並列向き前処理を用いた反復解法による構造解析

抄録

In this work, we describe the implementation of a parallel iterative solver in a large structural analysis program. To achieve maximum performance and stability for the solver, we needed to develop a parallel preconditioner. The preconditioner we use is Factorized Sparse Approximate Inverse. This preconditioner requires only the same amount of storage as the stiffness matrix and has excellent scalability. We devoloped a parallel structural analysis system using this iterative solver and look at a large industrial problem. The numerical results show the effectiveness of this solver.