is not the case on your machine,
4022: this macro should be defined to be the name of a function (either
4023: @code{data_section} or a function defined in @code{EXTRA_SECTIONS}) that
4024: switches to the section to be used for read-only items.
4025:
4026: If these items should be placed in the text section, this macro should
4027: not be defined.
4028:
4029: @findex SELECT_SECTION
4030: @item SELECT_SECTION (@var{exp}, @var{reloc})
4031: A C statement or statements to switch to the appropriate section for
4032: output of @var{exp}. You can assume that @var{exp} is either a
4033: @code{VAR_DECL} node or a constant of some sort. @var{reloc}
4034: indicates whether the initial value of @var{exp} requires link-time
4035: relocations. Select the section by calling @code{text_section} or one
4036: of the alternatives for other sections.
4037:
4038: Do not define this macro if you put all read-only variables and
4039: constants in the read-only data section (usually the text section).
4040:
4041: @findex SELECT_RTX_SECTION
4042: @item SELECT_RTX_SECTION (@var{mode}, @var{rtx})
4043: A C statement or statements to switch to the appropriate section for
4044: output of @var{rtx} in mode @var{mode}. You can assume that @var{rtx}
4045: is some kind of constant in RTL. The argument @var{mode} is redundant
4046: except in the case of a @code{const_int} rtx. Select the section by
4047: calling @code{text_section} or one of the alternatives for other
4048: sections.
4049:
4050: Do not define this macro if you put all constants in the read-only
4051: data section.
4052:
4053: @findex JUMP_TABLES_IN_TEXT_SECTION
4054: @item JUMP_TABLES_IN_TEXT_SECTION
4055: Define this macro if jump tables (for @code{tablejump} insns) should be
4056: output in the text section, along with the assembler instructions.
4057: Otherwise, the readonly data section is used.
4058:
4059: This macro is irrelevant if there is no separate readonly data section.
4060:
4061: @findex ENCODE_SECTION_INFO
4062: @item ENCODE_SECTION_INFO (@var{decl})
4063: Define this macro if references to a symbol must be treated differently
4064: depending on something about the variable or function named by the
4065: symbol (such as what section it is in).
4066:
4067: The macro definition, if any, is executed immediately after the rtl for
4068: @var{decl} has been created and stored in @code{DECL_RTL (@var{decl})}.
4069: The value of the rtl will be a @code{mem} whose address is a
4070: @code{symbol_ref}.
4071:
4072: @cindex @code{SYMBOL_REF_FLAG}, in @code{ENCODE_SECTION_INFO}
4073: The usual thing for this macro to do is to record a flag in the
4074: @code{symbol_ref} (such as @code{SYMBOL_REF_FLAG}) or to store a
4075: modified name string in the @code{symbol_ref} (if one bit is not enough
4076: information).
4077:
4078: @findex STRIP_NAME_ENCODING
4079: @item STRIP_NAME_ENCODING (@var{var}, @var{sym_name})
4080: Decode @var{sym_name} and store the real name part in @var{var}, sans
4081: the characters that encode section info. Define this macro if
4082: @code{ENCODE_SECTION_INFO} alters the symbol's name string.
4083: @end table
4084:
4085: @node PIC
4086: @section Position Independent Code
4087: @cindex position independent code
4088: @cindex PIC
4089:
4090: This section describes macros that help implement generation of position
4091: independent code. Simply defining these macros is not enough to
4092: generate valid PIC; you must also add support to the macros
4093: @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as
4094: well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of
4095: @samp{movsi} to do something appropriate when the source operand
4096: contains a symbolic address. You may also need to alter the handling of
4097: switch statements so that they use relative addresses.
4098: @c i rearranged the order of the macros above to try to force one of
4099: @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
4100:
4101: @table @code
4102: @findex PIC_OFFSET_TABLE_REGNUM
4103: @item PIC_OFFSET_TABLE_REGNUM
4104: The register number of the register used to address a table of static
4105: data addresses in memory. In some cases this register is defined by a
4106: processor's ``application binary interface'' (ABI). When this macro
4107: is defined, RTL is generated for this register once, as with the stack
4108: pointer and frame pointer registers. If this macro is not defined, it
4109: is up to the machine-dependent files to allocate such a register (if
4110: necessary).
4111:
4112: @findex FINALIZE_PIC
4113: @item FINALIZE_PIC
4114: By generating position-independent code, when two different programs (A
4115: and B) share a common library (libC.a), the text of the library can be
4116: shared whether or not the library is linked at the same address for both
4117: programs. In some of these environments, position-independent code
4118: requires not only the use of different addressing modes, but also
4119: special code to enable the use of these addressing modes.
4120:
4121: The @code{FINALIZE_PIC} macro serves as a hook to emit these special
4122: codes once the function is being compiled into assembly code, but not
4123: before. (It is not done before, because in the case of compiling an
4124: inline function, it would lead to multiple PIC prologues being
4125: included in functions which used inline functions and were compiled to
4126: assembly language.)
4127:
4128: @findex LEGITIMATE_PIC_OPERAND_P
4129: @item LEGITIMATE_PIC_OPERAND_P (@var{x})
4130: A C expression that is nonzero if @var{x} is a legitimate immediate
4131: operand on the target machine when generating position independent code.
4132: You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
4133: check this. You can also assume @var{flag_pic} is true, so you need not
4134: check it either. You need not define this macro if all constants
4135: (including @code{SYMBOL_REF}) can be immediate operands when generating
4136: position independent code.
4137: @end table
4138:
4139: @node Assembler Format
4140: @section Defining the Output Assembler Language
4141:
4142: This section describes macros whose principal purpose is to describe how
4143: to write instructions in assembler language--rather than what the
4144: instructions do.
4145:
4146: @menu
4147: * File Framework:: Structural information for the assembler file.
4148: * Data Output:: Output of constants (numbers, strings, addresses).
4149: * Uninitialized Data:: Output of uninitialized variables.
4150: * Label Output:: Output and generation of labels.
4151: * Initialization:: General principles of initialization
4152: and termination routines.
4153: * Macros for Initialization::
4154: Specific macros that control the handling of
4155: initialization and termination routines.
4156: * Instruction Output:: Output of actual instructions.
4157: * Dispatch Tables:: Output of jump tables.
4158: * Alignment Output:: Pseudo ops for alignment and skipping data.
4159: @end menu
4160:
4161: @node File Framework
4162: @subsection The Overall Framework of an Assembler File
4163: @cindex assembler format
4164: @cindex output of assembler code
4165:
4166: @table @code
4167: @findex ASM_FILE_START
4168: @item ASM_FILE_START (@var{stream})
4169: A C expression which outputs to the stdio stream @var{stream}
4170: some appropriate text to go at the start of an assembler file.
4171:
4172: Normally this macro is defined to output a line containing
4173: @samp{#NO_APP}, which is a comment that has no effect on most
4174: assemblers but tells the GNU assembler that it can save time by not
4175: checking for certain assembler constructs.
4176:
4177: On systems that use SDB, it is necessary to output certain commands;
4178: see @file{attasm.h}.
4179:
4180: @findex ASM_FILE_END
4181: @item ASM_FILE_END (@var{stream})
4182: A C expression which outputs to the stdio stream @var{stream}
4183: some appropriate text to go at the end of an assembler file.
4184:
4185: If this macro is not defined, the default is to output nothing
4186: special at the end of the file. Most systems don't require any
4187: definition.
4188:
4189: On systems that use SDB, it is necessary to output certain commands;
4190: see @file{attasm.h}.
4191:
4192: @findex ASM_IDENTIFY_GCC
4193: @item ASM_IDENTIFY_GCC (@var{file})
4194: A C statement to output assembler commands which will identify
4195: the object file as having been compiled with GNU CC (or another
4196: GNU compiler).
4197:
4198: If you don't define this macro, the string @samp{gcc_compiled.:}
4199: is output. This string is calculated to define a symbol which,
4200: on BSD systems, will never be defined for any other reason.
4201: GDB checks for the presence of this symbol when reading the
4202: symbol table of an executable.
4203:
4204: On non-BSD systems, you must arrange communication with GDB in
4205: some other fashion. If GDB is not used on your system, you can
4206: define this macro with an empty body.
4207:
4208: @findex ASM_COMMENT_START
4209: @item ASM_COMMENT_START
4210: A C string constant describing how to begin a comment in the target
4211: assembler language. The compiler assumes that the comment will end at
4212: the end of the line.
4213:
4214: @findex ASM_APP_ON
4215: @item ASM_APP_ON
4216: A C string constant for text to be output before each @code{asm}
4217: statement or group of consecutive ones. Normally this is
4218: @code{"#APP"}, which is a comment that has no effect on most
4219: assemblers but tells the GNU assembler that it must check the lines
4220: that follow for all valid assembler constructs.
4221:
4222: @findex ASM_APP_OFF
4223: @item ASM_APP_OFF
4224: A C string constant for text to be output after each @code{asm}
4225: statement or group of consecutive ones. Normally this is
4226: @code{"#NO_APP"}, which tells the GNU assembler to resume making the
4227: time-saving assumptions that are valid for ordinary compiler output.
4228:
4229: @findex ASM_OUTPUT_SOURCE_FILENAME
4230: @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
4231: A C statement to output COFF information or DWARF debugging information
4232: which indicates that filename @var{name} is the current source file to
4233: the stdio stream @var{stream}.
4234:
4235: This macro need not be defined if the standard form of output
4236: for the file format in use is appropriate.
4237:
4238: @findex ASM_OUTPUT_SOURCE_LINE
4239: @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
4240: A C statement to output DBX or SDB debugging information before code
4241: for line number @var{line} of the current source file to the
4242: stdio stream @var{stream}.
4243:
4244: This macro need not be defined if the standard form of debugging
4245: information for the debugger in use is appropriate.
4246:
4247: @findex ASM_OUTPUT_IDENT
4248: @item ASM_OUTPUT_IDENT (@var{stream}, @var{string})
4249: A C statement to output something to the assembler file to handle a
4250: @samp{#ident} directive containing the text @var{string}. If this
4251: macro is not defined, nothing is output for a @samp{#ident} directive.
4252:
4253: @findex OBJC_PROLOGUE
4254: @item OBJC_PROLOGUE
4255: A C statement to output any assembler statements which are required to
4256: precede any Objective C object definitions or message sending. The
4257: statement is executed only when compiling an Objective C program.
4258: @end table
4259:
4260: @need 2000
4261: @node Data Output
4262: @subsection Output of Data
4263:
4264: @table @code
4265: @findex ASM_OUTPUT_LONG_DOUBLE
4266: @findex ASM_OUTPUT_DOUBLE
4267: @findex ASM_OUTPUT_FLOAT
4268: @item ASM_OUTPUT_LONG_DOUBLE (@var{stream}, @var{value})
4269: @itemx ASM_OUTPUT_DOUBLE (@var{stream}, @var{value})
4270: @itemx ASM_OUTPUT_FLOAT (@var{stream}, @var{value})
4271: A C statement to output to the stdio stream @var{stream} an assembler
4272: instruction to assemble a floating-point constant of @code{TFmode},
4273: @code{DFmode} or @code{SFmode}, respectively, whose value is
4274: @var{value}. @var{value} will be a C expression of type
4275: @code{REAL_VALUE_TYPE}. Macros such as
4276: @code{REAL_VALUE_TO_TARGET_DOUBLE} are useful for writing these
4277: definitions.
