Hierarchical Program Analysis in High Performance Compilers for Parallel Computers

本研究将探索编制大型数值程序的新策略。使用分层控制流图来促进信息的快速传播将是采用的主要技术。这将利用程序控制中的层次结构，从而允许高性能计算中的自动并行化。为了证明这种方法的重要性，将开发一个新的编译器。

New strategies for compiling large numerical programs will be explored in this research. The use of hierarchical control flow graphs to facilitate fast information propagation will be the major technique employed. This will exploit hierarchies in program control thus allowing automatic parallelization in high performance computing. A new compiler will be developed in order to demonstrate the significance of this approach.