基于GPU与CPU协作的实时波束形成实现方法

摘要

This paper proposed a new method to implement the real-time broadband beamforming with CPU and GPU based on CUDA. Its processing speed can improve one or two magnitude compared with MATLAB and CPU. And it has many advantages compared with multi-DSP platform having the same processing speed, such as shorter development period, lower cost,higher reliability, smaller amount of work and so on.%采用基于CUDA(compute Unified device architecture,统一计算设备架构)的CPU(grapluc processing unit,图形处理器)与CPU协作处理方法,实现了宽带波束形成的实时处理.本方法的处理速度相较于MATLAB和CPU平台可以提高一至两个数量级,相较于同等处理速度的多DSP平台则体现了开发周期短、费用低、工作量小和可靠性高等众多优势.