Skip to main navigation Skip to search Skip to main content

Research on parallel architecture design of radar real-time signal processing based on CPU-GPU heterogeneous platform

  • Henan Dong
  • , Cheng Zheng
  • , Weiming Tian*
  • *Corresponding author for this work
  • Beijing Institute of Technology

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

Abstract

in view of the central processing unit (CPU) platform processing performance limits on computationally intensive tasks, as well as the FPGA + DSP[1-3] platform of software and hardware of high coupling and difficult defects, using the graphics processor (GPU) powerful parallel computing power, the CPU and GPU pipeline technology, heterogeneous systems with multithreading, GPU more and more flow, such as strategy, puts forward the real-time signal processing system of the multiple perspectives analysis methods, make full use of the independence between modules and signal processing algorithm concurrency features within the GPU.The experimental results show that, compared with the traditional MATLAB platform, the maximum acceleration ratio based on CPU-GPU heterogeneous platform can reach more than 20000, and the parallel architecture designed can fully meet the real-time requirements of radar signal processing.

Original languageEnglish
Title of host publicationIET Conference Proceedings
PublisherInstitution of Engineering and Technology
Pages323-329
Number of pages7
Volume2020
Edition9
ISBN (Electronic)9781839535406
DOIs
Publication statusPublished - 2020
Event5th IET International Radar Conference, IET IRC 2020 - Virtual, Online
Duration: 4 Nov 20206 Nov 2020

Conference

Conference5th IET International Radar Conference, IET IRC 2020
CityVirtual, Online
Period4/11/206/11/20

Keywords

  • GRAPHICS PROCESSING UNIT (GPU)
  • REAL-TIME SIGNAL PROCESSING
  • UNIVERSAL PARALLEL COMPUTING ARCHITECTURE (CUDA)

Fingerprint

Dive into the research topics of 'Research on parallel architecture design of radar real-time signal processing based on CPU-GPU heterogeneous platform'. Together they form a unique fingerprint.

Cite this