NEST - Research Summaries

Summary of Research

The Netherlands holds a strong position in the development and production of a wide variety of electronic equipment. Examples are radar systems (Thales), medical systems (Philips Medical Systems, PMS), copiers (Océ) AND INfotainment systems (NXP). These products have in common that they operate on streams of data and are therefore called streaming applications. To survive international competition, future applications of these companies require processing support for real-time and high-performance streaming applications. For example, future medical imaging systems need to reveal much more detail with lower latencies. Latencies within the millisecond range are the ultimate goal where minutes are feasible with state-of-the-art processors. Only then, real-time medical imaging during surgery is feasible. This requires a dramatic increase of computational capacity against acceptable production and operational costs. Similar requirements hold for streaming applications in the other application domains mentioned above. Furthermore, because of the energy awareness of the general public, there is more and more pressure to produce energy efficient systems.
Due to power consumption, performance and complexity barriers, known processing approaches will not lead to the necessary product innovations. Besides that, current approaches lack the ability to cope with the increasingly dynamic behaviour of architectures and applications in a predictable way. A different approach is inevitable. By limiting the application domain to streaming applications, processors can be more specialized compared to General Purpose Processors, which will increase their energy-efficiency dramatically. Multiple processors will be combined onto a single integrated circuit leading to a so-called Multi-Processor System on Chip (MP-SoC). Tools are required to program and operate these devices efficiently. Because the scope within NEST is limited to streaming applications, these multi-processor systems can be programmed and operated efficiently as preliminary research has indicated.

Within the NEST project, the approach will be to research and exploit:

MP-SoC architectures with means for low power, composability, and reconfigurability
A design flow for MP-SoC based systems using high level synthesis
An MP-SoC run-time system management. By means of dynamic reconfiguration, the run-time system is capable of dealing with adaptive service requirements and platform variability

To realize a major breakthrough requires a broad spectrum of disciplines. The NEST project combines the required expertise on streaming applications available in the Netherlands, resulting in a consortium of research groups from all three Technical Universities in the Netherlands and from Leiden University. Industry is emphasizing the relevance of this project by allocating five PhDs to NEST.

Summary of Utilization Plan

The efficient and predictable design of streaming based systems is crucial to the Dutch embedded systems industry. This is reflected by its great interest in the NEST project. NXP is planning to use the results of the NEST project to improve its software defined radio portfolio. Thales considers Multi-Processor Systems on Chip (MP-SoCs) as the enablers for the new generation SMART antenna systems. The tools and methods of the NEST project are necessary to reveal the raw computing performance to these types of antenna systems. Philips Medical Systems is considering the use of MP-SoCs in its equipment for Image Processing for Clinical Applications in X Ray Systems, while Océ wants to evaluate the use of MP-SoCs for digital image processing. Eventually, the project will result in four demonstrators, implemented using a demonstrator platform, a run-time resource manager and a design-time toolflow, realized within NEST.
The companies mentioned above actively support the NEST project by financing 5 PhDs and their daily guidance. To ensure that the NEST results will indeed be used within the companies, a stakeholder will be appointed for each company. A stakeholder has to ensure that developments within NEST are consistent with the roadmap of the company. The project will follow a phased approach where at the end of each phase, the results of this phase are evaluated and goals for the next phase can be set. This way, stakeholders can influence the course of the project. Furthermore, a stakeholder is responsible for utilization of the NEST result within the company.
To facilitate cooperation, bi-weekly video conferences will be organized. During these meetings, general coordination aspects are dealt with but also at least one PhD student will be presenting his/her work. Besides contributing partners, associate partners are actively involved. Associate partners are members of the user committee/advisory board and are allowed access to the prototype tools. An ASCI winterschool for the Dutch academic community is planned in 2010. To encourage dissemination towards the embedded community in general, the NEST consortium will organize each year a meeting called 'Annual NEST Research Day (ANRD)', open to both academic institutes and industry, a website will be launched from which all relevant documents and the NEST GPL toolchain can be downloaded.
Besides the multinationals within the project, SME companies have shown great interest. For that reason, companies like Salland Electronics, Devlab, Recore Systems, Compaan Design and ACE will provide a contribution to the consortium if appropriate.
The NEST project will contribute to the Dutch knowledge-base with respect to Embedded Systems by educating many PhDs and Master of Science students in this highly relevant area.