Debian Codegraph @

crypto_kem/348864f/avx/vec.c

raw
1 // 20240805 djb: more cryptoint usage
2 // 20240508 djb: switch to vec_mul_gf_using_64
3 // 20221230 djb: add linker line
4
5 // linker define vec_GF_mul
6 // linker use vec_mul_asm
7
8 #include "vec.h"
9 #include "crypto_int64.h"
10
11 /* input: v, an element in GF(2^m)[y]/(y^64+y^3+y+z) in bitsliced form */
12 /* input: a, an element in GF(2^m)[y]/(y^64+y^3+y+z) as an array of coefficients */
13 /* output: out, the product of v and a in bitsliced form */
14 void vec_GF_mul(vec out[ GFBITS ], vec v[ GFBITS ], gf a[ SYS_T ])
15 {
16 int i, j;
17 vec buf[GFBITS][2], prod[GFBITS], tmp[GFBITS];
18
19 const vec allone = -1;
20
21 // polynomial multiplication
22
23 for (i = 0; i < GFBITS; i++)
24 {
25 buf[i][0] = 0;
26 buf[i][1] = 0;
27 }
28
29 for (i = SYS_T-1; i >= 0; i--)
30 {
31 for (j = 0; j < GFBITS; j++)
32 {
33 buf[j][1] <<= 1;
34 buf[j][1] |= crypto_int64_negative_01(buf[j][0]);
35 buf[j][0] <<= 1;
36 }
37
38 vec_mul_gf_using_64(prod, v, a[i]);
39
40 for (j = 0; j < GFBITS; j++)
41 buf[j][0] ^= prod[j];
42 }
43
44 // reduction modulo y^64 + y^3 + y + z
45
46 for (i = 0; i < GFBITS; i++)
47 prod[i] = buf[i][1] & (allone << 3);
48
49 vec_mul_gf_using_64(tmp, prod, 2);
50
51 for (i = 0; i < GFBITS; i++)
52 {
53 buf[i][1] ^= prod[i] >> (SYS_T - 3);
54 buf[i][0] ^= prod[i] << (64 - (SYS_T - 3));
55 buf[i][1] ^= prod[i] >> (SYS_T - 1);
56 buf[i][0] ^= prod[i] << (64 - (SYS_T - 1));
57 buf[i][0] ^= tmp[i];
58 }
59
60 //
61
62 for (i = 0; i < GFBITS; i++)
63 prod[i] = buf[i][1] & 0x7;
64
65 vec_mul_gf_using_64(tmp, prod, 2);
66
67 for (i = 0; i < GFBITS; i++)
68 {
69 buf[i][0] ^= prod[i] << (64 - (SYS_T - 3));
70 buf[i][0] ^= prod[i] << (64 - (SYS_T - 1));
71 buf[i][0] ^= tmp[i];
72 }
73
74 //
75
76 for (i = 0; i < GFBITS; i++)
77 out[i] = buf[i][0];
78 }
79