PIERS timings on various parallel supercomputers

Abstract

A parallel implicit research reservoir simulator (PIERS), designed for experimentation rather than production yet with many of the features of realistic petroleum reservoir models, was made available to the authors as a test program. This article reports results of timing PIERS on five parallel supercomputers. Graphs illustrate performance as a function of the number of processors for several problem sizes. Results indicate that parallel computation provides substantial speed improvement in reservoir simulator codes. Also, the raw timings provide an interesting comparison of machine performance, though one should remember that they apply directly only to PIERS and may vary with different simulators, compilers, operating systems, and hardware configurations.