GPUs are being increasingly adopted as compute accelerators in many domains, spanning environments from mobile systems to cloud computing. These systems are usually running multiple applications, from one or several users. However GPUs do not provide the support for resource sharing traditionally expected in these scenarios. Thus, such systems are unable to provide key multiprogrammed workload requirements, such as responsiveness, fairness or quality of service. In this paper, we propose a set of hardware extensions that allow GPUs to efficiently support multiprogrammed GPU workloads. We argue for preemptive multitasking and design two preemption mechanisms that can be used to implement GPU scheduling policies. We extend the architecture to allow concurrent execution of GPU kernels from different user processes and implement a scheduling policy that dynamically distributes the GPU cores among concurrently running kernels, according to their priorities. We extend the NVIDIA GK110 (Kepler) like GPU architecture with our proposals and evaluate them on a set of multiprogrammed workloads with up to eight concurrent processes. Our proposals improve execution time of high-priority processes by 15.6x, the average application turnaround time between 1.5x to 2x, and system fairness up to 3.4x.

We would like to thank the anonymous reviewers, Alexan- der Veidenbaum, Carlos Villavieja, Lluis Vilanova, Lluc Al- varez, and Marc Jorda on their comments and help improving our work and this paper. This work is supported by Euro- pean Commission through TERAFLUX (FP7-249013), Mont- Blanc (FP7-288777), and RoMoL (GA-321253) projects, NVIDIA through the CUDA Center of Excellence program, Spanish Government through Programa Severo Ochoa (SEV-2011-0067) and Spanish Ministry of Science and Technology through TIN2007-60625 and TIN2012-34557 projects.

Peer Reviewed