Open Access
- 9 May 2023
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 52-62
Abstract
Graph neural networks (GNNs) are powerful tools for exploring and learning from graph structures and features. As such, achieving high-performance execution for GNNs becomes crucially important. Prior works have proposed to explore the sparsity (i.e., low density) in the input graph to accelerate GNNs, which uses the full-graph-level or block-level sparsity format. We show that they fail to balance the sparsity benefit and kernel execution efficiency. In this paper, we propose a novel system, referred to as AdaptGear, that addresses the challenge of optimizing GNNs performance by leveraging kernels tailored to the density characteristics at the subgraph level. Meanwhile, we also propose a method that dynamically chooses the optimal set of kernels for a given input graph. Our evaluation shows that AdaptGear can achieve a significant performance improvement, up to 6.49× (1.87× on average), over the state-of-the-art works on two mainstream NVIDIA GPUs across various datasets.Keywords
Funding Information
- The National Natural Science Foundation of China (62222210, U21B2017, and 62072297)
- The National Key R&D Program of China (2021ZD0110104)
This publication has 39 references indexed in Scilit:
- Dissecting GPU Memory Hierarchy Through MicrobenchmarkingIEEE Transactions on Parallel and Distributed Systems, 2016
- Gunrock2016
- CuSha2014
- X-Stream2013
- Spectral methods for community detection and graph partitioningPhysical Review E, 2013
- Mizan2013
- Mechanism Design for Stochastic Virtual Resource Allocation in Non-cooperative Cloud Systems2011
- Balanced graph partitioning2004
- R-MAT: A Recursive Model for Graph Mining2004
- CiteSeer1998