Commit | Line | Data |
---|---|---|
f7e7a1b8 MD |
1 | #ifndef _JHASH_H |
2 | #define _JHASH_H | |
3 | ||
8eefcfaf MJ |
4 | #if defined(__FreeBSD__) |
5 | #include <sys/endian.h> | |
6 | #endif | |
7 | ||
f7e7a1b8 MD |
8 | /* |
9 | * jhash.h | |
10 | * | |
11 | * Example hash function. | |
12 | * | |
13 | * Copyright 2009-2012 - Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca> | |
14 | * | |
15 | * THIS MATERIAL IS PROVIDED AS IS, WITH ABSOLUTELY NO WARRANTY EXPRESSED | |
16 | * OR IMPLIED. ANY USE IS AT YOUR OWN RISK. | |
17 | * | |
18 | * Permission is hereby granted to use or copy this program for any | |
19 | * purpose, provided the above notices are retained on all copies. | |
20 | * Permission to modify the code and to distribute modified code is | |
21 | * granted, provided the above notices are retained, and a notice that | |
22 | * the code was modified is included with the above copyright notice. | |
23 | */ | |
24 | ||
25 | /* | |
26 | * Hash function | |
27 | * Source: http://burtleburtle.net/bob/c/lookup3.c | |
28 | * Originally Public Domain | |
29 | */ | |
30 | ||
31 | #define rot(x, k) (((x) << (k)) | ((x) >> (32 - (k)))) | |
32 | ||
33 | #define mix(a, b, c) \ | |
34 | do { \ | |
35 | a -= c; a ^= rot(c, 4); c += b; \ | |
36 | b -= a; b ^= rot(a, 6); a += c; \ | |
37 | c -= b; c ^= rot(b, 8); b += a; \ | |
38 | a -= c; a ^= rot(c, 16); c += b; \ | |
39 | b -= a; b ^= rot(a, 19); a += c; \ | |
40 | c -= b; c ^= rot(b, 4); b += a; \ | |
41 | } while (0) | |
42 | ||
43 | #define final(a, b, c) \ | |
44 | { \ | |
45 | c ^= b; c -= rot(b, 14); \ | |
46 | a ^= c; a -= rot(c, 11); \ | |
47 | b ^= a; b -= rot(a, 25); \ | |
48 | c ^= b; c -= rot(b, 16); \ | |
49 | a ^= c; a -= rot(c, 4); \ | |
50 | b ^= a; b -= rot(a, 14); \ | |
51 | c ^= b; c -= rot(b, 24); \ | |
52 | } | |
53 | ||
54 | #if (BYTE_ORDER == LITTLE_ENDIAN) | |
55 | #define HASH_LITTLE_ENDIAN 1 | |
56 | #else | |
57 | #define HASH_LITTLE_ENDIAN 0 | |
58 | #endif | |
59 | ||
60 | /* | |
61 | * | |
62 | * hashlittle() -- hash a variable-length key into a 32-bit value | |
63 | * k : the key (the unaligned variable-length array of bytes) | |
64 | * length : the length of the key, counting by bytes | |
65 | * initval : can be any 4-byte value | |
66 | * Returns a 32-bit value. Every bit of the key affects every bit of | |
67 | * the return value. Two keys differing by one or two bits will have | |
68 | * totally different hash values. | |
69 | * | |
70 | * The best hash table sizes are powers of 2. There is no need to do | |
71 | * mod a prime (mod is sooo slow!). If you need less than 32 bits, | |
72 | * use a bitmask. For example, if you need only 10 bits, do | |
73 | * h = (h & hashmask(10)); | |
74 | * In which case, the hash table should have hashsize(10) elements. | |
75 | * | |
76 | * If you are hashing n strings (uint8_t **)k, do it like this: | |
77 | * for (i = 0, h = 0; i < n; ++i) h = hashlittle(k[i], len[i], h); | |
78 | * | |
79 | * By Bob Jenkins, 2006. bob_jenkins@burtleburtle.net. You may use this | |
80 | * code any way you wish, private, educational, or commercial. It's free. | |
81 | * | |
82 | * Use for hash table lookup, or anything where one collision in 2^^32 is | |
83 | * acceptable. Do NOT use for cryptographic purposes. | |
84 | */ | |
85 | static | |
