Dynamic Distributed Data (DDD) in a Parallel Programming Environment - Specification and Functionali

The parallel implementation of algorithms based on dynamic data structures (e.g. adaptive PDE solvers, sparse matrices) on architectures with distributed memory will unavoidably lead to various technical problems as for redundancy, consistency and load balancing. The module DDD is presented, which makes the development of dynamic parallel algorithms much easier. DDD consists of a formal model for describing distributed data, a functionality speci cation and a portable library implementation. The main goal of DDD is to provide a general and e cient interface connecting SPMD-style applications with various architectures and programming models.