-
Notifications
You must be signed in to change notification settings - Fork 12
/
calc_addrs.cl
1268 lines (1063 loc) · 30.4 KB
/
calc_addrs.cl
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
/*
* Vanitygen ETH, vanity ETH address generator
* Copyright (C) 2018 <[email protected]>
*
* Vanitygen is free software: you can redistribute it and/or modify
* it under the terms of the GNU Affero General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* any later version.
*
* Vanitygen is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU Affero General Public License for more details.
*
* You should have received a copy of the GNU Affero General Public License
* along with Vanitygen. If not, see <http://www.gnu.org/licenses/>.
*/
/*
* This file contains an OpenCL kernel for performing certain parts of
* the EOS address calculation process.
*
* Kernel: ec_add_grid
*
* Inputs:
* - Row: Array of (sequential) EC points
* - Column: Array of column increment EC points (= rowsize * Pgenerator)
*
* Steps:
* - Compute P = Row[x] + Column[y]
* P is computed as numerator/denominator components Pxj, Pyj, Pz
* Final values are: Px = Pxj / (Pz^2), Py = Pyj / (Pz^3)
*
* The modular inverse of Pz is required to compute Px and Py, and
* can be computed more efficiently in large batches. This is done in
* the next kernel heap_invert.
*
* - Store Pxj, Pyj to intermediate point buffer
* - Store Pz to z_heap
*
* Outputs:
* - Intermediate point buffer
* - Denominator buffer (z_heap)
*
* -------------------------------
* Kernel: heap_invert
*
* Inputs:
* - Denominator buffer (z_heap)
* - N = Batch size (power of 2)
*
* Steps:
* - Compute the product tree for N values in the denominator buffer
* - Compute the modular inverse of the root of the product tree
* - Multiply down the tree to compute the modular inverse of each leaf
*
* Outputs:
* - Modular inverse denominator buffer (z_heap)
*
* -------------------------------
* Kernel: hash_ec_point_get
*
* Inputs:
* - Intermediate point buffer
* - Modular inverse denominator buffer (z_heap)
*
* Steps:
* - Compute Px = Pxj * (1/Pz)^2
* - Compute Py = Pyj * (1/Pz)^3
* - Compute H = SHA3 (Px, Py)
*
* Output:
* - Array of 20-byte address values
*
* -------------------------------
* Kernel: hash_ec_point_search_prefix
*
* Like hash_ec_point_get, but instead of storing the complete hash
* value to an output buffer, it searches a sorted list of ranges,
* and if a match is found, writes a flag to an output buffer.
*/
/* Byte-swapping and endianness */
#define bswap32(v) \
(((v) >> 24) | (((v) >> 8) & 0xff00) | \
(((v) << 8) & 0xff0000) | ((v) << 24))
static inline ulong bswap64(ulong v){
uchar *bb=(uchar *)&v;
uchar by;
#pragma unroll 4
for(int i=0;i<4;i++) {
by = bb[7-i];
bb[7-i] = bb[i];
bb[i] = by;
}
return *(ulong *)bb;
}
#if __ENDIAN_LITTLE__ != 1
#define load_le32(v) bswap32(v)
#define load_be32(v) (v)
#else
#define load_le32(v) (v)
#define load_be32(v) bswap32(v)
#endif
/*
* Loop unrolling macros
*
* In most cases, preprocessor unrolling works best.
* The exception is NVIDIA's compiler, which seems to take unreasonably
* long to compile a loop with a larger iteration count, or a loop with
* a body of >50 PTX instructions, with preprocessor unrolling.
* However, it does not seem to take as long with pragma unroll, and
* produces good output.
