1 | // Copyright (c) 2014 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 | // FarmHash, by Geoff Pike |
22 | |
23 | // |
24 | // http://code.google.com/p/farmhash/ |
25 | // |
26 | // This file provides a few functions for hashing strings and other |
27 | // data. All of them are high-quality functions in the sense that |
28 | // they do well on standard tests such as Austin Appleby's SMHasher. |
29 | // They're also fast. FarmHash is the successor to CityHash. |
30 | // |
31 | // Functions in the FarmHash family are not suitable for cryptography. |
32 | // |
33 | // WARNING: This code has been only lightly tested on big-endian platforms! |
34 | // It is known to work well on little-endian platforms that have a small penalty |
35 | // for unaligned reads, such as current Intel and AMD moderate-to-high-end CPUs. |
36 | // It should work on all 32-bit and 64-bit platforms that allow unaligned reads; |
37 | // bug reports are welcome. |
38 | // |
39 | // By the way, for some hash functions, given strings a and b, the hash |
40 | // of a+b is easily derived from the hashes of a and b. This property |
41 | // doesn't hold for any hash functions in this file. |
42 | |
43 | #ifndef FARM_HASH_H_ |
44 | #define FARM_HASH_H_ |
45 | |
46 | #include <assert.h> |
47 | #include <stdint.h> |
48 | #include <stdlib.h> |
49 | #include <string.h> // for memcpy and memset |
50 | #include <utility> |
51 | |
52 | #ifndef NAMESPACE_FOR_HASH_FUNCTIONS |
53 | #define NAMESPACE_FOR_HASH_FUNCTIONS util |
54 | #endif |
55 | |
56 | namespace NAMESPACE_FOR_HASH_FUNCTIONS { |
57 | |
58 | #if defined(FARMHASH_UINT128_T_DEFINED) |
59 | #if defined(__clang__) |
60 | #if !defined(uint128_t) |
61 | #define uint128_t __uint128_t |
62 | #endif |
63 | #endif |
64 | inline uint64_t Uint128Low64(const uint128_t x) { |
65 | return static_cast<uint64_t>(x); |
66 | } |
67 | inline uint64_t Uint128High64(const uint128_t x) { |
68 | return static_cast<uint64_t>(x >> 64); |
69 | } |
70 | inline uint128_t Uint128(uint64_t lo, uint64_t hi) { |
71 | return lo + (((uint128_t)hi) << 64); |
72 | } |
73 | #else |
74 | typedef std::pair<uint64_t, uint64_t> uint128_t; |
75 | inline uint64_t Uint128Low64(const uint128_t x) { return x.first; } |
76 | inline uint64_t Uint128High64(const uint128_t x) { return x.second; } |
77 | inline uint128_t Uint128(uint64_t lo, uint64_t hi) { return uint128_t(lo, hi); } |
78 | #endif |
79 | |
80 | |
81 | // BASIC STRING HASHING |
82 | |
83 | // Hash function for a byte array. |
84 | // May change from time to time, may differ on different platforms, may differ |
85 | // depending on NDEBUG. |
86 | size_t Hash(const char* s, size_t len); |
87 | |
88 | // Hash function for a byte array. Most useful in 32-bit binaries. |
89 | // May change from time to time, may differ on different platforms, may differ |
90 | // depending on NDEBUG. |
91 | uint32_t Hash32(const char* s, size_t len); |
92 | |
93 | // Hash function for a byte array. For convenience, a 32-bit seed is also |
94 | // hashed into the result. |
95 | // May change from time to time, may differ on different platforms, may differ |
96 | // depending on NDEBUG. |
97 | uint32_t Hash32WithSeed(const char* s, size_t len, uint32_t seed); |
98 | |
99 | // Hash function for a byte array. |
100 | // May change from time to time, may differ on different platforms, may differ |
101 | // depending on NDEBUG. |
102 | uint64_t Hash64(const char* s, size_t len); |
103 | |
104 | // Hash function for a byte array. For convenience, a 64-bit seed is also |
105 | // hashed into the result. |
106 | // May change from time to time, may differ on different platforms, may differ |
107 | // depending on NDEBUG. |
108 | uint64_t Hash64WithSeed(const char* s, size_t len, uint64_t seed); |
109 | |
110 | // Hash function for a byte array. For convenience, two seeds are also |
111 | // hashed into the result. |
112 | // May change from time to time, may differ on different platforms, may differ |
113 | // depending on NDEBUG. |
114 | uint64_t Hash64WithSeeds(const char* s, size_t len, |
115 | uint64_t seed0, uint64_t seed1); |
116 | |
117 | // Hash function for a byte array. |
118 | // May change from time to time, may differ on different platforms, may differ |
