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
|
#pragma once
#include <emmintrin.h>
#include <tmmintrin.h>
#include "blake2b.h"
#include <library/cpp/digest/argonish/internal/rotations/rotations_ssse3.h>
namespace NArgonish {
template <>
void* TBlake2B<EInstructionSet::SSSE3>::GetIV_() const {
static const __m128i Iv[4] = {
_mm_set_epi64x(0xbb67ae8584caa73bULL, 0x6a09e667f3bcc908ULL),
_mm_set_epi64x(0xa54ff53a5f1d36f1ULL, 0x3c6ef372fe94f82bULL),
_mm_set_epi64x(0x9b05688c2b3e6c1fULL, 0x510e527fade682d1ULL),
_mm_set_epi64x(0x5be0cd19137e2179ULL, 0x1f83d9abfb41bd6bULL)};
return (void*)Iv;
}
static const ui32 Sigma[12][16] = {
{0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15},
{14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3},
{11, 8, 12, 0, 5, 2, 15, 13, 10, 14, 3, 6, 7, 1, 9, 4},
{7, 9, 3, 1, 13, 12, 11, 14, 2, 6, 5, 10, 4, 0, 15, 8},
{9, 0, 5, 7, 2, 4, 10, 15, 14, 1, 11, 12, 6, 8, 3, 13},
{2, 12, 6, 10, 0, 11, 8, 3, 4, 13, 7, 5, 15, 14, 1, 9},
{12, 5, 1, 15, 14, 13, 4, 10, 0, 7, 6, 3, 9, 2, 8, 11},
{13, 11, 7, 14, 12, 1, 3, 9, 5, 0, 15, 4, 8, 6, 2, 10},
{6, 15, 14, 9, 11, 3, 0, 8, 12, 2, 13, 7, 1, 4, 10, 5},
{10, 2, 8, 4, 7, 6, 1, 5, 15, 11, 9, 14, 3, 12, 13, 0},
{0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15},
{14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3}};
static inline void G1(
__m128i& row1l, __m128i& row2l, __m128i& row3l, __m128i& row4l,
__m128i& row1h, __m128i& row2h, __m128i& row3h, __m128i& row4h,
__m128i& b0, __m128i& b1) {
row1l = _mm_add_epi64(_mm_add_epi64(row1l, b0), row2l);
row1h = _mm_add_epi64(_mm_add_epi64(row1h, b1), row2h);
row4l = _mm_xor_si128(row4l, row1l);
row4h = _mm_xor_si128(row4h, row1h);
row4l = Rotr32(row4l);
row4h = Rotr32(row4h);
row3l = _mm_add_epi64(row3l, row4l);
row3h = _mm_add_epi64(row3h, row4h);
row2l = _mm_xor_si128(row2l, row3l);
row2h = _mm_xor_si128(row2h, row3h);
row2l = Rotr24(row2l);
row2h = Rotr24(row2h);
}
static inline void G2(
__m128i& row1l, __m128i& row2l, __m128i& row3l, __m128i& row4l,
__m128i& row1h, __m128i& row2h, __m128i& row3h, __m128i& row4h,
__m128i& b0, __m128i& b1) {
row1l = _mm_add_epi64(_mm_add_epi64(row1l, b0), row2l);
row1h = _mm_add_epi64(_mm_add_epi64(row1h, b1), row2h);
row4l = _mm_xor_si128(row4l, row1l);
row4h = _mm_xor_si128(row4h, row1h);
row4l = Rotr16(row4l);
row4h = Rotr16(row4h);
row3l = _mm_add_epi64(row3l, row4l);
row3h = _mm_add_epi64(row3h, row4h);
row2l = _mm_xor_si128(row2l, row3l);
row2h = _mm_xor_si128(row2h, row3h);
row2l = Rotr63(row2l);
row2h = Rotr63(row2h);
}
static inline void Diagonalize(
__m128i& row2l, __m128i& row3l, __m128i& row4l,
__m128i& row2h, __m128i& row3h, __m128i& row4h) {
__m128i t0 = _mm_alignr_epi8(row2h, row2l, 8);
__m128i t1 = _mm_alignr_epi8(row2l, row2h, 8);
row2l = t0;
row2h = t1;
t0 = row3l;
row3l = row3h;
row3h = t0;
t0 = _mm_alignr_epi8(row4h, row4l, 8);
t1 = _mm_alignr_epi8(row4l, row4h, 8);
row4l = t1;
row4h = t0;
}
static inline void Undiagonalize(
__m128i& row2l, __m128i& row3l, __m128i& row4l,
__m128i& row2h, __m128i& row3h, __m128i& row4h) {
__m128i t0 = _mm_alignr_epi8(row2l, row2h, 8);
__m128i t1 = _mm_alignr_epi8(row2h, row2l, 8);
row2l = t0;
