DPSKEL:: A skeleton based tool for parallel dynamic programming

Skeleton based libraries are considered one of the alternatives to reduce the distance between end users and parallel architectures. Algorithmic skeletons are based in general procedures describing the method to be implemented. Although a gap between general formalizations for dynamic programming and software components can be found, we develop a skeleton tool for dynamic programming problems. The design strategy is general enough to consider a wide rage of dynamic programming recurrences. As usual in skeleton approaches, the parallelism is provided in a transparent manner, so that, sequential users may access to the system. A set of tests problems representative of different classes of dynamic programming formulations has been used to validate the distributed memory implementation on an IBM-SP.