Accelerating many-nucleon basis generation for high performance computing enabled ab initio nuclear structure studies

0505344 - ÚJF 2020 RIV GB eng J - Journal Article
Langr, D. - Dytrych, Tomáš - Launey, K. D. - Draayer, J. P.
Accelerating many-nucleon basis generation for high performance computing enabled ab initio nuclear structure studies.
International Journal of High Performance Computing Applications. Roč. 33, č. 3 (2019), s. 522-533. ISSN 1094-3420. E-ISSN 1741-2846
R&D Projects: GA ČR GA16-16772S
Institutional support: RVO:61389005
Keywords : Ab initio * high performance computing * nuclear structure * many-nucleon basis generation * parallel algorithm * symmetry-adapted no-core shell model
OECD category: Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)
Impact factor: 2.365, year: 2019
Method of publishing: Limited access
https://doi.org/10.1177/1094342019838314

We present the problem of generating a many-nucleon basis in SU(3)-scheme for ab initio nuclear structure calculations in a symmetry-adapted no-core shell model framework. We first discuss and analyze the basis construction algorithm whose baseline implementation quickly becomes a significant bottleneck for large model spaces and heavier nuclei. The outcomes of this analysis are utilized to propose a new scalable version of the algorithm. Its performance is consequently studied empirically using the Blue Waters supercomputer. The measurements show significant acceleration achieved with over two orders of magnitude speedups realized for larger model spaces.
Permanent Link: http://hdl.handle.net/11104/0296791