ATTAINABLE COMPUTATION RATES ONCOMPUTERS WITH SERIAL ARCHITECTURE ARE A SERIOUS LIMITATION FOR THE SIMULATION OF THE HYDRODYNAMIC BEHAVIOR OF FLUIDS.

ATTAINABLE COMPUTATION RATES ONCOMPUTERS WITH SERIAL ARCHITECTURE ARE A SERIOUS LIMITATION FOR THE SIMULATION OF THE HYDRODYNAMIC BEHAVIOR OF FLUIDS. COMPUTERS WITH PARALLEL PROCESSING ARCHITECTURE CAN ACHIEVE MUCH HIGHER RATES. THE COMPUTER RESEARCH AND APPLICA TIONS GROUP AT LOS ALAMOS NATIONAL LABORATORIES IS MAKING THEIR DENELCOR HEP COMPUTER AVAILABLE FOR RESEARCH PROJECTS CONDUCTED BY OTHER INSTITUTIONS. THIS MACHINE IS A PARALLEL PROCESSOR. FOR A PHASE I RESEARCH CONTRACT, KMS FUSION PROPOSES A THREE-STEP PLAN FOR THE DEVELOPMENT OF HYDROCODES ON THE HEP. FIRST, WE WILL WRITE AN EXPLICIT, ONE-DIMENSIONAL LAGRANGIAN CODE. IN THIS CASE, THE DIFFERENCE EQUATIONS CAN BE READILY FORMULATED FOR CONCURRENT PROCESSING. SECOND, FAN IMPLICIT VERSIONWILL BE DEVELOPED IN ORDER TO ACHIEVE UNCONDITIONAL STABILITY. PARALLELIZATION WILL BE ACCOMPLISHEDVIA AN A PRIORI, SYMBOLIC INVERSION OF THE UNDERLYING LINEAR SYSTEM. THIRD, A REZONING ALGORITHM WILL BE USED AFTER LAGRANGIAN TIME STEP TO CONVERT THE PROGRAM TO AN IMPLICIT EULERIAN HYDROCODE.