UC Davis

The graphics processing unit (GPU) has become an integral part oftoday's mainstream computing systems. Over the past six years, therehas been a marked increase in the performance and capabilities ofGPUs. The modern GPU is not only a powerful graphics engine but also ahighly-parallel programmable processor featuring peak arithmetic andmemory bandwidth that substantially outpaces its CPU counterpart. TheGPU's rapid increase in both programmability and capability hasspawned a research community that has successfully mapped a broadrange of computationally demanding, complex problems to the GPU. Thiseffort in general-purpose computing on the GPU (GPGPU), alsoknown as GPU computing, has positioned the GPU as a compellingalternative to traditional microprocessors in high-performancecomputer systems of the future. We describe the background, hardware,and programming model for GPU computing, summarize the state of theart in tools and techniques, and present four GPU computing successesin game physics and computational biophysics that deliverorder-of-magnitude performance gains over optimized CPU applications.