Gaussian Basis Set Hartree–Fock, Density Functional Theory, and Beyond on GPUs

GPU acceleration of all-electron electronic structure theory using localized numeric atom-centered basis functions

“…GPGPUs are well suited for evaluation of highly vectorizable, compute-intensive algorithms due to their unique massively-parallel design, prompting early work to demonstrate matrix multiplications [38] and FFTs [39] on commodity hardware. The feasibility of GPU-accelerated electronic structure calculations was demonstrated in works by Ufimtsev and Martínez [40,41,42,43] which would form the basis for the TeraChem electronic structure package [44], as well as by Yasuda using the Gaussian package [45,46]. Since then, a number of electronic structure packages have incorporated GPU acceleration, including ABINIT [47], ADF [48], BigDFT [49,50], CP2K [51], GPAW [52,53,54], LS3DF [55], octopus [56,57,58], ONETEP [59], PEtot (now PWmat) [60,61,62,63], Q-Chem [64,65,15], Quantum ESPRESSO [66,67], RMG [68,69], VASP [70,71,72,73], and FHI-aims (this work).…”

Achieving performance portability in Gaussian basis set density functional theory on accelerator based architectures in NWChemEx

“…For atom-centered bases, the formation of the KS-DFT Fock matrix is dominated by the formation of the Coulomb matrix, exchangecorrelation (XC) matrix, and the exact-exchange matrix in the case of hybrid KS-DFT methods. Recently, a significant amount of research has been directed to the effecient evalution of the Coulomb and exact exchange matrices in Gaussian basis sets on GPU architectures [27,[31][32][33][34][35][36][37][38], while relatively little effort has been afforded to the evaluation of the XC matrix [24,25,27,28]. As such, we focus on the portable implementation of the XC matrix in this work.…”

On the Efficient Evaluation of the Exchange Correlation Potential on Graphics Processing Unit Clusters

“…Over the years, there has been a considerable amount of research devoted to porting implementations of Gaussian basis set KS-DFT to the GPU (Yasuda, 2008 ; Brown et al, 2010 ; Titov et al, 2013 ; Luehr et al, 2016 ; Kussmann and Ochsenfeld, 2017 ; Manathunga et al, 2020 ; Peters et al, 2020 ); however, the vast majority of this work has been centered around the evaluation and digestion of the ERIs in the construction of the Fock matrix (Ufimtsev and Martinez, 2008 , 2009a , b ; Asadchev et al, 2010 ; Miao and Merz, 2013 ; Kalinowski et al, 2017 ; Kussmann and Ochsenfeld, 2017 ; Laqua et al, 2020 ). On the other hand, the XC potential has received much less treatment in the literature in this regard (Yasuda, 2008 ; Luehr et al, 2016 ; Manathunga et al, 2020 ). This disparity is understandable due to the fact that for large systems, the ERI-related contributions to the Fock matrix are computationally dominant and the most challenging to parallelize.…”

“…Over the years, there has been a considerable amount of research devoted to porting implementations of Gaussian basis set KS-DFT to the GPU [45,53,66,51,83,72,11]; however, the vast majority of this work has been centered around the evaluation and digestion of the ERIs in the construction of the Fock matrix [39,45,77,75,76,54,6,47]. On the other hand, the XC potential has received much less treatment in the literature in this regard [83,53,51]. This disparity is understandable due to the fact that for large systems, the ERI related contributions to the Fock matrix are computationally dominant and the most challenging to parallelize.…”

“…Of the required integrals, the electron repulsion integrals (ERIs) and the exchange-correlation (XC) potential are among the most costly and the most challenging to port to GPU architectures. Over the years, there has been a considerable amount of research devoted to porting implementations of Gaussian basis set KS-DFT to the GPU [45,53,66,51,83,72,11]; however, the vast majority of this work has been centered around the evaluation and digestion of the ERIs in the construction of the Fock matrix [39,45,77,75,76,54,6,47]. On the other hand, the XC potential has received much less treatment in the literature in this regard [83,53,51].…”