At the University of Amsterdam, we are currently targeting at

developing system level models and algorithms in the design, analysis

and performance evaluation of platform-based heterogeneous

architectures for stream processing embedded systems applications.

Such applications, although at their infancy today, are gaining more

popularity and are expected to dominate the media market in the near

future (within a 10 years time). The characteristics of the streaming

applications include data parallelism to the highest extent, minimal

usage of global memory, and high computation demands. To meet these

high requirements, future embedded systems probably will consist of

system-on-chips that are integrating ten to several hundreds of,

possible heterogeneous, computing elements connected by an on-chip

scalable (point-to-point) network. The design and analysis of these

future embedded systems are not yet fully supported by today's system

level tools and methodologies which are aimed towards current

applications and chip technology. Currently, we strive for addressing

this shortcoming i) by formal models conceptually (theory), ii) by

practical methods computationally (algorithms), and iii) by

architecture exploration structurally (hardware/software

partitioning).



There is a very closely related project at the Center for Advanced

Design Technologies of Koc University in Istanbul on real-time

applications multiprocessor systems such as streaming multimedia. The

focus is on power consumption and minimization using stochastic

application models and intelligent energy management schemes which can

exploit the information captured by such models. They explore

algorithmic solutions that will be implemented in hardware and

software both at the application and operating system levels.



We want to explore whether these two teams, both working in the system

level performance/power evaluation of stream processing embedded

systems applications, can exchange/share information and cooperate in

terms of models, tools and methodologies. To be more precies, for

example, we want to know whether we can use their stochastic

application models in our modeling and simulation framework in order

to generate realistic application workloads. On the other hand, Koc

University may want to use our architecture library models to verify

their analytical power estimation models in terms of simulation. We

strongly believe that both teams have closely related research

interests and will benefit from a possible cooperation to a great

extent.

To realize a possible cooperation, we need funding for short 2-3 day

visits for exchanging ideas and bringing relevant research topics into

discussion. We plan to make 2 visits within the next 3 months for

which we ask a total of 1800 Euros seed funding.

Serdar Tasiran, Koc University, Istanbul TR

Alper Demir, Koc University, Istanbul TR