Utilization of Parallel Solver Libraries to solve Structural and Fluid problems

“With the processors becoming MULTICORE, even desktop operations are becoming increasingly parallel. DUAL core computers are becoming the norm and QUAD Core is very affordable for even Small to Medium Enterprises (SME). FEA being commonly used for solving large size problems (millions of Degrees of Freedom) even on desktops, there is a need for FEA Vendors to upgrade the solvers to exploit the multi-core capabilities. If one were to develop and maintain these solutions, it would involve considerable cost and time. However, same methodologies like OpenMP that were used for multi processor shared memory computers are equally valid for the MULTICORE systems. It is in this context that the Parallel Math and Matrix libraries provided by the hardware vendors like INTEL, which are fine tuned for performance on their processors and memory configurations, offer a significant advantage. It is profitable to exploit the availability of these libraries to cut short the development time and cost. This approach would allow the FEA Vendors to quickly adapt and exploit the newer hardware and the associated technologies. In this paper we discuss our experience of using the INTEL MATH KERNEL LIBRARY for extending NISA Solvers on to multi core systems. We would also define a framework for the users to incorporate the hardware vendor specific solvers for best performance”.