Is it possible to compile the NVIDIA kernel module on a GNU/Linux machine without GPUs?
The reason why I need it is the following:
I manage a small HPC cluster where compute nodes are diskless machines and mount their root filesystem from a master node.
The compute nodes are equipped with NVIDIA K80 GPUs, while the master node does not have any GPU at all.
Every time a new NVIDIA driver is available I would like to be able to compile a new kernel module for the current kernel on the master node, so then all the compute nodes can load it through NFS.
Thus, is there a way to force the installer to just compile the module without trying to load it end / or to check if a GPU is installed or not?
From the documentation I was not able to find any flags/args/etc…
Thanks in advance.