生命不息 折腾不止
0%

Posted on In ,

摘要

I_MPI_FABRICS参数设置

This topic is an excerpt from the Intel® MPI Library Reference Manual for Linux OS which provides further details on the I_MPI_FABRICS environment variable.

Select a particular network fabric to be used for communication.

Read more »

Posted on In

摘要

HPCG (High-Performance Conjugate Gradient) 是如今主要的 HPC 系统性能测试基准之一，是一个精简的大型稀疏方程组并行求解器，包含了区域分解（Domain Decomposition）、多重网格（Multigrid）和预条件子（Preconditioner）这些重要的线性方程组求解技术。我在学完 MATH 6644 这门课以及自己动手写了 3D Multigrid 以后，我便一直想看看 HPCG 是如何实现的（虽然我还没学过 DD）。下面的记录是我阅读 HPCG 3.0 官方参考实现的源代码的笔记，源代码参见 GitHub .

Read more »

Posted on

English Version

教育经历

  • 2019.09-2025.06(预计) 中山大学 计算机学院 博士研究生(直博生)

  • 2015.09-2019.06 中山大学 数据科学与计算机学院 计算机科学与技术(超算方向) 学士学位

    • 平均绩点：3.7/4.0 排名：4/22

研究领域

高性能网络：拓扑、路由、拥塞控制、InfiniBand、RoCE、聚合通信算法、容器、DPU、在网计算、可重构网络、光电混合网络、Benchmark

已录用论文

Guangnan Feng, Jiabin Xie, Dezun Dong, Yutong Lu. 2024. UNR: Unified Notifiable RMA Library for HPC. In SC ’24. (To be published).

Guangnan Feng, Dezun Dong, Shizhen Zhao, Yutong Lu. 2023. GRAP: Group-level Resource Allocation Policy for Reconfigurable Dragonfly Network in HPC. In 2023 International Conference on Supercomputing (ICS ’23), June 21–23, 2023, Orlando, FL, USA. ACM, New York, NY, USA, 13 pages.

Guangnan Feng, Dezun Dong, Yutong Lu. 2022. Optimized MPI Collective Algorithms for Dragonfly Topology. In 2022 International Conference on Supercomputing (ICS ‘22), June 28–30, 2022, Virtual Event, USA. ACM, New York, NY, USA, 11 pages.

Jiabin Xie, Guangnan Feng, Han Huang, Junxuan Feng, Yutong Lu. 2024. Extreme-scale Direct Numerical Simulation of Incompressible Turbulence on the Heterogeneous Many-core System. In 29th ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming (PPoPP 2024)

Jiabin Xie, Guangnan Feng, Han Huang, Junxuan Feng, Yutong Lu. 2023. Parallel Optimization Methods for Direct Numerical Simulation of High Reynolds Number Wall Turbulence with a Grid Size of 100 Billion. Research Poster at the International Conference for High Performance Computing, Networking, Storage and Analysis (SC’23), presented in Denver, CO, USA.

Yangle Zeng, Guangnan Feng, Zhiguang Chen, Yutong Lu and Nong Xiao. 2024. ATM: Area-Based Partition and Topology-Aware Mapping for Large-Scale SNN Simulation. In ISPA ‘24. (To be published).

Han Huang, Tengyang Zheng, Tianxing Yang, Yang Ye, Siran Liu, Zhe Tang, Shengyou Lu, Guangnan Feng, Zhiguang Chen, Dan Huang. 2024. Critique of “Productivity, Portability, Performance Data-Centric Python” by SCC Team From Sun Yat-sen University. In IEEE Transactions on Parallel & Distributed Systems (TPDS), vol. , no. 01, pp. 1-4, 5555.

专利

  • 一种用于Dragonfly拓扑的Allgather方法及装置 (已授权，202210707577.0)
  • 适用于Dragonfly拓扑的环形通信方法 (已授权，202210678883.6)
  • 用于Dragonfly拓扑的Allreduce方法 (已授权，202210678611.6)
  • 用于Dragonfly拓扑的Alltoall方法 (已公开，202310256746.8)
  • 基于可重构网络的资源分配方法、系统、设备以及介质 (已公开，202310594898.9)

获奖经历

研究经历

  • RoCE协议在HPC中的应用分析 2018.09-2019.09
  • RDMA在容器中的应用分析 2019.09-2020.01
  • 天河二号路由优化 2020.02-2020.10
  • 天河二号网络故障分析 2020.11-2021.02
  • DPU在HPC上应用的调研与分析 2021.03-2021.05
  • 面向Dragonfly拓扑优化的MPI聚合通信算法 2021.07-2022.07
  • 超大规模槽道湍流直接数值模拟中的通信优化 2022.10-2023.04
  • 面向高性能计算的可重构网络 2022.08-2023.04
  • 超大规模脉冲神经网络模拟中的聚合通信方法优化 2023.05-至今
  • 编译中的通信自动优化方法研究 2023.10-至今
  • 超算网络评测方法研究 2023.09-至今

