Abstract
This paper compares two parallelization methods for data flow networks. A strict decoupling between the networks' structural components and their evaluation is achieved using a pattern oriented architecture with generalized traverser objects. This architecture provides a clean basis for the implementation of parallel execution code using a) an explicit parallelization based on pthreads and b) an implicit parallelization using OpenMP. Both methods are then evaluated and compared to each other for different traversal heuristics.
Users
Please
log in to take part in the discussion (add own reviews or comments).