* tags.h (scm_tc7_variable): New.
[bpt/guile.git] / libguile / tags.h
1 /* classes: h_files */
2
3 #ifndef SCM_TAGS_H
4 #define SCM_TAGS_H
5 /* Copyright (C) 1995,1996,1997,1998,1999,2000,2001 Free Software Foundation, Inc.
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2, or (at your option)
10 * any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this software; see the file COPYING. If not, write to
19 * the Free Software Foundation, Inc., 59 Temple Place, Suite 330,
20 * Boston, MA 02111-1307 USA
21 *
22 * As a special exception, the Free Software Foundation gives permission
23 * for additional uses of the text contained in its release of GUILE.
24 *
25 * The exception is that, if you link the GUILE library with other files
26 * to produce an executable, this does not by itself cause the
27 * resulting executable to be covered by the GNU General Public License.
28 * Your use of that executable is in no way restricted on account of
29 * linking the GUILE library code into it.
30 *
31 * This exception does not however invalidate any other reasons why
32 * the executable file might be covered by the GNU General Public License.
33 *
34 * This exception applies only to the code released by the
35 * Free Software Foundation under the name GUILE. If you copy
36 * code from other Free Software Foundation releases into a copy of
37 * GUILE, as the General Public License permits, the exception does
38 * not apply to the code that you add in this way. To avoid misleading
39 * anyone as to the status of such modified files, you must delete
40 * this exception notice from them.
41 *
42 * If you write modifications of your own for GUILE, it is your choice
43 * whether to permit this exception to apply to your modifications.
44 * If you do not wish that, delete this exception notice. */
45
46
47 \f
48
49 /** This file defines the format of SCM values and cons pairs.
50 ** It is here that tag bits are assigned for various purposes.
51 **/
52
53 #ifdef HAVE_STDINT_H
54 #include <stdint.h>
55 #endif
56
57 #ifdef HAVE_INTTYPES_H
58 #include <inttypes.h>
59 #endif
60
61 \f
62
63 /* In the beginning was the Word:
64 */
65 #ifdef HAVE_UINTPTR_T
66 typedef uintptr_t scm_t_bits;
67 typedef intptr_t scm_t_signed_bits;
68 #else
69 typedef unsigned long scm_t_bits;
70 typedef signed long scm_t_signed_bits;
71 #endif
72
73 /* But as external interface, we use SCM, which may, according to the desired
74 * level of type checking, be defined in several ways:
75 */
76 #if (SCM_DEBUG_TYPING_STRICTNESS == 2)
77 typedef union { struct { scm_t_bits n; } n; } SCM;
78 static SCM scm_pack(scm_t_bits b) { SCM s; s.n.n = b; return s; }
79 # define SCM_UNPACK(x) ((x).n.n)
80 # define SCM_PACK(x) (scm_pack ((scm_t_bits) (x)))
81 #elif (SCM_DEBUG_TYPING_STRICTNESS == 1)
82 /* This is the default, which provides an intermediate level of compile time
83 * type checking while still resulting in very efficient code.
84 */
85 typedef struct scm_unused_struct * SCM;
86 # define SCM_UNPACK(x) ((scm_t_bits) (x))
87 # define SCM_PACK(x) ((SCM) (x))
88 #else
89 /* This should be used as a fall back solution for machines on which casting
90 * to a pointer may lead to loss of bit information, e. g. in the three least
91 * significant bits.
92 */
93 typedef scm_t_bits SCM;
94 # define SCM_UNPACK(x) (x)
95 # define SCM_PACK(x) ((scm_t_bits) (x))
96 #endif
97
98
99 /* SCM values can not be compared by using the operator ==. Use the following
100 * macro instead, which is the equivalent of the scheme predicate 'eq?'.
101 */
102 #define SCM_EQ_P(x, y) (SCM_UNPACK (x) == SCM_UNPACK (y))
103
104 \f
105
106 /* SCM variables can contain:
107 *
108 * Non-objects -- meaning that the tag-related macros don't apply to them
109 * in the usual way.