row2h = t1;
t0 = row3l;
row3l = row3h;
row3h = t0;
t0 = _mm_alignr_epi8(row4l, row4h, 8);
t1 = _mm_alignr_epi8(row4h, row4l, 8);
row4l = t1;
row4h = t0;
}
static inline void Round(int r, const ui64* block_ptr,
__m128i& row1l, __m128i& row2l, __m128i& row3l, __m128i& row4l,
__m128i& row1h, __m128i& row2h, __m128i& row3h, __m128i& row4h) {
__m128i b0, b1;
b0 = _mm_set_epi64x(block_ptr[Sigma[r][2]], block_ptr[Sigma[r][0]]);
b1 = _mm_set_epi64x(block_ptr[Sigma[r][6]], block_ptr[Sigma[r][4]]);
G1(row1l, row2l, row3l, row4l, row1h, row2h, row3h, row4h, b0, b1);
b0 = _mm_set_epi64x(block_ptr[Sigma[r][3]], block_ptr[Sigma[r][1]]);
b1 = _mm_set_epi64x(block_ptr[Sigma[r][7]], block_ptr[Sigma[r][5]]);
G2(row1l, row2l, row3l, row4l, row1h, row2h, row3h, row4h, b0, b1);
Diagonalize(row2l, row3l, row4l, row2h, row3h, row4h);
b0 = _mm_set_epi64x(block_ptr[Sigma[r][10]], block_ptr[Sigma[r][8]]);
b1 = _mm_set_epi64x(block_ptr[Sigma[r][14]], block_ptr[Sigma[r][12]]);
G1(row1l, row2l, row3l, row4l, row1h, row2h, row3h, row4h, b0, b1);
b0 = _mm_set_epi64x(block_ptr[Sigma[r][11]], block_ptr[Sigma[r][9]]);
b1 = _mm_set_epi64x(block_ptr[Sigma[r][15]], block_ptr[Sigma[r][13]]);
G2(row1l, row2l, row3l, row4l, row1h, row2h, row3h, row4h, b0, b1);
Undiagonalize(row2l, row3l, row4l, row2h, row3h, row4h);
}
template <>
void TBlake2B<EInstructionSet::SSSE3>::InitialXor_(ui8* h, const ui8* p) {
__m128i* m_res = (__m128i*)h;
const __m128i* m_p = (__m128i*)p;
__m128i* iv = (__m128i*)GetIV_();
_mm_storeu_si128(m_res + 0, _mm_xor_si128(iv[0], _mm_loadu_si128(m_p + 0)));
_mm_storeu_si128(m_res + 1, _mm_xor_si128(iv[1], _mm_loadu_si128(m_p + 1)));
_mm_storeu_si128(m_res + 2, _mm_xor_si128(iv[2], _mm_loadu_si128(m_p + 2)));
_mm_storeu_si128(m_res + 3, _mm_xor_si128(iv[3], _mm_loadu_si128(m_p + 3)));
}
template <>
void TBlake2B<EInstructionSet::SSSE3>::Compress_(const ui64 block[BLAKE2B_BLOCKQWORDS]) {
__m128i* iv = (__m128i*)GetIV_();
__m128i row1l = _mm_loadu_si128((__m128i*)&State_.H[0]);
__m128i row1h = _mm_loadu_si128((__m128i*)&State_.H[2]);
__m128i row2l = _mm_loadu_si128((__m128i*)&State_.H[4]);
__m128i row2h = _mm_loadu_si128((__m128i*)&State_.H[6]);
__m128i row3l = iv[0];
__m128i row3h = iv[1];
__m128i row4l = _mm_xor_si128(iv[2], _mm_loadu_si128((__m128i*)&State_.T[0]));
__m128i row4h = _mm_xor_si128(iv[3], _mm_loadu_si128((__m128i*)&State_.F[0]));
for (int r = 0; r < 12; ++r)
Round(r, block, row1l, row2l, row3l, row4l, row1h, row2h, row3h, row4h);
_mm_storeu_si128((__m128i*)&State_.H[0],
_mm_xor_si128(_mm_loadu_si128((__m128i*)&State_.H[0]), _mm_xor_si128(row3l, row1l)));
_mm_storeu_si128((__m128i*)&State_.H[2],
_mm_xor_si128(_mm_loadu_si128((__m128i*)&State_.H[2]), _mm_xor_si128(row3h, row1h)));
_mm_storeu_si128((__m128i*)&State_.H[4],
_mm_xor_si128(_mm_loadu_si128((__m128i*)&State_.H[4]), _mm_xor_si128(row4l, row2l)));
_mm_storeu_si128((__m128i*)&State_.H[6],
_mm_xor_si128(_mm_loadu_si128((__m128i*)&State_.H[6]), _mm_xor_si128(row4h, row2h)));
}
}
|
