摘要
针对系统级二进制翻译器在虚拟/物理地址转换中出现的性能瓶颈问题,提出一种基于指令、数据和堆栈等访问区域特征的快速地址转换方法.该方法在翻译态时识别不同区域的内存访问指令,并根据区域特征采取不同的优化算法,对于指令区域和数据区域采用编译时地址转换算法,对于堆栈区域采用连续页面映射算法.在开源高速指令模拟器QEMU上运行嵌人式CPU测试基准程序PowerStone.结果表明,运行态地址转换过程减少了1%~65%,运行态执行周期数减少了11%~38%.
To accelerate the speed of virtual/physical address translation in system-level binary translation, a high speed address translation method based on the memory access region attribute was proposed. Memory access instructions of different region was identified in the translation state and corresponding optimization algorithm for different memory region was implemented. Compiling based translation algorithm was adopted for instruction and data region. Consistency page translation algorithm was adopted for stack region. The results of running powerstone on the QEMU emulator shows that the address translation process in execute state decreases by 1%-65% and the total executed cycle decreases by 11% to 38 %.
出处
《浙江大学学报(工学版)》
EI
CAS
CSCD
北大核心
2014年第2期348-353,共6页
Journal of Zhejiang University:Engineering Science
基金
中央高校基本科研业务资助项目(2012QNA5004)
关键词
内存访问区域属性
系统级二进制翻译器
快速地址转换
memory access region attribute; system level binary translation; high speed address translation