Automatic Generation of Interprocedural Data—Flow Analyzers and Optimizers

摘要

Interprocedural analysis and optimization are very important for compilers toexploit 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 anautomatic generator of interprocedural analyzers and optimizers - IGEN. In IGEN, first a newmodel to describe the interprocedural data-flow problem is designed. It is fit for all traditionaldata-flow problems and can be used in context-sensitive algorithms. Then, the structure andthe working environment of IGEN are described in detail. Finally, the implementation of IGENand our experimental result are given.