中国科学院机构知识库网格
Chinese Academy of Sciences Institutional Repositories Grid
首页 机构 成果 学者
  
  1. CAS IR Grid
  2. 计算技术研究所
  3. 中国科学院计算技术研究所
  4. 中国科学院计算技术研究所期刊论文
  5. 英文
SqueezeFlow: A Sparse CNN Accelerator Exploiting Concise Convolution Rules

文献类型：期刊论文

作者Li, Jiajun1,2; Jiang, Shuhao1,2; Gong, Shijun1,2; Wu, Jingya1,2; Yan, Junchao1,2; Yan, Guihai1,2; Li, Xiaowei1,2
刊名IEEE TRANSACTIONS ON COMPUTERS
出版日期2019-11-01
卷号68期号:11页码:1663-1677
关键词Convolutional neural networks accelerator architecture hardware acceleration
ISSN号0018-9340
DOI10.1109/TC.2019.2924215
英文摘要Convolutional Neural Networks (CNNs) have been widely used in machine learning tasks. While delivering state-of-the-art accuracy, CNNs are known as both compute- and memory-intensive. This paper presents the SqueezeFlow accelerator architecture that exploits sparsity of CNN models for increased efficiency. Unlike prior accelerators that trade complexity for flexibility, SqueezeFlow exploits concise convolution rules to benefit from the reduction of computation and memory accesses as well as the acceleration of existing dense architectures without intrusive PE modifications. Specifically, SqueezeFlow employs a PT-OS-sparse dataflow that removes the ineffective computations while maintaining the regularity of CNN computations. We present a full design down to the layout at 65 nm, with an area of 4.80mm2 and power of 536.09mW. The experiments show that SqueezeFlow achieves a speedup of 2:9 on VGG16 compared to the dense architectures, with an area and power overhead of only 8.8 and 15.3 percent, respectively. On three representative sparse CNNs, SqueezeFlow improves the performance and energy efficiency by 1:8 and 1:5 over the state-of-the-art sparse accelerators.
资助项目National Natural Science Foundation of China[61872336] ; National Natural Science Foundation of China[61532017] ; National Natural Science Foundation of China[61572470] ; National Natural Science Foundation of China[61432017] ; National Natural Science Foundation of China[61521092] ; National Natural Science Foundation of China[61376043] ; Youth Innovation Promotion Association, CAS[Y404441000]
WOS研究方向Computer Science ; Engineering
语种英语
WOS记录号WOS:000491426600009
出版者IEEE COMPUTER SOC
源URL[http://119.78.100.204/handle/2XEOYT63/14910]  
专题中国科学院计算技术研究所期刊论文_英文
通讯作者Yan, Guihai; Li, Xiaowei
作者单位1.Chinese Acad Sci, Inst Comp Technol, State Key Lab Comp Architecture, Beijing 100190, Peoples R China
2.Univ Chinese Acad Sci, Beijing 100190, Peoples R China
推荐引用方式
GB/T 7714		 Li, Jiajun,Jiang, Shuhao,Gong, Shijun,et al. SqueezeFlow: A Sparse CNN Accelerator Exploiting Concise Convolution Rules[J]. IEEE TRANSACTIONS ON COMPUTERS,2019,68(11):1663-1677.
APA Li, Jiajun.,Jiang, Shuhao.,Gong, Shijun.,Wu, Jingya.,Yan, Junchao.,...&Li, Xiaowei.(2019).SqueezeFlow: A Sparse CNN Accelerator Exploiting Concise Convolution Rules.IEEE TRANSACTIONS ON COMPUTERS,68(11),1663-1677.
MLA Li, Jiajun,et al."SqueezeFlow: A Sparse CNN Accelerator Exploiting Concise Convolution Rules".IEEE TRANSACTIONS ON COMPUTERS 68.11(2019):1663-1677.

入库方式： OAI收割

来源：计算技术研究所

浏览0
下载0
收藏0
其他版本

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。