基于CUDA-GPU的宽带高速频谱分析系统的研究

摘要

This paper presents the design and implementation of a high-speed wideband Spectrometer system based on the compute unified device architecture ( CUDA) and graphics processing unit ( GPU) for radio signal observation. This system executes the Cooley-Tukey FFT algorithm and the spectrum analysis algorithm to achieve high-speed a-nalysis of a wideband spectrum. This was accomplished by the algorithm design in CUDA, with an optimized combination of Multi-Memory access, parallel and pipeline processing to reduce the redundant calculation. In particular , this system was designed for the HRMSFF project ( high resolution microwave survey for the radio telescope FAST) , and it also can be utilized on other radio signal processing and related computationally intensive calculation.%针对射电信号观测的需求,设计并实现了一种基于统一计算设备架构(CUDA)和图形显示卡(GPU)的宽带高速频谱分析系统.该系统通过运用库利-图基(Cooley-Tukey)快速傅立叶变换算法与谱分析算法实现实时宽带高速频谱分析.系统的关键部分在于通过CUDA来完成运用线程合并算法对数据在CPU与GPU之间传递时的转换,并使用并行流水算法在总线中共享多核GPU来降低实时运算时间.该系统主要为500m口径射电望远镜工程的高分辨率微波巡视项目而设计,实测中满足目标需求,并可应用在射电信号观测或类似的高速密集数据运算中.