Repository navigation
Expand file tree
/
Copy pathripemd160_avx2.cpp
More file actions
270 lines (238 loc) · 12.5 KB
/
Copy pathripemd160_avx2.cpp
File metadata and controls
270 lines (238 loc) · 12.5 KB
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
#include "ripemd160_avx2.h"
#include <immintrin.h>
#include <cstring>
#include <cstdint>
namespace ripemd160avx2 {
#ifdef WIN64
static const __declspec(align(32)) uint32_t _init[] = {
#else
static const uint32_t _init[] __attribute__((aligned(32))) = {
#endif
// 8 copies of A
0x67452301ul, 0x67452301ul, 0x67452301ul, 0x67452301ul,
0x67452301ul, 0x67452301ul, 0x67452301ul, 0x67452301ul,
// 8 copies of B
0xEFCDAB89ul, 0xEFCDAB89ul, 0xEFCDAB89ul, 0xEFCDAB89ul,
0xEFCDAB89ul, 0xEFCDAB89ul, 0xEFCDAB89ul, 0xEFCDAB89ul,
// 8 copies of C
0x98BADCFEul, 0x98BADCFEul, 0x98BADCFEul, 0x98BADCFEul,
0x98BADCFEul, 0x98BADCFEul, 0x98BADCFEul, 0x98BADCFEul,
// 8 copies of D
0x10325476ul, 0x10325476ul, 0x10325476ul, 0x10325476ul,
0x10325476ul, 0x10325476ul, 0x10325476ul, 0x10325476ul,
// 8 copies of E
0xC3D2E1F0ul, 0xC3D2E1F0ul, 0xC3D2E1F0ul, 0xC3D2E1F0ul,
0xC3D2E1F0ul, 0xC3D2E1F0ul, 0xC3D2E1F0ul, 0xC3D2E1F0ul
};
// AVX2 operations
#define _mm256_not_si256(x) _mm256_xor_si256((x), _mm256_set1_epi32(-1))
#define ROL(x, n) _mm256_or_si256(_mm256_slli_epi32(x, n), _mm256_srli_epi32(x, 32 - n))
// RIPEMD-160 functions
#define f1(x, y, z) _mm256_xor_si256(x, _mm256_xor_si256(y, z))
#define f2(x, y, z) _mm256_or_si256(_mm256_and_si256(x, y), _mm256_andnot_si256(x, z))
#define f3(x, y, z) _mm256_xor_si256(_mm256_or_si256(x, _mm256_not_si256(y)), z)
#define f4(x, y, z) _mm256_or_si256(_mm256_and_si256(x, z), _mm256_andnot_si256(z, y))
#define f5(x, y, z) _mm256_xor_si256(x, _mm256_or_si256(y, _mm256_not_si256(z)))
// Adding helpers
#define add3(x0, x1, x2) _mm256_add_epi32(_mm256_add_epi32(x0, x1), x2)
#define add4(x0, x1, x2, x3) _mm256_add_epi32(_mm256_add_epi32(x0, x1), _mm256_add_epi32(x2, x3))
// Round function
#define Round(a, b, c, d, e, f, x, k, r) \
do { \
__m256i u = add4(a, f, x, _mm256_set1_epi32(k)); \
a = _mm256_add_epi32(ROL(u, r), e); \
c = ROL(c, 10); \
} while(0)
// Macroses for each round
#define R11(a, b, c, d, e, x, r) Round(a, b, c, d, e, f1(b, c, d), x, 0, r)
#define R21(a, b, c, d, e, x, r) Round(a, b, c, d, e, f2(b, c, d), x, 0x5A827999ul, r)
#define R31(a, b, c, d, e, x, r) Round(a, b, c, d, e, f3(b, c, d), x, 0x6ED9EBA1ul, r)