4278:
4279: @findex ASM_OUTPUT_QUADRUPLE_INT
4280: @findex ASM_OUTPUT_DOUBLE_INT
4281: @findex ASM_OUTPUT_INT
4282: @findex ASM_OUTPUT_SHORT
4283: @findex ASM_OUTPUT_CHAR
4284: @findex output_addr_const
4285: @item ASM_OUTPUT_QUADRUPLE_INT (@var{stream}, @var{exp})
4286: @itemx ASM_OUTPUT_DOUBLE_INT (@var{stream}, @var{exp})
4287: @itemx ASM_OUTPUT_INT (@var{stream}, @var{exp})
4288: @itemx ASM_OUTPUT_SHORT (@var{stream}, @var{exp})
4289: @itemx ASM_OUTPUT_CHAR (@var{stream}, @var{exp})
4290: A C statement to output to the stdio stream @var{stream} an assembler
4291: instruction to assemble an integer of 16, 8, 4, 2 or 1 bytes,
4292: respectively, whose value is @var{value}. The argument @var{exp} will
4293: be an RTL expression which represents a constant value. Use
4294: @samp{output_addr_const (@var{stream}, @var{exp})} to output this value
4295: as an assembler expression.@refill
4296:
4297: For sizes larger than @code{UNITS_PER_WORD}, if the action of a macro
4298: would be identical to repeatedly calling the macro corresponding to
4299: a size of @code{UNITS_PER_WORD}, once for each word, you need not define
4300: the macro.
4301:
4302: @findex ASM_OUTPUT_BYTE
4303: @item ASM_OUTPUT_BYTE (@var{stream}, @var{value})
4304: A C statement to output to the stdio stream @var{stream} an assembler
4305: instruction to assemble a single byte containing the number @var{value}.
4306:
4307: @findex ASM_BYTE_OP
4308: @item ASM_BYTE_OP
4309: A C string constant giving the pseudo-op to use for a sequence of
4310: single-byte constants. If this macro is not defined, the default is
4311: @code{"byte"}.
4312:
4313: @findex ASM_OUTPUT_ASCII
4314: @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
4315: A C statement to output to the stdio stream @var{stream} an assembler
4316: instruction to assemble a string constant containing the @var{len}
4317: bytes at @var{ptr}. @var{ptr} will be a C expression of type
4318: @code{char *} and @var{len} a C expression of type @code{int}.
4319:
4320: If the assembler has a @code{.ascii} pseudo-op as found in the
4321: Berkeley Unix assembler, do not define the macro
4322: @code{ASM_OUTPUT_ASCII}.
4323:
4324: @findex ASM_OUTPUT_POOL_PROLOGUE
4325: @item ASM_OUTPUT_POOL_PROLOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
4326: A C statement to output assembler commands to define the start of the
4327: constant pool for a function. @var{funname} is a string giving
4328: the name of the function. Should the return type of the function
4329: be required, it can be obtained via @var{fundecl}. @var{size}
4330: is the size, in bytes, of the constant pool that will be written
4331: immediately after this call.
4332:
4333: If no constant-pool prefix is required, the usual case, this macro need
4334: not be defined.
4335:
4336: @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY
4337: @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
4338: A C statement (with or without semicolon) to output a constant in the
4339: constant pool, if it needs special treatment. (This macro need not do
4340: anything for RTL expressions that can be output normally.)
4341:
4342: The argument @var{file} is the standard I/O stream to output the
4343: assembler code on. @var{x} is the RTL expression for the constant to
4344: output, and @var{mode} is the machine mode (in case @var{x} is a
4345: @samp{const_int}). @var{align} is the required alignment for the value
4346: @var{x}; you should output an assembler directive to force this much
4347: alignment.
4348:
4349: The argument @var{labelno} is a number to use in an internal label for
4350: the address of this pool entry. The definition of this macro is
4351: responsible for outputting the label definition at the proper place.
4352: Here is how to do this:
4353:
4354: @example
4355: ASM_OUTPUT_INTERNAL_LABEL (@var{file}, "LC", @var{labelno});
4356: @end example
4357:
4358: When you output a pool entry specially, you should end with a
4359: @code{goto} to the label @var{jumpto}. This will prevent the same pool
4360: entry from being output a second time in the usual manner.
4361:
4362: You need not define this macro if it would do nothing.
4363:
4364: @findex ASM_OPEN_PAREN
4365: @findex ASM_CLOSE_PAREN
4366: @item ASM_OPEN_PAREN
4367: @itemx ASM_CLOSE_PAREN
4368: These macros are defined as C string constant, describing the syntax
4369: in the assembler for grouping arithmetic expressions. The following
4370: definitions are correct for most assemblers:
4371:
4372: @example
4373: #define ASM_OPEN_PAREN "("
4374: #define ASM_CLOSE_PAREN ")"
4375: @end example
4376: @end table
4377:
4378: These macros are provided by @file{real.h} for writing the definitions
4379: of @code{ASM_OUTPUT_DOUBLE} and the like:
4380:
4381: @table @code
4382: @item REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
4383: @itemx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
4384: @itemx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
4385: @findex REAL_VALUE_TO_TARGET_SINGLE
4386: @findex REAL_VALUE_TO_TARGET_DOUBLE
4387: @findex REAL_VALUE_TO_TARGET_LONG_DOUBLE
4388: These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's
4389: floating point representation, and store its bit pattern in the array of
4390: @code{long int} whose address is @var{l}. The number of elements in the
4391: output array is determined by the size of the desired target floating
4392: point data type: 32 bits of it go in each @code{long int} array
4393: element. Each array element holds 32 bits of the result, even if
4394: @code{long int} is wider than 32 bits on the host machine.
4395:
4396: The array element values are designed so that you can print them out
4397: using @code{fprintf} in the order they should appear in the target
4398: machine's memory.
4399:
4400: @item REAL_VALUE_TO_DECIMAL (@var{x}, @var{format}, @var{string})
4401: @findex REAL_VALUE_TO_DECIMAL
4402: This macro converts @var{x}, of type @code{REAL_VALUE_TYPE}, to a
4403: decimal number and stores it as a string into @var{string}.
4404: You must pass, as @var{string}, the address of a long enough block
4405: of space to hold the result.
4406:
4407: The argument @var{format} is a @code{printf}-specification that serves
4408: as a suggestion for how to format the output string.
4409: @end table
4410:
4411: @node Uninitialized Data
4412: @subsection Output of Uninitialized Variables
4413:
4414: Each of the macros in this section is used to do the whole job of
4415: outputting a single uninitialized variable.
4416:
4417: @table @code
4418: @findex ASM_OUTPUT_COMMON
4419: @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
4420: A C statement (sans semicolon) to output to the stdio stream
4421: @var{stream} the assembler definition of a common-label named
4422: @var{name} whose size is @var{size} bytes. The variable @var{rounded}
4423: is the size rounded up to whatever alignment the caller wants.
4424:
4425: Use the expression @code{assemble_name (@var{stream}, @var{name})} to
4426: output the name itself; before and after that, output the additional
4427: assembler syntax for defining the name, and a newline.
4428:
4429: This macro controls how the assembler definitions of uninitialized
4430: global variables are output.
4431:
4432: @findex ASM_OUTPUT_ALIGNED_COMMON
4433: @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
4434: Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
4435: separate, explicit argument. If you define this macro, it is used in
4436: place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
4437: handling the required alignment of the variable.
4438:
4439: @findex ASM_OUTPUT_SHARED_COMMON
4440: @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
4441: If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it
4442: is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON}
4443: will be used.
4444:
4445: @findex ASM_OUTPUT_LOCAL
4446: @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
4447: A C statement (sans semicolon) to output to the stdio stream
4448: @var{stream} the assembler definition of a local-common-label named
4449: @var{name} whose size is @var{size} bytes. The variable @var{rounded}
4450: is the size rounded up to whatever alignment the caller wants.
4451:
4452: Use the expression @code{assemble_name (@var{stream}, @var{name})} to
4453: output the name itself; before and after that, output the additional
4454: assembler syntax for defining the name, and a newline.
4455:
4456: This macro controls how the assembler definitions of uninitialized
4457: static variables are output.
4458:
4459: @findex ASM_OUTPUT_ALIGNED_LOCAL
4460: @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
4461: Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
4462: separate, explicit argument. If you define this macro, it is used in
4463: place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
4464: handling the required alignment of the variable.
4465:
4466: @findex ASM_OUTPUT_SHARED_LOCAL
4467: @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
4468: If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it
4469: is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL}
4470: will be used.
4471: @end table
4472:
4473: @node Label Output
4474: @subsection Output and Generation of Labels
4475:
4476: @table @code
4477: @findex ASM_OUTPUT_LABEL
4478: @findex assemble_name
4479: @item ASM_OUTPUT_LABEL (@var{stream}, @var{name})
4480: A C statement (sans semicolon) to output to the stdio stream
4481: @var{stream} the assembler definition of a label named @var{name}.
4482: Use the expression @code{assemble_name (@var{stream}, @var{name})} to
4483: output the name itself; before and after that, output the additional
4484: assembler syntax for defining the name, and a newline.
4485:
4486: @findex ASM_DECLARE_FUNCTION_NAME
4487: @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
4488: A C statement (sans semicolon) to output to the stdio stream
4489: @var{stream} any text necessary for declaring the name @var{name} of a
4490: function which is being defined. This macro is responsible for
4491: outputting the label definition (perhaps using
4492: @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
4493: @code{FUNCTION_DECL} tree node representing the function.
4494:
4495: If this macro is not defined, then the function name is defined in the
4496: usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
4497:
4498: @findex ASM_DECLARE_FUNCTION_SIZE
4499: @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
4500: A C statement (sans semicolon) to output to the stdio stream
4501: @var{stream} any text necessary for declaring the size of a function
4502: which is being defined. The argument @var{name} is the name of the
4503: function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
4504: representing the function.
4505:
4506: If this macro is not defined, then the function size is not defined.
4507:
4508: @findex ASM_DECLARE_OBJECT_NAME
4509: @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
4510: A C statement (sans semicolon) to output to the stdio stream
4511: @var{stream} any text necessary for declaring the name @var{name} of an
4512: initialized variable which is being defined. This macro must output the
4513: label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
4514: @var{decl} is the @code{VAR_DECL} tree node representing the variable.
4515:
4516: If this macro is not defined, then the variable name is defined in the
4517: usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
4518:
4519: @findex ASM_FINISH_DECLARE_OBJECT
4520: @item ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
4521: A C statement (sans semicolon) to finish up declaring a variable name
4522: once the compiler has processed its initializer fully and thus has had a
4523: chance to determine the size of an array when controlled by an
4524: initializer. This is used on systems where it's necessary to declare
4525: something about the size of the object.
4526:
4527: If you don't define this macro, that is equivalent to defining it to do
4528: nothing.
4529:
4530: @findex ASM_GLOBALIZE_LABEL
4531: @item ASM_GLOBALIZE_LABEL (@var{stream}, @var{name})
4532: A C statement (sans semicolon) to output to the stdio stream
4533: @var{stream} some commands that will make the label @var{name} global;
4534: that is, available for reference from other files. Use the expression
4535: @code{assemble_name (@var{stream}, @var{name})} to output the name
4536: itself; before and after that, output the additional assembler syntax
4537: for making that name global, and a newline.
