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
|
/******************************************************************************
* Copyright (c) 2013, 2014, 2017 Pieter Wuille, Andrew Poelstra, Jonas Nick *
* Distributed under the MIT software license, see the accompanying *
* file COPYING or https://www.opensource.org/licenses/mit-license.php. *
******************************************************************************/
#ifndef SECP256K1_ECMULT_IMPL_H
#define SECP256K1_ECMULT_IMPL_H
#include <string.h>
#include <stdint.h>
#include "util.h"
#include "group.h"
#include "scalar.h"
#include "ecmult.h"
#include "ecmult_static_pre_g.h"
#if defined(EXHAUSTIVE_TEST_ORDER)
/* We need to lower these values for exhaustive tests because
* the tables cannot have infinities in them (this breaks the
* affine-isomorphism stuff which tracks z-ratios) */
# if EXHAUSTIVE_TEST_ORDER > 128
# define WINDOW_A 5
# elif EXHAUSTIVE_TEST_ORDER > 8
# define WINDOW_A 4
# else
# define WINDOW_A 2
# endif
#else
/* optimal for 128-bit and 256-bit exponents. */
# define WINDOW_A 5
/** Larger values for ECMULT_WINDOW_SIZE result in possibly better
* performance at the cost of an exponentially larger precomputed
* table. The exact table size is
* (1 << (WINDOW_G - 2)) * sizeof(secp256k1_ge_storage) bytes,
* where sizeof(secp256k1_ge_storage) is typically 64 bytes but can
* be larger due to platform-specific padding and alignment.
* Two tables of this size are used (due to the endomorphism
* optimization).
*/
#endif
#define WNAF_BITS 128
#define WNAF_SIZE_BITS(bits, w) (((bits) + (w) - 1) / (w))
#define WNAF_SIZE(w) WNAF_SIZE_BITS(WNAF_BITS, w)
/* The number of objects allocated on the scratch space for ecmult_multi algorithms */
#define PIPPENGER_SCRATCH_OBJECTS 6
#define STRAUSS_SCRATCH_OBJECTS 7
#define PIPPENGER_MAX_BUCKET_WINDOW 12
/* Minimum number of points for which pippenger_wnaf is faster than strauss wnaf */
#define ECMULT_PIPPENGER_THRESHOLD 88
#define ECMULT_MAX_POINTS_PER_BATCH 5000000
/** Fill a table 'prej' with precomputed odd multiples of a. Prej will contain
* the values [1*a,3*a,...,(2*n-1)*a], so it space for n values. zr[0] will
* contain prej[0].z / a.z. The other zr[i] values = prej[i].z / prej[i-1].z.
* Prej's Z values are undefined, except for the last value.
*/
static void secp256k1_ecmult_odd_multiples_table(int n, secp256k1_gej *prej, secp256k1_fe *zr, const secp256k1_gej *a) {
secp256k1_gej d;
secp256k1_ge a_ge, d_ge;
int i;
VERIFY_CHECK(!a->infinity);
secp256k1_gej_double_var(&d, a, NULL);
/*
* Perform the additions on an isomorphism where 'd' is affine: drop the z coordinate
* of 'd', and scale the 1P starting value's x/y coordinates without changing its z.
*/
d_ge.x = d.x;
d_ge.y = d.y;
d_ge.infinity = 0;
secp256k1_ge_set_gej_zinv(&a_ge, a, &d.z);
prej[0].x = a_ge.x;
prej[0].y = a_ge.y;
prej[0].z = a->z;
prej[0].infinity = 0;
zr[0] = d.z;
for (i = 1; i < n; i++) {
secp256k1_gej_add_ge_var(&prej[i], &prej[i-1], &d_ge, &zr[i]);
}
/*
* Each point in 'prej' has a z coordinate too small by a factor of 'd.z'. Only
* the final point's z coordinate is actually used though, so just update that.
*/
secp256k1_fe_mul(&prej[n-1].z, &prej[n-1].z, &d.z);
}
/** Fill a table 'pre' with precomputed odd multiples of a.
*
* The resulting point set is brought to a single constant Z denominator, stores the X and Y
* coordinates as ge_storage points in pre, and stores the global Z in rz.
* It only operates on tables sized for WINDOW_A wnaf multiples.
*
* To compute a*P + b*G, we compute a table for P using this function,
* and use the precomputed table in <ecmult_static_pre_g.h> for G.
