Data-Oriented Language Implementation of Lattice-Boltzmann Method for Dense and Sparse Geometries

TL;DR

This paper introduces a data-oriented language approach for implementing the lattice-Boltzmann method, achieving high performance on GPUs for dense and sparse geometries with simplified code.

Contribution

The work presents a novel data-oriented language technique that simplifies complex memory access patterns while maintaining high performance in lattice-Boltzmann simulations.

Findings

Achieved 0.6-0.7 of maximum memory bandwidth on GPU

Performance of over 2.5 and 5.0 GLUPS for double and single precision

Close to state-of-the-art performance for large meshes

Abstract

The performance of lattice-Boltzmann solver implementations usually depends mainly on memory access patterns. Achieving high performance requires then complex code which handles careful data placement and ordering of memory transactions. In this work, we analyse the performance of an implementation based on a new approach called the data-oriented language, which allows the combining of complex memory access patterns with simple source code. As a use case, we present and provide the source code of a solver for D2Q9 lattice and show its performance on GTX Titan Xp GPU for dense and sparse geometries up to 4096 2 nodes. The obtained results are promising, around 1000 lines of code allowed us to achieve performance in the range of 0.6 to 0.7 of maximum theoretical memory bandwidth (over 2.5 and 5.0 GLUPS for double and single precision, respectively) for meshes of size above 1024 2 nodes, which is close to the current state-of-the-art. However, we also observed relatively high and sometimes difficult to predict overheads, especially for sparse data structures. The additional issue was also a rather long compilation, which extended the time of short simulations, and a lack of access to low-level optimisation mechanisms.