Controller node driven hop count based data distribution algorithm in ring connected binary tree network-on-chip for parallel processing

An effective distribution of packets is particularly important for the performance of a network-on-chip system. Ring connected binary tree structures do not have a defined distribution method and this paper elaborates the node architecture and a new distribution algorithm for ring connected binary tree structures. The distribution method described in this paper shows how optimized distribution of packets takes place for complex computations in ring connected binary tree. The controller node acts as a centralized node which creates the sub processing tasks and distributes to the computation nodes. The distribution of sub processing tasks is performed after figuring out the proximity of the computation nodes to the controller node in terms of hop count. In this work, the number of clock cycles for parallel processing is reduced by improving the synchronization between the transmit and receive operations thereby reducing the power consumption of the system.