ABSTRACT
Large-scale multi-physics computational simulations often provide insight into complex problems that both complement experiments and help define future physical and computational experiments [13]. A good example of a production-strength code is Uintah [13, 12]. The code is designed to solve reacting fluid-structure problems involving large deformations and fragmentation. The underlying methods inside Uintah are a combination of standard fluid-flow methods and material point (particle) methods. In the case of codes, like Uintah, which solve large systems of partial differential equations on a mesh, refining the mesh increases the accuracy of the simulation. Unfortunately refining a mesh by a factor of two increases the work by a factor of 2d, where d is the dimensionality of the problem. This rapid increase in
computational effort severely limits the accuracy attainable on a particular parallel machine.
