Algorithm Development for Multi-Core GPU-Based Computing Architectures

For the Phase I effort, a team comprised of experts in the areas of alternative high performance computing architectures such as the GPU, IBM Cell BE and Field Programmable Gate Arrays (Research Institution), and finite element analysis and continuum mechanics (Global Engineering Research and Technologies) has been formed to design and develop a framework and a library of algorithms optimized for the GPU architecture and targeted for solving partial differential equations. These algorithms will address problems relevant and important to aeronautics industry, which include solid mechanics, heat transfer and fluid dynamics. In solving specific PDEs, most suitable method will be selected. Newly developed algorithms on the GPU will be evaluated through a performance comparison against CPU platforms. Scalability of the developed algorithms will be tested. BENEFIT: The Phase I effort will establish the validity, merit and feasibility of the proposed approach for solving specific partial differential equations (PDEs). This effort will serve as the foundation for the Phase II development effort. During the Phase II, the framework developed during Phase I will be extended to include additional PDEs that are commonly used in research related to Air Force and other branches of DOD. This research will lead to obtaining solutions to PDEs faster with greater accuracy. Also, solutions to problems previously discarded due to being deemed computationally prohibitive will be within reach. Considering the fact that numerical solution of PDEs is one of the essential components in design, manufacturing and analysis of machines and structures, the vast majority of manufacturing sector (aerospace, automobile, electronics, etc.) will find this research highly attractive.