aboutsummaryrefslogtreecommitdiff
path: root/src/secp256k1/src/scalar_8x32_impl.h
blob: 22b31d41125daeb942d2db22f80089856b611af1 (plain)
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
/**********************************************************************
 * Copyright (c) 2014 Pieter Wuille                                   *
 * Distributed under the MIT software license, see the accompanying   *
 * file COPYING or http://www.opensource.org/licenses/mit-license.php.*
 **********************************************************************/

#ifndef _SECP256K1_SCALAR_REPR_IMPL_H_
#define _SECP256K1_SCALAR_REPR_IMPL_H_

/* Limbs of the secp256k1 order. */
#define SECP256K1_N_0 ((uint32_t)0xD0364141UL)
#define SECP256K1_N_1 ((uint32_t)0xBFD25E8CUL)
#define SECP256K1_N_2 ((uint32_t)0xAF48A03BUL)
#define SECP256K1_N_3 ((uint32_t)0xBAAEDCE6UL)
#define SECP256K1_N_4 ((uint32_t)0xFFFFFFFEUL)
#define SECP256K1_N_5 ((uint32_t)0xFFFFFFFFUL)
#define SECP256K1_N_6 ((uint32_t)0xFFFFFFFFUL)
#define SECP256K1_N_7 ((uint32_t)0xFFFFFFFFUL)

/* Limbs of 2^256 minus the secp256k1 order. */
#define SECP256K1_N_C_0 (~SECP256K1_N_0 + 1)
#define SECP256K1_N_C_1 (~SECP256K1_N_1)
#define SECP256K1_N_C_2 (~SECP256K1_N_2)
#define SECP256K1_N_C_3 (~SECP256K1_N_3)
#define SECP256K1_N_C_4 (1)

/* Limbs of half the secp256k1 order. */
#define SECP256K1_N_H_0 ((uint32_t)0x681B20A0UL)
#define SECP256K1_N_H_1 ((uint32_t)0xDFE92F46UL)
#define SECP256K1_N_H_2 ((uint32_t)0x57A4501DUL)
#define SECP256K1_N_H_3 ((uint32_t)0x5D576E73UL)
#define SECP256K1_N_H_4 ((uint32_t)0xFFFFFFFFUL)
#define SECP256K1_N_H_5 ((uint32_t)0xFFFFFFFFUL)
#define SECP256K1_N_H_6 ((uint32_t)0xFFFFFFFFUL)
#define SECP256K1_N_H_7 ((uint32_t)0x7FFFFFFFUL)

SECP256K1_INLINE static void secp256k1_scalar_clear(secp256k1_scalar_t *r) {
    r->d[0] = 0;
    r->d[1] = 0;
    r->d[2] = 0;
    r->d[3] = 0;
    r->d[4] = 0;
    r->d[5] = 0;
    r->d[6] = 0;
    r->d[7] = 0;
}

SECP256K1_INLINE static void secp256k1_scalar_set_int(secp256k1_scalar_t *r, unsigned int v) {
    r->d[0] = v;
    r->d[1] = 0;
    r->d[2] = 0;
    r->d[3] = 0;
    r->d[4] = 0;
    r->d[5] = 0;
    r->d[6] = 0;
    r->d[7] = 0;
}

SECP256K1_INLINE static unsigned int secp256k1_scalar_get_bits(const secp256k1_scalar_t *a, unsigned int offset, unsigned int count) {
    VERIFY_CHECK((offset + count - 1) >> 5 == offset >> 5);
    return (a->d[offset >> 5] >> (offset & 0x1F)) & ((1 << count) - 1);
}

SECP256K1_INLINE static unsigned int secp256k1_scalar_get_bits_var(const secp256k1_scalar_t *a, unsigned int offset, unsigned int count) {
    VERIFY_CHECK(count < 32);
    VERIFY_CHECK(offset + count <= 256);
    if ((offset + count - 1) >> 5 == offset >> 5) {
        return secp256k1_scalar_get_bits(a, offset, count);
    } else {
        VERIFY_CHECK((offset >> 5) + 1 < 8);
        return ((a->d[offset >> 5] >> (offset & 0x1F)) | (a->d[(offset >> 5) + 1] << (32 - (offset & 0x1F)))) & ((((uint32_t)1) << count) - 1);
    }
}