#define R41(a, b, c, d, e, x, r) Round(a, b, c, d, e, f4(b, c, d), x, 0x8F1BBCDCul, r)
#define R51(a, b, c, d, e, x, r) Round(a, b, c, d, e, f5(b, c, d), x, 0xA953FD4Eul, r)
#define R12(a, b, c, d, e, x, r) Round(a, b, c, d, e, f5(b, c, d), x, 0x50A28BE6ul, r)
#define R22(a, b, c, d, e, x, r) Round(a, b, c, d, e, f4(b, c, d), x, 0x5C4DD124ul, r)
#define R32(a, b, c, d, e, x, r) Round(a, b, c, d, e, f3(b, c, d), x, 0x6D703EF3ul, r)
#define R42(a, b, c, d, e, x, r) Round(a, b, c, d, e, f2(b, c, d), x, 0x7A6D76E9ul, r)
#define R52(a, b, c, d, e, x, r) Round(a, b, c, d, e, f1(b, c, d), x, 0, r)
// Macros to load words from the message blocks
#define LOADW(i) _mm256_set_epi32(*((uint32_t *)blk[0] + i), *((uint32_t *)blk[1] + i), \
*((uint32_t *)blk[2] + i), *((uint32_t *)blk[3] + i), \
*((uint32_t *)blk[4] + i), *((uint32_t *)blk[5] + i), \
*((uint32_t *)blk[6] + i), *((uint32_t *)blk[7] + i))
inline void Initialize(__m256i *s) {
_mm256_store_si256(s + 0, _mm256_load_si256((const __m256i*)(_init + 0)));
_mm256_store_si256(s + 1, _mm256_load_si256((const __m256i*)(_init + 8)));
_mm256_store_si256(s + 2, _mm256_load_si256((const __m256i*)(_init + 16)));
_mm256_store_si256(s + 3, _mm256_load_si256((const __m256i*)(_init + 24)));
_mm256_store_si256(s + 4, _mm256_load_si256((const __m256i*)(_init + 32)));
}
// Transform function processes one block for each message
void Transform(__m256i *s, uint8_t *blk[8]) {
// Load state variables
__m256i a1 = _mm256_load_si256(s + 0);
__m256i b1 = _mm256_load_si256(s + 1);
__m256i c1 = _mm256_load_si256(s + 2);
__m256i d1 = _mm256_load_si256(s + 3);
__m256i e1 = _mm256_load_si256(s + 4);
// Initialize second set of variables
__m256i a2 = a1;
__m256i b2 = b1;
__m256i c2 = c1;
__m256i d2 = d1;
__m256i e2 = e1;
__m256i u;
__m256i w[16];
for (int i = 0; i < 16; ++i) {
w[i] = LOADW(i);
}
// Main rounds 0-15 of Riepmd160
// Round 1
R11(a1, b1, c1, d1, e1, w[0], 11); R12(a2, b2, c2, d2, e2, w[5], 8);
R11(e1, a1, b1, c1, d1, w[1], 14); R12(e2, a2, b2, c2, d2, w[14], 9);
R11(d1, e1, a1, b1, c1, w[2], 15); R12(d2, e2, a2, b2, c2, w[7], 9);
R11(c1, d1, e1, a1, b1, w[3], 12); R12(c2, d2, e2, a2, b2, w[0], 11);
R11(b1, c1, d1, e1, a1, w[4], 5); R12(b2, c2, d2, e2, a2, w[9], 13);
R11(a1, b1, c1, d1, e1, w[5], 8); R12(a2, b2, c2, d2, e2, w[2], 15);
R11(e1, a1, b1, c1, d1, w[6], 7); R12(e2, a2, b2, c2, d2, w[11], 15);
