Some of the technical details of Redis source are worth learning 02

1. Implementation of the hash function in Redis:

Redis uses a different hash function for integer key and string key

For integer key,redis, the Dict.c/dictinthashfunction function is implemented with the use of Thomas Wang's all-bit Mix functions:

1 /*Thomas Wang's + bit Mix Function*/2UnsignedintDictinthashfunction (unsignedintkey)3 {4Key + = ~ (Key << the);5Key ^= (Key >>Ten);6Key + = (Key <<3);7Key ^= (Key >>6);8Key + = ~ (Key << One);9Key ^= (Key >> -);Ten     returnkey; One}

The beauty of this code I did not have time to study carefully, and so the study will be here to make up, but found a two at the beginning of a good link:

The first is the link to the Thomas Wang God Himself:

Http://web.archive.org/web/20071223173210/http://www.concentric.net/~Ttwang/tech/inthash.htm

The other is the summary written by others based on the links above and other information

http://blog.csdn.net/jasper_xulei/article/details/18364313

For Key,redis in string form, the MURMURHASH2 algorithm and the DJB algorithm are used:

The MURMURHASH2 algorithm is case sensitive to key and produces inconsistent results on big-endian machines and small-end machines.

The dict.c/dictgenhashfunction of Redis is the C language implementation of the MURMURHASH2 algorithm:

1UnsignedintDictgenhashfunction (Const void*key,intLen) {2     /*' m ' and ' R ' are mixing constants generated offline.3 they ' re not really ' magic ', they just happen to work well. */4uint32_t seed =Dict_hash_function_seed;5     Constuint32_t m =0x5bd1e995;6     Const intR = -;7 8     /*Initialize the hash to a ' random ' value*/9uint32_t h = seed ^Len;Ten  One     /*Mix 4 bytes at a time into the hash*/ A     ConstUnsignedChar*data = (ConstUnsignedChar*) key; -  -      while(Len >=4) { theuint32_t k = * (uint32_t*) data; -  -K *=m; -K-^= K >>R; +K *=m; -  +H *=m; AH ^=K; at  -Data + =4; -Len-=4; -     } -  -     /*Handle The last few bytes of the input array*/ in     Switch(len) { -      Case 3: H ^= data[2] << -; to      Case 2: H ^= data[1] <<8; +      Case 1: H ^= data[0]; H *=m; -     }; the  *     /*Do a few final mixes of the "hash to ensure" the last few $ * Bytes is well-incorporated.*/Panax NotoginsengH ^= H >> -; -H *=m; theH ^= H >> the; +  A     return(unsignedint) H; the}

Redis, by virtue of the DJB function, implements a case-insensitive hash function dict.c/dictgencasehashfunction:

1UnsignedintDictgencasehashfunction (ConstUnsignedChar*buf,intLen) {2Unsignedinthash = (unsignedint) Dict_hash_function_seed;3 4      while(len--)5hash = (Hash <<5+ hash) + (ToLower (*buf++));/*Hash * + C*/6     returnHash;7}

The above three hash functions (Dictinthashfunction, Dictinthashfunction, dictgencasehashfunction) are used in different areas of Redis to achieve the hash requirements in different situations, This will be explained in more detail next.

The use of several different hashing functions in different situations in 2.Redis

Some of the technical details of Redis source are worth learning 02