Efficiency results for Gaussian elimination on the iPSC/2 hypercube

摘要

We compare two methods for solving banded linear systems on a hypercube multiprocessor. Both methods are based on Gaussian elimination. The differences in the methods are due to different allocation schemes to distribute the data among the nodes. We implemented both methods on the Intel iPSC/2 hypercube. Timing results and efficiency results obtained on this multiprocessor are discussed.