*/
static void secp256k1_ecmult_odd_multiples_table_globalz_windowa(secp256k1_ge *pre, secp256k1_fe *globalz, const secp256k1_gej *a) {
secp256k1_gej prej[ECMULT_TABLE_SIZE(WINDOW_A)];
secp256k1_fe zr[ECMULT_TABLE_SIZE(WINDOW_A)];
/* Compute the odd multiples in Jacobian form. */
secp256k1_ecmult_odd_multiples_table(ECMULT_TABLE_SIZE(WINDOW_A), prej, zr, a);
/* Bring them to the same Z denominator. */
secp256k1_ge_globalz_set_table_gej(ECMULT_TABLE_SIZE(WINDOW_A), pre, globalz, prej, zr);
}
/** The following two macro retrieves a particular odd multiple from a table
* of precomputed multiples. */
#define ECMULT_TABLE_GET_GE(r,pre,n,w) do { \
VERIFY_CHECK(((n) & 1) == 1); \
VERIFY_CHECK((n) >= -((1 << ((w)-1)) - 1)); \
VERIFY_CHECK((n) <= ((1 << ((w)-1)) - 1)); \
if ((n) > 0) { \
*(r) = (pre)[((n)-1)/2]; \
} else { \
*(r) = (pre)[(-(n)-1)/2]; \
secp256k1_fe_negate(&((r)->y), &((r)->y), 1); \
} \
} while(0)
#define ECMULT_TABLE_GET_GE_STORAGE(r,pre,n,w) do { \
VERIFY_CHECK(((n) & 1) == 1); \
VERIFY_CHECK((n) >= -((1 << ((w)-1)) - 1)); \
VERIFY_CHECK((n) <= ((1 << ((w)-1)) - 1)); \
if ((n) > 0) { \
secp256k1_ge_from_storage((r), &(pre)[((n)-1)/2]); \
} else { \
secp256k1_ge_from_storage((r), &(pre)[(-(n)-1)/2]); \
secp256k1_fe_negate(&((r)->y), &((r)->y), 1); \
} \
} while(0)
/** Convert a number to WNAF notation. The number becomes represented by sum(2^i * wnaf[i], i=0..bits),
* with the following guarantees:
* - each wnaf[i] is either 0, or an odd integer between -(1<<(w-1) - 1) and (1<<(w-1) - 1)
* - two non-zero entries in wnaf are separated by at least w-1 zeroes.
* - the number of set values in wnaf is returned. This number is at most 256, and at most one more
* than the number of bits in the (absolute value) of the input.
*/
static int secp256k1_ecmult_wnaf(int *wnaf, int len, const secp256k1_scalar *a, int w) {
secp256k1_scalar s;
int last_set_bit = -1;
int bit = 0;
int sign = 1;
int carry = 0;
VERIFY_CHECK(wnaf != NULL);
VERIFY_CHECK(0 <= len && len <= 256);
VERIFY_CHECK(a != NULL);
VERIFY_CHECK(2 <= w && w <= 31);
memset(wnaf, 0, len * sizeof(wnaf[0]));
s = *a;
if (secp256k1_scalar_get_bits(&s, 255, 1)) {
secp256k1_scalar_negate(&s, &s);
sign = -1;
}
while (bit < len) {
int now;
int word;
if (secp256k1_scalar_get_bits(&s, bit, 1) == (unsigned int)carry) {
bit++;
continue;
}
now = w;
if (now > len - bit) {
now = len - bit;
}
word = secp256k1_scalar_get_bits_var(&s, bit, now) + carry;
carry = (word >> (w-1)) & 1;
word -= carry << w;
wnaf[bit] = sign * word;
last_set_bit = bit;
bit += now;
}
#ifdef VERIFY
CHECK(carry == 0);
while (bit < 256) {
CHECK(secp256k1_scalar_get_bits(&s, bit++, 1) == 0);
}
#endif
return last_set_bit + 1;
}
struct secp256k1_strauss_point_state {
secp256k1_scalar na_1, na_lam;
int wnaf_na_1[129];
int wnaf_na_lam[129];
int bits_na_1;
int bits_na_lam;
size_t input_pos;
};
struct secp256k1_strauss_state {
secp256k1_gej* prej;
secp256k1_fe* zr;
secp256k1_ge* pre_a;
secp256k1_ge* pre_a_lam;
struct secp256k1_strauss_point_state* ps;
};
static void secp256k1_ecmult_strauss_wnaf(const struct secp256k1_strauss_state *state, secp256k1_gej *r, size_t num, const secp256k1_gej *a, const secp256k1_scalar *na, const secp256k1_scalar *ng) {
secp256k1_ge tmpa;
secp256k1_fe Z;
/* Split G factors. */
secp256k1_scalar ng_1, ng_128;
int wnaf_ng_1[129];
int bits_ng_1 = 0;
int wnaf_ng_128[129];
int bits_ng_128 = 0;
int i;
int bits = 0;
size_t np;
size_t no = 0;
for (np = 0; np < num; ++np) {
if (secp256k1_scalar_is_zero(&na[np]) || secp256k1_gej_is_infinity(&a[np])) {
continue;
}
state->ps[no].input_pos = np;
/* split na into na_1 and na_lam (where na = na_1 + na_lam*lambda, and na_1 and na_lam are ~128 bit) */
secp256k1_scalar_split_lambda(&state->ps[no].na_1, &state->ps[no].na_lam, &na[np]);
/* build wnaf representation for na_1 and na_lam. */
state->ps[no].bits_na_1 = secp256k1_ecmult_wnaf(state->ps[no].wnaf_na_1, 129, &state->ps[no].na_1, WINDOW_A);
state->ps[no].bits_na_lam = secp256k1_ecmult_wnaf(state->ps[no].wnaf_na_lam, 129, &state->ps[no].na_lam, WINDOW_A);
VERIFY_CHECK(state->ps[no].bits_na_1 <= 129);
VERIFY_CHECK(state->ps[no].bits_na_lam <= 129);
if (state->ps[no].bits_na_1 > bits) {
bits = state->ps[no].bits_na_1;
}
if (state->ps[no].bits_na_lam > bits) {
bits = state->ps[no].bits_na_lam;
}
++no;
}
/* Calculate odd multiples of a.
