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
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
base / third_party / cityhash_v103 / src / city_v103.cc [blame]
// Copyright (c) 2011 Google, Inc.
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in
// all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
// THE SOFTWARE.
//
// CityHash, by Geoff Pike and Jyrki Alakuijala
//
// This file provides CityHash64() and related functions.
//
// It's probably possible to create even faster hash functions by
// writing a program that systematically explores some of the space of
// possible hash functions, by using SIMD instructions, or by
// compromising on hash quality.
#include "base/third_party/cityhash_v103/src/city_v103.h"
#include <string.h> // for memcpy and memset
#include <algorithm>
#include "build/build_config.h"
namespace base {
namespace internal {
namespace cityhash_v103 {
static uint64 UNALIGNED_LOAD64(const char* p) {
uint64 result;
memcpy(&result, p, sizeof(result));
return result;
}
static uint32 UNALIGNED_LOAD32(const char* p) {
uint32 result;
memcpy(&result, p, sizeof(result));
return result;
}
#if defined(ARCH_CPU_LITTLE_ENDIAN)
#define uint32_in_expected_order(x) (x)
#define uint64_in_expected_order(x) (x)
#else
#if defined(__clang__)
// Use builtins where possible. On Windows for instance, this may prevent a
// function call instead of emitting a single instruction.
#define bswap_32(x) __builtin_bswap32(x)
#define bswap_64(x) __builtin_bswap64(x)
#elif _MSC_VER
#include <stdlib.h>
#define bswap_32(x) _byteswap_ulong(x)
#define bswap_64(x) _byteswap_uint64(x)
#elif defined(__APPLE__)
// Mac OS X / Darwin features
#include <libkern/OSByteOrder.h>
#define bswap_32(x) OSSwapInt32(x)
#define bswap_64(x) OSSwapInt64(x)
#else
#include <byteswap.h>
#endif
#define uint32_in_expected_order(x) (bswap_32(x))
#define uint64_in_expected_order(x) (bswap_64(x))
#endif // defined(ARCH_CPU_LITTLE_ENDIAN)
#if !defined(LIKELY)
#if HAVE_BUILTIN_EXPECT
#define LIKELY(x) (__builtin_expect(!!(x), 1))
#else
#define LIKELY(x) (x)
#endif
#endif
static uint64 Fetch64(const char* p) {
return uint64_in_expected_order(UNALIGNED_LOAD64(p));
}
static uint32 Fetch32(const char* p) {
return uint32_in_expected_order(UNALIGNED_LOAD32(p));
}
// Some primes between 2^63 and 2^64 for various uses.
static const uint64 k0 = 0xc3a5c85c97cb3127ULL;
static const uint64 k1 = 0xb492b66fbe98f273ULL;
static const uint64 k2 = 0x9ae16a3b2f90404fULL;
static const uint64 k3 = 0xc949d7c7509e6557ULL;
// Bitwise right rotate. Normally this will compile to a single
// instruction, especially if the shift is a manifest constant.
static uint64 Rotate(uint64 val, int shift) {
// Avoid shifting by 64: doing so yields an undefined result.
return shift == 0 ? val : ((val >> shift) | (val << (64 - shift)));
}
// Equivalent to Rotate(), but requires the second arg to be non-zero.
// On x86-64, and probably others, it's possible for this to compile
// to a single instruction if both args are already in registers.
static uint64 RotateByAtLeast1(uint64 val, size_t shift) {
return (val >> shift) | (val << (64 - shift));
}
static uint64 ShiftMix(uint64 val) {
return val ^ (val >> 47);
}
static uint64 HashLen16(uint64 u, uint64 v) {
return Hash128to64(uint128(u, v));
}
static uint64 HashLen0to16(const char* s, size_t len) {
if (len > 8) {
uint64 a = Fetch64(s);
uint64 b = Fetch64(s + len - 8);
return HashLen16(a, RotateByAtLeast1(b + len, len)) ^ b;
}
if (len >= 4) {
uint64 a = Fetch32(s);
return HashLen16(len + (a << 3), Fetch32(s + len - 4));
}
if (len > 0) {
uint8 a = static_cast<uint8>(s[0]);
uint8 b = static_cast<uint8>(s[len >> 1]);
uint8 c = static_cast<uint8>(s[len - 1]);
uint32 y = static_cast<uint32>(a) + (static_cast<uint32>(b) << 8);
uint32 z = static_cast<uint32>(len) + (static_cast<uint32>(c) << 2);
return ShiftMix(y * k2 ^ z * k3) * k2;
}
return k2;
}
// This probably works well for 16-byte strings as well, but it may be overkill
// in that case.
static uint64 HashLen17to32(const char* s, size_t len) {
uint64 a = Fetch64(s) * k1;
uint64 b = Fetch64(s + 8);
uint64 c = Fetch64(s + len - 8) * k2;
uint64 d = Fetch64(s + len - 16) * k0;
return HashLen16(Rotate(a - b, 43) + Rotate(c, 30) + d,
a + Rotate(b ^ k3, 20) - c + len);
}
// Return a 16-byte hash for 48 bytes. Quick and dirty.
