Benchmark on Intel(R) Core(TM) i5-2500S CPU @ 2.70GHz (albeit in a VM)

benchmark              old ns/op    new ns/op    delta
BenchmarkHash8Bytes         1606          699  -56.48%
BenchmarkHash1K            21920         7268  -66.84%
BenchmarkHash8K           165696        53756  -67.56%

benchmark               old MB/s     new MB/s  speedup
BenchmarkHash8Bytes         4.98        11.44    2.30x
BenchmarkHash1K            46.72       140.88    3.02x
BenchmarkHash8K            49.44       152.39    3.08x

R=agl, cldorian
CC=golang-codereviews
https://golang.org/cl/44800044