* All multiples are brought to the same Z 'denominator', which is stored
* in Z. Due to secp256k1' isomorphism we can do all operations pretending
* that the Z coordinate was 1, use affine addition formulae, and correct
* the Z coordinate of the result once at the end.
* The exception is the precomputed G table points, which are actually
* affine. Compared to the base used for other points, they have a Z ratio
* of 1/Z, so we can use secp256k1_gej_add_zinv_var, which uses the same
* isomorphism to efficiently add with a known Z inverse.
*/
if (no > 0) {
/* Compute the odd multiples in Jacobian form. */
secp256k1_ecmult_odd_multiples_table(ECMULT_TABLE_SIZE(WINDOW_A), state->prej, state->zr, &a[state->ps[0].input_pos]);
for (np = 1; np < no; ++np) {
secp256k1_gej tmp = a[state->ps[np].input_pos];
#ifdef VERIFY
secp256k1_fe_normalize_var(&(state->prej[(np - 1) * ECMULT_TABLE_SIZE(WINDOW_A) + ECMULT_TABLE_SIZE(WINDOW_A) - 1].z));
#endif
secp256k1_gej_rescale(&tmp, &(state->prej[(np - 1) * ECMULT_TABLE_SIZE(WINDOW_A) + ECMULT_TABLE_SIZE(WINDOW_A) - 1].z));
secp256k1_ecmult_odd_multiples_table(ECMULT_TABLE_SIZE(WINDOW_A), state->prej + np * ECMULT_TABLE_SIZE(WINDOW_A), state->zr + np * ECMULT_TABLE_SIZE(WINDOW_A), &tmp);
secp256k1_fe_mul(state->zr + np * ECMULT_TABLE_SIZE(WINDOW_A), state->zr + np * ECMULT_TABLE_SIZE(WINDOW_A), &(a[state->ps[np].input_pos].z));
}
/* Bring them to the same Z denominator. */
secp256k1_ge_globalz_set_table_gej(ECMULT_TABLE_SIZE(WINDOW_A) * no, state->pre_a, &Z, state->prej, state->zr);
} else {
secp256k1_fe_set_int(&Z, 1);
}
for (np = 0; np < no; ++np) {
for (i = 0; i < ECMULT_TABLE_SIZE(WINDOW_A); i++) {
secp256k1_ge_mul_lambda(&state->pre_a_lam[np * ECMULT_TABLE_SIZE(WINDOW_A) + i], &state->pre_a[np * ECMULT_TABLE_SIZE(WINDOW_A) + i]);
}
}
if (ng) {
/* split ng into ng_1 and ng_128 (where gn = gn_1 + gn_128*2^128, and gn_1 and gn_128 are ~128 bit) */
secp256k1_scalar_split_128(&ng_1, &ng_128, ng);
/* Build wnaf representation for ng_1 and ng_128 */
bits_ng_1 = secp256k1_ecmult_wnaf(wnaf_ng_1, 129, &ng_1, WINDOW_G);
bits_ng_128 = secp256k1_ecmult_wnaf(wnaf_ng_128, 129, &ng_128, WINDOW_G);
if (bits_ng_1 > bits) {
bits = bits_ng_1;
}
if (bits_ng_128 > bits) {
bits = bits_ng_128;
}
}
secp256k1_gej_set_infinity(r);
for (i = bits - 1; i >= 0; i--) {
int n;
secp256k1_gej_double_var(r, r, NULL);
for (np = 0; np < no; ++np) {
if (i < state->ps[np].bits_na_1 && (n = state->ps[np].wnaf_na_1[i])) {
ECMULT_TABLE_GET_GE(&tmpa, state->pre_a + np * ECMULT_TABLE_SIZE(WINDOW_A), n, WINDOW_A);
secp256k1_gej_add_ge_var(r, r, &tmpa, NULL);
}
if (i < state->ps[np].bits_na_lam && (n = state->ps[np].wnaf_na_lam[i])) {
ECMULT_TABLE_GET_GE(&tmpa, state->pre_a_lam + np * ECMULT_TABLE_SIZE(WINDOW_A), n, WINDOW_A);
secp256k1_gej_add_ge_var(r, r, &tmpa, NULL);
}
}
if (i < bits_ng_1 && (n = wnaf_ng_1[i])) {
ECMULT_TABLE_GET_GE_STORAGE(&tmpa, secp256k1_pre_g, n, WINDOW_G);
secp256k1_gej_add_zinv_var(r, r, &tmpa, &Z);
}
if (i < bits_ng_128 && (n = wnaf_ng_128[i])) {
ECMULT_TABLE_GET_GE_STORAGE(&tmpa, secp256k1_pre_g_128, n, WINDOW_G);
secp256k1_gej_add_zinv_var(r, r, &tmpa, &Z);
}
}
if (!r->infinity) {
secp256k1_fe_mul(&r->z, &r->z, &Z);
}
}
static void secp256k1_ecmult(secp256k1_gej *r, const secp256k1_gej *a, const secp256k1_scalar *na, const secp256k1_scalar *ng) {
secp256k1_gej prej[ECMULT_TABLE_SIZE(WINDOW_A)];
secp256k1_fe zr[ECMULT_TABLE_SIZE(WINDOW_A)];
secp256k1_ge pre_a[ECMULT_TABLE_SIZE(WINDOW_A)];
struct secp256k1_strauss_point_state ps[1];
secp256k1_ge pre_a_lam[ECMULT_TABLE_SIZE(WINDOW_A)];
