Performance Issues for Frontal Schemes on a Cache-Based High Performance Computer

摘要

The authors consider the implementation of a frontal code for the solution oflarge sparse unsymmetric linear systems on a high performance computer where data must be in the cache before arithmetic operations can be performed on it. In particular, the authors show how they can modify the frontal solution algorithm to enhance the proportion of data in the cache. The authors illustrate the effects of this on Silicon Graphics Power Challenge machines using problems which arise in real engineering and industrial applications.