Jarmatz, Piet

Load balancing in particle simulations is a well-researched field, but its effect on molecular-continuum coupled simulations is comparatively less explored. In this work, we implement static load balancing into the macro-micro-coupling tool (MaMiCo), a software for molecular-continuum coupling, and demonstrate its effectiveness in two classes of experiments by coupling with the particle simulation software ls1 mardyn. The first class comprises a liquid-vapour multiphase scenario, modelling evaporation of a liquid into vacuum and requiring load balancing due to heterogeneous particle distributions in space. The second class considers execution of molecular-continuum simulations on heterogeneous hardware, running at very different efficiencies. After a series of experiments with balanced and unbalanced setups, we find that, with our balanced configurations, we achieve a reduction in runtime by 44% and 55% respectively.

In the scope of the dtec.bw project hpc.bw, innovative HPC hardware resources were procured to investigate their performance for HSU-relevant compute-intensive software. Benchmarks for different software packages were conducted, and respective results are reported and documented in the following, considering the Intel Xeon architecture used in the HPC cluster HSUper, AMD EPYC 7763 and ARM FX700.

The macro–micro-coupling tool (MaMiCo) is an open source C++ research software framework, designed to create molecular-continuum flow simulations in a modular way, i.e. with exchangeable solvers. It can be used to couple discrete particle dynamics codes with computational fluid dynamics solvers while retaining performance, especially for parallel execution on supercomputers. We present a new version of MaMiCo that extends its functionality by a multitude of new features, notably with dynamic handling of molecular dynamics simulation instances, automated error estimation, coupling interfaces to the community codes ls1 mardyn and OpenFOAM, a Python interface, support for machine learning modules and enhanced two-way coupling. These features of the new MaMiCo version impact several fields of computational science and can be employed to tackle open research questions in the future, such as efficient multiscale numerical simulation of multi-phase flows, or fault tolerance of coupled simulations on large-scale cluster systems.

Noise filtering in fluid dynamics enables stable, strongly-coupled molecular-continuum coupling despite potentially high thermal fluctuations. In this extended version of our conference paper “MaMiCo: Non-local Means Filtering with Flexible Data-Flow for Coupling MD and CFD” (ICCS 2021), we apply Non-Local Means filtering (NLM) in a novel space–time formulation to particle data. Our implementation in the Macro–Micro-Coupling tool (MaMiCo) features a flexible filter chain execution for HPC systems. We test it on 3D simulation data with multiple flow scenarios and continuum solvers, including a coupling to OpenFOAM. Our filtering results demonstrate that NLM not only has an excellent signal-to-noise ratio gain when extracting macroscopic flow information from particle ensembles, but also yields benefits for transient two-way coupling.

Transient molecular-continuum coupled flow simulations often suffer from high thermal noise, created by fluctuating hydrodynamics within the molecular dynamics (MD) simulation. Multi-instance MD computations are an approach to extract smooth flow field quantities on rather short time scales, but they require a huge amount of computational resources. Filtering particle data using signal processing methods to reduce numerical noise can significantly reduce the number of instances necessary. This leads to improved stability and reduced computational cost in the molecular-continuum setting. We extend the Macro-Micro-Coupling tool (MaMiCo) – a software to couple arbitrary continuum and MD solvers – by a new parallel interface for universal MD data analytics and post-processing, especially for noise reduction. It is designed modularly and compatible with multi-instance sampling. We present a Proper Orthogonal Decomposition (POD) implementation of the interface, capable of massively parallel noise filtering. The resulting coupled simulation is validated using a three-dimensional Couette flow scenario. We quantify the denoising, conduct performance benchmarks and scaling tests on a supercomputing platform. We thus demonstrate that the new interface enables massively parallel data analytics and post-processing in conjunction with any MD solver coupled to MaMiCo.