Hi
I have a question regarding the working of the prediction model. I want to actually check to see if my understanding of the concept is correct! I think the prediction model has to work independently from the clock. It actually receives the input transactions and creates the output transactions, one at a time. Actually, it has the ability to receive lots of transactions and produce the corresponding output transactions in zero simulation time. But because there is a handshaking mechanism between the UVM sequence and the DUT, the transactions are actually produced by the sequence at a rate suitable for the DUT, and this causes an implicit dependence of the rate of transaction production on the DUT clock. Therefore, although the predictor is capable to process an infinite number of transactions during zero simulation time, this will not actually happen, and the predictor will be forced to work at the same rate as the DUT.
Is there some problem with my understanding of the concept?
Thank you!
In reply to Farhad:
What you say is correct for most DUTs, but not all. Sometimes prediction models are created from an algorithmic point of view, like encryption/decryption, and you have to wait for all transactions to complete before you can make a prediction.