search

rapidsai / wholegraph

WholeGraph - large scale Graph Neural Networks
https://docs.rapids.ai/api/cugraph/stable/wholegraph/
Apache License 2.0
100 stars 38 forks source link

Add a new memory type: Hierarchy #227

Closed zhuofan1123 closed 1 month ago

zhuofan1123 commented 1 month ago

This PR adds a new memory type called Hierarchy, which exploits power-law degree distribution in graph to reduce inter-node communication.

  1. Hierarchy WholeMemory has the same storage pattern as Distributed WholeMemory and only optimizes the gather function.
  2. Hierarchy WholeMemory can achieve 1.5x-2.0x speedup in multi-node gather, compared with Distributed.
  3. For intra-node host memory location, Hierarchy can still achieve 1.5x-2.0x speedup.
copy-pr-bot[bot] commented 1 month ago

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

BradReesWork commented 1 month ago

/okay to test

BradReesWork commented 1 month ago

/okay to test

BradReesWork commented 1 month ago

/merge