4538:
4539: @findex ASM_OUTPUT_EXTERNAL
4540: @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
4541: A C statement (sans semicolon) to output to the stdio stream
4542: @var{stream} any text necessary for declaring the name of an external
4543: symbol named @var{name} which is referenced in this compilation but
4544: not defined. The value of @var{decl} is the tree node for the
4545: declaration.
4546:
4547: This macro need not be defined if it does not need to output anything.
4548: The GNU assembler and most Unix assemblers don't require anything.
4549:
4550: @findex ASM_OUTPUT_EXTERNAL_LIBCALL
4551: @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref})
4552: A C statement (sans semicolon) to output on @var{stream} an assembler
4553: pseudo-op to declare a library function name external. The name of the
4554: library function is given by @var{symref}, which has type @code{rtx} and
4555: is a @code{symbol_ref}.
4556:
4557: This macro need not be defined if it does not need to output anything.
4558: The GNU assembler and most Unix assemblers don't require anything.
4559:
4560: @findex ASM_OUTPUT_LABELREF
4561: @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
4562: A C statement (sans semicolon) to output to the stdio stream
4563: @var{stream} a reference in assembler syntax to a label named
4564: @var{name}. This should add @samp{_} to the front of the name, if that
4565: is customary on your operating system, as it is in most Berkeley Unix
4566: systems. This macro is used in @code{assemble_name}.
4567:
4568: @ignore @c Seems not to exist anymore.
4569: @findex ASM_OUTPUT_LABELREF_AS_INT
4570: @item ASM_OUTPUT_LABELREF_AS_INT (@var{file}, @var{label})
4571: Define this macro for systems that use the program @code{collect2}.
4572: The definition should be a C statement to output a word containing
4573: a reference to the label @var{label}.
4574: @end ignore
4575:
4576: @findex ASM_OUTPUT_INTERNAL_LABEL
4577: @item ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{prefix}, @var{num})
4578: A C statement to output to the stdio stream @var{stream} a label whose
4579: name is made from the string @var{prefix} and the number @var{num}.
4580:
4581: It is absolutely essential that these labels be distinct from the labels
4582: used for user-level functions and variables. Otherwise, certain programs
4583: will have name conflicts with internal labels.
4584:
4585: It is desirable to exclude internal labels from the symbol table of the
4586: object file. Most assemblers have a naming convention for labels that
4587: should be excluded; on many systems, the letter @samp{L} at the
4588: beginning of a label has this effect. You should find out what
4589: convention your system uses, and follow it.
4590:
4591: The usual definition of this macro is as follows:
4592:
4593: @example
4594: fprintf (@var{stream}, "L%s%d:\n", @var{prefix}, @var{num})
4595: @end example
4596:
4597: @findex ASM_GENERATE_INTERNAL_LABEL
4598: @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
4599: A C statement to store into the string @var{string} a label whose name
4600: is made from the string @var{prefix} and the number @var{num}.
4601:
4602: This string, when output subsequently by @code{assemble_name}, should
4603: produce the output that @code{ASM_OUTPUT_INTERNAL_LABEL} would produce
4604: with the same @var{prefix} and @var{num}.
4605:
4606: If the string begins with @samp{*}, then @code{assemble_name} will
4607: output the rest of the string unchanged. It is often convenient for
4608: @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
4609: string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
4610: to output the string, and may change it. (Of course,
4611: @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
4612: you should know what it does on your machine.)
4613:
4614: @findex ASM_FORMAT_PRIVATE_NAME
4615: @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
4616: A C expression to assign to @var{outvar} (which is a variable of type
4617: @code{char *}) a newly allocated string made from the string
4618: @var{name} and the number @var{number}, with some suitable punctuation
4619: added. Use @code{alloca} to get space for the string.
4620:
4621: The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
4622: produce an assembler label for an internal static variable whose name is
4623: @var{name}. Therefore, the string must be such as to result in valid
4624: assembler code. The argument @var{number} is different each time this
4625: macro is executed; it prevents conflicts between similarly-named
4626: internal static variables in different scopes.
4627:
4628: Ideally this string should not be a valid C identifier, to prevent any
4629: conflict with the user's own symbols. Most assemblers allow periods
4630: or percent signs in assembler symbols; putting at least one of these
4631: between the name and the number will suffice.
4632:
4633: @findex OBJC_GEN_METHOD_LABEL
4634: @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
4635: Define this macro to override the default assembler names used for
4636: Objective C methods.
4637:
4638: The default name is a unique method number followed by the name of the
4639: class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
4640: the category is also included in the assembler name (e.g.@:
4641: @samp{_1_Foo_Bar}).
4642:
4643: These names are safe on most systems, but make debugging difficult since
4644: the method's selector is not present in the name. Therefore, particular
4645: systems define other ways of computing names.
4646:
4647: @var{buf} is an expression of type @code{char *} which gives you a
4648: buffer in which to store the name; its length is as long as
4649: @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
4650: 50 characters extra.
4651:
4652: The argument @var{is_inst} specifies whether the method is an instance
4653: method or a class method; @var{class_name} is the name of the class;
4654: @var{cat_name} is the name of the category (or NULL if the method is not
4655: in a category); and @var{sel_name} is the name of the selector.
4656:
4657: On systems where the assembler can handle quoted names, you can use this
4658: macro to provide more human-readable names.
4659: @end table
4660:
4661: @node Initialization
4662: @subsection How Initialization Functions Are Handled
4663: @cindex initialization routines
4664: @cindex termination routines
4665: @cindex constructors, output of
4666: @cindex destructors, output of
4667:
4668: The compiled code for certain languages includes @dfn{constructors}
4669: (also called @dfn{initialization routines})---functions to initialize
4670: data in the program when the program is started. These functions need
4671: to be called before the program is ``started''---that is to say, before
4672: @code{main} is called.
4673:
4674: Compiling some languages generates @dfn{destructors} (also called
4675: @dfn{termination routines}) that should be called when the program
4676: terminates.
4677:
4678: To make the initialization and termination functions work, the compiler
4679: must output something in the assembler code to cause those functions to
4680: be called at the appropriate time. When you port the compiler to a new
4681: system, you need to specify how to do this.
4682:
4683: There are two major ways that GCC currently supports the execution of
4684: initialization and termination functions. Each way has two variants.
4685: Much of the structure is common to all four variations.
4686:
4687: @findex __CTOR_LIST__
4688: @findex __DTOR_LIST__
4689: The linker must build two lists of these functions---a list of
4690: initialization functions, called @code{__CTOR_LIST__}, and a list of
4691: termination functions, called @code{__DTOR_LIST__}.
4692:
4693: Each list always begins with an ignored function pointer (which may hold
4694: 0, @minus{}1, or a count of the function pointers after it, depending on
4695: the environment). This is followed by a series of zero or more function
4696: pointers to constructors (or destructors), followed by a function
4697: pointer containing zero.
4698:
4699: Depending on the operating system and its executable file format, either
4700: @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
4701: time and exit time. Constructors are called in forward order of the
4702: list; destructors in reverse order.
4703:
4704: The best way to handle static constructors works only for object file
4705: formats which provide arbitrarily-named sections. A section is set
4706: aside for a list of constructors, and another for a list of destructors.
4707: Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
4708: object file that defines an initialization function also puts a word in
4709: the constructor section to point to that function. The linker
4710: accumulates all these words into one contiguous @samp{.ctors} section.
4711: Termination functions are handled similarly.
4712:
4713: To use this method, you need appropriate definitions of the macros
4714: @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR}. Usually
4715: you can get them by including @file{svr4.h}.
4716:
4717: When arbitrary sections are available, there are two variants, depending
4718: upon how the code in @file{crtstuff.c} is called. On systems that
4719: support an @dfn{init} section which is executed at program startup,
4720: parts of @file{crtstuff.c} are compiled into that section. The
4721: program is linked by the @code{gcc} driver like this:
4722:
4723: @example
4724: ld -o @var{output_file} crtbegin.o @dots{} crtend.o -lgcc
4725: @end example
4726:
4727: The head of a function (@code{__do_global_ctors}) appears in the init
4728: section of @file{crtbegin.o}; the remainder of the function appears in
4729: the init section of @file{crtend.o}. The linker will pull these two
4730: parts of the section together, making a whole function. If any of the
4731: user's object files linked into the middle of it contribute code, then that
4732: code will be executed as part of the body of @code{__do_global_ctors}.
4733:
4734: To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
4735: macro properly.
4736:
4737: If no init section is available, do not define
4738: @code{INIT_SECTION_ASM_OP}. Then @code{__do_global_ctors} is built into
4739: the text section like all other functions, and resides in
4740: @file{libgcc.a}. When GCC compiles any function called @code{main}, it
4741: inserts a procedure call to @code{__main} as the first executable code
4742: after the function prologue. The @code{__main} function, also defined
4743: in @file{libgcc2.c}, simply calls @file{__do_global_ctors}.
4744:
4745: In file formats that don't support arbitrary sections, there are again
4746: two variants. In the simplest variant, the GNU linker (GNU @code{ld})
4747: and an `a.out' format must be used. In this case,
4748: @code{ASM_OUTPUT_CONSTRUCTOR} is defined to produce a @code{.stabs}
4749: entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
4750: and with the address of the void function containing the initialization
4751: code as its value. The GNU linker recognizes this as a request to add
4752: the value to a ``set''; the values are accumulated, and are eventually
4753: placed in the executable as a vector in the format described above, with
4754: a leading (ignored) count and a trailing zero element.
4755: @code{ASM_OUTPUT_DESTRUCTOR} is handled similarly. Since no init
4756: section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
4757: the compilation of @code{main} to call @code{__main} as above, starting
4758: the initialization process.
4759:
4760: The last variant uses neither arbitrary sections nor the GNU linker.
4761: This is preferable when you want to do dynamic linking and when using
4762: file formats which the GNU linker does not support, such as `ECOFF'. In
4763: this case, @code{ASM_OUTPUT_CONSTRUCTOR} does not produce an
4764: @code{N_SETT} symbol; initialization and termination functions are
4765: recognized simply by their names. This requires an extra program in the
4766: linkage step, called @code{collect2}. This program pretends to be the
4767: linker, for use with GNU CC; it does its job by running the ordinary
4768: linker, but also arranges to include the vectors of initialization and
4769: termination functions. These functions are called via @code{__main} as
4770: described above.
4771:
4772: Choosing among these configuration options has been simplified by a set
4773: of operating-system-dependent files in the @file{config} subdirectory.
4774: These files define all of the relevant parameters. Usually it is
4775: sufficient to include one into your specific machine-dependent
4776: configuration file. These files are:
4777:
4778: @table @file
4779: @item aoutos.h
4780: For operating systems using the `a.out' format.
4781:
4782: @item next.h
4783: For operating systems using the `MachO' format.
4784:
4785: @item svr3.h
4786: For System V Release 3 and similar systems using `COFF' format.
4787:
4788: @item svr4.h
4789: For System V Release 4 and similar systems using `ELF' format.
4790:
4791: @item vms.h
4792: For the VMS operating system.
4793: @end table
4794:
4795: @ifinfo
4796: The following section describes the specific macros that control and
4797: customize the handling of initialization and termination functions.
