Building an object oriented problem solving environment for the parallel numerical solution of PDEs

Traditionally, the development of parallel implementations of algorithms for the Numerical Solution of Partial Differential Equations (PDE) and Linear Algebra Problems is based on software libraries; typically, Fortran libraries. Due to the lack of abstraction of Fortran-like libraries, two main problems are encountered: the libraries are complex to develop and the library interfaces are difficult to use. This PhD project proposes to investigate (1) the advantages and disadvantages that object oriented design and programming can bring to the Numerical Solution of PDEs and Numerical Linear Algebra (NLA) and (2) the integration of PDE and NLA object oriented libraries with intelligent Agents to form the basis of a Problem Solving Environment (PSE) for Numerical Computation.