| As the core component of aircraft,airborne computers have the characteristics of high reliability,high safety and high performance.With the rapid development of aeronautical computing technology,more requirements have been put forward for the new generation of airborne computers.The USA,for instance,are taking the lead in the development of the latest generation of airborne computers.It has put forward the concept of Integrated Core Processor(ICP),which has been applied successfully to F-35 and other important models.The emergence of ICP indicates that the airborne embedded system is developing from the discrete,hybrid and joint structure to the highly integrated one.ICP adopts many new technologies,such as the partition operating system,the heterogeneous multi-core processing,and the multi-class bus hybrid communication,etc.,whose requirements cannot be satisfied by the traditional performance evaluation index and capability measurement methods.Therefore,system performance evaluation methods and capability measurement models are urgently needed to be further studied.In view of the above problems,this thesis proposes a multi-dimensional comprehensive evaluation method of embedded system performance based on an in-depth analysis of the characteristics of current airborne computers and ICP.The proposed method focuses on the evaluation of the intelligent control capability of embedded systems,analyzes the indicators affecting their performance at all levels,thus putting forward corresponding evaluation methods and measurement models.Specifically,the contributions of this thesis are as follows:1)This thesis presents a comprehensive evaluation method of the five-dimension embedded system performance based on the grey theory.Considering the five aspects of intelligent control,network interconnection,comprehensive calculation,security protection and energy consumption control of embedded system,the proposed method evaluates the system performance in a multi-dimensional and multi-level way and renders the evaluation results through radar charts.In this way,the system can be better selected and optimized.Through the grey correlation analysis,the problem of inaccurate evaluation results caused by high correlation between indicators can be solved.In addition,about intelligent control capability of embedded system,a performance evaluation method based on the deviation intelligence quotient and a corresponding performance evaluation model based on capability curve are proposed.Consequently,the problem that the evaluation results are difficult to measure uniformly due to the different upper limits of the index scores in the evaluation of intelligent control capability can be solved.A case study is conducted to illustrate the feasibility and the effectiveness of the proposed evaluation method.2)A dynamic communication strategy between heterogeneous multi-core processing units based on GCM factor and its adaptive capability evaluation method are proposed.Traditional heterogeneous multi-core processing units usually adopt static communication strategy among cores.Because of problems such as changeable operating environment,limited resources and unstable communication performance,this thesis proposes a dynamic communication strategy model between heterogeneous multi-cores based on system memory and time constraints.It also introduces the influence factors of communication granularity,communication buffer and message transmission mechanism(GCM)to study the model's impacts on the performance of inter-core communication in different stages and valuate the adaptive capability of inter-core communication.The experimental results show that the dynamic communication strategy optimizes the transmission efficiency of the inter-core communication by selecting GCM factor reasonably and reduces the task execution time by 5%-30% compared with the static communication strategies with different parameters.Therefore,it can be concluded that the dynamic communication strategy has superior adaptability and stability.Meanwhile,the proposed method can accurately give the adaptive scores under different strategies,thus making the evaluation results more comprehensive and adaptable.3)A partitioned task scheduling algorithm with dynamic periodic execution time(DCET)is designed and implemented,and a self-optimizing capability evaluation method for task scheduling is proposed.About the problem of excessive idle time slices in fixed period time window of partition task scheduling,the residual time slice management mechanism is adopted to calculate the remaining time slices after each partition completes the periodic task,and the task execution sequence in the partition is dynamically planned in real time to improve the utilization of processor.Through the analysis of the schedulability and real-time performance of DCET algorithm,the evaluation method of task scheduling self-optimizing ability is extracted.The experimental results show that compared with APS and its advanced algorithm,DCET algorithm reduces the average task switching time by 0.015 um(about 0.4%)and the average task execution time by 2.585 MS(about 9.14%).The proposed evaluation method effectively evaluates the self-optimization ability of different task scheduling algorithms.4)A self-regulation mechanism of embedded system energy consumption based on Roofline model and the evaluation method are proposed.Due to the problem that there are various kinds of energy-saving control technologies and the energy-saving effect of different technologies cannot be measured uniformly,according to the characteristics of airborne embedded computer,an energy-saving self-regulation capability evaluation system is proposed.Such system includes key indicators such as average energy-saving rate,performance loss rate,energy-consuming performance ratio,and energy-saving intensity etc.It measures the relationship between different energy consumption control strategies by drawing the Roofline model through performance,energy consumption and energy-saving intensity as well as elements like roof line and ridge point.The experimental results show that DPM and DVS strategy provides higher computing power per unit energy consumption,that is 2.37% lower than that the strategy without energy saving,3.8% lower than DPM strategy and 2.5% lower than DVS strategy.Also,DVS strategy plays a leading role in saving energy.The proposed evaluation method effectively quantifies the relationship between system energy consumption and performance under different strategies,and the evaluation results can correctly reflect the self-regulation ability of system energy consumption. |
PDF Full Text Request