-
Notifications
You must be signed in to change notification settings - Fork 1.9k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[META] Explore usage of SIMD in OpenSearch #9473
Comments
@reta That issue I was reading is specifically discussing around KNN improvements and probably also getting lucene optimizations. I wanted to keep this issue to discuss if there is potential to make improvements in OpenSearch codebase itself. |
@reta Earlier based on my search on issue title I didn't came across any other issue for SIMD exploration. My thought was having a parent issue to discuss broadly on SIMD optimization which can keep track of different specific discussions like enabling the lucene optimizations can be sub issues/tracks (hence linked it here). That sub-issue seems like is also tracked for 2.10.0 whereas this is more open ended to figure out what other potential improvements we can do in OpenSearch itself. What do you think ? |
this is already enabled by default on 3.0.0 and we be backported to 2.x once we get JDK-21, would OpenSearch or any other plugin use it - this is different question.
Up to you, #9423 is general ticket which could repurposed as I mentioned (the one for 2.11.0 would be JDK-21 update that would provide Project Panama). |
Closing this as I have repurposed the issue #9423 |
Correcting the reference link for enabling incubating api. #8601 |
Describe the solution you'd like
With the Project Panama there is support for Vector API to utilize the available SSE and AVX instructions sets on hardware for vectorized processing. The API is currently in 6th incubation but we are seeing some on-demand usage of such APIs in Lucene already. I would think similar to lucene there will be potential to utilize these in OpenSearch codebase itself (may be in Aggregations or other places (TBD)). I wanted to start a discussion to see if any community member has explored it or has any thought around it.
Additional context
The text was updated successfully, but these errors were encountered: