Arm 基于性能监控单元（PMU）的性能分析实践

所属专题：解析性能优化

嘉宾 : (1) 陈颖 Tony Chen | Arm 中国基础设施软件生态总监(2) 蒋良 Alan Jiang | Arm 基础设施事业部 主任性能工程师

会议室 : 宴会厅1

讲师介绍

专题演讲嘉宾：陈颖 Tony Chen

Arm 中国基础设施软件生态总监

Tony Chen is Director of Infrastructure Software Ecosystem at Arm China. He joined Arm China in 2015, responsible for building and leading Arm global open source software teams focusing on server and networking related open source software enablement and optimization on Arm64 architecture, covering kernel, virtualization, container; runtime languages and libraries; cloud computing, big data, database, storage; networking and cloud native related areas. Currently Tony is focusing on Infrastructure Software Ecosystem development for Arm in China and engaging with related open source communities. Tony has 24 years of experience in system software development and team management. Prior to Arm, Tony had spent 18 years at Intel holding various technical and management positions at Intel’s Software & Service Group, Data Center Group and Client Computing Group etc.

陈颖（Tony）是 Arm 中国基础设施软件生态总监。2015 年加入 Arm 中国，负责组建和带领Arm 全球开源软件团队从事与服务器和网络相关的开源软件在Arm64 架构上的使能和优化，涉及内核、虚拟化、容器、运行时语言和库、云计算、大数据、数据库、存储、网络和云原生等领域。目前 Tony 主要从事 与 Arm 基础设施相关的软件生态在国内的拓展，并与相关的开源社区进行合作。Tony 在系统软件开发和团队管理方面有 24 年的经验。在加入 Arm 之前，Tony 在 Intel 工作了 18 年，在 Intel 的软件与服务、数据中心和客户计算等部门担任过各种技术和管理职位。

专题演讲嘉宾：蒋良 Alan Jiang

Arm 基础设施事业部 主任性能工程师

Alan Jiang is system performance engineer, kernel developer, cloud architect and inventor with over 21 years of experience in building complex Enterprise Unix Systems, Multi-Petabyte Object Storage Systems and Multi-Tenant public/private clouds across continents. Currently at Arm, Alan is focusing on performance analysis and architectural exploration to optimize performance of cloud workloads running on Arm Architecture based products. Prior to Arm, Alan was Chief Cloud and Storage Architect at State Street Corp, led the effort of the Data Center and Storage Transformation initiative, and partnered with IBM to develop its first all-flash-based object storage system. Prior to State Street, Alan had spent 17 years at IBM holding varies technical leadership and engineering roles in the following areas – IBM Power System and AIX Operating System development, IBM Cleversafe Object Storage research and development, IBM Public Cloud performance monitoring and next generation control plane development.

蒋良（Alan），系统性能工程师、内核开发人员、云架构师和发明家，在构建复杂的 企业 Unix 系统、PB 级对象存储系统和跨洲的多租户公共/私有云方面拥有超过 21 年的经验。Alan 目前在 Arm 美国从事基于 Arm 架构产品和云工作负载相关的性能分析和优化。在加入 Arm 之前，Alan 是道富银行（State Street Corp）的首席云和存储架构师，领导了数据中心和存储转型，并与 IBM 合作开发了其首个全闪存对象存储系统。在加入道富银行之前，Alan 在 IBM 工作了 17 年，在 IBM Power System 和 AIX 操作系统、IBM Cleversafe 对象存储系统、IBM 公共云性能监控和下一代控制平面等领域担任过不同的技术领导和工程职务。

议题介绍

地点：宴会厅1
所属专题：解析性能优化

演讲：Arm 基于性能监控单元（PMU）的性能分析实践

Performance Monitor Unit (PMU) is a commonly available subsystem on most micro-processors nowadays that monitors micro-architectural events like elapsed cycles, cache hits, cache misses, etc. It helps in analyzing how applications or operating systems are performing on the processor. There is no exception for Arm to provide a comprehensive PMU subsystem for its wide range of IP Cores. For this talk, we will introduce PMU based performance analysis methodology and performance metrics by focusing on Arm Neoverse N1 based scalable compute SoCs. Furthermore, a detailed case study of DynamoRIO Stride benchmark analysis will help unveil the power of PMU profiling by analyzing Arm PMU events and metrics.

性能监控单元（Performance Monitor Unit，PMU）是目前大多数微处理器上常用的子系统，用于监控微架构事件，如运行周期、缓存命中率、缓存丢失等。它有助于分析应用程序或操作系统在处理器上的执行情况。Arm 也不例外，为其广泛的 IP 核提供全面的 PMU 子系统。在本次演讲中，我们将重点介绍在 Arm Neoverse N1（可扩展计算平台）上基于 PMU 的性能分析方法和性能指标。此外，会通过一个有关 DynamoRIO Stride 基准分析的详细案例，来展示如何通过 Arm PMU 事件和指标来做性能分析。

演讲提纲：
1. Introduction （简介）

  • Arm Architecture and Neoverse Scalable Compute Platform（Arm 架构和 Neoverse 可扩展计算平台）
  • Arm Neoverse N1 micro-architecture（Arm Neoverse N1 微架构）

2. Performance Monitor Unit (PMU) based Analysis（基于性能监控单元（PMU）的分析）

  • Methodology（方法）
  • Metrics（指标）
  • Tooling （工具）

3. Case Study – DynamoRIO Stride Benchmark Analysis（案例研究 - DynamoRIO Stride 基准分析）
4. Q & A（问答）

你将获得：

  • Help audience better know Arm Neoverse platform and N1 micro-architecture.
  • Help audience engage PMU based performance analysis methodology to profile workload on Arm platform and deepen the understanding by walking through DynamoRIO Stride Benchmark analysis based on Arm PMU metrics.
  • 帮助观众更好地了解 Arm Neoverse 平台和 N1 微架构。
  • 帮助观众了解如何基于 PMU 来分析 Arm 平台上的工作负载性能，并通过 DynamoRIO Stride基 准分析这个具体实例以加深对 Arm PMU 指标的理解。

本专题下其他演讲

联系我们

交通指南

上海宝华万豪酒店

Shanghai Marriott Hotel Parkview
地址：上海市静安区广中西路 333 号
Copyright © 2026, Geekbang Technology Ltd. All rights reserved. 极客邦控股（北京）有限公司 | 京 ICP 备 16027448 号 - 5京公网安备京公网安备 11010502039052号 | 产品资质

  • 微信咨询

  • 电话咨询

    联系电话：+86 18514549229

微信联系我们

小姐姐 小倩