forked from ocaml-flambda/flambda-backend
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathbigarray.c
1443 lines (1311 loc) · 43.9 KB
/
bigarray.c
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/**************************************************************************/
/* */
/* OCaml */
/* */
/* Manuel Serrano and Xavier Leroy, INRIA Rocquencourt */
/* */
/* Copyright 2000 Institut National de Recherche en Informatique et */
/* en Automatique. */
/* */
/* All rights reserved. This file is distributed under the terms of */
/* the GNU Lesser General Public License version 2.1, with the */
/* special exception on linking described in the file LICENSE. */
/* */
/**************************************************************************/
#define CAML_INTERNALS
#include <stddef.h>
#include <stdarg.h>
#include <string.h>
#include "caml/alloc.h"
#include "caml/bigarray.h"
#include "caml/custom.h"
#include "caml/fail.h"
#include "caml/intext.h"
#include "caml/hash.h"
#include "caml/memory.h"
#include "caml/mlvalues.h"
#include "caml/signals.h"
#define int8 caml_ba_int8
#define uint8 caml_ba_uint8
#define int16 caml_ba_int16
#define uint16 caml_ba_uint16
/* Half-precision floating point numbers */
#if defined(__GNUC__) && defined(__aarch64__)
union float16_bits { uint16_t i; _Float16 f; };
Caml_inline float caml_float16_to_float(uint16 d)
{
union float16_bits u;
u.i = d; return u.f;
}
Caml_inline uint16 caml_float_to_float16(float d)
{
union float16_bits u;
u.f = d; return u.i;
}
#elif defined(__GNUC__) && defined(__F16C__)
#include <immintrin.h>
Caml_inline float caml_float16_to_float(uint16 d)
{ return _cvtsh_ss(d); }
Caml_inline uint16 caml_float_to_float16(float d)
{ return _cvtss_sh(d, (_MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC)); }
#else
union float_bits {
uint32_t i;
float f;
};
/*
* half_to_float_fast5
* https://gist.github.com/rygorous/2144712
*/
static float caml_float16_to_float(uint16 d)
{
static const union float_bits magic = { (254 - 15) << 23 };
static const union float_bits was_infnan = { (127 + 16) << 23 };
union float_bits o;
o.i = (d & 0x7fff) << 13; /* exponent/mantissa bits */
o.f *= magic.f; /* exponent adjust */
if (o.f >= was_infnan.f) /* make sure Inf/NaN survive */
o.i |= 255 << 23;
o.i |= (d & 0x8000) << 16; /* sign bit */
return o.f;
}
/*
* float_to_half_fast3_rtne
* https://gist.github.com/rygorous/2156668
*/
static uint16 caml_float_to_float16(float d)
{
static const union float_bits f32infty = { 255 << 23 };
static const union float_bits f16max = { (127 + 16) << 23 };
static const union float_bits denorm_magic =
{ ((127 - 15) + (23 - 10) + 1) << 23 };
static const uint32_t sign_mask = 0x80000000u;
union float_bits f;
uint16 o = 0;
uint32_t sign;
f.f = d;
sign = f.i & sign_mask;
f.i ^= sign;
// NOTE all the integer compares in this function can be safely
// compiled into signed compares since all operands are below
// 0x80000000. Important if you want fast straight SSE2 code
// (since there's no unsigned PCMPGTD).
if (f.i >= f16max.i) // result is Inf or NaN (all exponent bits set)
o = (f.i > f32infty.i) ? 0x7e00 : 0x7c00; // NaN->qNaN and Inf->Inf
else // (De)normalized number or zero
{
if (f.i < (113 << 23)) // resulting FP16 is subnormal or zero
{
// use a magic value to align our 10 mantissa bits at the bottom of
// the float. as long as FP addition is round-to-nearest-even this
// just works.
f.f += denorm_magic.f;
// and one integer subtract of the bias later, we have our final float!
o = f.i - denorm_magic.i;
}
else
{
uint32_t mant_odd = (f.i >> 13) & 1; // resulting mantissa is odd
// update exponent, rounding bias part 1
f.i += ((15 - 127) << 23) + 0xfff;
// rounding bias part 2
f.i += mant_odd;
// take the bits!
