Parallel Reconstruction System Of Global Flow Analysis

Parallel compiling has become an essential means to exploit the vast computing power of modern supercomputers. The dependence analysis Is the key technique of a parallel compiler. The global flow analysis affects the accuracy of dependence analysis, the strategy of optimization and the efficiency of parallel transforming, so it Is Important too. This thesis deals with the flow analysis In the front-end of a parallel compiler.First, it illustrates the principle and algorithm of flow analysis.second, it Introduces a practical global flow analysis (GFA) In the front-end of the Parallelizing Optimizing Restructuring Tools (PORT), which was developed on SGI power (4 cpu), by my advisor professor Zhang Zhao qing and her fellow people, at the Institute of Computing Technology, the Chinese Academy of Science. GFA absorbed other parallel compiler's quintessence, converted control dependence Into data dependence, so It deleted GOTO statements and could parallelize do loops which contain IF statements. In the meanwhile, GFA had its own features: It realized accurate interprocedural data flow analysis, which was avoided by many other parallel compilers, it used new measure to handle common space, so it could parallelize DO loops which contain CALL statements; it employed optimized algorithm and technique to minimize the time and space expense; It contained many functions which other parallel compilers did not have, for example, the scalar renaming chain. Above all, the GFA is a practical, accurate, global and extending flow analysis. This thesis concentrates on its components, implementation and some advanced technologies.Last, a conclusion and prospection of parallel compiling Is given.