Implementation of spectral-finite difference method for simulation of stratified turbulent flows on distributed memory multiprocessors
Rajat P. Garg, Joel H. Ferziger, Stephen G. Monismith · PPSC · 1995
The parallel implementation of a spectral finite difference algorithm for simulation of stratified turbulent flows on iPSC/860 Hypercube and Paragon XP/S parallel computers is presented. A single-program multiple data abstraction is used in conjunction with static data partitioning scheme. Performance measurements of the overall algorithm are presented for three different uni-partitioning schemes and are discussed in the context of associated dependency and communication overheads. The timing measurements show that Paragon is about 60% faster than iPSC/860 but the iPSC/860 shows a better speedup efficiency. Unscaled speedup efficiency of up to 91% was obtained on iPSC/860.