Research On Optimization Of Dynamic Binary Translation

Nowadays, X86processors occupy many market shares and a wealth of software resources. Domestic processors are difficult to commercialize because of the lack of software resources. Dynamic binary translation can effectively solve the problem of software compatibility caused by the difference of architecture. It has a great significance for the development and commercial applications of RISC processors, especially for the domestic processors.In this paper, there is a research for a new dynamic binary translation system AB(Architecture Bridge),:whose goal is to directly run X86operating system or application on the Shenwei platform.Due to the big differences between the source and target architectures, the commonly used dynamic binary translation optimization techniques are difficult to achieve the optimization of the AB system, which seriously affected the efficiency of the AB system.This paper analyzes the performance bottlenecks in dynamic binary translation, and finds that it is too much redundant codes and frequent context switchings that affect the AB system.What’s more,dynamic instruction translation optimization technique based on decoding guided and the block chain optimization technique based on jumping instruction translation are proposed for solving the problems. Dynamic instruction translation optimization technique based on decoding guided is designed that the AB system can directly generate the optimized target code, achieve the desired optimization effect, and at the same time effectively reduce the overhead on the target code optimization. It designs the block chain optimization technique based on jumping instruction translation.By modifying the translation cache structure, it achieves the direct jump between the target code blocks, which reduces the system context switches and improves the efficiency of the implementation of the AB system.The test results show that, comparing with traditional optimization, dynamic instruction translation optimization technique based on decoding guided can increase11.21%n the efficiency of the AB system, and22.46%n the performance of the AB system by using the block chain optimization technique based on jump instruction translation. It can improve the performance successfully by using dynamic instruction translation optimization technique based on decoding guided and the block chain optimization technique based on jumping instruction translation. Morever, it provides a new idea for the research of dynamic binary translation.