Parallel programming of a resevoir simulator

Abstract

This study concerns applying parallel programming to reservoir simulation using a 32-Mbyte, 12-processor parallel computer. The effects of number of processes, granularity, load balancing and program structure were studied. The model simulated was a two-dimensionals, two-phase, black oil model with a fully-implicit formulation. The differenced equations were solved by the Newton-Raphson method and, Gaussian elimination was used to solve the Jacobian matrix. Matrix generation was parallelized using monitors as macros to synchronize calculation. The performance of the simulator was measured by the speed up. The speed ups of the matrix generation time increased almost linearly with increasing number of processes. For all of the models tested, the speed ups ranged from 3.5 to 4.0 for four processes and 7.0 to 7.9 for eight proceses.