86 | uint32_t hashlittle(const void *key, size_t length, uint32_t initval) | |
87 | { | |
88 | uint32_t a, b, c; /* internal state */ | |
89 | union { | |
90 | const void *ptr; | |
91 | size_t i; | |
92 | } u; | |
93 | ||
94 | /* Set up the internal state */ | |
95 | a = b = c = 0xdeadbeef + ((uint32_t)length) + initval; | |
96 | ||
97 | u.ptr = key; | |
98 | if (HASH_LITTLE_ENDIAN && ((u.i & 0x3) == 0)) { | |
99 | const uint32_t *k = (const uint32_t *) key; /* read 32-bit chunks */ | |
100 | ||
101 | /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */ | |
102 | while (length > 12) { | |
103 | a += k[0]; | |
104 | b += k[1]; | |
105 | c += k[2]; | |
106 | mix(a, b, c); | |
107 | length -= 12; | |
108 | k += 3; | |
109 | } | |
110 | ||
111 | /*----------------------------- handle the last (probably partial) block */ | |
112 | /* | |
113 | * "k[2]&0xffffff" actually reads beyond the end of the string, but | |
114 | * then masks off the part it's not allowed to read. Because the | |
115 | * string is aligned, the masked-off tail is in the same word as the | |
116 | * rest of the string. Every machine with memory protection I've seen | |
117 | * does it on word boundaries, so is OK with this. But VALGRIND will | |
118 | * still catch it and complain. The masking trick does make the hash | |
119 | * noticably faster for short strings (like English words). | |
120 | */ | |
121 | #ifndef VALGRIND | |
122 | ||
123 | switch (length) { | |
124 | case 12: c+=k[2]; b+=k[1]; a+=k[0]; break; | |
125 | case 11: c+=k[2]&0xffffff; b+=k[1]; a+=k[0]; break; | |
126 | case 10: c+=k[2]&0xffff; b+=k[1]; a+=k[0]; break; | |
127 | case 9 : c+=k[2]&0xff; b+=k[1]; a+=k[0]; break; | |
128 | case 8 : b+=k[1]; a+=k[0]; break; | |
129 | case 7 : b+=k[1]&0xffffff; a+=k[0]; break; | |
130 | case 6 : b+=k[1]&0xffff; a+=k[0]; break; | |
131 | case 5 : b+=k[1]&0xff; a+=k[0]; break; | |
132 | case 4 : a+=k[0]; break; | |
133 | case 3 : a+=k[0]&0xffffff; break; | |
134 | case 2 : a+=k[0]&0xffff; break; | |
135 | case 1 : a+=k[0]&0xff; break; | |
136 | case 0 : return c; /* zero length strings require no mixing */ | |
137 | } | |
138 | ||
139 | #else /* make valgrind happy */ | |
140 | { | |
141 | const uint8_t *k8; | |
142 | ||
143 | k8 = (const uint8_t *) k; | |
144 | switch (length) { | |
145 | case 12: c+=k[2]; b+=k[1]; a+=k[0]; break; | |
146 | case 11: c+=((uint32_t) k8[10])<<16; /* fall through */ | |
147 | case 10: c+=((uint32_t) k8[9])<<8; /* fall through */ | |
148 | case 9 : c+=k8[8]; /* fall through */ | |
149 | case 8 : b+=k[1]; a+=k[0]; break; | |
150 | case 7 : b+=((uint32_t) k8[6])<<16; /* fall through */ | |
151 | case 6 : b+=((uint32_t) k8[5])<<8; /* fall through */ | |
152 | case 5 : b+=k8[4]; /* fall through */ | |
153 | case 4 : a+=k[0]; break; | |
154 | case 3 : a+=((uint32_t) k8[2])<<16; /* fall through */ | |
155 | case 2 : a+=((uint32_t) k8[1])<<8; /* fall through */ | |
156 | case 1 : a+=k8[0]; break; | |
157 | case 0 : return c; | |
158 | } | |
159 | } | |
160 | #endif /* !valgrind */ | |
161 | ||
162 | } else if (HASH_LITTLE_ENDIAN && ((u.i & 0x1) == 0)) { | |
163 | const uint16_t *k = (const uint16_t *) key; /* read 16-bit chunks */ | |
164 | const uint8_t *k8; | |
165 | ||
166 | /*--------------- all but last block: aligned reads and different mixing */ | |
167 | while (length > 12) | |
168 | { | |
169 | a += k[0] + (((uint32_t) k[1])<<16); | |