4798: @end ifinfo
4799:
4800: @node Macros for Initialization
4801: @subsection Macros Controlling Initialization Routines
4802:
4803: Here are the macros that control how the compiler handles initialization
4804: and termination functions:
4805:
4806: @table @code
4807: @findex INIT_SECTION_ASM_OP
4808: @item INIT_SECTION_ASM_OP
4809: If defined, a C string constant for the assembler operation to identify
4810: the following data as initialization code. If not defined, GNU CC will
4811: assume such a section does not exist. When you are using special
4812: sections for initialization and termination functions, this macro also
4813: controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to run the
4814: initialization functions.
4815:
4816: @item ASM_OUTPUT_CONSTRUCTOR (@var{stream}, @var{name})
4817: @findex ASM_OUTPUT_CONSTRUCTOR
4818: Define this macro as a C statement to output on the stream @var{stream}
4819: the assembler code to arrange to call the function named @var{name} at
4820: initialization time.
4821:
4822: Assume that @var{name} is the name of a C function generated
4823: automatically by the compiler. This function takes no arguments. Use
4824: the function @code{assemble_name} to output the name @var{name}; this
4825: performs any system-specific syntactic transformations such as adding an
4826: underscore.
4827:
4828: If you don't define this macro, nothing special is output to arrange to
4829: call the function. This is correct when the function will be called in
4830: some other manner---for example, by means of the @code{collect2} program,
4831: which looks through the symbol table to find these functions by their
4832: names.
4833:
4834: @item ASM_OUTPUT_DESTRUCTOR (@var{stream}, @var{name})
4835: @findex ASM_OUTPUT_DESTRUCTOR
4836: This is like @code{ASM_OUTPUT_CONSTRUCTOR} but used for termination
4837: functions rather than initialization functions.
4838: @end table
4839:
4840: If your system uses @code{collect2} as the means of processing
4841: constructors, then that program normally uses @code{nm} to scan an
4842: object file for constructor functions to be called. On certain kinds of
4843: systems, you can define these macros to make @code{collect2} work faster
4844: (and, in some cases, make it work at all):
4845:
4846: @table @code
4847: @findex OBJECT_FORMAT_COFF
4848: @item OBJECT_FORMAT_COFF
4849: Define this macro if the system uses COFF (Common Object File Format)
4850: object files, so that @code{collect2} can assume this format and scan
4851: object files directly for dynamic constructor/destructor functions.
4852:
4853: @findex OBJECT_FORMAT_ROSE
4854: @item OBJECT_FORMAT_ROSE
4855: Define this macro if the system uses ROSE format object files, so that
4856: @code{collect2} can assume this format and scan object files directly
4857: for dynamic constructor/destructor functions.
4858: @end table
4859:
4860: These macros are effective only in a native compiler; @code{collect2} as
4861: part of a cross compiler always uses @code{nm}.
4862:
4863: @table @code
4864: @findex REAL_NM_FILE_NAME
4865: @item REAL_NM_FILE_NAME
4866: Define this macro as a C string constant containing the file name to use
4867: to execute @code{nm}. The default is to search the path normally for
4868: @code{nm}.
4869: @end table
4870:
4871: @node Instruction Output
4872: @subsection Output of Assembler Instructions
4873:
4874: @table @code
4875: @findex REGISTER_NAMES
4876: @item REGISTER_NAMES
4877: A C initializer containing the assembler's names for the machine
4878: registers, each one as a C string constant. This is what translates
4879: register numbers in the compiler into assembler language.
4880:
4881: @findex ADDITIONAL_REGISTER_NAMES
4882: @item ADDITIONAL_REGISTER_NAMES
4883: If defined, a C initializer for an array of structures containing a name
4884: and a register number. This macro defines additional names for hard
4885: registers, thus allowing the @code{asm} option in declarations to refer
4886: to registers using alternate names.
4887:
4888: @findex ASM_OUTPUT_OPCODE
4889: @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
4890: Define this macro if you are using an unusual assembler that
4891: requires different names for the machine instructions.
4892:
4893: The definition is a C statement or statements which output an
4894: assembler instruction opcode to the stdio stream @var{stream}. The
4895: macro-operand @var{ptr} is a variable of type @code{char *} which
4896: points to the opcode name in its ``internal'' form---the form that is
4897: written in the machine description. The definition should output the
4898: opcode name to @var{stream}, performing any translation you desire, and
4899: increment the variable @var{ptr} to point at the end of the opcode
4900: so that it will not be output twice.
4901:
4902: In fact, your macro definition may process less than the entire opcode
4903: name, or more than the opcode name; but if you want to process text
4904: that includes @samp{%}-sequences to substitute operands, you must take
4905: care of the substitution yourself. Just be sure to increment
4906: @var{ptr} over whatever text should not be output normally.
4907:
4908: @findex recog_operand
4909: If you need to look at the operand values, they can be found as the
4910: elements of @code{recog_operand}.
4911:
4912: If the macro definition does nothing, the instruction is output
4913: in the usual way.
4914:
4915: @findex FINAL_PRESCAN_INSN
4916: @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
4917: If defined, a C statement to be executed just prior to the output of
4918: assembler code for @var{insn}, to modify the extracted operands so
4919: they will be output differently.
4920:
4921: Here the argument @var{opvec} is the vector containing the operands
4922: extracted from @var{insn}, and @var{noperands} is the number of
4923: elements of the vector which contain meaningful data for this insn.
4924: The contents of this vector are what will be used to convert the insn
4925: template into assembler code, so you can change the assembler output
4926: by changing the contents of the vector.
4927:
4928: This macro is useful when various assembler syntaxes share a single
4929: file of instruction patterns; by defining this macro differently, you
4930: can cause a large class of instructions to be output differently (such
4931: as with rearranged operands). Naturally, variations in assembler
4932: syntax affecting individual insn patterns ought to be handled by
4933: writing conditional output routines in those patterns.
4934:
4935: If this macro is not defined, it is equivalent to a null statement.
4936:
4937: @findex PRINT_OPERAND
4938: @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
4939: A C compound statement to output to stdio stream @var{stream} the
4940: assembler syntax for an instruction operand @var{x}. @var{x} is an
4941: RTL expression.
4942:
4943: @var{code} is a value that can be used to specify one of several ways
4944: of printing the operand. It is used when identical operands must be
4945: printed differently depending on the context. @var{code} comes from
4946: the @samp{%} specification that was used to request printing of the
4947: operand. If the specification was just @samp{%@var{digit}} then
4948: @var{code} is 0; if the specification was @samp{%@var{ltr}
4949: @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
4950:
4951: @findex reg_names
4952: If @var{x} is a register, this macro should print the register's name.
4953: The names can be found in an array @code{reg_names} whose type is
4954: @code{char *[]}. @code{reg_names} is initialized from
4955: @code{REGISTER_NAMES}.
4956:
4957: When the machine description has a specification @samp{%@var{punct}}
4958: (a @samp{%} followed by a punctuation character), this macro is called
4959: with a null pointer for @var{x} and the punctuation character for
4960: @var{code}.
4961:
4962: @findex PRINT_OPERAND_PUNCT_VALID_P
4963: @item PRINT_OPERAND_PUNCT_VALID_P (@var{code})
4964: A C expression which evaluates to true if @var{code} is a valid
4965: punctuation character for use in the @code{PRINT_OPERAND} macro. If
4966: @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
4967: punctuation characters (except for the standard one, @samp{%}) are used
4968: in this way.
4969:
4970: @findex PRINT_OPERAND_ADDRESS
4971: @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
4972: A C compound statement to output to stdio stream @var{stream} the
4973: assembler syntax for an instruction operand that is a memory reference
4974: whose address is @var{x}. @var{x} is an RTL expression.
4975:
4976: @cindex @code{ENCODE_SECTION_INFO} usage
4977: On some machines, the syntax for a symbolic address depends on the
4978: section that the address refers to. On these machines, define the macro
4979: @code{ENCODE_SECTION_INFO} to store the information into the
4980: @code{symbol_ref}, and then check for it here. @xref{Assembler Format}.
4981:
4982: @findex DBR_OUTPUT_SEQEND
4983: @findex dbr_sequence_length
4984: @item DBR_OUTPUT_SEQEND(@var{file})
4985: A C statement, to be executed after all slot-filler instructions have
4986: been output. If necessary, call @code{dbr_sequence_length} to
4987: determine the number of slots filled in a sequence (zero if not
4988: currently outputting a sequence), to decide how many no-ops to output,
4989: or whatever.
4990:
4991: Don't define this macro if it has nothing to do, but it is helpful in
4992: reading assembly output if the extent of the delay sequence is made
4993: explicit (e.g. with white space).
4994:
4995: @findex final_sequence
4996: Note that output routines for instructions with delay slots must be
4997: prepared to deal with not being output as part of a sequence (i.e.
4998: when the scheduling pass is not run, or when no slot fillers could be
4999: found.) The variable @code{final_sequence} is null when not
5000: processing a sequence, otherwise it contains the @code{sequence} rtx
5001: being output.
5002:
5003: @findex REGISTER_PREFIX
5004: @findex LOCAL_LABEL_PREFIX
5005: @findex USER_LABEL_PREFIX
5006: @findex IMMEDIATE_PREFIX
5007: @findex asm_fprintf
5008: @item REGISTER_PREFIX
5009: @itemx LOCAL_LABEL_PREFIX
5010: @itemx USER_LABEL_PREFIX
5011: @itemx IMMEDIATE_PREFIX
5012: If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
5013: @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
5014: @file{final.c}). These are useful when a single @file{md} file must
5015: support multiple assembler formats. In that case, the various @file{tm.h}
5016: files can define these macros differently.
5017:
5018: @findex ASSEMBLER_DIALECT
5019: @item ASSEMBLER_DIALECT
5020: If your target supports multiple dialects of assembler language (such as
5021: different opcodes), define this macro as a C expression that gives the
5022: numeric index of the assembler langauge dialect to use, with zero as the
5023: first variant.
5024:
5025: If this macro is defined, you may use
5026: @samp{@{option0|option1|option2@dots{}@}} constructs in the output
5027: templates of patterns (@pxref{Output Template}) or in the first argument
5028: of @code{asm_fprintf}. This construct outputs @samp{option0},
5029: @samp{option1} or @samp{option2}, etc., if the value of
5030: @code{ASSEMBLER_DIALECT} is zero, one or two, etc. Any special
5031: characters within these strings retain their usual meaning.
5032:
5033: If you do not define this macro, the characters @samp{@{}, @samp{|} and
5034: @samp{@}} do not have any special meaning when used in templates or
5035: operands to @code{asm_fprintf}.
5036:
5037: Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
5038: @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
5039: the variations in assemble language syntax with that mechanism. Define
5040: @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
5041: if the syntax variant are larger and involve such things as different
5042: opcodes or operand order.
5043:
5044: @findex ASM_OUTPUT_REG_PUSH
5045: @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
5046: A C expression to output to @var{stream} some assembler code
5047: which will push hard register number @var{regno} onto the stack.
5048: The code need not be optimal, since this macro is used only when
5049: profiling.
5050:
5051: @findex ASM_OUTPUT_REG_POP
5052: @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
5053: A C expression to output to @var{stream} some assembler code
5054: which will pop hard register number @var{regno} off of the stack.
5055: The code need not be optimal, since this macro is used only when
5056: profiling.
5057: @end table
5058:
5059: @node Dispatch Tables
5060: @subsection Output of Dispatch Tables
5061:
5062: @table @code
5063: @cindex dispatch table
5064: @findex ASM_OUTPUT_ADDR_DIFF_ELT
5065: @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{value}, @var{rel})
5066: This macro should be provided on machines where the addresses
5067: in a dispatch table are relative to the table's own address.
