Scaling the Hartree-Fock Matrix Build on Summit

Giuseppe M.J. Barca, David L. Poole, Jorge L. Galvez Vallejo, Melisa Alkan, Colleen Bertoni, Alistair P. Rendell, Mark S. Gordon

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Citation (Scopus)

Abstract

Usage of Graphics Processing Units (GPU) has become strategic for simulating the chemistry of large molecular systems, with the majority of top supercomputers utilizing GPUs as their main source of computational horsepower. In this paper, a new fragmentation-based Hartree-Fock matrix build algorithm designed for scaling on many-GPU architectures is presented. The new algorithm uses a novel dynamic load balancing scheme based on a binned shell-pair container to distribute batches of significant shell quartets with the same code path to different GPUs. This maximizes computational throughput and load balancing, and eliminates GPU thread divergence due to integral screening. Additionally, the code uses a novel Fock digestion algorithm to contract electron repulsion integrals into the Fock matrix, which exploits all forms of permutational symmetry and eliminates thread synchronization requirements. The implementation demonstrates excellent scalability on the Summit computer, achieving good strong scaling performance up to 4096 nodes, and linear weak scaling up to 612 nodes.

Original languageEnglish
Title of host publicationProceedings of SC20
Subtitle of host publicationThe International Conference for High Performance Computing, Networking, Storage and Analysis: Virtual Event, November 9-19, 2020
PublisherInstitute of Electrical and Electronics Engineers Inc.
Number of pages14
ISBN (Electronic)9781728199986
DOIs
Publication statusPublished - 2020
Event2020 International Conference for High Performance Computing, Networking, Storage and Analysis - Virtual Conference, Atlanta, United States
Duration: 9 Nov 202019 Nov 2020

Conference

Conference2020 International Conference for High Performance Computing, Networking, Storage and Analysis
Abbreviated titleSC20
Country/TerritoryUnited States
CityAtlanta
Period9/11/2019/11/20

Keywords

  • GPU
  • Hartree-Fock
  • Summit

Fingerprint

Dive into the research topics of 'Scaling the Hartree-Fock Matrix Build on Summit'. Together they form a unique fingerprint.

Cite this