Difference between revisions of "NVIDIA GPU configuration"

From Colettapedia
Jump to navigation Jump to search
Line 14: Line 14:
 
# libglvnd
 
# libglvnd
 
#* sudo yum install libglvnd-devel
 
#* sudo yum install libglvnd-devel
 +
# Add cuda bin to path
 +
<pre>
 +
cat > /etc/profile.d/cuda.sh <<EOF
 +
# 2019-10-30 Coletta: Add cuda to the path for sh compatible users
 +
 +
if ! echo $PATH | grep -q /usr/local/cuda/bin ; then
 +
  export PATH=$PATH:/usr/local/cuda/bin
 +
fi
 +
EOF
 +
</pre>
 
# Install NVIDIA docker
 
# Install NVIDIA docker
 
#* [https://github.com/NVIDIA/nvidia-docker nvidia-docker]
 
#* [https://github.com/NVIDIA/nvidia-docker nvidia-docker]

Revision as of 19:16, 30 October 2019

Steps

  1. Download GPU and CUDA drivers
  2. Disable Nouveau
    • modprobe configuration files
    • Installer creates these files:
      • /usr/lib/modprobe.d/nvidia-installer-disable-nouveau.conf
      • /etc/modprobe.d/nvidia-installer-disable-nouveau.conf
  3. Kernel source Tree
    • yum install kernel-devel - install kernel source files on CentOS 7
  4. (Unnecessary) Install pkg-config/X SDK
  5. libglvnd
    • sudo yum install libglvnd-devel
  6. Add cuda bin to path
cat > /etc/profile.d/cuda.sh <<EOF
# 2019-10-30 Coletta: Add cuda to the path for sh compatible users

if ! echo $PATH | grep -q /usr/local/cuda/bin ; then
  export PATH=$PATH:/usr/local/cuda/bin
fi
EOF
  1. Install NVIDIA docker