o = f.i >> 13;
}
}
o |= sign >> 16;
return o;
}
#endif /* defined(__GNUC__) && defined(__F16C__) */
CAMLexport double caml_double_of_float16(intnat x)
{
return (double) caml_float16_to_float((uint16) x);
}
CAMLexport intnat caml_float16_of_double(double x)
{
return (intnat) caml_float_to_float16((float) x);
}
Caml_inline uint32_t caml_hash_mix_float16(uint32_t hash, uint16 d)
{
/* Normalize NaNs */
if ((d & 0x7c00) == 0x7c00 && (d & 0x03ff) != 0) {
d = 0x7c01;
}
/* Normalize -0 into +0 */
else if (d == 0x8000) {
d = 0;
}
return caml_hash_mix_uint32(hash, d);
}
/* Compute the number of elements of a big array */
CAMLexport uintnat caml_ba_num_elts(struct caml_ba_array * b)
{
uintnat num_elts;
int i;
num_elts = 1;
for (i = 0; i < b->num_dims; i++) num_elts = num_elts * b->dim[i];
return num_elts;
}
/* Size in bytes of a bigarray element, indexed by bigarray kind */
CAMLexport int caml_ba_element_size[] =
{ 4 /*FLOAT32*/, 8 /*FLOAT64*/,
1 /*SINT8*/, 1 /*UINT8*/,
2 /*SINT16*/, 2 /*UINT16*/,
4 /*INT32*/, 8 /*INT64*/,
sizeof(value) /*CAML_INT*/, sizeof(value) /*NATIVE_INT*/,
8 /*COMPLEX32*/, 16 /*COMPLEX64*/,
1 /*CHAR*/, 2 /*FLOAT16*/
};
/* Compute the number of bytes for the elements of a big array */
CAMLexport uintnat caml_ba_byte_size(struct caml_ba_array * b)
{
return caml_ba_num_elts(b)
* caml_ba_element_size[b->flags & CAML_BA_KIND_MASK];
}
/* Operation table for bigarrays */
CAMLexport struct custom_operations caml_ba_ops = {
"_bigarr02",
caml_ba_finalize,
caml_ba_compare,
caml_ba_hash,
caml_ba_serialize,
caml_ba_deserialize,
custom_compare_ext_default,
custom_fixed_length_default
};
/* Allocation of a big array */
/* [caml_ba_alloc] will allocate a new bigarray object in the heap.
If [data] is NULL, the memory for the contents is also allocated
(with [malloc]) by [caml_ba_alloc].
[data] cannot point into the OCaml heap.
[dim] may point into an object in the OCaml heap.
*/
CAMLexport value
caml_ba_alloc(int flags, int num_dims, void * data, intnat * dim)
{
uintnat num_elts, asize, size;
int i;
value res;
struct caml_ba_array * b;
intnat dimcopy[CAML_BA_MAX_NUM_DIMS];
CAMLassert(num_dims >= 0 && num_dims <= CAML_BA_MAX_NUM_DIMS);
CAMLassert((flags & CAML_BA_KIND_MASK) < CAML_BA_FIRST_UNIMPLEMENTED_KIND);
for (i = 0; i < num_dims; i++) dimcopy[i] = dim[i];
size = 0;
if (data == NULL) {
num_elts = 1;
for (i = 0; i < num_dims; i++) {
if (caml_umul_overflow(num_elts, dimcopy[i], &num_elts))
caml_raise_out_of_memory();
}
if (caml_umul_overflow(num_elts,
caml_ba_element_size[flags & CAML_BA_KIND_MASK],
&size))
caml_raise_out_of_memory();
data = malloc(size);
if (data == NULL && size != 0) caml_raise_out_of_memory();
flags |= CAML_BA_MANAGED;
}
asize = SIZEOF_BA_ARRAY + num_dims * sizeof(intnat);
res = caml_alloc_custom_mem(&caml_ba_ops, asize, size);
b = Caml_ba_array_val(res);
b->data = data;
b->num_dims = num_dims;
b->flags = flags;
b->proxy = NULL;
for (i = 0; i < num_dims; i++) b->dim[i] = dimcopy[i];
return res;
}
/* Same as caml_ba_alloc, but dimensions are passed as a list of
arguments */
CAMLexport value caml_ba_alloc_dims(int flags, int num_dims, void * data, ...)
