基于GPU的连续卷积算法

Research on Continual Convolution Algorithm Based on GPU

介绍了目前最新的图形处理器(GPU)编程模型,以数字信号处理中最常用的卷积计算为例,分析了常规卷积算法的计算量和快速卷积算法的使用局限性,并在此基础上提出了基于GPU的分段卷积算法实现,通过与当前主流CPU平台进行实测对比,通过性能对比分析,探讨GPU编程技术应用在数字信号处理领域中的优势,及需要注意的主要问题。

This paper introduces the newest Programming Model based on graphics processing units （GPU）, and analyzes the arithmetic operations of the general algorithm for convolution and the limitation of the algorithm for fast convolution in digital signal processing. A GPU-based algorithm of partition convolution is proposed, and the performance comparison between the GPU-based platform and the CPU-based platform is described. According to the performance comparison results, the application advantages of GPU programming technology in the digital signal processing field and the problem needed to be noticed are discussed.