Approximate Simulation

5.6 Summary

This chapter presented two approximation solutions for symbolic simulation. Ideally, both cycle-based symbolic simulation and quasi-symbolic simulation improve scalability and performance of simulation by running an approximation of the state space and exploiting the symbolic information within. CBSS solves this problem using reparametrization, while quasi-symbolic simulation has an automatic and adaptive detection technique to select the symbols to approximate.

CBSS has shown to improve the scalability of symbolic simulation by providing a quick and memory-friendly parametrization technique for the state equations. It can find quickly a large subset of the frontier set which can be represented with great efficiency. The experimental results show that, in most cases, we can achieve 10–20 orders of magnitude (or more) better efficiency over a compiled logic simulator. However, in some cases we notice that many variables in the support of the state vector are complex, or shared, and need to be evaluated to a constant value. In those situations the performance is no longer competitive with logic simulation and the breadth of the state exploration is limited. In order to improve on the quality of the parametrization, we need to explore better techniques to represent the state vector through parameters. To this end, the next chapter introduces alternative exact parametrization techniques.