Performance of Coupled-Cluster Singles and Doubles on Modern Stream Processing Architectures

B Scott Fales; Ethan R Curtis; K Grace Johnson; Dean Lahana; Stefan Seritan; Yuanheng Wang; Hayley Weir; Todd J Martínez; Edward G Hohenstein

doi:10.1021/acs.jctc.0c00336

Performance of Coupled-Cluster Singles and Doubles on Modern Stream Processing Architectures

J Chem Theory Comput. 2020 Jul 14;16(7):4021-4028. doi: 10.1021/acs.jctc.0c00336. Epub 2020 Jun 29.

Authors

B Scott Fales^{1

2}, Ethan R Curtis^{1

2}, K Grace Johnson^{1

2}, Dean Lahana^{1

2}, Stefan Seritan^{1

2}, Yuanheng Wang^{1

2}, Hayley Weir^{1

2}, Todd J Martínez^{1

2}, Edward G Hohenstein^{1

2}

Affiliations

¹ Department of Chemistry and The PULSE Institute, Stanford University, Stanford, California 94305, United States.
² SLAC National Accelerator Laboratory, 2575 Sand Hill Road, Menlo Park, California 94025, United States.

PMID: 32567305
DOI: 10.1021/acs.jctc.0c00336

Abstract

We develop a new implementation of coupled-cluster singles and doubles (CCSD) optimized for the most recent graphical processing unit (GPU) hardware. We find that a single node with 8 NVIDIA V100 GPUs is capable of performing CCSD computations on roughly 100 atoms and 1300 basis functions in less than 1 day. Comparisons against massively parallel implementations of CCSD suggest that more than 64 CPU-based nodes (each with 16 cores) are required to match this performance.