电子学报2023，Vol.51Issue(11)：3354-3364,11.DOI:10.12263/DZXB.20230281
嵌入式异构智能计算系统并行多流水线设计
Parallel Multi Pipeline Design of Embedded Heterogeneous AI Computing Systems
摘要
Abstract
Due to the limited power consumption and the need for real-time intelligent processing of multiple sensors,embedded AI computing systems desire for higher energy efficiency and more parallel intelligent computing services simul-taneously.The digital signal processing architecture DSP+FPGA commonly used in traditional embedded computing sys-tems is not suitable for multiple ANN models inference acceleration.Based on embedded heterogeneous intelligent comput-ing architecture ARM+DLP+SRIO,this paper proposes a parallel multi pipeline design method by taking advantage of the characteristics of multi chip,multi-core and multi memory channels of deep learning processors.Considering the time cost of data transmission,copy,reference and feedback,this method allocates intelligent computing resources for different neu-ral network models to achieve the maximum end-to-end throughput.The experimental results show that the utilization of the deep learning processor using the parallel multi pipeline design method is about 25.2%higher than that of a single pipeline,and about 30.7%higher than that without pipeline.It meets the real-time intelligent processing requirements of visible light,infrared and SAR images,and is valuable for practical applications.关键词
嵌入式智能计算系统/异构计算架构/神经网络模型/并行多流水线/深度学习处理器Key words
embedded AI computing systems/heterogeneous computing architecture/neural network model/paral-lel multi pipeline/deep learning processor(DLP)分类
信息技术与安全科学引用本文复制引用
赵二虎,吴济文,肖思莹,晋振杰,徐勇军..嵌入式异构智能计算系统并行多流水线设计[J].电子学报,2023,51(11):3354-3364,11.基金项目
中国科学院技术支撑人才项目 （）
北航杭州创新研究院钱江实验室开放基金(No.2020-Y8-A-023)CAS Technology Support Talent Program （No.2020-Y8-A-023）
Hangzhou Innovation Institute,Beihang University(No.2020-Y8-A-023) （No.2020-Y8-A-023）