119 | // depending on NDEBUG. |
120 | uint128_t Hash128(const char* s, size_t len); |
121 | |
122 | // Hash function for a byte array. For convenience, a 128-bit seed is also |
123 | // hashed into the result. |
124 | // May change from time to time, may differ on different platforms, may differ |
125 | // depending on NDEBUG. |
126 | uint128_t Hash128WithSeed(const char* s, size_t len, uint128_t seed); |
127 | |
128 | // BASIC NON-STRING HASHING |
129 | |
130 | // Hash 128 input bits down to 64 bits of output. |
131 | // This is intended to be a reasonably good hash function. |
132 | // May change from time to time, may differ on different platforms, may differ |
133 | // depending on NDEBUG. |
134 | inline uint64_t Hash128to64(uint128_t x) { |
135 | // Murmur-inspired hashing. |
136 | const uint64_t kMul = 0x9ddfea08eb382d69ULL; |
137 | uint64_t a = (Uint128Low64(x) ^ Uint128High64(x)) * kMul; |
138 | a ^= (a >> 47); |
139 | uint64_t b = (Uint128High64(x) ^ a) * kMul; |
140 | b ^= (b >> 47); |
141 | b *= kMul; |
142 | return b; |
143 | } |
144 | |
145 | // FINGERPRINTING (i.e., good, portable, forever-fixed hash functions) |
146 | |
147 | // Fingerprint function for a byte array. Most useful in 32-bit binaries. |
148 | uint32_t Fingerprint32(const char* s, size_t len); |
149 | |
150 | // Fingerprint function for a byte array. |
151 | uint64_t Fingerprint64(const char* s, size_t len); |
152 | |
153 | // Fingerprint function for a byte array. |
154 | uint128_t Fingerprint128(const char* s, size_t len); |
155 | |
156 | // This is intended to be a good fingerprinting primitive. |
157 | // See below for more overloads. |
158 | inline uint64_t Fingerprint(uint128_t x) { |
159 | // Murmur-inspired hashing. |
160 | const uint64_t kMul = 0x9ddfea08eb382d69ULL; |
161 | uint64_t a = (Uint128Low64(x) ^ Uint128High64(x)) * kMul; |
162 | a ^= (a >> 47); |
163 | uint64_t b = (Uint128High64(x) ^ a) * kMul; |
164 | b ^= (b >> 44); |
165 | b *= kMul; |
166 | b ^= (b >> 41); |
167 | b *= kMul; |
168 | return b; |
169 | } |
170 | |
171 | // This is intended to be a good fingerprinting primitive. |
172 | inline uint64_t Fingerprint(uint64_t x) { |
173 | // Murmur-inspired hashing. |
174 | const uint64_t kMul = 0x9ddfea08eb382d69ULL; |
175 | uint64_t b = x * kMul; |
176 | b ^= (b >> 44); |
177 | b *= kMul; |
178 | b ^= (b >> 41); |
179 | b *= kMul; |
180 | return b; |
181 | } |
182 | |
183 | #ifndef FARMHASH_NO_CXX_STRING |
184 | |
185 | // Convenience functions to hash or fingerprint C++ strings. |
186 | // These require that Str::data() return a pointer to the first char |
187 | // (as a const char*) and that Str::length() return the string's length; |
188 | // they work with std::string, for example. |
189 | |
190 | // Hash function for a byte array. |
191 | // May change from time to time, may differ on different platforms, may differ |
192 | // depending on NDEBUG. |
193 | template <typename Str> |
194 | inline size_t Hash(const Str& s) { |
195 | assert(sizeof(s[0]) == 1); |
196 | return Hash(s.data(), s.length()); |
197 | } |
198 | |
199 | // Hash function for a byte array. Most useful in 32-bit binaries. |
200 | // May change from time to time, may differ on different platforms, may differ |
201 | // depending on NDEBUG. |
202 | template <typename Str> |
203 | inline uint32_t Hash32(const Str& s) { |
204 | assert(sizeof(s[0]) == 1); |
205 | return Hash32(s.data(), s.length()); |
206 | } |
207 | |
208 | // Hash function for a byte array. For convenience, a 32-bit seed is also |
209 | // hashed into the result. |
210 | // May change from time to time, may differ on different platforms, may differ |
211 | // depending on NDEBUG. |
212 | template <typename Str> |
213 | inline uint32_t Hash32WithSeed(const Str& s, uint32_t seed) { |
214 | assert(sizeof(s[0]) == 1); |
215 | return Hash32WithSeed(s.data(), s.length(), seed); |
216 | } |
217 | |
218 | // Hash 128 input bits down to 64 bits of output. |
219 | // Hash function for a byte array. |
220 | // May change from time to time, may differ on different platforms, may differ |
221 | // depending on NDEBUG. |
222 | template <typename Str> |
223 | inline uint64_t Hash64(const Str& s) { |
224 | assert(sizeof(s[0]) == 1); |