*/
/* Explicit loop unrolling */
#define unroll_5(a) do { a(0) a(1) a(2) a(3) a(4) } while (0)
#define unroll_8(a) do { a(0) a(1) a(2) a(3) a(4) a(5) a(6) a(7) } while (0)
#define unroll_1_7(a) do { a(1) a(2) a(3) a(4) a(5) a(6) a(7) } while (0)
#define unroll_7(a) do { a(0) a(1) a(2) a(3) a(4) a(5) a(6) } while (0)
#define unroll_7_0(a) do { a(7) a(6) a(5) a(4) a(3) a(2) a(1) a(0) } while (0)
#define unroll_7_1(a) do { a(7) a(6) a(5) a(4) a(3) a(2) a(1) } while (0)
#define unroll_16(a) do { \
a(0) a(1) a(2) a(3) a(4) a(5) a(6) a(7) \
a(8) a(9) a(10) a(11) a(12) a(13) a(14) a(15) \
} while (0)
#define unroll_64(a) do { \
a(0) a(1) a(2) a(3) a(4) a(5) a(6) a(7) \
a(8) a(9) a(10) a(11) a(12) a(13) a(14) a(15) \
a(16) a(17) a(18) a(19) a(20) a(21) a(22) a(23) \
a(24) a(25) a(26) a(27) a(28) a(29) a(30) a(31) \
a(32) a(33) a(34) a(35) a(36) a(37) a(38) a(39) \
a(40) a(41) a(42) a(43) a(44) a(45) a(46) a(47) \
a(48) a(49) a(50) a(51) a(52) a(53) a(54) a(55) \
a(56) a(57) a(58) a(59) a(60) a(61) a(62) a(63) \
} while (0)
/* Conditional loop unrolling */
#if defined(DEEP_PREPROC_UNROLL)
#define iter_5(a) unroll_5(a)
#define iter_8(a) unroll_8(a)
#define iter_16(a) unroll_16(a)
#define iter_64(a) unroll_64(a)
#else
#define iter_5(a) do {int _i; for (_i = 0; _i < 5; _i++) { a(_i) }} while (0)
#define iter_8(a) do {int _i; for (_i = 0; _i < 8; _i++) { a(_i) }} while (0)
#define iter_16(a) do {int _i; for (_i = 0; _i < 16; _i++) { a(_i) }} while (0)
#define iter_64(a) do {int _i; for (_i = 0; _i < 64; _i++) { a(_i) }} while (0)
#endif
/*
* BIGNUM mini-library
* This module deals with fixed-size 256-bit bignums.
* Where modular arithmetic is performed, the SECP256k1 prime
* modulus (below) is assumed.
*
* Methods include:
* - bn_is_zero/bn_is_one/bn_is_odd/bn_is_even/bn_is_bit_set
* - bn_rshift[1]/bn_lshift[1]
* - bn_neg
* - bn_uadd/bn_uadd_p
* - bn_usub/bn_usub_p
*/
typedef uint bn_word;
#define BN_NBITS 256
#define BN_WSHIFT 5
#define BN_WBITS (1 << BN_WSHIFT)
#define BN_NWORDS ((BN_NBITS/8) / sizeof(bn_word))
#define BN_WORDMAX 0xffffffff
#define MODULUS_BYTES \
0xfffffc2f, 0xfffffffe, 0xffffffff, 0xffffffff, \
0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff
typedef struct {
bn_word d[BN_NWORDS];
} bignum;
__constant bn_word modulus[] = { MODULUS_BYTES };
__constant bignum bn_zero = {{0x0}};
__constant bn_word mont_rr[BN_NWORDS] = { 0xe90a1, 0x7a2, 0x1, 0, };
__constant bn_word mont_n0[2] = { 0xd2253531, 0xd838091d };
#define bn_is_odd(bn) (bn.d[0] & 1)
#define bn_is_even(bn) (!bn_is_odd(bn))
#define bn_is_zero(bn) (!bn.d[0] && !bn.d[1] && !bn.d[2] && \
!bn.d[3] && !bn.d[4] && !bn.d[5] && \
!bn.d[6] && !bn.d[7])
#define bn_is_one(bn) ((bn.d[0] == 1) && !bn.d[1] && !bn.d[2] && \
!bn.d[3] && !bn.d[4] && !bn.d[5] && \
!bn.d[6] && !bn.d[7])
#define bn_is_bit_set(bn, n) \
((((bn_word*)&bn)[n >> BN_WSHIFT]) & (1 << (n & (BN_WBITS-1))))
#define bn_unroll(e) unroll_8(e)
#define bn_unroll_sf(e) unroll_1_7(e)
#define bn_unroll_sl(e) unroll_7(e)
#define bn_unroll_reverse(e) unroll_7_0(e)
#define bn_unroll_reverse_sl(e) unroll_7_1(e)
