Research On Low Power Map Algorithm For Three - Dimensional On - Chip Network Based On Quantum Particle Swarm

In order to overcome the limitations of System on Chip (SoC) in power consumption, communication bandwidth, physical design and so on, Two-Dimensional Network on Chip (2D NoC) was proposed. However, with further increasing of chip integration,2D NoC reached the bottleneck in layout, area, density, power consumption and other aspects. Therefore, Three-Dimensional Network on Chip (3D NoC) came into being.3D NoC has many advantages, such as lower loss of interconnection, shorter global interconnects, smaller volume, higher packing density and higher performance. In the study of 3D NoC, how to mapping computational tasks into the nodes of 3D NoC is one of the key problems.3D NoC mapping has great impact on power consumption and delay of system. Mapping optimization has become an important method to solve the problem of power consumption and heat dissipation. It is necessary to do research on better 3D NoC mapping algorithms.In this paper,3D NoC mapping algorithms are studied in detail and the works completed are as follows. First, a novel global convergent algorithm named Quantum-behaved particle swarm optimization algorithm which has faster convergent speed is applied to 3D NoC low-power mapping problem for the first time. Simulation results show that the mapping algorithm based on quantum-behaved particle swarm algorithm has faster convergence speed with maximum case as 90.48% compared with the basic mapping algorithm based on particle swarm algorithm. It also can effectively reduce the power consumption of 3D NoC mapping with maximum case as 20.99% when the cores of Application Characteristic Graph is between 20 and 80. Secondly, to solve the mapping problem of large-scale 3D NoC, a low-power mapping algorithm for 3D NoC based on Diversity-controlled Quantum-behaved Particle Swarm Optimization is proposed in this paper. Simulation results show that for large-scale application characteristic graph with more than 120 cores, this algorithm is able to maintain stable power optimization efficiency (4.08%-8.04%) compared with mapping algorithm based on Quantum-behaved particle swarm optimization algorithm and converges much faster with maximum case as 66.7%.