high performance computing on graphics processing units: hgpu.org

Graphics Processing Units (GPUs) have been extensively applied in the High Performance Computing (HPC) community. HPC applications require additional special programming environments to improve the utilization of GPUs, for example, NVIDIA’s CUDA and Khronos group’s OpenCL. This thesis will introduce a preprocessor framework called HPC.NET, which is deployed on the Microsoft .NET platform to meet demands of GPU programmers while utilizing familiar languages. Multiple .NET languages are supported as well as detected loops and .NET TPL methods are converted into NVIDIA CUDA code automatically using common language processing techniques. Moreover, HPC.NET is able to scale to utilize multiple GPUs in multiple machines based on Windows HPC Server 2008 R2. This framework takes care of all task scheduling and communication in order to achieve better programmability and high performance in .NET platform.