high performance computing on graphics processing units: hgpu.org

Graphics processing units (GPUs) have emerged as a powerful platform for high-performance computation. They have been successfully used to accelerate many scientific workloads. Typically, the computationally intensive parts of the application are offloaded to the GPU, which serves as the CPU’s parallel coprocessor. The key to effective utilization of GPUs for scientific computing is the design and implementation of efficient data-parallel algorithms that can scale to hundreds of tightly coupled processing units. Many compute intensive scientific applications are well suited to GPUs, due to their extensive computational requirements, and because they lend themselves to parallel processing implementations. The use of multiple GPUs can bring even more computational power to bear on highly parallelizable computational problems. This paper discusses performance results for some fundamental cores of scientific applications such as fft, smith-waterman sequence alignment algorithm, and data encryption standard (DES) on the Nvidia GPUs using the CUDA programming model. Results have demonstrated acceleration up to 25 times speedup using a single G80 Nvidia GPU.