Tony-X commented on issue #16556: URL: https://github.com/apache/lucene/issues/16556#issuecomment-5502246997
> this optimization amortizes the cost of loading chunks of the query vector across N documents, instead of loading it once per document Yes, we should leverage it more! Instead of moving, it'd be simpler to think about extending [VectorUtilSupport](https://github.com/apache/lucene/blob/2822b28b5ab042672d6cb569589128e5dcafa648/lucene/core/src/java/org/apache/lucene/internal/vectorization/VectorUtilSupport.java#L25) with methods that target single-query-many-doc scoring. Then the [RandomVectorScorer](https://github.com/apache/lucene/blob/2822b28b5ab042672d6cb569589128e5dcafa648/lucene/core/src/java/org/apache/lucene/util/hnsw/RandomVectorScorer.java#L37-L54) implementors can start consuming those new methods. This way we minimize the API surface churn. -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
