Performance analysis of a BiCGSTAB solver for multiple-marine-propeller simulation with several MPI libraries and platforms

Authors:
Pengfei Liu;Kun Li
Affiliations:
Research Officer, National Research Council Canada, Institute for Marine Dynamics, 1 Kerwin Place, St. John's, NF Canada A1B 3T5;Work Term Student, Department of Electrical and Computer Engineering, Memorial University of Newfoundland, St. John's, NF Canada A1B 3X5
Venue:
High performance scientific and engineering computing
Year:
2004

Citing 3
Cited 0

Programming the Bi-CGSTAB Matrix Solver for HPC and Benchmarking IBM SP3 and Alpha ES40

IPDPS '02 Proceedings of the 16th International Parallel and Distributed Processing Symposium
The Improved BiCG Method for Large and Sparse Linear Systems on Parallel Distributed Memory Architectures

IPDPS '02 Proceedings of the 16th International Parallel and Distributed Processing Symposium
Performance comparison of LAM/MPI, MPICH, and MVICH on a linux cluster connected by a gigabit ethernet network

ALS'00 Proceedings of the 4th annual Linux Showcase & Conference - Volume 4

Quantified Score

Hi-index	0.00

Visualization

Abstract

Panel method, or boundary element method is widely used in marine hydrodynamics computations, especially for lifting flows in which forces are to be obtained for foil or blade sections. For simulation of multiple propellers and foils, a couple of large, dense and asymmetric matrices are to be generated and to be solved a number of times during a numerical iteration process. A parallel BiCGSTAB matrix solver is developed for this engineering application and tested to run under four different MPI software libraries (MPICH, MPI/Pro, WMPI and Compaq Tru64 MPI) and three different operating systems (Windows 2000 Pro, Red Hat Linux 7.2 and Tru64) under a Compaq Proliant sever and a Compaq Alpha ES40 Server. Performance of the matrix solver, for this application case, for each configuration is evaluated in terms of computational time, speed, and price performance ratio for three different matrix sizes: 256MB, 1GB and 4GB.