225 | return Hash64(s.data(), s.length()); |
226 | } |
227 | |
228 | // Hash function for a byte array. For convenience, a 64-bit seed is also |
229 | // hashed into the result. |
230 | // May change from time to time, may differ on different platforms, may differ |
231 | // depending on NDEBUG. |
232 | template <typename Str> |
233 | inline uint64_t Hash64WithSeed(const Str& s, uint64_t seed) { |
234 | assert(sizeof(s[0]) == 1); |
235 | return Hash64WithSeed(s.data(), s.length(), seed); |
236 | } |
237 | |
238 | // Hash function for a byte array. For convenience, two seeds are also |
239 | // hashed into the result. |
240 | // May change from time to time, may differ on different platforms, may differ |
241 | // depending on NDEBUG. |
242 | template <typename Str> |
243 | inline uint64_t Hash64WithSeeds(const Str& s, uint64_t seed0, uint64_t seed1) { |
244 | assert(sizeof(s[0]) == 1); |
245 | return Hash64WithSeeds(s.data(), s.length(), seed0, seed1); |
246 | } |
247 | |
248 | // Hash function for a byte array. |
249 | // May change from time to time, may differ on different platforms, may differ |
250 | // depending on NDEBUG. |
251 | template <typename Str> |
252 | inline uint128_t Hash128(const Str& s) { |
253 | assert(sizeof(s[0]) == 1); |
254 | return Hash128(s.data(), s.length()); |
255 | } |
256 | |
257 | // Hash function for a byte array. For convenience, a 128-bit seed is also |
258 | // hashed into the result. |
259 | // May change from time to time, may differ on different platforms, may differ |
260 | // depending on NDEBUG. |
261 | template <typename Str> |
262 | inline uint128_t Hash128WithSeed(const Str& s, uint128_t seed) { |
263 | assert(sizeof(s[0]) == 1); |
264 | return Hash128(s.data(), s.length(), seed); |
265 | } |
266 | |
267 | // FINGERPRINTING (i.e., good, portable, forever-fixed hash functions) |
268 | |
269 | // Fingerprint function for a byte array. Most useful in 32-bit binaries. |
270 | template <typename Str> |
271 | inline uint32_t Fingerprint32(const Str& s) { |
272 | assert(sizeof(s[0]) == 1); |
273 | return Fingerprint32(s.data(), s.length()); |
274 | } |
275 | |
276 | // Fingerprint 128 input bits down to 64 bits of output. |
277 | // Fingerprint function for a byte array. |
278 | template <typename Str> |
279 | inline uint64_t Fingerprint64(const Str& s) { |
280 | assert(sizeof(s[0]) == 1); |
281 | return Fingerprint64(s.data(), s.length()); |
282 | } |
283 | |
284 | // Fingerprint function for a byte array. |
285 | template <typename Str> |
286 | inline uint128_t Fingerprint128(const Str& s) { |
287 | assert(sizeof(s[0]) == 1); |
288 | return Fingerprint128(s.data(), s.length()); |
289 | } |
290 | |
291 | #endif |
292 | |
293 | } // namespace NAMESPACE_FOR_HASH_FUNCTIONS |
294 | |
295 | /* gently define FARMHASH_BIG_ENDIAN when detected big-endian machine */ |
296 | #if defined(__BIG_ENDIAN__) |
297 | #if !defined(FARMHASH_BIG_ENDIAN) |
298 | #define FARMHASH_BIG_ENDIAN |
299 | #endif |
300 | #elif defined(__LITTLE_ENDIAN__) |
301 | // nothing for little-endian |
302 | #elif defined(__BYTE_ORDER__) && defined(__ORDER_LITTLE_ENDIAN__) && (__BYTE_ORDER == __ORDER_LITTLE_ENDIAN__) |
303 | // nothing for little-endian |
304 | #elif defined(__BYTE_ORDER__) && defined(__ORDER_BIG_ENDIAN__) && (__BYTE_ORDER == __ORDER_BIG_ENDIAN__) |
305 | #if !defined(FARMHASH_BIG_ENDIAN) |
306 | #define FARMHASH_BIG_ENDIAN |
307 | #endif |
308 | #elif defined(__linux__) || defined(__CYGWIN__) || defined( __GNUC__ ) && !defined(_WIN32) || defined( __GNU_LIBRARY__ ) |
309 | #include <endian.h> // libc6-dev, GLIBC |
310 | #if BYTE_ORDER == BIG_ENDIAN |
311 | #if !defined(FARMHASH_BIG_ENDIAN) |
312 | #define FARMHASH_BIG_ENDIAN |
313 | #endif |
314 | #endif |
315 | #elif defined(__OpenBSD__) || defined(__NetBSD__) || defined(__FreeBSD__) || defined(__DragonFly__) || defined(__s390x__) |
316 | #include <sys/endian.h> |
317 | #if BYTE_ORDER == BIG_ENDIAN |
318 | #if !defined(FARMHASH_BIG_ENDIAN) |
319 | #define FARMHASH_BIG_ENDIAN |
320 | #endif |
321 | #endif |
322 | #elif defined(_WIN32) |
323 | // Windows is (currently) little-endian |
324 | #else |
325 | #error "Unable to determine endianness!" |
326 | #endif /* __BIG_ENDIAN__ */ |
327 | |
328 | #endif // FARM_HASH_H_ |
329 | |