CentOS 8 Advanced Install¶
Before starting, follow the steps in Generic SeaRay Installation.
Warning
Installing video drivers on Linux can sometimes cause you to lose your display. Recovery is often difficult. If you cannot afford for this to happen you should take steps to backup your system.
Support for NVIDIA Graphics¶
Prepare for EPEL:
- For CentOS, type
sudo dnf config-manager --set-enabled PowerTools - For RHEL, type
ARCH=$( /bin/arch )followed bysudo subscription-manager repos --enable "codeready-builder-for-rhel-8-${ARCH}-rpms"
- For CentOS, type
Go to EPEL and install. As of this writing there is a link,
epel-release-latest-8, that runs a graphical installer.Go to RPM Fusion and install the
nonfreerepository for RHEL 8 or compatible (there is no charge,nonfreerefers to license restrictions). The link runs a graphical installer.Type
sudo dnf install akmod-nvidia- This automatic kernel module recompiles automatically when a new Linux kernel is installed (e.g. during a system update). After restarting you must allow extra time for the kernel module to compile. There could be a long delay before the login screen appears.
Restart the system, allow extra time for this restart.
sudo dnf install xorg-x11-drv-nvidia-cudasudo dnf install ocl-icd-develActivate the conda environment (if not already active)
conda install -c conda-forge ocl-icd-system
Support for AMD Graphics¶
Download the driver for your GPU from the AMD website.
Follow the instructions from AMD to install the driver with OpenCL.
- As of this writing, the link is here. You must specify a command line option
--opencl=palor--opencl=legacydepending on your GPU. You must also use theamdgpu-pro-installvariant of the install script.Activate the conda environment (if not already active)
conda install -c conda-forge ocl-icd-system
Advanced 3D Plotting¶
The SeaRay plotter supports matplotlib and/or mayavi for 3d plotting. The 3D capabilities of matplotlib are at present nonideal (e.g., depth is not properly rendered in all cases). If you want robust 3D plots you should install mayavi.
In some cases mayavi and matplotlib step on each other. If this happens you may need separate environments for each. The plotter is written to sense which library is available and react accordingly.
- Activate your environment.
conda install -c conda-forge mayavi
Interactive Notebooks¶
If your environment is not already activated, activate it as above.
conda install jupyter ipymplCreate a directory
~/.jupyter/custom/and copyraysroot/extras/custom.cssto the new directory.If there are problems with Jupyter notebooks any or all of the following may be tried:
conda install widgetsnbextension=n, wherenis some older version.conda install ipywidgetsjupyter nbextension install --py --sys-prefix widgetsnbextensionjupyter nbextension enable --py --sys-prefix widgetsnbextension