MAPS: An Efficient Parallel Language for Scientific Computing

We propose to construct MAPS, a collection-oriented SPMD programming environment for writing parallel programs that will simplify the development and porting of many scientific algorithms, both structured and unstructured. This very general class of applications includes finite-difference, finite-element and finite-volume methods, multiblock and multigrid algorithms, PIC codes for simulating plasma physics, quadtree and octree algorithms, graph algorithms such as circuit simulation, sparse matrix algorithms, and many others.