Open Collections

Kebriti, Sajedeh

University of British Columbia

Mesh generation requires substantial computational resources in terms of both CPU time and memory. Computational cost becomes especially pronounced when considering the scale of advanced industrial applications, in which mesh sizes can reach ten billion cells and growing. Examesh, introduced in 2019, aimed to develop a fast and reliable methodology for generating large-scale meshes. Presently, the software has demonstrated successful generation of 347 billion cells. However, the resulting mesh files for such large cell counts are expected to be on the order of terabytes (TB). Storing such massive mesh files is not practical. To address this challenge, we have enhanced the software’s performance by parallelizing through the integration of the Message Passing Interface (MPI) framework. This speeds up mesh generation, but perhaps more importantly generates the parallel mesh in situ, removing the requirement to store it on disk or transport it between machines. The most challenging part of parallelizing mesh generation software that we addressed in this research work is establishing topological relationships between mesh entities at the part boundaries. We successfully generated meshes in parallel with as many as hundreds of billions of cells on as many as thousands of processors.

Text

2024-06-24

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/88503

Mechanical Engineering

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/