[GitHub] [lucene] mayya-sharipova commented on pull request #992: LUCENE-10592 Build HNSW Graph on indexing

[GitBox <gi...@apache.org>]

mayya-sharipova commented on PR #992:
URL: https://github.com/apache/lucene/pull/992#issuecomment-1171306876

   > I am a bit surprised about the benchmark results. In [LUCENE-10375](https://issues.apache.org/jira/browse/LUCENE-10375), we found that writing all vectors to disk before building the graph sped up indexing (not just merging). This change goes back to the strategy of using on-heap vectors to build the graph, so I'd expect a slowdown. Here are the benchmark results from that issue:
   
   @jtibshirani Thank for the initial review. "write vectors" here means the time for the whole flush operation. In for the main branch, as we build a graph during flush it takes a lot of time (840392 msec); while for this PR flush operation is fast (1017 msec )
   
   


-- 
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.

To unsubscribe, e-mail: issues-unsubscribe@lucene.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@lucene.apache.org
For additional commands, e-mail: issues-help@lucene.apache.org