arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2303.16459cs.ARcs.AIcs.LG
收藏

GNNBuilder:用于通用图神经网络加速器生成、仿真和优化的自动化框架

GNNBuilder: An Automated Framework for Generic Graph Neural Network Accelerator Generation, Simulation, and Optimization

School of Electrical and Computer Engineering, Georgia Institute of Technology(电气与计算机工程学院,佐治亚理工学院);Georgia Tech Research Institute(佐治亚理工研究机构)

Stefan Abi-Karam, Cong Hao

AI总结:

本文提出GNNBuilder,首个自动化、通用、端到端的GNN加速器生成框架,支持任意GNN模型、PyTorch接口、一键式设计,性能模型误差小,加速比CPU 6.33倍、GPU 6.87倍。

摘要:

目前已有大量图神经网络(GNN)加速器被提出。然而,它们高度依赖用户的硬件专业知识,并且通常针对特定GNN模型进行优化,使其难以实际应用。因此,在这项工作中,我们提出了GNNBuilder,这是第一个自动化、通用、端到端的GNN加速器生成框架。它具有四个优势:(1)GNNBuilder可以自动为用户任意定义的广泛GNN模型生成GNN加速器;(2)GNNBuilder采用标准PyTorch编程接口,为算法开发者带来零开销;(3)GNNBuilder支持端到端的代码生成、仿真、加速器优化和硬件部署,实现GNN加速器设计的“一键式”方式;(4)GNNBuilder配备了其生成加速器的精确性能模型,支持快速灵活的设计空间探索(DSE)。在实验中,首先,我们展示了我们的加速器性能模型在延迟预测方面的误差在36%以内,在BRAM数量预测方面的误差在18%以内。其次,我们展示了我们生成的加速器可以比CPU快6.33倍,比GPU快6.87倍。该框架是开源的,代码可在https://github.com/sharc-lab/gnn-builder获取。

英文摘要:

There are plenty of graph neural network (GNN) accelerators being proposed. However, they highly rely on users' hardware expertise and are usually optimized for one specific GNN model, making them challenging for practical use. Therefore, in this work, we propose GNNBuilder, the first automated, generic, end-to-end GNN accelerator generation framework. It features four advantages: (1) GNNBuilder can automatically generate GNN accelerators for a wide range of GNN models arbitrarily defined by users; (2) GNNBuilder takes standard PyTorch programming interface, introducing zero overhead for algorithm developers; (3) GNNBuilder supports end-to-end code generation, simulation, accelerator optimization, and hardware deployment, realizing a push-button fashion for GNN accelerator design; (4) GNNBuilder is equipped with accurate performance models of its generated accelerator, enabling fast and flexible design space exploration (DSE). In the experiments, first, we show that our accelerator performance model has errors within $36\%$ for latency prediction and $18\%$ for BRAM count prediction. Second, we show that our generated accelerators can outperform CPU by $6.33\times$ and GPU by $6.87\times$. This framework is open-source, and the code is available at https://github.com/sharc-lab/gnn-builder.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