Research On Critical Techniques In Unified Translation For Multi-source Binary Code

Binary translation plays an important role in maintaining legacy code, promoting hardware innovation and protecting computer security, and thus has a vital significance to the prosperity of the national computer industry. Due to the complexity, the development of a new binary translation system is labor-costing and time-consuming. Therefore, an increasing number of researchers dedicate to seeking a unified approach to binary translation for a variety of platforms, in order to shorten development cycle and reduce development costs.Aiming at migrating software from multiple platforms to a certain Chinese-made CPU in a fast way, this thesis investigates several key techniques in multi-source binary translation. The main contents and contributions of this thesis are as follows:First of all, for achieving platform independence and scalability, the multi-source translation architecture based on a virtual framework is designed, and the instruction atomic semantic description language is proposed after in-depth study on the similarity of basic elements in different instruction sets, to build a unified and platform-independent intermediate representation. Thereafter, in order to ensure the completeness, correctness and simplicity of semantic equivalence conversion, the algorithm combined with the semantic hierarchical structure is proposed and implemented. Besides that, the methods of flags simulation and optimization translation using semantic tree integration are presented to improve the quality of generated code. Finally, considering the incompleteness of translation caused by indirect jump and limitations of existing approaches, a jump table recovery method based on critical semantic subtree is proposed and implemented. By extracting the formal semantics characteristics, jump table which is the typical form of indirect jump can be solved in a platform and compiler independent way.The strategies and algorithms proposed above have been implemented in the Multi-source Binary Translator MBT. Through the tests on the related modules in the MBT system, the correctness, validity and generality have been well proved.