Objective: To develop a wireless processing platform that is energy efficient, yet capable to process streaming audio and video data. In the SENSA project an important part of the processing is done in or close to sensors. Audio and video are typical examples of streaming data that will be used in SENSA. This means that efficient stream processing platforms are needed that can be built into such sensors. Because of the flexibility, the programming platforms need to be programmable (or reconfigurable), and should enable easy and efficient execution of stream processing algorithms. Energy-efficiency is one of the major design constraints of such high bandwidth and processing intensive sensor nodes.
It is expected that, in the coming years, semiconductor technology will advance in such a way that devices with hundreds of cores can be designed and will become available. Multiple cores of different type (e.g. low-power RISC, SIMD, accelerators, etc.) can be placed onto one (heterogeneous) platform interconnected by a Network on Chip (NoC). Such architectures are interesting because images from cameras and sound from microphones can be processed in parallel on a SIMD processor.
Techniques will be needed to scale performance and power-consumption of individual cores through voltage- and frequency scaling or to isolate cores by switching them off completely. This last technique can also be used to switch-off faulty cores, increasing reliability of MP-SoCs. WP2 will not be able to deliver a real chip as prototype. Prototypes are functional FPGA implementations without the desired energy consumption profile that can be used in WP5. The results of WP2 are used in WP5.
- Task 2.1 Advanced power saving technologies to reduce the power consumption of processors for streaming applications (UT/CAES)
- Task 2.2. Development of low-power architectures for streaming data, with an implementation and evaluation on FPGAs (TU/ES)
- Task 2.3 Development of streaming data algorithms that can run on the developed platforms (VDG)