5068:
5069: The definition should be a C statement to output to the stdio stream
5070: @var{stream} an assembler pseudo-instruction to generate a difference
5071: between two labels. @var{value} and @var{rel} are the numbers of two
5072: internal labels. The definitions of these labels are output using
5073: @code{ASM_OUTPUT_INTERNAL_LABEL}, and they must be printed in the same
5074: way here. For example,
5075:
5076: @example
5077: fprintf (@var{stream}, "\t.word L%d-L%d\n",
5078: @var{value}, @var{rel})
5079: @end example
5080:
5081: @findex ASM_OUTPUT_ADDR_VEC_ELT
5082: @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
5083: This macro should be provided on machines where the addresses
5084: in a dispatch table are absolute.
5085:
5086: The definition should be a C statement to output to the stdio stream
5087: @var{stream} an assembler pseudo-instruction to generate a reference to
5088: a label. @var{value} is the number of an internal label whose
5089: definition is output using @code{ASM_OUTPUT_INTERNAL_LABEL}.
5090: For example,
5091:
5092: @example
5093: fprintf (@var{stream}, "\t.word L%d\n", @var{value})
5094: @end example
5095:
5096: @findex ASM_OUTPUT_CASE_LABEL
5097: @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
5098: Define this if the label before a jump-table needs to be output
5099: specially. The first three arguments are the same as for
5100: @code{ASM_OUTPUT_INTERNAL_LABEL}; the fourth argument is the
5101: jump-table which follows (a @code{jump_insn} containing an
5102: @code{addr_vec} or @code{addr_diff_vec}).
5103:
5104: This feature is used on system V to output a @code{swbeg} statement
5105: for the table.
5106:
5107: If this macro is not defined, these labels are output with
5108: @code{ASM_OUTPUT_INTERNAL_LABEL}.
5109:
5110: @findex ASM_OUTPUT_CASE_END
5111: @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
5112: Define this if something special must be output at the end of a
5113: jump-table. The definition should be a C statement to be executed
5114: after the assembler code for the table is written. It should write
5115: the appropriate code to stdio stream @var{stream}. The argument
5116: @var{table} is the jump-table insn, and @var{num} is the label-number
5117: of the preceding label.
5118:
5119: If this macro is not defined, nothing special is output at the end of
5120: the jump-table.
5121: @end table
5122:
5123: @node Alignment Output
5124: @subsection Assembler Commands for Alignment
5125:
5126: @table @code
5127: @findex ASM_OUTPUT_ALIGN_CODE
5128: @item ASM_OUTPUT_ALIGN_CODE (@var{file})
5129: A C expression to output text to align the location counter in the way
5130: that is desirable at a point in the code that is reached only by
5131: jumping.
5132:
5133: This macro need not be defined if you don't want any special alignment
5134: to be done at such a time. Most machine descriptions do not currently
5135: define the macro.
5136:
5137: @findex ASM_OUTPUT_LOOP_ALIGN
5138: @item ASM_OUTPUT_LOOP_ALIGN (@var{file})
5139: A C expression to output text to align the location counter in the way
5140: that is desirable at the beginning of a loop.
5141:
5142: This macro need not be defined if you don't want any special alignment
5143: to be done at such a time. Most machine descriptions do not currently
5144: define the macro.
5145:
5146: @findex ASM_OUTPUT_SKIP
5147: @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
5148: A C statement to output to the stdio stream @var{stream} an assembler
5149: instruction to advance the location counter by @var{nbytes} bytes.
5150: Those bytes should be zero when loaded. @var{nbytes} will be a C
5151: expression of type @code{int}.
5152:
5153: @findex ASM_NO_SKIP_IN_TEXT
5154: @item ASM_NO_SKIP_IN_TEXT
5155: Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
5156: text section because it fails put zeros in the bytes that are skipped.
5157: This is true on many Unix systems, where the pseudo--op to skip bytes
5158: produces no-op instructions rather than zeros when used in the text
5159: section.
5160:
5161: @findex ASM_OUTPUT_ALIGN
5162: @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
5163: A C statement to output to the stdio stream @var{stream} an assembler
5164: command to advance the location counter to a multiple of 2 to the
5165: @var{power} bytes. @var{power} will be a C expression of type @code{int}.
5166: @end table
5167:
5168: @need 3000
5169: @node Debugging Info
5170: @section Controlling Debugging Information Format
5171: @c this should have some text in here.... --mew 10feb93
5172:
5173: @menu
5174: * All Debuggers:: Macros that affect all debugging formats uniformly.
5175: * DBX Options:: Macros enabling specific options in DBX format.
5176: * DBX Hooks:: Hook macros for varying DBX format.
5177: * File Names and DBX:: Macros controlling output of file names in DBX format.
5178: * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
5179: @end menu
5180:
5181: @node All Debuggers
5182: @subsection Macros Affecting All Debugging Formats
5183:
5184: @table @code
5185: @findex DBX_REGISTER_NUMBER
5186: @item DBX_REGISTER_NUMBER (@var{regno})
5187: A C expression that returns the DBX register number for the compiler
5188: register number @var{regno}. In simple cases, the value of this
5189: expression may be @var{regno} itself. But sometimes there are some
5190: registers that the compiler knows about and DBX does not, or vice
5191: versa. In such cases, some register may need to have one number in
5192: the compiler and another for DBX.
5193:
5194: If two registers have consecutive numbers inside GNU CC, and they can be
5195: used as a pair to hold a multiword value, then they @emph{must} have
5196: consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
5197: Otherwise, debuggers will be unable to access such a pair, because they
5198: expect register pairs to be consecutive in their own numbering scheme.
5199:
5200: If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
5201: does not preserve register pairs, then what you must do instead is
5202: redefine the actual register numbering scheme.
5203:
5204: @findex DEBUGGER_AUTO_OFFSET
5205: @item DEBUGGER_AUTO_OFFSET (@var{x})
5206: A C expression that returns the integer offset value for an automatic
5207: variable having address @var{x} (an RTL expression). The default
5208: computation assumes that @var{x} is based on the frame-pointer and
5209: gives the offset from the frame-pointer. This is required for targets
5210: that produce debugging output for DBX or COFF-style debugging output
5211: for SDB and allow the frame-pointer to be eliminated when the
5212: @samp{-g} options is used.
5213:
5214: @findex DEBUGGER_ARG_OFFSET
5215: @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
5216: A C expression that returns the integer offset value for an argument
5217: having address @var{x} (an RTL expression). The nominal offset is
5218: @var{offset}.
5219:
5220: @findex PREFERRED_DEBUGGING_TYPE
5221: @item PREFERRED_DEBUGGING_TYPE
5222: A C expression that returns the type of debugging output GNU CC produces
5223: when the user specifies @samp{-g} or @samp{-ggdb}. Define this if you
5224: have arranged for GNU CC to support more than one format of debugging
5225: output. Currently, the allowable values are @code{DBX_DEBUG},
5226: @code{SDB_DEBUG}, @code{DWARF_DEBUG}, and @code{XCOFF_DEBUG}.
5227:
5228: The value of this macro only affects the default debugging output; the
5229: user can always get a specific type of output by using @samp{-gstabs},
5230: @samp{-gcoff}, @samp{-gdwarf}, or @samp{-gxcoff}.
5231: @end table
5232:
5233: @node DBX Options
5234: @subsection Specific Options for DBX Output
5235:
5236: @table @code
5237: @findex DBX_DEBUGGING_INFO
5238: @item DBX_DEBUGGING_INFO
5239: Define this macro if GNU CC should produce debugging output for DBX
5240: in response to the @samp{-g} option.
5241:
5242: @findex XCOFF_DEBUGGING_INFO
5243: @item XCOFF_DEBUGGING_INFO
5244: Define this macro if GNU CC should produce XCOFF format debugging output
5245: in response to the @samp{-g} option. This is a variant of DBX format.
5246:
5247: @findex DEFAULT_GDB_EXTENSIONS
5248: @item DEFAULT_GDB_EXTENSIONS
5249: Define this macro to control whether GNU CC should by default generate
5250: GDB's extended version of DBX debugging information (assuming DBX-format
5251: debugging information is enabled at all). If you don't define the
5252: macro, the default is 1: always generate the extended information
5253: if there is any occasion to.
5254:
5255: @findex DEBUG_SYMS_TEXT
5256: @item DEBUG_SYMS_TEXT
5257: Define this macro if all @code{.stabs} commands should be output while
5258: in the text section.
5259:
5260: @findex ASM_STABS_OP
5261: @item ASM_STABS_OP
5262: A C string constant naming the assembler pseudo op to use instead of
5263: @code{.stabs} to define an ordinary debugging symbol. If you don't
5264: define this macro, @code{.stabs} is used. This macro applies only to
5265: DBX debugging information format.
5266:
5267: @findex ASM_STABD_OP
5268: @item ASM_STABD_OP
5269: A C string constant naming the assembler pseudo op to use instead of
5270: @code{.stabd} to define a debugging symbol whose value is the current
5271: location. If you don't define this macro, @code{.stabd} is used.
5272: This macro applies only to DBX debugging information format.
5273:
5274: @findex ASM_STABN_OP
5275: @item ASM_STABN_OP
5276: A C string constant naming the assembler pseudo op to use instead of
5277: @code{.stabn} to define a debugging symbol with no name. If you don't
5278: define this macro, @code{.stabn} is used. This macro applies only to
5279: DBX debugging information format.
5280:
5281: @findex DBX_NO_XREFS
5282: @item DBX_NO_XREFS
5283: Define this macro if DBX on your system does not support the construct
5284: @samp{xs@var{tagname}}. On some systems, this construct is used to
5285: describe a forward reference to a structure named @var{tagname}.
5286: On other systems, this construct is not supported at all.
5287:
5288: @findex DBX_CONTIN_LENGTH
5289: @item DBX_CONTIN_LENGTH
5290: A symbol name in DBX-format debugging information is normally
5291: continued (split into two separate @code{.stabs} directives) when it
5292: exceeds a certain length (by default, 80 characters). On some
5293: operating systems, DBX requires this splitting; on others, splitting
5294: must not be done. You can inhibit splitting by defining this macro
5295: with the value zero. You can override the default splitting-length by
5296: defining this macro as an expression for the length you desire.
5297:
5298: @findex DBX_CONTIN_CHAR
5299: @item DBX_CONTIN_CHAR
5300: Normally continuation is indicated by adding a @samp{\} character to
5301: the end of a @code{.stabs} string when a continuation follows. To use
5302: a different character instead, define this macro as a character
5303: constant for the character you want to use. Do not define this macro
5304: if backslash is correct for your system.
5305:
5306: @findex DBX_STATIC_STAB_DATA_SECTION
5307: @item DBX_STATIC_STAB_DATA_SECTION
5308: Define this macro if it is necessary to go to the data section before
5309: outputting the @samp{.stabs} pseudo-op for a non-global static
5310: variable.
5311:
5312: @findex DBX_TYPE_DECL_STABS_CODE
5313: @item DBX_TYPE_DECL_STABS_CODE
5314: The value to use in the ``code'' field of the @code{.stabs} directive
5315: for a typedef. The default is @code{N_LSYM}.
