high performance computing on graphics processing units: hgpu.org

Accelerators, such as Graphic Processing Units (GPUs), are popular components of modern parallel systems. Their energy-efficient performance make them attractive components for modern data center nodes. However, they lack control for fair resource sharing amongst multiple users. This paper presents a runtime and Just In Time compiler that enables resource sharing control and software managed scheduling on accelerators. It is portable and transparent, requiring no modification or recompilation of existing systems or user applications. We provide an extensive evaluation of our scheme with over 40,000 different workloads on 2 platforms and we deliver fairness improvements ranging from 6.8x to 13.66x. In addition, we also deliver system throughput speedups ranging from 1.13x to 1.31x.