数据与计算发展前沿 ›› 2020, Vol. 2 ›› Issue (1): 70-84.

doi: 10.11871/jfdc.issn.2096-742X.2020.01.006

所属专题: “高性能与高通量计算及应用”专刊

• 专刊:高性能与高通量计算及应用 • 上一篇    下一篇

高通量众核处理器设计

叶笑春1,李文明1,张洋1,张浩1,王达1,范东睿1,2,*()   

  1. 1. 中国科学院计算技术研究所,计算机体系结构国家重点实验室,北京 100190
    2. 中国科学院大学,北京 100049
  • 收稿日期:2019-11-30 出版日期:2020-02-20 发布日期:2020-03-28
  • 通讯作者: 范东睿
  • 作者简介:叶笑春,中国科学院计算机技术研究所,博士,副研究员,计算机学会会员,主要研究领域为高通量处理器体系结构及软件模拟技术。
    本文主要承担的工作为众核处理器架构设计和文章修改。
    Ye Xiaochun, Ph.D., is an associate professor. Member of China Computer Federation. His research interests include high throughput processor architecture and software simulation.
    The main contributions to this paper is the design of many-core architecture and paper revision.
    E-mail: yexiaochun@ict.ac.cn|李文明,中国科学院计算机技术研究所,博士,副研究员,计算机学会会员,主要研究领域为高通量处理器设计及软件模拟技术。
    本文主要承担的工作为高密度路网设计、片上存储设计及文章编写。
    Li Wenming, Ph.D., is an associate professor. Member of China Computer Federation. His research interests include high throughput computing architecture and software simulation.
    The main contributions to this paper is the design of high-density NoC, on-chip memory and paper writing.
    E-mail: liwenming@ict.ac.cn|张洋,中国科学院计算机技术研究所,博士,工程师,计算机学会会员,主要研究领域为高通量处理器设计及实时系统。
    本文主要承担的工作为任务实时性调度机制设计及实验分析。
    Zhang Yang, Ph.D., is an engineer. Member of China Computer Federation. His research interests include high throughput computer architecture, real-time system.
    The main contributions to this paper is the design of real-time scheduler and experiments analysis.
    E-mail: zhangyang@ict.ac.cn|张浩,中国科学院计算机技术研究所,博士,高级工程师,计算机学会会员,主要研究领域为高通量计算机体系结构。
    本文主要承担的工作为结构设计的可行性分析及实验设计。
    Zhang Hao, Ph.D., is an associate professor. Member of China Computer Federation. His research interests include high throughput computer architecture.
    The main contributions to this paper is conducting the feasibility analysis and experiments design.
    E-mail: zhanghao@ict.ac.cn|王达,中国科学院计算机技术研究所,博士,副研究员,计算机学会会员,主要研究领域为处理器微体系结构。
    本文主要承担的工作为处理器核内多线程微结构设计及线程实时性任务调度结构设计。
    Wang Da, Ph.D., is an associate professor. Member of China Computer Federation. Her research interest focuses on processor microarchitecture design.
    The main contributions to this paper is the design of multi-threaded microstructure and real-time task scheduling architecture.
    E-mail: wangda@ict.ac.cn|范东睿,中国科学院计算机技术研究所,博士,研究员,博士生导师,计算机学会高级会员,主要研究方向为高通量/高性能处理器体系结构。
    本文主要承担的工作为研究项目负责人及整体架构设计。
    Fan Dongrui, Ph.D., is a professor. Senior Member of China Computer Federation. His research interests include high throughput and High performance many-core processor microarchitecture.
    The main contributions to this paper is the leader of the project and design of the overall architecture.
  • 基金资助:
    国家自然科学基金重点项目“后E级时代的新型高能效处理器体系结构”(61732018);国家自然科学基金面上项目“据流众核体系结构中的数据通路优化研究”(61872335);国家自然科学基金青年基金项目“高通量众核处理器访存数据通路优化研究”(61802367)

High-Throughput Many-Core Processor Design

Ye Xiaochun1,Li Wenming1,Zhang Yang1,Zhang Hao1,Wang Da1,Fan Dongrui1,2,*()   

  1. 1. SKL of Computer Architecture, Institute of Computing Technology, Chinese Academy of Sciences,Beijing 100190, China
    2. University of Chinese Academy of Sciences, Beijing 100049, China
  • Received:2019-11-30 Online:2020-02-20 Published:2020-03-28
  • Contact: Fan Dongrui

摘要:

【目的】随着云计算、物联网以及人工智能等新型高通量应用的迅速兴起,高性能计算的主要应用从传统的科学与工程计算为主逐步演变为以新兴数据处理为核心,这给传统处理器带来了巨大的挑战,而高通量众核处理器作为面向此类应用的新型处理器结构成为重要的研究方向。【方法】针对上述问题,本文分析了高通量典型应用特征,从数据处理端、传输端以及存储端三个核心环节开展了高通量众核处理器关键技术设计探讨,包括实时任务动态调度、高密度片上网络设计、片上存储层次优化等。【结果】实验结果显示上述机制可以有效确保任务的服务质量,提升网络的数据吞吐率,以及简化片上存储层次。【结论】随着万物互联时代对高并发强实时处理的迫切需求,高通量众核处理器有望成为未来数据中心的核心处理引擎。

关键词: 高通量计算, 众核处理器, 数据通路

Abstract:

[Objective] With the rapid growth of new high-throughput applications such as cloud computing, the Internet of Things, and artificial intelligence, the main applications of high-performance computing have gradually evolved from traditional scientific and engineering computing to emerging data processing, which brought huge challenges to traditional processors. High-throughput many-core processors are becoming a new type of processor architecture dealing with such applications and therefore an important research direction. [Method] In view of the above problems, this paper analyzes the typical characteristics of high-throughput applications, and discusses the key design of high-throughput many-core processors from the three core aspects of data processing, transmission, and storage. The design includes real-time task dynamic scheduling, high-density on-chip network design and on-chip storage hierarchy optimization, etc. [Results] The experimental results show that the above mechanism can effectively ensure the service quality of tasks, improve the data throughput rates, and simplify the on-chip memory hierarchy. [Conclusion] With the urgent demand for high concurrency and strong real-time processing in the era of Internet of Everything, high-throughput many-core processors are expected to become the main processing engine in future data centers.

Key words: high throughput computing, many-core processor, data path