I'm having some trouble solving a issue in my code, I hope you can help me.
I have two modules, A and B. Module A does requests to B, and after a number of cycles B sends a multi-cycle response to A. A can hold up to 8 requests waiting to be responded, and the responses from B don't necessarily come back ordered. That's why we use an ID, to identify the returning data.
To verify this behaviour, I have a scoreboard with several checkers. One of the checkings that I do is if the ID used for a request is free or not. To do that, I keep an associative array with the IDs pending to be responded, and I insert, check and delete items as needed. I control this from two interfaces and monitors, one for the requests and another one for the responses. The response monitor, being the responses more than one cycle long, waits until it has all the data to send a transaction to the scoreboard, where I update my structs.
The problem comes the moment that once that A sees that it is actually getting a valid response from B, frees the ID and can use it for a new request. That is happening in some of my simulations, and since I won't receive the transaction until all the response is complete, block A is doing a new request with an ID I won't know it's legit to use until I get the complete transaction from the monitor.
Any ideas on how to solve this? Thanks!
In the cycle that you see a response from B, why don't you move the request from A into another associative array, one that represents responses that have been initiated. That way you'll have a free slot in original array to handle new requests from A, but now you'll have the new, second array to handle the multi-cycle responses that have already begun.