Multiprocessor systems are rapidly entering various high-performance computing segments, like multimedia processing. Instead of an increase in processor clock frequency, the new trend is enabling multiple cores in performing processing, e.g. dual or quadrapule CPUs in one subsystem. In this contribution, we address the problem of modeling the resource requirements of multimedia applications for a distributed computation on a multiprocessor system. This paper shows that the estimation of resource requirements based on input data enables the dynamic activation of tasks and run-time redistribution of application tasks. We also formally specify the optimal selection of the co-executed application with aim to provide the most optimal end-results of such streaming applications within one networks-on-chip (NoC) system. We present a new concept for system optimization which involves the major system parameters and resource usage. Experimental results are based on mapping an arbitrary-shaped MPEG-4 video decoder onto a multiprocessor NoC.
|Title of host publication||Proceedings of 27th symposium on Information Theory in the Benelux, June 8-9, 2006, Noordwijk, The Netherlands|
|Editors||R. Lagendijk, Jos Weber, H., A. Berg, van den, F.|
|Place of Publication||Eefde|
|Publisher||Werkgemeenschap voor Informatie- en Communicatietheorie (WIC)|
|Publication status||Published - 2006|