Re: [ANN] LuaJIT-2.0.0-beta1

["Juris Kalnins" <juris@...>]

On Mon, 02 Nov 2009 15:25:16 +0200, Mike Pall <mikelu-0911@mike.de> wrote:

> Not with the hash per-se but with the selection of the characters
> to hash. It badly hashed strings that contained URLs.

Thanks, this is useful information.

I was thinking how much it is possible to gain by searching for hash  
function
that gives least number of lower bit collisions on the set of keys that
are actually used together in the application. (e.g. names of modules,  
names of metamethods, ...)

Also, if I understand correctly, LJ 2 can much optimize table lookups if  
the key lives at is 'primary' position?