Abstract
High-resolution simulations, such as those of plasma dynamics in the Earth's magnetosphere, generate vast amounts of data that challenge storage and processing capabilities, especially in the Exascale computing era. Managing these data volumes requires innovative solutions that balance compression efficiency with the accuracy required for scientific analysis. In this work, we present a novel approach for lossy runtime data compression tailored for high-dimensional velocity distribution functions (VDFs) using the Vlasiator code. Our method integrates implicit neural representations, specifically Multi-Layer Perceptrons (MLPs), to train on and compress VDFs dynamically during the simulation runtime. We address the challenges of preserving physical correctness while achieving high compression ratios by employing techniques such as Fourier feature encoding and sinusoidal activation functions to mitigate spectral bias. We compare our method with industry-standard methods such as ZFP and demonstrate significant improvements in compression efficiency without compromising physical correctness.
| Original language | English |
|---|---|
| Article number | 113905 |
| Journal | Physics of Plasmas |
| Volume | 32 |
| Issue number | 11 |
| DOIs | |
| Publication status | Published - 1 Nov 2025 |
| MoE publication type | A1 Journal article-refereed |
Funding
This work was conducted within the project Adaptive Strategies Towards Expedient Recovery In eXascale (ASTERIX). Innovation Study ASTERIX has received funding through the Inno4scale project, which is funded by the European High-Performance Computing Joint Undertaking (JU) under Grant Agreement No. 101118139. The JU receives support from the European Union's Horizon Europe Programme. The work presented in this paper would not have been possible without the high-performance computing resources provided by the CSC—IT Center for Science (CSC). The verification of the methods was conducted on the LUMI supercomputer. The authors also wish to acknowledge the Oregon Advanced Computing Institute for Science and Society (OACISS). The performance tests for NVIDIA hardware were run on the Saturn and Hopper1/2 supercomputing nodes. The Odyssey node was also used for profiling runs on AMD hardware. The work of AK was partially supported by the Research Council of Finland Grant No. 355460. MP acknowledges the Research Council of Finland Grant Nos. 361901, 347795, and 352846. MA acknowledges the Research Council of Finland Grant Nos. 352846 and 361901 and the Inno4Scale ASTERIX project. KP also wants to thank AK and Ivan Zaitsev for the very helpful and inspiring conversations during this work.