A Machine Learning Approach for Optimizing Parallel Logic Simulation

摘要: Parallel discrete event simulation can be applied as a fast and cost effective approach for the gate level of current VLSI circuits. In this paper we combine dynamic load balancing algorithm bounded window optimistic simulation. The time prevents from being too excessive rollbacks. We utilize machine learning (Qlearning) to effect combination. introduce two load-balancing algorithms communication computational use agents these algorithms. One agent combines learns their corresponding parameters, while second optimizes value window. Experimental results show up 46% improvement in using combined several open source To best our knowledge, is first that Q-learning has been used optimize an