[Ecls-list] string hashing

[Dustin Long]

Andrew Schein wrote:
> Hi all -
>
> I was having a look at some common lisp benchmarks using a 2004 release of ECL:
>
> http://www.cliki.net/Performance%20Benchmarks
>
> and I see that ECL was relatively slow in some areas of interest, namely
> string hashing.  I am wondering to what extent hash table performance
> has improved since then?  Has there been any rewrite of this functionality?
>
> Thanks,
>
> Andrew
>   
Looking at the code in hash.d, the hashing looks really complicated, 
with many operations per byte. I didn't do any analysis or testing 
myself, but maybe if someone would like to improve it, I've heard that 
Jenkin's One-at-a-Time works really well:

http://burtleburtle.net/bob/hash/doobs.html
http://en.wikipedia.org/wiki/Hash_table#Choosing_a_good_hash_function

I've used it in some of my own code. It's speedy, simple, and leads to 
very good results.

Dustin