CCA-LISI: On Designing A CCA Parallel Sparse Linear Solver Interface

Abstract

Sparse linear solvers account for much of the execution time in many High Performance Computing applications, and not every solver works on all problems. Hence choosing a suitable solver is crucial step for an efficient application. Unfortunately, the best solver cannot be determined during the application development. Experiments on finding best suitable solver require a plug and play mechanism. This work is part of the Common Component Architecture effort on designing common interface among various parallel high performance linear solver libraries,hence componenizing them and enabling dynamical switching. The implementation of this interface provides a CCA toolkit and is reusable under CCA-compliant framework such as Ccaffeine.