R11(d1, e1, a1, b1, c1, w[7], 9); R12(d2, e2, a2, b2, c2, w[4], 5);
R11(c1, d1, e1, a1, b1, w[8], 11); R12(c2, d2, e2, a2, b2, w[13], 7);
R11(b1, c1, d1, e1, a1, w[9], 13); R12(b2, c2, d2, e2, a2, w[6], 7);
R11(a1, b1, c1, d1, e1, w[10], 14);R12(a2, b2, c2, d2, e2, w[15], 8);
R11(e1, a1, b1, c1, d1, w[11], 15);R12(e2, a2, b2, c2, d2, w[8], 11);
R11(d1, e1, a1, b1, c1, w[12], 6); R12(d2, e2, a2, b2, c2, w[1], 14);
R11(c1, d1, e1, a1, b1, w[13], 7); R12(c2, d2, e2, a2, b2, w[10], 14);
R11(b1, c1, d1, e1, a1, w[14], 9); R12(b2, c2, d2, e2, a2, w[3], 12);
R11(a1, b1, c1, d1, e1, w[15], 8); R12(a2, b2, c2, d2, e2, w[12], 6);
// Round 2
R21(e1, a1, b1, c1, d1, w[7], 7); R22(e2, a2, b2, c2, d2, w[6], 9);
R21(d1, e1, a1, b1, c1, w[4], 6); R22(d2, e2, a2, b2, c2, w[11], 13);
R21(c1, d1, e1, a1, b1, w[13], 8); R22(c2, d2, e2, a2, b2, w[3], 15);
R21(b1, c1, d1, e1, a1, w[1], 13); R22(b2, c2, d2, e2, a2, w[7], 7);
R21(a1, b1, c1, d1, e1, w[10], 11);R22(a2, b2, c2, d2, e2, w[0], 12);
R21(e1, a1, b1, c1, d1, w[6], 9); R22(e2, a2, b2, c2, d2, w[13], 8);
R21(d1, e1, a1, b1, c1, w[15], 7); R22(d2, e2, a2, b2, c2, w[5], 9);
R21(c1, d1, e1, a1, b1, w[3], 15);R22(c2, d2, e2, a2, b2, w[10], 11);
R21(b1, c1, d1, e1, a1, w[12], 7); R22(b2, c2, d2, e2, a2, w[14], 7);
R21(a1, b1, c1, d1, e1, w[0], 12);R22(a2, b2, c2, d2, e2, w[15], 7);
R21(e1, a1, b1, c1, d1, w[9], 15); R22(e2, a2, b2, c2, d2, w[8], 12);
R21(d1, e1, a1, b1, c1, w[5], 9); R22(d2, e2, a2, b2, c2, w[12], 7);
R21(c1, d1, e1, a1, b1, w[2], 11);R22(c2, d2, e2, a2, b2, w[4], 6);
R21(b1, c1, d1, e1, a1, w[14], 7);R22(b2, c2, d2, e2, a2, w[9], 15);
R21(a1, b1, c1, d1, e1, w[11], 13);R22(a2, b2, c2, d2, e2, w[1], 13);
R21(e1, a1, b1, c1, d1, w[8], 12); R22(e2, a2, b2, c2, d2, w[2], 11);
// Round 3
R31(d1, e1, a1, b1, c1, w[3], 11); R32(d2, e2, a2, b2, c2, w[15], 9);
R31(c1, d1, e1, a1, b1, w[10], 13);R32(c2, d2, e2, a2, b2, w[5], 7);
R31(b1, c1, d1, e1, a1, w[14], 6); R32(b2, c2, d2, e2, a2, w[1], 15);
R31(a1, b1, c1, d1, e1, w[4], 7); R32(a2, b2, c2, d2, e2, w[3], 11);
R31(e1, a1, b1, c1, d1, w[9], 14); R32(e2, a2, b2, c2, d2, w[7], 8);
R31(d1, e1, a1, b1, c1, w[15], 9); R32(d2, e2, a2, b2, c2, w[14], 6);
R31(c1, d1, e1, a1, b1, w[8], 13); R32(c2, d2, e2, a2, b2, w[6], 6);
R31(b1, c1, d1, e1, a1, w[1], 15); R32(b2, c2, d2, e2, a2, w[9], 14);
R31(a1, b1, c1, d1, e1, w[2], 14); R32(a2, b2, c2, d2, e2, w[11], 12);
R31(e1, a1, b1, c1, d1, w[7], 8); R32(e2, a2, b2, c2, d2, w[8], 13);
