目前,类脑计算所面临的最具挑战性的问题之一是如何高性能且低功耗地进行大规模类脑仿真。本文选用应用生态完整、支持大规模仿真的NEST类脑仿真器,针对NEST类脑仿真器可移植性差、仿真速度慢等问题,设计了一种ARM+FPGA的类脑计算平台的通用性系统架构。本设计采用硬件加速神经元计算模块、通用数据传输接口设计、软硬件协同设计等方法提升了NEST类脑仿真器的性能。在3款类脑计算平台上证明了该架构的可行性,为类脑计算平台提供了一种通用解决方案。
Abstract
Currently,one of the most challenging problems among brain-like computing areas is how to perform large-scale brain-like simulations with both higher performance and lower power consumption.In the paper,the NEST brain-like simulator which not only has complete application ecology but also supports for large-scale simulation is employed.Aiming at the problems of poor portability and slow running speed of NEST brain-like simulator,a general system architecture of ARM+FPGA based brain-like computing platform is designed.This design uses hardware-accelerated neuron computing modules,general data transmission interface design,software-hardware co-design and other methods to improve the performance of the brain-like simulator.The feasibility of the architecture is verified on three brain-like computing platforms,which provides a general solution for brain-like computing platforms.
关键词
类脑计算 /
脉冲神经网络 /
软硬件协同设计 /
可编程逻辑门阵列
Key words
brain-like computing /
spiking neural network /
software and hardware co-design /
FPGA
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] Bargmann C I,Newsome W T.The brain research through advancing innovative neurotechnologies(BRAIN) initiative and neurology[J].JAMA neurology,2014,71(6):675-676.
[2] 黄铁军,施路平,唐华锦,等.多媒体技术研究:2015—类脑计算的研究进展与发展趋势[J].中国图象图形学报,2018,21(11):1411-1424.
[3] Debole M V,Taba B,Amir A,et al.TrueNorth: Accelerating from zero to 64 million neurons in 10 years[J].Computer,2019,52(5):20-29.
[4] Fischl K D,Andreou A G,Stewart T C,et al.Implementation of the neural engineering framework on the TrueNorth neurosynaptic system[C]//2018 IEEE Biomedical Circuits and Systems Conference(BioCAS),2018:1-4.
[5] Quang Nguyen,Philipp Andelfinger,Wentong Cai,et al.Transitioning Spiking Neural Network Simulators to Heterogeneous Hardware[C]//Proceedings of the 2019 ACM SIGSIM Conference on Principles of Advanced Discrete Simulation.New York,NY,USA:ACM,2019.
[6] A Podobas,S Matsuoka,Luk Wayne.Designing and accelerating spiking neural networks using OpenCL for FPGAs[C]//International Conference on Field Programmable Technology (ICFPT).Melbourne,VIC, Australia:IEEE,2017.
[7] Ju X,Fang B,Yan R,et al.An FPGA Implementation of Deep Spiking Neural Networks for Low-Power and Fast Classification[J].Neural Computation,2019,32(1):1-23.
[8] Gewaltig M-O,Diesmann M.Nest (neural simulation tool)[J].Scholarpedia,2007,2(4):1430.
基金
*基于工作负载表征的类脑体系结构基准测试模型与自动映射方法研究(61972180)。
PDF(989 KB)
