Presentation of Master's thesis

Mapping and Scheduling of MultiKPN Applications - A Case Study:
Multi-processor systems are increasingly becoming common these days, both in general purpose computers and embedded systems. Applications need to be parallelized to take advantage of such systems. One common way to achieve this, especially in embedded systems, is to model the application as a data flow model. The thesis concentrates on applications modeled as Kahn Process Networks (KPN), a flexible data flow model.
The number of applications executed on an embedded system is increasing. This raises a need to analyze, for performance and constraints, the different combinations of applications running on the system. Static analysis for different combinations of applications in such an analysis can quickly become infeasible because of state-space explosion. Moreover, a KPN application cannot be analyzed statically. The composability analysis thus requires a different execution-based approach.
The presentation will describe the contribution of the thesis towards single KPN and multi-KPN analysis. First, it will cover the two data flow models, a Synchronous Data Flow (SDF) graph, presented by E. A. Lee (87), and a Kahn Process Network (KPN), presented by G. Kahn (71). Then, it will describe the related work done on scheduling the data flow models and their composability analysis. It will then present the contribution of the thesis towards scheduling of a single KPN and composability analysis of multi-KPN applications. The presentation will conclude by presenting some results.

Speaker: Aamer Shah, Institute for Integrated Signal Processing Systems, RWHT Aachen

Date: Monday, September 13, 2010

Time: 10:00

Location: Lecture Room 001, German Research School for Simulation Sciences, Schinkelstr. 2a, 52062 Aachen