struct secp256k1_strauss_state state;
state.prej = prej;
state.zr = zr;
state.pre_a = pre_a;
state.pre_a_lam = pre_a_lam;
state.ps = ps;
secp256k1_ecmult_strauss_wnaf(&state, r, 1, a, na, ng);
}
static size_t secp256k1_strauss_scratch_size(size_t n_points) {
static const size_t point_size = (2 * sizeof(secp256k1_ge) + sizeof(secp256k1_gej) + sizeof(secp256k1_fe)) * ECMULT_TABLE_SIZE(WINDOW_A) + sizeof(struct secp256k1_strauss_point_state) + sizeof(secp256k1_gej) + sizeof(secp256k1_scalar);
return n_points*point_size;
}
static int secp256k1_ecmult_strauss_batch(const secp256k1_callback* error_callback, secp256k1_scratch *scratch, secp256k1_gej *r, const secp256k1_scalar *inp_g_sc, secp256k1_ecmult_multi_callback cb, void *cbdata, size_t n_points, size_t cb_offset) {
secp256k1_gej* points;
secp256k1_scalar* scalars;
struct secp256k1_strauss_state state;
size_t i;
const size_t scratch_checkpoint = secp256k1_scratch_checkpoint(error_callback, scratch);
secp256k1_gej_set_infinity(r);
if (inp_g_sc == NULL && n_points == 0) {
return 1;
}
/* We allocate STRAUSS_SCRATCH_OBJECTS objects on the scratch space. If these
* allocations change, make sure to update the STRAUSS_SCRATCH_OBJECTS
* constant and strauss_scratch_size accordingly. */
points = (secp256k1_gej*)secp256k1_scratch_alloc(error_callback, scratch, n_points * sizeof(secp256k1_gej));
scalars = (secp256k1_scalar*)secp256k1_scratch_alloc(error_callback, scratch, n_points * sizeof(secp256k1_scalar));
state.prej = (secp256k1_gej*)secp256k1_scratch_alloc(error_callback, scratch, n_points * ECMULT_TABLE_SIZE(WINDOW_A) * sizeof(secp256k1_gej));
state.zr = (secp256k1_fe*)secp256k1_scratch_alloc(error_callback, scratch, n_points * ECMULT_TABLE_SIZE(WINDOW_A) * sizeof(secp256k1_fe));
state.pre_a = (secp256k1_ge*)secp256k1_scratch_alloc(error_callback, scratch, n_points * ECMULT_TABLE_SIZE(WINDOW_A) * sizeof(secp256k1_ge));
state.pre_a_lam = (secp256k1_ge*)secp256k1_scratch_alloc(error_callback, scratch, n_points * ECMULT_TABLE_SIZE(WINDOW_A) * sizeof(secp256k1_ge));
state.ps = (struct secp256k1_strauss_point_state*)secp256k1_scratch_alloc(error_callback, scratch, n_points * sizeof(struct secp256k1_strauss_point_state));
if (points == NULL || scalars == NULL || state.prej == NULL || state.zr == NULL || state.pre_a == NULL || state.pre_a_lam == NULL || state.ps == NULL) {
secp256k1_scratch_apply_checkpoint(error_callback, scratch, scratch_checkpoint);
return 0;
}
for (i = 0; i < n_points; i++) {
secp256k1_ge point;
if (!cb(&scalars[i], &point, i+cb_offset, cbdata)) {
secp256k1_scratch_apply_checkpoint(error_callback, scratch, scratch_checkpoint);
return 0;
}
secp256k1_gej_set_ge(&points[i], &point);
}
secp256k1_ecmult_strauss_wnaf(&state, r, n_points, points, scalars, inp_g_sc);
secp256k1_scratch_apply_checkpoint(error_callback, scratch, scratch_checkpoint);
return 1;
}
/* Wrapper for secp256k1_ecmult_multi_func interface */
static int secp256k1_ecmult_strauss_batch_single(const secp256k1_callback* error_callback, secp256k1_scratch *scratch, secp256k1_gej *r, const secp256k1_scalar *inp_g_sc, secp256k1_ecmult_multi_callback cb, void *cbdata, size_t n) {
return secp256k1_ecmult_strauss_batch(error_callback, scratch, r, inp_g_sc, cb, cbdata, n, 0);
}
static size_t secp256k1_strauss_max_points(const secp256k1_callback* error_callback, secp256k1_scratch *scratch) {
return secp256k1_scratch_max_allocation(error_callback, scratch, STRAUSS_SCRATCH_OBJECTS) / secp256k1_strauss_scratch_size(1);
}
/** Convert a number to WNAF notation.
