[QST] single/multi-cta support for topk larger than 1024
Open
Nobody has claimed this yet.
question
- Dominant language
- Cuda
- Stars
- 854
- Forks
- 236
- Avg merge
- 3d 3h
- Merged PRs (30d)
- 62
Description
What is your question?
We benchmark the performance of cuvs for large topk(k > 1024). We find that the speed is much slower compared to the cases for k < 1024. We find that the search algorithm used for k > 1024 is MULTI_KERNEL instead of single/multi-cta.
We have three questions.
- Will you please explain why single/multi-cta algorithms cannnot support large k?
- How can we improve the performance for k > 1024?
- Do you have a plan to support k > 1024 for single/multi-cta algorithms in the near future?
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
Start by reproducing the reported benchmark comparison for k below and above 1024, then inspect where the MULTI_KERNEL path is selected instead of single/multi-cta. Done would require explaining the algorithm limit, identifying a supported performance path for larger k, and clarifying whether single/multi-cta support is planned.
Written by the indexing model from the issue text.
Assessment
- Domain
- performance, search
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100