Design And Implemention Of Single Core Single Thread Of CPU Perforamance Analysis Tool For ARM-android's Application

In recent years,ARM Android has become a mobile intelligent terminal platform with the largest market share.With the surging of various Android applications,the scale and complexity of them have witnessed continuous improvement.The interactive modes and hardware properties of Android mobile terminal require that applications should take full advantage of hardware resources;Only by doing this,they can provide excellent user experience.For the native libraries of Android applications,performance analysis and optimization of CPU micro architecture will be greatly helpful for developers to create smart applications,which make full use of CPU resources.The market,however,is short of efficient performance analysis tools.The thesis mainly studied the hardware performance counter-based performance analysis tool of ARM Android platform.It made in-depth analysis of the structure and operating methods of Oprofile performance analysis tool.With a thorough understanding of its operating principle,it made reference to its realization process design,and realized the hardware performance counter-based performance analysis tool.The thesis also made research of data acquisition module,realized the terminal sampling of combination of time and Linux kernel static system tracepoint It supports the collection of processor micro architecture behavior characteristics data in the whole system.It improved the Varint coding method in Google Protobuf,made effective compression of collected data,and saved a large number of memory space.On the basis of TCP/IP protocol,it designed application layer communication protocol,realized the communication module of host machine terminal and target machine terminal.It constructed a safe and efficient data transmission channel between host machine and target machine.Fundamentally,the thesis focused on data analysis module.It made integrated utilization of software and hardware performance data,and adopted top-down micro architecture performance analysis method.It established CPI-Stack performance analysis model for application hot spots,and applied it to users' control interface.It thus made it easier for application developers to take performance analysis at processor micro architecture level,to detect performance bottleneck quickly and accurately,and to make efficient performance analysis and optirmization.In addition,the aiding resources applied to this thesis took an extra 5.07%beyond the original budget,which is acceptable for the total expenditure.