I’ve noticed that the video driver for OpenCL is different from the CUDA driver under linux.
The version numbers are different.
Does this mean that you can only run either one or the other on a particular host? That is,
either you can develop and run OpenCL code or CUDA code but going between the two would require
uninstalling the driver and then installing the other driver? If so will the OpenCL and CUDA capabilities eventually
be supported in just a single driver, maybe when OpenCL stabilized.