Optimization of modeling the water-oil mixture filtration in elastic porous media for clusters with Xeon Phi accelerators

S. E. Kireev

Institute of Computational Mathematics and Mathematical Geophysics of Siberian Branch of Russian Academy of Sciences, Novosibirsk

Abstract: On the basis of a previously developed program for modeling the multiphase flows in finite-deformed porous media, a new parallel program optimized for clusters with Intel Xeon Phi accelerators is implemented. Several optimization techniques specific for such accelerators are considered and their effect on the program execution time is discussed. A comparison of the symmetric and offload programming models for the accelerators is performed. The parallelization speedup and efficiency are estimatedwhen using various numbers of cluster's nodes.

Keywords: high-performance computing, parallel programming, program optimization, Intel Xeon Phi accelerators, scalability.

UDC: 004.021

Received: 10.02.2015