I’m looking for a way to associate the numeric cuda device identifier with a physical device. I’m working on writing diagnostics for GPU using CUDA. It really does not help to let the user know that device 0 is not working as one will not be able to map it a physical device. This is a major problem when there are multiple GPUs in a system.
Can you please help me relate the numeric cuda device identifier to either a unique windows device identifier or to the unique PCI bus and slot location.