Conclusion and Future Work

The significant conclusions demonstrated herein are:

1. That there is a common underlying computational infrastructure for a wide family of parallel adaptive computation algorithms.

2. That a substantial decrease in effort in implementation for these important algorithms can be attained without sacrifice of performance through use of this computational infrastructure.

There is much further research needed to complete development of a robust and supportable computational infrastructure for adaptive algorithms. The existing versions of SDDA require extension and engineering. The programming abstractions for each solution method require enriching. It is hoped to extend the programming abstraction layer to support other adaptive methods such as wavelet methods. There is a need to do many more applications to define the requirements for the programming abstraction layer interfaces.