I have correctly started and segregated applications using uvmm but this approach leads to a noticeable performance prediction.
I say this because by running for cycles on the ‘NOP’ instruction, the times obtained with uvmm are increased by 100 times.