传输触发架构的可编程神经网络处理器设计

doi:10.3969/j.issn.1001-2400.2018.04.017

摘要/Abstract

摘要：

卷积神经网络算法存在着算法结构多样和数据交换计算量大的问题．为此，提出了一种基于传输触发体系架构的可编程卷积神经网络处理器．系统采用多通道直接存储器访问通道、多端口存储器和专用池化数据通道组成数据传输网络解决了数据交换问题．实验表明，该系统在实现卷积神经网络的加速计算方面，虽然吞吐速率比并行流水线方案慢11%，但与之相比具备可编程、适应不同神经网络的特性，节省了46.5%硬件乘法器资源，比其他非流水线实现方案吞吐速率至少快40%．该方案具有系统并行度大、可编程、可在线配置和处理速度较高的特点．

关键词: 深度学习, 卷积神经网络, 并行处理, 现场可编程门阵列

Abstract:

The convolutional neural networks have the problems of structure diversity and large amounts of data exchange and computation. A transport triggered architecture based convolutional neural network processor is presented in this paper. The data transport network is constructed with multi-channel direct memory access channels, the multi-port memory and the specialized pooling data path, which solves the inefficient data exchange problem. Experimental results show that, although the proposed architecture is 11% slower than the streamline structure, it can adapt to a variety of convolutional neural networks and save 46.5% multipliers. Compared with the schemes presented in other papers except pipeline implementation, our design improves the data throughput rate by 40% at least. Besides, this system has advantages of parallel efficiency, programmable flexibility, online architecture reconfiguration, high processing speed, etc.

Key words: deep learning, convolutional neural networks, parallel computing, field programmable gate array

赵博然;张犁;石光明;黄蓉;徐欣冉. 传输触发架构的可编程神经网络处理器设计[J]. 西安电子科技大学学报, 2018, 45(4): 92-98.

ZHAO Boran;ZHANG Li;SHI Guangming;HUANG Rong;XU Xinran. Design of the programmable neural network processor based on the transport triggered architecture[J]. Journal of Xidian University, 2018, 45(4): 92-98.

参考文献

［1］ LECUN Y, BOTTOU L, BENGIO Y, et al. Gradient-based Learning Applied to Document Recognition［J］. Proceedings of the IEEE, 1998, 86(11): 2278-2324.
［2］ KRIZHEVSKY A, SUTSKEVER I, HINTON G E. ImageNet Classification with Deep Convolutional Neural Networks［C］//Proceedings of the Advances in Neural Information Processing Systems. Vancouver, Canada: Neural Information Processing System Foundation, 2012: 1097-1105.
［3］ SIMONYAN K, ZISSERMAN A. Very Deep Convolutional Networks for Large-scale Image Recognition［J］. Computer Science, 2014, 41(5): 1409-1556.
［4］ SUNG W, PARK J. Architecture Exploration of a Programmable Neural Network Processor for Embedded Systems［C］//Proceedings of the 2016 16th International Conference on Embedded Computer Systems: Architectures, Modeling and Simulation. Piscataway: IEEE, 2017: 124-131.
［5］ SUDA N, CHANDRA V, DASIKA G, et al. Throughput-optimized OpenCL-based FPGA Accelerator for Large-scale Convolutional Neural Networks［C］//Proceedings of the 2016 ACM/SIGDA International Symposium on Field-programmable Gate Arrays. New York: ACM, 2016: 16-25.
［6］ QIU J, WANG J, YAO S, et al. Going Deeper with Embedded FPGA Platform for Convolutional Neural Network［C］//Proceedings of the 2016 ACM/SIGDA International Symposium on Field-programmable Gate Arrays. New York: ACM, 2016: 26-35.
［7］ LIU Z, DOU Y, JIANG J, et al. Automatic Code Generation of Convolutional Neural Networks in FPGA Implementation［C］//Proceedings of the 2016 International Conference on Field-programmable Technology. Piscataway: IEEE, 2017: 61-68.
［8］ ZHANG C, LI P, SUN G, et al. Optimizing FPGA-based Accelerator Design for Deep Convolutional Neural Networks［C］//Proceedings of the 2015 ACM/SIGDA International Symposium on Field-programmable Gate Arrays. New York: ACM, 2015: 161-170.
［9］ GOKHALE V, JIN J, DUNDAR A, et al. A 240 G-ops/s Mobile Coprocessor for Deep Neural Networks［C］//Proceedings of the 2014 IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops. Washington: IEEE Computer Society, 2014: 696-701.
［10］ CHEN T, DU Z, SUN N, et al. DianNao: a Small-footprint High-throughput Accelerator for Ubiquitous Machine-learning［C］//Proceedings of the 2014 International Conference on Architectural Support for Programming Languages and Operating Systems. New York: ACM, 2014: 269-284.
［11］ PEEMEN M, SETIO A A A, MESMAN B, et al. Memory-centric Accelerator Design for Convolutional Neural Networks［C］//Proceedings of the 2013 31st IEEE International Conference on Computer Design. Piscataway: IEEE, 2013: 13-19.
［12］ CHAKRADHAR S, SANKARADAS M, JAKKULA V, et al. A Dynamically Configurable Coprocessor for Convolutional Neural Networks［C］//Proceedings of the 2010 International Symposium on Computer Architecture. Piscataway: IEEE, 2010: 247-257.
［13］ CADAMBI S, MAJUMDAR A, BECCHI M, et al. A Programmable Parallel Accelerator for Learning and Classification［C］//Proceedings of the 2017 International Conference on Parallel Architectures and Compilation Techniques. Piscataway: IEEE, 2017: 273-284.
［14］ FARABET C, POULET C, HAN J Y, et al. CNP: an FPGA-based Processor for Convolutional Networks［C］//Proceedings of the 2009 19th International Conference on Field Programmable Logic and Applications. Piscataway: IEEE, 2009: 32-37.
［15］ SANKARADAS M, JAKKULA V, CADAMBI S, et al. A Massively Parallel Coprocessor for Convolutional Neural Networks［C］//Proceedings of the 2009 International Conference on Application-Specific Systems, Architectures and Processors. Piscataway: IEEE, 2009: 53-60.
［16］ ZHANG C, WU D, SUN J, et al. Energy-efficient CNN Implementation on a Deeply Pipelined FPGA Cluster［C］//Proceedings of the 2016 International Symposium on Low Power Electronics and Design. New York: ACM, 2016:326-331.
［17］ KEUTZER K, MALIK S, NEWTON A R. From ASIC to ASIP: the Next Design Discontinuity［C］//Proceedings of the 2002 IEEE International Conference on Computer Design: VLSI in Computers and Processors: 2002. Piscataway: IEEE, 2002: 84-90.
［18］ CORPORAAL H. Microprocessor Architectures-from VLIW to TTA［M］. Hoboken: Wiley, 1997.
［19］ MORENO J H, MOUDGILL M, EBCIOGLU K, et al. Architecture, Compiler and Simulation of a Tree-based VLIW Processor: IBM Research Report RC20495 ［R］. New York: IBM, 1996.
［20］刘镇弢, 李涛, 黄虎才, 等. 一种用于实时图像处理的众核结构设计［J］. 西安电子科技大学学报, 2015, 42(2): 95-101.
LIU Zhentao, LI Tao, HUANG Hucai, et al. Novel Many-core Architecture Design for Real-time Image Processing ［J］. Journal of Xidian University, 2015, 42(2): 95-101.
［21］邓军勇, 李涛, 蒋林等. 面向 OpenGL 的图形加速器设计与实现［J］. 西安电子科技大学学报, 2015, 42(6): 124-130.
DENG Junyong, LI Tao, JIANG Lin, et al. Design and Implementation of the Graphics Accelerator Oriented to OpenGL ［J］. Journal of Xidian University, 2015, 42(6): 124-130.

