论文部分内容阅读
本文作者用自行研制的一台使用 4片TMS32 0C4 0的、具有特殊四面体结构的数字信号高速并行处理机 ,实现了MUSIC算法的高速并行计算。在算法实现中 ,认真分析了算法中对性能影响很大的实对称矩阵特征值提取算法的并行实现问题。本文讨论了实对称矩阵特征值提取的串行算法和一种并行算法的特点 ,结合二者的特点 ,提出了一种新的高效的并行算法 ,并通过实验证明文中提出的实对称矩阵特征值提取的并行算法确实有效 ,为MUSIC算法的并行计算取得较好的结果打下基础
The author uses a self-developed digital signal high-speed parallel processor with four TMS32 0C40s and a special tetrahedron structure to realize the high-speed parallel computation of the MUSIC algorithm. In the implementation of the algorithm, the parallel implementation of the eigenvalue extraction algorithm for real symmetric matrices, which greatly affects the performance of the algorithm, is carefully analyzed. This paper discusses the serial algorithm of eigenvalue extraction of real symmetric matrix and the characteristics of a parallel algorithm. Combining the characteristics of the two, a new efficient parallel algorithm is proposed. The eigenvalues of real symmetric matrix The parallel algorithm extracted is indeed valid, which paves the way for better results for the parallel computation of the MUSIC algorithm