R31(d1, e1, a1, b1, c1, w[0], 13);R32(d2, e2, a2, b2, c2, w[12], 5);
R31(c1, d1, e1, a1, b1, w[6], 6); R32(c2, d2, e2, a2, b2, w[2], 14);
R31(b1, c1, d1, e1, a1, w[13], 5);R32(b2, c2, d2, e2, a2, w[10], 13);
R31(a1, b1, c1, d1, e1, w[11], 12);R32(a2, b2, c2, d2, e2, w[0], 13);
R31(e1, a1, b1, c1, d1, w[5], 7); R32(e2, a2, b2, c2, d2, w[4], 7);
R31(d1, e1, a1, b1, c1, w[12], 5); R32(d2, e2, a2, b2, c2, w[13], 5);
// Round 4
R41(c1, d1, e1, a1, b1, w[1], 11); R42(c2, d2, e2, a2, b2, w[8], 15);
R41(b1, c1, d1, e1, a1, w[9], 12); R42(b2, c2, d2, e2, a2, w[6], 5);
R41(a1, b1, c1, d1, e1, w[11], 14);R42(a2, b2, c2, d2, e2, w[4], 8);
R41(e1, a1, b1, c1, d1, w[10], 15);R42(e2, a2, b2, c2, d2, w[1], 11);
R41(d1, e1, a1, b1, c1, w[0], 14); R42(d2, e2, a2, b2, c2, w[3], 14);
R41(c1, d1, e1, a1, b1, w[8], 15); R42(c2, d2, e2, a2, b2, w[11], 14);
R41(b1, c1, d1, e1, a1, w[12], 9); R42(b2, c2, d2, e2, a2, w[15], 6);
R41(a1, b1, c1, d1, e1, w[4], 8); R42(a2, b2, c2, d2, e2, w[0], 14);
R41(e1, a1, b1, c1, d1, w[13], 9);R42(e2, a2, b2, c2, d2, w[5], 6);
R41(d1, e1, a1, b1, c1, w[3], 14);R42(d2, e2, a2, b2, c2, w[12], 9);
R41(c1, d1, e1, a1, b1, w[7], 5); R42(c2, d2, e2, a2, b2, w[2], 12);
R41(b1, c1, d1, e1, a1, w[15], 6);R42(b2, c2, d2, e2, a2, w[13], 9);
R41(a1, b1, c1, d1, e1, w[14], 8); R42(a2, b2, c2, d2, e2, w[9], 12);
R41(e1, a1, b1, c1, d1, w[5], 6); R42(e2, a2, b2, c2, d2, w[7], 5);
R41(d1, e1, a1, b1, c1, w[6], 5); R42(d2, e2, a2, b2, c2, w[10], 15);
R41(c1, d1, e1, a1, b1, w[2], 12); R42(c2, d2, e2, a2, b2, w[14], 8);
// Round 5
R51(b1, c1, d1, e1, a1, w[4], 9); R52(b2, c2, d2, e2, a2, w[12], 8);
R51(a1, b1, c1, d1, e1, w[0], 15); R52(a2, b2, c2, d2, e2, w[15], 5);
R51(e1, a1, b1, c1, d1, w[5], 5); R52(e2, a2, b2, c2, d2, w[10], 12);
R51(d1, e1, a1, b1, c1, w[9], 11); R52(d2, e2, a2, b2, c2, w[4], 9);
R51(c1, d1, e1, a1, b1, w[7], 6); R52(c2, d2, e2, a2, b2, w[1], 12);
R51(b1, c1, d1, e1, a1, w[12], 8); R52(b2, c2, d2, e2, a2, w[5], 5);
R51(a1, b1, c1, d1, e1, w[2], 13); R52(a2, b2, c2, d2, e2, w[8], 14);
R51(e1, a1, b1, c1, d1, w[10], 12);R52(e2, a2, b2, c2, d2, w[7], 6);
R51(d1, e1, a1, b1, c1, w[14], 5);R52(d2, e2, a2, b2, c2, w[6], 8);
R51(c1, d1, e1, a1, b1, w[1], 12); R52(c2, d2, e2, a2, b2, w[2], 13);
R51(b1, c1, d1, e1, a1, w[3], 13); R52(b2, c2, d2, e2, a2, w[13], 6);
R51(a1, b1, c1, d1, e1, w[8], 14); R52(a2, b2, c2, d2, e2, w[14], 5);