// Callers do best to use "random-looking" values for a and b.
static std::pair<uint64, uint64> WeakHashLen32WithSeeds(uint64 w,
uint64 x,
uint64 y,
uint64 z,
uint64 a,
uint64 b) {
a += w;
b = Rotate(b + a + z, 21);
uint64 c = a;
a += x;
a += y;
b += Rotate(a, 44);
return std::make_pair(a + z, b + c);
}
// Return a 16-byte hash for s[0] ... s[31], a, and b. Quick and dirty.
static std::pair<uint64, uint64> WeakHashLen32WithSeeds(const char* s,
uint64 a,
uint64 b) {
return WeakHashLen32WithSeeds(Fetch64(s), Fetch64(s + 8), Fetch64(s + 16),
Fetch64(s + 24), a, b);
}
// Return an 8-byte hash for 33 to 64 bytes.
static uint64 HashLen33to64(const char* s, size_t len) {
uint64 z = Fetch64(s + 24);
uint64 a = Fetch64(s) + (len + Fetch64(s + len - 16)) * k0;
uint64 b = Rotate(a + z, 52);
uint64 c = Rotate(a, 37);
a += Fetch64(s + 8);
c += Rotate(a, 7);
a += Fetch64(s + 16);
uint64 vf = a + z;
uint64 vs = b + Rotate(a, 31) + c;
a = Fetch64(s + 16) + Fetch64(s + len - 32);
z = Fetch64(s + len - 8);
b = Rotate(a + z, 52);
c = Rotate(a, 37);
a += Fetch64(s + len - 24);
c += Rotate(a, 7);
a += Fetch64(s + len - 16);
uint64 wf = a + z;
uint64 ws = b + Rotate(a, 31) + c;
uint64 r = ShiftMix((vf + ws) * k2 + (wf + vs) * k0);
return ShiftMix(r * k0 + vs) * k2;
}
uint64 CityHash64(const char* s, size_t len) {
if (len <= 32) {
if (len <= 16) {
return HashLen0to16(s, len);
} else {
return HashLen17to32(s, len);
}
} else if (len <= 64) {
return HashLen33to64(s, len);
}
// For strings over 64 bytes we hash the end first, and then as we
// loop we keep 56 bytes of state: v, w, x, y, and z.
uint64 x = Fetch64(s + len - 40);
uint64 y = Fetch64(s + len - 16) + Fetch64(s + len - 56);
uint64 z = HashLen16(Fetch64(s + len - 48) + len, Fetch64(s + len - 24));
std::pair<uint64, uint64> v = WeakHashLen32WithSeeds(s + len - 64, len, z);
std::pair<uint64, uint64> w = WeakHashLen32WithSeeds(s + len - 32, y + k1, x);
x = x * k1 + Fetch64(s);
// Decrease len to the nearest multiple of 64, and operate on 64-byte chunks.
len = (len - 1) & ~static_cast<size_t>(63);
do {
x = Rotate(x + y + v.first + Fetch64(s + 8), 37) * k1;
y = Rotate(y + v.second + Fetch64(s + 48), 42) * k1;
x ^= w.second;
y += v.first + Fetch64(s + 40);
z = Rotate(z + w.first, 33) * k1;
v = WeakHashLen32WithSeeds(s, v.second * k1, x + w.first);
w = WeakHashLen32WithSeeds(s + 32, z + w.second, y + Fetch64(s + 16));
std::swap(z, x);
s += 64;
len -= 64;
} while (len != 0);
return HashLen16(HashLen16(v.first, w.first) + ShiftMix(y) * k1 + z,
HashLen16(v.second, w.second) + x);
}
uint64 CityHash64WithSeed(const char* s, size_t len, uint64 seed) {
return CityHash64WithSeeds(s, len, k2, seed);
}
uint64 CityHash64WithSeeds(const char* s,
size_t len,
uint64 seed0,
uint64 seed1) {
return HashLen16(CityHash64(s, len) - seed0, seed1);
}
// A subroutine for CityHash128(). Returns a decent 128-bit hash for strings
// of any length representable in signed long. Based on City and Murmur.
