Giters

rapidsai / raft

RAFT contains fundamental widely-used algorithms and primitives for machine learning and information retrieval. The algorithms are CUDA-accelerated and form building blocks for more easily writing high performance applications.

Home Page:https://docs.rapids.ai/api/raft/stable/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

[QST]How to ensure that the returned neighbor IDs in cagra multi-cta algorithm are non-duplicate?

GatePara opened this issue · comments

commented

How to ensure that the returned neighbor IDs in cagra multi-cta mode are non-duplicate?
The hash table used by the cagra multi-cta algorithm to record distance calculation records is stored in device memory and is shared by multiple CTAs. Is there a mechanism for mutually exclusive access to the hash table by different CTAs in the algorithm to ensure that neighbors will not be repeatedly calculated and then added to the TOP-M queue of different CTAs. Or does the algorithm use other methods to ensure that the neighbor IDs returned in the end are non-duplicate, such as deduplicating the intermediate results before returning the results?
@cjnolet @aschaffer @divyegala