Study And Applications Of Analyser Of Android Executable Programs

Android can quickly become the mainstream mobile device operatingsystem,Because it has some outstanding characteristics,such as open source,goodportability and the application which on the system has unrelated with the hardwareplatform.As the limitations of the hardware, computing power and storage capacity ofthe phone is very limited,if we migration the run time-consuming function of theAndroid applications which on the phone to server-side execution which has highperformance and large storage, that can speed up the execution speed of the application,and provide a better user experience.There are two key issues when we migrate the time-consuming functions ofAndroid application: first, how to known which function can be migrated; second,howto know which function is a relatively time-consuming function. The method that solvethe first key problem is: create the dex file analysis model, then get the callingrelationships between functions in the program, and filter out those functions that usethe local resource are not suitable to migrate (the local resources, such as UI, hardware,network functions and local (NATIVE) function),and the direct calling functions,indirect calling functions can not migrate too. Compared with existing the analysis toolsTraceview and dmtracedump, the method is based on the compiled executable programfiles,without having to modify or fetch the source code,can be achieved function callrelationship analysis, has a wide range of practical value in real applications. Themethod that solve the second key problem is: modifying the source code of the Dalvikvirtual machine, we can add the start and end time function at the the inlet and outlet ofDalvik virtual machine interpreter to execute the function. Compared with existingmethod, we cannot need or modify source code,only need to put the text which recodethe functions that need to calculate the execute time into a mobile device.To illustratethe effectiveness of the proposed method, the thesis provides some examples toexperiment,and the results that the method of analysis are correct, and has a higherefficiency. This thesis provides a method for obtaining a migration function execution time,overcoming the prior method which must get the source code of application or modifythe dex file before we receive the functions calling relations and the execution time ofthe functions, it has greater applicabilityand when the same function run on the x86platform which compare with run on mobile devices, the functions which run long timeon a mobile device or consume memory or CPU, migrate to x86, that can improve theefficiency of15-20times.