Performance evaluation of structural analysis solver for many RHS problem on multicore processor

Yamada, Tomonori; Kawai, Hiroshi*

In this study, computational performance of iterative solver of large linear equations with many right hand side vectors derived from finite element discretization in structural mechanics is evaluated. The current multicore architecture has memory wall problem and many iterative solvers suffer for this problem. Hence, high performance computing with iterative solver on multicore architecture is one of key issues in recent computer science. Because the performance of sparse matrix single vector multiplication has well surveyed, the performance of sparse matrix multiple vectors multiplication is investigated in this paper.