Research On Processor Performance Prediction Method Based On Belief Rule Base

With the continuous improvement of processor performance,its design and manufacture become more complex.The performance analysis and prediction of processor become a reference target of processor design,and also a basic index to measure the level of design.While some architectural enhancements have brought significant improvements in processor performance,they have also increased the complexity of processor design.Generally,the evaluation index of processor performance consists of many structural parameters that directly or indirectly affect the whole system,including program characteristic parameters,processor-application related parameters,etc.Due to the complex interaction between the structure and application characteristics of the processor,it is difficult for the architecture designer to establish an intuitive and accurate performance analysis model and have a deeply understanding of the processor structure.So,it is also hard to evaluate the impact of structural improvements on performance.Affected by the system structure and environmental factors,it is difficult for traditional performance prediction modeling method to improve the interpretability of the model while ensuring the accuracy of the model.In addition,output process cannot be derived and proven when using traditional model.Therefore,it is necessary to propose an accurate and reliable method to analyze processor performance as a feasible solution.To solve the above problems,a reasonable and accurate performance prediction method is proposed in this paper,which can help to understand the factors related to processor resource usage and provide a reliable reference for performance analysis.The main research contents of this paper include the following aspects:(1)To solve the problems of complex analysis index and model accuracy in processor performance analysis,a processor performance analysis model combining Evidence Reasoning(ER)and Hierarchical Belief Rule Base(HBRB)is proposed.In this method,the ER algorithm is used to evaluate the influencing factors of the processor in different layers,and then the performance is comprehensively analyzed by hierarchical BRB.Finally,the model parameters are optimized by the intelligent optimization algorithm.The experimental results show that the accuracy of the model can be effectively improved by training and parameter optimization.(2)Although the performance analysis model can initially solve the prediction accuracy problem,the interpretability of the model needs to be further explained.Based on the study of processor performance analysis models,a new processor performance prediction model is proposed based on Interpretable Hierarchical Belief Rule Base(HBRB-I)and Sensitivity Analysis(SA).Combined with the mechanism analysis of the input index and the general explainable criteria,the explainable prediction model was defined to ensure the transparency of the reasoning process.Finally,the global sensitivity analysis method was used to verify the explainability of the model.The method studied in this paper has been effectively verified on the hardware data set.(3)In order to better apply the predictive method of the processor performance of the belief rule base to the actual system,this paper designs and implements the predictive system of the processor performance based on the interpretable belief rule base.The system includes man-machine interaction,index evaluation,performance prediction,sensitivity analysis function.The core algorithm of ER&HBRB-I is embedded in the system to ensure the accuracy and interpretability of prediction results,and complete the process of theoretical innovation,experimental verification and concrete implementation.According to the function analysis,the system design in this paper includes man-machine interaction,parameter setting,algorithm execution and sensitivity analysis module.In summary,aiming at the problems of complex processor analysis indicators,low model accuracy and lack of explainability,this paper adopts rule modeling methods such as ER and BRB to build an interpretable processor performance prediction model.At the same time,it combines the intelligent optimization algorithm to optimize the model accuracy and analyzes the feasibility of the method from a theoretical perspective.Combined with the computer hardware data set,the effectiveness of the proposed method is verified,and the actual application system is developed,which provides a reasonable solution for the performance prediction of the processor.