Content deleted Content added
Line 12:
The fastest implementation[http://murmurhash.googlepages.com/MurmurHash2.cpp] of MurmurHash2 reads four bytes at a time and generates canonical values only on a processor with Intel [[little-endian]] byte order. For processors that require aligned reads[http://murmurhash.googlepages.com/MurmurHashAligned2.cpp] or have Motorola [[big-endian]] byte order[http://murmurhash.googlepages.com/MurmurHashNeutral2.cpp], there are versions that work correctly but at approximately half speed. The algorithm generates a 32-bit hash value and is optimized for processors with 32-bit or larger registers.
There are also two versions[http://code.google.com/p/maatkit/issues/detail?id=19#c4] available which generate 64-bit hash values, suitable for differentiating among tens of thousands of items without false positives caused by the [[birthday paradox]]. The first, MurmurHash64A, is MurmurHash2 scaled up directly to 64 bits, for maximum performance on 64-bit processors. The second, MurmurHash64B, is designed for 32-bit processors,
===MurmurHash2===
|