<span class="var-sub_title">Optimization of a Lattice Boltzmann Program</span> SC18 Proceedings

The International Conference for High Performance Computing, Networking, Storage, and Analysis

Women in HPC: Diversifying the HPC Community

Optimization of a Lattice Boltzmann Program

Authors: Muhong Zhou (Numerical Algorithms Group, BP Center For High Performance Computing)

Abstract: In this project, I optimized a Lattice Boltzmann (LB) program developed internally by the digital rock team at BP to simulate single-phase flow and predict permeability of porous rocks. LB algorithm is simple to formulate and implement, and it is easily parallelizable on computer clusters. My work in this project includes adding parallel input/output, building solid side-wall boundary condition, and optimizing code runtime through improving input load balance and automatic loop vectorization. These runtime optimizations lead to a 3x speedup over the current (now previous) production LB code.

Archive Materials

Back to Women in HPC: Diversifying the HPC Community Archive Listing

Back to Full Workshop Archive Listing