当前位置: X-MOL 学术IEEE Trans. Comput. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
EnGN: A High-Throughput and Energy-Efficient Accelerator for Large Graph Neural Networks
IEEE Transactions on Computers ( IF 3.6 ) Pub Date : 2020-08-06 , DOI: 10.1109/tc.2020.3014632
Shengwen Liang , Ying Wang , Cheng Liu , Lei He , Huawei Li , Dawen Xu , Xiaowei Li

Graph neural networks (GNNs) emerge as a powerful approach to process non-euclidean data structures and have been proved powerful in various application domains such as social networks and e-commerce. While such graph data maintained in real-world systems can be extremely large and sparse, thus employing GNNs to deal with them requires substantial computational and memory overhead, which induces considerable energy and resource cost on CPUs and GPUs. In this article, we present a specialized accelerator architecture, EnGN, to enable high-throughput and energy-efficient processing of large-scale GNNs. The proposed EnGN is designed to accelerate the three key stages of GNN propagation, which is abstracted as common computing patterns shared by typical GNNs. To support the key stages simultaneously, we propose the ring-edge-reduce(RER) dataflow that tames the poor locality of sparsely-and-randomly connected vertices, and the RER PE-array to practice RER dataflow. In addition, we utilize a graph tiling strategy to fit large graphs into EnGN and make good use of the hierarchical on-chip buffers through adaptive computation reordering and tile scheduling. Overall, EnGN achieves performance speedup by 1802.9X, 19.75X, and 2.97X and energy efficiency by 1326.35X, 304.43X, and 6.2X on average compared to CPU, GPU, and a state-of-the-art GCN accelerator HyGCN, respectively.

中文翻译:

EnGN:用于大型图神经网络的高吞吐量和高能效加速器

图神经网络 (GNN) 作为处理非欧几里得数据结构的强大方法出现,并已在社交网络和电子商务等各种应用领域中被证明是强大的。虽然在现实世界系统中维护的此类图数据可能非常大且非常稀疏,因此使用 GNN 来处理它们需要大量的计算和内存开销,这会导致 CPU 和 GPU 的大量能源和资源成本。在本文中,我们展示了一种专门的加速器架构 EnGN,以实现大规模 GNN 的高吞吐量和节能处理。所提出的 EnGN 旨在加速 GNN 传播的三个关键阶段,将其抽象为典型 GNN 共享的通用计算模式。同时支持关键阶段,我们提出了环边减少(RER)数据流来驯服稀疏和随机连接顶点的不良局部性,以及 RER PE 阵列来练习 RER 数据流。此外,我们利用图平铺策略将大图拟合到 EnGN 中,并通过自适应计算重新排序和平铺调度充分利用分层的片上缓冲区。总体而言,与 CPU、GPU 和最先进的 GCN 加速器 HyGCN 相比,EnGN 平均实现了 1802.9X、19.75X 和 2.97X 的性能加速和 1326.35X、304.43X 和 6.2X 的能源效率提升,分别。我们利用图平铺策略将大图拟合到 EnGN 中,并通过自适应计算重新排序和平铺调度充分利用分层的片上缓冲区。总体而言,与 CPU、GPU 和最先进的 GCN 加速器 HyGCN 相比,EnGN 平均实现了 1802.9X、19.75X 和 2.97X 的性能加速和 1326.35X、304.43X 和 6.2X 的能效提升,分别。我们利用图平铺策略将大图拟合到 EnGN 中,并通过自适应计算重新排序和平铺调度充分利用分层的片上缓冲区。总体而言,与 CPU、GPU 和最先进的 GCN 加速器 HyGCN 相比,EnGN 平均实现了 1802.9X、19.75X 和 2.97X 的性能加速和 1326.35X、304.43X 和 6.2X 的能源效率提升,分别。
更新日期:2020-08-06
down
wechat
bug