* The number becomes represented by sum(2^{wi} * wnaf[i], i=0..WNAF_SIZE(w)+1) - return_val.
* It has the following guarantees:
* - each wnaf[i] is either 0 or an odd integer between -(1 << w) and (1 << w)
* - the number of words set is always WNAF_SIZE(w)
* - the returned skew is 0 or 1
*/
static int secp256k1_wnaf_fixed(int *wnaf, const secp256k1_scalar *s, int w) {
int skew = 0;
int pos;
int max_pos;
int last_w;
const secp256k1_scalar *work = s;
if (secp256k1_scalar_is_zero(s)) {
for (pos = 0; pos < WNAF_SIZE(w); pos++) {
wnaf[pos] = 0;
}
return 0;
}
if (secp256k1_scalar_is_even(s)) {
skew = 1;
}
wnaf[0] = secp256k1_scalar_get_bits_var(work, 0, w) + skew;
/* Compute last window size. Relevant when window size doesn't divide the
* number of bits in the scalar */
last_w = WNAF_BITS - (WNAF_SIZE(w) - 1) * w;
/* Store the position of the first nonzero word in max_pos to allow
* skipping leading zeros when calculating the wnaf. */
for (pos = WNAF_SIZE(w) - 1; pos > 0; pos--) {
int val = secp256k1_scalar_get_bits_var(work, pos * w, pos == WNAF_SIZE(w)-1 ? last_w : w);
if(val != 0) {
break;
}
wnaf[pos] = 0;
}
max_pos = pos;
pos = 1;
while (pos <= max_pos) {
int val = secp256k1_scalar_get_bits_var(work, pos * w, pos == WNAF_SIZE(w)-1 ? last_w : w);
if ((val & 1) == 0) {
wnaf[pos - 1] -= (1 << w);
wnaf[pos] = (val + 1);
} else {
wnaf[pos] = val;
}
/* Set a coefficient to zero if it is 1 or -1 and the proceeding digit
* is strictly negative or strictly positive respectively. Only change
* coefficients at previous positions because above code assumes that
* wnaf[pos - 1] is odd.
*/
if (pos >= 2 && ((wnaf[pos - 1] == 1 && wnaf[pos - 2] < 0) || (wnaf[pos - 1] == -1 && wnaf[pos - 2] > 0))) {
if (wnaf[pos - 1] == 1) {
wnaf[pos - 2] += 1 << w;
} else {
wnaf[pos - 2] -= 1 << w;
}
wnaf[pos - 1] = 0;
}
++pos;
}
return skew;
}
struct secp256k1_pippenger_point_state {
int skew_na;
size_t input_pos;
};
struct secp256k1_pippenger_state {
int *wnaf_na;
struct secp256k1_pippenger_point_state* ps;
};
/*
* pippenger_wnaf computes the result of a multi-point multiplication as
* follows: The scalars are brought into wnaf with n_wnaf elements each. Then
* for every i < n_wnaf, first each point is added to a "bucket" corresponding
* to the point's wnaf[i]. Second, the buckets are added together such that
* r += 1*bucket[0] + 3*bucket[1] + 5*bucket[2] + ...
*/
static int secp256k1_ecmult_pippenger_wnaf(secp256k1_gej *buckets, int bucket_window, struct secp256k1_pippenger_state *state, secp256k1_gej *r, const secp256k1_scalar *sc, const secp256k1_ge *pt, size_t num) {
size_t n_wnaf = WNAF_SIZE(bucket_window+1);
size_t np;
size_t no = 0;
int i;
int j;
for (np = 0; np < num; ++np) {
if (secp256k1_scalar_is_zero(&sc[np]) || secp256k1_ge_is_infinity(&pt[np])) {
continue;
}
state->ps[no].input_pos = np;
state->ps[no].skew_na = secp256k1_wnaf_fixed(&state->wnaf_na[no*n_wnaf], &sc[np], bucket_window+1);
no++;
}
secp256k1_gej_set_infinity(r);
if (no == 0) {
return 1;
}
for (i = n_wnaf - 1; i >= 0; i--) {
secp256k1_gej running_sum;
for(j = 0; j < ECMULT_TABLE_SIZE(bucket_window+2); j++) {
secp256k1_gej_set_infinity(&buckets[j]);
}
for (np = 0; np < no; ++np) {
int n = state->wnaf_na[np*n_wnaf + i];
struct secp256k1_pippenger_point_state point_state = state->ps[np];
secp256k1_ge tmp;
int idx;
if (i == 0) {
/* correct for wnaf skew */
int skew = point_state.skew_na;
if (skew) {
secp256k1_ge_neg(&tmp, &pt[point_state.input_pos]);
secp256k1_gej_add_ge_var(&buckets[0], &buckets[0], &tmp, NULL);
}
}
if (n > 0) {
idx = (n - 1)/2;
secp256k1_gej_add_ge_var(&buckets[idx], &buckets[idx], &pt[point_state.input_pos], NULL);
} else if (n < 0) {
idx = -(n + 1)/2;
secp256k1_ge_neg(&tmp, &pt[point_state.input_pos]);
secp256k1_gej_add_ge_var(&buckets[idx], &buckets[idx], &tmp, NULL);
}
}
for(j = 0; j < bucket_window; j++) {
secp256k1_gej_double_var(r, r, NULL);
}
secp256k1_gej_set_infinity(&running_sum);
/* Accumulate the sum: bucket[0] + 3*bucket[1] + 5*bucket[2] + 7*bucket[3] + ...
