基于递归耦合方法的三对角线性方程组分布式并行算法

Parallel algorithm based on recursive decoupling method for tridiagonal linear equations

提出了一种在分布式计算机上用递归倍增方法解三对角线性方程组的并行算法。通过研究算法中的额外开销达到优化标量算法的执行和通讯,并减少了存储开销。当三对角线性方程组的系数矩阵满足对角占优时,该算法在运行过程中不会中断。最后,在采用消息传递编程模型的基于局域网MPI并行环境下对算法进行了评价。数值实验结果表明,该算法是高效的。

A parallelization ofa recursive decoupling method for sovling tridiagonal linear system on distributed computer is presented. The extra spending in the algorithm was studied to optimize the execution and communication of the algorithm and reduce the cost of memory. If the tridiagonal linear linear equations＇ coefficient matrix is diagonally dominant, the process will not break down. Finally, the algorithm on MPI parallel environment was evaluated based on local net. The experiment results show that the new algorithm is highly efficient.