110 *
111 * Immediates -- meaning that the variable contains an entire Scheme object.
112 *
113 * Non-immediates -- meaning that the variable holds a (possibly
114 * tagged) pointer into the cons pair heap.
115 *
116 * Non-objects are distinguished from other values by careful coding
117 * only (i.e., programmers must keep track of any SCM variables they
118 * create that don't contain ordinary scheme values).
119 *
120 * All immediates and non-immediates must have a 0 in bit 0. Only
121 * non-object values can have a 1 in bit 0. In some cases, bit 0 of a
122 * word in the heap is used for the GC tag so during garbage
123 * collection, that bit might be 1 even in an immediate or
124 * non-immediate value. In other cases, bit 0 of a word in the heap
125 * is used to tag a pointer to a GLOC (VM global variable address) or
126 * the header of a struct. But whenever an SCM variable holds a
127 * normal Scheme value, bit 0 is 0.
128 *
129 * Immediates and non-immediates are distinguished by bits two and four.
130 * Immediate values must have a 1 in at least one of those bits. Does
131 * this (or any other detail of tagging) seem arbitrary? Try changing it!
132 * (Not always impossible but it is fair to say that many details of tags
133 * are mutually dependent). */
134
135 #define SCM_IMP(x) (6 & SCM_UNPACK (x))
136 #define SCM_NIMP(x) (!SCM_IMP (x))
137
138 /* Here is a summary of tagging in SCM values as they might occur in
139 * SCM variables or in the heap.
140 *
141 * low bits meaning
142 *
143 *
144 * 0 Most objects except...
145 * 1 ...glocs and structs (this tag valid only in a SCM_CAR or
146 * in the header of a struct's data).
147 *
148 * 00 heap addresses and many immediates (not integers)
149 * 01 glocs/structs, some tc7_ codes
150 * 10 immediate integers
151 * 11 various tc7_ codes including, tc16_ codes.
152 *
153 *
154 * 000 heap address
155 * 001 glocs/structs
156 * 010 integer
157 * 011 closure
158 * 100 immediates
159 * 101 tc7_
160 * 110 integer
161 * 111 tc7_
162 *
163 *
164 * 100 --- IMMEDIATES
165 *
166 * Looking at the seven final bits of an immediate:
167 *
168 * 0000-100 short instruction
169 * 0001-100 short instruction
170 * 0010-100 short instruction
171 * 0011-100 short instruction
172 * 0100-100 short instruction
173 * 0101-100 short instruction
174 * 0110-100 various immediates and long instructions
175 * 0111-100 short instruction
176 * 1000-100 short instruction
177 * 1001-100 short instruction
178 * 1010-100 short instruction
179 * 1011-100 short instruction
180 * 1100-100 short instruction
181 * 1101-100 short instruction
182 * 1110-100 immediate characters
183 * 1111-100 ilocs
184 *
185 * Some of the 0110100 immediates are long instructions (they dispatch
186 * in two steps compared to one step for a short instruction).
187 * The two steps are, (1) dispatch on 7 bits to the long instruction
188 * handler, (2) dispatch on 7 additional bits.
189 *
190 * One way to think of it is that there are 128 short instructions,
191 * with the 13 immediates above being some of the most interesting.
192 *
193 * Also noteworthy are the groups of 16 7-bit instructions implied by
194 * some of the 3-bit tags. For example, closure references consist
195 * of an 8-bit aligned address tagged with 011. There are 16 identical 7-bit
196 * instructions, all ending 011, which are invoked by evaluating closures.
197 *
198 * In other words, if you hand the evaluator a closure, the evaluator
199 * treats the closure as a graph of virtual machine instructions.
200 * A closure is a pair with a pointer to the body of the procedure
201 * in the CDR and a pointer to the environment of the closure in the CAR.
202 * The environment pointer is tagged 011 which implies that the least
203 * significant 7 bits of the environment pointer also happen to be
204 * a virtual machine instruction we could call "SELF" (for self-evaluating
205 * object).