SECP256K1_INLINE static int secp256k1_scalar_check_overflow(const secp256k1_scalar_t *a) {
    int yes = 0;
    int no = 0;
    no |= (a->d[7] < SECP256K1_N_7); /* No need for a > check. */
    no |= (a->d[6] < SECP256K1_N_6); /* No need for a > check. */
    no |= (a->d[5] < SECP256K1_N_5); /* No need for a > check. */
    no |= (a->d[4] < SECP256K1_N_4);
    yes |= (a->d[4] > SECP256K1_N_4) & ~no;
    no |= (a->d[3] < SECP256K1_N_3) & ~yes;
    yes |= (a->d[3] > SECP256K1_N_3) & ~no;
    no |= (a->d[2] < SECP256K1_N_2) & ~yes;
    yes |= (a->d[2] > SECP256K1_N_2) & ~no;
    no |= (a->d[1] < SECP256K1_N_1) & ~yes;
    yes |= (a->d[1] > SECP256K1_N_1) & ~no;
    yes |= (a->d[0] >= SECP256K1_N_0) & ~no;
    return yes;
}

SECP256K1_INLINE static int secp256k1_scalar_reduce(secp256k1_scalar_t *r, uint32_t overflow) {
    uint64_t t;
    VERIFY_CHECK(overflow <= 1);
    t = (uint64_t)r->d[0] + overflow * SECP256K1_N_C_0;
    r->d[0] = t & 0xFFFFFFFFUL; t >>= 32;
    t += (uint64_t)r->d[1] + overflow * SECP256K1_N_C_1;
    r->d[1] = t & 0xFFFFFFFFUL; t >>= 32;
    t += (uint64_t)r->d[2] + overflow * SECP256K1_N_C_2;
    r->d[2] = t & 0xFFFFFFFFUL; t >>= 32;
    t += (uint64_t)r->d[3] + overflow * SECP256K1_N_C_3;
    r->d[3] = t & 0xFFFFFFFFUL; t >>= 32;
    t += (uint64_t)r->d[4] + overflow * SECP256K1_N_C_4;
    r->d[4] = t & 0xFFFFFFFFUL; t >>= 32;
    t += (uint64_t)r->d[5];
    r->d[5] = t & 0xFFFFFFFFUL; t >>= 32;
    t += (uint64_t)r->d[6];
    r->d[6] = t & 0xFFFFFFFFUL; t >>= 32;
    t += (uint64_t)r->d[7];
    r->d[7] = t & 0xFFFFFFFFUL;
    return overflow;
}

static int secp256k1_scalar_add(secp256k1_scalar_t *r, const secp256k1_scalar_t *a, const secp256k1_scalar_t *b) {
    int overflow;
    uint64_t t = (uint64_t)a->d[0] + b->d[0];
    r->d[0] = t & 0xFFFFFFFFULL; t >>= 32;
    t += (uint64_t)a->d[1] + b->d[1];
    r->d[1] = t & 0xFFFFFFFFULL; t >>= 32;
    t += (uint64_t)a->d[2] + b->d[2];
    r->d[2] = t & 0xFFFFFFFFULL; t >>= 32;
    t += (uint64_t)a->d[3] + b->d[3];
    r->d[3] = t & 0xFFFFFFFFULL; t >>= 32;
    t += (uint64_t)a->d[4] + b->d[4];
    r->d[4] = t & 0xFFFFFFFFULL; t >>= 32;
    t += (uint64_t)a->d[5] + b->d[5];
    r->d[5] = t & 0xFFFFFFFFULL; t >>= 32;
    t += (uint64_t)a->d[6] + b->d[6];
    r->d[6] = t & 0xFFFFFFFFULL; t >>= 32;
    t += (uint64_t)a->d[7] + b->d[7];
    r->d[7] = t & 0xFFFFFFFFULL; t >>= 32;
    overflow = t + secp256k1_scalar_check_overflow(r);
    VERIFY_CHECK(overflow == 0 || overflow == 1);
    secp256k1_scalar_reduce(r, overflow);
    return overflow;
}

static void secp256k1_scalar_add_bit(secp256k1_scalar_t *r, unsigned int bit) {
    uint64_t t;
    VERIFY_CHECK(bit < 256);
    t = (uint64_t)r->d[0] + (((uint32_t)((bit >> 5) == 0)) << (bit & 0x1F));
    r->d[0] = t & 0xFFFFFFFFULL; t >>= 32;
    t += (uint64_t)r->d[1] + (((uint32_t)((bit >> 5) == 1)) << (bit & 0x1F));
    r->d[1] = t & 0xFFFFFFFFULL; t >>= 32;
    t += (uint64_t)r->d[2] + (((uint32_t)((bit >> 5) == 2)) << (bit & 0x1F));
    r->d[2] = t & 0xFFFFFFFFULL; t >>= 32;
    t += (uint64_t)r->d[3] + (((uint32_t)((bit >> 5) == 3)) << (bit & 0x1F));
    r->d[3] = t & 0xFFFFFFFFULL; t >>= 32;
    t += (uint64_t)r->d[4] + (((uint32_t)((bit >> 5) == 4)) << (bit & 0x1F));
    r->d[4] = t & 0xFFFFFFFFULL; t >>= 32;
    t += (uint64_t)r->d[5] + (((uint32_t)((bit >> 5) == 5)) << (bit & 0x1F));
    r->d[5] = t & 0xFFFFFFFFULL; t >>= 32;
    t += (uint64_t)r->d[6] + (((uint32_t)((bit >> 5) == 6)) << (bit & 0x1F));
    r->d[6] = t & 0xFFFFFFFFULL; t >>= 32;
    t += (uint64_t)r->d[7] + (((uint32_t)((bit >> 5) == 7)) << (bit & 0x1F));
    r->d[7] = t & 0xFFFFFFFFULL;
#ifdef VERIFY
    VERIFY_CHECK((t >> 32) == 0);
    VERIFY_CHECK(secp256k1_scalar_check_overflow(r) == 0);
#endif
}

