Project Plan RECONFIGURABLE SYSTEM ARCHITECTURES

1 Summary This project proposal suggests research in a key area of embedded computer systems, namely how to obtain modularity and flexibility together with the necessary performance and cost efficiency. The research focuses on reconfigurable solutions at all system levels, taking advantage of technological developments and breakthroughs that have emerged recently. We will study the applicability of modular, reconfigurable processor arrays as well as modular, reconfigurable interconnection solutions. Together, these technologies have the potential to provide scalable solutions for the future. It is a specific advantage – and uniqueness – of the project that it takes a unified view of reconfigurability – on the building block level as well as on the system architecture/communication level. The project will evaluate configurable gate-and processor-array based hardware architectures through the implementation of advanced algorithms taken from the application areas of the industrial partners as well as from the University's research in intelligent signal processing. On the system level, we will study how static topolo-gies, through the use of optics, can be reconfigured in less than a millisecond, thus giving many of the advantages of dynamic topologies. Although, on the system level, the potential and opportunities of optically interconnected systems will be our primary focus, we will also study how reconfigurable interconnections can be obtained through, e.g., electronic FPGA circuits. A very important outcome of the research is the knowledge of the process from algorithm to efficient flexible hardware. We will gain knowledge about how the hardware designs can adapt to the applications and how the applications can be redeveloped to draw benefits from the new computational processes. This will be demonstrated through a number of representative applications, but will have a general applicability. Since the use of parallel programmable hardware and high-speed interconnection solutions, including optics-based ones, will continue to grow in importance, knowledge of this design process and its potential has great significance. The overall goal of the project is to identify and evaluate the various ways of utilising reconfigurability and a modular approach in the design of embedded computer and communication systems. As part of this, we will evaluate the efficiency of some emerging, reconfigurable, parallel processing paradigms in selected embedded applications. Of specific interest is the granularity of the parallel processor in relation to the parallelism of the application. An important part of this will be to find ways of mapping the algorithms efficiently onto the parallel architectures. We will …