Abstract
Since General structural analysis program like NASTRAN(NASA Structural Analysis Program)[1] performs 'Out-of-Core' method, it is required to configure highly balanced system with CPU performance , memory bandwidth and I/O capability to realize high-performance computing. The Watson Sparse Matrix Package,WSMP[2][3] ,is a software package for solving linear equations which has being developed by IBM Watson Research Center . WSMP enables to solve large scale sparse matrices to achieve robust and outstanding performance using 'In-Core' method. This paper describes techniques how to apply WSMP to ISV(Independent Software Vendor) package software exemplified by NASTRAN and also describes it enables to modify program characteristics to realize performance improvement. Many CAE(Computer Aided Engineering) software including NSTARAN provide functionality of external user interface that allows users to encapsulates matrices solving part. That is , internal matrices are possible to solve with combination of the user interface and user program written by FORTRAN andlor c. As the result of applying this methodology, it enable simple system configuration for NASTRAN. Also it is possible to improve total elapsed time by eliminating I/O requests and fully utilized parallel processing capability.