矩池云上安装OpenCL教程
本教程租用的是K80,Pytorch1.11 镜像。
添加nvidia-cuda的阿里源
curl -fsSL https://mirrors.aliyun.com/nvidia-cuda/ubuntu1804/x86_64/3bf863cc.pub | apt-key add -
echo "deb https://mirrors.aliyun.com/nvidia-cuda/ubuntu1804/x86_64/ /" > /etc/apt/sources.list.d/cuda.list
apt update
安装 opencl
apt install ocl-icd-opencl-dev -y
创建链接
mkdir -p /etc/OpenCL/vendors/
cd /etc/OpenCL/vendors
vim nvidia.icd
在 nvidia.icd 里填入以下链接
/usr/lib/x86_64-linux-gnu/libnvidia-opencl.so.1
验证是否正确链接
cat nvidia.icd
ll /usr/lib/x86_64-linux-gnu/libnvidia-opencl.so.1
测试opencl
clGetPlatformIDs.c 内容
#include <stdio.h>
#include <CL/opencl.h>
int main(int argc, char **argv)
{
int status;
cl_uint n_platform;
status = clGetPlatformIDs(0, NULL, &n_platform);
if(status != CL_SUCCESS)
{
fprintf(stderr, "no platforms %d\n", status);
return -1;
} else
{
fprintf(stderr, "number of CL platforms %d\n", n_platform);
return 0;
}
}
进入项目目录,我这里是/mnt/testme/opencl_t
,新建一个clGetPlatformIDs.c
文件,然后写入上面内容。
cd /mnt/testme/opencl_t
vim clGetPlatformIDs.c
gcc clGetPlatformIDs.c -lOpenCL -o main
chmod +x main
./main
出现number of CL platforms 1,安装及测试就完成了