206 *
207 * A less trivial example are the 16 instructions ending 000. If those
208 * bits tag the CAR of a pair, then evidently the pair is an ordinary
209 * cons pair and should be evaluated as a procedure application. The sixteen,
210 * 7-bit 000 instructions are all "NORMAL-APPLY" (Things get trickier.
211 * For example, if the CAR of a procedure application is a symbol, the NORMAL-APPLY
212 * instruction will, as a side effect, overwrite that CAR with a new instruction
213 * that contains a cached address for the variable named by the symbol.)
214 *
215 * Here is a summary of tags in the CAR of a non-immediate:
216 *
217 * HEAP CELL: G=gc_mark; 1 during mark, 0 other times.
218 *
219 * cons ..........SCM car..............0 ...........SCM cdr.............G
220 * gloc ..........SCM vcell..........001 ...........SCM cdr.............G
221 * struct ..........void * type........001 ...........void * data.........G
222 * closure ..........SCM code...........011 ...........SCM env.............G
223 * tc7 ......24.bits of data...Gxxxx1S1 ..........void *data............
224 *
225 *
226 *
227 * 101 & 111 --- tc7_ types
228 *
229 * tc7_tags are 7 bit tags ending in 1x1. These tags
230 * occur only in the CAR of heap cells, and have the
231 * handy property that all bits of the CAR above the
232 * bottom eight can be used to store some data, thus
233 * saving a word in the body itself. Thus, we use them
234 * for strings and vectors (among other things).
235 *
236 * TYP7(X) returns bits 0...6 of CELL_TYPE (X)
237 *
238 * Sometimes we choose the bottom seven bits carefully,
239 * so that the 2-valued bit (called S bit) can be masked
240 * off to reveal a common type.
241 *
242 * TYP7S(X) returns TYP7, but masking out the option bit S.
243 *
244 * For example, all strings have 0010 in the 'xxxx' bits
245 * in the diagram above, the S bit says whether it's a
246 * substring.
247 *
248 * for example:
249 * S
250 * scm_tc7_string = G0010101
251 * scm_tc7_substring = G0010111
252 *
253 * TYP7S turns both string tags into tc7_string; thus,
254 * testing TYP7S against tc7_string is a quick way to
255 * test for any kind of string, shared or unshared.
256 *
257 * Some TC7 types are subdivided into 256 subtypes giving
258 * rise to the macros:
259 *
260 * TYP16
261 * TYP16S
262 *
263 * TYP16S functions similarly wrt to TYP16 as TYP7S to TYP7,
264 * but a different option bit is used (bit 2 for TYP7S,
265 * bit 8 for TYP16S).
266 * */
267
268
269
270 \f
271 /* {Non-immediate values.}
272 *
273 * If X is non-immediate, it is necessary to look at SCM_CAR (X) to
274 * figure out Xs type. X may be a cons pair, in which case the value
275 * SCM_CAR (x) will be either an immediate or non-immediate value. X
276 * may be something other than a cons pair, in which case the value
277 * SCM_CAR (x) will be a non-object value.
278 *
279 * All immediates and non-immediates have a 0 in bit 0. We
280 * additionally preserve the invariant that all non-object values
281 * stored in the SCM_CAR of a non-immediate object have a 1 in bit 1:
282 */
283
284 #define SCM_CONSP(x) (!SCM_IMP (x) && ((1 & SCM_CELL_TYPE (x)) == 0))
285 #define SCM_NCONSP(x) (!SCM_CONSP (x))
286
287
288 /* SCM_ECONSP should be used instead of SCM_CONSP at places where GLOCS
289 * can be expected to occur.
290 */
291 #define SCM_ECONSP(x) \
292 (!SCM_IMP (x) \
293 && (SCM_CONSP (x) \
294 || (SCM_TYP3 (x) == 1 \
295 && (SCM_STRUCT_VTABLE_DATA (x)[scm_vtable_index_vcell] != 0))))
296 #define SCM_NECONSP(x) (!SCM_ECONSP (x))
297
298 \f
299
300 #define SCM_CELLP(x) (((sizeof (scm_cell) - 1) & SCM_UNPACK (x)) == 0)
301 #define SCM_NCELLP(x) (!SCM_CELLP (x))
302
303 /* See numbers.h for macros relating to immediate integers.
