1. 9
  1. 1

    Mr. Wellons posted an update:

    About one week after publishing this article I found an even better hash function. I believe this is the least biased 32-bit integer hash function *of this form* ever devised. It’s even less biased than the MurmurHash3 finalizer.


    If you’re willing to use an additional round of multiply-xorshift, this next function actually reaches the theoretical bias limit (bias = ~0.021) as exhibited by a perfect integer hash function

    …I think he’s a bot. /s :)