A programmable processing element dedicated as building block for a large area integrated multiprocessor system

The architecture and implementation of a programmable processing element dedicated as building block for a large area integrated multiprocessor system is presented. The processor element allows an efficient implementation of video coding standards. It consists of a RISC processor supplemented by a low level coprocessor for computation intensive convolution-like tasks. Several of the processing elements can be bus-connected to built a coarse-grained MIMD based large area integrated multiprocessor system. Each of the processor elements provides high performance for video coding tasks in order to keep the overall number of required processing elements for complex real-time video coding applications small. A key feature of the processing element to support large area integration is a unidirectional data supply which allows reconfiguration strategies for the interconnection network. Furthermore on-chip program and data memories have been implemented in order to relax the overall bandwidth requirements of the large area integrated multiprocessor system.