Technology Of Target C Code Generation And Optimization In IA-64 Binary Translation

Binary translation technology is a branch of software reverse engineering. It can translate binary code from one architecture to another. Accordingly implement software immigration between diffirent architecture. This paper studies on the techniques of target code generation and optimization in the binary translation from IA-64 to Alpha and presents related implementations, focusing on the technology of high level control flow recovery.High level control flow recovery is a process of extracting and recognizing high level language control structure from intermediate code. The result of control flow recovery compresses the space of analysis phases, promotes the abstract level of programs, reduces the difficulties of the latter analysis phases, and is more useful to understand target code. In addition, when we use binary translator to analysize executable programs, the target C code generated by recovering high level control flow will be easier in recognizing program's behaviour. This paper firstly introduces the data structure of the intermediate code, furthermore analysizes the representation forms of the various high level control structure in intermediate code, according to the related theories, improves the relevant algorithms, presents three suitable recovery algorithms of high level language control structure to IA-64 architecture, correctly recovers the high level control flow structure which is similar to the original executable programs code in functions, solves the problems of large numbers of 'goto' statements in the target C code and the disadvantage of the indigestibility of the code, and is verified in the ITA system. Finally, this paper executes the testing in functions and efficiencies for the high level control flow generatation parts. The system has passed the testing of over 1000 programs from several testing benchmarks such as SPEC 2000. The testing results show that the generated target C code is correct in functions and the compiled binary codes in ALPHA have high efficiencies.