面向Android系统库文件访存的汇编优化策略

Assembly Optimization Strategy for Memory Accesses of Android System Libraries

以自主嵌入式处理器为平台,对Android系统性能进行分析。通过Oprofile工具采集Android系统下的访存热点函数,结合处理器架构特点,并充分考虑传统Cache特性,重点针对Android系统的Bionic C库及Libcutils库中的热点访存函数提出优化算法进行汇编优化。实验表明:优化后的Bionic C库和Libcutils库与优化前相比,访存带宽分别提升8.91%和12.3%,系统性能分别提升1.54%和3.81%;Android系统整体性能提升5.35%。

Android system performance is analyzed based on sen-designed processor embedded platform.The hot-spot functions are monitored and sampled which is assisted by Oprofile tools,combined with the processor structure and consideration of traditional Cache hierarchy,the corresponding hot-spot functions in Bionic C library of Android system and Libcutils library are analyzed and optimized through the optimization algorithms.The experimental results indicate that compared with the performance before optimization,Bionic C lib and Libcutils lib could contribute 8.91%and 12.3%bandwidth profit and 1.54%and 3.81%system performance profit respectively,the whole performance of Android system can be promoted 5.35%.