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
|
package xxh3
import (
"math/bits"
)
// Hash128 returns the 128-bit hash of the byte slice.
func Hash128(b []byte) Uint128 {
return hashAny128(*(*str)(ptr(&b)))
}
// HashString128 returns the 128-bit hash of the string slice.
func HashString128(s string) Uint128 {
return hashAny128(*(*str)(ptr(&s)))
}
func hashAny128(s str) (acc u128) {
p, l := s.p, s.l
switch {
case l <= 16:
switch {
case l > 8: // 9-16
const bitflipl = key64_032 ^ key64_040
const bitfliph = key64_048 ^ key64_056
input_lo := readU64(p, 0)
input_hi := readU64(p, ui(l)-8)
m128_h, m128_l := bits.Mul64(input_lo^input_hi^bitflipl, prime64_1)
m128_l += uint64(l-1) << 54
input_hi ^= bitfliph
m128_h += input_hi + uint64(uint32(input_hi))*(prime32_2-1)
m128_l ^= bits.ReverseBytes64(m128_h)
acc.Hi, acc.Lo = bits.Mul64(m128_l, prime64_2)
acc.Hi += m128_h * prime64_2
acc.Lo = xxh3Avalanche(acc.Lo)
acc.Hi = xxh3Avalanche(acc.Hi)
return acc
case l > 3: // 4-8
const bitflip = key64_016 ^ key64_024
input_lo := readU32(p, 0)
input_hi := readU32(p, ui(l)-4)
input_64 := u64(input_lo) + u64(input_hi)<<32
keyed := input_64 ^ bitflip
acc.Hi, acc.Lo = bits.Mul64(keyed, prime64_1+(uint64(l)<<2))
acc.Hi += acc.Lo << 1
acc.Lo ^= acc.Hi >> 3
acc.Lo ^= acc.Lo >> 35
acc.Lo *= 0x9fb21c651e98df25
acc.Lo ^= acc.Lo >> 28
acc.Hi = xxh3Avalanche(acc.Hi)
return acc
case l == 3: // 3
c12 := u64(readU16(p, 0))
c3 := u64(readU8(p, 2))
acc.Lo = c12<<16 + c3 + 3<<8
case l > 1: // 2
c12 := u64(readU16(p, 0))
acc.Lo = c12*(1<<24+1)>>8 + 2<<8
case l == 1: // 1
c1 := u64(readU8(p, 0))
acc.Lo = c1*(1<<24+1<<16+1) + 1<<8
default: // 0
return u128{0x99aa06d3014798d8, 0x6001c324468d497f}
}
acc.Hi = uint64(bits.RotateLeft32(bits.ReverseBytes32(uint32(acc.Lo)), 13))
acc.Lo ^= uint64(key32_000 ^ key32_004)
acc.Hi ^= uint64(key32_008 ^ key32_012)
acc.Lo = xxh64AvalancheSmall(acc.Lo)
acc.Hi = xxh64AvalancheSmall(acc.Hi)
return acc
case l <= 128:
acc.Lo = u64(l) * prime64_1
if l > 32 {
if l > 64 {
if l > 96 {
in8, in7 := readU64(p, ui(l)-8*8), readU64(p, ui(l)-7*8)
i6, i7 := readU64(p, 6*8), readU64(p, 7*8)
acc.Hi += mulFold64(in8^key64_112, in7^key64_120)
acc.Hi ^= i6 + i7
acc.Lo += mulFold64(i6^key64_096, i7^key64_104)
acc.Lo ^= in8 + in7
} // 96
in6, in5 := readU64(p, ui(l)-6*8), readU64(p, ui(l)-5*8)
i4, i5 := readU64(p, 4*8), readU64(p, 5*8)
acc.Hi += mulFold64(in6^key64_080, in5^key64_088)
acc.Hi ^= i4 + i5
acc.Lo += mulFold64(i4^key64_064, i5^key64_072)
acc.Lo ^= in6 + in5
} // 64
in4, in3 := readU64(p, ui(l)-4*8), readU64(p, ui(l)-3*8)
i2, i3 := readU64(p, 2*8), readU64(p, 3*8)
acc.Hi += mulFold64(in4^key64_048, in3^key64_056)
acc.Hi ^= i2 + i3
acc.Lo += mulFold64(i2^key64_032, i3^key64_040)
acc.Lo ^= in4 + in3
} // 32
in2, in1 := readU64(p, ui(l)-2*8), readU64(p, ui(l)-1*8)
i0, i1 := readU64(p, 0*8), readU64(p, 1*8)
acc.Hi += mulFold64(in2^key64_016, in1^key64_024)
acc.Hi ^= i0 + i1
acc.Lo += mulFold64(i0^key64_000, i1^key64_008)
