EffiSha: A Software Framework for Enabling Effficient Preemptive Scheduling of GPU
EffiSha: A Software Framework for Enabling Effficient Preemptive Scheduling of GPU
复制标题
DOI：
10.1145/3018743.3018748
复制
发表时间：
2017-01
期刊：
影响因子：
--
通讯作者：
Guoyang Chen;Yue Zhao;Xipeng Shen;Huiyang Zhou
中科院分区：
文献类型：
--
作者：
Guoyang Chen;Yue Zhao;Xipeng Shen;Huiyang Zhou
现代gpu被广泛应用于许多多任务环境，包括数据中心和智能手机。然而，目前对多个GPU内核（来自不同应用程序）调度的支持是有限的，这是GPU满足许多实际需求的主要障碍。这项工作首次证明了在现有的GPU上，即使没有特殊的硬件支持，GPU内核的高效抢占调度也是可能的。具体来说，它提出了EffiSha，这是一个纯软件框架，可以以非常低的开销实现GPU内核的抢先调度。启用的抢占式调度器为不同优先级的内核提供了灵活的支持，并展示了减少平均周转时间和提高共享现代GPU的程序的系统总体吞吐量的巨大潜力。
Modern GPUs are broadly adopted in many multitasking environments, including data centers and smartphones. However, the current support for the scheduling of multiple GPU kernels (from different applications) is limited, forming a major barrier for GPU to meet many practical needs. This work for the first time demonstrates that on existing GPUs, efficient preemptive scheduling of GPU kernels is possible even without special hardware support. Specifically, it presents EffiSha, a pure software framework that enables preemptive scheduling of GPU kernels with very low overhead. The enabled preemptive scheduler offers flexible support of kernels of different priorities, and demonstrates significant potential for reducing the average turnaround time and improving the system overall throughput of programs that time share a modern GPU.