#define bn_unroll_arg(e, arg) \
e(arg, 0) e(arg, 1) e(arg, 2) e(arg, 3) \
e(arg, 4) e(arg, 5) e(arg, 6) e(arg, 7)
#define bn_unroll_arg_sf(e, arg) \
e(arg, 1) e(arg, 2) e(arg, 3) \
e(arg, 4) e(arg, 5) e(arg, 6) e(arg, 7)
#define bn_iter(e) iter_8(e)
/*
* Bitwise shift
*/
void
bn_lshift1(bignum *bn)
{
#define bn_lshift1_inner1(i) \
bn->d[i] = (bn->d[i] << 1) | (bn->d[i-1] >> 31);
bn_unroll_reverse_sl(bn_lshift1_inner1);
bn->d[0] <<= 1;
}
void
bn_rshift(bignum *bn, int shift)
{
int wd, iws, iwr;
bn_word ihw, ilw;
iws = (shift & (BN_WBITS-1));
iwr = BN_WBITS - iws;
wd = (shift >> BN_WSHIFT);
ihw = (wd < BN_WBITS) ? bn->d[wd] : 0;
#define bn_rshift_inner1(i) \
wd++; \
ilw = ihw; \
ihw = (wd < BN_WBITS) ? bn->d[wd] : 0; \
bn->d[i] = (ilw >> iws) | (ihw << iwr);
bn_unroll_sl(bn_rshift_inner1);
bn->d[BN_NWORDS-1] = (ihw >> iws);
}
void
bn_rshift1(bignum *bn)
{
#define bn_rshift1_inner1(i) \
bn->d[i] = (bn->d[i+1] << 31) | (bn->d[i] >> 1);
bn_unroll_sl(bn_rshift1_inner1);
bn->d[BN_NWORDS-1] >>= 1;
}
void
bn_rshift1_2(bignum *bna, bignum *bnb)
{
#define bn_rshift1_2_inner1(i) \
bna->d[i] = (bna->d[i+1] << 31) | (bna->d[i] >> 1); \
bnb->d[i] = (bnb->d[i+1] << 31) | (bnb->d[i] >> 1);
bn_unroll_sl(bn_rshift1_2_inner1);
bna->d[BN_NWORDS-1] >>= 1;
bnb->d[BN_NWORDS-1] >>= 1;
}
/*
* Unsigned comparison
*/
int
bn_ucmp_ge(bignum *a, bignum *b)
{
int l = 0, g = 0;
#define bn_ucmp_ge_inner1(i) \
if (a->d[i] < b->d[i]) l |= (1 << i); \
if (a->d[i] > b->d[i]) g |= (1 << i);
bn_unroll_reverse(bn_ucmp_ge_inner1);
return (l > g) ? 0 : 1;
}
int
bn_ucmp_ge_c(bignum *a, __constant bn_word *b)
{
int l = 0, g = 0;
#define bn_ucmp_ge_c_inner1(i) \
if (a->d[i] < b[i]) l |= (1 << i); \
if (a->d[i] > b[i]) g |= (1 << i);
bn_unroll_reverse(bn_ucmp_ge_c_inner1);
return (l > g) ? 0 : 1;
}
/*
* Negate
*/
void
bn_neg(bignum *n)
{
int c = 1;
#define bn_neg_inner1(i) \
c = (n->d[i] = (~n->d[i]) + c) ? 0 : c;
bn_unroll(bn_neg_inner1);
}
/*
* Add/subtract
*/
#define bn_add_word(r, a, b, t, c) do { \
t = a + b; \
c = (t < a) ? 1 : 0; \
r = t; \
} while (0)
#define bn_addc_word(r, a, b, t, c) do { \
t = a + b + c; \
c = (t < a) ? 1 : ((c & (t == a)) ? 1 : 0); \
r = t; \
} while (0)
bn_word
bn_uadd_words_seq(bn_word *r, bn_word *a, bn_word *b)
{
bn_word t, c = 0;
#define bn_uadd_words_seq_inner1(i) \
bn_addc_word(r[i], a[i], b[i], t, c);
bn_add_word(r[0], a[0], b[0], t, c);
bn_unroll_sf(bn_uadd_words_seq_inner1);
return c;
}
bn_word
bn_uadd_words_c_seq(bn_word *r, bn_word *a, __constant bn_word *b)
{
bn_word t, c = 0;
bn_add_word(r[0], a[0], b[0], t, c);
bn_unroll_sf(bn_uadd_words_seq_inner1);
return c;
}
#define bn_sub_word(r, a, b, t, c) do { \
t = a - b; \
c = (a < b) ? 1 : 0; \
r = t; \
} while (0)
#define bn_subb_word(r, a, b, t, c) do { \
t = a - (b + c); \
c = (!(a) && c) ? 1 : 0; \
c |= (a < b) ? 1 : 0; \
r = t; \
} while (0)
bn_word
bn_usub_words_seq(bn_word *r, bn_word *a, bn_word *b)
{
bn_word t, c = 0;
#define bn_usub_words_seq_inner1(i) \
bn_subb_word(r[i], a[i], b[i], t, c);
bn_sub_word(r[0], a[0], b[0], t, c);
bn_unroll_sf(bn_usub_words_seq_inner1);
return c;
}
bn_word
