Publications

MNEMEE: Memory management technology for adaptive and efficient design of embedded systems

The theme of intelligent ubiquitous devices will dominate future embedded system designs and speed up the integration of multimedia and communication applications, thus creating very complex, dynamic source code. Today it is increasingly impossible for designers to map applications cost- efficiently to any platform without significant optimization of the initial source code. The MNEMEE project will address this key challenge with innovative source-to-source optimizations for data management. Novel features include: i) multi-objective explorations that allow trade-offs which designers highly need to rightly position their product in the huge search space, ii) a combination of design-time and run-time techniques to boost the cost-efficiency and performance, and iii) automated optimizations that are applied once to reduce design effort and can handle very complex code. These optimizations and automation support will result in reduced exploration design time by at least a factor of 2, decreased memory footprint and memory bandwidth requirements of 30%, and improved energy and power efficiency by a factor of 2. The project will enable customized solutions for programmable MPSoC platforms involving the necessary system trade-offs, while meeting all real-time specifications.