FlowGNN: A Dataflow Architecture for Universal Graph Neural Network Inference via Multi-Queue Streaming
Rishov Sarkar, Stefan Abi-Karam, Yuqi He, Lakshmi Sathidevi, Cong Hao
School of Electrical and Computer Engineering, Georgia Institute of Technology
Overview
This is FlowGNN, a generic dataflow architecture for GNN acceleration which can flexibly support the majority of message-passing GNNs. Read the paper on arXiv.