Characterizing convolutional neural network workloads on a detailed GPU simulator

2017 International SoC Design Conference (ISOCC) Pub Date : 2017-11-01 DOI:10.1109/ISOCC.2017.8368781

Kwanghee Chang, Minsik Kim, Kyungah Kim, W. Ro

引用次数: 1

Abstract

Recent frameworks on convolutional neural networks (CNNs) such as Caffe and MXNet have focused primarily on being compatible with CUDA software and hardware application. However, it was designed for GPU architecture of compute capability 3.0 and above. Therefore, it needs verification of function to perform GPGPU-Sim which is implemented as NVIDIA compute capability devices 2.x. We developed a framework which can make inferencing AlexNet on GPGPU-Sim. We also analyze the execution results of the GPGPU-Sim. The number of lines in one set of the L1 data cache is sensitive to influence performance of AlexNet inference.

查看原文

微信好友朋友圈 QQ好友复制链接

本刊更多论文

在详细的GPU模拟器上表征卷积神经网络工作负载

卷积神经网络(cnn)的最新框架，如Caffe和MXNet，主要关注与CUDA软件和硬件应用的兼容。然而，它是为计算能力3.0及以上的GPU架构设计的。因此，GPGPU-Sim作为NVIDIA计算能力器件2.x实现，需要进行功能验证。我们开发了一个可以在GPGPU-Sim上对AlexNet进行推理的框架。本文还对GPGPU-Sim的执行结果进行了分析。一组L1数据缓存中的行数对AlexNet推理性能的影响非常敏感。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文去求助

来源期刊

2017 International SoC Design Conference (ISOCC)

自引率

0.00%

发文量

期刊最新文献

Memory efficient self guided image filtering A fully-digital phase modulator with phase calibration loop for high data-rate systems Development of SoC virtual platform for IoT terminals based on OneM2M Hardware feasible offset and gain error correction for time-interleaved ADC A design of ultra-low noise LDO using noise reduction network techniques