Monitoring GPU core usage by CUDA on Linux

Hello guys.

I have a proplem when monitoring GPU core usage on CentOS 6.2
I searched for hours, and I knew that NVAPI and CUDA-Z would not be my solution.
(NVAPI just for windows and CUDA-Z seems no info I want)

Is it possible to monitoring GPU core usage by CUDA, such as API or something else?
Or any other solution?

Thank a lot.


What exactly are you trying to monitor?
I have a system with a GPU only for CUDA programming, the only thing I actually monitor is it’s core temperature (if there’s something running, the temperature can go really up), I don’t know if you can actually monitor more advanced stuff, withouth actually using the GPU (which would be used for what you are executing already).
In short, what do you want to monitor?