A Domain Decomposition Based Parallel Implementation
Most implementations of parallel adaptive codes for distributed memory computers are based on the domain decomposition (DD) paradigm. The computational mesh is partitioned into subdomains, with each subdomain assigned to a single processor. Each processor is executing a sequential code with minimal upgrades to support a necessary communication betweeen the subdomains. To maintain scalability of the parallel code, only local subdomain information may be stored in each processor’s memory.