动态二进制翻译的多线程架构及关键优化技术研究.doc

动态二进制翻译的多线程架构及关键优化技术研究.docABSTRACT
Dynamic binary translation technology (DBT) achieves the binary executable program
cross-platform transplantation from one processor to the other one. This technology is of great
significance in the software design and development, vulnerability mining and other aspects,
especially in dealing with the cross-platform software transplantation issue from X86 to the
domestic processors. This technique has an irreplaceable role in marketing and development of
the domestic processors. As the traditional dynamic binary translation platform belonging to the
serial program structure, optimization techniques brings too heavy dynamic binary overhead and
plays a negative impact on the performance of optimized effectiveness. With the development of
the maturity and multi-core multi-thread technology, how to take advantage of multi-threading
parallelization tools for dynamic binary translation performance optimization has e a hot
topic in the field of binary translation research in recent years.
To solve the problem of traditional dynamic binary translation platform optimized
inefficiencies and to enhance the efficiency of dynamic binary translation, we develop a
multithreaded dynamic binary translation optimization system based on in-depth exploration of
the works and the flaws of traditional dynamic binary translation platform and the key
technologies of the multi-threading. The main research and innovation include:
1. TEXE-HPOP model is proposed to face the issues on the problem that serial program
cannot take advantage of multicore resources, and multi-threaded dynamic binary translation
optimization system (MDOS) is designed to achieve multithreaded parallel execution between
the translation execution and hot paths optimization
2. The one-to-one mapping strategy of register from X86 to the domestic processors is
designed, and intermediate semantic equivalence transformation rules are proposed. The work
takes advantage of host machines registers and promotes