304 */
305
306 #define SCM_ITAG3(x) (7 & SCM_UNPACK (x))
307 #define SCM_TYP3(x) (7 & SCM_CELL_TYPE (x))
308 #define scm_tc3_cons 0
309 #define scm_tc3_cons_gloc 1
310 #define scm_tc3_int_1 2
311 #define scm_tc3_closure 3
312 #define scm_tc3_imm24 4
313 #define scm_tc3_tc7_1 5
314 #define scm_tc3_int_2 6
315 #define scm_tc3_tc7_2 7
316
317
318 /*
319 * Do not change the three bit tags.
320 */
321
322
323 #define SCM_ITAG7(x) (127 & SCM_UNPACK (x))
324 #define SCM_TYP7(x) (0x7f & SCM_CELL_TYPE (x))
325 #define SCM_TYP7S(x) ((0x7f & ~2) & SCM_CELL_TYPE (x))
326
327
328 #define SCM_TYP16(x) (0xffff & SCM_CELL_TYPE (x))
329 #define SCM_TYP16S(x) (0xfeff & SCM_CELL_TYPE (x))
330
331 #define SCM_TYP16_PREDICATE(tag,x) (SCM_NIMP (x) && SCM_TYP16 (x) == (tag))
332
333 \f
334
335 #define scm_tc7_symbol 5
336 #define scm_tc7_variable 7
337
338 /* couple */
339 #define scm_tc7_vector 13
340 #define scm_tc7_wvect 15
341
342 /* couple */
343 #define scm_tc7_string 21
344 #define scm_tc7_substring 23
345
346 /* Many of the following should be turned
347 * into structs or smobs. We need back some
348 * of these 7 bit tags!
349 */
350 #define scm_tc7_pws 31
351
352 #ifdef HAVE_ARRAYS
353 #define scm_tc7_llvect 29
354 #define scm_tc7_uvect 37
355 /* free 39 */
356 #define scm_tc7_fvect 45
357 #define scm_tc7_dvect 47
358 #define scm_tc7_cvect 53
359 #define scm_tc7_svect 55
360 #define scm_tc7_bvect 71
361 #define scm_tc7_byvect 77
362 #define scm_tc7_ivect 79
363 #endif
364
365 /* free 61 */
366 #define scm_tc7_cclo 63
367 #define scm_tc7_rpsubr 69
368 #define scm_tc7_subr_0 85
369 #define scm_tc7_subr_1 87
370 #define scm_tc7_cxr 93
371 #define scm_tc7_subr_3 95
372 #define scm_tc7_subr_2 101
373 #define scm_tc7_asubr 103
374 #define scm_tc7_subr_1o 109
375 #define scm_tc7_subr_2o 111
376 #define scm_tc7_lsubr_2 117
377 #define scm_tc7_lsubr 119
378
379
380 /* There are 256 port subtypes.
381 */
382 #define scm_tc7_port 125
383
384
385 /* There are 256 smob subtypes. Here are the first four.
386 */
387
388 #define scm_tc7_smob 127 /* DO NOT CHANGE [**] */
389
390 /* [**] If you change scm_tc7_smob, you must also change
391 * the places it is hard coded in this file and possibly others.
392 */
393
394
395 /* scm_tc_free_cell is also the 0th smob type. We place this
396 * in free cells to tell the conservative marker not to trace it.
397 */
398 #define scm_tc_free_cell (scm_tc7_smob + 0 * 256L)
399
400 /* Smob type 1 to 3 (note the dependency on the predicate SCM_NUMP)
401 */
402 #define scm_tc16_big (scm_tc7_smob + 1 * 256L)
403 #define scm_tc16_real (scm_tc7_smob + 2 * 256L)
404 #define scm_tc16_complex (scm_tc7_smob + 3 * 256L)
405
406 \f
407 /* {Immediate Values}
408 */
409
410 enum scm_tags
411 {
412 scm_tc8_char = 0xf4,
413 scm_tc8_iloc = 0xfc
414 };
415
416 #define SCM_ITAG8(X) (SCM_UNPACK (X) & 0xff)
417 #define SCM_MAKE_ITAG8(X, TAG) SCM_PACK (((X) << 8) + TAG)
418 #define SCM_ITAG8_DATA(X) (SCM_UNPACK (X) >> 8)
419
420
421 \f
422 /* Immediate Symbols, Special Symbols, Flags (various constants).
