Energy and Performance Model of DME

Abstract

As a part of MOSART, an international EU project, an efficient and high-level framework is needed to help rapid and accurate estimation of energy and performance for McNoc (Multi-core Network on Chip). This thesis is about characterization of the Data Management Engine (DME) which is triggered by the Leon3 processor. In this thesis, a power model was established and used to characterize the micro-functions in the Data Management Engine (DME) to be able to get a quick estimation of energy and performance without having to run slow gate-level simulation and slow power extraction. The result was validated by being tested on a FFT program.