志愿者经历

  • 中山大学数据科学与计算机学院第三届学生会主席 @2017.06-2018.06
  • 广西防城港三下乡 @2017.08
  • 中山大学数据科学与计算机学院第二届学生会副主席 @2016.06-2017.06
  • 广东省湛江市安铺镇安铺中学支教 数学老师兼任副班主任 @2016.07-2016.08

专业技能

  • ROSS模拟器(一款乐观的并行离散事件模拟器)
  • CODES模拟器(一款基于ROSS的网络模拟器)
  • C, Python
  • OpenMP, MPI
  • Linux

兴趣爱好

NAS、路由器、HTPC、VPS、装机、服务器维修、自建各种服务

看动漫、打游戏

联系方式

ad at mail dot sysu dot tech

Posted on

中文版(Chinese Version)

Education

Research Interests

High Performance Network: Topology, Dragonfly, Reconfigurable Network, Routing, Congestion Control, Infiniband, RoCE, Collective Algorithm, DPU, Benchmark.

Publications

Guangnan Feng, Dezun Dong, and Yutong Lu. 2022. Optimized MPI Collective Algorithms for Dragonfly Topology. In 2022 International Conference on Supercomputing (ICS ‘22), June 28–30, 2022, Virtual Event, USA. ACM, New York, NY, USA, 11 pages.

Guangnan Feng, Dezun Dong, Shizhen Zhao, and Yutong Lu. 2023. GRAP: Group-level Resource Allocation Policy for Reconfigurable Dragonfly Network in HPC. In 2023 International Conference on Supercomputing (ICS ’23), June 21–23, 2023, Orlando, FL, USA. ACM, New York, NY, USA, 13 pages.

Jiabin Xie, Guangnan Feng, Han Huang. 2023. Parallel Optimization Methods for Direct Numerical Simulation of High Reynolds Number Wall Turbulence with a Grid Size of 100 Billion. Research Poster at the International Conference for High Performance Computing, Networking, Storage and Analysis (SC’23), presented in Denver, CO, USA.

Jiabin Xie, Guangnan Feng, Han Huang, Junxuan Feng, Yutong Lu. 2024 Extreme-scale Direct Numerical Simulation of Incompressible Turbulence on the Heterogeneous Many-core System. In 29th ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming (PPoPP 2024) (Accepted, not published yet. If you need the preview edition, please contact me!)

Honors and Awards

  • Sun Yat-sen University President’s Scholarship for Doctoral Candidates (Special Class) @2023.10

  • PAC22, 2nd Place in Application Track @ Jinan, China, 2022.11

  • 2022 Kunpeng Application Innovation Competition, 1st Prize in HPC track @ Hangzhou, China 2022.10

  • ISC19, 4th Place (team captain) @ Frankfurt, Germany, 2019.06

  • ASC19, Highest HPL, e-Prize, Group Competition Champion, 3rd Place (student advisor) @ Dalian, China, 2019.04

  • Yuerou Li Foundation Scholarship (top prize scholarship in SYSU) @ 2018.12

  • 1st Prize Student Scholarship in SYSU @ 2018.10

  • CPC18, 4th Place (team member) @ Wuxi, China, 2018.10

  • ASC18, 6th Place (team captain) @ Nanchang, China, 2018.05

  • 3rd Prize Student Scholarship in SYSU × 2 @ 2017.10, 2016.10

  • Outstanding Student Leader in SYSU × 2 @ 2017.12, 2016.12

Abbreviation

PAC: Parallel Application Challenge

ISC: ISC Student Cluster Competition

ASC: Asia Supercomputer Community - Student Supercomputer Challenge

CPC: China Parallel Application Challenge on Domestic CPU

SYSU: Sun Yat-sen University

Research Experience

  • The Analysis of RoCE for HPC application, 2018.09-2019.09
  • The Analysis of RDMA in container, 2019.09-2020.01
  • Routing optimization for Tianhe-2, 2020.02-2020.10
  • Fault analysis for Tianhe-2 2020.11-2021.02
  • Investigation and analysis of DPU for HPC application 2021.03-2021.05
  • Optimized MPI Collective Algorithms for Dragonfly Topology 2021.07-2022.06
  • Communication Optimization for Direct Numerical Simulation of High Reynolds Number Wall Turbulence with a Grid Size of 100 Billion, 2022.10-2023.04
  • Reconfigurable Dragonfly Network for HPC, 2022.07-2023.04
  • Collective Communication Optimization for Large Scale Spiking Neural Network Simulations, 2023.05-present
  • Auto Communication Optimization in Compiler, 2023.10-present
  • Benchmark for HPC Network, 2023.09-present

Skills

Contact Me

ad at mail dot sysu dot tech