Nvidia Cuda安装指南

超、凢脫俗 2022-05-26 07:59 254阅读 0赞

## 1. 检查是否存在显卡信息, 检查nvidia ##

> $ lspci | grep -i nvidia

输出结果类似的情况如下：  
`02:00.0 3D controller: NVIDIA Corporation Device 1b38 (rev a1) 04:00.0 3D controller: NVIDIA Corporation Device 1b38 (rev a1) 83:00.0 3D controller: NVIDIA Corporation Device 1b38 (rev a1) 84:00.0 3D controller: NVIDIA Corporation Device 1b38 (rev a1)`

## 检查os的版本信息 ##

> uname -m && cat /etc/\*release

输出结果类似：

x86_64
    CentOS Linux release 7.2.1511 (Core)
    NAME="CentOS Linux"
    VERSION="7 (Core)"
    ID="centos"
    ID_LIKE="rhel fedora"
    VERSION_ID="7"
    PRETTY_NAME="CentOS Linux 7 (Core)"
    ANSI_COLOR="0;31"
    CPE_NAME="cpe:/o:centos:centos:7"
    HOME_URL="https://www.centos.org/"
    BUG_REPORT_URL="https://bugs.centos.org/"
    
    CENTOS_MANTISBT_PROJECT="CentOS-7"
    CENTOS_MANTISBT_PROJECT_VERSION="7"
    REDHAT_SUPPORT_PRODUCT="centos"
    REDHAT_SUPPORT_PRODUCT_VERSION="7"
    
    CentOS Linux release 7.2.1511 (Core)
    CentOS Linux release 7.2.1511 (Core)

## 检查是否存在gcc ##

> gcc –version

输出结果类似：

gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-4)
    Copyright (C) 2015 Free Software Foundation, Inc.
    This is free software; see the source for copying conditions.  There is NO
    warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

## 安装内核header信息 ##

> uname -r 获取os版本  
> sudo yum install kernel-devel-(*u**n**a**m**e*−*r*)*k**e**r**n**e**l*−*h**e**a**d**e**r**s*−  ( u n a m e − r ) k e r n e l − h e a d e r s − (uname -r) 安装header

## 下载cuda toolkit安装包 ##

> download cuda-repo-rhel7-8.0.61-1.x86\_64.rpm from the following url:  
> 这里的是由于需要下载8.0的版本，目前最新的是9.1的版本。

参考资料： [https://developer.nvidia.com/cuda-80-ga2-download-archive][https_developer.nvidia.com_cuda-80-ga2-download-archive]  
下载最新的cuda-toolkit ， 请参阅：  
[https://developer.nvidia.com/cuda-downloads][https_developer.nvidia.com_cuda-downloads]

然后安装，并更新cache

> sudo rpm -i cuda-repo-rhel7-8.0.61-1.x86\_64.rpm  
> sudo yum clean all

## 配置代理\[optional\] ##

> vi /etc/yum.conf

在文件中新增：

> proxy=[http://192.168.1.1:8888][http_192.168.1.1_8888]

