The impact of interprocedural analysis and optimization in the Rn programming environment

In spite of substantial progress in the theory of interprocedural data flow analysis, few practical compiling systems can afford to apply it to produce more efficient object programs. To perform interprocedural analysis, a compiler needs not only the source code of the module being compiled, but also information about the side effects of every procedure in the program containing that module, even separately compiled procedures. In a conventional batch compiler system, the increase in compilation time required to gather this information would make the whole process impractical. In an integrated programming environment, however, other tools can cooperate with the compiler to compute the necessary interprocedural information incrementally. as the program is being developed, decreasing both the overall cost of the analysis and the cost of individual compilations. A central goal of the Rn project at Rice University is to construct a prototype software development environment that is designed to build whole programs, rather than just individual modules. It employs interprocedural analysis and optimization to produce high-quality machine code for whole programs. This paper presents an overview of the methods used by the environment to accomplish this task and discusses the impact of these methods on the various environment components. The responsibilities of each component of the environment for the preparation and use of interprocedural information are presented in detail.