Linux下编译安装SuperLU

SuperLU用于求解大规模稀疏线性方程组，本文记录在远程Linux服务器下编译安装SuperLU的流程。

一、配置VS Code

2.1 安装VS Code Extensions

在本地打开VS Code, 安装以下扩展插件，

2.2 Remote SSH

通过"Remote Development"扩展插件登录远程Linux服务器，并安装以下插件

二、安装与配置oneAPI

1.1 配置包管理器

wget -O- https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB \
| gpg --dearmor | sudo tee /usr/share/keyrings/oneapi-archive-keyring.gpg > /dev/null

echo "deb [signed-by=/usr/share/keyrings/oneapi-archive-keyring.gpg] https://apt.repos.intel.com/oneapi all main" | sudo tee /etc/apt/sources.list.d/oneAPI.list

sudo apt update

1.2 安装

sudo apt -y install cmake pkg-config build-essential

1.3 安装oneAPI

sudo apt install intel-basekit
sudo apt install intel-hpckit

1.4 配置CMake Kits

打开VS Code， 执行"CMake: Edit User-Local CMake Kits"，编辑“~/.local/share/CMakeTools/cmake-tools-kits.json”文件，

[
  {
    "name": "GCC 11.4.0 x86_64-linux-gnu",
    "compilers": {
      "C": "/usr/bin/gcc",
      "CXX": "/usr/bin/g++"
    },
    "isTrusted": true
  },
  {
    "name": "Intel(R) oneAPI DPC++/C++ Compiler 2024.2.1",
    "compilers": {
      "C": "/opt/intel/oneapi/compiler/2024.2/bin/icx",
      "CXX": "/opt/intel/oneapi/compiler/2024.2/bin/icpx"
    },
    "isTrusted": true
  }
]

1.5 oneAPI环境变量配置

编辑"~/.bashrc"文件，添加以下内容，

source /opt/intel/oneapi/setvars.sh --force

然后重启远程服务器，在本地重新通过"VS Code"连接远程Linux服务器。

三、编译安装SuperLU

3.1 下载SuperLU

git https://github.com/xiaoyeli/superlu.git
cd ./superlu
git checkout v7.0.0

3.2 编译安装SuperLU

可以使用CMake Kits或者CMake Presets来构建SuperLU，

方法1： 使用CMake Kits

选择"Intel(R) oneAPI DPC++/C++ Compiler 2024.2.1", 使用在VS Code中使用CMake Tools完成构建，同时做如下配置修改，

方法2： 使用CMake Presets

在superlu目录下，编辑CMakeUserPresets.json文件，

{
    "version": 4,
    "configurePresets": [
        {
            "name": "linux_default",
            "hidden": true,
            "displayName": "Default Linux Config",
            "description": "Default build using Unix Makefiles generator",
            "generator": "Unix Makefiles",
            "binaryDir": "${sourceDir}/out/build/${presetName}",
            "cacheVariables": {
                "CMAKE_INSTALL_PREFIX": "${sourceDir}/out/install/${presetName}",
                "CMAKE_BUILD_TYPE": "Debug"
            }
        },
        {
            "name": "linux_gcc",
            "inherits": "linux_default",
            "displayName": "GCC 11.4.0 x86_64-linux-gnu",
            "description": "Using compilers: C = /usr/bin/gcc, CXX = /usr/bin/g++",
            "cacheVariables": {
                "CMAKE_C_COMPILER": "/usr/bin/gcc",
                "CMAKE_CXX_COMPILER": "/usr/bin/g++",
                "CMAKE_MAKE_PROGRAM": "/usr/bin/make"
            }
        },
        {
            "name": "linux_intel",
            "inherits": "linux_default",
            "displayName": "Intel(R) oneAPI DPC++/C++ Compiler 2024.0.2",
            "description": "Using compilers: C = /opt/intel/oneapi/compiler/latest/bin/icx, CXX = /opt/intel/oneapi/compiler/latest/bin/icpx",
            "cacheVariables": {
                "CMAKE_C_COMPILER": "/opt/intel/oneapi/compiler/latest/bin/icx",
                "CMAKE_CXX_COMPILER": "/opt/intel/oneapi/compiler/latest/bin/icpx"
            }
        }
    ],
    "buildPresets": [
        {
            "name": "linux_gcc",
            "description": "",
            "displayName": "",
            "configurePreset": "linux_gcc"
        },
        {
            "name": "linux_intel",
            "description": "",
            "displayName": "",
            "configurePreset": "linux_intel"
        }
    ]
}