static void secp256k1_scalar_set_b32(secp256k1_scalar_t *r, const unsigned char *b32, int *overflow) {
    int over;
    r->d[0] = (uint32_t)b32[31] | (uint32_t)b32[30] << 8 | (uint32_t)b32[29] << 16 | (uint32_t)b32[28] << 24;
    r->d[1] = (uint32_t)b32[27] | (uint32_t)b32[26] << 8 | (uint32_t)b32[25] << 16 | (uint32_t)b32[24] << 24;
    r->d[2] = (uint32_t)b32[23] | (uint32_t)b32[22] << 8 | (uint32_t)b32[21] << 16 | (uint32_t)b32[20] << 24;
    r->d[3] = (uint32_t)b32[19] | (uint32_t)b32[18] << 8 | (uint32_t)b32[17] << 16 | (uint32_t)b32[16] << 24;
    r->d[4] = (uint32_t)b32[15] | (uint32_t)b32[14] << 8 | (uint32_t)b32[13] << 16 | (uint32_t)b32[12] << 24;
    r->d[5] = (uint32_t)b32[11] | (uint32_t)b32[10] << 8 | (uint32_t)b32[9] << 16 | (uint32_t)b32[8] << 24;
    r->d[6] = (uint32_t)b32[7] | (uint32_t)b32[6] << 8 | (uint32_t)b32[5] << 16 | (uint32_t)b32[4] << 24;
    r->d[7] = (uint32_t)b32[3] | (uint32_t)b32[2] << 8 | (uint32_t)b32[1] << 16 | (uint32_t)b32[0] << 24;
    over = secp256k1_scalar_reduce(r, secp256k1_scalar_check_overflow(r));
    if (overflow) {
        *overflow = over;
    }
}

static void secp256k1_scalar_get_b32(unsigned char *bin, const secp256k1_scalar_t* a) {
    bin[0] = a->d[7] >> 24; bin[1] = a->d[7] >> 16; bin[2] = a->d[7] >> 8; bin[3] = a->d[7];
    bin[4] = a->d[6] >> 24; bin[5] = a->d[6] >> 16; bin[6] = a->d[6] >> 8; bin[7] = a->d[6];
    bin[8] = a->d[5] >> 24; bin[9] = a->d[5] >> 16; bin[10] = a->d[5] >> 8; bin[11] = a->d[5];
    bin[12] = a->d[4] >> 24; bin[13] = a->d[4] >> 16; bin[14] = a->d[4] >> 8; bin[15] = a->d[4];
    bin[16] = a->d[3] >> 24; bin[17] = a->d[3] >> 16; bin[18] = a->d[3] >> 8; bin[19] = a->d[3];
    bin[20] = a->d[2] >> 24; bin[21] = a->d[2] >> 16; bin[22] = a->d[2] >> 8; bin[23] = a->d[2];
    bin[24] = a->d[1] >> 24; bin[25] = a->d[1] >> 16; bin[26] = a->d[1] >> 8; bin[27] = a->d[1];
    bin[28] = a->d[0] >> 24; bin[29] = a->d[0] >> 16; bin[30] = a->d[0] >> 8; bin[31] = a->d[0];
}

SECP256K1_INLINE static int secp256k1_scalar_is_zero(const secp256k1_scalar_t *a) {
    return (a->d[0] | a->d[1] | a->d[2] | a->d[3] | a->d[4] | a->d[5] | a->d[6] | a->d[7]) == 0;
}

