Node-level optimization by caching data and choosing the optimal tile size for parallel dense linear algebra on distributed systems: student research abstract.