当前位置>主页 > 期刊在线 > 信息技术 >

信息技术22年8期

基于 PYNQ 集群的内存负载分析系统设计
华夏¹,柴志雷 ¹,²,张曦煌¹
(1.江南大学 人工智能与计算机学院,江苏 无锡 214122;2.江苏省模式识别与计算智能工程实验室,江苏 无锡 214122)

摘  要:在分布式计算平台上研究脉冲神经网络(SNN)的工作负载特性时,快速确定 SNN 模型构建所需的内存消耗以及平台的网络承载能力,是提高工作负载研究效率的重要手段。针对该问题,文章搭建了 PYNQ 集群分布式计算平台,设计了集群内存负载分析系统。实验表明:内存负载分析系统在内存消耗的预测方面取得了 97.98% 的平均准确率,在预测集群网络承载能力方面取得了 97.19% 的准确率,通过分析集群承载 SNN 模型时的内存负载,有效提升了集群上的 SNN 工作负载研究效率。


关键词:脉冲神经网络(SNN);分布式计算平台;计算能效;NEST 仿真器



DOI:10.19850/j.cnki.2096-4706.2022.08.001


基金项目:国家自然科学基金资助项目(61972180)


中图分类号:TP302                                      文献标识码:A                             文章编号:2096-4706(2022)08-0001-05


Design of Memory Load Analysis System Based on PYNQ Cluster

HUA Xia 1, CHAI Zhilei 1,2, ZHANG Xihuang1

(1.School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi 214122, China; 2.Jiangsu Provincial Engineering Laboratory of Pattern Recognition and Computational Intelligence, Wuxi 214122, China)

Abstract: When studying the workload characteristics of Spiking Neural Network (SNN) on distributed computing platform, it is an important means to improve the efficiency of workload research to quickly determine the memory consumption required for SNN model construction and the network carrying capacity of the platform. To solve this problem, PYNQ cluster distributed computing platform is built and a cluster memory load analysis system is designed. The experimental results show that the average accuracy of the memory load analysis system is 97.98% in the prediction of memory consumption and 97.19% in the prediction of cluster network carrying capacity. By analyzing the memory load when the cluster carries the SNN model, the research efficiency of SNN workload on cluster is effectively improved.

Keywords: spiking neural network (SNN); distributed computing platform; computing efficiency; NEST simulator


参考文献:

[1] MAASS W. Networks of spiking neurons:the third generation of neural network models [J].Neural networks,1997,10(9):1659-1671.

[2] 张铁林,徐波 . 脉冲神经网络研究现状及展望 [J]. 计算机学报,2021,44(9):1767-1785.

[3] DENG L,WANG G R,LI G Q,et al. Tianjic:A Unified and Scalable Chip Bridging Spike-Based and Continuous Neural Computation [J].IEEE Journal of Solid-State Circuits,2020,55(8):2228-2246.

[4] DAVIES M,SRINIVASA N,LIN T H,et al. Loihi:A Neuromorphic Manycore Processor with On-Chip Learning [J].IEEE Micro,2018,38(1):82-99.

[5] YAKOPCIC C,RAHMAN N,ATAHARY T,et al. Solving constraint satisfaction problems using the loihi spiking neuromorphic processor [C]//2020 Design,Automation & Test in Europe Conference & Exhibition(DATE).Grenoble:IEEE,2020:1079-1084.

[6] KULKARNI S R,PARSA M,MITCHELL J P,et al. Benchmarking the performance of neuromorphic and spiking neural network simulators [J].Neurocomputing,2021,447:145-160.

[7] 刘家华,陈靖宇.多核并行脉冲神经网络模拟器的设计 [J].计算机工程与应用,2020,56(22):244-250.

[8] NGUYEN Q A P,ANDELFINGER P,TAN W J,et al.Transitioning Spiking Neural Network Simulators to Heterogeneous Hardware [J].ACM Transactions on Modeling and Computer Simulation (TOMACS),2021,31(2):1-26.

[9] GEWALTIG M O,DIESMANN M. NEST (neural simulation tool) [J].Scholarpedia,2007,2(4):1430.

[10] pynq.io [EB/OL].(2021-11-22) [2022-03-25].http://discuss. pynq.io/.

[11] MORRISON A,AERTSEN A,DIESMANN M. Spiketiming-dependent plasticity in balanced random networks [J].Neural computation,2007,19(6):1437-1467.


作者简介:华夏(1997—),男,汉族,江苏无锡人,硕士研究生,研究方向:类脑计算、嵌入式系统;柴志雷(1975—),男,汉族,山西新绛人,教授,博士,研究方向:软件定义的高效计算机系统、嵌入式系统、软硬件协同设计等;张曦煌(1962—),男,汉族,江苏无锡人,教授,博士,研究方向:嵌入式系统、分布式系统与应用等。