Effects of Easy Hybrid Parallelization with CUDA for Numerical-Atomic-Orbital Density Functional Theory Calculation
QoLT IIDC and School of Earth and Environmental Sciences, Seoul National University, Seoul 151-747, Korea
arXiv:1402.4247 [cs.DC], (18 Feb 2014)
@article{2014arXiv1402.4247P,
author={Parq}, J.-H. and {Sevre}, E. and {Lee}, S.-M.},
title={"{Effects of Easy Hybrid Parallelization with CUDA for Numerical-Atomic-Orbital Density Functional Theory Calculation}"},
journal={ArXiv e-prints},
archivePrefix={"arXiv"},
eprint={1402.4247},
primaryClass={"cs.DC"},
keywords={Computer Science – Distributed, Parallel, and Cluster Computing},
year={2014},
month={feb},
adsurl={http://adsabs.harvard.edu/abs/2014arXiv1402.4247P},
adsnote={Provided by the SAO/NASA Astrophysics Data System}
}
We modified a MPI-friendly density functional theory (DFT) source code within hybrid parallelization including CUDA. Our objective is to find out how simple conversions within the hybrid parallelization with mid-range GPUs affect DFT code not originally suitable to CUDA. We settled several rules of hybrid parallelization for numerical-atomic-orbital (NAO) DFT codes. The test was performed on a magnetite material system with OpenMX code by utilizing a hardware system containing 2 Xeon E5606 CPUs and 2 Quadro 4000 GPUs. 3-way hybrid routines obtained a speedup of 7.55 while 2-way hybrid speedup by 10.94. GPUs with CUDA complement the efficiency of OpenMP and compensate CPUs’ excessive competition within MPI.
February 21, 2014 by hgpu
