Reproduce “A three-dimensional domain decomposition method for large-scale DFT electronic structure calculations” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Truong Vinh Truong Duy, Taisuke Ozaki
This work introduces a three-dimensional domain decomposition method for Density Functional Theory (DFT) calculations, integrating both atom and grid partitioning for massively parallel computing. The approach achieves high parallel efficiency on supercomputers like the K computer, enabling simulations of systems up to 131,072 atoms with robust scalability and reduced communication overhead for 3D Fast Fourier Transforms.
Key insights
- The combination of inertia tensor-based recursive bisection for atom distribution and a tailored grid decomposition effectively balances workload and preserves data locality across diverse atomic systems.
- A hybrid 1D/2D grid decomposition strategy for 3D FFTs significantly improves communication efficiency compared to traditional 2D methods, particularly for intermediate numbers of processes.
- The proposed decomposition is broadly applicable to both conventional O(N^3) and linear scaling O(N) DFT methods, enabling robust and efficient calculations on extreme-scale computing resources.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.