-
-
Notifications
You must be signed in to change notification settings - Fork 178
/
aesni-hash.h
executable file
·151 lines (135 loc) · 4 KB
/
aesni-hash.h
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
// Copyright (c) 2023, Ruan Kunliang.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
#pragma once
#ifndef AESNI_HASH_H
#define AESNI_HASH_H
#include <stdint.h>
#include <immintrin.h>
#ifdef __cplusplus
extern "C" {
#endif
static __m128i AESNI_Hash128(const uint8_t* msg, unsigned len, uint32_t seed = 0) {
auto a = _mm_set1_epi32(seed);
auto b = _mm_set1_epi32(len);
auto m = _mm_set_epi32(0xdeadbeef, 0xffff0000, 0x01234567, 0x89abcdef);
auto s = _mm_set_epi8(3, 7, 11, 15, 2, 6, 10, 14, 1, 5, 9, 13, 0, 4, 8, 12);
bool greed = (((uintptr_t)msg + (len-1)) & 0xfffUL) >= 15; // do not cross page
if (len > 80) {
auto c = _mm_aesenc_si128(b, m);
auto d = _mm_aesdec_si128(a, m);
a = _mm_aesenc_si128(a, m);
b = _mm_aesdec_si128(b, m);
do {
a = _mm_xor_si128(a, _mm_lddqu_si128((const __m128i*)msg));
b = _mm_xor_si128(b, _mm_lddqu_si128((const __m128i*)(msg + 16)));
c = _mm_xor_si128(c, _mm_lddqu_si128((const __m128i*)(msg + 32)));
d = _mm_xor_si128(d, _mm_lddqu_si128((const __m128i*)(msg + 48)));
a = _mm_shuffle_epi8(_mm_aesenc_si128(a, m), s);
b = _mm_shuffle_epi8(_mm_aesdec_si128(b, m), s);
c = _mm_shuffle_epi8(_mm_aesenc_si128(c, m), s);
d = _mm_shuffle_epi8(_mm_aesdec_si128(d, m), s);
msg += 64;
len -= 64;
} while (len > 80);
c = _mm_aesenc_si128(a, c);
d = _mm_aesdec_si128(b, d);
a = _mm_aesenc_si128(c, d);
b = _mm_aesdec_si128(d, c);
}
auto mix = [&a, &b, m, s](__m128i x) {
a = _mm_aesenc_si128(x, a);
a = _mm_aesenc_si128(a, m);
b = _mm_shuffle_epi8(_mm_xor_si128(x, b), s);
b = _mm_shuffle_epi8(_mm_aesdec_si128(b, m), s);
};
while (len >= 16) {
mix(_mm_lddqu_si128((const __m128i*)msg));
msg += 16;
len -= 16;
}
if (greed) {
#define GREEDILY_READ(n, addr) \
_mm_bsrli_si128(_mm_bslli_si128(_mm_lddqu_si128((const __m128i*)addr), (16-(n))), (16-(n)))
switch (len) {
case 15: mix(GREEDILY_READ(15,msg)); break;
case 14: mix(GREEDILY_READ(14,msg)); break;
case 13: mix(GREEDILY_READ(13,msg)); break;
case 12: mix(GREEDILY_READ(12,msg)); break;
case 11: mix(GREEDILY_READ(11,msg)); break;
case 10: mix(GREEDILY_READ(10,msg)); break;
case 9: mix(GREEDILY_READ(9,msg)); break;
case 8: mix((__m128i)_mm_load_sd((const double*)msg)); break;
case 7: mix(GREEDILY_READ(7,msg)); break;
case 6: mix(GREEDILY_READ(6,msg)); break;
case 5: mix(GREEDILY_READ(5,msg)); break;
case 4: mix((__m128i)_mm_load_ss((const float*)msg)); break;
case 3: mix(GREEDILY_READ(3,msg)); break;
case 2: mix(GREEDILY_READ(2,msg)); break;
case 1: mix(GREEDILY_READ(1,msg)); break;
case 0:
default: // try to keep m & s from register spilling
a = _mm_add_epi8(a, s);
b = _mm_add_epi8(b, m);
}
#undef GREEDILY_READ
return _mm_aesenc_si128(a, b);
}
uint64_t x = 0;
switch (len) {
case 15:
x |= ((uint64_t)msg[14]) << 48U;
case 14:
x |= ((uint64_t)msg[13]) << 40U;
case 13:
x |= ((uint64_t)msg[12]) << 32U;
case 12:
x |= *(const uint32_t*)(msg + 8);
mix(_mm_set_epi64x(x, *(const uint64_t*)msg));
break;
case 11:
x |= ((uint32_t)msg[10]) << 16U;
case 10:
x |= ((uint32_t)msg[9]) << 8U;
case 9:
x |= msg[8];
case 8:
mix(_mm_set_epi64x(x, *(const uint64_t*)msg));
break;
case 7:
x |= ((uint64_t)msg[6]) << 48U;
case 6:
x |= ((uint64_t)msg[5]) << 40U;
case 5:
x |= ((uint64_t)msg[4]) << 32U;
case 4:
x |= *(const uint32_t*)msg;
mix(_mm_set_epi64x(0, x));
break;
case 3:
x |= ((uint32_t)msg[2]) << 16U;
case 2:
x |= ((uint32_t)msg[1]) << 8U;
case 1:
x |= msg[0];
mix(_mm_set_epi64x(0, x));
break;
case 0:
default: // try to keep m & s from register spilling
a = _mm_add_epi8(a, s);
b = _mm_add_epi8(b, m);
}
return _mm_aesenc_si128(a, b);
}
static inline uint64_t AESNI_Hash64(const uint8_t* msg, unsigned len, uint32_t seed = 0) {
union {
uint64_t x[2];
__m128i v;
} t;
t.v = AESNI_Hash128(msg, len, seed);
return t.x[0];
}
#ifdef __cplusplus
} // C
#endif
#endif // AESNI_HASH_H