static void secp256k1_scalar_negate(secp256k1_scalar_t *r, const secp256k1_scalar_t *a) {
    uint32_t nonzero = 0xFFFFFFFFUL * (secp256k1_scalar_is_zero(a) == 0);
    uint64_t t = (uint64_t)(~a->d[0]) + SECP256K1_N_0 + 1;
    r->d[0] = t & nonzero; t >>= 32;
    t += (uint64_t)(~a->d[1]) + SECP256K1_N_1;
    r->d[1] = t & nonzero; t >>= 32;
    t += (uint64_t)(~a->d[2]) + SECP256K1_N_2;
    r->d[2] = t & nonzero; t >>= 32;
    t += (uint64_t)(~a->d[3]) + SECP256K1_N_3;
    r->d[3] = t & nonzero; t >>= 32;
    t += (uint64_t)(~a->d[4]) + SECP256K1_N_4;
    r->d[4] = t & nonzero; t >>= 32;
    t += (uint64_t)(~a->d[5]) + SECP256K1_N_5;
    r->d[5] = t & nonzero; t >>= 32;
    t += (uint64_t)(~a->d[6]) + SECP256K1_N_6;
    r->d[6] = t & nonzero; t >>= 32;
    t += (uint64_t)(~a->d[7]) + SECP256K1_N_7;
    r->d[7] = t & nonzero;
}

SECP256K1_INLINE static int secp256k1_scalar_is_one(const secp256k1_scalar_t *a) {
    return ((a->d[0] ^ 1) | a->d[1] | a->d[2] | a->d[3] | a->d[4] | a->d[5] | a->d[6] | a->d[7]) == 0;
}

static int secp256k1_scalar_is_high(const secp256k1_scalar_t *a) {
    int yes = 0;
    int no = 0;
    no |= (a->d[7] < SECP256K1_N_H_7);
    yes |= (a->d[7] > SECP256K1_N_H_7) & ~no;
    no |= (a->d[6] < SECP256K1_N_H_6) & ~yes; /* No need for a > check. */
    no |= (a->d[5] < SECP256K1_N_H_5) & ~yes; /* No need for a > check. */
    no |= (a->d[4] < SECP256K1_N_H_4) & ~yes; /* No need for a > check. */
    no |= (a->d[3] < SECP256K1_N_H_3) & ~yes;
    yes |= (a->d[3] > SECP256K1_N_H_3) & ~no;
    no |= (a->d[2] < SECP256K1_N_H_2) & ~yes;
    yes |= (a->d[2] > SECP256K1_N_H_2) & ~no;
    no |= (a->d[1] < SECP256K1_N_H_1) & ~yes;
    yes |= (a->d[1] > SECP256K1_N_H_1) & ~no;
    yes |= (a->d[0] > SECP256K1_N_H_0) & ~no;
    return yes;
}

/* Inspired by the macros in OpenSSL's crypto/bn/asm/x86_64-gcc.c. */

/** Add a*b to the number defined by (c0,c1,c2). c2 must never overflow. */
#define muladd(a,b) { \
    uint32_t tl, th; \
    { \
        uint64_t t = (uint64_t)a * b; \
        th = t >> 32;         /* at most 0xFFFFFFFE */ \
        tl = t; \
    } \
    c0 += tl;                 /* overflow is handled on the next line */ \
    th += (c0 < tl) ? 1 : 0;  /* at most 0xFFFFFFFF */ \
    c1 += th;                 /* overflow is handled on the next line */ \
    c2 += (c1 < th) ? 1 : 0;  /* never overflows by contract (verified in the next line) */ \
    VERIFY_CHECK((c1 >= th) || (c2 != 0)); \
}

/** Add a*b to the number defined by (c0,c1). c1 must never overflow. */
#define muladd_fast(a,b) { \
    uint32_t tl, th; \
    { \
        uint64_t t = (uint64_t)a * b; \
        th = t >> 32;         /* at most 0xFFFFFFFE */ \
        tl = t; \
    } \
    c0 += tl;                 /* overflow is handled on the next line */ \
    th += (c0 < tl) ? 1 : 0;  /* at most 0xFFFFFFFF */ \
    c1 += th;                 /* never overflows by contract (verified in the next line) */ \
    VERIFY_CHECK(c1 >= th); \
}

/** Add 2*a*b to the number defined by (c0,c1,c2). c2 must never overflow. */
#define muladd2(a,b) { \
    uint32_t tl, th, th2, tl2; \
    { \
        uint64_t t = (uint64_t)a * b; \
        th = t >> 32;               /* at most 0xFFFFFFFE */ \
        tl = t; \
    } \
    th2 = th + th;                  /* at most 0xFFFFFFFE (in case th was 0x7FFFFFFF) */ \
    c2 += (th2 < th) ? 1 : 0;       /* never overflows by contract (verified the next line) */ \
    VERIFY_CHECK((th2 >= th) || (c2 != 0)); \
    tl2 = tl + tl;                  /* at most 0xFFFFFFFE (in case the lowest 63 bits of tl were 0x7FFFFFFF) */ \
    th2 += (tl2 < tl) ? 1 : 0;      /* at most 0xFFFFFFFF */ \
    c0 += tl2;                      /* overflow is handled on the next line */ \
    th2 += (c0 < tl2) ? 1 : 0;      /* second overflow is handled on the next line */ \
    c2 += (c0 < tl2) & (th2 == 0);  /* never overflows by contract (verified the next line) */ \
    VERIFY_CHECK((c0 >= tl2) || (th2 != 0) || (c2 != 0)); \
    c1 += th2;                      /* overflow is handled on the next line */ \
    c2 += (c1 < th2) ? 1 : 0;       /* never overflows by contract (verified the next line) */ \
    VERIFY_CHECK((c1 >= th2) || (c2 != 0)); \
}

