Between Suif2-based Process Analysis And Optimization

With the great development of processors technology, compiler optimization techniques take more and more important role in the research of modern processors. Not only can appropriate optimization algorithms generate high efficient target code, but also simplify the hardware control mechanism for processors. Moreover, generations of new language features and techniques for analyzing programs present new challenges to compilers, which need be adapted to these new language features. We should use new methods to improve code quality and resolve new problems during compiling.The traditional intraprocedural analysis and optimization handle the objects in only one procedure. But interprocedural analysis and optimization cover all over the program. For dealing with the whole program, they can get more context using the more complicated methods. It needs other analyzing techniques, distinguished from those of intraprocedural analysis, to find useful information across each procedure which compilers use to optimize code subsequently. Firstly, this dissertation introduces the conventional optimization techniques concisely and compares those with interprocedural techniques. Then, it explains the impact of interprocedural analysis and optimization, which results in the motive of our research. We present some classical theories about interprocedural analysis and optimization, including call multigraph construction, inline substitution and procedure cloning, and interprocedural information (side-effect summary sets, alias information and interprocedural constant). Furthermore, we give the overview of SUIF2, our research environment, which has modular structure and extensible program representation to maximize code reuse. With the modules and tools provided by SUIF2, we implemented an algorithm to construct call multigraph, and verified results. Lastly, we summarized our research work.Interprocedural analysis and optimization has great significance. Although they introduce some large compiler-time cost, they represent a trend of compiler optimization. Along with the more developed and effective methods of interprocedural analysis and optimization, there must be more compilers to be applied with these techniques.