Re: Do we know why Lucene's HNSW may be slower than other HNSW implementations?
Chris Hegarty <[email protected]>
| Newsgroups | gmane.comp.jakarta.lucene.devel |
|---|---|
| Message-ID | <[email protected]> |
> On 23 Jun 2025, at 14:09, Benjamin Trent <[email protected]> wrote: > > - Reading floats onto heap float[] instead of accessing Memory Segments directly when scoring > This is on my list of things to do, now that we’re moving main to a minimum of Java 25, it helps a lot. -Chris.