论文信息 - Numerical Evaluation of Linear Recurrences on High Performance Computers and Clusters of Workstations

Numerical Evaluation of Linear Recurrences on High Performance Computers and Clusters of Workstations

The aim of this paper is to show that the recently developed algorithm for solving linear recurrence systems on shared-memory parallel computers can be efficiently implemented on high performance message passing parallel and vector computers and clusters of workstations. We analyze the performance of the algorithm using two well known models: BSP and Hockney-Jesshope. Finally, we present the results of experiments performed of a Cray X1 and cluster of 15 personal computers running under Linux.

Przemyslaw Stpiczynski

[1] Roberto Barrio,et al. On the numerical evaluation of linear recurrences , 2003 .

[2] Przemyslaw Stpiczynski,et al. Fast parallel algorithms for computing trigonometric sums , 2002, Proceedings. International Conference on Parallel Computing in Electrical Engineering.

[3] D. Sorensen,et al. LAPACK Working Note No. 2: Block reduction of matrices to condensed forms for eigenvalue computations , 1987 .

[4] Przemyslaw Stpiczynski,et al. Fast Parallel Algorithm for Polynomial Evaluation , 2003, Parallel Algorithms Appl..

[5] Almerico Murli,et al. Algorithm 682: Talbot's method of the Laplace inversion problems , 1990, TOMS.

[6] Przemyslaw Stpiczynski,et al. Solving Linear Recurrence Systems Using Level 2 and 3 BLAS Routines , 2003, PPAM.

[7] David B. Skillicorn,et al. Portability of performance with the BSPLib communications library , 1997, Proceedings. Third Working Conference on Massively Parallel Programming Models (Cat. No.97TB100228).

[8] Przemyslaw Stpiczynski,et al. A New Message Passing Algorithm for Solving Linear Recurrence Systems , 2001, PPAM.