GATe: Efficient Graph Attention Network Acceleration With Near-Memory Processing

Graph Attention Network (GAT) has gained widespread adoption thanks to its exceptional performance in processing non-Euclidean graphs. The critical components of a GAT model involve aggregation and attention, which cause numerous main-memory access, occupying significant inference time. Recently, much research has proposed near-memory processing (NMP) architectures to accelerate aggregation. However, graph attention requires additional operations distinct from aggregation, making previous NMP architectures less suitable for supporting GAT, as they typically target aggregation-only workloads. In this paper, we propose GATe, a practical and efficient GAT accelerator with NMP architecture. To the best of our knowledge, this is the first time that accelerates both attention and aggregation computation on DIMM. We unify feature vector access to eliminate the two repetitive memory accesses to source nodes caused by the sequential phase-by-phase execution of attention and aggregation. Next, we refine the computation flow to reduce data dependencies in concatenation and softmax, which lowers on-chip memory usage and communication overhead. Additionally, we introduce a novel sharding method that enhances data reusability of high-degree nodes. Experiments show that GATe achieves substantial speedup of GAT attention and aggregation phases up to 6.77${\boldsymbol\times}$mml:mathmml:mrow<mml:mo mathvariant="bold">× and 2.46${\boldsymbol\times}$mml:mathmml:mrow<mml:mo mathvariant="bold">×, with average to 3.69${\boldsymbol\times}$mml:mathmml:mrow<mml:mo mathvariant="bold">× and 2.24${\boldsymbol\times}$mml:mathmml:mrow<mml:mo mathvariant="bold">×, respectively, compared to state-of-the-art NMP works GNNear and GraNDe.