基于FPGA的卷积神经网络加速模块设计
梅志伟,王维东

Design of Convolutional Neural Network acceleration module based on FPGA
Zhiwei Mei,Weidong Wang
表6 与其他加速器在VGG?16卷积层加速效果对比
Table 6 The acceleration performance of convolutional layers and other accelerators in VGG?16
VGG?16加速[13][14][15]本文方案
年份2016201820182020
FPGAStratix?V GSD8Zynq XC7Z045Virtex7 VX690TVirtex7 VX485T
时钟120 MHz150 MHz150 MHz100 MHz
数据16 bits16 bits16 bits16 bits
DSP19637802833588(21%)
BRAM25674861248288(28%)
性能(GOPS)136.50187.80488.00189.03

DSP性能效率

(GOPS/DSP)

0.0700.2410.1720.321