Toward Parallel Modeling of Solidification Based on the Generalized Finite Difference Method Using Intel Xeon Phi

Modern heterogeneous computing platforms have become powerful HPC solutions, which could be applied for a wide range of applications. In particular, the hybrid platforms equipped with Intel Xeon Phi coprocessors offers performance advantages over conventional homogeneous solutions based on CPUs, while supporting practically the same parallel programming model. However, there is still an open issue how scientific applications can utilize efficiently the hybrid platforms equipped with Intel coprocessors. In this paper we propose a method for porting a real-life scientific application to computing platforms with Intel Xeon Phi. We focus on the parallel implementation of a numerical model of solidification, which is based on the generalized finite difference method. We develop a sequence of steps that are necessary for porting this application to platforms with accelerators, assuming no significant modifications of the code. The proposed method considers not only efficient data transfers that allow for overlapping computations with data movements, but also takes into account an adequate utilization of cores/threads and vector units. The developed approach allows us to execute the whole application 3.45 times faster than the original parallel version running on two CPUs.