{
va_list ap;
intnat dim[CAML_BA_MAX_NUM_DIMS];
int i;
value res;
CAMLassert(num_dims <= CAML_BA_MAX_NUM_DIMS);
va_start(ap, data);
for (i = 0; i < num_dims; i++) dim[i] = va_arg(ap, intnat);
va_end(ap);
res = caml_ba_alloc(flags, num_dims, data, dim);
return res;
}
/* Finalization of a big array */
CAMLexport void caml_ba_finalize(value v)
{
struct caml_ba_array * b = Caml_ba_array_val(v);
switch (b->flags & CAML_BA_MANAGED_MASK) {
case CAML_BA_EXTERNAL:
break;
case CAML_BA_MANAGED:
if (b->proxy == NULL) {
free(b->data);
} else {
if (-- b->proxy->refcount == 0) {
free(b->proxy->data);
free(b->proxy);
}
}
break;
case CAML_BA_MAPPED_FILE:
/* Bigarrays for mapped files use a different finalization method */
/* fallthrough */
default:
CAMLassert(0);
}
}
/* Comparison of two big arrays */
CAMLexport int caml_ba_compare(value v1, value v2)
{
struct caml_ba_array * b1 = Caml_ba_array_val(v1);
struct caml_ba_array * b2 = Caml_ba_array_val(v2);
uintnat n, num_elts;
intnat flags1, flags2;
int i;
/* Compare kind & layout in case the arguments are of different types */
flags1 = b1->flags & (CAML_BA_KIND_MASK | CAML_BA_LAYOUT_MASK);
flags2 = b2->flags & (CAML_BA_KIND_MASK | CAML_BA_LAYOUT_MASK);
if (flags1 != flags2) return flags2 - flags1;
/* Compare number of dimensions */
if (b1->num_dims != b2->num_dims) return b2->num_dims - b1->num_dims;
/* Same number of dimensions: compare dimensions lexicographically */
for (i = 0; i < b1->num_dims; i++) {
intnat d1 = b1->dim[i];
intnat d2 = b2->dim[i];
if (d1 != d2) return d1 < d2 ? -1 : 1;
}
/* Same dimensions: compare contents lexicographically */
num_elts = caml_ba_num_elts(b1);
#define DO_INTEGER_COMPARISON(type) \
{ type * p1 = b1->data; type * p2 = b2->data; \
for (n = 0; n < num_elts; n++) { \
type e1 = *p1++; type e2 = *p2++; \
if (e1 < e2) return -1; \
if (e1 > e2) return 1; \
} \
return 0; \
}
#define DO_GENERIC_UNORDERED_COMPARISON(ptype, etype, conv) \
{ ptype * p1 = b1->data; ptype * p2 = b2->data; \
for (n = 0; n < num_elts; n++) { \
etype e1 = conv(*p1++); etype e2 = conv(*p2++); \
if (e1 < e2) return -1; \
if (e1 > e2) return 1; \
if (e1 != e2) { \
Caml_state->compare_unordered = 1; \
if (e1 == e1) return 1; \
if (e2 == e2) return -1; \
} \
} \
return 0; \
}
#define DO_FLOAT_COMPARISON(type) \
DO_GENERIC_UNORDERED_COMPARISON(type, type, )
switch (b1->flags & CAML_BA_KIND_MASK) {
case CAML_BA_FLOAT16:
DO_GENERIC_UNORDERED_COMPARISON(uint16, float, caml_float16_to_float);
case CAML_BA_COMPLEX32:
num_elts *= 2; /*fallthrough*/
case CAML_BA_FLOAT32:
DO_FLOAT_COMPARISON(float);
case CAML_BA_COMPLEX64:
num_elts *= 2; /*fallthrough*/
case CAML_BA_FLOAT64:
DO_FLOAT_COMPARISON(double);
case CAML_BA_CHAR:
DO_INTEGER_COMPARISON(caml_ba_uint8);
case CAML_BA_SINT8:
DO_INTEGER_COMPARISON(caml_ba_int8);
case CAML_BA_UINT8:
DO_INTEGER_COMPARISON(caml_ba_uint8);
case CAML_BA_SINT16:
DO_INTEGER_COMPARISON(caml_ba_int16);
case CAML_BA_UINT16:
DO_INTEGER_COMPARISON(caml_ba_uint16);
case CAML_BA_INT32:
DO_INTEGER_COMPARISON(int32_t);
case CAML_BA_INT64:
DO_INTEGER_COMPARISON(int64_t);
case CAML_BA_CAML_INT:
case CAML_BA_NATIVE_INT:
DO_INTEGER_COMPARISON(intnat);
default:
CAMLassert(0);
return 0; /* should not happen */
}
#undef DO_INTEGER_COMPARISON
#undef DO_FLOAT_COMPARISON
}
/* Hashing of a bigarray */
CAMLexport intnat caml_ba_hash(value v)
{
struct caml_ba_array * b = Caml_ba_array_val(v);
intnat num_elts, n;