R51(e1, a1, b1, c1, d1, w[11], 11);R52(e2, a2, b2, c2, d2, w[0], 15);
R51(d1, e1, a1, b1, c1, w[6], 8); R52(d2, e2, a2, b2, c2, w[3], 13);
R51(c1, d1, e1, a1, b1, w[15], 5);R52(c2, d2, e2, a2, b2, w[9], 11);
R51(b1, c1, d1, e1, a1, w[13], 6);R52(b2, c2, d2, e2, a2, w[11], 11);
// Combine results and update state
__m256i t = _mm256_load_si256(s + 0);
_mm256_store_si256(s + 0, add3(_mm256_load_si256(s + 1), c1, d2));
_mm256_store_si256(s + 1, add3(_mm256_load_si256(s + 2), d1, e2));
_mm256_store_si256(s + 2, add3(_mm256_load_si256(s + 3), e1, a2));
_mm256_store_si256(s + 3, add3(_mm256_load_si256(s + 4), a1, b2));
_mm256_store_si256(s + 4, add3(t, b1, c2));
}
#ifdef WIN64
#define DEPACK(d, i) \
((uint32_t *)d)[0] = _mm256_extract_epi32(s[0], i); \
((uint32_t *)d)[1] = _mm256_extract_epi32(s[1], i); \
((uint32_t *)d)[2] = _mm256_extract_epi32(s[2], i); \
((uint32_t *)d)[3] = _mm256_extract_epi32(s[3], i); \
((uint32_t *)d)[4] = _mm256_extract_epi32(s[4], i);
#else
#define DEPACK(d, i) \
((uint32_t *)d)[0] = ((uint32_t *)&s[0])[i]; \
((uint32_t *)d)[1] = ((uint32_t *)&s[1])[i]; \
((uint32_t *)d)[2] = ((uint32_t *)&s[2])[i]; \
((uint32_t *)d)[3] = ((uint32_t *)&s[3])[i]; \
((uint32_t *)d)[4] = ((uint32_t *)&s[4])[i];
#endif
static const uint64_t sizedesc_32 = 32 << 3;
static const unsigned char pad[64] = { 0x80 };
// Main function to compute Rimepd160 hash for 8 messages of 32 bytes each
void ripemd160avx2_32(
unsigned char *i0, unsigned char *i1,
unsigned char *i2, unsigned char *i3,
unsigned char *i4, unsigned char *i5,
unsigned char *i6, unsigned char *i7,
unsigned char *d0, unsigned char *d1,
unsigned char *d2, unsigned char *d3,
unsigned char *d4, unsigned char *d5,
unsigned char *d6, unsigned char *d7)
{
__m256i s[5];
uint8_t *bs[] = { i0, i1, i2, i3, i4, i5, i6, i7 };
// Initialize state
ripemd160avx2::Initialize(s);
// Add padding and length
for (int i = 0; i < 8; ++i) {
memcpy(bs[i] + 32, pad, 24);
memcpy(bs[i] + 56, &sizedesc_32, 8);
}
// Process message blocks
ripemd160avx2::Transform(s, bs);
#ifndef WIN64
uint32_t *s0 = (uint32_t *)&s[0];
uint32_t *s1 = (uint32_t *)&s[1];
uint32_t *s2 = (uint32_t *)&s[2];
uint32_t *s3 = (uint32_t *)&s[3];
uint32_t *s4 = (uint32_t *)&s[4];
#endif
// Unpack the hash values to the output buffers
DEPACK(d0, 7);
DEPACK(d1, 6);
DEPACK(d2, 5);
DEPACK(d3, 4);
DEPACK(d4, 3);
DEPACK(d5, 2);
DEPACK(d6, 1);
DEPACK(d7, 0);
}
} // namespace ripemd160avx2