bn_usub_words_c_seq(bn_word *r, bn_word *a, __constant bn_word *b)
{
bn_word t, c = 0;
bn_sub_word(r[0], a[0], b[0], t, c);
bn_unroll_sf(bn_usub_words_seq_inner1);
return c;
}
/*
* Add/subtract better suited for AMD's VLIW architecture
*/
bn_word
bn_uadd_words_vliw(bn_word *r, bn_word *a, bn_word *b)
{
bignum x;
bn_word c = 0, cp = 0;
#define bn_uadd_words_vliw_inner1(i) \
x.d[i] = a[i] + b[i];
#define bn_uadd_words_vliw_inner2(i) \
c |= (a[i] > x.d[i]) ? (1 << i) : 0; \
cp |= (!~x.d[i]) ? (1 << i) : 0;
#define bn_uadd_words_vliw_inner3(i) \
r[i] = x.d[i] + ((c >> i) & 1);
bn_unroll(bn_uadd_words_vliw_inner1);
bn_unroll(bn_uadd_words_vliw_inner2);
c = ((cp + (c << 1)) ^ cp);
r[0] = x.d[0];
bn_unroll_sf(bn_uadd_words_vliw_inner3);
return c >> BN_NWORDS;
}
bn_word
bn_uadd_words_c_vliw(bn_word *r, bn_word *a, __constant bn_word *b)
{
bignum x;
bn_word c = 0, cp = 0;
bn_unroll(bn_uadd_words_vliw_inner1);
bn_unroll(bn_uadd_words_vliw_inner2);
c = ((cp + (c << 1)) ^ cp);
r[0] = x.d[0];
bn_unroll_sf(bn_uadd_words_vliw_inner3);
return c >> BN_NWORDS;
}
bn_word
bn_usub_words_vliw(bn_word *r, bn_word *a, bn_word *b)
{
bignum x;
bn_word c = 0, cp = 0;
#define bn_usub_words_vliw_inner1(i) \
x.d[i] = a[i] - b[i];
#define bn_usub_words_vliw_inner2(i) \
c |= (a[i] < b[i]) ? (1 << i) : 0; \
cp |= (!x.d[i]) ? (1 << i) : 0;
#define bn_usub_words_vliw_inner3(i) \
r[i] = x.d[i] - ((c >> i) & 1);
bn_unroll(bn_usub_words_vliw_inner1);
bn_unroll(bn_usub_words_vliw_inner2);
c = ((cp + (c << 1)) ^ cp);
r[0] = x.d[0];
bn_unroll_sf(bn_usub_words_vliw_inner3);
return c >> BN_NWORDS;
}
bn_word
bn_usub_words_c_vliw(bn_word *r, bn_word *a, __constant bn_word *b)
{
bignum x;
bn_word c = 0, cp = 0;
bn_unroll(bn_usub_words_vliw_inner1);
bn_unroll(bn_usub_words_vliw_inner2);
c = ((cp + (c << 1)) ^ cp);
r[0] = x.d[0];
bn_unroll_sf(bn_usub_words_vliw_inner3);
return c >> BN_NWORDS;
}
#if defined(DEEP_VLIW)
#define bn_uadd_words bn_uadd_words_vliw
#define bn_uadd_words_c bn_uadd_words_c_vliw
#define bn_usub_words bn_usub_words_vliw
#define bn_usub_words_c bn_usub_words_c_vliw
#else
#define bn_uadd_words bn_uadd_words_seq
#define bn_uadd_words_c bn_uadd_words_c_seq
#define bn_usub_words bn_usub_words_seq
#define bn_usub_words_c bn_usub_words_c_seq
#endif
#define bn_uadd(r, a, b) bn_uadd_words((r)->d, (a)->d, (b)->d)
#define bn_uadd_c(r, a, b) bn_uadd_words_c((r)->d, (a)->d, b)
#define bn_usub(r, a, b) bn_usub_words((r)->d, (a)->d, (b)->d)
#define bn_usub_c(r, a, b) bn_usub_words_c((r)->d, (a)->d, b)
/*
* Modular add/sub
*/
void
bn_mod_add(bignum *r, bignum *a, bignum *b)
{
if (bn_uadd(r, a, b) ||
(bn_ucmp_ge_c(r, modulus)))
bn_usub_c(r, r, modulus);
}
void
bn_mod_sub(bignum *r, bignum *a, bignum *b)
{
if (bn_usub(r, a, b))
bn_uadd_c(r, r, modulus);
}
void
bn_mod_lshift1(bignum *bn)
{
bn_word c = (bn->d[BN_NWORDS-1] & 0x80000000);
bn_lshift1(bn);
if (c || (bn_ucmp_ge_c(bn, modulus)))
bn_usub_c(bn, bn, modulus);
}
/*
* Montgomery multiplication
*
* This includes normal multiplication of two "Montgomeryized"
* bignums, and bn_from_mont for de-Montgomeryizing a bignum.
