CentOS 8.5高性能计算开发环境配置备忘

CentOS 8.5环境配置

在CentOS 8.5系统下配置高性能计算环境。主要包含编译器套件、分布式并行套件、各种数值计算库、Python环境、辅助管理工具、GPU开发环境等。利用CentOS自带的RPM仓库，加上OpenHPC和Intel oneAPI仓库,大体上能满足yum和dnf安装需求!

安装配置过程零零散散，特汇集一处，随时增量完善，以备忘!

---

工具命令

• yum
• dnf
• module

---

安装源仓库

• repolist

  # dnf repolist
  仓库 id                             仓库名称
  OpenHPC                             OpenHPC-2 - Base
  OpenHPC-updates                     OpenHPC-2 - Updates
  appstream                           CentOS Linux 8 - AppStream
  baseos                              CentOS Linux 8 - BaseOS
  cudnn-local-rhel8-8.5.0.96          cudnn-local-rhel8-8.5.0.96
  epel                                Extra Packages for Enterprise Linux 8 - x86_64
  extras                              CentOS Linux 8 - Extras
  oneAPI                              Intel(R) oneAPI repository
  
  

• OpenHPC仓库包含很多HPC开发所用的工具和数学库(比如,SuperLU和MUMPS)

• 安装Intel oneAPI 仓库(包含rpm安装包)

  yum install intel-oneapi-toolkit-release-ohpc
  

• 搜索powertools会有很多rpm安装包

  dnf --enablerepo=powertools search openblas-devel
  

• 安装environment-modules可以使用module管理配置开发环境

  $ rpm -qa | grep env
  environment-modules-4.5.2-1.el8.x86_64
  

• 安装conda-4.10.3, 用conda管理python环境

  # rpm -qa | grep conda
  anaconda-gui-33.16.5.6-1.el8.x86_64
  conda-4.10.3-1.el8.noarch
  libreport-anaconda-2.9.5-15.el8.x86_64
  python3-conda-4.10.3-1.el8.noarch
  anaconda-core-33.16.5.6-1.el8.x86_64
  anaconda-widgets-33.16.5.6-1.el8.x86_64
  anaconda-tui-33.16.5.6-1.el8.x86_64
  anaconda-user-help-8.3.3-1.el8.centos.noarch
  python3-conda-package-handling-1.7.3-2.el8.x86_64
  

• ~/.bashrc内容

  $ cat  .bashrc
  # .bashrc
  
  # Source global definitions
  if [ -f /etc/bashrc ]; then
          . /etc/bashrc
  fi
  
  # User specific environment
  PATH="$HOME/.local/bin:$HOME/bin:$PATH"
  export PATH
  
  
  
  export PS1='\u@\h \w \n\$ '
  
  # Uncomment the following line if you don't like systemctl's auto-paging feature:
  # export SYSTEMD_PAGER=
  
  
  # User specific aliases and functions
  
  #load mpich
  #module load /usr/share/modulefiles/mpi/mpich-x86_64
  
  #load openmpi env
  module load /usr/share/modulefiles/mpi/openmpi-x86_64
  
  #load nvhpc env
  #module load /opt/nvidia/hpc_sdk/modulefiles/nvhpc/22.2
  
  #export KOKKOS_PATH=~/local.opt
  
  export PATH=$PATH:/usr/local/cuda/bin
  export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib64
  
  # >>> conda initialize >>>
  # !! Contents within this block are managed by 'conda init' !!
  __conda_setup="$('/usr/bin/conda' 'shell.bash' 'hook' 2> /dev/null)"
  if [ $? -eq 0 ]; then
      eval "$__conda_setup"
  else
      if [ -f "/usr/etc/profile.d/conda.sh" ]; then
          . "/usr/etc/profile.d/conda.sh"
      else
          export PATH="/usr/bin:$PATH"
      fi
  fi
  unset __conda_setup
  # <<< conda initialize <<<
  

• 其他版本gcc-toolset

  $ yum search gcc-toolset
  上次元数据过期检查：1:03:17 前，执行于 2023年03月07日 星期二 13时12分30秒。
  ============================================================== 名称 和 概况 匹配：gcc-toolset ==============================================================
  gcc-toolset-10.x86_64 : Package that installs gcc-toolset-10
  gcc-toolset-10-runtime.x86_64 : Package that handles gcc-toolset-10 Software Collection.
  gcc-toolset-11.x86_64 : Package that installs gcc-toolset-11
  gcc-toolset-11-runtime.x86_64 : Package that handles gcc-toolset-11 Software Collection.
  gcc-toolset-9.x86_64 : Package that installs gcc-toolset-9
  gcc-toolset-9-runtime.x86_64 : Package that handles gcc-toolset-9 Software Collection.
  ================================================================== 名称 匹配：gcc-toolset ==================================================================
  gcc-toolset-10-annobin.x86_64 : Annotate and examine compiled binary files
  gcc-toolset-10-binutils.x86_64 : A GNU collection of binary utilities
  gcc-toolset-10-binutils-devel.i686 : BFD and opcodes static and dynamic libraries and header files
  gcc-toolset-10-binutils-devel.x86_64 : BFD and opcodes static and dynamic libraries and header files
  gcc-toolset-10-build.x86_64 : Package shipping basic build configuration
  gcc-toolset-10-dwz.x86_64 : DWARF optimization and duplicate removal tool
  gcc-toolset-10-dyninst.x86_64 : An API for Run-time Code Generation
  gcc-toolset-10-dyninst-devel.x86_64 : Header files for compiling programs with Dyninst
  gcc-toolset-10-elfutils.x86_64 : A collection of utilities and DSOs to handle ELF files and DWARF data
  gcc-toolset-10-elfutils-debuginfod-client.i686 : Library and command line client for build-id HTTP ELF/DWARF server
  gcc-toolset-10-elfutils-debuginfod-client.x86_64 : Library and command line client for build-id HTTP ELF/DWARF server
  gcc-toolset-10-elfutils-debuginfod-client-devel.i686 : Libraries and headers to build debuginfod client applications
  gcc-toolset-10-elfutils-debuginfod-client-devel.x86_64 : Libraries and headers to build debuginfod client applications
  gcc-toolset-10-elfutils-devel.i686 : Development libraries to handle compiled objects
  ...
  

• NVHPC和CUDA开发安装包需要去NVIDIA官网下载安装，用module配置加载

---

网络资料

• OPENHPC搭建集群

AMGX