• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

计算机工程与科学 ›› 2023, Vol. 45 ›› Issue (04): 590-598.

• 高性能计算 • 上一篇    下一篇

一种面向蒙特卡洛程序的128核可扩展体系结构

张立,黎铁军,张建民   

  1. (国防科技大学计算机学院,湖南 长沙 410073)
  • 收稿日期:2022-09-24 修回日期:2022-11-16 接受日期:2023-04-25 出版日期:2023-04-25 发布日期:2023-04-13
  • 基金资助:
    国家重点研发计划(2022YFB2803405);国家自然科学基金(62072464);装备预先研究项目(31511010101)

A 128-core scalable architecture for Monte Carlo application

ZHANG Li,LI Tie-jun,ZHANG Jian-min   

  1. (College of Computer Science and Technology,National University of Defense Technology,Changsha 410073,China)
  • Received:2022-09-24 Revised:2022-11-16 Accepted:2023-04-25 Online:2023-04-25 Published:2023-04-13

摘要: 蒙特卡洛方法是研究粒子输运问题的重要方法,为蒙特卡洛方法设计定制加速体系结构成为粒子输运模拟的研究热点。分析了一种典型的蒙特卡洛方法代理程序Quicksilver的程序特征,并对存储层次和簇规模等影响可扩展性的结构参数进行了探索,提出了一种面向蒙特卡洛程序的128核可扩展体系结构,在128核配置下实现了相比单核90倍的加速比和70.1%的扩展效率。

关键词: 蒙特卡洛方法, 粒子输运, 众核体系结构, 可扩展性,

Abstract: Monte Carlo method is an important method to study particle transport problems. Designing a customized acceleration architecture for Monte Carlo method has become a research hotspot in particle transport simulation. This paper profiles Quicksilver, a typical proxy application using Monte Carlo method, and makes an architectural exploration to the structural parameters that affect scalability, such as storage hierarchy and cluster size. Finally, a 128-core scalable architecture for Monte Carlo program is proposed, which achieves 90× speedup compared to a single core and a scaling efficiency of 70.1% under 128 cores.

Key words: Monte Carlo method, particle transport, manycore architecture, scalability, cluster