欢迎访问中国科学院大学学报,今天是

中国科学院大学学报 ›› 2021, Vol. 38 ›› Issue (5): 696-701.DOI: 10.7523/j.issn.2095-6134.2021.05.014

• 计算机科学 • 上一篇    下一篇

基于EDA仿真软件的多资源调度算法

王静1,2, 陈岚1, 张贺1, 王海永1   

  1. 1. 中国科学院微电子研究所 三维及纳米集成电路设计自动化技术北京市重点实验室, 北京 100029;
    2. 中国科学院大学, 北京 100049
  • 收稿日期:2019-12-12 修回日期:2020-01-21 发布日期:2021-09-13
  • 通讯作者: 陈岚
  • 基金资助:
    国家重点研发计划高性能计算专项(2017YFB0203501)和北京市科技新星与领军人才专项(Z171100001117147)资助

Multi-resource scheduling algorithm based on EDA simulation software

WANG Jing1,2, CHEN Lan1, ZHANG He1, WANG Haiyong1   

  1. 1. Beijing Key Laboratory of Three-dimensional and Nanometer Integrated Circuit Design Automation Technology, Institute of Microelectronics, Chinese Academy of Sciences, Beijing 100029, China;
    2. University of Chinese Academy of Sciences, Beijing 100049, China
  • Received:2019-12-12 Revised:2020-01-21 Published:2021-09-13

摘要: 为提高电子设计自动化(electronic design automation,EDA)并行仿真任务的资源利用率并保证公平性,在占优资源公平分配机制(dominant resource fariness,DRF)的基础上,提出考虑license的占优资源公平分配(dominant resource fairness allocation algorithm considering license,LDRF)算法。一方面考虑多类型资源的公平调度问题,满足EDA任务对资源种类多样性的需求;另一方面考虑对EDA工具的license资源调度,避免任务占有硬件资源但是没有获得license授权不能运行,导致资源利用率下降。实验仿真结果表明,在license有限的条件下,LDRF的平均CPU资源利用率比DRF算法提高60%,平均内存资源利用率比DRF算法提高34%。

关键词: 资源分配, DRF算法, 公平性, EDA软件, license调度

Abstract: In order to improve the resource utilization of electronic design automation(EDA) parallel simulation tasks and ensure fairness, on the basis of dominant resource fairness allocation(DRF), dominant resource fairness allocation algorithm considering license(LDRF) is proposed. On the one hand, the problem of fair scheduling of multiple types of resources is considered, satisfying the diversity of resource requirements of EDA tasks. On the other hand, the license resource scheduling of EDA tool is considered to avoid the reduction of utilization caused by the lack of license for hardware resources. The experimental simulation results show that under the condition of limited license, the CPU resource utilization of LDRF is increased by 60% compared with DRF, and the memory resource utilization rate is increased by 34%.

Key words: resource allocation, DRF algorithm, fairness, EDA software, license scheduling

中图分类号: