Computational Acceleration on Density Functional Approach Code Using GPU
Authors:
Open Access
- 28 September 2026
- book chapter
- Published by Springer Nature
- p. 315-322
Abstract
The density functional theory (DFT) calculations, promising tools for analyzing and predicting the physical properties of molecules and solids, are currently used in various fields of material science. The acceleration of DFT calculations is crucial to promote the investigations. For this purpose, we have developed a GPU-accelerated DFT code with the plane wave basis set and the ultrasoft pseudopotential scheme. Through the analysis of the efficiency of the GPU implementation, we have shown that reducing data transfer between CPU and GPU is vital for achieving high efficiency. The reduction of data transfer has been achieved by utilizing the GPU memory with the data GPUization implementation.Keywords
This publication has 7 references indexed in Scilit:
- Evaluating Emerging AI/ML Accelerators: IPU, RDU, and NVIDIA/AMD GPUs2024
- On the Use of GPUs in Density Functional Theory Atomistic Simulations: A Case of Acceleration SuccessInternational Journal of Earth & Environmental Sciences, 2017
- The analysis of a plane wave pseudopotential density functional theory code on a GPU machineComputer Physics Communications, 2013
- Accelerating Density Functional Calculations with Graphics Processing UnitJournal of Chemical Theory and Computation, 2008
- Ab initio investigation of twin boundary motion in the magnetic shape memory Heusler alloy Ni2MnGaJournal of Materials Science, 2008
- Noncollinear magnetism in liquid oxygen: A first-principles molecular dynamics studyPhysical Review B, 2004
- Car-Parrinello molecular dynamics with Vanderbilt ultrasoft pseudopotentialsPhysical Review B, 1993