High-Performance, Graphics Processing Unit-Accelerated Fock Build Algorithm

Giuseppe M.J. Barca; Jorge L. Galvez-Vallejo; David L. Poole; Alistair P. Rendell; Mark S. Gordon

doi:10.1021/acs.jctc.0c00768

High-Performance, Graphics Processing Unit-Accelerated Fock Build Algorithm

Giuseppe M.J. Barca^*, Jorge L. Galvez-Vallejo, David L. Poole, Alistair P. Rendell, Mark S. Gordon

^*Corresponding author for this work

Research output: Contribution to journal › Article › peer-review

28 Citations (Scopus)

Abstract

We present a high-performance, GPU (graphics processing unit)-accelerated algorithm for building the Fock matrix. The algorithm is designed for efficient calculations on large molecular systems and uses a novel dynamic load balancing scheme that maximizes the GPU throughput and avoids thread divergence that could occur due to integral screening. Additionally, the code adopts a novel ERI digestion algorithm that exploits all forms of permutational symmetry, combines efficiently the evaluation of both Coulomb and exchange terms together, and eliminates explicit thread synchronization requirements. Performance results obtained using a number of large molecules reveal remarkable speedups up to 24.4× with respect to the QUICK GPU code and up to 237× with respect to the GAMESS CPU parallel code.

Original language	English
Pages (from-to)	7232-7238
Number of pages	7
Journal	Journal of Chemical Theory and Computation
Volume	16
Issue number	12
DOIs	https://doi.org/10.1021/acs.jctc.0c00768
Publication status	Published - 8 Dec 2020

Access to Document

10.1021/acs.jctc.0c00768

Cite this

@article{127aa70f64294223bef69b857d3d1a39,

title = "High-Performance, Graphics Processing Unit-Accelerated Fock Build Algorithm",

abstract = "We present a high-performance, GPU (graphics processing unit)-accelerated algorithm for building the Fock matrix. The algorithm is designed for efficient calculations on large molecular systems and uses a novel dynamic load balancing scheme that maximizes the GPU throughput and avoids thread divergence that could occur due to integral screening. Additionally, the code adopts a novel ERI digestion algorithm that exploits all forms of permutational symmetry, combines efficiently the evaluation of both Coulomb and exchange terms together, and eliminates explicit thread synchronization requirements. Performance results obtained using a number of large molecules reveal remarkable speedups up to 24.4× with respect to the QUICK GPU code and up to 237× with respect to the GAMESS CPU parallel code.",

author = "Barca, {Giuseppe M.J.} and Galvez-Vallejo, {Jorge L.} and Poole, {David L.} and Rendell, {Alistair P.} and Gordon, {Mark S.}",

note = "Publisher Copyright: {\textcopyright} ",

year = "2020",

month = dec,

day = "8",

doi = "10.1021/acs.jctc.0c00768",

language = "English",

volume = "16",

pages = "7232--7238",

journal = "Journal of Chemical Theory and Computation",

issn = "1549-9618",

publisher = "American Chemical Society",

number = "12",

}

TY - JOUR

T1 - High-Performance, Graphics Processing Unit-Accelerated Fock Build Algorithm

AU - Barca, Giuseppe M.J.

AU - Galvez-Vallejo, Jorge L.

AU - Poole, David L.

AU - Rendell, Alistair P.

AU - Gordon, Mark S.

PY - 2020/12/8

Y1 - 2020/12/8

N2 - We present a high-performance, GPU (graphics processing unit)-accelerated algorithm for building the Fock matrix. The algorithm is designed for efficient calculations on large molecular systems and uses a novel dynamic load balancing scheme that maximizes the GPU throughput and avoids thread divergence that could occur due to integral screening. Additionally, the code adopts a novel ERI digestion algorithm that exploits all forms of permutational symmetry, combines efficiently the evaluation of both Coulomb and exchange terms together, and eliminates explicit thread synchronization requirements. Performance results obtained using a number of large molecules reveal remarkable speedups up to 24.4× with respect to the QUICK GPU code and up to 237× with respect to the GAMESS CPU parallel code.

AB - We present a high-performance, GPU (graphics processing unit)-accelerated algorithm for building the Fock matrix. The algorithm is designed for efficient calculations on large molecular systems and uses a novel dynamic load balancing scheme that maximizes the GPU throughput and avoids thread divergence that could occur due to integral screening. Additionally, the code adopts a novel ERI digestion algorithm that exploits all forms of permutational symmetry, combines efficiently the evaluation of both Coulomb and exchange terms together, and eliminates explicit thread synchronization requirements. Performance results obtained using a number of large molecules reveal remarkable speedups up to 24.4× with respect to the QUICK GPU code and up to 237× with respect to the GAMESS CPU parallel code.

UR - http://www.scopus.com/inward/record.url?scp=85096624713&partnerID=8YFLogxK

U2 - 10.1021/acs.jctc.0c00768

DO - 10.1021/acs.jctc.0c00768

M3 - Article

SN - 1549-9618

VL - 16

SP - 7232

EP - 7238

JO - Journal of Chemical Theory and Computation

JF - Journal of Chemical Theory and Computation

IS - 12

ER -

High-Performance, Graphics Processing Unit-Accelerated Fock Build Algorithm

Abstract

Access to Document

Other files and links

Fingerprint

Cite this