然后在VS Code CMake Tools中，修改CMake Cache,

3.3 检查SuperLU连接引用

检查编译生成的libsuperlu.so是否正常链接到了oneAPI MKL，

ldd /data/3rdparty/install/intel-opt-dto/lib/libsuperlu.so
        linux-vdso.so.1 (0x00007ffc24be1000)
        libmkl_intel_lp64.so.2 => /opt/intel/oneapi/mkl/2024.2/lib/libmkl_intel_lp64.so.2 (0x00007dd914c00000)
        libmkl_intel_thread.so.2 => /opt/intel/oneapi/mkl/2024.2/lib/libmkl_intel_thread.so.2 (0x00007dd912800000)
        libmkl_core.so.2 => /opt/intel/oneapi/mkl/2024.2/lib/libmkl_core.so.2 (0x00007dd90e800000)
        libiomp5.so => /opt/intel/oneapi/compiler/2024.2/lib/libiomp5.so (0x00007dd90e200000)
        libimf.so => /opt/intel/oneapi/compiler/2024.2/lib/libimf.so (0x00007dd90dc00000)
        libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007dd915b88000)
        libsvml.so => /opt/intel/oneapi/compiler/2024.2/lib/libsvml.so (0x00007dd90c400000)
        libirng.so => /opt/intel/oneapi/compiler/2024.2/lib/libirng.so (0x00007dd914b07000)
        libintlc.so.5 => /opt/intel/oneapi/compiler/2024.2/lib/libintlc.so.5 (0x00007dd915b25000)
        libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007dd90c000000)
        libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007dd915b20000)
        libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007dd915b1b000)
        librt.so.1 => /lib/x86_64-linux-gnu/librt.so.1 (0x00007dd915b16000)
        /lib64/ld-linux-x86-64.so.2 (0x00007dd915d2e000)

可以看到，SuperLU实际上是链接到了LP64接口的MKL。若想SuperLU链接到ILP64接口的MKL，需要做如下CMake配置，

四、FAQs

Q. 在主程序中同时引用oneMKL、SuperLU，运行时调用oneMKL API报错。

A. 这个可能是由于libmkl_intel_lp64.so/libmkl_intel_ilp64.so混合调用所致。可考虑将CMake Cache “MKL_INTERFACE”设置为"lp64"，这样会在主程序中使用"LP64"接口的MKL。

网络资料

Intel® oneAPI Toolkits Installation Guideshttps://www.intel.com/content/www/us/en/developer/articles/guide/installation-guide-for-oneapi-toolkits.html

Get Started with the Intel® oneAPI Base Toolkit for Linux*https://www.intel.com/content/www/us/en/docs/oneapi-base-toolkit/get-started-guide-linux/2024-0/overview.html

Get Started with the Intel® HPC Toolkit for Linux*https://www.intel.com/content/www/us/en/docs/oneapi-hpc-toolkit/get-started-guide-linux/2024-0/overview.html

Using Visual Studio Code with Intel® oneAPI Toolkits User Guidehttps://www.intel.com/content/www/us/en/docs/oneapi/user-guide-vs-code/2024-2/overview.html

How to use CMake with Intel® oneAPI Toolkithttps://www.intel.com/content/www/us/en/developer/articles/technical/how-to-use-cmake-with-intel-oneapi-toolkits.html?wapkw=cmake

SuperLU: Home Page (nersc.gov)https://portal.nersc.gov/project/sparse/superlu/#superluUsing the ILP64 Interface vs. LP64 Interfacehttps://www.intel.com/content/www/us/en/docs/onemkl/developer-guide-windows/2023-0/using-the-ilp64-interface-vs-lp64-interface.html