进入WSL2

wsl -d Ubuntu-22.04

$ nvidia-smi Sun Jun2808:40:582026+-----------------------------------------------------------------------------------------+|NVIDIA-SMI610.43.02 KMD Version:610.47CUDA UMD Version:13.3|+-----------------------------------------+------------------------+----------------------+|GPU Name Persistence-M|Bus-Id Disp.A|Volatile Uncorr. ECC||Fan Temp Perf Pwr:Usage/Cap|Memory-Usage|GPU-Util Compute M.||||MIG M.||=========================================+========================+======================||0NVIDIA RTX PRO1000Blac... On|00000000:01:00.0 Off|N/A||N/A 50C P4 10W / 55W|0MiB / 8151MiB|0% Default||||N/A|+-----------------------------------------+------------------------+----------------------+ +-----------------------------------------------------------------------------------------+|Processes:||GPU GI CI PID Type Process name GPU Memory||ID ID Usage||=========================================================================================||No running processes found|+-----------------------------------------------------------------------------------------+

添加cuda-toolkit源

wgethttps://developer.download.nvidia.com/compute/cuda/repos/wsl-ubuntu/x86_64/cuda-keyring_1.1-1_all.debsudodpkg-icuda-keyring_1.1-1_all.debsudoapt-getupdate

查找源中支持哪些

(base)hugh@HUGH:~$aptlist-acuda-toolkit-12-4 Listing... Done cuda-toolkit-12-4/unknown12.4.1-1 amd64 cuda-toolkit-12-4/unknown12.4.0-1 amd64(base)hugh@HUGH:~$aptlist-acuda-toolkit-12-8 Listing... Done cuda-toolkit-12-8/unknown12.8.2-1 amd64 cuda-toolkit-12-8/unknown12.8.1-1 amd64 cuda-toolkit-12-8/unknown12.8.0-1 amd64(base)hugh@HUGH:~$aptlist-acuda-toolkit-13-1 Listing... Done cuda-toolkit-13-1/unknown13.1.2-1 amd64 cuda-toolkit-13-1/unknown13.1.1-1 amd64 cuda-toolkit-13-1/unknown13.1.0-1 amd64(base)hugh@HUGH:~$aptlist-acuda-toolkit-13-3 Listing... Done cuda-toolkit-13-3/unknown13.3.0-1 amd64

issue

Command ‘nvcc’ not found, but can be installed with: sudo apt install nvidia-cuda-toolkit

1. check dpkg

dpkg-l|grepcuda-toolkit-12-8 ii cuda-toolkit-12-812.8.2-1 amd64 CUDA Toolkit12.8meta-package ii cuda-toolkit-12-8-config-common12.8.90-1 all Common config packageforCUDA Toolkit12.8.

2. try run nvcc version

$exportPATH=/usr/local/cuda-12.8/bin:$PATH$exportLD_LIBRARY_PATH=/usr/local/cuda-12.8/lib64:$LD_LIBRARY_PATH(base)hugh@HUGH:~$ nvcc--versionnvcc: NVIDIA(R)Cuda compiler driver Copyright(c)2005-2025 NVIDIA Corporation Built on Fri_Feb_21_20:23:50_PST_2025 Cuda compilation tools, release12.8, V12.8.93 Build cuda_12.8.r12.8/compiler.35583870_0

3. add to environment

$echo'export PATH=/usr/local/cuda-12.8/bin:$PATH'>>~/.bashrc $echo'export LD_LIBRARY_PATH=/usr/local/cuda-12.8/lib64:$LD_LIBRARY_PATH'>>~/.bashrc $source~/.bashrc $ nvcc--versionnvcc: NVIDIA(R)Cuda compiler driver Copyright(c)2005-2025 NVIDIA Corporation Built on Fri_Feb_21_20:23:50_PST_2025 Cuda compilation tools, release12.8, V12.8.93 Build cuda_12.8.r12.8/compiler.35583870_0

尝试CUDA编程

编写hello.cu程序

#include<stdio.h>// 这是将在GPU上执行的函数// __global__ 关键字告诉编译器这是一个内核函数 (kernel)__global__voidhello_from_gpu(){printf("Hello World from GPU!\n");}intmain(){// 这是将在CPU上执行的代码printf("Hello World from CPU!\n");// 启动GPU内核// <<<1, 1>>> 表示启动1个线程块，每个线程块包含1个线程hello_from_gpu<<<1,1>>>();// 等待GPU执行完毕，这样才能在终端看到GPU的打印信息// 如果不加这一行，程序可能在GPU打印前就结束了[reference:5]cudaDeviceSynchronize();return0;}

编译

$ nvcc hello.cu-ohello nvcc warning:Supportforoffline compilationforarchitectures prior to'<compute/sm/lto>_75'will be removedina future release(Use -Wno-deprecated-gpu-targets to suppress warning).

执行第一个CUDA程序

$ ./hello Hello World from CPU!Hello World from GPU!

Reference

CUDA Toolkit 12.4 Downloads