Dresden 2026 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
MA: Fachverband Magnetismus
MA 34: Computational Magnetism I
MA 34.1: Vortrag
Mittwoch, 11. März 2026, 15:00–15:15, HSZ/0004
GPU acceleration of atomistic spin dynamics simulations — Mariia Mohylna1, Anders Bergman2, Arkadijs Slobodkins2, Olle Eriksson2, Anna Delin1, and •Johan Hellsvik1 — 1KTH Royal Institute of Technology, Stockholm, Sweden — 2Uppsala University, Uppsala, Sweden
The Uppsala Atomistic Spin Dynamics (UppASD) [1,2] code implements algorithms for modelling magnetic materials on microscopic and mesoscopic length scales. In our recent implementation, kernels for magnetic field calculation for bilinear spin-spin pair interaction and one numerical integrator have been implemented with the CUDA programming model for use on NVIDIA GPUs, and the HIP programming model for AMD GPUs. Benchmarks for the simulation of bulk bcc Fe and the spin glass CuMn establish firmly the proposition of high speed-up for magnetization dynamics simulation on GPU hardware. The superior performance of the GPU implementations highlights the compatibility of spin-related problems with GPU architecture.
[1] The UppASD code https://github.com/UppASD/UppASD [2] Atomistic spin-dynamics; foundations and applications, O. Eriksson, et al. (Oxford University Press, 2017).
Keywords: atomistic spin dynamics; GPU acceleration; spin glass