Cooperative kernels: GPU multitasking for blocking algorithms

Abstract

There is growing interest in accelerating irregular data-parallel algorithms on GPUs. These algorithms are typically blocking, so they require fair scheduling. But GPU programming models (e.g. OpenCL) do not mandate fair scheduling, and GPU schedulers are unfair in practice. Current approaches avoid this issue by exploiting scheduling quirks of today's GPUs in a manner that does not allow the GPU to be shared with other workloads (such as graphics rendering tasks). We propose cooperative kernels, an extension to the traditional GPU programming model geared towards writing blocking algorithms. Workgroups of a cooperative kernel are fairly scheduled, and multitasking is supported via a small set of language extensions through which the kernel and scheduler cooperate. We describe a prototype implementation of a cooperative kernel framework implemented in OpenCL 2.0 and evaluate our approach by porting a set of blocking GPU applications to cooperative kernels and examining their performance under multitasking.

MCC’2017 – The Seventh Model Checking Contest

“…Finally, FlexibleBarrier describes a special barrier used in the context of cooperative kernels, so as to enable multitasking on GPUs [39]. In this programming model, processes can offer to be killed or forked by the scheduler at precise points of their execution; hence, the number of alive processes varies dynamically and the flexible barrier must interact with the scheduler to know how many processes have to be synchronised.…”