基于LLVM的多目标高性能动态二进制翻译框架  

LLVM-Based Multiple Targets High-Performance Dynamic Binary Translation Framework

作　　者：石磊[1] 田娜 康烁 Shi Lei;Tian Na;Kang Shuo(China Air to Air Missile Research Institute,Luoyang 471000,China;Beijing Digiproto Technology Co.,Ltd.,Beijing 100085,China;Research Institute of Information Technology,Tsinghua University,Beijing 100084,China)

机构地区：[1]中国空空导弹研究院,河南洛阳471000 [2]北京迪捷数原科技有限公司,北京100085 [3]清华大学信息技术研究院,北京100084

出　　处：《航空科学技术》2020年第8期73-78,共6页Aeronautical Science & Technology

基　　金：航空科学基金(2017ZD12013)。

摘　　要：动态二进制翻译技术是构造高性能异构虚拟机的关键技术,而代码翻译的质量则是动态二进制翻译性能的关键。本文实现了一种基于LLVM动态二进制翻译框架,利用LLVM优化技术以及多目标编译的特点,实现了可以针对多个目标体系结构的高性能动态二进制翻译。基于开源Skyeye实现了这种翻译框架,并在两种目标体系结构ARM和PowerPC上验证了框架的可移植性和运行效率,与QEMU在ARM目标平台上做了性能对比,结果表明该模拟器比Qemu性能平均快10%以上。Dynamic binary translation is a key technology of constructing high-performance heterogeneous virtual machine, while the quality of the code translation is the key of dynamic binary translation performance. An LLVMbased dynamic binary translation framework is implemented. Utilizing LLVM optimization technology and the features of multiple target compiling, the high performance dynamic binary translation for multiple target architectures is implemented. The dynamic binary translation framework is implemented based on open source software Skyeye. The portability and performance were verified on two target architectures ARM and PowerPC. Compared with Qemu on ARM target platform, experiment results show that the average performance is faster than QEMU by over 10%.

关 键 词：动态二进制翻译 异构虚拟机 翻译性能 LLVM 多目标编译 

分 类 号：TP314[自动化与计算机技术—计算机软件与理论]