基于GPU的高速FHT计算及性能分析

GPU-based High-speed FHT Computation and Performance Analysis

针对传统FHT算法在处理海量数据时不能很好的满足实时性需求,该文提出了一种基于CUDA高效的并行FHT算法。通过分析FHT算法的分治特性及CUDA的编程模型,采用了将数据映射到多线程并行运算的方法,实现了对FHT算法的加速和优化。实验结果表明,新的并行算法可以有效地提升FHT处理速度,且随着数据规模的增长,加速效果越明显。

The traditional fast Hartley transform （ FHT） algorithm can not processing large amounts of data to meet real-time requirements .This paper proposed a CUDA-based efficient parallel FHT algorithm .By analyzing the algorithm FHT divide and conquer feature and the CUDA programming model , brought a strategy used multi-threaded to parallel computing by mapped the data to each thread , and explored optimization in memory organization .Experimental results show that the algorithm is efficient , and with the growth in the size of data to accelerate the effect is more obvious .