OS evaluation for real time image processing on a multicore architecture

Abstract

Traditionally embedded applications were limited to dedicated hardware with limited functionalities controlled by a microcontroller. With the advancement of technology and the rising needs embedded applications now run on Consumer off the shelf (COTS) products. Using general-purpose operating systems makes the development cycle for these applications shorter by providing necessary hardware support and thus allowing the application to be developed on a higher abstraction. In this thesis, we develop a model to evaluate different operating systems for such an embedded application. The goal was to be able to predict the application performance on different operating systems without having to port or code the application for each operating system. The model uses micro-benchmarks on operating systems and then consolidates the results to give an overall score. The model thus developed was used to test and evaluate operating systems for a medical image processing application used in cardio vascular intervention procedures.