/** Add a to the number defined by (c0,c1,c2). c2 must never overflow. */
#define sumadd(a) { \
    unsigned int over; \
    c0 += (a);                  /* overflow is handled on the next line */ \
    over = (c0 < (a)) ? 1 : 0; \
    c1 += over;                 /* overflow is handled on the next line */ \
    c2 += (c1 < over) ? 1 : 0;  /* never overflows by contract */ \
}

/** Add a to the number defined by (c0,c1). c1 must never overflow, c2 must be zero. */
#define sumadd_fast(a) { \
    c0 += (a);                 /* overflow is handled on the next line */ \
    c1 += (c0 < (a)) ? 1 : 0;  /* never overflows by contract (verified the next line) */ \
    VERIFY_CHECK((c1 != 0) | (c0 >= (a))); \
    VERIFY_CHECK(c2 == 0); \
}

/** Extract the lowest 32 bits of (c0,c1,c2) into n, and left shift the number 32 bits. */
#define extract(n) { \
    (n) = c0; \
    c0 = c1; \
    c1 = c2; \
    c2 = 0; \
}

/** Extract the lowest 32 bits of (c0,c1,c2) into n, and left shift the number 32 bits. c2 is required to be zero. */
#define extract_fast(n) { \
    (n) = c0; \
    c0 = c1; \
    c1 = 0; \
    VERIFY_CHECK(c2 == 0); \
}

static void secp256k1_scalar_reduce_512(secp256k1_scalar_t *r, const uint32_t *l) {
    uint64_t c;
    uint32_t n0 = l[8], n1 = l[9], n2 = l[10], n3 = l[11], n4 = l[12], n5 = l[13], n6 = l[14], n7 = l[15];
    uint32_t m0, m1, m2, m3, m4, m5, m6, m7, m8, m9, m10, m11, m12;
    uint32_t p0, p1, p2, p3, p4, p5, p6, p7, p8;

    /* 96 bit accumulator. */
    uint32_t c0, c1, c2;

    /* Reduce 512 bits into 385. */
    /* m[0..12] = l[0..7] + n[0..7] * SECP256K1_N_C. */
    c0 = l[0]; c1 = 0; c2 = 0;
    muladd_fast(n0, SECP256K1_N_C_0);
    extract_fast(m0);
    sumadd_fast(l[1]);
    muladd(n1, SECP256K1_N_C_0);
    muladd(n0, SECP256K1_N_C_1);
    extract(m1);
    sumadd(l[2]);
    muladd(n2, SECP256K1_N_C_0);
    muladd(n1, SECP256K1_N_C_1);
    muladd(n0, SECP256K1_N_C_2);
    extract(m2);
    sumadd(l[3]);
    muladd(n3, SECP256K1_N_C_0);
    muladd(n2, SECP256K1_N_C_1);
    muladd(n1, SECP256K1_N_C_2);
    muladd(n0, SECP256K1_N_C_3);
    extract(m3);
    sumadd(l[4]);
    muladd(n4, SECP256K1_N_C_0);
    muladd(n3, SECP256K1_N_C_1);
    muladd(n2, SECP256K1_N_C_2);
    muladd(n1, SECP256K1_N_C_3);
    sumadd(n0);
    extract(m4);
    sumadd(l[5]);
    muladd(n5, SECP256K1_N_C_0);
    muladd(n4, SECP256K1_N_C_1);
    muladd(n3, SECP256K1_N_C_2);
    muladd(n2, SECP256K1_N_C_3);
    sumadd(n1);
    extract(m5);
    sumadd(l[6]);
    muladd(n6, SECP256K1_N_C_0);
    muladd(n5, SECP256K1_N_C_1);
    muladd(n4, SECP256K1_N_C_2);
    muladd(n3, SECP256K1_N_C_3);
    sumadd(n2);
    extract(m6);
    sumadd(l[7]);
    muladd(n7, SECP256K1_N_C_0);
    muladd(n6, SECP256K1_N_C_1);
    muladd(n5, SECP256K1_N_C_2);
    muladd(n4, SECP256K1_N_C_3);
    sumadd(n3);
    extract(m7);
    muladd(n7, SECP256K1_N_C_1);
    muladd(n6, SECP256K1_N_C_2);
    muladd(n5, SECP256K1_N_C_3);
    sumadd(n4);
    extract(m8);
    muladd(n7, SECP256K1_N_C_2);
    muladd(n6, SECP256K1_N_C_3);
    sumadd(n5);
    extract(m9);
    muladd(n7, SECP256K1_N_C_3);
    sumadd(n6);
    extract(m10);
    sumadd_fast(n7);
    extract_fast(m11);
    VERIFY_CHECK(c0 <= 1);
    m12 = c0;

