Open-source data structure for distributed unstructured homogeneous meshes with GPU support

What is it about?

The paper describes a general data structure for an efficient representation of unstructured meshes for scientific computations. The data structure is provided as open-source software as part of the TNL library (https://tnl-project.org/). The main features are configurability to different applications, support for computations on CPUs as well as GPUs, and support for distributed computing using MPI. The paper describes the design decisions and implementation details, internal storage layout, and analysis of the efficiency on several benchmark problems. The applicability to advanced numerical methods is demonstrated with an example problem of two-phase flow in porous media using a numerical scheme based on the mixed-hybrid finite element method (MHFEM).

Featured Image

Photo by Nick Fewings on Unsplash

Photo by Nick Fewings on Unsplash

Why is it important?

The presented data structure can be used as a fundamental building block in advanced numerical simulation codes written in C++ and/or CUDA. The data structure natively supports modern GPUs which can greatly accelerate the time to solution of many problems. For several benchmark problems, we show GPU speed-ups that rise above 20 in 2D and 50 in 3D when compared to sequential CPU computations, and above 2 in 2D and 9 in 3D when compared to 12-threaded CPU computations.