Project Team Moais

Members
Overall Objectives
Scientific Foundations
Application Domains
Software
New Results
Contracts and Grants with Industry
Partnerships and Cooperations
Dissemination
Bibliography
PDF e-pub XML


Section: Application Domains

Embedded Systems

Participants : Jean-Louis Roch, Guillaume Huard, Denis Trystram, Vincent Danjean.

To improve the performance of current embedded systems, Multiprocessor System-on-Chip (MPSoC) offers many advantages, especially in terms of flexibility and low cost. Multimedia applications, such as video encoding, require more and more intensive computations. The system should be able to exploit the resources as much as possible to save power and time. This challenge may be addressed by parallel computing coupled with performant scheduling. On-going work focuses on reusing the scheduling technologies developed in MOAIS for embedded systems.

In the framework of our cooperation with STM (Serge de Paoli, Miguel Santana) and within the SCEPTRE project (global competitiveness cluster MINALOGIC/EMSOC), Julien Bernard in his thesis (grant cofunded by STM and CNRS) provides a specialized version of Kaapi for adaptive stream computations, named AWS, on MPSoCs platforms. AWS has been implemented and is being evaluated on two platforms: STM-8010 (3 processors on chip) and a cycle-approximate simulation (TIMA, Frédéric Pétrot). We are also studying self-specialized implementation of work-stealing from an abstract description (from SPIRIT standard) of the MPSoC architecture. Since those applications are developed based on component models, we are developing adaptive schedules for such component applications within the Nano2012 HiPeCoMP contract.

We are also considering adaptive algorithms to take advantage of the new trend of computers to integrate several computing units that may have different computing abilities. For instance today machines can be built with several dual-core processors and graphical processing units. New architectures, like the Cell processors, also integrate several computing units. First works concern balancing work load on multi GPU and CPU architectures workload balancing for scientific visualization problems.