08-28-2017, 09:12 AM
I understand you, however how do you explain this? As far as I know opencl from nvidia was first broadcast in cuda and then into machine code. Thus logically it is impossible to achieve high performance. (opencl over cuda)