Speed up OptimizedScalarQuantizer #131599
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters
elasticsearchmachine added the Team:Search Relevance 
Collaborator
| Pinging @elastic/es-search-relevance (Team:Search Relevance) |
Collaborator
| Hi @iverase, I've created a changelog YAML for you. |
benwtrent approved these changes Jul 22, 2025
Member
benwtrent left a comment
benwtrent left a comment There was a problem hiding this comment.
This optimization makes sense to me.
We don't need to keep the legacy interface.
My only concern is making sure recall is unchanged. Looking at the code, all the paths already did a "Math.round" except now some of the paths are using int instead of rounding floats. Which is fine.
The speed ups are hilarious!
Contributor Author
I am pretty sure the new code is equivalent to the old one, we are just caching the results from the resulls of |
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit. This suggestion is invalid because no changes were made to the code. Suggestions cannot be applied while the pull request is closed. Suggestions cannot be applied while viewing a subset of changes. Only one suggestion per line can be applied in a batch. Add this suggestion to a batch that can be applied as a single commit. Applying suggestions on deleted lines is not supported. You must change the existing code in this line in order to create a valid suggestion. Outdated suggestions cannot be applied. This suggestion has been applied or marked resolved. Suggestions cannot be applied from pending reviews. Suggestions cannot be applied on multi-line comments. Suggestions cannot be applied while the pull request is queued to merge. Suggestion cannot be applied right now. Please check back later.
While reviewing the code in OptimizedScalarQuantizer, I noticed that we are quantizing the same vector a few times, once when we computing the loss and then again when computing the next grid points.I wondered if we could reuse the valu between those two calls and avoid that repeated computation.
This PR does that, it uses the destination array to keep the quantize value during the loss computation and give to the method computing the grid points. In addition we can skip the final quantization of the vector if the method that optimize the intervals finishes without computing a worst loss.
The only side effect is that we need to remove the legacy method on osq. That's ok as it was only used for benchmark comparison.
The results how a clear speed up in both, scalar and vector variants.
Current values with 128 bits preferred size:
With this PR: