# HPL筆記 環境:RTX3060 ubuntu24.04 memory4GB 2cores 要關上 Nouveau 驅動 ## 安裝nvidia驅動 先安裝gcc和build-essential 查詢推薦的driver ``` ubuntu-drivers devices ``` 並以推薦的driver去做安裝這邊選用 ``` sudo apt install nvidia-driver-535 ``` ## 下載cuda 安裝cuda toolkit選用runfile可同時安裝cuda以及nvidia driver也可選定安裝特定檔案，通常只選用cuda不裝driver,driver另外裝 這邊版本選擇cuda toolkit 12.8 ``` wget https://developer.download.nvidia.com/compute/cuda/12.8.1/local_installers/cuda_12.8.1_570.124.06_linux.run sudo sh cuda_12.8.1_570.124.06_linux.run ``` 設置環境變數 ``` echo 'export PATH=/usr/local/cuda-12.8/bin${PATH:+:${PATH}}' >> ~/.bashrc echo 'export LD_LIBRARY_PATH=/usr/local/cuda-12.8/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}' >> ~/.bashrc source ~/.bashrc ``` 測試是否安裝成功 ``` nvcc --version ``` ## 下載 Linpack Link : https://developer.nvidia.com/rdp/assets/cuda-accelerated-linpack-linux64 從上面的連結，登入CUDA註冊開發者會員，下載linpack for Linux64版本，這裡下載到的版本為hpl-2.0_FERMI_v15.tgz。 ## 下載 Intel® oneAPI HPC Toolkit link : https://software.intel.com/en-us/qualify-for-free-software 按右上角搜尋 HPC Toolkit ![image](https://hackmd.io/_uploads/rkXWXTnkll.png) 選第一個 ![image](https://hackmd.io/_uploads/rJBG763Jgx.png) 選擇 `Linux` `Offline Installer`，點右下角的 `Continue as a Guest` ![image](https://hackmd.io/_uploads/BJfr7631el.png) 下載完成後 ``` sh intel-oneapi-hpc-toolkit-2025.1.0.666_offline.sh ``` ![image](https://hackmd.io/_uploads/HkP0YZAJxe.png) 接著設定環境變數 ``` /home/hpc/intel/oneapi source setvars.sh ``` ## 安裝 mpich2 ``` wget http://www.mpich.org/static/downloads/3.2.1/mpich-3.2.1.tar.gz ``` ![image](https://hackmd.io/_uploads/BkkANpnkle.png) ``` tar zxvf mpich-3.2.1.tar.gz cd mpich-3.2.1 ``` ![image](https://hackmd.io/_uploads/SJBrBT3kgg.png) ``` ./configure -prefix=/home/<username>/mpich --disable-fortran ``` ![image](https://hackmd.io/_uploads/SkaCFanJlg.png) ``` make make install ``` 打開環境變數檔案 ``` vim /etc/environment ``` 裡面加上 ``` PATH="/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/usr/local/cuda-9.2/bin:/home/<username>/mpich/bin" ``` ![image](https://hackmd.io/_uploads/SJRenThJle.png) ## 安裝openmpi 我下載最新版本 ``` wget https://download.open-mpi.org/release/open-mpi/v5.0/openmpi-5.0.7.tar.gz ``` 解壓縮他 ``` tar zxvf openmpi-5.0.7.tar.gz cd openmpi-5.0.7 ./configure --prefix=/opt/openmpi ``` 編譯 ``` make make install ``` 配置環境變數 ``` PATH="/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/usr/local/cuda-9.2/bin:/home/<username>/mpich/bin" ``` ## 更改ldconfig ``` cd /etc/ld.so.conf.d/ sudo vim hpl.conf ``` 輸入內容 ``` /usr/local/cuda-9.2/lib64 /lib /opt/intel/mkl/lib/intel64 /opt/intel/lib/intel64 /home/ubuntu/hpl/src/cuda ``` - 最後一行的ubuntu改成自己的user name 執行並檢測 ``` sudo ldconfig sudo ldconfig -v | grep cuda ``` ## 開始編譯Linpack benchmark for CUDA 建立 hpl 資料夾 ``` mkdir -p ~/hpl tar -xvf hpl-2.0_FERMI_v15.tgz -C ~/hpl ``` ![image](https://hackmd.io/_uploads/ByWvCp2ygx.png) ``` cd ~/hpl/hpl-2.0_FERMI_v15 ls ``` ![image](https://hackmd.io/_uploads/ry4a062yle.png) 編輯 Make.CUDA ![image](https://hackmd.io/_uploads/Bko_0Z0Jxl.png) ``` make arch=CUDA ``` 修改~/hpl/bin/CUDA/run_linpack中的HPL_DIR為你hpl的路徑 ``` HPL_DIR=/home/hpc/hpl/hpl-2.0_FERMI_v15 ```