摘要
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 .
出处
《杭州电子科技大学学报(自然科学版)》
2013年第6期62-65,共4页
Journal of Hangzhou Dianzi University:Natural Sciences
基金
国家自然科学基金资助项目(61272391)
关键词
图像处理器
快速哈特利变换
并行计算
graphic processing unit
fast Hartley transform
parallel computing