*/
#define bn_mul_word(r, a, w, c, p, s) do { \
r = (a * w) + c; \
p = mul_hi(a, w); \
c = (r < c) ? p + 1 : p; \
} while (0)
#define bn_mul_add_word(r, a, w, c, p, s) do { \
s = r + c; \
p = mul_hi(a, w); \
r = (a * w) + s; \
c = (s < c) ? p + 1 : p; \
if (r < s) c++; \
} while (0)
void
bn_mul_mont(bignum *r, bignum *a, bignum *b)
{
bignum t;
bn_word tea, teb, c, p, s, m;
#if !defined(VERY_EXPENSIVE_BRANCHES)
int q;
#endif
c = 0;
#define bn_mul_mont_inner1(j) \
bn_mul_word(t.d[j], a->d[j], b->d[0], c, p, s);
bn_unroll(bn_mul_mont_inner1);
tea = c;
teb = 0;
c = 0;
m = t.d[0] * mont_n0[0];
bn_mul_add_word(t.d[0], modulus[0], m, c, p, s);
#define bn_mul_mont_inner2(j) \
bn_mul_add_word(t.d[j], modulus[j], m, c, p, s); \
t.d[j-1] = t.d[j];
bn_unroll_sf(bn_mul_mont_inner2);
t.d[BN_NWORDS-1] = tea + c;
tea = teb + ((t.d[BN_NWORDS-1] < c) ? 1 : 0);
#define bn_mul_mont_inner3_1(i, j) \
bn_mul_add_word(t.d[j], a->d[j], b->d[i], c, p, s);
#define bn_mul_mont_inner3_2(i, j) \
bn_mul_add_word(t.d[j], modulus[j], m, c, p, s); \
t.d[j-1] = t.d[j];
#define bn_mul_mont_inner3(i) \
c = 0; \
bn_unroll_arg(bn_mul_mont_inner3_1, i); \
tea += c; \
teb = ((tea < c) ? 1 : 0); \
c = 0; \
m = t.d[0] * mont_n0[0]; \
bn_mul_add_word(t.d[0], modulus[0], m, c, p, s); \
bn_unroll_arg_sf(bn_mul_mont_inner3_2, i); \
t.d[BN_NWORDS-1] = tea + c; \
tea = teb + ((t.d[BN_NWORDS-1] < c) ? 1 : 0);
/*
* The outer loop here is quite long, and we won't unroll it
* unless VERY_EXPENSIVE_BRANCHES is set.