uint32_t h, w;
int i;
num_elts = 1;
for (i = 0; i < b->num_dims; i++) num_elts = num_elts * b->dim[i];
h = 0;
switch (b->flags & CAML_BA_KIND_MASK) {
case CAML_BA_CHAR:
case CAML_BA_SINT8:
case CAML_BA_UINT8: {
caml_ba_uint8 * p = b->data;
if (num_elts > 256) num_elts = 256;
for (n = 0; n + 4 <= num_elts; n += 4, p += 4) {
w = p[0] | (p[1] << 8) | (p[2] << 16) | (p[3] << 24);
h = caml_hash_mix_uint32(h, w);
}
w = 0;
switch (num_elts & 3) {
case 3: w = p[2] << 16; /* fallthrough */
case 2: w |= p[1] << 8; /* fallthrough */
case 1: w |= p[0];
h = caml_hash_mix_uint32(h, w);
}
break;
}
case CAML_BA_SINT16:
case CAML_BA_UINT16: {
caml_ba_uint16 * p = b->data;
if (num_elts > 128) num_elts = 128;
for (n = 0; n + 2 <= num_elts; n += 2, p += 2) {
w = p[0] | (p[1] << 16);
h = caml_hash_mix_uint32(h, w);
}
if ((num_elts & 1) != 0)
h = caml_hash_mix_uint32(h, p[0]);
break;
}
case CAML_BA_INT32:
{
uint32_t * p = b->data;
if (num_elts > 64) num_elts = 64;
for (n = 0; n < num_elts; n++, p++) h = caml_hash_mix_uint32(h, *p);
break;
}
case CAML_BA_CAML_INT:
case CAML_BA_NATIVE_INT:
{
intnat * p = b->data;
if (num_elts > 64) num_elts = 64;
for (n = 0; n < num_elts; n++, p++) h = caml_hash_mix_intnat(h, *p);
break;
}
case CAML_BA_INT64:
{
int64_t * p = b->data;
if (num_elts > 32) num_elts = 32;
for (n = 0; n < num_elts; n++, p++) h = caml_hash_mix_int64(h, *p);
break;
}
case CAML_BA_FLOAT16:
{
uint16 * p = b->data;
if (num_elts > 128) num_elts = 128;
for (n = 0; n < num_elts; n++, p++) h = caml_hash_mix_float16(h, *p);
break;
}
case CAML_BA_COMPLEX32:
num_elts *= 2; /* fallthrough */
case CAML_BA_FLOAT32:
{
float * p = b->data;
if (num_elts > 64) num_elts = 64;
for (n = 0; n < num_elts; n++, p++) h = caml_hash_mix_float(h, *p);
break;
}
case CAML_BA_COMPLEX64:
num_elts *= 2; /* fallthrough */
case CAML_BA_FLOAT64:
{
double * p = b->data;
if (num_elts > 32) num_elts = 32;
for (n = 0; n < num_elts; n++, p++) h = caml_hash_mix_double(h, *p);
break;
}
}
return h;
}
static void caml_ba_serialize_longarray(void * data,
intnat num_elts,
intnat min_val, intnat max_val)
{
#ifdef ARCH_SIXTYFOUR
int overflow_32 = 0;
intnat * p, n;
for (n = 0, p = data; n < num_elts; n++, p++) {
if (*p < min_val || *p > max_val) { overflow_32 = 1; break; }
}
if (overflow_32) {
caml_serialize_int_1(1);
caml_serialize_block_8(data, num_elts);
} else {
caml_serialize_int_1(0);
for (n = 0, p = data; n < num_elts; n++, p++)
caml_serialize_int_4((int32_t) *p);
}
#else
caml_serialize_int_1(0);
caml_serialize_block_4(data, num_elts);
#endif
}
CAMLexport void caml_ba_serialize(value v,
uintnat * wsize_32,
uintnat * wsize_64)
{
struct caml_ba_array * b = Caml_ba_array_val(v);
intnat num_elts;
int i;
/* Serialize header information */
caml_serialize_int_4(b->num_dims);
caml_serialize_int_4(b->flags & (CAML_BA_KIND_MASK | CAML_BA_LAYOUT_MASK));
for (i = 0; i < b->num_dims; i++) {
intnat len = b->dim[i];
if (len < 0xffff) {
caml_serialize_int_2(len);
} else {
caml_serialize_int_2(0xffff);
caml_serialize_int_8(len);
}
}
/* Compute total number of elements */
num_elts = 1;
for (i = 0; i < b->num_dims; i++) num_elts = num_elts * b->dim[i];
/* Serialize elements */
switch (b->flags & CAML_BA_KIND_MASK) {
case CAML_BA_CHAR:
case CAML_BA_SINT8:
case CAML_BA_UINT8:
caml_serialize_block_1(b->data, num_elts); break;
case CAML_BA_SINT16:
case CAML_BA_UINT16:
case CAML_BA_FLOAT16:
