HPC MSU

Publication Abstract

Large Scale Parallel Lattice Boltzmann Model of Dendritic Growth

Jelinek, B., Eshraghi, M., & Felicelli, S. D. (2013). Large Scale Parallel Lattice Boltzmann Model of Dendritic Growth. San Antonio, TX: TMS.

Abstract

We present a parallel lattice Boltzmann - cellular automaton model for the simulation of two-dimensional dendritic growth during solidification of binary alloys. The model incorporates the effects of solute and energy transport under melt convection. The Message Passing Interface (MPI) standard is applied to parallelize the model, together with state-of-the-art Hierarchical Data Format 5 (HDF5) to improve output performance and visualization. The parallel performance of the algorithm is assessed utilizing the NSF Extreme Science and Engineering Discovery Environment (XSEDE) supercomputing resources. A loss in the parallel efficiency is observed on multi-core architectures when all cores of a computer node are utilized. Very good strong scaling up to thousands of cores is obtained across the nodes of a computer cluster, along with near-perfect weak scaling up to ten thousand cores. The developed solidification model shows good scalability up to centimeter-size domains containing millions of dendrites.