423 */
424
425 /* SCM_ISYMP tests for ISPCSYM and ISYM */
426 #define SCM_ISYMP(n) ((0x187 & SCM_UNPACK (n)) == 4)
427
428 /* SCM_IFLAGP tests for ISPCSYM, ISYM and IFLAG */
429 #define SCM_IFLAGP(n) ((0x87 & SCM_UNPACK (n)) == 4)
430 #define SCM_ISYMNUM(n) (SCM_UNPACK (n) >> 9)
431 #define SCM_ISYMCHARS(n) (scm_isymnames[SCM_ISYMNUM (n)])
432 #define SCM_MAKSPCSYM(n) SCM_PACK (((n) << 9) + ((n) << 3) + 4L)
433 #define SCM_MAKISYM(n) SCM_PACK (((n) << 9) + 0x74L)
434 #define SCM_MAKIFLAG(n) SCM_PACK (((n) << 9) + 0x174L)
435
436 extern char *scm_isymnames[]; /* defined in print.c */
437
438 /* This table must agree with the declarations
439 * in repl.c: {Names of immediate symbols}.
440 *
441 * These are used only in eval but their values
442 * have to be allocated here.
443 *
444 */
445
446 #define SCM_IM_AND SCM_MAKSPCSYM (0)
447 #define SCM_IM_BEGIN SCM_MAKSPCSYM (1)
448 #define SCM_IM_CASE SCM_MAKSPCSYM (2)
449 #define SCM_IM_COND SCM_MAKSPCSYM (3)
450 #define SCM_IM_DO SCM_MAKSPCSYM (4)
451 #define SCM_IM_IF SCM_MAKSPCSYM (5)
452 #define SCM_IM_LAMBDA SCM_MAKSPCSYM (6)
453 #define SCM_IM_LET SCM_MAKSPCSYM (7)
454 #define SCM_IM_LETSTAR SCM_MAKSPCSYM (8)
455 #define SCM_IM_LETREC SCM_MAKSPCSYM (9)
456 #define SCM_IM_OR SCM_MAKSPCSYM (10)
457 #define SCM_IM_QUOTE SCM_MAKSPCSYM (11)
458 #define SCM_IM_SET_X SCM_MAKSPCSYM (12)
459 #define SCM_IM_DEFINE SCM_MAKSPCSYM (13)
460 #define SCM_IM_APPLY SCM_MAKISYM (14)
461 #define SCM_IM_CONT SCM_MAKISYM (15)
462 #define SCM_BOOL_F SCM_MAKIFLAG (16)
463 #define SCM_BOOL_T SCM_MAKIFLAG (17)
464 #define SCM_UNDEFINED SCM_MAKIFLAG (18)
465 #define SCM_EOF_VAL SCM_MAKIFLAG (19)
466 #define SCM_EOL SCM_MAKIFLAG (20)
467 #define SCM_UNSPECIFIED SCM_MAKIFLAG (21)
468 #define SCM_IM_DISPATCH SCM_MAKISYM (22)
469 #define SCM_IM_SLOT_REF SCM_MAKISYM (23)
470 #define SCM_IM_SLOT_SET_X SCM_MAKISYM (24)
471
472 /* Multi-language support */
473
474 #define SCM_IM_NIL_COND SCM_MAKISYM (25)
475 #define SCM_IM_NIL_IFY SCM_MAKISYM (26)
476 #define SCM_IM_T_IFY SCM_MAKISYM (27)
477 #define SCM_IM_0_COND SCM_MAKISYM (28)
478 #define SCM_IM_0_IFY SCM_MAKISYM (29)
479 #define SCM_IM_1_IFY SCM_MAKISYM (30)
480 #define SCM_IM_BIND SCM_MAKISYM (31)
481
482 #define SCM_IM_DELAY SCM_MAKISYM (32)
483 #define SCM_IM_CALL_WITH_VALUES SCM_MAKISYM (33)
484
485 /* When a variable is unbound this is marked by the SCM_UNDEFINED
486 * value. The following is an unbound value which can be handled on
487 * the Scheme level, i.e., it can be stored in and retrieved from a
488 * Scheme variable. This value is only intended to mark an unbound
489 * slot in GOOPS. It is needed now, but we should probably rewrite
490 * the code which handles this value in C so that SCM_UNDEFINED can be
491 * used instead. It is not ideal to let this kind of unique and
492 * strange values loose on the Scheme level.