acc.Lo ^= in2 + in1
acc.Hi, acc.Lo = (acc.Lo*prime64_1)+(acc.Hi*prime64_4)+(u64(l)*prime64_2), acc.Hi+acc.Lo
acc.Hi = -xxh3Avalanche(acc.Hi)
acc.Lo = xxh3Avalanche(acc.Lo)
return acc
case l <= 240:
acc.Lo = u64(l) * prime64_1
{
i0, i1, i2, i3 := readU64(p, 0*8), readU64(p, 1*8), readU64(p, 2*8), readU64(p, 3*8)
acc.Hi += mulFold64(i2^key64_016, i3^key64_024)
acc.Hi ^= i0 + i1
acc.Lo += mulFold64(i0^key64_000, i1^key64_008)
acc.Lo ^= i2 + i3
}
{
i0, i1, i2, i3 := readU64(p, 4*8), readU64(p, 5*8), readU64(p, 6*8), readU64(p, 7*8)
acc.Hi += mulFold64(i2^key64_048, i3^key64_056)
acc.Hi ^= i0 + i1
acc.Lo += mulFold64(i0^key64_032, i1^key64_040)
acc.Lo ^= i2 + i3
}
{
i0, i1, i2, i3 := readU64(p, 8*8), readU64(p, 9*8), readU64(p, 10*8), readU64(p, 11*8)
acc.Hi += mulFold64(i2^key64_080, i3^key64_088)
acc.Hi ^= i0 + i1
acc.Lo += mulFold64(i0^key64_064, i1^key64_072)
acc.Lo ^= i2 + i3
}
{
i0, i1, i2, i3 := readU64(p, 12*8), readU64(p, 13*8), readU64(p, 14*8), readU64(p, 15*8)
acc.Hi += mulFold64(i2^key64_112, i3^key64_120)
acc.Hi ^= i0 + i1
acc.Lo += mulFold64(i0^key64_096, i1^key64_104)
acc.Lo ^= i2 + i3
}
// avalanche
acc.Hi = xxh3Avalanche(acc.Hi)
acc.Lo = xxh3Avalanche(acc.Lo)
// trailing groups after 128
top := ui(l) &^ 31
for i := ui(4 * 32); i < top; i += 32 {
i0, i1, i2, i3 := readU64(p, i+0), readU64(p, i+8), readU64(p, i+16), readU64(p, i+24)
k0, k1, k2, k3 := readU64(key, i-125), readU64(key, i-117), readU64(key, i-109), readU64(key, i-101)
acc.Hi += mulFold64(i2^k2, i3^k3)
acc.Hi ^= i0 + i1
acc.Lo += mulFold64(i0^k0, i1^k1)
acc.Lo ^= i2 + i3
}
// last 32 bytes
{
i0, i1, i2, i3 := readU64(p, ui(l)-32), readU64(p, ui(l)-24), readU64(p, ui(l)-16), readU64(p, ui(l)-8)
acc.Hi += mulFold64(i0^key64_119, i1^key64_127)
acc.Hi ^= i2 + i3
acc.Lo += mulFold64(i2^key64_103, i3^key64_111)
acc.Lo ^= i0 + i1
}
acc.Hi, acc.Lo = (acc.Lo*prime64_1)+(acc.Hi*prime64_4)+(u64(l)*prime64_2), acc.Hi+acc.Lo
acc.Hi = -xxh3Avalanche(acc.Hi)
acc.Lo = xxh3Avalanche(acc.Lo)
return acc
default:
acc.Lo = u64(l) * prime64_1
acc.Hi = ^(u64(l) * prime64_2)
accs := [8]u64{
prime32_3, prime64_1, prime64_2, prime64_3,
prime64_4, prime32_2, prime64_5, prime32_1,
}
if hasAVX512 && l >= avx512Switch {
accumAVX512(&accs, p, key, u64(l))
} else if hasAVX2 {
accumAVX2(&accs, p, key, u64(l))
} else if hasSSE2 {
accumSSE(&accs, p, key, u64(l))
} else {
accumScalar(&accs, p, key, u64(l))
}
// merge accs
acc.Lo += mulFold64(accs[0]^key64_011, accs[1]^key64_019)
acc.Hi += mulFold64(accs[0]^key64_117, accs[1]^key64_125)
acc.Lo += mulFold64(accs[2]^key64_027, accs[3]^key64_035)
acc.Hi += mulFold64(accs[2]^key64_133, accs[3]^key64_141)
acc.Lo += mulFold64(accs[4]^key64_043, accs[5]^key64_051)
acc.Hi += mulFold64(accs[4]^key64_149, accs[5]^key64_157)
acc.Lo += mulFold64(accs[6]^key64_059, accs[7]^key64_067)
acc.Hi += mulFold64(accs[6]^key64_165, accs[7]^key64_173)
acc.Lo = xxh3Avalanche(acc.Lo)
acc.Hi = xxh3Avalanche(acc.Hi)
return acc
}
}
|
