Automatic generation of interprocedural data-flow analyzers and optimizers

Interprocedural analysis and optimization are very important for compilers to exploit parallelism of modern high-performance computer systems. But it is very complicated, easy to make mistakes and difficult to maintain and port. To solve the problem, we construct an automatic generator of interprocedural analyzers and optimizers—IGEN. In IGEN, first a new model to describe the interprocedural data-flow problem is designed. It is fit for all traditional data-flow problems and can be used in context-sensitive algorithms. Then, the structure and the working environment of IGEN are described in detail. Finally, the implementation of IGEN and our experimental result are given.