2 * Copyright (C) 2011 Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
4 * This library is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Lesser General Public
6 * License as published by the Free Software Foundation;
7 * version 2.1 of the License.
9 * This library is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * Lesser General Public License for more details.
14 * You should have received a copy of the GNU Lesser General Public
15 * License along with this library; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 #include <urcu/compiler.h>
20 #include <lttng/ust-endian.h>
24 * Source: http://burtleburtle.net/bob/c/lookup3.c
25 * Originally Public Domain
28 #define rot(x, k) (((x) << (k)) | ((x) >> (32 - (k))))
30 #define mix(a, b, c) \
32 a -= c; a ^= rot(c, 4); c += b; \
33 b -= a; b ^= rot(a, 6); a += c; \
34 c -= b; c ^= rot(b, 8); b += a; \
35 a -= c; a ^= rot(c, 16); c += b; \
36 b -= a; b ^= rot(a, 19); a += c; \
37 c -= b; c ^= rot(b, 4); b += a; \
40 #define final(a, b, c) \
42 c ^= b; c -= rot(b, 14); \
43 a ^= c; a -= rot(c, 11); \
44 b ^= a; b -= rot(a, 25); \
45 c ^= b; c -= rot(b, 16); \
46 a ^= c; a -= rot(c, 4);\
47 b ^= a; b -= rot(a, 14); \
48 c ^= b; c -= rot(b, 24); \
51 #if (BYTE_ORDER == LITTLE_ENDIAN)
52 #define HASH_LITTLE_ENDIAN 1
54 #define HASH_LITTLE_ENDIAN 0
59 * hashlittle() -- hash a variable-length key into a 32-bit value
60 * k : the key (the unaligned variable-length array of bytes)
61 * length : the length of the key, counting by bytes
62 * initval : can be any 4-byte value
63 * Returns a 32-bit value. Every bit of the key affects every bit of
64 * the return value. Two keys differing by one or two bits will have
65 * totally different hash values.
67 * The best hash table sizes are powers of 2. There is no need to do
68 * mod a prime (mod is sooo slow!). If you need less than 32 bits,
69 * use a bitmask. For example, if you need only 10 bits, do
70 * h = (h & hashmask(10));
71 * In which case, the hash table should have hashsize(10) elements.
73 * If you are hashing n strings (uint8_t **)k, do it like this:
74 * for (i = 0, h = 0; i < n; ++i) h = hashlittle(k[i], len[i], h);
76 * By Bob Jenkins, 2006. bob_jenkins@burtleburtle.net. You may use this
77 * code any way you wish, private, educational, or commercial. It's free.
79 * Use for hash table lookup, or anything where one collision in 2^^32 is
80 * acceptable. Do NOT use for cryptographic purposes.
83 uint32_t hashlittle(const void *key
, size_t length
, uint32_t initval
)
85 uint32_t a
, b
, c
; /* internal state */
91 /* Set up the internal state */
92 a
= b
= c
= 0xdeadbeef + ((uint32_t)length
) + initval
;
95 if (HASH_LITTLE_ENDIAN
&& ((u
.i
& 0x3) == 0)) {
96 const uint32_t *k
= (const uint32_t *) key
; /* read 32-bit chunks */
98 /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */
108 /*----------------------------- handle the last (probably partial) block */
110 * "k[2]&0xffffff" actually reads beyond the end of the string, but
111 * then masks off the part it's not allowed to read. Because the
112 * string is aligned, the masked-off tail is in the same word as the
113 * rest of the string. Every machine with memory protection I've seen
114 * does it on word boundaries, so is OK with this. But VALGRIND will
115 * still catch it and complain. The masking trick does make the hash
116 * noticably faster for short strings (like English words).
