Two Efficient Parallel Algorithms for Solving Symmetric Pentadiagonal Linear Systems of Equations


Abstract in English

In this paper, we described tow parallel algorithms for finding the solution of symmetric pentadiagonal linear systems of equations of order n . The proposed algorithms require 2 processors; each of both possesses      N O n local memory. The first algorithm includes writing the pentadiagonal matrix in the form of product of tow tridiagonal matrices. We suggested a parallel algorithm for solving tridiagonal linear systems of equations. The second algorithm consists of decomposition of the pentadiagonal matrix in a form such that we can carry out the resulting linear systems of equations by using parallel algorithm. We carried out many numerical experiments to illustrate the efficiency, speeding up and accuracy for solving symmetric pentadiagonal linear systems of equations. The numerical experiments showed that the proposed algorithms were efficient and one of both was much faster in factor of 2 than the other one for solving the same test problems.

References used

C.W. Groetsch, J.T. King, Matrix Methods and Applications, Prentice Hall, Englewood Cliffs, NJ, 1988
A. Quarteroni, R. Sacco and F. Saleri, Numerical Mathematics, Springer-Verlag, 2000
Arnt H. Veenstra, H.X. Lin and E.A.H. Vollebregt, A comparison of scability of different parallel iterative methods for shallow water equations, Contemp. Math. 218 (1998) 357–364

Download