Hierarchical Parallel Placement Using a Genetic Algorithm for Realizing Low Power Consumption

TL;DRAbstract

With portable information devices now widely disseminated, low power consumption LSIs are increasingly needed in battery, implementation, and heat generation applications. We propose parallel placement to realize low power consumption and confirm its effectiveness in experiments with a commercial electronic design automation (EDA) tool for designing LSIs. Our proposal hierarchically combines outline and detail placement based on genetic algorithm. In selection operators, new evaluation functions are introduced for realizing the reduction of power consumption focusing on the signal transition probability. Considering a parallel processing, in which a processing speed has a scalable relation with the number of processors, and by implementing it in a parallel computer, its effect is demonstrated.

Chat with Paper

AI Agents for this Paper