A software technique to enhance register utilization of Convolutional Neural Networks on GPGPUs

Che Huai Lin, An Ting Cheng, Bo-Cheng Lai

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

2 Scopus citations

Abstract

CNNs (Convolutional Neural Networks) have demonstrated superior results in a wide range of applications. However, the time-consuming convolution operations required by CNNs pose great challenges to designers. GPGPUs (General Purpose Graphic Processing Units) have been widely used to exploiting the massive parallelism of convolution operations. This paper proposes a software-based loop-unrolling technique to enhance the data usage on the registers and significantly improve the overall performance. The experimental results on a cycle-Accurate GPGPU simulator have shown that the proposed technique can achieve up to 71% performance enhancement when compared with the reference design.

Original languageEnglish
Title of host publicationProceedings of the 2017 IEEE International Conference on Applied System Innovation
Subtitle of host publicationApplied System Innovation for Modern Technology, ICASI 2017
EditorsTeen-Hang Meen, Artde Donald Kin-Tak Lam, Stephen D. Prior
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages614-617
Number of pages4
ISBN (Electronic)9781509048977
DOIs
StatePublished - 21 Jul 2017
Event2017 IEEE International Conference on Applied System Innovation, ICASI 2017 - Sapporo, Japan
Duration: 13 May 201717 May 2017

Publication series

NameProceedings of the 2017 IEEE International Conference on Applied System Innovation: Applied System Innovation for Modern Technology, ICASI 2017

Conference

Conference2017 IEEE International Conference on Applied System Innovation, ICASI 2017
Country/TerritoryJapan
CitySapporo
Period13/05/1717/05/17

Keywords

  • CNN
  • Design and optimization
  • GPU

Fingerprint

Dive into the research topics of 'A software technique to enhance register utilization of Convolutional Neural Networks on GPGPUs'. Together they form a unique fingerprint.

Cite this