*/
#if defined(VERY_EXPENSIVE_BRANCHES)
bn_unroll_sf(bn_mul_mont_inner3);
c = tea | !bn_usub_c(r, &t, modulus);
if (!c)
*r = t;
#else
for (q = 1; q < BN_NWORDS; q++) {
bn_mul_mont_inner3(q);
}
c = tea || (t.d[BN_NWORDS-1] >= modulus[BN_NWORDS-1]);
if (c) {
c = tea | !bn_usub_c(r, &t, modulus);
if (c)
return;
}
*r = t;
#endif
}
void
bn_from_mont(bignum *rb, bignum *b)
{
#define WORKSIZE ((2*BN_NWORDS) + 1)
bn_word r[WORKSIZE];
bn_word m, c, p, s;
#if defined(PRAGMA_UNROLL)
int i;
#endif
/* Copy the input to the working area */
/* Zero the upper words */
#define bn_from_mont_inner1(i) \
r[i] = b->d[i];
#define bn_from_mont_inner2(i) \
r[BN_NWORDS+i] = 0;
bn_unroll(bn_from_mont_inner1);
bn_unroll(bn_from_mont_inner2);
r[WORKSIZE-1] = 0;
/* Multiply (long) by modulus */
#define bn_from_mont_inner3_1(i, j) \
bn_mul_add_word(r[i+j], modulus[j], m, c, p, s);
#if !defined(VERY_EXPENSIVE_BRANCHES)
#define bn_from_mont_inner3_2(i) \
if (r[BN_NWORDS + i] < c) \
r[BN_NWORDS + i + 1] += 1;
#else
#define bn_from_mont_inner3_2(i) \
r[BN_NWORDS + i + 1] += (r[BN_NWORDS + i] < c) ? 1 : 0;
#endif
#define bn_from_mont_inner3(i) \
m = r[i] * mont_n0[0]; \
c = 0; \
bn_unroll_arg(bn_from_mont_inner3_1, i); \
r[BN_NWORDS + i] += c; \
bn_from_mont_inner3_2(i)
/*
* The outer loop here is not very long, so we will unroll
* it by default. However, it's just complicated enough to
* cause NVIDIA's compiler to take unreasonably long to compile
* it, unless we use pragma unroll.
*/
#if !defined(PRAGMA_UNROLL)
bn_iter(bn_from_mont_inner3);
#else
#pragma unroll 8
for (i = 0; i < BN_NWORDS; i++) { bn_from_mont_inner3(i) }
#endif
/*
* Make sure the result is less than the modulus.
* Subtracting is not much more expensive than compare, so
* subtract always and assign based on the carry out value.
*/
c = bn_usub_words_c(rb->d, &r[BN_NWORDS], modulus);
if (c) {
#define bn_from_mont_inner4(i) \
rb->d[i] = r[BN_NWORDS + i];
bn_unroll(bn_from_mont_inner4);
}
}
/*
* Modular inversion
*/
void
bn_mod_inverse(bignum *r, bignum *n)
{
bignum a, b, x, y;
int shift;
bn_word xc, yc;
for (shift = 0; shift < BN_NWORDS; shift++) {
a.d[shift] = modulus[shift];
x.d[shift] = 0;
y.d[shift] = 0;
}
b = *n;
x.d[0] = 1;
xc = 0;
yc = 0;
while (!bn_is_zero(b)) {
shift = 0;
while (!bn_is_odd(b)) {
if (bn_is_odd(x))
xc += bn_uadd_c(&x, &x, modulus);
bn_rshift1_2(&x, &b);
x.d[7] |= (xc << 31);
xc >>= 1;
}
while (!bn_is_odd(a)) {
if (bn_is_odd(y))
yc += bn_uadd_c(&y, &y, modulus);
bn_rshift1_2(&y, &a);
y.d[7] |= (yc << 31);
yc >>= 1;
}
if (bn_ucmp_ge(&b, &a)) {
xc += yc + bn_uadd(&x, &x, &y);
bn_usub(&b, &b, &a);
} else {
yc += xc + bn_uadd(&y, &y, &x);
bn_usub(&a, &a, &b);
}
}
if (!bn_is_one(a)) {
/* no modular inverse */
*r = bn_zero;
} else {
/* Compute y % m as cheaply as possible */
while (yc < 0x80000000)
yc -= bn_usub_c(&y, &y, modulus);
bn_neg(&y);
*r = y;
}
}
/*
* HASH FUNCTIONS
*
* BYTE ORDER NOTE: None of the hash functions below deal with byte
* order. The caller is expected to be aware of this when it stuffs
* data into in the native integer.
*
* NOTE #2: Endianness of the OpenCL device makes no difference here.