* = bucket[0] + bucket[1] + bucket[2] + bucket[3] + ...
* + 2 * (bucket[1] + 2*bucket[2] + 3*bucket[3] + ...)
* using an intermediate running sum:
* running_sum = bucket[0] + bucket[1] + bucket[2] + ...
*
* The doubling is done implicitly by deferring the final window doubling (of 'r').
*/
for(j = ECMULT_TABLE_SIZE(bucket_window+2) - 1; j > 0; j--) {
secp256k1_gej_add_var(&running_sum, &running_sum, &buckets[j], NULL);
secp256k1_gej_add_var(r, r, &running_sum, NULL);
}
secp256k1_gej_add_var(&running_sum, &running_sum, &buckets[0], NULL);
secp256k1_gej_double_var(r, r, NULL);
secp256k1_gej_add_var(r, r, &running_sum, NULL);
}
return 1;
}
/**
* Returns optimal bucket_window (number of bits of a scalar represented by a
* set of buckets) for a given number of points.
*/
static int secp256k1_pippenger_bucket_window(size_t n) {
if (n <= 1) {
return 1;
} else if (n <= 4) {
return 2;
} else if (n <= 20) {
return 3;
} else if (n <= 57) {
return 4;
} else if (n <= 136) {
return 5;
} else if (n <= 235) {
return 6;
} else if (n <= 1260) {
return 7;
} else if (n <= 4420) {
return 9;
} else if (n <= 7880) {
return 10;
} else if (n <= 16050) {
return 11;
} else {
return PIPPENGER_MAX_BUCKET_WINDOW;
}
}
/**
* Returns the maximum optimal number of points for a bucket_window.
*/
static size_t secp256k1_pippenger_bucket_window_inv(int bucket_window) {
switch(bucket_window) {
case 1: return 1;
case 2: return 4;
case 3: return 20;
case 4: return 57;
case 5: return 136;
case 6: return 235;
case 7: return 1260;
case 8: return 1260;
case 9: return 4420;
case 10: return 7880;
case 11: return 16050;
case PIPPENGER_MAX_BUCKET_WINDOW: return SIZE_MAX;
}
return 0;
}
SECP256K1_INLINE static void secp256k1_ecmult_endo_split(secp256k1_scalar *s1, secp256k1_scalar *s2, secp256k1_ge *p1, secp256k1_ge *p2) {
secp256k1_scalar tmp = *s1;
secp256k1_scalar_split_lambda(s1, s2, &tmp);
secp256k1_ge_mul_lambda(p2, p1);
if (secp256k1_scalar_is_high(s1)) {
secp256k1_scalar_negate(s1, s1);
secp256k1_ge_neg(p1, p1);
}
if (secp256k1_scalar_is_high(s2)) {
secp256k1_scalar_negate(s2, s2);
secp256k1_ge_neg(p2, p2);
}
}
/**
* Returns the scratch size required for a given number of points (excluding
* base point G) without considering alignment.
