基于单DSP实现并行扩维离散Hartley变换算法的研究

On a parallel DHT algorithm on expanded-dimension parallelism based on single DSP

提出了一种将N点的一维离散Hartley变换(简称DHT)分解成N0×N1点的二维DHT(其中N=N0×N1)和一些运算量很小的附加运算的并行扩维DHT算法,此算法通过减少数据相关性的方法突破了DSP高效求解快速离散Hartley变换(简称FHT)时问题规模受片内内存容量限制问题,降低了编程复杂性,并在TMS320C80的单处理单元上进行了该算法实现方法的研究.结果表明,理论分析和试验结果吻合,该算法适合在单DSP上实现.

A parallel DHT algorithm is put forward based on the expanded-dimension parallelism. By the algorithm, one dimension DHT of length N=N0N1 can be decomposed into two dimension DHT with length N0 × N1 with some small extra-operation processes. Based on a processor unit of multiprocessor platform-TMS320C80, the experimental study on the parallel DHT algorithm is carried out. The algorithm reduces programming complexity, data relativity and limitation of processor unit＇s onchip memory capacity. The experiment result is consistent with the theory analysis, which proves that the method can be easily realized on a single DSP.