benwbooth / Set-IntervalTree

Perform range-based lookups on sets of ranges.
3 stars 3 forks source link

Performance regression after version 0.10 #10

Open RoelKluin opened 4 years ago

RoelKluin commented 4 years ago

In version 0.12 we've noticed for some time that there's a severe performance degradation, from a few minutes (v 0.10) to hours (v 0.12) for a typical RNAseq sample. We use Set::IntervalTree in our tool for RNAseq gene counting, probably mostly in-house, but there is a public repository here:

https://github.com/NKI-GCF/itreecount

benwbooth commented 4 years ago

Hi @RoelKluin

Set::IntervalTree is not maintained anymore. If you want I can add you as a maintainer so you can fix the performance issues. It was designed as an XS wrapper around a C++ Interval Tree implementation I found on the internet (I forgot where!). To be honest I never dug into how the underlying implementation works. So I wouldn't be much help in figuring out the performance problems.