Introducing Chaotic Mechanism Of Enhanced 2 D Mesh Network Mapping Algorithm Research

Network-on-Chip, which is a network-based multi-core system solution thatdifferent from bus-based architecture, is proposed to solve bus competition issuesbrought by more and more integrated resources in traditional System-on-Chip. Lowpower consumption and low delay are still considered to be critical problems in futureNoC system design. So this thesis makes low power consumption and low delay asentries, researches mapping problem which is a key step in NoC system-level design.Power and delay constraints are converted to a single constraint to establish the unifiedevaluation function. An effective algorithm for searching the optimal solution based oncombined power-delay constraint is designed and implemented in this thesis.The verifications are based on2D Mesh platform because this topology which hasparticular advantages has been widely used in NoC key issues studying. In thisdissertation, after analyzing the composition of power consumption and delay in NoCarchitecture, a power-delay model combined with the2D Mesh characteristics isproposed, and this model is only related to the task size, communication amountbetween tasks and the length of transmission path. Based on this, a mathematicaldescription of NoC mapping under combined power-delay constraint is given. Andwhat’s more, this description in not only applies to the2D Mesh topology, but can beextended to others.NoC mapping has been proved to be a sort of NP-Complete problem, which meansit beyond the computation ability to solve it if the problem size is too large. As a result,approaches in this area are focused on heuristic algorithms to find solutions metengineering requirements. After reviewing a series of heuristic algorithms, we foundthat not all existed algorithms are suitable here due to the features of NoC mappingproblem. Artificial bee colony (ABC) algorithm, which carries out two-stepoptimization based on probabilities, is selected as the basic algorithm after a wide rangeof contrasts. We analyzed the features of the original ABC algorithm and modified it tofit the discretization characteristics of NoC mapping, and then the defect of originalABC algorithm was improved. New discrete artificial bee colony algorithm inherits the advantages of simple for understanding and easy for implementation. What’s more, theproposed ECABC algorithm converges quickly and avoids prematurity by enhancedchaos mechanism.Finally, after a series of simulation comparison, we verified the algorithm’sfeasibility, approached a balance between search speed and quality, and improved thedefect of original ABC algorithm. The simulation results prove that the proposeddiscrete artificial bee colony algorithm with enhanced chaos mechanism is high efficientfor NoC mapping problem.