5316:
5317: @findex DBX_STATIC_CONST_VAR_CODE
5318: @item DBX_STATIC_CONST_VAR_CODE
5319: The value to use in the ``code'' field of the @code{.stabs} directive
5320: for a static variable located in the text section. DBX format does not
5321: provide any ``right'' way to do this. The default is @code{N_FUN}.
5322:
5323: @findex DBX_REGPARM_STABS_CODE
5324: @item DBX_REGPARM_STABS_CODE
5325: The value to use in the ``code'' field of the @code{.stabs} directive
5326: for a parameter passed in registers. DBX format does not provide any
5327: ``right'' way to do this. The default is @code{N_RSYM}.
5328:
5329: @findex DBX_REGPARM_STABS_LETTER
5330: @item DBX_REGPARM_STABS_LETTER
5331: The letter to use in DBX symbol data to identify a symbol as a parameter
5332: passed in registers. DBX format does not customarily provide any way to
5333: do this. The default is @code{'P'}.
5334:
5335: @findex DBX_MEMPARM_STABS_LETTER
5336: @item DBX_MEMPARM_STABS_LETTER
5337: The letter to use in DBX symbol data to identify a symbol as a stack
5338: parameter. The default is @code{'p'}.
5339:
5340: @findex DBX_FUNCTION_FIRST
5341: @item DBX_FUNCTION_FIRST
5342: Define this macro if the DBX information for a function and its
5343: arguments should precede the assembler code for the function. Normally,
5344: in DBX format, the debugging information entirely follows the assembler
5345: code.
5346:
5347: @findex DBX_LBRAC_FIRST
5348: @item DBX_LBRAC_FIRST
5349: Define this macro if the @code{N_LBRAC} symbol for a block should
5350: precede the debugging information for variables and functions defined in
5351: that block. Normally, in DBX format, the @code{N_LBRAC} symbol comes
5352: first.
5353: @end table
5354:
5355: @node DBX Hooks
5356: @subsection Open-Ended Hooks for DBX Format
5357:
5358: @table @code
5359: @findex DBX_OUTPUT_LBRAC
5360: @item DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
5361: Define this macro to say how to output to @var{stream} the debugging
5362: information for the start of a scope level for variable names. The
5363: argument @var{name} is the name of an assembler symbol (for use with
5364: @code{assemble_name}) whose value is the address where the scope begins.
5365:
5366: @findex DBX_OUTPUT_RBRAC
5367: @item DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
5368: Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
5369:
5370: @findex DBX_OUTPUT_ENUM
5371: @item DBX_OUTPUT_ENUM (@var{stream}, @var{type})
5372: Define this macro if the target machine requires special handling to
5373: output an enumeration type. The definition should be a C statement
5374: (sans semicolon) to output the appropriate information to @var{stream}
5375: for the type @var{type}.
5376:
5377: @findex DBX_OUTPUT_FUNCTION_END
5378: @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function})
5379: Define this macro if the target machine requires special output at the
5380: end of the debugging information for a function. The definition should
5381: be a C statement (sans semicolon) to output the appropriate information
5382: to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for
5383: the function.
5384:
5385: @findex DBX_OUTPUT_STANDARD_TYPES
5386: @item DBX_OUTPUT_STANDARD_TYPES (@var{syms})
5387: Define this macro if you need to control the order of output of the
5388: standard data types at the beginning of compilation. The argument
5389: @var{syms} is a @code{tree} which is a chain of all the predefined
5390: global symbols, including names of data types.
5391:
5392: Normally, DBX output starts with definitions of the types for integers
5393: and characters, followed by all the other predefined types of the
5394: particular language in no particular order.
5395:
5396: On some machines, it is necessary to output different particular types
5397: first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output
5398: those symbols in the necessary order. Any predefined types that you
5399: don't explicitly output will be output afterward in no particular order.
5400:
5401: Be careful not to define this macro so that it works only for C. There
5402: are no global variables to access most of the built-in types, because
5403: another language may have another set of types. The way to output a
5404: particular type is to look through @var{syms} to see if you can find it.
5405: Here is an example:
5406:
5407: @smallexample
5408: @{
5409: tree decl;
5410: for (decl = syms; decl; decl = TREE_CHAIN (decl))
5411: if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)),
5412: "long int"))
5413: dbxout_symbol (decl);
5414: @dots{}
5415: @}
5416: @end smallexample
5417:
5418: @noindent
5419: This does nothing if the expected type does not exist.
5420:
5421: See the function @code{init_decl_processing} in @file{c-decl.c} to find
5422: the names to use for all the built-in C types.
5423:
5424: Here is another way of finding a particular type:
5425:
5426: @c this is still overfull. --mew 10feb93
5427: @smallexample
5428: @{
5429: tree decl;
5430: for (decl = syms; decl; decl = TREE_CHAIN (decl))
5431: if (TREE_CODE (decl) == TYPE_DECL
5432: && (TREE_CODE (TREE_TYPE (decl))
5433: == INTEGER_CST)
5434: && TYPE_PRECISION (TREE_TYPE (decl)) == 16
5435: && TYPE_UNSIGNED (TREE_TYPE (decl)))
5436: @group
5437: /* @r{This must be @code{unsigned short}.} */
5438: dbxout_symbol (decl);
5439: @dots{}
5440: @}
5441: @end group
5442: @end smallexample
5443: @end table
5444:
5445: @node File Names and DBX
5446: @subsection File Names in DBX Format
5447:
5448: @table @code
5449: @findex DBX_WORKING_DIRECTORY
5450: @item DBX_WORKING_DIRECTORY
5451: Define this if DBX wants to have the current directory recorded in each
5452: object file.
5453:
5454: Note that the working directory is always recorded if GDB extensions are
5455: enabled.
5456:
5457: @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME
5458: @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
5459: A C statement to output DBX debugging information to the stdio stream
5460: @var{stream} which indicates that file @var{name} is the main source
5461: file---the file specified as the input file for compilation.
5462: This macro is called only once, at the beginning of compilation.
5463:
5464: This macro need not be defined if the standard form of output
5465: for DBX debugging information is appropriate.
5466:
5467: @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY
5468: @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name})
5469: A C statement to output DBX debugging information to the stdio stream
5470: @var{stream} which indicates that the current directory during
5471: compilation is named @var{name}.
5472:
5473: This macro need not be defined if the standard form of output
5474: for DBX debugging information is appropriate.
5475:
5476: @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END
5477: @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
5478: A C statement to output DBX debugging information at the end of
5479: compilation of the main source file @var{name}.
5480:
5481: If you don't define this macro, nothing special is output at the end
5482: of compilation, which is correct for most machines.
5483:
5484: @findex DBX_OUTPUT_SOURCE_FILENAME
5485: @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
5486: A C statement to output DBX debugging information to the stdio stream
5487: @var{stream} which indicates that file @var{name} is the current source
5488: file. This output is generated each time input shifts to a different
5489: source file as a result of @samp{#include}, the end of an included file,
5490: or a @samp{#line} command.
5491:
5492: This macro need not be defined if the standard form of output
5493: for DBX debugging information is appropriate.
5494: @end table
5495:
5496: @need 2000
5497: @node SDB and DWARF
5498: @subsection Macros for SDB and DWARF Output
5499:
5500: @table @code
5501: @findex SDB_DEBUGGING_INFO
5502: @item SDB_DEBUGGING_INFO
5503: Define this macro if GNU CC should produce COFF-style debugging output
5504: for SDB in response to the @samp{-g} option.
5505:
5506: @findex DWARF_DEBUGGING_INFO
5507: @item DWARF_DEBUGGING_INFO
5508: Define this macro if GNU CC should produce dwarf format debugging output
5509: in response to the @samp{-g} option.
5510:
5511: @findex PUT_SDB_@dots{}
5512: @item PUT_SDB_@dots{}
5513: Define these macros to override the assembler syntax for the special
5514: SDB assembler directives. See @file{sdbout.c} for a list of these
5515: macros and their arguments. If the standard syntax is used, you need
5516: not define them yourself.
5517:
5518: @findex SDB_DELIM
5519: @item SDB_DELIM
5520: Some assemblers do not support a semicolon as a delimiter, even between
5521: SDB assembler directives. In that case, define this macro to be the
5522: delimiter to use (usually @samp{\n}). It is not necessary to define
5523: a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
5524: required.
5525:
5526: @findex SDB_GENERATE_FAKE
5527: @item SDB_GENERATE_FAKE
5528: Define this macro to override the usual method of constructing a dummy
5529: name for anonymous structure and union types. See @file{sdbout.c} for
5530: more information.
5531:
5532: @findex SDB_ALLOW_UNKNOWN_REFERENCES
5533: @item SDB_ALLOW_UNKNOWN_REFERENCES
5534: Define this macro to allow references to unknown structure,
5535: union, or enumeration tags to be emitted. Standard COFF does not
5536: allow handling of unknown references, MIPS ECOFF has support for
5537: it.
5538:
5539: @findex SDB_ALLOW_FORWARD_REFERENCES
5540: @item SDB_ALLOW_FORWARD_REFERENCES
5541: Define this macro to allow references to structure, union, or
5542: enumeration tags that have not yet been seen to be handled. Some
5543: assemblers choke if forward tags are used, while some require it.
5544: @end table
5545:
5546: @node Cross-compilation
5547: @section Cross Compilation and Floating Point
5548: @cindex cross compilation and floating point
5549: @cindex floating point and cross compilation
5550:
5551: While all modern machines use 2's complement representation for integers,
5552: there are a variety of representations for floating point numbers. This
5553: means that in a cross-compiler the representation of floating point numbers
5554: in the compiled program may be different from that used in the machine
5555: doing the compilation.
5556:
5557: @findex atof
5558: Because different representation systems may offer different amounts of
5559: range and precision, the cross compiler cannot safely use the host
5560: machine's floating point arithmetic. Therefore, floating point constants
5561: must be represented in the target machine's format. This means that the
5562: cross compiler cannot use @code{atof} to parse a floating point constant;
5563: it must have its own special routine to use instead. Also, constant
5564: folding must emulate the target machine's arithmetic (or must not be done
5565: at all).
5566:
5567: The macros in the following table should be defined only if you are cross
5568: compiling between different floating point formats.
5569:
5570: Otherwise, don't define them. Then default definitions will be set up which
5571: use @code{double} as the data type, @code{==} to test for equality, etc.
5572:
5573: You don't need to worry about how many times you use an operand of any
5574: of these macros. The compiler never uses operands which have side effects.
5575:
5576: @table @code
5577: @findex REAL_VALUE_TYPE
5578: @item REAL_VALUE_TYPE
5579: A macro for the C data type to be used to hold a floating point value
5580: in the target machine's format. Typically this would be a
5581: @code{struct} containing an array of @code{int}.
5582:
5583: @findex REAL_VALUES_EQUAL
5584: @item REAL_VALUES_EQUAL (@var{x}, @var{y})
5585: A macro for a C expression which compares for equality the two values,
5586: @var{x} and @var{y}, both of type @code{REAL_VALUE_TYPE}.
5587:
5588: @findex REAL_VALUES_LESS
5589: @item REAL_VALUES_LESS (@var{x}, @var{y})
5590: A macro for a C expression which tests whether @var{x} is less than
5591: @var{y}, both values being of type @code{REAL_VALUE_TYPE} and
5592: interpreted as floating point numbers in the target machine's
5593: representation.
