Acceleration of norm-conserving Pseudopotential Plane-Wave-Based DFT Calculation on GPU using CUDA

In present study, acceleration of density functional theory calculation using norm-conserving pseudopotential and plane wave (NCPP-PW) basis set has been performed. It did not use or parallelize commonly program packages (such as ABINIT, VASP, PWSCF, etc.) but propose prototypical program to carry out self-consistent field calculations to solve Kohn-Sham equation and focus on Hamiltonian diagonalization part by using CUDA to utilize a graphical processing unit (GPU) accelerator. The results showed that acceleration up to 10 times speed-ups for certain type of GPU, namely NVidia GTX 460, for three systems: 8 silicon atoms in cubic unit cell (small), 10 water molecules in a box (medium), and 64 silicon atoms in 2 × 2 × 2 cubic supercell (large).