Publication Details
Weighted Decomposition in High-Performance Lattice-Boltzmann Simulations: Are Some Lattice Sites More Equal than Others?
Chacra David A. (UCL)
Nash Rupert W. (UCL)
Jaroš Jiří, doc. Ing., Ph.D. (DCSY FIT BUT)
Bernabeu Miguel O. (UCL)
Coveney Peter V. (UCL)
high performance computing; lattice-Boltzmann; domain decomposition
Obtaining a good load balance is a significant challenge in scaling up lattice-Boltzmann simulations of realistic sparse problems to the exascale. Here we analyse the effect of two optimization techniques on the performance of the HemeLB lattice-Boltzmann simulation environment, when applied to sparse domains. Both techniques are applied prior to partitioning our domain using a third party library. One technique is to provide wall and in/outlet sites with increased weights, and the other is to sort the lattice sites according to a space filling curve. We tested these strategies on a sparse bifurcation and very sparse aneurysm geometry, and find that using weights reduces calculation load imbalance by up to 85%, although the overall communication overhead is higher is some of our runs.
@INPROCEEDINGS{FITPUB10510,
author = "Derek Groen and A. David Chacra and W. Rupert Nash and Ji\v{r}\'{i} Jaro\v{s} and O. Miguel Bernabeu and V. Peter Coveney",
title = "Weighted Decomposition in High-Performance Lattice-Boltzmann Simulations: Are Some Lattice Sites More Equal than Others?",
pages = "28--38",
booktitle = "Solving Software Challengesfor Exascale",
series = "Lecture Notes in Computer Science",
volume = 8759,
year = 2015,
location = "Stockholm, SE",
publisher = "Springer Verlag",
ISBN = "978-3-319-15975-1",
doi = "10.1007/978-3-319-15976-8\_2",
language = "english",
url = "https://www.fit.vut.cz/research/publication/10510"
}