5594:
5595: @findex REAL_VALUE_LDEXP
5596: @findex ldexp
5597: @item REAL_VALUE_LDEXP (@var{x}, @var{scale})
5598: A macro for a C expression which performs the standard library
5599: function @code{ldexp}, but using the target machine's floating point
5600: representation. Both @var{x} and the value of the expression have
5601: type @code{REAL_VALUE_TYPE}. The second argument, @var{scale}, is an
5602: integer.
5603:
5604: @findex REAL_VALUE_FIX
5605: @item REAL_VALUE_FIX (@var{x})
5606: A macro whose definition is a C expression to convert the target-machine
5607: floating point value @var{x} to a signed integer. @var{x} has type
5608: @code{REAL_VALUE_TYPE}.
5609:
5610: @findex REAL_VALUE_UNSIGNED_FIX
5611: @item REAL_VALUE_UNSIGNED_FIX (@var{x})
5612: A macro whose definition is a C expression to convert the target-machine
5613: floating point value @var{x} to an unsigned integer. @var{x} has type
5614: @code{REAL_VALUE_TYPE}.
5615:
5616: @findex REAL_VALUE_RNDZINT
5617: @item REAL_VALUE_RNDZINT (@var{x})
5618: A macro whose definition is a C expression to round the target-machine
5619: floating point value @var{x} towards zero to an integer value (but still
5620: as a floating point number). @var{x} has type @code{REAL_VALUE_TYPE},
5621: and so does the value.
5622:
5623: @findex REAL_VALUE_UNSIGNED_RNDZINT
5624: @item REAL_VALUE_UNSIGNED_RNDZINT (@var{x})
5625: A macro whose definition is a C expression to round the target-machine
5626: floating point value @var{x} towards zero to an unsigned integer value
5627: (but still represented as a floating point number). @var{x} has type
5628: @code{REAL_VALUE_TYPE}, and so does the value.
5629:
5630: @findex REAL_VALUE_ATOF
5631: @item REAL_VALUE_ATOF (@var{string}, @var{mode})
5632: A macro for a C expression which converts @var{string}, an expression of
5633: type @code{char *}, into a floating point number in the target machine's
5634: representation for mode @var{mode}. The value has type
5635: @code{REAL_VALUE_TYPE}.
5636:
5637: @findex REAL_INFINITY
5638: @item REAL_INFINITY
5639: Define this macro if infinity is a possible floating point value, and
5640: therefore division by 0 is legitimate.
5641:
5642: @findex REAL_VALUE_ISINF
5643: @findex isinf
5644: @item REAL_VALUE_ISINF (@var{x})
5645: A macro for a C expression which determines whether @var{x}, a floating
5646: point value, is infinity. The value has type @code{int}.
5647: By default, this is defined to call @code{isinf}.
5648:
5649: @findex REAL_VALUE_ISNAN
5650: @findex isnan
5651: @item REAL_VALUE_ISNAN (@var{x})
5652: A macro for a C expression which determines whether @var{x}, a floating
5653: point value, is a ``nan'' (not-a-number). The value has type
5654: @code{int}. By default, this is defined to call @code{isnan}.
5655: @end table
5656:
5657: @cindex constant folding and floating point
5658: Define the following additional macros if you want to make floating
5659: point constant folding work while cross compiling. If you don't
5660: define them, cross compilation is still possible, but constant folding
5661: will not happen for floating point values.
5662:
5663: @table @code
5664: @findex REAL_ARITHMETIC
5665: @item REAL_ARITHMETIC (@var{output}, @var{code}, @var{x}, @var{y})
5666: A macro for a C statement which calculates an arithmetic operation of
5667: the two floating point values @var{x} and @var{y}, both of type
5668: @code{REAL_VALUE_TYPE} in the target machine's representation, to
5669: produce a result of the same type and representation which is stored
5670: in @var{output} (which will be a variable).
5671:
5672: The operation to be performed is specified by @var{code}, a tree code
5673: which will always be one of the following: @code{PLUS_EXPR},
5674: @code{MINUS_EXPR}, @code{MULT_EXPR}, @code{RDIV_EXPR},
5675: @code{MAX_EXPR}, @code{MIN_EXPR}.@refill
5676:
5677: @cindex overflow while constant folding
5678: The expansion of this macro is responsible for checking for overflow.
5679: If overflow happens, the macro expansion should execute the statement
5680: @code{return 0;}, which indicates the inability to perform the
5681: arithmetic operation requested.
5682:
5683: @findex REAL_VALUE_NEGATE
5684: @item REAL_VALUE_NEGATE (@var{x})
5685: A macro for a C expression which returns the negative of the floating
5686: point value @var{x}. Both @var{x} and the value of the expression
5687: have type @code{REAL_VALUE_TYPE} and are in the target machine's
5688: floating point representation.
5689:
5690: There is no way for this macro to report overflow, since overflow
5691: can't happen in the negation operation.
5692:
5693: @findex REAL_VALUE_TRUNCATE
5694: @item REAL_VALUE_TRUNCATE (@var{mode}, @var{x})
5695: A macro for a C expression which converts the floating point value
5696: @var{x} to mode @var{mode}.
5697:
5698: Both @var{x} and the value of the expression are in the target machine's
5699: floating point representation and have type @code{REAL_VALUE_TYPE}.
5700: However, the value should have an appropriate bit pattern to be output
5701: properly as a floating constant whose precision accords with mode
5702: @var{mode}.
5703:
5704: There is no way for this macro to report overflow.
5705:
5706: @findex REAL_VALUE_TO_INT
5707: @item REAL_VALUE_TO_INT (@var{low}, @var{high}, @var{x})
5708: A macro for a C expression which converts a floating point value
5709: @var{x} into a double-precision integer which is then stored into
5710: @var{low} and @var{high}, two variables of type @var{int}.
5711:
5712: @item REAL_VALUE_FROM_INT (@var{x}, @var{low}, @var{high})
5713: @findex REAL_VALUE_FROM_INT
5714: A macro for a C expression which converts a double-precision integer
5715: found in @var{low} and @var{high}, two variables of type @var{int},
5716: into a floating point value which is then stored into @var{x}.
5717: @end table
5718:
5719: @node Misc
5720: @section Miscellaneous Parameters
5721: @cindex parameters, miscellaneous
5722:
5723: @table @code
5724: @item PREDICATE_CODES
5725: @findex PREDICATE_CODES
5726: Define this if you have defined special-purpose predicates in the file
5727: @file{@var{machine}.c}. This macro is called within an initializer of an
5728: array of structures. The first field in the structure is the name of a
5729: predicate and the second field is an array of rtl codes. For each
5730: predicate, list all rtl codes that can be in expressions matched by the
5731: predicate. The list should have a trailing comma. Here is an example
5732: of two entries in the list for a typical RISC machine:
5733:
5734: @smallexample
5735: #define PREDICATE_CODES \
5736: @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \
5737: @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@},
5738: @end smallexample
5739:
5740: Defining this macro does not affect the generated code (however,
5741: incorrect definitions that omit an rtl code that may be matched by the
5742: predicate can cause the compiler to malfunction). Instead, it allows
5743: the table built by @file{genrecog} to be more compact and efficient,
5744: thus speeding up the compiler. The most important predicates to include
5745: in the list specified by this macro are thoses used in the most insn
5746: patterns.
5747:
5748: @findex CASE_VECTOR_MODE
5749: @item CASE_VECTOR_MODE
5750: An alias for a machine mode name. This is the machine mode that
5751: elements of a jump-table should have.
5752:
5753: @findex CASE_VECTOR_PC_RELATIVE
5754: @item CASE_VECTOR_PC_RELATIVE
5755: Define this macro if jump-tables should contain relative addresses.
5756:
5757: @findex CASE_DROPS_THROUGH
5758: @item CASE_DROPS_THROUGH
5759: Define this if control falls through a @code{case} insn when the index
5760: value is out of range. This means the specified default-label is
5761: actually ignored by the @code{case} insn proper.
5762:
5763: @findex CASE_VALUES_THRESHOLD
5764: @item CASE_VALUES_THRESHOLD
5765: Define this to be the smallest number of different values for which it
5766: is best to use a jump-table instead of a tree of conditional branches.
5767: The default is four for machines with a @code{casesi} instruction and
5768: five otherwise. This is best for most machines.
5769:
5770: @findex WORD_REGISTER_OPERATIONS
5771: @item WORD_REGISTER_OPERATIONS
5772: Define this macro if operations between registers with integral mode
5773: smaller than a word are always performed on the entire register.
5774: Most RISC machines have this property and most CISC machines do not.
5775:
5776: @findex LOAD_EXTEND_OP
5777: @item LOAD_EXTEND_OP (@var{mode})
5778: Define this macro to be a C expression indicating when insns that read
5779: memory in @var{mode}, an integral mode narrower than a word, set the
5780: bits outside of @var{mode} to be either the sign-extension or the
5781: zero-extension of the data read. Return @code{SIGN_EXTEND} for values
5782: of @var{mode} for which the
5783: insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
5784: @code{NIL} for other modes.
5785:
5786: This macro is not called with @var{mode} non-integral or with a width
5787: greater than or equal to @code{BITS_PER_WORD}, so you may return any
5788: value in this case. Do not define this macro if it would always return
5789: @code{NIL}. On machines where this macro is defined, you will normally
5790: define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
5791:
5792: @findex IMPLICIT_FIX_EXPR
5793: @item IMPLICIT_FIX_EXPR
5794: An alias for a tree code that should be used by default for conversion
5795: of floating point values to fixed point. Normally,
5796: @code{FIX_ROUND_EXPR} is used.@refill
5797:
5798: @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC
5799: @item FIXUNS_TRUNC_LIKE_FIX_TRUNC
5800: Define this macro if the same instructions that convert a floating
5801: point number to a signed fixed point number also convert validly to an
5802: unsigned one.
5803:
5804: @findex EASY_DIV_EXPR
5805: @item EASY_DIV_EXPR
5806: An alias for a tree code that is the easiest kind of division to
5807: compile code for in the general case. It may be
5808: @code{TRUNC_DIV_EXPR}, @code{FLOOR_DIV_EXPR}, @code{CEIL_DIV_EXPR} or
5809: @code{ROUND_DIV_EXPR}. These four division operators differ in how
5810: they round the result to an integer. @code{EASY_DIV_EXPR} is used
5811: when it is permissible to use any of those kinds of division and the
5812: choice should be made on the basis of efficiency.@refill
5813:
5814: @findex MOVE_MAX
5815: @item MOVE_MAX
5816: The maximum number of bytes that a single instruction can move quickly
5817: from memory to memory.
5818:
5819: @findex MAX_MOVE_MAX
5820: @item MAX_MOVE_MAX
5821: The maximum number of bytes that a single instruction can move quickly
5822: from memory to memory. If this is undefined, the default is
5823: @code{MOVE_MAX}. Otherwise, it is the constant value that is the
5824: largest value that @code{MOVE_MAX} can have at run-time.
5825:
5826: @findex SHIFT_COUNT_TRUNCATED
5827: @item SHIFT_COUNT_TRUNCATED
5828: A C expression that is nonzero if on this machine the number of bits
5829: actually used for the count of a shift operation is equal to the number
5830: of bits needed to represent the size of the object being shifted. When
5831: this macro is non-zero, the compiler will assume that it is safe to omit
5832: a sign-extend, zero-extend, and certain bitwise `and' instructions that
5833: truncates the count of a shift operation. On machines that have
5834: instructions that act on bitfields at variable positions, which may
5835: include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
5836: also enables deletion of truncations of the values that serve as
5837: arguments to bitfield instructions.
