This question is dealt with multiple forums. However, I did not get clear picture as people have conflicting opinions.So I am asking it again.
Suppose my current kernel is not using GPU resources completely, say only 10 threads are running at a time. Hence to harness the full computational power can I run another independent kernel along with the current kernel on GPUs?
My guess is that it is not possible. Somewhere in AMD's GPU literature I read that if I try to run two kernels simultaneously they will automatically be serialized. However in OpenCL there is an option to create multiple independent command queues which dont require any synchronization. So if I put first kernel in the first queue and the second in the another would it make both of them running simultaneously?
I have Radeon 5450 graphics card just if anyone is curious.