On Fri, Dec 6, 2013 at 1:45 PM, Ketil Malde
Then I thought I'd look at hash tables, using the 'hashtables' package. I haven't tested it much yet, but it appears to be a lot slower than Judy (maybe as much as 10x), and uses a lot more memory (also perhaps a factor of 10). I guess I might be able to improve things a bit by judiciously applying strictness, but it seems to be storing both keys and values unboxed, so I don't expect to come close to Judy - I guess there isn't any unboxed hash table implementations around?
If you want to try the git master version of the hashtables library, I've
made some performance and memory overhead improvements that haven't been
released yet (I still need to run more benchmarks before release). Try both
the "basic" and "cuckoo" hash tables (cuckoo might be better). IIRC we
force keys stored in the hash tables but not the values -- you might want
to confirm you're not building up value thunks.
G
--
Gregory Collins