msokolov edited a comment on pull request #2108: URL: https://github.com/apache/lucene-solr/pull/2108#issuecomment-741790004
I found some problems with the first implementation posted here, which used a tricky data structure combining parallel arrays with a separate sortable key held in an (int) priority queue. It had bugs, and I realized was wasteful too. This replaces the int queue with a long queue, where the parallel arrays (float score + int node id) are folded together into a single (sortable by score) long. I measured 20% index time improvement (over the baseline, not the previous impl here). My plan is to push as soon as we have an indexing time datapoint available in lucene nightly benchmarks. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