caml_serialize_block_2(b->data, num_elts); break;
case CAML_BA_FLOAT32:
case CAML_BA_INT32:
caml_serialize_block_4(b->data, num_elts); break;
case CAML_BA_COMPLEX32:
caml_serialize_block_4(b->data, num_elts * 2); break;
case CAML_BA_FLOAT64:
case CAML_BA_INT64:
caml_serialize_block_8(b->data, num_elts); break;
case CAML_BA_COMPLEX64:
caml_serialize_block_8(b->data, num_elts * 2); break;
case CAML_BA_CAML_INT:
caml_ba_serialize_longarray(b->data, num_elts, -0x40000000, 0x3FFFFFFF);
break;
case CAML_BA_NATIVE_INT:
caml_ba_serialize_longarray(b->data, num_elts, -0x80000000, 0x7FFFFFFF);
break;
}
/* Compute required size in OCaml heap. Assumes struct caml_ba_array
is exactly 4 + num_dims words */
CAMLassert(SIZEOF_BA_ARRAY == 4 * sizeof(value));
*wsize_32 = (4 + b->num_dims) * 4;
*wsize_64 = (4 + b->num_dims) * 8;
}
static void caml_ba_deserialize_longarray(void * dest, intnat num_elts)
{
int sixty = caml_deserialize_uint_1();
#ifdef ARCH_SIXTYFOUR
if (sixty) {
caml_deserialize_block_8(dest, num_elts);
} else {
intnat * p, n;
for (n = 0, p = dest; n < num_elts; n++, p++)
*p = caml_deserialize_sint_4();
}
#else
if (sixty)
caml_deserialize_error("input_value: cannot read bigarray "
"with 64-bit OCaml ints");
caml_deserialize_block_4(dest, num_elts);
#endif
}
CAMLexport uintnat caml_ba_deserialize(void * dst)
{
struct caml_ba_array * b = dst;
int i;
uintnat num_elts, size;
/* Read back header information */
b->num_dims = caml_deserialize_uint_4();
if (b->num_dims < 0 || b->num_dims > CAML_BA_MAX_NUM_DIMS)
caml_deserialize_error("input_value: wrong number of bigarray dimensions");
b->flags = caml_deserialize_uint_4() | CAML_BA_MANAGED;
b->proxy = NULL;
for (i = 0; i < b->num_dims; i++) {
intnat len = caml_deserialize_uint_2();
if (len == 0xffff) len = caml_deserialize_uint_8();
b->dim[i] = len;
}
/* Compute total number of elements. Watch out for overflows (MPR#7765). */
num_elts = 1;
for (i = 0; i < b->num_dims; i++) {
if (caml_umul_overflow(num_elts, b->dim[i], &num_elts))
caml_deserialize_error("input_value: size overflow for bigarray");
}
/* Determine array size in bytes. Watch out for overflows (MPR#7765). */
if ((b->flags & CAML_BA_KIND_MASK) >= CAML_BA_FIRST_UNIMPLEMENTED_KIND)
caml_deserialize_error("input_value: bad bigarray kind");
if (caml_umul_overflow(num_elts,
caml_ba_element_size[b->flags & CAML_BA_KIND_MASK],
&size))
caml_deserialize_error("input_value: size overflow for bigarray");
/* Allocate room for data */
b->data = malloc(size);
if (b->data == NULL)
caml_deserialize_error("input_value: out of memory for bigarray");
/* Read data */
switch (b->flags & CAML_BA_KIND_MASK) {
case CAML_BA_CHAR:
case CAML_BA_SINT8:
case CAML_BA_UINT8:
caml_deserialize_block_1(b->data, num_elts); break;
case CAML_BA_SINT16:
case CAML_BA_UINT16:
case CAML_BA_FLOAT16:
caml_deserialize_block_2(b->data, num_elts); break;
case CAML_BA_FLOAT32:
case CAML_BA_INT32:
caml_deserialize_block_4(b->data, num_elts); break;
case CAML_BA_COMPLEX32:
caml_deserialize_block_4(b->data, num_elts * 2); break;
case CAML_BA_FLOAT64:
case CAML_BA_INT64:
caml_deserialize_block_8(b->data, num_elts); break;
case CAML_BA_COMPLEX64:
caml_deserialize_block_8(b->data, num_elts * 2); break;
case CAML_BA_CAML_INT:
case CAML_BA_NATIVE_INT:
caml_ba_deserialize_longarray(b->data, num_elts); break;
}
/* PR#5516: use C99's flexible array types if possible */
return SIZEOF_BA_ARRAY + b->num_dims * sizeof(intnat);
