深度神经网络硬件基准测试现状及发展趋势

信息通信技术与政策 ›› 2019, Vol. 45 ›› Issue (12): 74-78.

深度神经网络硬件基准测试现状及发展趋势

Deep neural network hardware benchmark status and development trend

出版日期:2019-12-15 发布日期:2020-11-19
作者简介:
张蔚敏:中国信息通信研究云计算与大数据研究所工程师

Online:2019-12-15 Published:2020-11-19

摘要/Abstract

摘要： 基准测试作为一种客观的评价方式，在计算机体系架构的发展中扮演着重要的角色，有效地推动着面向不同方向的硬件和软件设计的演进。针对当前深度神经网络处理器的应用场景、发展现状和与通用处理器基准测试差异性进行了分析，给出基准测试参考架构设计，并对深度神经网络处理器基准测试的重要性与所面临的挑战进行了分析。

关键词: 深度神经网络, AI芯片, 基准测试

Abstract: As an objective evaluation method, benchmarking plays an important role in the development of computer architecture,and effectively promoting the evolution of hardware and software design for different directions. In this paper, the application scenarios, development status and differences between general processor benchmarks of the deep neural network (DNN) processor were analyzed. The reference test architecture design ofDNNprocessor was proposed.And the importance and challenges ofDNNprocessor benchmarking were analyzed.

Key words: deep neural network, AIchip, benchmark

张蔚敏. 深度神经网络硬件基准测试现状及发展趋势[J]. 信息通信技术与政策, 2019, 45(12): 74-78.

ZHANGWeimin. Deep neural network hardware benchmark status and development trend[J]. Information and Communications Technology and Policy, 2019, 45(12): 74-78.

导出引用管理器 EndNote|Ris|BibTeX

链接本文: http://ictp.caict.ac.cn/CN/Y2019/V45/I12/74

[1]	杜子东. 寒武纪:智能处理器和基准测试集[J]. 人工智能,2018(2).
[2]	HanY,Wang X, Leung V, et al. Convergence of edge computing and deep learning:Acomprehensive survey[J]. arXiv Preprint arXiv:1907.08349,2019.
[3]	Ignatov A, Timofte R, ChouW, et al. Ai benchmark: Running deep neural networks on android smartphones[C]//Proceedings of the European ConferenceonComputerVision(ECCV),2018:0-0.
[4]	Verma S,Wu Q, Hanindhito B, et al. Demystifying the MLPerf benchmark suite[J]. arXiv Preprint arXiv:1908.09207,2019.
[5]	Coleman, Cody, et al. Dawnbench:An end-to-end deep learning benchmark and competition.Training100.101 (2017):102.
[6]	Tao J H, Du ZD, Guo Q, et al.Bench IP: Benchmarking intelligence processors[J]. Journal of Computer Science and Technology,2018, 33(1):1-23.
[7]	ZhangW,WeiW,XuL, et al.AI Matrix:Adeep learning benchmark for alibaba data centers[J]. arXiv Preprint arXiv: 1909.10562, 2019.

深度神经网络硬件基准测试现状及发展趋势

Deep neural network hardware benchmark status and development trend

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献 7

相关文章 0

编辑推荐

Metrics

本文评价