Research On Self-similarity Characteristics Of Network On Chip

With the rapid development of integrated circuit, the traditional System on Chip (SoC) has been increasingly unable to meet the current design requirements. A new concept of Network-on-Chip (NoC) has been proposed by academic community to solve the problems with regards to scalability, communication efficiency and clock synchroni-zation with which SoC is facing. Based on the in-depth study of NoC architecture, routing algorithm and switching mechanism, the self-similarity characteristic are found in the traffic of NoC just as that of in computer network. At the same time, applications mapped on NoC, such as multimedia applications, also have distinct self-similarity feature, which enhances the self-similarity of network traffic in NoC. The main works of this paper are as following:A simulation platform featuring configurable the mesh dimension, routing algo-rithms, buffer depth of each router channel, and so on, is built with SystemC. And it has passed through the function verification correctly.To lower the influence caused by self-similarity of network traffic on network per-formance, a routing algorithm, Distance Prediction XY (DPXY) routing algorithm, is in-troduced based on the congestion information. It can process the congestion information of router while the inFSM module processing the head flit. So the optimum direction is selected in advance, and success rate of data transmission link established by NoC router is improved.Finally, a H.264 decoder algorithm is mapped on a NoC simulation platform featur-ing DPXY routing algorithm, the decoding performance increases 13.46% compared with that using XY routing algorithm.