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
|
/* Copyright (c) 2014 Cryptography Research, Inc.
* Released under the MIT License. See LICENSE.txt for license information.
*/
#include "f_field.h"
#if (defined(__OPTIMIZE__) && !defined(__OPTIMIZE_SIZE__) && !I_HATE_UNROLLED_LOOPS) \
|| defined(DECAF_FORCE_UNROLL)
#define REPEAT8(_x) _x _x _x _x _x _x _x _x
#define FOR_LIMB(_i,_start,_end,_x) do { _i=_start; REPEAT8( if (_i<_end) { _x; } _i++;) } while (0)
#else
#define FOR_LIMB(_i,_start,_end,_x) do { for (_i=_start; _i<_end; _i++) _x; } while (0)
#endif
void gf_mul (gf_s *__restrict__ cs, const gf as, const gf bs) {
const uint32_t *a = as->limb, *b = bs->limb;
uint32_t *c = cs->limb;
uint64_t accum0 = 0, accum1 = 0, accum2 = 0;
uint32_t mask = (1<<28) - 1;
uint32_t aa[8], bb[8];
int i,j;
for (i=0; i<8; i++) {
aa[i] = a[i] + a[i+8];
bb[i] = b[i] + b[i+8];
}
FOR_LIMB(j,0,8,{
accum2 = 0;
FOR_LIMB (i,0,j+1,{
accum2 += widemul(a[j-i],b[i]);
accum1 += widemul(aa[j-i],bb[i]);
accum0 += widemul(a[8+j-i], b[8+i]);
});
accum1 -= accum2;
accum0 += accum2;
accum2 = 0;
FOR_LIMB (i,j+1,8,{
accum0 -= widemul(a[8+j-i], b[i]);
accum2 += widemul(aa[8+j-i], bb[i]);
accum1 += widemul(a[16+j-i], b[8+i]);
});
accum1 += accum2;
accum0 += accum2;
c[j] = ((uint32_t)(accum0)) & mask;
c[j+8] = ((uint32_t)(accum1)) & mask;
accum0 >>= 28;
accum1 >>= 28;
});
accum0 += accum1;
accum0 += c[8];
accum1 += c[0];
c[8] = ((uint32_t)(accum0)) & mask;
c[0] = ((uint32_t)(accum1)) & mask;
accum0 >>= 28;
accum1 >>= 28;
c[9] += ((uint32_t)(accum0));
c[1] += ((uint32_t)(accum1));
}
void gf_mulw_unsigned (gf_s *__restrict__ cs, const gf as, uint32_t b) {
assert(b<1<<28);
const uint32_t *a = as->limb;
uint32_t *c = cs->limb;
uint64_t accum0 = 0, accum8 = 0;
uint32_t mask = (1ull<<28)-1;
int i;
FOR_LIMB(i,0,8,{
accum0 += widemul(b, a[i]);
accum8 += widemul(b, a[i+8]);
c[i] = accum0 & mask; accum0 >>= 28;
c[i+8] = accum8 & mask; accum8 >>= 28;
});
accum0 += accum8 + c[8];
c[8] = accum0 & mask;
c[9] += accum0 >> 28;
accum8 += c[0];
c[0] = accum8 & mask;
c[1] += accum8 >> 28;
}
void gf_sqr (gf_s *__restrict__ cs, const gf as) {
gf_mul(cs,as,as); /* Performs better with a dedicated square */
}
|
