An overhead reducing technique for Time Warp

In this paper, we introduce a technique to reduce the number of state savings and the event queue size of Time Warp. By reducing the state saving and the sizes of event queues, we can decrease the overhead and the maximum memory requirement in Time Warp. We exploit the look-ahead technique to get a lower bound time stamp of the next event and to determine if an event is safe to be executed. No State saving is carried out when the event execution is safe. This lower bound can be used to discard saved states even though the time stamps are greater than the global virtual time (GVT). We prove that the proposed technique is correct under both aggressive and lazy cancellation schemes. This technique can be implemented with minimal additional overhead. Benchmark results on logic simulation show that the mechanism can reduce the number of state savings and memory size requirements significantly.