NVIDIA / NVIDIA/cuvs

[BUG] Lucene: route explicit MULTI_KERNEL queries through per-segment search

Open
#2,620 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Cuda
Stars
854
Forks
236
Avg merge
3d 3h
Merged PRs (30d)
62

Description

Problem

GPUKnnFloatVectorQuery.rewrite() can send explicit MULTI_KERNEL searches to cuVS multi-partition search, although that native path rejects MULTI_KERNEL.

Proposed change

Route explicit MULTI_KERNEL through Lucene's per-segment rewrite path so each CAGRA reader executes the supported algorithm. Preserve AUTO, SINGLE_CTA, supported multi-partition behavior, and the selective-filter fallback being developed in #2612.

Add a multi-segment CAGRA-only regression test covering rank-one self matches, exact hit count, unique IDs, and absence of brute-force indexes.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start at GPUKnnFloatVectorQuery.rewrite() and trace how explicit MULTI_KERNEL queries are routed for multi-segment searches. Add the CAGRA-only regression test described in the issue, verifying rank-one self matches, exact hit counts, unique IDs, and no brute-force indexes while preserving the stated AUTO, SINGLE_CTA, multi-partition, and selective-filter behavior.

Written by the indexing model from the issue text.

Assessment

Domain
search
Issue type
Bug
Difficulty
3/5
Estimated time
1-2 days
Activity status
Active
Clarity
Mostly clear
Newbie friendliness
72/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.