Live Webcast 15th Annual Charm++ Workshop

Parallel computing has enjoyed considerable growth in recent years and parallel architectures are assuming an increasingly central role in information processing. The stimulus behind this growth has been the ever increasing demand for compute power to solve complex problems in the shortest possible time. This sustained growth has also manifested itself in the evolution of a diverse design space for parallel machines. In order to exploit the benefits offered by the multitude of architectures available it is important for parallel applications to be developed in a machine independent fashion making only the most generally applicable assumptions about the underlying architecture. This requirement has been addressed by the developement of the Converse parallel runtime system and the Charm++ parallel programming language. Though the growth in the field of parallel computing has been tremendous, it has not been optimal because effort is wasted in duplicating the capabilities required for parallel computing. Thus there is a need for writing more reusable libraries and modules, which would enhance the growth in the area manyfold. This thesis is an effort in this direction. It provides a library for the Adaptive Mesh Refinement (AMR) technique which is used in a variety of applications like cosmology, global atmospheric modelling, hyperbolic partial differential equations, shock hydrodynamics, etc. The library described in this thesis facilitates the implementation of applications using 1D, 2D and 3D AMR in parallel.