linux 配置cuda、cudnn环境

1、安装cuda

https://developer.nvidia.com/zh-cn/cuda-toolkit

以11.2为例，选择版本后，进行下载安装：

wget https://developer.download.nvidia.com/compute/cuda/11.5.0/local_installers/cuda_11.5.0_495.29.05_linux_sbsa.run

sudo sh cuda_11.5.0_495.29.05_linux_sbsa.run

然后配置环境变量：

export PATH=/usr/local/cuda/bin:$PATH
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib64
export CUDA_HOME=$CUDA_HOME:/usr/local/cuda

不然使用cmake编译会报CMAKE_CUDA_COMPILER找不到的错误

nvcc -V检查安装是否成功

2、下载cudnn

下载地址：https://developer.nvidia.com/rdp/cudnn-download

这里使用8.1.1

tar zxvf cudnn-11.2-linux-aarch64sbsa-v8.1.1.33.tgz

sudo cp cuda/include/* /usr/local/cudn/include/
sudo cp cuda/lib64/libcudnn* /usr/local/cuda/lib64/
sudo chmod a+r /usr/local/include/cudnn.h
sudo chmod a+r /usr/local/cuda/lib64/libcudnn*

完成后，通过如下命令查看安装情况，

cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2

3、下载nccl

https://developer.nvidia.com/nccl

例如下载nccl-<version>.txz

tar xvf nccl-<version>.txz 

cp -r nccl-<version> /usr/local/

cp nccl-<version>/include/* /usr/local/include

或者

sudo rpm -i nccl-repo-<version>.rpm

sudo yum update

sudo yum install libnccl libnccl-devel libnccl-static