{"language":[{"iso":"eng"}],"publication":"arXiv:2104.08245","date_updated":"2022-06-28T07:49:31Z","type":"preprint","author":[{"first_name":"Robert","last_name":"Schade","full_name":"Schade, Robert"},{"last_name":"Kenter","full_name":"Kenter, Tobias","first_name":"Tobias"},{"first_name":"Hossam","full_name":"Elgabarty, Hossam","last_name":"Elgabarty"},{"full_name":"Lass, Michael","last_name":"Lass","first_name":"Michael"},{"first_name":"Ole","full_name":"Schütt, Ole","last_name":"Schütt"},{"first_name":"Alfio","full_name":"Lazzaro, Alfio","last_name":"Lazzaro"},{"first_name":"Hans","full_name":"Pabst, Hans","last_name":"Pabst"},{"first_name":"Stephan","full_name":"Mohr, Stephan","last_name":"Mohr"},{"first_name":"Jürg","last_name":"Hutter","full_name":"Hutter, Jürg"},{"full_name":"Kühne, Thomas D.","last_name":"Kühne","first_name":"Thomas D."},{"last_name":"Plessl","full_name":"Plessl, Christian","first_name":"Christian"}],"abstract":[{"lang":"eng","text":"We push the boundaries of electronic structure-based \\textit{ab-initio}\r\nmolecular dynamics (AIMD) beyond 100 million atoms. This scale is otherwise\r\nbarely reachable with classical force-field methods or novel neural network and\r\nmachine learning potentials. We achieve this breakthrough by combining\r\ninnovations in linear-scaling AIMD, efficient and approximate sparse linear\r\nalgebra, low and mixed-precision floating-point computation on GPUs, and a\r\ncompensation scheme for the errors introduced by numerical approximations. The\r\ncore of our work is the non-orthogonalized local submatrix method (NOLSM),\r\nwhich scales very favorably to massively parallel computing systems and\r\ntranslates large sparse matrix operations into highly parallel, dense matrix\r\noperations that are ideally suited to hardware accelerators. We demonstrate\r\nthat the NOLSM method, which is at the center point of each AIMD step, is able\r\nto achieve a sustained performance of 324 PFLOP/s in mixed FP16/FP32 precision\r\ncorresponding to an efficiency of 67.7% when running on 1536 NVIDIA A100 GPUs."}],"year":"2021","project":[{"name":"PC2: Computing Resources Provided by the Paderborn Center for Parallel Computing","_id":"52"}],"status":"public","date_created":"2022-06-28T07:48:31Z","citation":{"apa":"Schade, R., Kenter, T., Elgabarty, H., Lass, M., Schütt, O., Lazzaro, A., Pabst, H., Mohr, S., Hutter, J., Kühne, T. D., &#38; Plessl, C. (2021). Towards Electronic Structure-Based Ab-Initio Molecular Dynamics  Simulations with Hundreds of Millions of Atoms. In <i>arXiv:2104.08245</i>.","mla":"Schade, Robert, et al. “Towards Electronic Structure-Based Ab-Initio Molecular Dynamics  Simulations with Hundreds of Millions of Atoms.” <i>ArXiv:2104.08245</i>, 2021.","ama":"Schade R, Kenter T, Elgabarty H, et al. Towards Electronic Structure-Based Ab-Initio Molecular Dynamics  Simulations with Hundreds of Millions of Atoms. <i>arXiv:210408245</i>. Published online 2021.","ieee":"R. Schade <i>et al.</i>, “Towards Electronic Structure-Based Ab-Initio Molecular Dynamics  Simulations with Hundreds of Millions of Atoms,” <i>arXiv:2104.08245</i>. 2021.","chicago":"Schade, Robert, Tobias Kenter, Hossam Elgabarty, Michael Lass, Ole Schütt, Alfio Lazzaro, Hans Pabst, et al. “Towards Electronic Structure-Based Ab-Initio Molecular Dynamics  Simulations with Hundreds of Millions of Atoms.” <i>ArXiv:2104.08245</i>, 2021.","short":"R. Schade, T. Kenter, H. Elgabarty, M. Lass, O. Schütt, A. Lazzaro, H. Pabst, S. Mohr, J. Hutter, T.D. Kühne, C. Plessl, ArXiv:2104.08245 (2021).","bibtex":"@article{Schade_Kenter_Elgabarty_Lass_Schütt_Lazzaro_Pabst_Mohr_Hutter_Kühne_et al._2021, title={Towards Electronic Structure-Based Ab-Initio Molecular Dynamics  Simulations with Hundreds of Millions of Atoms}, journal={arXiv:2104.08245}, author={Schade, Robert and Kenter, Tobias and Elgabarty, Hossam and Lass, Michael and Schütt, Ole and Lazzaro, Alfio and Pabst, Hans and Mohr, Stephan and Hutter, Jürg and Kühne, Thomas D. and et al.}, year={2021} }"},"external_id":{"arxiv":["2104.08245"]},"user_id":"15278","department":[{"_id":"27"}],"title":"Towards Electronic Structure-Based Ab-Initio Molecular Dynamics  Simulations with Hundreds of Millions of Atoms","_id":"32244"}