I am unable to find if it is possible to measure for how many clock cycles has a given CUDA context been running on the GPU. Does the GPU context scheduler keep track of this?
Usually operating systems would compute total CPU time for each thread/context. I want something equivalent for the GPU.