Research On Autonomous Temperature Sensing Technique For Optical Network-on-chip

Optical network-on-chip is an emerging communication architecture,which has been widely studied and used for its low latency,ultra-high bandwidth,and good scalability.However,with the increasing number of processor cores integrated on a single chip,the optical network-on-chip system often suffers from overheating problem.Overheating problem can affect the life and performance of the chip,making the system unreliable.How to precisely monitor the temperature of each processor core on a single chip to reschedule and assign tasks is critical to ensuring the reliability of the optical network-on-chip system.Recent on-chip temperature sensing techniques are typically implemented in either hardware-based or software-based.For specifically,recent hardware-based on-chip temperature sensing techniques possess high accuracy and efficiency at the expense of additional overhead on-chip area and cost.The existing software-based on-chip temperature sensing techniques require no additional hardware support,but possess the features of low speed and large overhead,cannot meet the requirement of system performance.Based on this,and considering the unique advantages of software techniques,a software-based temperature sensing technique is proposed in this thesis.The specific work is as follows:Initially,in this thesis,the thermo-optic effect of the micro-ring resonator is introduced in detail.And then,the thermal sensitivity model of the micro-ring resonator is proposed and some reasonable assumptions are made to better study the autonomous temperature sensing technique in this thesis.Secondly,based on the thermal sensitivity model of the micro-ring resonator,this thesis proposes a temperature sensing technique model based on the optical network-on-chip,thus the problem of measuring the temperature of the chip is transformed into a mathematical problem of solving the equations.In addition,in order to further improve the success rate of finding the unique solution of the model,the linear model theory and ring model theory are proposed in this thesis.Thirdly,based on the autonomous temperature sensing model of optical network-on-chip,this thesis proposes an autonomous temperature sensing algorithm,which is realized by building additional auxiliary paths by using the idle communication resources on optical network-on-chip.Finally,in this thesis,simulation experiments are carried out on the simulation software,and the experimental results show that the thermal sensitivity model of micro-ring resonator proposed in this thesis is correct.In addition,simulation experiments are carried out on a large number of randomly generated data communication instances of optical network-on-chip,and the experimental results show that the algorithm proposed in this thesis can find the unique solution of the model with a success rate of 94%,which is much higher than the success rate of depth-first search algorithm.Finally,when compared with other existing methods,the proposed method in this thesis possess the advantages of high efficiency,high accuracy,and low power consumption.