5838:
5839: If both types of instructions truncate the count (for shifts) and
5840: position (for bitfield operations), or if no variable-position bitfield
5841: instructions exist, you should define this macro.
5842:
5843: However, on some machines, such as the 80386 and the 680x0, truncation
5844: only applies to shift operations and not the (real or pretended)
5845: bitfield operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
5846: such machines. Instead, add patterns to the @file{md} file that include
5847: the implied truncation of the shift instructions.
5848:
5849: You need not define this macro if it would always have the value of zero.
5850:
5851: @findex TRULY_NOOP_TRUNCATION
5852: @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
5853: A C expression which is nonzero if on this machine it is safe to
5854: ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
5855: bits (where @var{outprec} is smaller than @var{inprec}) by merely
5856: operating on it as if it had only @var{outprec} bits.
5857:
5858: On many machines, this expression can be 1.
5859:
5860: @c rearranged this, removed the phrase "it is reported that". this was
5861: @c to fix an overfull hbox. --mew 10feb93
5862: When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
5863: modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
5864: If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
5865: such cases may improve things.
5866:
5867: @findex STORE_FLAG_VALUE
5868: @item STORE_FLAG_VALUE
5869: A C expression describing the value returned by a comparison operator
5870: with an integral mode and stored by a store-flag instruction
5871: (@samp{s@var{cond}}) when the condition is true. This description must
5872: apply to @emph{all} the @samp{s@var{cond}} patterns and all the
5873: comparison operators whose results have a @code{MODE_INT} mode.
5874:
5875: A value of 1 or -1 means that the instruction implementing the
5876: comparison operator returns exactly 1 or -1 when the comparison is true
5877: and 0 when the comparison is false. Otherwise, the value indicates
5878: which bits of the result are guaranteed to be 1 when the comparison is
5879: true. This value is interpreted in the mode of the comparison
5880: operation, which is given by the mode of the first operand in the
5881: @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
5882: @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
5883: the compiler.
5884:
5885: If @code{STORE_FLAG_VALUE} is neither 1 or -1, the compiler will
5886: generate code that depends only on the specified bits. It can also
5887: replace comparison operators with equivalent operations if they cause
5888: the required bits to be set, even if the remaining bits are undefined.
5889: For example, on a machine whose comparison operators return an
5890: @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
5891: @samp{0x80000000}, saying that just the sign bit is relevant, the
5892: expression
5893:
5894: @smallexample
5895: (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
5896: @end smallexample
5897:
5898: @noindent
5899: can be converted to
5900:
5901: @smallexample
5902: (ashift:SI @var{x} (const_int @var{n}))
5903: @end smallexample
5904:
5905: @noindent
5906: where @var{n} is the appropriate shift count to move the bit being
5907: tested into the sign bit.
5908:
5909: There is no way to describe a machine that always sets the low-order bit
5910: for a true value, but does not guarantee the value of any other bits,
5911: but we do not know of any machine that has such an instruction. If you
5912: are trying to port GNU CC to such a machine, include an instruction to
5913: perform a logical-and of the result with 1 in the pattern for the
5914: comparison operators and let us know
5915: @ifset USING
5916: (@pxref{Bug Reporting,,How to Report Bugs}).
5917: @end ifset
5918: @ifclear USING
5919: (@pxref{Bug Reporting,,How to Report Bugs,gcc.info,Using GCC}).
5920: @end ifclear
5921:
5922: Often, a machine will have multiple instructions that obtain a value
5923: from a comparison (or the condition codes). Here are rules to guide the
5924: choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
5925: to be used:
5926:
5927: @itemize @bullet
5928: @item
5929: Use the shortest sequence that yields a valid definition for
5930: @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
5931: ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
5932: comparison operators to do so because there may be opportunities to
5933: combine the normalization with other operations.
5934:
5935: @item
5936: For equal-length sequences, use a value of 1 or -1, with -1 being
5937: slightly preferred on machines with expensive jumps and 1 preferred on
5938: other machines.
5939:
5940: @item
5941: As a second choice, choose a value of @samp{0x80000001} if instructions
5942: exist that set both the sign and low-order bits but do not define the
5943: others.
5944:
5945: @item
5946: Otherwise, use a value of @samp{0x80000000}.
5947: @end itemize
5948:
5949: Many machines can produce both the value chosen for
5950: @code{STORE_FLAG_VALUE} and its negation in the same number of
5951: instructions. On those machines, you should also define a pattern for
5952: those cases, e.g., one matching
5953:
5954: @smallexample
5955: (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
5956: @end smallexample
5957:
5958: Some machines can also perform @code{and} or @code{plus} operations on
5959: condition code values with less instructions than the corresponding
5960: @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
5961: machines, define the appropriate patterns. Use the names @code{incscc}
5962: and @code{decscc}, respectively, for the the patterns which perform
5963: @code{plus} or @code{minus} operations on condition code values. See
5964: @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
5965: find such instruction sequences on other machines.
5966:
5967: You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
5968: instructions.
5969:
5970: @findex FLOAT_STORE_FLAG_VALUE
5971: @item FLOAT_STORE_FLAG_VALUE
5972: A C expression that gives a non-zero floating point value that is
5973: returned when comparison operators with floating-point results are true.
5974: Define this macro on machine that have comparison operations that return
5975: floating-point values. If there are no such operations, do not define
5976: this macro.
5977:
5978: @findex Pmode
5979: @item Pmode
5980: An alias for the machine mode for pointers. Normally the definition
5981: can be
5982:
5983: @smallexample
5984: #define Pmode SImode
5985: @end smallexample
5986:
5987: @findex FUNCTION_MODE
5988: @item FUNCTION_MODE
5989: An alias for the machine mode used for memory references to functions
5990: being called, in @code{call} RTL expressions. On most machines this
5991: should be @code{QImode}.
5992:
5993: @findex INTEGRATE_THRESHOLD
5994: @item INTEGRATE_THRESHOLD (@var{decl})
5995: A C expression for the maximum number of instructions above which the
5996: function @var{decl} should not be inlined. @var{decl} is a
5997: @code{FUNCTION_DECL} node.
5998:
5999: The default definition of this macro is 64 plus 8 times the number of
6000: arguments that the function accepts. Some people think a larger
6001: threshold should be used on RISC machines.
6002:
6003: @findex SCCS_DIRECTIVE
6004: @item SCCS_DIRECTIVE
6005: Define this if the preprocessor should ignore @code{#sccs} directives
6006: and print no error message.
6007:
6008: @findex NO_IMPLICIT_EXTERN_C
6009: @item NO_IMPLICIT_EXTERN_C
6010: Define this macro if the system header files support C++ as well as C.
6011: This macro inhibits the usual method of using system header files in
6012: C++, which is to pretend that the file's contents are enclosed in
6013: @samp{extern "C" @{@dots{}@}}.
6014:
6015: @findex HANDLE_PRAGMA
6016: @findex #pragma
6017: @findex pragma
6018: @item HANDLE_PRAGMA (@var{stream})
6019: Define this macro if you want to implement any pragmas. If defined, it
6020: should be a C statement to be executed when @code{#pragma} is seen. The
6021: argument @var{stream} is the stdio input stream from which the source
6022: text can be read.
6023:
6024: It is generally a bad idea to implement new uses of @code{#pragma}. The
6025: only reason to define this macro is for compatibility with other
6026: compilers that do support @code{#pragma} for the sake of any user
6027: programs which already use it.
6028:
6029: @findex DOLLARS_IN_IDENTIFIERS
6030: @item DOLLARS_IN_IDENTIFIERS
6031: Define this macro to control use of the character @samp{$} in identifier
6032: names. The value should be 0, 1, or 2. 0 means @samp{$} is not allowed
6033: by default; 1 means it is allowed by default if @samp{-traditional} is
6034: used; 2 means it is allowed by default provided @samp{-ansi} is not used.
6035: 1 is the default; there is no need to define this macro in that case.
6036:
6037: @findex NO_DOLLAR_IN_LABEL
6038: @item NO_DOLLAR_IN_LABEL
6039: Define this macro if the assembler does not accept the character
6040: @samp{$} in label names. By default constructors and destructors in
6041: G++ have @samp{$} in the identifiers. If this macro is defined,
6042: @samp{.} is used instead.
6043:
6044: @findex NO_DOT_IN_LABEL
6045: @item NO_DOT_IN_LABEL
6046: Define this macro if the assembler does not accept the character
6047: @samp{.} in label names. By default constructors and destructors in G++
6048: have names that use @samp{.}. If this macro is defined, these names
6049: are rewritten to avoid @samp{.}.
6050:
6051: @findex DEFAULT_MAIN_RETURN
6052: @item DEFAULT_MAIN_RETURN
6053: Define this macro if the target system expects every program's @code{main}
6054: function to return a standard ``success'' value by default (if no other
6055: value is explicitly returned).
6056:
6057: The definition should be a C statement (sans semicolon) to generate the
6058: appropriate rtl instructions. It is used only when compiling the end of
6059: @code{main}.
6060:
6061: @item HAVE_ATEXIT
6062: @findex HAVE_ATEXIT
6063: Define this if the target system supports the function
6064: @code{atexit} from the ANSI C standard. If this is not defined,
6065: and @code{INIT_SECTION_ASM_OP} is not defined, a default
6066: @code{exit} function will be provided to support C++.
6067:
6068: @item EXIT_BODY
6069: @findex EXIT_BODY
6070: Define this if your @code{exit} function needs to do something
6071: besides calling an external function @code{_cleanup} before
6072: terminating with @code{_exit}. The @code{EXIT_BODY} macro is
6073: only needed if netiher @code{HAVE_ATEXIT} nor
6074: @code{INIT_SECTION_ASM_OP} are defined.
6075:
6076: @findex INSN_SETS_ARE_DELAYED
6077: @item INSN_SETS_ARE_DELAYED (@var{insn})
6078: Define this macro as a C expression that is nonzero if it is safe for the
6079: delay slot scheduler to place instructions in the delay slot of @var{insn},
6080: even if they appear to use a resource set or clobbered in @var{insn}.
6081: @var{insn} is always a @code{jump_insn} or an @code{insn}; GNU CC knows that
6082: every @code{call_insn} has this behavior. On machines where some @code{insn}
6083: or @code{jump_insn} is really a function call and hence has this behavior,
6084: you should define this macro.
6085:
6086: You need not define this macro if it would always return zero.
6087:
6088: @findex INSN_REFERENCES_ARE_DELAYED
6089: @item INSN_REFERENCES_ARE_DELAYED (@var{insn})
6090: Define this macro as a C expression that is nonzero if it is safe for the
6091: delay slot scheduler to place instructions in the delay slot of @var{insn},
6092: even if they appear to set or clobber a resource referenced in @var{insn}.
6093: @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
6094: some @code{insn} or @code{jump_insn} is really a function call and its operands
6095: are registers whose use is actually in the subroutine it calls, you should
6096: define this macro. Doing so allows the delay slot scheduler to move
6097: instructions which copy arguments into the argument registers into the delay
6098: slot of @var{insn}.
6099:
6100: You need not define this macro if it would always return zero.
6101: @end table
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.