}
/* Allocate a bigarray from OCaml */
CAMLprim value caml_ba_create(value vkind, value vlayout, value vdim)
{
intnat dim[CAML_BA_MAX_NUM_DIMS];
mlsize_t num_dims;
int i, flags;
num_dims = Wosize_val(vdim);
/* here num_dims is unsigned (mlsize_t) so no need to check (num_dims >= 0) */
if (num_dims > CAML_BA_MAX_NUM_DIMS)
caml_invalid_argument("Bigarray.create: bad number of dimensions");
for (i = 0; i < num_dims; i++) {
dim[i] = Long_val(Field(vdim, i));
if (dim[i] < 0)
caml_invalid_argument("Bigarray.create: negative dimension");
}
flags = Caml_ba_kind_val(vkind) | Caml_ba_layout_val(vlayout);
return caml_ba_alloc(flags, num_dims, NULL, dim);
}
/* Given a big array and a vector of indices, check that the indices
are within the bounds and return the offset of the corresponding
array element in the data part of the array. */
intnat caml_ba_offset(struct caml_ba_array * b, intnat * index)
{
intnat offset;
int i;
offset = 0;
if ((b->flags & CAML_BA_LAYOUT_MASK) == CAML_BA_C_LAYOUT) {
/* C-style layout: row major, indices start at 0 */
for (i = 0; i < b->num_dims; i++) {
if ((uintnat) index[i] >= (uintnat) b->dim[i])
caml_array_bound_error();
offset = offset * b->dim[i] + index[i];
}
} else {
/* Fortran-style layout: column major, indices start at 1 */
for (i = b->num_dims - 1; i >= 0; i--) {
if ((uintnat) (index[i] - 1) >= (uintnat) b->dim[i])
caml_array_bound_error();
offset = offset * b->dim[i] + (index[i] - 1);
}
}
return offset;
}
/* Helper function to allocate a record of two double floats */
static value copy_two_doubles(double d0, double d1)
{
value res = caml_alloc_small(2 * Double_wosize, Double_array_tag);
Store_double_field(res, 0, d0);
Store_double_field(res, 1, d1);
return res;
}
/* Generic code to read from a big array */
value caml_ba_get_N(value vb, value * vind, int nind)
{
struct caml_ba_array * b = Caml_ba_array_val(vb);
intnat index[CAML_BA_MAX_NUM_DIMS];
int i;
intnat offset;
/* Check number of indices = number of dimensions of array
(maybe not necessary if ML typing guarantees this) */
if (nind != b->num_dims)
caml_invalid_argument("Bigarray.get: wrong number of indices");
/* Compute offset and check bounds */
for (i = 0; i < b->num_dims; i++) index[i] = Long_val(vind[i]);
offset = caml_ba_offset(b, index);
/* Perform read */
switch ((b->flags) & CAML_BA_KIND_MASK) {
default:
CAMLassert(0);
case CAML_BA_FLOAT16:
return caml_copy_double(
(double) caml_float16_to_float(((uint16 *) b->data)[offset]));
case CAML_BA_FLOAT32:
return caml_copy_double((double) ((float *) b->data)[offset]);
case CAML_BA_FLOAT64:
return caml_copy_double(((double *) b->data)[offset]);
case CAML_BA_SINT8:
return Val_int(((int8 *) b->data)[offset]);
case CAML_BA_UINT8:
return Val_int(((uint8 *) b->data)[offset]);
case CAML_BA_SINT16:
return Val_int(((int16 *) b->data)[offset]);
case CAML_BA_UINT16:
return Val_int(((uint16 *) b->data)[offset]);
case CAML_BA_INT32:
return caml_copy_int32(((int32_t *) b->data)[offset]);
case CAML_BA_INT64:
return caml_copy_int64(((int64_t *) b->data)[offset]);
case CAML_BA_NATIVE_INT:
return caml_copy_nativeint(((intnat *) b->data)[offset]);
case CAML_BA_CAML_INT:
return Val_long(((intnat *) b->data)[offset]);
case CAML_BA_COMPLEX32:
{ float * p = ((float *) b->data) + offset * 2;
return copy_two_doubles((double) p[0], (double) p[1]); }
case CAML_BA_COMPLEX64:
{ double * p = ((double *) b->data) + offset * 2;
