search

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.
https://docs.rapids.ai/api/raft/stable/
Apache License 2.0
764 stars 194 forks source link

[FEA] integrate block-level HashGraph hash table #626

Open cjnolet opened 2 years ago

cjnolet commented 2 years ago

This should provide a nice optimization for lookups on (sparse) algorithms which rely on hash table lookups at the block-level. The original HashGraph paper outlines an implementation at the device level but it shouldn’t be too hard to make a device function that builds the hash table in shared memory. This can be useful in raft’s sparse pairwise distances, cugraph, and even graphblas.

If we do end up replacing the current linear probed hash table with HashGraph then this will also remove the need for a dependency on cucollections.

Another benefit to having this approach is potentially optimizing shuffling / sampling w/o replacement for non-weighted cases.

cjnolet commented 2 years ago

cc @ogreen and @jeaton32

github-actions[bot] commented 2 years ago

This issue has been labeled inactive-30d due to no recent activity in the past 30 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed. This issue will be labeled inactive-90d if there is no activity in the next 60 days.

github-actions[bot] commented 2 years ago

This issue has been labeled inactive-90d due to no recent activity in the past 90 days. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed.