Research On Garbage Collection Based On ART

Google developed the ART system to make the Android system more smoothly,which rarely appear pause.From the point of view of garbage collection,it has been greatly improved that the garbage collection mechanism in ART,but it still can not be a good solution to non-removable garbage collection in the memory fragmentation.The thesis is aimed to the optimization and improvement of the original garbage collection mechanism in ART.In this thesis,the garbage collection mechanism of ART is studied deeply.Aiming at the problem of memory fragmentation caused by the mechanism of non-removable garbage collection in ART,an incremental garbage collection algorithm is proposed.This algorithm can compress the part of Alloc heap in a relatively short time,which leads to reduce the memory fragmentation and improve the memory utilization.The ART has been used to refine the heap space structure and increase the large objects,so the small objects are allocated in the Alloc heap,only.The replication cost of small objects is relatively low,which can effectively reduce pause time of the increasing generational garbage collection.The execution time of incremental garbage collection algorithm is studied in this thesis,optimizing the garbage collection strategy of ART.Because the incremental garbage collection algorithm deal with new objects by Sticky GC and enhance the age of objects by Partial GC,the execution time of the algorithm is determined by the currently running CMS algorithm and the threshold of address together in the thesis.This strategy can not only guarantee the efficiency of the original Sticky and Partial GC in the ART,but also improve the efficiency of the incremental garbage collection algorithm.The incremental generation garbage collection algorithm is designed to solve the problem of memory fragmentation.But the most direct data to evaluate the performance of garbage collection is the garbage collection pause time and garbage collection start times.Only when the algorithm reduced the pause time and the number of triggers it is can users enjoy experience.This thesis uses Antutu and CaffeineMark to test the incremental garbage collection algorithm on Nexus5 phone.The results show that the algorithm effectively reduces the pause time of garbage collection and the number of triggers.The optimized garbage collection mechanism decreased the average pause time by 14.0% compared to the native,and decreased the number of triggers by an average of 18.05%.