return copy_two_doubles(p[0], p[1]); }
case CAML_BA_CHAR:
return Val_int(((unsigned char *) b->data)[offset]);
}
}
CAMLprim value caml_ba_get_1(value vb, value vind1)
{
return caml_ba_get_N(vb, &vind1, 1);
}
CAMLprim value caml_ba_get_2(value vb, value vind1, value vind2)
{
value vind[2];
vind[0] = vind1; vind[1] = vind2;
return caml_ba_get_N(vb, vind, 2);
}
CAMLprim value caml_ba_get_3(value vb, value vind1, value vind2, value vind3)
{
value vind[3];
vind[0] = vind1; vind[1] = vind2; vind[2] = vind3;
return caml_ba_get_N(vb, vind, 3);
}
CAMLprim value caml_ba_get_generic(value vb, value vind)
{
return caml_ba_get_N(vb, &Field(vind, 0), Wosize_val(vind));
}
CAMLprim value caml_ba_uint8_get16(value vb, value vind)
{
intnat res;
unsigned char b1, b2;
intnat idx = Long_val(vind);
struct caml_ba_array * b = Caml_ba_array_val(vb);
if (idx < 0 || idx >= b->dim[0] - 1) caml_array_bound_error();
b1 = ((unsigned char*) b->data)[idx];
b2 = ((unsigned char*) b->data)[idx+1];
#ifdef ARCH_BIG_ENDIAN
res = b1 << 8 | b2;
#else
res = b2 << 8 | b1;
#endif
return Val_int(res);
}
CAMLprim value caml_ba_uint8_get32(value vb, value vind)
{
uint32_t res;
unsigned char b1, b2, b3, b4;
intnat idx = Long_val(vind);
struct caml_ba_array * b = Caml_ba_array_val(vb);
if (idx < 0 || idx >= b->dim[0] - 3) caml_array_bound_error();
b1 = ((unsigned char*) b->data)[idx];
b2 = ((unsigned char*) b->data)[idx+1];
b3 = ((unsigned char*) b->data)[idx+2];
b4 = ((unsigned char*) b->data)[idx+3];
#ifdef ARCH_BIG_ENDIAN
res = b1 << 24 | b2 << 16 | b3 << 8 | b4;
#else
res = b4 << 24 | b3 << 16 | b2 << 8 | b1;
#endif
return caml_copy_int32(res);
}
CAMLprim value caml_ba_uint8_get64(value vb, value vind)
{
uint64_t res;
unsigned char b1, b2, b3, b4, b5, b6, b7, b8;
intnat idx = Long_val(vind);
struct caml_ba_array * b = Caml_ba_array_val(vb);
if (idx < 0 || idx >= b->dim[0] - 7) caml_array_bound_error();
b1 = ((unsigned char*) b->data)[idx];
b2 = ((unsigned char*) b->data)[idx+1];
b3 = ((unsigned char*) b->data)[idx+2];
b4 = ((unsigned char*) b->data)[idx+3];
b5 = ((unsigned char*) b->data)[idx+4];
b6 = ((unsigned char*) b->data)[idx+5];
b7 = ((unsigned char*) b->data)[idx+6];
b8 = ((unsigned char*) b->data)[idx+7];
#ifdef ARCH_BIG_ENDIAN
res = (uint64_t) b1 << 56 | (uint64_t) b2 << 48
| (uint64_t) b3 << 40 | (uint64_t) b4 << 32
| (uint64_t) b5 << 24 | (uint64_t) b6 << 16
| (uint64_t) b7 << 8 | (uint64_t) b8;
#else
res = (uint64_t) b8 << 56 | (uint64_t) b7 << 48
| (uint64_t) b6 << 40 | (uint64_t) b5 << 32
| (uint64_t) b4 << 24 | (uint64_t) b3 << 16
| (uint64_t) b2 << 8 | (uint64_t) b1;
#endif
return caml_copy_int64(res);
}
/* Generic write to a big array */
static value caml_ba_set_aux(value vb, value * vind, intnat nind, value newval)
{
struct caml_ba_array * b = Caml_ba_array_val(vb);
intnat index[CAML_BA_MAX_NUM_DIMS];
int i;
intnat offset;
/* Check number of indices = number of dimensions of array
(maybe not necessary if ML typing guarantees this) */
if (nind != b->num_dims)
caml_invalid_argument("Bigarray.set: wrong number of indices");
/* Compute offset and check bounds */
for (i = 0; i < b->num_dims; i++) index[i] = Long_val(vind[i]);
offset = caml_ba_offset(b, index);
/* Perform write */
switch (b->flags & CAML_BA_KIND_MASK) {
default:
CAMLassert(0);
case CAML_BA_FLOAT16:
((uint16 *) b->data)[offset] =
caml_float_to_float16(Double_val(newval)); break;
