An Ant Colony metaheuristic for energy aware application mapping on NoCs

The increasing number and complexity of tasks running on embedded systems demands more and more processors. The traditional bus based communication architecture are not scalable for high performance applications requiring a large number of processors. In this sense NoC-based MPSoC have been proposed to address these limitations. One critical problem in NoC based MPSoC is the communication costs. This cost depends on the task mapping into the processors. This paper presents an approach that uses an Ant Colony meta-heuristic to assign a set of tasks to a set of processors in a NoC-based MPSoC. The proposed approach evaluates tasks mappings while optimizing the traffic and the power consumption in the NoC. The result of applying our approach in MPEG and image segmentation applications shows that our technique was able to find a task mapping that reduces the traffic and energy consumption. Compared with the worst case solutions, our approach was able to find a task allocation that reduces the traffic and also the energy consumption of about 42% and 69% for the MPEG; and 51% and 69% for the image segmentation.