Annotation of embedaddon/trafshow/lookupa.c, revision 1.1.1.1
1.1 misho 1: /*
2: --------------------------------------------------------------------
3: lookupa.c, by Bob Jenkins, December 1996. Same as lookup2.c
4: Use this code however you wish. Public Domain. No warranty.
5: Source is http://burtleburtle.net/bob/c/lookupa.c
6: --------------------------------------------------------------------
7: */
8: #ifndef STANDARD
9: #include "standard.h"
10: #endif
11: #ifndef LOOKUPA
12: #include "lookupa.h"
13: #endif
14:
15: /*
16: --------------------------------------------------------------------
17: mix -- mix 3 32-bit values reversibly.
18: For every delta with one or two bit set, and the deltas of all three
19: high bits or all three low bits, whether the original value of a,b,c
20: is almost all zero or is uniformly distributed,
21: * If mix() is run forward or backward, at least 32 bits in a,b,c
22: have at least 1/4 probability of changing.
23: * If mix() is run forward, every bit of c will change between 1/3 and
24: 2/3 of the time. (Well, 22/100 and 78/100 for some 2-bit deltas.)
25: mix() was built out of 36 single-cycle latency instructions in a
26: structure that could supported 2x parallelism, like so:
27: a -= b;
28: a -= c; x = (c>>13);
29: b -= c; a ^= x;
30: b -= a; x = (a<<8);
31: c -= a; b ^= x;
32: c -= b; x = (b>>13);
33: ...
34: Unfortunately, superscalar Pentiums and Sparcs can't take advantage
35: of that parallelism. They've also turned some of those single-cycle
36: latency instructions into multi-cycle latency instructions. Still,
37: this is the fastest good hash I could find. There were about 2^^68
38: to choose from. I only looked at a billion or so.
39: --------------------------------------------------------------------
40: */
41: #define mix(a,b,c) \
42: { \
43: a -= b; a -= c; a ^= (c>>13); \
44: b -= c; b -= a; b ^= (a<<8); \
45: c -= a; c -= b; c ^= (b>>13); \
46: a -= b; a -= c; a ^= (c>>12); \
47: b -= c; b -= a; b ^= (a<<16); \
48: c -= a; c -= b; c ^= (b>>5); \
49: a -= b; a -= c; a ^= (c>>3); \
50: b -= c; b -= a; b ^= (a<<10); \
51: c -= a; c -= b; c ^= (b>>15); \
52: }
53:
54: /*
55: --------------------------------------------------------------------
56: lookup() -- hash a variable-length key into a 32-bit value
57: k : the key (the unaligned variable-length array of bytes)
58: len : the length of the key, counting by bytes
59: level : can be any 4-byte value
60: Returns a 32-bit value. Every bit of the key affects every bit of
61: the return value. Every 1-bit and 2-bit delta achieves avalanche.
62: About 6len+35 instructions.
63:
64: The best hash table sizes are powers of 2. There is no need to do
65: mod a prime (mod is sooo slow!). If you need less than 32 bits,
66: use a bitmask. For example, if you need only 10 bits, do
67: h = (h & hashmask(10));
68: In which case, the hash table should have hashsize(10) elements.
69:
70: If you are hashing n strings (ub1 **)k, do it like this:
71: for (i=0, h=0; i<n; ++i) h = lookup( k[i], len[i], h);
72:
73: By Bob Jenkins, 1996. bob_jenkins@burtleburtle.net. You may use this
74: code any way you wish, private, educational, or commercial.
75:
76: See http://burtleburtle.net/bob/hash/evahash.html
77: Use for hash table lookup, or anything where one collision in 2^32 is
78: acceptable. Do NOT use for cryptographic purposes.
79: --------------------------------------------------------------------
80: */
81:
82: ub4 lookup( k, length, level)
83: register ub1 *k; /* the key */
84: register ub4 length; /* the length of the key */
85: register ub4 level; /* the previous hash, or an arbitrary value */
86: {
87: register ub4 a,b,c,len;
88:
89: /* Set up the internal state */
90: len = length;
91: a = b = 0x9e3779b9; /* the golden ratio; an arbitrary value */
92: c = level; /* the previous hash value */
93:
94: /*---------------------------------------- handle most of the key */
95: while (len >= 12)
96: {
97: a += (k[0] +((ub4)k[1]<<8) +((ub4)k[2]<<16) +((ub4)k[3]<<24));
98: b += (k[4] +((ub4)k[5]<<8) +((ub4)k[6]<<16) +((ub4)k[7]<<24));
99: c += (k[8] +((ub4)k[9]<<8) +((ub4)k[10]<<16)+((ub4)k[11]<<24));
100: mix(a,b,c);
101: k += 12; len -= 12;
102: }
103:
104: /*------------------------------------- handle the last 11 bytes */
105: c += length;
106: switch(len) /* all the case statements fall through */
107: {
108: case 11: c+=((ub4)k[10]<<24);
109: case 10: c+=((ub4)k[9]<<16);
110: case 9 : c+=((ub4)k[8]<<8);
111: /* the first byte of c is reserved for the length */
112: case 8 : b+=((ub4)k[7]<<24);
113: case 7 : b+=((ub4)k[6]<<16);
114: case 6 : b+=((ub4)k[5]<<8);
115: case 5 : b+=k[4];
116: case 4 : a+=((ub4)k[3]<<24);
117: case 3 : a+=((ub4)k[2]<<16);
118: case 2 : a+=((ub4)k[1]<<8);
119: case 1 : a+=k[0];
120: /* case 0: nothing left to add */
121: }
122: mix(a,b,c);
123: /*-------------------------------------------- report the result */
124: return c;
125: }
126:
127:
128: /*
129: --------------------------------------------------------------------
130: mixc -- mixc 8 4-bit values as quickly and thoroughly as possible.
