ARM GPU的多任务调度设计与实现

丑文龙; 梅魁志; 高增辉; 李博良

doi:10.7652/xjtuxb201412014

您当前的位置：

首页 >

文章列表页 >

ARM GPU的多任务调度设计与实现

更新时间：2025-07-09

- ARM GPU的多任务调度设计与实现
- Design and Implementation of Multitask Scheduling for Embedded ARM GPU
- 西安交通大学学报 2014年48卷第12期页码：87-92
- 作者机构：
  
  西安交通大学电子与信息工程学院,西安,710049
- 作者简介：
- 基金信息：
  
  : 2014-06-23。作者简介: 丑文龙(1989—),男,硕士生;梅魁志(通信作者),男,副教授。基金项目: 国家高技术研究发展计划资助项目(2012AA010904);国家自然科学基金资助项目(61375023)
- DOI：10.7652/xjtuxb201412014
  中图分类号： TP316
- 网络首发：2014-12-10，
  
  纸质出版：2014
- 稿件说明：
移动端阅览
丑文龙, 梅魁志, 高增辉, 等. ARM GPU的多任务调度设计与实现[J]. 西安交通大学学报, 2014,48(12):87-92.

Design and Implementation of Multitask Scheduling for Embedded ARM GPU[J]. 2014, 48(12): 87-92.
丑文龙, 梅魁志, 高增辉, 等. ARM GPU的多任务调度设计与实现[J]. 西安交通大学学报, 2014,48(12):87-92. DOI： 10.7652/xjtuxb201412014.

Design and Implementation of Multitask Scheduling for Embedded ARM GPU[J]. 2014, 48(12): 87-92. DOI： 10.7652/xjtuxb201412014.

摘要

针对现有GPU任务调度系统在多任务环境下不能保证图形任务响应时间的问题

提出基于分类和多优先级队列(CPMQ)的调度方案

并在ARM的嵌入式GPU上实现验证。该方案中

将GPU的多任务划分为图形任务、通用计算任务和实时图形3类任务并分别建立队列排队

其中图形任务和通用计算任务按照优先级在各自队列中排队

实时图形按照任务截止时间排队。面向多队列的任务调度

优先从实时任务队列中选择任务

并按照加权公平算法分别在图形任务队列和通用计算队列中选择任务。实验结果表明:相比于ARM GPU的原有调度系统

CPMQ在不显著增加通用计算任务的执行时间和调度开销的情况下

将实时图形任务的帧率提升了5%～20%。

Abstract

A scheduling solution of class priority multiple queue(CPMQ)is proposed to solve the problem that the response time to graphic tasks is not ensured by existing task scheduling systems of GPU under multitask conditions

and the schedule is implemented on an embedded system. Multiple tasks on GPU are firstly classified into three classes of tasks

that is

graphic tasks

real-time graphic tasks and general purpose computing tasks. These three classes of tasks then queued respectively with different queuing policy. Graphic tasks and general purpose computing tasks are queued by their priorities

while real-time graphic tasks are queued by their deadlines. When the multi-class tasks are scheduled

real-time graphic tasks are selected at first

and then graphic tasks and general purpose computing tasks are selected out using a weighted fair queuing algorithm. Experimental results and comparisons with the original scheduling system of ARM's GPU show that CPMQ increases the frame rate of reel time graphic tasks by 5%-20% without significant increase in execution time of general purpose computing tasks and scheduling expense.

关键词

Keywords

references

JOG A, KAYIRAN O, NACHIAPPAN N C, et al. OWL: cooperative thread array aware scheduling techniques for improving GPGPU performance[C]∥ Proceedings of the 18th International Conference on Architectural Support for Programming Languages and Operating Systems. New York, NY, USA: ACM, 2013: 395-406.

PAUL B. Introduction to the direct rendering infrastructure[EB/OL].(2000-08-10)[2014-03-23]. http:∥dri.sourceforge.net/doc/DRIintro.html.

KATO S, LAKSHMANAN K, RAJKUMAR R, et al. TimeGraph: GPU scheduling for real-time multi-tasking environments[C]∥ Proceedings of the 2011 USENIX Conference on USENIX Annual Technical Conference. Berkeley, CA, USA: USENIX Association, 2011: 17-30.

MARROQUIM R, MAXIMO A. Introduction to GPU programming with GLSL[C]∥ Proceedings of the 2009 Tutorials of the 22nd Brazilian Symposium on Computer Graphics and Image Processing. Washington, DC, USA: IEEE Computer Society, 2009: 3-16.[5] BAUTIN M, DWARAKINATH A, CHIUEH T. Graphic engine resource management[C]∥Proceedings of the International Society for Optics and Photonics. Bellingham, WA, USA: SPIE, 2008: 68180O.

PRONOVOST S. Windows display driver model(WDDM)v2 and beyond[C/OL]∥ Proceedings of the Windows Hardware Engineering Conference.[2014-03-23].http:∥ci.nii.ac.jp/naid/10018383501/.

WONG C S, TAN I, KUMARI R D, et al. Towards achieving fairness in the Linux scheduler[J]. Operating Systems Review, 2008, 42(5): 34-43.

BENNETT J C, ZHANG Hui. WF2Q: worst-case fair weighted fair queuing[C]∥ Proceedings of the 15th Annual Joint Conference of the IEEE Computer Societies on Networking the next Generation. Piscataway, NJ, USA: IEEE, 1996: 120-128.

WONG H T. Packet scheduling using dual weight single priority queue: USA, 6570883[P]. 2003-05-27.

DOYTCHINOV B, LEHOCZKY J, SHREVE S. Real-time queues in heavy traffic with earliest-deadline-first queue discipline[J]. The Annals of Applied Probability, 2001, 11(2): 332-378.

张虹,郑霄,赵丹.GPU加速窦房结计算机仿真的实现及优化.2014,48(7):60-64.[doi:10.7652/xjtuxb201407011]

周秦武,隋芳芳,白平,等.嵌入式无接触视频心率检测方法.2013,47(12):55-60.[doi:10.7652/xjtuxb201312010]

李亮,王恩东,朱正东,等.应用动态生成树的GPU显存数据复用优化.2013,47(10):44-50.[doi:10.7652/xjtuxb2013 10008]

张保,董小社,白秀秀,等.CPU-GPU系统中基于剖分的全局性能优化方法.2012,46(2):17-23.[doi:10.7652/xjtuxb 201202004]

向坤,陈娟,张安学,等.提高喇叭天线增益的超介质构建方法.2011,45(2):92-96.[doi:10.7652/xjtuxb201102019]

邹华,高新波,吕新荣.一种八叉树编码加速的3D纹理体绘制算法.2008,42(12):1490-1494.[doi:10.7652/xjtuxb2008 12012]

刘晓东,寻亮,马栋,等.基于球体追踪的动态视差遮挡映射算法.2007,41(12):1401-1405.[doi:10.7652/xjtuxb200712 004]

赵保华,张炜,林华辉,等.一种通信有限状态机的被动测试及其错误诊断.2007,41(6):640-644.[doi:10.7652/xjtuxb 200706003]

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

一种5G网络低时延资源调度算法

电动车辆间接热系统快速除湿响应迟滞机理及优化控制策略

采用机器学习的大规模并行程序I/O性能建模与预测方法

张角凹坑气膜孔对双层壁气膜冷却特性的影响

超临界水异常区域热力学特性研究