high performance computing on graphics processing units: hgpu.org

Over the past few years, there has been a shift towards multi-core processors, driven partially by physical limitations. Mistaken assumptions of how effective and useful parallel systems can be have also provided motivation for this change. In this paper, we seek to directly identify the barriers to parallel computation. The barriers are not, as conventional wisdom might hold, a lack of programmer ingenuity or research funding. Rather, the limitations are fundamental to the problems addressed; new computer architectures and compilers offer limited benefit.