Chinese Journal of Liquid Crystals and Displays, Volume. 40, Issue 3, 448(2025)

Design of heterogeneous FPGA hardware accelerator based on CNN

Haolin JI1,2,3, Wei XU1,3、*, Yongjie PIAO1,3, Xiaobin WU1,2,3, and Tan GAO1,2,3
Author Affiliations
  • 1Changchun Institute of Optics, Fine Mechanics and Physics, Chinese Academy of Sciences, Changchun 130033, China
  • 2University of Chinese Academy of Sciences, Beijing 100049, China
  • 3Key Laboratory of Space-Based Dynamic & Rapid Optical Imaging Technology, Chinese Academy of Sciences, Changchun 130033,China
  • show less
    Figures & Tables(9)
    Schematic diagram of convolutional layers
    Schematic diagram of CNN accelerator architecture
    SIMD multiplication using common operand C
    SIMD CU structure diagram
    (a) Data flow of convolutional kernels in PE units; (b)Input feature map data flow in PE unit.
    SIMD buffer structure diagram
    Analysis of parallelism in XC7Z020 Device
    Implementation of embedded systems on ZYNQ devices
    • Table 1. Comparison of implementation results of CNN accelerator proposed in this paper with that of other accelerators

      View table
      View in Article

      Table 1. Comparison of implementation results of CNN accelerator proposed in this paper with that of other accelerators

      FPGA平台

      (DSP总数)

      频率/MHzGopsDSPs

      DE/

      (Gops/DSPs)

      CONVs

      Time/ms

      PEff/

      (Gops·W-1

      本文XC7Z020(220)15090.1216(98.1%)0.417382.238.5
      文献[12XC7Z020(220)15084.3190(86.3%)0.44336424.1
      文献[9XC7Z020(220)12548.53220(100%)0.2263327.7
      本文XC7Z045(900)167425.32864(96.0%)0.4891.2102
      文献[10XC7Z045(900)15036.8197(21.8%)0.1861639.379.7
      文献[12XC7Z045(900)214137780(86.6%)0.175224.614.2
      文献[9XC7Z045(900)125155.81855(95.0%)0.182249.538.8
      文献[13XC7Z045(900)150374.98900(100%)0.41682.03NA1
      文献[14XC7Z045(900)140169864(96.0%)0.195181.816.9
      文献[11GX1150(1518)200715.91518(100%)0.4743.2NA
    Tools

    Get Citation

    Copy Citation Text

    Haolin JI, Wei XU, Yongjie PIAO, Xiaobin WU, Tan GAO. Design of heterogeneous FPGA hardware accelerator based on CNN[J]. Chinese Journal of Liquid Crystals and Displays, 2025, 40(3): 448

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Circuit Design

    Received: Jul. 12, 2024

    Accepted: --

    Published Online: Apr. 27, 2025

    The Author Email: Wei XU (xwciomp@163.com)

    DOI:10.37188/CJLCD.2024-0198

    Topics