john%ghostwheel.unm.edu@ariel.unm.edu (John Prentice) (12/14/90)
My group at Amparo Corporation is currently developing an Eulerian solid dynamics wavecode for the Air Force. Our main focus in this effort is on developing domain decomposition methods for solving the continuum equations of elastic/plastic solid dynamics using an operator splitting method which is fairly common in the condensed matter shock physics community. We are developing the code to run on the SunSparc and the Cray series of computers (as a serial code). Although not explicitly part of our current contract, we would very much like to design and write the code in such a way that it can be ported to either MIMD or SIMD parallel machines with a minimum of effort. We have a fair amount of experience with parallelism for other types of codes (mostly quantum mechanics related stuff), but less with finite difference parallelism. In effect, our code is a CFD code even though the physics is different. We are solving essentially the same Euler equations one would in CFD. So, I am interested in any references people might have about the use of parallel computers to do CFD calculations. In particular I am interested in articles which describe the algorithms selected, programming models used, performance specs, etc... If anyone knows of such references or has direct experience with this type of thing, we would very much appreciate the info. Thanks. John Prentice Computational Physics Groups Amparo Corporation Albuquerque, NM john@unmfys.unm.edu