131: Repeating mix() three times achieves avalanche.
132: Repeating mix() four times eliminates all funnels and all
133: characteristics stronger than 2^{-11}.
134: --------------------------------------------------------------------
135: */
136: #define mixc(a,b,c,d,e,f,g,h) \
137: { \
138: a^=b<<11; d+=a; b+=c; \
139: b^=c>>2; e+=b; c+=d; \
140: c^=d<<8; f+=c; d+=e; \
141: d^=e>>16; g+=d; e+=f; \
142: e^=f<<10; h+=e; f+=g; \
143: f^=g>>4; a+=f; g+=h; \
144: g^=h<<8; b+=g; h+=a; \
145: h^=a>>9; c+=h; a+=b; \
146: }
147:
148: /*
149: --------------------------------------------------------------------
150: checksum() -- hash a variable-length key into a 256-bit value
151: k : the key (the unaligned variable-length array of bytes)
152: len : the length of the key, counting by bytes
153: state : an array of CHECKSTATE 4-byte values (256 bits)
154: The state is the checksum. Every bit of the key affects every bit of
155: the state. There are no funnels. About 112+6.875len instructions.
156:
157: If you are hashing n strings (ub1 **)k, do it like this:
158: for (i=0; i<8; ++i) state[i] = 0x9e3779b9;
159: for (i=0, h=0; i<n; ++i) checksum( k[i], len[i], state);
160:
161: (c) Bob Jenkins, 1996. bob_jenkins@burtleburtle.net. You may use this
162: code any way you wish, private, educational, or commercial, as long
163: as this whole comment accompanies it.
164:
165: See http://burtleburtle.net/bob/hash/evahash.html
166: Use to detect changes between revisions of documents, assuming nobody
167: is trying to cause collisions. Do NOT use for cryptography.
168: --------------------------------------------------------------------
169: */
170: void checksum( k, len, state)
171: register ub1 *k;
172: register ub4 len;
173: register ub4 *state;
174: {
175: register ub4 a,b,c,d,e,f,g,h,length;
176:
177: /* Use the length and level; add in the golden ratio. */
178: length = len;
179: a=state[0]; b=state[1]; c=state[2]; d=state[3];
180: e=state[4]; f=state[5]; g=state[6]; h=state[7];
181:
182: /*---------------------------------------- handle most of the key */
183: while (len >= 32)
184: {
185: a += (k[0] +(k[1]<<8) +(k[2]<<16) +(k[3]<<24));
186: b += (k[4] +(k[5]<<8) +(k[6]<<16) +(k[7]<<24));
187: c += (k[8] +(k[9]<<8) +(k[10]<<16)+(k[11]<<24));
188: d += (k[12]+(k[13]<<8)+(k[14]<<16)+(k[15]<<24));
189: e += (k[16]+(k[17]<<8)+(k[18]<<16)+(k[19]<<24));
190: f += (k[20]+(k[21]<<8)+(k[22]<<16)+(k[23]<<24));
191: g += (k[24]+(k[25]<<8)+(k[26]<<16)+(k[27]<<24));
192: h += (k[28]+(k[29]<<8)+(k[30]<<16)+(k[31]<<24));
193: mixc(a,b,c,d,e,f,g,h);
194: mixc(a,b,c,d,e,f,g,h);
195: mixc(a,b,c,d,e,f,g,h);
196: mixc(a,b,c,d,e,f,g,h);
197: k += 32; len -= 32;
198: }
199:
200: /*------------------------------------- handle the last 31 bytes */
201: h += length;
202: switch(len)
203: {
204: case 31: h+=(k[30]<<24);
205: case 30: h+=(k[29]<<16);
206: case 29: h+=(k[28]<<8);
207: case 28: g+=(k[27]<<24);
208: case 27: g+=(k[26]<<16);
209: case 26: g+=(k[25]<<8);
210: case 25: g+=k[24];
211: case 24: f+=(k[23]<<24);
212: case 23: f+=(k[22]<<16);
213: case 22: f+=(k[21]<<8);
214: case 21: f+=k[20];
215: case 20: e+=(k[19]<<24);
216: case 19: e+=(k[18]<<16);
217: case 18: e+=(k[17]<<8);
218: case 17: e+=k[16];
219: case 16: d+=(k[15]<<24);
220: case 15: d+=(k[14]<<16);
221: case 14: d+=(k[13]<<8);
222: case 13: d+=k[12];
223: case 12: c+=(k[11]<<24);
224: case 11: c+=(k[10]<<16);
225: case 10: c+=(k[9]<<8);
226: case 9 : c+=k[8];
227: case 8 : b+=(k[7]<<24);
228: case 7 : b+=(k[6]<<16);
229: case 6 : b+=(k[5]<<8);
230: case 5 : b+=k[4];
231: case 4 : a+=(k[3]<<24);
232: case 3 : a+=(k[2]<<16);
233: case 2 : a+=(k[1]<<8);
234: case 1 : a+=k[0];
235: }
236: mixc(a,b,c,d,e,f,g,h);
237: mixc(a,b,c,d,e,f,g,h);
238: mixc(a,b,c,d,e,f,g,h);
239: mixc(a,b,c,d,e,f,g,h);
240:
241: /*-------------------------------------------- report the result */
242: state[0]=a; state[1]=b; state[2]=c; state[3]=d;
243: state[4]=e; state[5]=f; state[6]=g; state[7]=h;
244: }
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>