    /* Reduce 385 bits into 258. */
    /* p[0..8] = m[0..7] + m[8..12] * SECP256K1_N_C. */
    c0 = m0; c1 = 0; c2 = 0;
    muladd_fast(m8, SECP256K1_N_C_0);
    extract_fast(p0);
    sumadd_fast(m1);
    muladd(m9, SECP256K1_N_C_0);
    muladd(m8, SECP256K1_N_C_1);
    extract(p1);
    sumadd(m2);
    muladd(m10, SECP256K1_N_C_0);
    muladd(m9, SECP256K1_N_C_1);
    muladd(m8, SECP256K1_N_C_2);
    extract(p2);
    sumadd(m3);
    muladd(m11, SECP256K1_N_C_0);
    muladd(m10, SECP256K1_N_C_1);
    muladd(m9, SECP256K1_N_C_2);
    muladd(m8, SECP256K1_N_C_3);
    extract(p3);
    sumadd(m4);
    muladd(m12, SECP256K1_N_C_0);
    muladd(m11, SECP256K1_N_C_1);
    muladd(m10, SECP256K1_N_C_2);
    muladd(m9, SECP256K1_N_C_3);
    sumadd(m8);
    extract(p4);
    sumadd(m5);
    muladd(m12, SECP256K1_N_C_1);
    muladd(m11, SECP256K1_N_C_2);
    muladd(m10, SECP256K1_N_C_3);
    sumadd(m9);
    extract(p5);
    sumadd(m6);
    muladd(m12, SECP256K1_N_C_2);
    muladd(m11, SECP256K1_N_C_3);
    sumadd(m10);
    extract(p6);
    sumadd_fast(m7);
    muladd_fast(m12, SECP256K1_N_C_3);
    sumadd_fast(m11);
    extract_fast(p7);
    p8 = c0 + m12;
    VERIFY_CHECK(p8 <= 2);

    /* Reduce 258 bits into 256. */
    /* r[0..7] = p[0..7] + p[8] * SECP256K1_N_C. */
    c = p0 + (uint64_t)SECP256K1_N_C_0 * p8;
    r->d[0] = c & 0xFFFFFFFFUL; c >>= 32;
    c += p1 + (uint64_t)SECP256K1_N_C_1 * p8;
    r->d[1] = c & 0xFFFFFFFFUL; c >>= 32;
    c += p2 + (uint64_t)SECP256K1_N_C_2 * p8;
    r->d[2] = c & 0xFFFFFFFFUL; c >>= 32;
    c += p3 + (uint64_t)SECP256K1_N_C_3 * p8;
    r->d[3] = c & 0xFFFFFFFFUL; c >>= 32;
    c += p4 + (uint64_t)p8;
    r->d[4] = c & 0xFFFFFFFFUL; c >>= 32;
    c += p5;
    r->d[5] = c & 0xFFFFFFFFUL; c >>= 32;
    c += p6;
    r->d[6] = c & 0xFFFFFFFFUL; c >>= 32;
    c += p7;
    r->d[7] = c & 0xFFFFFFFFUL; c >>= 32;

    /* Final reduction of r. */
    secp256k1_scalar_reduce(r, c + secp256k1_scalar_check_overflow(r));
}

static void secp256k1_scalar_mul_512(uint32_t *l, const secp256k1_scalar_t *a, const secp256k1_scalar_t *b) {
    /* 96 bit accumulator. */
    uint32_t c0 = 0, c1 = 0, c2 = 0;