static uint128 CityMurmur(const char* s, size_t len, uint128 seed) {
uint64 a = Uint128Low64(seed);
uint64 b = Uint128High64(seed);
uint64 c = 0;
uint64 d = 0;
if (len <= 16) {
a = ShiftMix(a * k1) * k1;
c = b * k1 + HashLen0to16(s, len);
d = ShiftMix(a + (len >= 8 ? Fetch64(s) : c));
} else {
c = HashLen16(Fetch64(s + len - 8) + k1, a);
d = HashLen16(b + len, c + Fetch64(s + len - 16));
a += d;
// len > 16 here, so do...while is safe
do {
a ^= ShiftMix(Fetch64(s) * k1) * k1;
a *= k1;
b ^= a;
c ^= ShiftMix(Fetch64(s + 8) * k1) * k1;
c *= k1;
d ^= c;
s += 16;
len -= 16;
} while (len > 16);
}
a = HashLen16(a, c);
b = HashLen16(d, b);
return uint128(a ^ b, HashLen16(b, a));
}
uint128 CityHash128WithSeed(const char* s, size_t len, uint128 seed) {
if (len < 128) {
return CityMurmur(s, len, seed);
}
// We expect len >= 128 to be the common case. Keep 56 bytes of state:
// v, w, x, y, and z.
std::pair<uint64, uint64> v, w;
uint64 x = Uint128Low64(seed);
uint64 y = Uint128High64(seed);
uint64 z = len * k1;
v.first = Rotate(y ^ k1, 49) * k1 + Fetch64(s);
v.second = Rotate(v.first, 42) * k1 + Fetch64(s + 8);
w.first = Rotate(y + z, 35) * k1 + x;
w.second = Rotate(x + Fetch64(s + 88), 53) * k1;
// This is the same inner loop as CityHash64(), manually unrolled.
do {
x = Rotate(x + y + v.first + Fetch64(s + 8), 37) * k1;
y = Rotate(y + v.second + Fetch64(s + 48), 42) * k1;
x ^= w.second;
y += v.first + Fetch64(s + 40);
z = Rotate(z + w.first, 33) * k1;
v = WeakHashLen32WithSeeds(s, v.second * k1, x + w.first);
w = WeakHashLen32WithSeeds(s + 32, z + w.second, y + Fetch64(s + 16));
std::swap(z, x);
s += 64;
x = Rotate(x + y + v.first + Fetch64(s + 8), 37) * k1;
y = Rotate(y + v.second + Fetch64(s + 48), 42) * k1;
x ^= w.second;
y += v.first + Fetch64(s + 40);
z = Rotate(z + w.first, 33) * k1;
v = WeakHashLen32WithSeeds(s, v.second * k1, x + w.first);
w = WeakHashLen32WithSeeds(s + 32, z + w.second, y + Fetch64(s + 16));
std::swap(z, x);
s += 64;
len -= 128;
} while (LIKELY(len >= 128));
x += Rotate(v.first + z, 49) * k0;
z += Rotate(w.first, 37) * k0;
// If 0 < len < 128, hash up to 4 chunks of 32 bytes each from the end of s.
for (size_t tail_done = 0; tail_done < len;) {
tail_done += 32;
y = Rotate(x + y, 42) * k0 + v.second;
w.first += Fetch64(s + len - tail_done + 16);
x = x * k0 + w.first;
z += w.second + Fetch64(s + len - tail_done);
w.second += v.first;
v = WeakHashLen32WithSeeds(s + len - tail_done, v.first + z, v.second);
}
// At this point our 56 bytes of state should contain more than
// enough information for a strong 128-bit hash. We use two
// different 56-byte-to-8-byte hashes to get a 16-byte final result.
x = HashLen16(x, v.first);
y = HashLen16(y + z, w.first);
return uint128(HashLen16(x + v.second, w.second) + y,
HashLen16(x + w.second, y + v.second));
}
uint128 CityHash128(const char* s, size_t len) {
if (len >= 16) {
return CityHash128WithSeed(s + 16, len - 16,
uint128(Fetch64(s) ^ k3, Fetch64(s + 8)));
} else if (len >= 8) {
return CityHash128WithSeed(
nullptr, 0,
uint128(Fetch64(s) ^ (len * k0), Fetch64(s + len - 8) ^ k1));
} else {
return CityHash128WithSeed(s, len, uint128(k0, k1));
}
}
} // namespace cityhash_v103
} // namespace internal
} // namespace base