121 case 12: c
+=k
[2]; b
+=k
[1]; a
+=k
[0]; break;
122 case 11: c
+=k
[2]&0xffffff; b
+=k
[1]; a
+=k
[0]; break;
123 case 10: c
+=k
[2]&0xffff; b
+=k
[1]; a
+=k
[0]; break;
124 case 9 : c
+=k
[2]&0xff; b
+=k
[1]; a
+=k
[0]; break;
125 case 8 : b
+=k
[1]; a
+=k
[0]; break;
126 case 7 : b
+=k
[1]&0xffffff; a
+=k
[0]; break;
127 case 6 : b
+=k
[1]&0xffff; a
+=k
[0]; break;
128 case 5 : b
+=k
[1]&0xff; a
+=k
[0]; break;
129 case 4 : a
+=k
[0]; break;
130 case 3 : a
+=k
[0]&0xffffff; break;
131 case 2 : a
+=k
[0]&0xffff; break;
132 case 1 : a
+=k
[0]&0xff; break;
133 case 0 : return c
; /* zero length strings require no mixing */
136 #else /* make valgrind happy */
140 k8
= (const uint8_t *) k
;
142 case 12: c
+=k
[2]; b
+=k
[1]; a
+=k
[0]; break;
143 case 11: c
+=((uint32_t) k8
[10])<<16; /* fall through */
144 case 10: c
+=((uint32_t) k8
[9])<<8; /* fall through */
145 case 9 : c
+=k8
[8]; /* fall through */
146 case 8 : b
+=k
[1]; a
+=k
[0]; break;
147 case 7 : b
+=((uint32_t) k8
[6])<<16; /* fall through */
148 case 6 : b
+=((uint32_t) k8
[5])<<8; /* fall through */
149 case 5 : b
+=k8
[4]; /* fall through */
150 case 4 : a
+=k
[0]; break;
151 case 3 : a
+=((uint32_t) k8
[2])<<16; /* fall through */
152 case 2 : a
+=((uint32_t) k8
[1])<<8; /* fall through */
153 case 1 : a
+=k8
[0]; break;
157 #endif /* !valgrind */
159 } else if (HASH_LITTLE_ENDIAN
&& ((u
.i
& 0x1) == 0)) {
160 const uint16_t *k
= (const uint16_t *) key
; /* read 16-bit chunks */
163 /*--------------- all but last block: aligned reads and different mixing */
166 a
+= k
[0] + (((uint32_t) k
[1])<<16);
167 b
+= k
[2] + (((uint32_t) k
[3])<<16);
168 c
+= k
[4] + (((uint32_t) k
[5])<<16);
174 /*----------------------------- handle the last (probably partial) block */
175 k8
= (const uint8_t *) k
;
178 case 12: c
+=k
[4]+(((uint32_t) k
[5])<<16);
179 b
+=k
[2]+(((uint32_t) k
[3])<<16);
180 a
+=k
[0]+(((uint32_t) k
[1])<<16);
182 case 11: c
+=((uint32_t) k8
[10])<<16; /* fall through */
184 b
+=k
[2]+(((uint32_t) k
[3])<<16);
185 a
+=k
[0]+(((uint32_t) k
[1])<<16);
187 case 9 : c
+=k8
[8]; /* fall through */
188 case 8 : b
+=k
[2]+(((uint32_t) k
[3])<<16);
189 a
+=k
[0]+(((uint32_t) k
[1])<<16);
191 case 7 : b
+=((uint32_t) k8
[6])<<16; /* fall through */
193 a
+=k
[0]+(((uint32_t) k
[1])<<16);
195 case 5 : b
+=k8
[4]; /* fall through */
196 case 4 : a
+=k
[0]+(((uint32_t) k
[1])<<16);
198 case 3 : a
+=((uint32_t) k8
[2])<<16; /* fall through */
203 case 0 : return c
; /* zero length requires no mixing */
206 } else { /* need to read the key one byte at a time */
207 const uint8_t *k
= (const uint8_t *)key
;
209 /*--------------- all but the last block: affect some 32 bits of (a, b, c) */
210 while (length
> 12) {
212 a
+= ((uint32_t) k
[1])<<8;
213 a
+= ((uint32_t) k
[2])<<16;
214 a
+= ((uint32_t) k
[3])<<24;
216 b
+= ((uint32_t) k
[5])<<8;
217 b
+= ((uint32_t) k
[6])<<16;
218 b
+= ((uint32_t) k
[7])<<24;
220 c
+= ((uint32_t) k
[9])<<8;
221 c
+= ((uint32_t) k
[10])<<16;
222 c
+= ((uint32_t) k
[11])<<24;
228 /*-------------------------------- last block: affect all 32 bits of (c) */
229 switch (length
) { /* all the case statements fall through */
230 case 12: c
+=((uint32_t) k
[11])<<24;
231 case 11: c
+=((uint32_t) k
[10])<<16;
232 case 10: c
+=((uint32_t) k
[9])<<8;
234 case 8 : b
+=((uint32_t) k
[7])<<24;
235 case 7 : b
+=((uint32_t) k
[6])<<16;
236 case 6 : b
+=((uint32_t) k
[5])<<8;
238 case 4 : a
+=((uint32_t) k
[3])<<24;
239 case 3 : a
+=((uint32_t) k
[2])<<16;
240 case 2 : a
+=((uint32_t) k
[1])<<8;
252 uint32_t jhash(const void *key
, size_t length
, uint32_t seed
)
254 return hashlittle(key
, length
, seed
);
This page took 0.036293 seconds and 4 git commands to generate.