Re: A good hash algorithm

[Luiz Henrique de Figueiredo <lhf@...>]

>variations on a simple old algorithm[1] turn out to have
>similar runtime performance, and significantly lower collision rates than
>Jenkins's
>
>[1] djb2 algorithm: http://www.cs.yorku.ca/~oz/hash.html

Lua's hash function for strings is similar to djb2.
The core of djb2 is
    h = ((h << 5) + h) + c

The core of Lua's function
    h = h ^ ((h<<5)+(h>>2)+c)

Of course, being similar does not guarantee anything about performance, but
we're very satisfied with the overall performance (code size and speed) of
Lua's hashing.
--lhf