    /* l[0..15] = a[0..7] * b[0..7]. */
    muladd_fast(a->d[0], b->d[0]);
    extract_fast(l[0]);
    muladd(a->d[0], b->d[1]);
    muladd(a->d[1], b->d[0]);
    extract(l[1]);
    muladd(a->d[0], b->d[2]);
    muladd(a->d[1], b->d[1]);
    muladd(a->d[2], b->d[0]);
    extract(l[2]);
    muladd(a->d[0], b->d[3]);
    muladd(a->d[1], b->d[2]);
    muladd(a->d[2], b->d[1]);
    muladd(a->d[3], b->d[0]);
    extract(l[3]);
    muladd(a->d[0], b->d[4]);
    muladd(a->d[1], b->d[3]);
    muladd(a->d[2], b->d[2]);
    muladd(a->d[3], b->d[1]);
    muladd(a->d[4], b->d[0]);
    extract(l[4]);
    muladd(a->d[0], b->d[5]);
    muladd(a->d[1], b->d[4]);
    muladd(a->d[2], b->d[3]);
    muladd(a->d[3], b->d[2]);
    muladd(a->d[4], b->d[1]);
    muladd(a->d[5], b->d[0]);
    extract(l[5]);
    muladd(a->d[0], b->d[6]);
    muladd(a->d[1], b->d[5]);
    muladd(a->d[2], b->d[4]);
    muladd(a->d[3], b->d[3]);
    muladd(a->d[4], b->d[2]);
    muladd(a->d[5], b->d[1]);
    muladd(a->d[6], b->d[0]);
    extract(l[6]);
    muladd(a->d[0], b->d[7]);
    muladd(a->d[1], b->d[6]);
    muladd(a->d[2], b->d[5]);
    muladd(a->d[3], b->d[4]);
    muladd(a->d[4], b->d[3]);
    muladd(a->d[5], b->d[2]);
    muladd(a->d[6], b->d[1]);
    muladd(a->d[7], b->d[0]);
    extract(l[7]);
    muladd(a->d[1], b->d[7]);
    muladd(a->d[2], b->d[6]);
    muladd(a->d[3], b->d[5]);
    muladd(a->d[4], b->d[4]);
    muladd(a->d[5], b->d[3]);
    muladd(a->d[6], b->d[2]);
    muladd(a->d[7], b->d[1]);
    extract(l[8]);
    muladd(a->d[2], b->d[7]);
    muladd(a->d[3], b->d[6]);
    muladd(a->d[4], b->d[5]);
    muladd(a->d[5], b->d[4]);
    muladd(a->d[6], b->d[3]);
    muladd(a->d[7], b->d[2]);
    extract(l[9]);
    muladd(a->d[3], b->d[7]);
    muladd(a->d[4], b->d[6]);
    muladd(a->d[5], b->d[5]);
    muladd(a->d[6], b->d[4]);
    muladd(a->d[7], b->d[3]);
    extract(l[10]);
    muladd(a->d[4], b->d[7]);
    muladd(a->d[5], b->d[6]);
    muladd(a->d[6], b->d[5]);
    muladd(a->d[7], b->d[4]);
    extract(l[11]);
    muladd(a->d[5], b->d[7]);
    muladd(a->d[6], b->d[6]);
    muladd(a->d[7], b->d[5]);
    extract(l[12]);
    muladd(a->d[6], b->d[7]);
    muladd(a->d[7], b->d[6]);
    extract(l[13]);
    muladd_fast(a->d[7], b->d[7]);
    extract_fast(l[14]);
    VERIFY_CHECK(c1 == 0);
    l[15] = c0;
}

static void secp256k1_scalar_sqr_512(uint32_t *l, const secp256k1_scalar_t *a) {
    /* 96 bit accumulator. */
    uint32_t c0 = 0, c1 = 0, c2 = 0;

    /* l[0..15] = a[0..7]^2. */
    muladd_fast(a->d[0], a->d[0]);
    extract_fast(l[0]);
    muladd2(a->d[0], a->d[1]);
    extract(l[1]);
    muladd2(a->d[0], a->d[2]);
    muladd(a->d[1], a->d[1]);
    extract(l[2]);
    muladd2(a->d[0], a->d[3]);
    muladd2(a->d[1], a->d[2]);
    extract(l[3]);
    muladd2(a->d[0], a->d[4]);
    muladd2(a->d[1], a->d[3]);
    muladd(a->d[2], a->d[2]);
    extract(l[4]);
    muladd2(a->d[0], a->d[5]);
    muladd2(a->d[1], a->d[4]);
    muladd2(a->d[2], a->d[3]);
    extract(l[5]);
    muladd2(a->d[0], a->d[6]);
    muladd2(a->d[1], a->d[5]);
    muladd2(a->d[2], a->d[4]);
    muladd(a->d[3], a->d[3]);
    extract(l[6]);
    muladd2(a->d[0], a->d[7]);
    muladd2(a->d[1], a->d[6]);
    muladd2(a->d[2], a->d[5]);
    muladd2(a->d[3], a->d[4]);
    extract(l[7]);
    muladd2(a->d[1], a->d[7]);
    muladd2(a->d[2], a->d[6]);
    muladd2(a->d[3], a->d[5]);
    muladd(a->d[4], a->d[4]);
    extract(l[8]);
    muladd2(a->d[2], a->d[7]);
    muladd2(a->d[3], a->d[6]);
    muladd2(a->d[4], a->d[5]);
    extract(l[9]);
    muladd2(a->d[3], a->d[7]);
    muladd2(a->d[4], a->d[6]);
    muladd(a->d[5], a->d[5]);
    extract(l[10]);
    muladd2(a->d[4], a->d[7]);
    muladd2(a->d[5], a->d[6]);
    extract(l[11]);
    muladd2(a->d[5], a->d[7]);
    muladd(a->d[6], a->d[6]);
    extract(l[12]);
    muladd2(a->d[6], a->d[7]);
    extract(l[13]);
    muladd_fast(a->d[7], a->d[7]);
    extract_fast(l[14]);
    VERIFY_CHECK(c1 == 0);
    l[15] = c0;
}

