Grid‐Based Projector‐Augmented Wave Method

Hakala, Samuli; Enkovaara, Jussi; Havu, Ville; Yan, Jun; Li, Lin; O'Grady, C.P.; Nieminen, Risto M.

doi:10.1002/9781118670712.ch9

Cited by 1 publication

(1 citation statement)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…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). Development cost can be alleviated by using dropin GPU-accelerated libraries such as cuBLAS [74], cuFFT [75], Thrust [76], ELPA [77,78,79], and MAGMA [80,81,82]…”

Section: Introductionmentioning

confidence: 99%

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

Huhn

Lange

et al. 2020

Computer Physics Communications

View full text Add to dashboard Cite

We present an implementation of all-electron density-functional theory for massively parallel GPGPU-based platforms, using localized atom-centered basis functions and real-space integration grids. Special attention is paid to domain decomposition of the problem on non-uniform grids, which enables compute-and memory-parallel execution across thousands of nodes for realspace operations, e.g. the update of the electron density, the integration of the real-space Hamiltonian matrix, and calculation of Pulay forces. To assess the performance of our GPGPU implementation, we performed benchmarks on three different architectures using a 103-material test set. We find that operations which rely on dense serial linear algebra show dramatic speedups from GPGPU acceleration: in particular, SCF iterations including force and stress calculations exhibit speedups ranging from 4.5 to 6.6. For the architectures and problem types investigated here, this translates to an expected overall speedup between 3-4 for the entire calculation (including non-GPU accelerated parts), for problems featuring several tens to hundreds of atoms. Additional calculations for a 375-atom Bi 2 Se 3 bilayer show that the present GPGPU strategy scales for large-scale distributed-parallel simulations.

show abstract