high performance computing on graphics processing units: hgpu.org

Graphics Processing Units (GPUs) are commodity chips primarily used as coprocessors for processing high definition graphics on a computer system. It possess faster processing power and efficiency in handling accurate single and double floating point numbers with less power consumption compared to CPUs. Realising its potential in general purpose computing manufacturers of these chips have added programmability feature so that enormous power of these chips can be utilized for general purpose computing. However, programming these chips has emerged new area of research and persuaded many companies into determining ways of flexibility and ease of programming. Presently there are few programming models which have their own advantages and disadvantages but most matured and widely used of its kind is CUDA developed by NVidia however it is architecture dependent on NVidia architecture. Another directive based programming model OpenMP for Accelerator is under development by OpenMP forum which is architecture independent. Evaluations of both these models are elaborated by porting serial Molecular Dynamics code on each of these models. It appears that directive based approach is easier and less time consuming however it is slower in performance compared to CUDA. Moreover when these models are compared with Host OpenMP, they proved widely scalable with additional advantages of economics over maintainability and hardware cost.