#undef sumadd
#undef sumadd_fast
#undef muladd
#undef muladd_fast
#undef muladd2
#undef extract
#undef extract_fast

static void secp256k1_scalar_mul(secp256k1_scalar_t *r, const secp256k1_scalar_t *a, const secp256k1_scalar_t *b) {
    uint32_t l[16];
    secp256k1_scalar_mul_512(l, a, b);
    secp256k1_scalar_reduce_512(r, l);
}

static void secp256k1_scalar_sqr(secp256k1_scalar_t *r, const secp256k1_scalar_t *a) {
    uint32_t l[16];
    secp256k1_scalar_sqr_512(l, a);
    secp256k1_scalar_reduce_512(r, l);
}

#ifdef USE_ENDOMORPHISM
static void secp256k1_scalar_split_128(secp256k1_scalar_t *r1, secp256k1_scalar_t *r2, const secp256k1_scalar_t *a) {
    r1->d[0] = a->d[0];
    r1->d[1] = a->d[1];
    r1->d[2] = a->d[2];
    r1->d[3] = a->d[3];
    r1->d[4] = 0;
    r1->d[5] = 0;
    r1->d[6] = 0;
    r1->d[7] = 0;
    r2->d[0] = a->d[4];
    r2->d[1] = a->d[5];
    r2->d[2] = a->d[6];
    r2->d[3] = a->d[7];
    r2->d[4] = 0;
    r2->d[5] = 0;
    r2->d[6] = 0;
    r2->d[7] = 0;
}
#endif

SECP256K1_INLINE static int secp256k1_scalar_eq(const secp256k1_scalar_t *a, const secp256k1_scalar_t *b) {
    return ((a->d[0] ^ b->d[0]) | (a->d[1] ^ b->d[1]) | (a->d[2] ^ b->d[2]) | (a->d[3] ^ b->d[3]) | (a->d[4] ^ b->d[4]) | (a->d[5] ^ b->d[5]) | (a->d[6] ^ b->d[6]) | (a->d[7] ^ b->d[7])) == 0;
}

SECP256K1_INLINE static void secp256k1_scalar_mul_shift_var(secp256k1_scalar_t *r, const secp256k1_scalar_t *a, const secp256k1_scalar_t *b, unsigned int shift) {
    uint32_t l[16];
    unsigned int shiftlimbs;
    unsigned int shiftlow;
    unsigned int shifthigh;
    VERIFY_CHECK(shift >= 256);
    secp256k1_scalar_mul_512(l, a, b);
    shiftlimbs = shift >> 5;
    shiftlow = shift & 0x1F;
    shifthigh = 32 - shiftlow;
    r->d[0] = shift < 512 ? (l[0 + shiftlimbs] >> shiftlow | (shift < 480 && shiftlow ? (l[1 + shiftlimbs] << shifthigh) : 0)) : 0;
    r->d[1] = shift < 480 ? (l[1 + shiftlimbs] >> shiftlow | (shift < 448 && shiftlow ? (l[2 + shiftlimbs] << shifthigh) : 0)) : 0;
    r->d[2] = shift < 448 ? (l[2 + shiftlimbs] >> shiftlow | (shift < 416 && shiftlow ? (l[3 + shiftlimbs] << shifthigh) : 0)) : 0;
    r->d[3] = shift < 416 ? (l[3 + shiftlimbs] >> shiftlow | (shift < 384 && shiftlow ? (l[4 + shiftlimbs] << shifthigh) : 0)) : 0;
    r->d[4] = shift < 384 ? (l[4 + shiftlimbs] >> shiftlow | (shift < 352 && shiftlow ? (l[5 + shiftlimbs] << shifthigh) : 0)) : 0;
    r->d[5] = shift < 352 ? (l[5 + shiftlimbs] >> shiftlow | (shift < 320 && shiftlow ? (l[6 + shiftlimbs] << shifthigh) : 0)) : 0;
    r->d[6] = shift < 320 ? (l[6 + shiftlimbs] >> shiftlow | (shift < 288 && shiftlow ? (l[7 + shiftlimbs] << shifthigh) : 0)) : 0;
    r->d[7] = shift < 288 ? (l[7 + shiftlimbs] >> shiftlow)  : 0;
    if ((l[(shift - 1) >> 5] >> ((shift - 1) & 0x1f)) & 1) {
        secp256k1_scalar_add_bit(r, 0);
    }
}

#endif