Search

The article focuses on stack simulation. Trace-driven simulation is a standard process for analyzing memory system performance. Given a sequence of memory addresses, trace-driven simulation is often used to produce performance metrics such as miss and/or write-back ratios. These results can then be incorporated into a system model to predict system performance under different memory design, choices. The stack algorithm works as long as the replacement policy, such as the LRU (least recently used) replacement policy, guarantees the inclusion property, that is, the contents of any size cache is a superset of those in a smaller cache. Trace-driven simulation takes a sequence of addresses as input. In many cases the addresses are either virtual addresses generated from a software simulator, or real addresses collected from a piece of hardware that monitors bus addresses.