*/
static size_t secp256k1_pippenger_scratch_size(size_t n_points, int bucket_window) {
size_t entries = 2*n_points + 2;
size_t entry_size = sizeof(secp256k1_ge) + sizeof(secp256k1_scalar) + sizeof(struct secp256k1_pippenger_point_state) + (WNAF_SIZE(bucket_window+1)+1)*sizeof(int);
return (sizeof(secp256k1_gej) << bucket_window) + sizeof(struct secp256k1_pippenger_state) + entries * entry_size;
}
static int secp256k1_ecmult_pippenger_batch(const secp256k1_callback* error_callback, secp256k1_scratch *scratch, secp256k1_gej *r, const secp256k1_scalar *inp_g_sc, secp256k1_ecmult_multi_callback cb, void *cbdata, size_t n_points, size_t cb_offset) {
const size_t scratch_checkpoint = secp256k1_scratch_checkpoint(error_callback, scratch);
/* Use 2(n+1) with the endomorphism, when calculating batch
* sizes. The reason for +1 is that we add the G scalar to the list of
* other scalars. */
size_t entries = 2*n_points + 2;
secp256k1_ge *points;
secp256k1_scalar *scalars;
secp256k1_gej *buckets;
struct secp256k1_pippenger_state *state_space;
size_t idx = 0;
size_t point_idx = 0;
int i, j;
int bucket_window;
secp256k1_gej_set_infinity(r);
if (inp_g_sc == NULL && n_points == 0) {
return 1;
}
bucket_window = secp256k1_pippenger_bucket_window(n_points);
/* We allocate PIPPENGER_SCRATCH_OBJECTS objects on the scratch space. If
* these allocations change, make sure to update the
* PIPPENGER_SCRATCH_OBJECTS constant and pippenger_scratch_size
* accordingly. */
points = (secp256k1_ge *) secp256k1_scratch_alloc(error_callback, scratch, entries * sizeof(*points));
scalars = (secp256k1_scalar *) secp256k1_scratch_alloc(error_callback, scratch, entries * sizeof(*scalars));
state_space = (struct secp256k1_pippenger_state *) secp256k1_scratch_alloc(error_callback, scratch, sizeof(*state_space));
if (points == NULL || scalars == NULL || state_space == NULL) {
secp256k1_scratch_apply_checkpoint(error_callback, scratch, scratch_checkpoint);
return 0;
}
state_space->ps = (struct secp256k1_pippenger_point_state *) secp256k1_scratch_alloc(error_callback, scratch, entries * sizeof(*state_space->ps));
state_space->wnaf_na = (int *) secp256k1_scratch_alloc(error_callback, scratch, entries*(WNAF_SIZE(bucket_window+1)) * sizeof(int));
buckets = (secp256k1_gej *) secp256k1_scratch_alloc(error_callback, scratch, (1<<bucket_window) * sizeof(*buckets));
if (state_space->ps == NULL || state_space->wnaf_na == NULL || buckets == NULL) {
secp256k1_scratch_apply_checkpoint(error_callback, scratch, scratch_checkpoint);
return 0;
}
if (inp_g_sc != NULL) {
scalars[0] = *inp_g_sc;
points[0] = secp256k1_ge_const_g;
idx++;
secp256k1_ecmult_endo_split(&scalars[0], &scalars[1], &points[0], &points[1]);
idx++;
}
while (point_idx < n_points) {
if (!cb(&scalars[idx], &points[idx], point_idx + cb_offset, cbdata)) {
secp256k1_scratch_apply_checkpoint(error_callback, scratch, scratch_checkpoint);
return 0;
}
idx++;
secp256k1_ecmult_endo_split(&scalars[idx - 1], &scalars[idx], &points[idx - 1], &points[idx]);
idx++;
point_idx++;
}
secp256k1_ecmult_pippenger_wnaf(buckets, bucket_window, state_space, r, scalars, points, idx);
/* Clear data */
for(i = 0; (size_t)i < idx; i++) {
secp256k1_scalar_clear(&scalars[i]);
state_space->ps[i].skew_na = 0;
for(j = 0; j < WNAF_SIZE(bucket_window+1); j++) {
state_space->wnaf_na[i * WNAF_SIZE(bucket_window+1) + j] = 0;
}
}
for(i = 0; i < 1<<bucket_window; i++) {
secp256k1_gej_clear(&buckets[i]);
}
secp256k1_scratch_apply_checkpoint(error_callback, scratch, scratch_checkpoint);
return 1;
}
/* Wrapper for secp256k1_ecmult_multi_func interface */
static int secp256k1_ecmult_pippenger_batch_single(const secp256k1_callback* error_callback, secp256k1_scratch *scratch, secp256k1_gej *r, const secp256k1_scalar *inp_g_sc, secp256k1_ecmult_multi_callback cb, void *cbdata, size_t n) {
return secp256k1_ecmult_pippenger_batch(error_callback, scratch, r, inp_g_sc, cb, cbdata, n, 0);
}
/**
* Returns the maximum number of points in addition to G that can be used with
* a given scratch space. The function ensures that fewer points may also be
* used.
