Non-Stack Single-Pass Simulation

Abstract

The previous chapter dealt with stack-based single-pass simulation. Stack-based single-pass simulation permits the simulation of a range of cache configurations in a time and space efficient manner. All stack-based simulation algorithms maintain multiple caches in a stack, exploiting inclusion properties between caches. During simulation they do a sequential search down the stack examining, modifying and moving entries as appropriate. Stack-based single-pass simulation is elegant and efficient relative to performing the simulations one at a time. However, taking a step back we see that the essential idea exploited in stack-based single-pass simulation is one of reducing simulation effort by simulating multiple configurations together and exploiting relations between the configurations to reduce simulation effort. This idea may be exploited to develop efficient single-pass algorithms in situations where stack simulation is not applicable. Even in situations where stack simulation is applicable non-stack single-pass simulation algorithms can be more efficient by avoiding the sequential search of the stack. In this chapter we discuss single-pass simulation algorithms that are not stack-based.