*/
#define hash256_unroll(a) unroll_8(a)
#define hash160_unroll(a) unroll_5(a)
#define hash256_iter(a) iter_8(a)
#define hash160_iter(a) iter_5(a)
/******** The Keccak-f[1600] permutation ********/
/*** Constants. ***/
static const __constant uchar rho[24] = \
{ 1, 3, 6, 10, 15, 21,
28, 36, 45, 55, 2, 14,
27, 41, 56, 8, 25, 43,
62, 18, 39, 61, 20, 44};
static const __constant uchar pi[24] = \
{10, 7, 11, 17, 18, 3,
5, 16, 8, 21, 24, 4,
15, 23, 19, 13, 12, 2,
20, 14, 22, 9, 6, 1};
static const __constant ulong RC[24] = \
{1UL, 0x8082UL, 0x800000000000808aUL, 0x8000000080008000UL,
0x808bUL, 0x80000001UL, 0x8000000080008081UL, 0x8000000000008009UL,
0x8aUL, 0x88UL, 0x80008009UL, 0x8000000aUL,
0x8000808bUL, 0x800000000000008bUL, 0x8000000000008089UL, 0x8000000000008003UL,
0x8000000000008002UL, 0x8000000000000080UL, 0x800aUL, 0x800000008000000aUL,
0x8000000080008081UL, 0x8000000000008080UL, 0x80000001UL, 0x8000000080008008UL};
/*** Helper macros to unroll the permutation. ***/
#define rol(x, s) (((x) << s) | ((x) >> (64 - s)))
#define REPEAT6(e) e e e e e e
#define REPEAT24(e) REPEAT6(e e e e)
#define REPEAT5(e) e e e e e
#define FOR5(v, s, e) \
v = 0; \
REPEAT5(e; v += s;)
/*** Keccak-f[1600] ***/
static inline void keccakf(void* state) {
ulong* a = (ulong*)state;
ulong b[5] = {0UL,0UL,0UL,0UL,0UL};
ulong t = 0UL;
uchar x, y;
for (int i = 0; i < 24; i++) {
// Theta
FOR5(x, 1,
b[x] = 0;
FOR5(y, 5,
b[x] ^= a[x + y]; ))
FOR5(x, 1,
FOR5(y, 5,
a[y + x] ^= b[(x + 4) % 5] ^ rol(b[(x + 1) % 5], 1); ))
// Rho and pi
t = a[1];
x = 0;
REPEAT24(b[0] = a[pi[x]];
a[pi[x]] = rol(t, rho[x]);
t = b[0];
x++; )
// Chi
FOR5(y,
5,
FOR5(x, 1,
b[x] = a[y + x];)
FOR5(x, 1,
a[y + x] = b[x] ^ ((~b[(x + 1) % 5]) & b[(x + 2) % 5]); ))
// Iota
a[0] ^= RC[i];
}
}
/******** The FIPS202-defined functions. ********/
/*** Some helper macros. ***/
#define _(S) do { S } while (0)
#define FOR(i, ST, L, S) \
_(for (size_t i = 0; i < L; i += ST) { S; })
#define mkapply_ds(NAME, S) \
static inline void NAME(uchar* dst, \
const uchar* src, \
size_t len) { \
FOR(i, 1, len, S); \
}
#define mkapply_sd(NAME, S) \
static inline void NAME(const uchar* src, \
uchar* dst, \
size_t len) { \
FOR(i, 1, len, S); \
}
mkapply_ds(xorin, dst[i] ^= src[i]) // xorin
mkapply_sd(setout, dst[i] = src[i]) // setout
#define P keccakf
#define Plen 200
// Fold P*F over the fUL blocks of an input.
#define foldP(I, L, F) \
while (L >= rate) { \
F(a, I, rate); \
P(a); \
I += rate; \
L -= rate; \
}
/** The sponge-based hash construction. **/
static inline void hash(uchar* out, size_t outlen,
const uchar* in, size_t inlen,
size_t rate, uchar delim) {
uchar a[Plen] = {0};
ulong *st=(ulong *)a;
for(size_t i=0;i<25;i++) *st++ = 0UL;
// Absorb input.
foldP(in, inlen, xorin);
// Xor in the DS and pad frame.
a[inlen] ^= delim;
a[rate - 1] ^= 0x80;
// Xor in the last block.
xorin(a, in, inlen);
// Apply P
P(a);
// Squeeze output.
