An iteration method for high order finite element schemes. Part III. On parallel implementation

The parallel variant of iterative method for hp-schemes obtained in the case of approximation of differential equations on unstructured grids with the help of finite element method is introduced. Represented parallel algorithm is intended for work on parallel computer with distributed memory (MIMD-architecture). Scalability of parallel code is tested on the first boundary problem for Poisson equation. Reasons why scalability on some stages of algorithm is degraded are analyzed and the ways of overcoming are offered. As parallel programming environments MPI is used.