Re: [TCLCORE] Upgrade of Tcl's String Hash Algorithm

[Donal K. F. <don...@ma...>]

On 16/02/2010 18:30, Tom Jackson wrote:
> Try out my Tcl hash functions in my previous email. The problem is not
> the hash function, it is generally the method of transforming the hash
> value into a bucket index. The better performing hashes shown in your
> post are the result of a lot of randomness in the low order bits. This
> is to be expected of a hash which mixes all the bits in with the new
> bits.
>
> If the entire hash value was used, it could make a lot of sense to
> increase the sophistication of the hash function, but it would be much
> easier to just take a prime modulus of hash value just once. This
> would take into account all the bits.

Actually, if we were using lookup3 then the entropy is distributed
pretty evenly across the entire key. It's a very sophisticated hash
function that's resistant to a lot of odd abuses.

Donal.