We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

physics.comp-ph

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Physics > Computational Physics

Title: Breaking the Exascale Barrier for the Electronic Structure Problem in Ab-Initio Molecular Dynamics

Abstract: The non-orthogonal local submatrix method applied to electronic-structure based molecular dynamics simulations is shown to exceed 1.1 EFLOP/s in FP16/FP32 mixed floating-point arithmetic when using 4,400 NVIDIA A100 GPUs of the Perlmutter system. This is enabled by a modification of the original method that pushes the sustained fraction of the peak performance to about 80%. Example calculations are performed for SARS-CoV-2 spike proteins with up to 83 million atoms.
Comments: 6 pages, 6 figures, 2 tables
Subjects: Computational Physics (physics.comp-ph); Materials Science (cond-mat.mtrl-sci); Chemical Physics (physics.chem-ph); Quantitative Methods (q-bio.QM)
Cite as: arXiv:2205.12182 [physics.comp-ph]
  (or arXiv:2205.12182v2 [physics.comp-ph] for this version)

Submission history

From: Robert Schade [view email]
[v1] Tue, 24 May 2022 16:29:49 GMT (27186kb,D)
[v2] Tue, 7 Jun 2022 12:28:57 GMT (28116kb,D)

Link back to: arXiv, form interface, contact.