Should reference models get the virtual interface handle of the DUT's interface?

UVM
,
1

Hi,
I have a doubt with respect to the reference model getting a handle of virtual interface, is this the correct procedure to adapt in our environments.
What impact will it show when we move levels up the hierarchy of verification from block level to integration level to subsystem level etc??

My understanding of reference model is, it is just a software model which is part of scoreboard and basically feeds on the packets generated by the monitor related to that block and gives us the expected output. It should be isolated from the virtual interfaces which will only be accessed by the DRIVERS to drive onto the interface and MONITORS to monitor the activity on the interface and to pack it into some transaction items which will inturn feed the testbench components.

Can any one correct me if I am wrong and also answer my query whether it is good to have virtual interface handle inside the refrence model and the impacts of having it in reference model??

Thanks,
Neith

2

Neith,

Ideally, you are correct. But there will always be some odd cases where it is not practical to create a monitor for every signal that your scoreboard requires. This could be to correct some severe performance issues, or in the volume of internal signals needed to keep the reference model accurate. But as soon as you do that you severely limit the re-usability of that scoreboard as the implementation changes, or as you go up levels of abstraction in your verification.

Any time you diverge from the recommend methodology, make sure you document why it is knowingly being done.