*/
static size_t secp256k1_pippenger_max_points(const secp256k1_callback* error_callback, secp256k1_scratch *scratch) {
size_t max_alloc = secp256k1_scratch_max_allocation(error_callback, scratch, PIPPENGER_SCRATCH_OBJECTS);
int bucket_window;
size_t res = 0;
for (bucket_window = 1; bucket_window <= PIPPENGER_MAX_BUCKET_WINDOW; bucket_window++) {
size_t n_points;
size_t max_points = secp256k1_pippenger_bucket_window_inv(bucket_window);
size_t space_for_points;
size_t space_overhead;
size_t entry_size = sizeof(secp256k1_ge) + sizeof(secp256k1_scalar) + sizeof(struct secp256k1_pippenger_point_state) + (WNAF_SIZE(bucket_window+1)+1)*sizeof(int);
entry_size = 2*entry_size;
space_overhead = (sizeof(secp256k1_gej) << bucket_window) + entry_size + sizeof(struct secp256k1_pippenger_state);
if (space_overhead > max_alloc) {
break;
}
space_for_points = max_alloc - space_overhead;
n_points = space_for_points/entry_size;
n_points = n_points > max_points ? max_points : n_points;
if (n_points > res) {
res = n_points;
}
if (n_points < max_points) {
/* A larger bucket_window may support even more points. But if we
* would choose that then the caller couldn't safely use any number
* smaller than what this function returns */
break;
}
}
return res;
}
/* Computes ecmult_multi by simply multiplying and adding each point. Does not
* require a scratch space */
static int secp256k1_ecmult_multi_simple_var(secp256k1_gej *r, const secp256k1_scalar *inp_g_sc, secp256k1_ecmult_multi_callback cb, void *cbdata, size_t n_points) {
size_t point_idx;
secp256k1_scalar szero;
secp256k1_gej tmpj;
secp256k1_scalar_set_int(&szero, 0);
secp256k1_gej_set_infinity(r);
secp256k1_gej_set_infinity(&tmpj);
/* r = inp_g_sc*G */
secp256k1_ecmult(r, &tmpj, &szero, inp_g_sc);
for (point_idx = 0; point_idx < n_points; point_idx++) {
secp256k1_ge point;
secp256k1_gej pointj;
secp256k1_scalar scalar;
if (!cb(&scalar, &point, point_idx, cbdata)) {
return 0;
}
/* r += scalar*point */
secp256k1_gej_set_ge(&pointj, &point);
secp256k1_ecmult(&tmpj, &pointj, &scalar, NULL);
secp256k1_gej_add_var(r, r, &tmpj, NULL);
}
return 1;
}
/* Compute the number of batches and the batch size given the maximum batch size and the
* total number of points */
static int secp256k1_ecmult_multi_batch_size_helper(size_t *n_batches, size_t *n_batch_points, size_t max_n_batch_points, size_t n) {
if (max_n_batch_points == 0) {
return 0;
}
if (max_n_batch_points > ECMULT_MAX_POINTS_PER_BATCH) {
max_n_batch_points = ECMULT_MAX_POINTS_PER_BATCH;
}
if (n == 0) {
*n_batches = 0;
*n_batch_points = 0;
return 1;
}
/* Compute ceil(n/max_n_batch_points) and ceil(n/n_batches) */
*n_batches = 1 + (n - 1) / max_n_batch_points;
*n_batch_points = 1 + (n - 1) / *n_batches;
return 1;
}
typedef int (*secp256k1_ecmult_multi_func)(const secp256k1_callback* error_callback, secp256k1_scratch*, secp256k1_gej*, const secp256k1_scalar*, secp256k1_ecmult_multi_callback cb, void*, size_t);
static int secp256k1_ecmult_multi_var(const secp256k1_callback* error_callback, secp256k1_scratch *scratch, secp256k1_gej *r, const secp256k1_scalar *inp_g_sc, secp256k1_ecmult_multi_callback cb, void *cbdata, size_t n) {
size_t i;
int (*f)(const secp256k1_callback* error_callback, secp256k1_scratch*, secp256k1_gej*, const secp256k1_scalar*, secp256k1_ecmult_multi_callback cb, void*, size_t, size_t);
size_t n_batches;
size_t n_batch_points;
secp256k1_gej_set_infinity(r);
if (inp_g_sc == NULL && n == 0) {
return 1;
} else if (n == 0) {
secp256k1_scalar szero;
secp256k1_scalar_set_int(&szero, 0);
secp256k1_ecmult(r, r, &szero, inp_g_sc);
return 1;
}
if (scratch == NULL) {
return secp256k1_ecmult_multi_simple_var(r, inp_g_sc, cb, cbdata, n);
}
/* Compute the batch sizes for Pippenger's algorithm given a scratch space. If it's greater than
* a threshold use Pippenger's algorithm. Otherwise use Strauss' algorithm.
* As a first step check if there's enough space for Pippenger's algo (which requires less space
* than Strauss' algo) and if not, use the simple algorithm. */
if (!secp256k1_ecmult_multi_batch_size_helper(&n_batches, &n_batch_points, secp256k1_pippenger_max_points(error_callback, scratch), n)) {
return secp256k1_ecmult_multi_simple_var(r, inp_g_sc, cb, cbdata, n);
}
if (n_batch_points >= ECMULT_PIPPENGER_THRESHOLD) {
f = secp256k1_ecmult_pippenger_batch;
} else {
if (!secp256k1_ecmult_multi_batch_size_helper(&n_batches, &n_batch_points, secp256k1_strauss_max_points(error_callback, scratch), n)) {
return secp256k1_ecmult_multi_simple_var(r, inp_g_sc, cb, cbdata, n);
}
f = secp256k1_ecmult_strauss_batch;
}
for(i = 0; i < n_batches; i++) {
size_t nbp = n < n_batch_points ? n : n_batch_points;
size_t offset = n_batch_points*i;
secp256k1_gej tmp;
if (!f(error_callback, scratch, &tmp, i == 0 ? inp_g_sc : NULL, cb, cbdata, nbp, offset)) {
return 0;
}
secp256k1_gej_add_var(r, r, &tmp, NULL);
n -= nbp;
}
return 1;
}
#endif /* SECP256K1_ECMULT_IMPL_H */
|