At this year's Haskell Implementors Workshop I gave talk about performance improvements to the unordered-containers package. I've talked about this package before. New since last time I gave this talk are performance numbers from the hash array mapped trie data implementation I've been working on. Here are the slides:
(You can also download the slides as a PDF.)
Any news on the HAMT branch of unordered-containers? Are you going to release it?
ReplyDeleteBill,
ReplyDeleteI still plan to release it. There are two outstanding issues:
1. The benchmark results show high variance. I'd like to understand why this is. Perhaps the GC doesn't work as well as it could on small arrays.
2. I need to optimize a bunch of functions (e.g. map, filter, union) so their performance doesn't regress compared to the release version of unordered-containers.
Unfortunately I haven't found the time to do this yet.