[FEA] Use `warp_select` inside of faiss_select inside of fusedL2KNN, haversine_knn and RBC
Open
Nobody has claimed this yet.
feature request
- Dominant language
- Cuda
- Stars
- 854
- Forks
- 236
- Avg merge
- 3d 3h
- Merged PRs (30d)
- 62
Description
We should upgrade the fusedL2KNN code, haversine_knn and RBC code to use our warp_select from raft (aka GridSelect from https://dl.acm.org/doi/10.1145/3581784.3607062 ), instead of using the existing faiss_select code.
Our warp_select code has shown itself to be substantially faster than faiss select, especially on smaller numbers of input rows.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Locate fusedL2KNN, haversine_knn, and RBC and trace their existing faiss_select calls. Compare those call sites with raft's warp_select, also called GridSelect in the linked paper, and confirm all three paths use it without changing results. Benchmark smaller input-row cases to check the intended speedup.
Written by the indexing model from the issue text.
Assessment
- Domain
- performance, search
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 38/100