System on Chip
A system on chip (SoC) is usually made up of a processor connected to a set of peripherals and dedicated hardware modules via a bus system. In video streaming applications in which large amount of data must be computed in real time, this traditional system architecture is no longer viable because the bus becomes a bottleneck.
The programmability and the seamless use of hardware accelerators in image processing application is insure through the design and implementation of a streaming data interface. By porting the Intel's OpenCV computer vision library to the system, software developers does not need to recompile their code in order to use the platform.
In the SoC design part, the major focus is on the implementation of a concept which connects physical external memory to the several implemented application oriented processor units, to supply them by required raw stream data and to restore processed results get back to memory without intervention of main processor (Power PC). This interface has been called Streaming Data Interface (SDI) and plays major role in the whole performance of system. Our goal is to build a complete solution for software developers to have an accelerated system on chip to perform their OpenCV algorithms. In hardware-software-co-design the data flow analysis plays an important role because the communication between modules can heavily influence the performance of the system. Our architecture offers a wide spectrum of possibilities to interconnect modules and to let data flow.
