Ready valid channels were not issuing reverse tokens (ready tokens from the enq port) irrevocably. The target backpressure seen by the enqueuer would vary as function of the host-time behavior of the dequeuer.
This PR fixes this behavior in single-clock ready-valid channels. I don't have time at the moment to fix it properly in channels that model CDC -- i think those should be rewritten from the ground up, and the cost of getting it wrong is too high.
Ready valid channels were not issuing reverse tokens (ready tokens from the enq port) irrevocably. The target backpressure seen by the enqueuer would vary as function of the host-time behavior of the dequeuer.
This PR fixes this behavior in single-clock ready-valid channels. I don't have time at the moment to fix it properly in channels that model CDC -- i think those should be rewritten from the ground up, and the cost of getting it wrong is too high.