Design of a Meta-Parallelizer for Large Scientific Applications

The "classical parallelizers" integrate more and more sophisticated and costly parallelization techniques. As a result, they are limited by the size of the program they are able to compile and are not well suited to parallelize real scientiic programs whose parallelism detection complexity may diier very much from one code fragment to another. The goal of our work is the construction of a meta-parallelizer for shared memory MIMD machines, LPARAD (Large applications, PARallelizer, ADaptative), able to eeciently parallelize large scientiic programs with minimum help from the user in a minimal time. The implementation of LPARAD within the scope of the PAF PAF90] project validates our approach.