摘要
为提高正交算法的运算速度,提出了一种将N点的一维正交变换分解成N0×N1点的二维正交变换(N=N0N1)和运算量较少的附加运算的并行扩维正交变换算法。在定义正交变换算法扩维并行性的基础上,讨论了离散傅里叶变换(DFT)、Hadamard变换和Hartley变换等算法的实现及性能。在TMS320C80多处理机平台上不同算法实现的试验结果表明:算法可有效减少数据的相关性、降低编程的复杂性,消除了处理单元片内内存容量的限制,适于以数字信号处理器(DSP)为处理单元的多处理机平台的并行实现。
To speed the calculation of orthogonal algorithm,a parallel orthogonal transform algorithm on expanded-dimension parallelism based on that one dimension orthogonal transform of length N=N_0N_1 would be decomposed into two dimension orthogonal transform of length N_0×N_1 with some little additional operations was put forward in this paper.According to the definition of expanded-dimension parallelism for parallel orthogonal transform algorithm,the realization and feature of DFT algorithm,Hadamard transform al...
出处
《上海航天》
北大核心
2008年第1期21-25,共5页
Aerospace Shanghai
关键词
正交算法
并行处理
扩维并行性
并行正交变换算法
TMS320C80芯片
Orthogonal algorithm
Parallel algorithm
Expanded-dimension parallelism
Parallel orthogonal transform algorithm processing
TMS320C80