case CAML_BA_FLOAT32:
((float *) b->data)[offset] = Double_val(newval); break;
case CAML_BA_FLOAT64:
((double *) b->data)[offset] = Double_val(newval); break;
case CAML_BA_CHAR:
case CAML_BA_SINT8:
case CAML_BA_UINT8:
((int8 *) b->data)[offset] = Int_val(newval); break;
case CAML_BA_SINT16:
case CAML_BA_UINT16:
((int16 *) b->data)[offset] = Int_val(newval); break;
case CAML_BA_INT32:
((int32_t *) b->data)[offset] = Int32_val(newval); break;
case CAML_BA_INT64:
((int64_t *) b->data)[offset] = Int64_val(newval); break;
case CAML_BA_NATIVE_INT:
((intnat *) b->data)[offset] = Nativeint_val(newval); break;
case CAML_BA_CAML_INT:
((intnat *) b->data)[offset] = Long_val(newval); break;
case CAML_BA_COMPLEX32:
{ float * p = ((float *) b->data) + offset * 2;
p[0] = Double_field(newval, 0);
p[1] = Double_field(newval, 1);
break; }
case CAML_BA_COMPLEX64:
{ double * p = ((double *) b->data) + offset * 2;
p[0] = Double_field(newval, 0);
p[1] = Double_field(newval, 1);
break; }
}
return Val_unit;
}
CAMLprim value caml_ba_set_1(value vb, value vind1, value newval)
{
return caml_ba_set_aux(vb, &vind1, 1, newval);
}
CAMLprim value caml_ba_set_2(value vb, value vind1, value vind2, value newval)
{
value vind[2];
vind[0] = vind1; vind[1] = vind2;
return caml_ba_set_aux(vb, vind, 2, newval);
}
CAMLprim value caml_ba_set_3(value vb, value vind1, value vind2, value vind3,
value newval)
{
value vind[3];
vind[0] = vind1; vind[1] = vind2; vind[2] = vind3;
return caml_ba_set_aux(vb, vind, 3, newval);
}
value caml_ba_set_N(value vb, value * vind, int nargs)
{
return caml_ba_set_aux(vb, vind, nargs - 1, vind[nargs - 1]);
}
CAMLprim value caml_ba_set_generic(value vb, value vind, value newval)
{
return caml_ba_set_aux(vb, &Field(vind, 0), Wosize_val(vind), newval);
}
CAMLprim value caml_ba_uint8_set16(value vb, value vind, value newval)
{
unsigned char b1, b2;
intnat val;
intnat idx = Long_val(vind);
struct caml_ba_array * b = Caml_ba_array_val(vb);
if (idx < 0 || idx >= b->dim[0] - 1) caml_array_bound_error();
val = Long_val(newval);
#ifdef ARCH_BIG_ENDIAN
b1 = 0xFF & val >> 8;
b2 = 0xFF & val;
#else
b2 = 0xFF & val >> 8;
b1 = 0xFF & val;
#endif
((unsigned char*) b->data)[idx] = b1;
((unsigned char*) b->data)[idx+1] = b2;
return Val_unit;
}
CAMLprim value caml_ba_uint8_set32(value vb, value vind, value newval)
{
unsigned char b1, b2, b3, b4;
intnat idx = Long_val(vind);
intnat val;
struct caml_ba_array * b = Caml_ba_array_val(vb);
if (idx < 0 || idx >= b->dim[0] - 3) caml_array_bound_error();
val = Int32_val(newval);
#ifdef ARCH_BIG_ENDIAN
b1 = 0xFF & val >> 24;
b2 = 0xFF & val >> 16;
b3 = 0xFF & val >> 8;
b4 = 0xFF & val;
#else
b4 = 0xFF & val >> 24;
b3 = 0xFF & val >> 16;
b2 = 0xFF & val >> 8;
b1 = 0xFF & val;
#endif
((unsigned char*) b->data)[idx] = b1;
((unsigned char*) b->data)[idx+1] = b2;
((unsigned char*) b->data)[idx+2] = b3;
((unsigned char*) b->data)[idx+3] = b4;
return Val_unit;
}
CAMLprim value caml_ba_uint8_set64(value vb, value vind, value newval)
{
unsigned char b1, b2, b3, b4, b5, b6, b7, b8;
intnat idx = Long_val(vind);
int64_t val;
struct caml_ba_array * b = Caml_ba_array_val(vb);
if (idx < 0 || idx >= b->dim[0] - 7) caml_array_bound_error();
val = Int64_val(newval);
#ifdef ARCH_BIG_ENDIAN
b1 = 0xFF & val >> 56;
b2 = 0xFF & val >> 48;
b3 = 0xFF & val >> 40;
b4 = 0xFF & val >> 32;
b5 = 0xFF & val >> 24;
b6 = 0xFF & val >> 16;
b7 = 0xFF & val >> 8;