Adaptive Framework for Automated Mapping and Architecture Trades for Embedded Heterogeneous Systems

摘要

This paper presents a framework that can be adapted to perform automated mapping/scheduling and architecture trades for embedded computing environments. The framework is based on a learning automata model whose adaptable nature makes it invaluable in tools that provide end-to-end solutions for embedded systems development. The framework has been incorporated into one such tool, Systems and Applications Genesis Environment (SAGE). The paper begins with an introduction to learning automata and the architecture of the framework. The following sections discuss how algorithms for automated mapping/scheduling and architecture trades are constructed within the framework.