foldP(out, outlen, setout);
setout(a, out, outlen);
}
void sha3_256(uchar * out, const uchar * in)
{
hash(out, 32, in, 64, 200 - (256 / 4), 0x01);
}
/*** FIPS202 SHA3 FOFs ***/
#ifdef TEST_KERNELS
/*
* Test kernels
*/
/* Montgomery multiplication test kernel */
__kernel void
test_mul_mont(__global bignum *products_out, __global bignum *nums_in)
{
bignum a, b, c;
int o;
o = get_global_id(0);
nums_in += (2*o);
a = nums_in[0];
b = nums_in[1];
bn_mul_mont(&c, &a, &b);
products_out[o] = c;
}
/* modular inversion test kernel */
__kernel void
test_mod_inverse(__global bignum *inv_out, __global bignum *nums_in,
int count)
{
bignum x, xp;
int i, o;
o = get_global_id(0) * count;
for (i = 0; i < count; i++) {
x = nums_in[o];
bn_mod_inverse(&xp, &x);
inv_out[o++] = xp;
}
}
#endif /* TEST_KERNELS */
#define ACCESS_BUNDLE 1024
#define ACCESS_STRIDE (ACCESS_BUNDLE/BN_NWORDS)
__kernel void
ec_add_grid(__global bn_word *points_out, __global bn_word *z_heap,
__global bn_word *row_in, __global bignum *col_in)
{
bignum rx, ry;
bignum x1, y1, a, b, c, d, e, z;
bn_word cy;
int i, cell, start;
/* Load the row increment point */
i = 2 * get_global_id(1);
rx = col_in[i];
ry = col_in[i+1];
cell = get_global_id(0);
start = ((((2 * cell) / ACCESS_STRIDE) * ACCESS_BUNDLE) +
(cell % (ACCESS_STRIDE/2)));
#define ec_add_grid_inner_1(i) \
x1.d[i] = row_in[start + (i*ACCESS_STRIDE)];
bn_unroll(ec_add_grid_inner_1);
start += (ACCESS_STRIDE/2);
#define ec_add_grid_inner_2(i) \
y1.d[i] = row_in[start + (i*ACCESS_STRIDE)];
bn_unroll(ec_add_grid_inner_2);
bn_mod_sub(&z, &x1, &rx);
cell += (get_global_id(1) * get_global_size(0));
start = (((cell / ACCESS_STRIDE) * ACCESS_BUNDLE) +
(cell % ACCESS_STRIDE));
#define ec_add_grid_inner_3(i) \
z_heap[start + (i*ACCESS_STRIDE)] = z.d[i];
bn_unroll(ec_add_grid_inner_3);
bn_mod_sub(&b, &y1, &ry);
bn_mod_add(&c, &x1, &rx);
bn_mod_add(&d, &y1, &ry);
bn_mul_mont(&y1, &b, &b);
bn_mul_mont(&x1, &z, &z);
bn_mul_mont(&e, &c, &x1);
bn_mod_sub(&y1, &y1, &e);
/*
* This disgusting code caters to the global memory unit on
* various GPUs, by giving it a nice contiguous patch to write
* per warp/wavefront.
*/
start = ((((2 * cell) / ACCESS_STRIDE) * ACCESS_BUNDLE) +
(cell % (ACCESS_STRIDE/2)));
#define ec_add_grid_inner_4(i) \
points_out[start + (i*ACCESS_STRIDE)] = y1.d[i];
bn_unroll(ec_add_grid_inner_4);
bn_mod_lshift1(&y1);
bn_mod_sub(&y1, &e, &y1);
bn_mul_mont(&y1, &y1, &b);
bn_mul_mont(&a, &x1, &z);
bn_mul_mont(&c, &d, &a);
bn_mod_sub(&y1, &y1, &c);
cy = 0;
if (bn_is_odd(y1))
cy = bn_uadd_c(&y1, &y1, modulus);
bn_rshift1(&y1);
y1.d[BN_NWORDS-1] |= (cy ? 0x80000000 : 0);
start += (ACCESS_STRIDE/2);
bn_unroll(ec_add_grid_inner_4);
}
__kernel void
heap_invert(__global bn_word *z_heap, int batch)
{
bignum a, b, c, z;
int i, off, lcell, hcell, start;
#define heap_invert_inner_load_a(j) \
a.d[j] = z_heap[start + j*ACCESS_STRIDE];
#define heap_invert_inner_load_b(j) \
b.d[j] = z_heap[start + j*ACCESS_STRIDE];
#define heap_invert_inner_load_z(j) \
z.d[j] = z_heap[start + j*ACCESS_STRIDE];