A Software Environment for High Performance Scientific Computing Applications

Efficient numerical algorithms play an increasingly important role in computational sciences to make large scale computer simulations tractable. Solution of very large sparse matrices has been one of the most time-consuming parts of many large scale high performance computer simulation problems. One of the important tasks in high performance scientific computing is to identify which solver is suitable for what class of applications (sparse matrices), and which sparse matrix can be solved by what solver. We will use the techniques and ideas in knowledge discovery and data mining to extract useful information and special features from unstructured sparse matrices and to design appropriate strategies to match sparse matrices and solvers. The outcome of this exploratory study is some important preliminary data and database to demonstrate the feasibility of building a software environment for high performance scientific computing applications based on mining sparse matrices and extracting features.