[1]	李源,崔玉爽,王伟. 一种基于字词双通道网络的文本情感分析方法[J]. 西安电子科技大学学报, 2021, 48(6): 179-186.
[2]	刘佳玮,张文辉,寇晓丽,李雁妮. 增强型深度对抗样本攻击防御算法[J]. 西安电子科技大学学报, 2021, 48(6): 23-31.
[3]	于浩洋,尹良,李书芳,吕顺. 生成对抗网络小样本雷达调制信号识别算法[J]. 西安电子科技大学学报, 2021, 48(6): 96-104.
[4]	董如婵,焦李成,赵进,沈维燕. 一种深度融合机制的遥感图像目标检测技术[J]. 西安电子科技大学学报, 2021, 48(5): 128-138.
[5]	李鹏,冯存前,许旭光,唐子翔. 一种利用贝叶斯优化的弹道目标微动分类网络[J]. 西安电子科技大学学报, 2021, 48(5): 139-148.
[6]	孙彦景,魏力,张年龙,云霄,董锴文,葛敏,程小舟,侯晓峰. 联合DD-GAN和全局特征的井下人员重识别方法[J]. 西安电子科技大学学报, 2021, 48(5): 201-211.
[7]	闫佳,曹玉东,任佳兴,陈冬昊,李晓会. 深度非对称压缩型哈希算法[J]. 西安电子科技大学学报, 2021, 48(5): 212-221.
[8]	宁阳,杜建超,韩硕,杨传凯. 改进DeeplabV3+的火焰分割与火情分析方法[J]. 西安电子科技大学学报, 2021, 48(5): 38-46.
[9]	周鹏,杨军. 采用神经网络架构搜索的遥感影像分割方法[J]. 西安电子科技大学学报, 2021, 48(5): 47-57.
[10]	戚艳军,孔月萍,王佳婧,朱旭东. 一种LSTM与CNN相结合的步态识别方法[J]. 西安电子科技大学学报, 2021, 48(5): 78-85.
[11]	杨云航,闵连权. 采用空洞卷积的多尺度融合草图识别模型[J]. 西安电子科技大学学报, 2021, 48(5): 92-99.
[12]	陈昌川,王海宁,黄炼,黄涛,李连杰,黄向康,代少升. 一种基于局部表征的面部表情识别算法[J]. 西安电子科技大学学报, 2021, 48(5): 100-109.
[13]	宋建锋,苗启广,王崇晓,徐浩,杨瑾. 注意力机制的多尺度单目标跟踪算法[J]. 西安电子科技大学学报, 2021, 48(5): 110-116.
[14]	张宇浩,程培涛,张书豪,王秀美. 一种自适应权重学习的轻量超分辨率重建网络[J]. 西安电子科技大学学报, 2021, 48(5): 15-22.
[15]	宋剑桥,王峰,牛锦,师泽洲,马军辉. 一种面向时空神经网络的潜在情绪识别方法[J]. 西安电子科技大学学报, 2021, 48(4): 159-167.