Computational Chemistry in Parallel

• Author / Creator

  Hennessey, Dylan

• General Purpose computing with Graphics Processing Units (GPGPU) is a method of com-putation that allows for calculations to be performed in parallel. Graphics Processing Units(GPU) allow a single basic computer to perform as well as a super computer cluster composedof only Central Processing Units (CPU), at a fraction of the cost. As quantum computa-tional chemistry scales very poorly with the size of systems, the use of GPGPU has been avery effective tool for quantum chemists.This thesis presents a program called cudaDFRATOM. This program, written in CUDAFortran, allows for the calculation and optimization of Well-Tempered Basis Sets (WTBS)at the four-component relativistic level of theory. Except for a few small initial calculationsand control flow, all work is performed entirely on the GPU.When calculating the two-electron integrals, cudaDFRATOM uses a very effective methodof utilizing as much of the GPU’s resources as possible and can calculate these integrals ap-proximately 20 times faster than a similar program running on only a CPU.Two different algorithms for forming the P and Q supermatrices are presented. Therelatively small size of these matrices and the need to use double-precision floating-pointoperations stunts the speedup of these algorithms such that they only attain a speedup ofabout 2 times over the CPU program. An overall speedup of about 4 times is achieved bycudaDFRATOM.In addition, new WTBS are calculated for non-relativistic levels of theory using therWTBS program and at the four-component relativistic level of theory using cudaDFRATOM.In both cases, a novel algorithm for automating the optimization of these basis sets was used.The algorithm finds a basis set that is as small a possible while still retaining sufficient ac-curacy to numerical calculations.

• Subjects / Keywords

  • CUDA
  • Chemistry
  • GPGPU

• Graduation date

  Fall 2018

• Type of Item

  Thesis

• Degree

  Master of Science

• DOI

  https://doi.org/10.7939/R3KH0FF7W

• License

  Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.