基于粒子群的HLS的自动化架构实现  

AUTOMATED ARCHITECTURE IMPLEMENTATION OF HLS BASED ON PARTICLE SWARM OPTIMIZATION

作　　者：吴家飞 黄晞[1] 施文灶[1] Wu Jiafei;Huang Xi;Shi Wenzao(Fujian Provinial Engineering Technology Research Center of Photoelectric Sensing Application,Fujian Normal University,Fuzhou 350007,Fujian,China)

机构地区：[1]福建师范大学福建省光电传感应用工程技术研究中心,福建福州350007

出　　处：《计算机应用与软件》2018年第9期22-26,37,共6页Computer Applications and Software

基　　金：国家自然科学基金项目(41701491);福建省自然科学基金项目(2017J01464)

摘　　要：随着数据挖掘、深度学习等领域的火热的发展,在嵌入式设备实现深度学习等计算量较大的算法已经成为一种趋势。由于传统CPU和GPU平台已经难以保证算法实时性的情况下,利用FPGA高性能低功耗的特点对复杂算法实现硬件加速,有着较好的优势。利用Xilinx公司的HLS工具,基于粒子群算法设计一套硬件加速方法。该方法可实现算法优化方案的自动化寻找,极大地提升了设计效率。实验结果表明,该优化方法在寻找一般算法的较优方案上具有一定程度上的通用性。With the hot development of data mining, deep learning and other fields, it has become a trend to realize deep learning in embedded devices. Because the traditional CPU and GPU plafform can＇t guarantee the real-time pefformance of the algorithm, it has a good advantage to realize the hardware acceleration of complex algorithm by using the characteristics of FPGA high performance and low power consumption. Using Xilinx HLS tools, a hardware acceleration method was designed based on particle swarm optimization. This method can automatically search the algorithm optimization scheme and greatly improve the design efficiency. The experimental results show that the optimization method has a certain degree of universality in searching for the better scheme of the general algorithm.

关 键 词：FPGA 模糊离散粒子群算法 高层次综合(HLS) 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]