Author

Ethirajan

Date of Award

4-1997

Degree Name

Master of Science

Department

Electrical and Computer Engineering

First Advisor

Dr. Sharon Hu

Second Advisor

Dr. Garrison Greenwood

Access Setting

Masters Thesis-Open Access

Abstract

We present our approach to hardware/software partitioning for embedded systems based upon Evolutionary Algorithms. We have implemented it in a CAD tool, EvoC - Evolutionary Codesign which does automatic hardware/software partitioning of real-time embedded systems at the system level. Our objective is to find good design configurations that are tuned towards user's preferences. We are able to explore a large, often intractable design space using Evolutionary Algorithms while evaluating solutions having multiple and sometimes conflicting attributes in the light of Multi-Attribute Utility Theory. EvoC provides a generic format for specifying a wide variety of design problems and the implementation assumes no target architecture. A multiple bus and shared memory communication scheme has been incorporated into EvoC which analyzes the behavior and produces connected systems. Two design examples are given to illustrate the capability of our tool. Two of the factors which caused the high execution time of EvoC were identified and the appropriate corrective measures taken are discussed.

Share

COinS