注解信息制导的动态二进制翻译及优化

动态二进制翻译器在运行时将源体系结构机器码翻译为目标体系结构机器码.这种即时编译技术使得源机器上的软件无需重编译就可以直接在目标机上较高效地运行.然而,利用动态二进制翻译器运行源软件的效率大大低于针对目标机器重新编译运行源软件的效率.本文在比较分析动态翻译生成的目的机器码的性能偏低的原因的基础上,提出了注解信息制导的动态二进制翻译及优化的方法.本文选取了三种注解信息,在英特尔的商用动态二进制翻译器"IA-32Execution Layer"和静态编译器"Intel(r)Compiler"上实现了注解信息制导的动态二进制编译及优化技术.实验结果表明该三种注解信息较大程度地提高了动态翻译码的执行效率.

Annotation Directed Dynamic Binary Translation and Optimization

Dynamic binary translator enables transparently running applications of source architecture on the top of target architecture without re-compilation. However, the efficiency of running applications using dynamic binary translator is much slower than directly running them on the target machine after re-compilation. This paper analyzes the reasons that lead to the low performance of the translated code, and proposes to narrow these gaps through annotating useful information to the source binary. The paper describes the overall framework of annotation directed binary translation and introduces an extendable annotation format. To prove the feasibility and efficiency of the framework, we selected three annotations and implemented the dynamic optimization based on these annotations. Our results show these annotations improve the performance of the translated code significantly with moderate overhead.