Cooperation with HemeLB

“HemeLB is a high performance lattice-Boltzmann solver optimized for simulating blood flow through sparse geometries, such as those found in the human vasculature.” [1] The code is used within the CompBioMed HPC Centre of Excellence H2020 project [2], and is already highly optimized for HPC usage. As a consequence of the initial workshop about the ALL library hosted by JSC in the context of E-CAM, a cooperation was set up in order analyse and test whether the use of ALL could improve the existing scalability of the code.

Purpose of Module

This module describes the cooperation between the ALL library and the HemeLB code, from the CompBioMed HPC Centre of Excellence. It provides details about the work performed an the results of the cooperation.

Background Information / Bibliography

More information about HemeLB can be found at [1], while more information about CompBioMed can be found at [2].

[1](1, 2), 03.02.2021
[2](1, 2)

Cooperation content

As ALL was designed to work with particle codes, it was interesting to apply the library to an lattice-Boltzmann solver, which usually is not particle-based. Therefore the different grid points of the solution grid were designated as particles and since each of the grid-points already was assigned a workload, the sum of grid-point workloads could be used as domain work load. Since the creation and change of domain boundaries during the simulation was deemed not feasible, an example code within ALL was used to create initial, balanced domain decompositions for various systems and the results compared to the already existing load-balancing solution within HemeLB. Since test runs were performed on a large scale, e.g. SuperMUC, it was necessary to also provide MPI-I/O based output for better parallel I/O efficiency.

Cooperation results

As a result it can be stated that the domain compositions provided by ALL show a better theoretical load distribution. Tests to check if this translates into better code performance are inconclusive as yet, due to hardware related issues on the testing platforms. These are currently under furtherinvestigation, and more definitive results about the performance of the ALL-provided domain decompositions can be expected in the near future.

The results were part of a publication about HemeLB, which was published in 2020 [Coveney].

[Coveney]Coveney, P. V. et al., Towards blood flow in the virtual human: efficient self-coupling of HemeLB, Interface focus 11(1), 20190119