Back-tracing and event-driven techniques in high-level simulation with decision diagrams

Abstract

The paper addresses the problem of the cycle-based simulation performance of synchronous digital systems modeled by High-Level Decision Diagrams (DDs). A new class of DD representation, called Register-Oriented DDs (RODD) is introduced. The RODD model appears to be an efficient and compact representation of the system behavior for the high-level cycle simulation. In order to fully exploit the advantages of RODDs a new simulation algorithm, which is a combination of cycle-based forward event-driven and recursive back-tracing techniques is proposed. The characteristics of the simulation algorithms used to efficiently execute the evaluation of the DD network are discussed. Further the experimental results carried out on the real case examples demonstrating the gain in simulation performance of the proposed approach and a comparison of four cycle-based simulation algorithms are presented. Additionally, a comparison with the commercial event-driven and cycle-based HDL simulation tools is included.