Automatic Generation of Nested, Fork-Join Parallelism

摘要: This paper presents an efficient algorithm that automatically generates a parallel program from dependence-based representation of sequential program. The resulting consists nested fork-join constructs, composed the loops and statements Data dependences are handled by two techniques. One technique implicitly satisfies them sequencing, thereby reducing parallelism. Where increased parallelism results, other eliminates privatization: introduction process-specific private instances variables. Additionally, determines when copying values such in out constructs results greater is first for generating general model. as much possible our model while minimizing privatization.