170 | b += k[2] + (((uint32_t) k[3])<<16); | |
171 | c += k[4] + (((uint32_t) k[5])<<16); | |
172 | mix(a, b, c); | |
173 | length -= 12; | |
174 | k += 6; | |
175 | } | |
176 | ||
177 | /*----------------------------- handle the last (probably partial) block */ | |
178 | k8 = (const uint8_t *) k; | |
179 | switch(length) | |
180 | { | |
181 | case 12: c+=k[4]+(((uint32_t) k[5])<<16); | |
182 | b+=k[2]+(((uint32_t) k[3])<<16); | |
183 | a+=k[0]+(((uint32_t) k[1])<<16); | |
184 | break; | |
185 | case 11: c+=((uint32_t) k8[10])<<16; /* fall through */ | |
186 | case 10: c+=k[4]; | |
187 | b+=k[2]+(((uint32_t) k[3])<<16); | |
188 | a+=k[0]+(((uint32_t) k[1])<<16); | |
189 | break; | |
190 | case 9 : c+=k8[8]; /* fall through */ | |
191 | case 8 : b+=k[2]+(((uint32_t) k[3])<<16); | |
192 | a+=k[0]+(((uint32_t) k[1])<<16); | |
193 | break; | |
194 | case 7 : b+=((uint32_t) k8[6])<<16; /* fall through */ | |
195 | case 6 : b+=k[2]; | |
196 | a+=k[0]+(((uint32_t) k[1])<<16); | |
197 | break; | |
198 | case 5 : b+=k8[4]; /* fall through */ | |
199 | case 4 : a+=k[0]+(((uint32_t) k[1])<<16); | |
200 | break; | |
201 | case 3 : a+=((uint32_t) k8[2])<<16; /* fall through */ | |
202 | case 2 : a+=k[0]; | |
203 | break; | |
204 | case 1 : a+=k8[0]; | |
205 | break; | |
206 | case 0 : return c; /* zero length requires no mixing */ | |
207 | } | |
208 | ||
209 | } else { /* need to read the key one byte at a time */ | |
210 | const uint8_t *k = (const uint8_t *)key; | |
211 | ||
212 | /*--------------- all but the last block: affect some 32 bits of (a, b, c) */ | |
213 | while (length > 12) { | |
214 | a += k[0]; | |
215 | a += ((uint32_t) k[1])<<8; | |
216 | a += ((uint32_t) k[2])<<16; | |
217 | a += ((uint32_t) k[3])<<24; | |
218 | b += k[4]; | |
219 | b += ((uint32_t) k[5])<<8; | |
220 | b += ((uint32_t) k[6])<<16; | |
221 | b += ((uint32_t) k[7])<<24; | |
222 | c += k[8]; | |
223 | c += ((uint32_t) k[9])<<8; | |
224 | c += ((uint32_t) k[10])<<16; | |
225 | c += ((uint32_t) k[11])<<24; | |
226 | mix(a,b,c); | |
227 | length -= 12; | |
228 | k += 12; | |
229 | } | |
230 | ||
231 | /*-------------------------------- last block: affect all 32 bits of (c) */ | |
232 | switch (length) { /* all the case statements fall through */ | |
447c9339 MJ |
233 | case 12: c+=((uint32_t) k[11])<<24; /* fall through */ |
234 | case 11: c+=((uint32_t) k[10])<<16; /* fall through */ | |
235 | case 10: c+=((uint32_t) k[9])<<8; /* fall through */ | |
236 | case 9 : c+=k[8]; /* fall through */ | |
237 | case 8 : b+=((uint32_t) k[7])<<24; /* fall through */ | |
238 | case 7 : b+=((uint32_t) k[6])<<16; /* fall through */ | |
239 | case 6 : b+=((uint32_t) k[5])<<8; /* fall through */ | |
240 | case 5 : b+=k[4]; /* fall through */ | |
241 | case 4 : a+=((uint32_t) k[3])<<24; /* fall through */ | |
242 | case 3 : a+=((uint32_t) k[2])<<16; /* fall through */ | |
243 | case 2 : a+=((uint32_t) k[1])<<8; /* fall through */ | |
f7e7a1b8 MD |
244 | case 1 : a+=k[0]; |
245 | break; | |
246 | case 0 : return c; | |
247 | } | |
248 | } | |
249 | ||
250 | final(a, b, c); | |
251 | return c; | |
252 | } | |
253 | ||
254 | static inline | |
255 | uint32_t jhash(const void *key, size_t length, uint32_t seed) | |
256 | { | |
257 | return hashlittle(key, length, seed); | |
258 | } | |
259 | ||
260 | #endif /* _JHASH_H */ |