Benchmarks#

Parallelization#

Very complex and expensive computations are parallelized in AMEP to also run efficiently on high-performance computing hardware. The following table gives an overview on all observables that can be calculated in parallel with AMEP.

Observable:

AMEP evaluate object:

Hexagonal order correlation function

amep.evaluate.HexOrderCor

2d pair correlation function

amep.evaluate.PCF2d

Angular pair correlation function

amep.evaluate.PFCangle

Radial pair distribution function

amep.evaluate.RDF

2d static structure factor

amep.evaluate.SF2d

Isotropic static structure factor

amep.evaluate.SFiso

Spatial velocity correlation function

amep.evaluate.SpatialVelCor

The scaling shown in the image below was run with two Intel® Xeon® Platinum 9242 Processors (96 cores in total) on the Lichtenberg high-performance computer at Technical University of Darmstadt. It is based on the total runtime for evaluating a Brownian dynamics simulation of 100,000 active Brownian particles done with LAMMPS.

Scaling behavior of parallel methods in AMEP on the Lichtenberg high-performance computer at Technical University of Darmstadt.