high performance computing on graphics processing units: hgpu.org

General Purpose Computation on Graphics Processing Units (GP-GPU) has been recognized as viable and inexpensive technique in recent trends of parallel computing. Earlier this technology has only been used as commodity processing units in video cards which have been used for generating graphics in High resolution. This technology provides greater computational power with its high performance many core computing units than commodity CPU. High computational jobs, like Many Physics experiments require simulations to understand and analyze the results. But these simulations are time consuming it requires days to weeks to get results from conventional CPU. In those cases GPU can be used as an accelerator with CPU to make a heterogeneous machine which will reduce that execution time in minutes from hours. This thesis investigates GPU architecture and CUDA Programming Model to understand GPU Computing. Some of Basic Linear Algebra Subroutines like Matrix Addition, Matrix Multiplication etc. also explored. The thesis also taken up acceleration of Sedimentary Rock structure simulation as a major problem which is to be ported on GPU architecture.