1 // Copyright (c) 2011 Google, Inc.
2 //
3 // Permission is hereby granted, free of charge, to any person obtaining a copy
4 // of this software and associated documentation files (the "Software"), to deal
5 // in the Software without restriction, including without limitation the rights
6 // to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
7 // copies of the Software, and to permit persons to whom the Software is
8 // furnished to do so, subject to the following conditions:
9 //
10 // The above copyright notice and this permission notice shall be included in
11 // all copies or substantial portions of the Software.
12 //
13 // THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
14 // IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
15 // FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
16 // AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
17 // LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
18 // OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
19 // THE SOFTWARE.
20 //
21 // CityHash, by Geoff Pike and Jyrki Alakuijala
22 //
23 // http://code.google.com/p/cityhash/
24 //
25 // This file provides a few functions for hashing strings. All of them are
26 // high-quality functions in the sense that they pass standard tests such
27 // as Austin Appleby's SMHasher. They are also fast.
28 //
29 // For 64-bit x86 code, on short strings, we don't know of anything faster than
30 // CityHash64 that is of comparable quality. We believe our nearest competitor
31 // is Murmur3. For 64-bit x86 code, CityHash64 is an excellent choice for hash
32 // tables and most other hashing (excluding cryptography).
33 //
34 // For 64-bit x86 code, on long strings, the picture is more complicated.
35 // On many recent Intel CPUs, such as Nehalem, Westmere, Sandy Bridge, etc.,
36 // CityHashCrc128 appears to be faster than all competitors of comparable
37 // quality. CityHash128 is also good but not quite as fast. We believe our
38 // nearest competitor is Bob Jenkins' Spooky. We don't have great data for
39 // other 64-bit CPUs, but for long strings we know that Spooky is slightly
40 // faster than CityHash on some relatively recent AMD x86-64 CPUs, for example.
41 // Note that CityHashCrc128 is declared in citycrc.h.
42 //
43 // For 32-bit x86 code, we don't know of anything faster than CityHash32 that
44 // is of comparable quality. We believe our nearest competitor is Murmur3A.
45 // (On 64-bit CPUs, it is typically faster to use the other CityHash variants.)
46 //
47 // Functions in the CityHash family are not suitable for cryptography.
48 //
49 // Please see CityHash's README file for more details on our performance
50 // measurements and so on.
51 //
52 // WARNING: This code has been only lightly tested on big-endian platforms!
53 // It is known to work well on little-endian platforms that have a small penalty
54 // for unaligned reads, such as current Intel and AMD moderate-to-high-end CPUs.
55 // It should work on all 32-bit and 64-bit platforms that allow unaligned reads;
56 // bug reports are welcome.
57 //
58 // By the way, for some hash functions, given strings a and b, the hash
59 // of a+b is easily derived from the hashes of a and b. This property
60 // doesn't hold for any hash functions in this file.
61
62 #ifndef BASE_THIRD_PARTY_CITYHASH_CITY_H_
63 #define BASE_THIRD_PARTY_CITYHASH_CITY_H_
64
65 #include <stdint.h>
66 #include <stdlib.h> // for size_t.
67 #include <utility>
68
69 // XXX(cavalcantii): Declaring it inside of the 'base' namespace allows to
70 // handle linker symbol clash error with deprecated CityHash from
71 // third_party/smhasher in a few unit tests.
72 namespace base {
73 namespace internal {
74 namespace cityhash_v111 {
75
76 typedef uint8_t uint8;
77 typedef uint32_t uint32;
78 typedef uint64_t uint64;
79 typedef std::pair<uint64, uint64> uint128;
80
Uint128Low64(const uint128 & x)81 inline uint64 Uint128Low64(const uint128& x) {
82 return x.first;
83 }
Uint128High64(const uint128 & x)84 inline uint64 Uint128High64(const uint128& x) {
85 return x.second;
86 }
87
88 // Hash function for a byte array.
89 uint64 CityHash64(const char* buf, size_t len);
90
91 // Hash function for a byte array. For convenience, a 64-bit seed is also
92 // hashed into the result.
93 uint64 CityHash64WithSeed(const char* buf, size_t len, uint64 seed);
94
95 // Hash function for a byte array. For convenience, two seeds are also
96 // hashed into the result.
97 uint64 CityHash64WithSeeds(const char* buf,
98 size_t len,
99 uint64 seed0,
100 uint64 seed1);
101
102 // Hash function for a byte array.
103 uint128 CityHash128(const char* s, size_t len);
104
105 // Hash function for a byte array. For convenience, a 128-bit seed is also
106 // hashed into the result.
107 uint128 CityHash128WithSeed(const char* s, size_t len, uint128 seed);
108
109 // Hash function for a byte array. Most useful in 32-bit binaries.
110 uint32 CityHash32(const char* buf, size_t len);
111
112 // Hash 128 input bits down to 64 bits of output.
113 // This is intended to be a reasonably good hash function.
Hash128to64(const uint128 & x)114 inline uint64 Hash128to64(const uint128& x) {
115 // Murmur-inspired hashing.
116 const uint64 kMul = 0x9ddfea08eb382d69ULL;
117 uint64 a = (Uint128Low64(x) ^ Uint128High64(x)) * kMul;
118 a ^= (a >> 47);
119 uint64 b = (Uint128High64(x) ^ a) * kMul;
120 b ^= (b >> 47);
121 b *= kMul;
122 return b;
123 }
124
125 } // namespace cityhash_v111
126 } // namespace internal
127 } // namespace base
128
129 #endif // CITY_HASH_H_
130