493 */
494 #define SCM_UNBOUND SCM_MAKIFLAG (34)
495
496 #define SCM_UNBNDP(x) (SCM_EQ_P ((x), SCM_UNDEFINED))
497
498 \f
499
500 /* Dispatching aids: */
501
502
503 /* For cons pairs with immediate values in the CAR
504 */
505
506 #define scm_tcs_cons_imcar 2:case 4:case 6:case 10:\
507 case 12:case 14:case 18:case 20:\
508 case 22:case 26:case 28:case 30:\
509 case 34:case 36:case 38:case 42:\
510 case 44:case 46:case 50:case 52:\
511 case 54:case 58:case 60:case 62:\
512 case 66:case 68:case 70:case 74:\
513 case 76:case 78:case 82:case 84:\
514 case 86:case 90:case 92:case 94:\
515 case 98:case 100:case 102:case 106:\
516 case 108:case 110:case 114:case 116:\
517 case 118:case 122:case 124:case 126
518
519 /* For cons pairs with non-immediate values in the SCM_CAR
520 */
521 #define scm_tcs_cons_nimcar 0:case 8:case 16:case 24:\
522 case 32:case 40:case 48:case 56:\
523 case 64:case 72:case 80:case 88:\
524 case 96:case 104:case 112:case 120
525
526 /* A CONS_GLOC occurs in code. It's CAR is a pointer to the
527 * CDR of a variable. The low order bits of the CAR are 001.
528 * The CDR of the gloc is the code continuation.
529 */
530 #define scm_tcs_cons_gloc 1:case 9:case 17:case 25:\
531 case 33:case 41:case 49:case 57:\
532 case 65:case 73:case 81:case 89:\
533 case 97:case 105:case 113:case 121
534
535 #define scm_tcs_closures 3:case 11:case 19:case 27:\
536 case 35:case 43:case 51:case 59:\
537 case 67:case 75:case 83:case 91:\
538 case 99:case 107:case 115:case 123
539
540 #define scm_tcs_subrs scm_tc7_asubr:case scm_tc7_subr_0:case scm_tc7_subr_1:case scm_tc7_cxr:\
541 case scm_tc7_subr_3:case scm_tc7_subr_2:case scm_tc7_rpsubr:case scm_tc7_subr_1o:\
542 case scm_tc7_subr_2o:case scm_tc7_lsubr_2:case scm_tc7_lsubr
543
544 \f
545
546 #if (SCM_DEBUG_DEPRECATED == 0)
547
548 #define SCM_SLOPPY_CONSP(x) ((1 & SCM_CELL_TYPE (x)) == 0)
549 #define SCM_SLOPPY_NCONSP(x) (!SCM_SLOPPY_CONSP(x))
550
551 #define scm_tc7_ssymbol scm_tc7_symbol
552 #define scm_tc7_msymbol scm_tc7_symbol
553 #define scm_tcs_symbols scm_tc7_symbol
554
555 #endif /* SCM_DEBUG_DEPRECATED == 0 */
556
557 #endif /* SCM_TAGS_H */
558
559 /*
560 Local Variables:
561 c-file-style: "gnu"
562 End:
563 */