之后执行该配置文件：

> source /etc/yum.conf  
> yum update

## 安装前置条件 ##

> yum install epel-release  
> yum install dkms

## 检查cuda是否在yum repository中 ##

> yum list | grep cuda

一般的输出结果为：

cuda-8-0.x86_64                           8.0.61-1                     @cuda    
    cuda-command-line-tools-8-0.x86_64        8.0.61-1                     @cuda    
    cuda-core-8-0.x86_64                      8.0.61-1                     @cuda    
    cuda-cublas-8-0.x86_64                    8.0.61.2-1                   @cuda    
    cuda-cublas-dev-8-0.x86_64                8.0.61.2-1                   @cuda    
    cuda-cudart-8-0.x86_64                    8.0.61-1                     @cuda    
    cuda-cudart-dev-8-0.x86_64                8.0.61-1                     @cuda    
    cuda-cufft-8-0.x86_64                     8.0.61-1                     @cuda    
    cuda-cufft-dev-8-0.x86_64                 8.0.61-1                     @cuda    
    cuda-curand-8-0.x86_64                    8.0.61-1                     @cuda    
    cuda-curand-dev-8-0.x86_64                8.0.61-1                     @cuda    
    cuda-cusolver-8-0.x86_64                  8.0.61-1                     @cuda    
    cuda-cusolver-dev-8-0.x86_64              8.0.61-1                     @cuda    
    cuda-cusparse-8-0.x86_64                  8.0.61-1                     @cuda    
    cuda-cusparse-dev-8-0.x86_64              8.0.61-1                     @cuda    
    cuda-demo-suite-8-0.x86_64                8.0.61-1                     @cuda    
    cuda-documentation-8-0.x86_64             8.0.61-1                     @cuda    
    cuda-driver-dev-8-0.x86_64                8.0.61-1                     @cuda    
    cuda-drivers.x86_64                       390.30-1                     @cuda    
    cuda-license-8-0.x86_64                   8.0.61-1                     @cuda    
    cuda-misc-headers-8-0.x86_64              8.0.61-1                     @cuda    
    cuda-npp-8-0.x86_64                       8.0.61-1                     @cuda    
    cuda-npp-dev-8-0.x86_64                   8.0.61-1                     @cuda    
    cuda-nvgraph-8-0.x86_64                   8.0.61-1                     @cuda    
    cuda-nvgraph-dev-8-0.x86_64               8.0.61-1                     @cuda    
    cuda-nvml-dev-8-0.x86_64                  8.0.61-1                     @cuda    
    cuda-nvrtc-8-0.x86_64                     8.0.61-1                     @cuda    
    cuda-nvrtc-dev-8-0.x86_64                 8.0.61-1                     @cuda    
    cuda-repo-rhel7.x86_64                    8.0.61-1                     installed
    cuda-runtime-8-0.x86_64                   8.0.61-1                     @cuda    
    cuda-samples-8-0.x86_64                   8.0.61-1                     @cuda    
    cuda-toolkit-8-0.x86_64                   8.0.61-1                     @cuda    
    cuda-visual-tools-8-0.x86_64              8.0.61-1                     @cuda 
    ..............

进行cuda安装，选取特定的版本：

> sudo yum install cuda-8-0.x86\_64

## 配置环境变量\[optional\] ##

> vi /etc/profile  
> 在文件中新增：  
> `export PATH=/usr/local/cuda-8.0/bin/:$PATH export LD_LIBRARY_PATH=/usr/local/cuda-9.1/lib64:$LD_LIBRARY_PATH`

生效配置信息：

> source /etc/profile

检查环境变量是否生效：

> echo $PATH

## 重启机器 \[optional\] ##

> reboot now

重启机器是为了让各种配置生效，比如docker之类。比如碰到了如下错误信息：

46c1799eafe28c759582c5a60ded455d484d375e9e378b7c62d82bf8cf802e54
    docker: Error response from daemon: driver failed programming external connectivity on endpoint adoring_yalow (e1f53253a8c0a5590d0767a54fd6be4c05a1351ca1033cad5bddd5c495f888e2): iptables failed: iptables --wait -t nat -A DOCKER -p tcp -d 0/0 --dport 9010 -j DNAT --to-destination 192.168.0.2:9010 ! -i docker0: iptables: No chain/target/match by that name.
     (exit status 1).

## 总结 ##

其中碰到了N多问题，不过这也是一个非常好的锻炼，发现问题，解决问题。

[https_developer.nvidia.com_cuda-80-ga2-download-archive]: https://developer.nvidia.com/cuda-80-ga2-download-archive
[https_developer.nvidia.com_cuda-downloads]: https://developer.nvidia.com/cuda-downloads
[http_192.168.1.1_8888]: http://192.168.1.1:8888