Samsung Advanced Institute of Technology

 We report a high performance multi graphics processing unit (GPU) implementation of the Kohn？Sham time-dependent density functional theory (TDDFT) method. Our new GPU algorithm scales well with materials dimensionality, significantly reducing the computational wall time and further expanding the materials horizons that can be considered by a full quantum chemical treatment. Benchmark TDDFT study on the green fluorescent protein complex composed of 4,353 atoms with 40,518 Gaussian functions from def2-SVP basis sets, as the largest molecule attempted to date to the best of our knowledge, demonstrated 37％ parallel efficiency with 256 GPUs with 884,736 FP64 cores on a custom-built state-of-the-art GPU computing system with Nvidia A100 GPUs. We believe our GPU-oriented algorithms that empower first-principle atomistic simulations regardless of the sizes and the complex combinations of the involved molecules may provide deeper understanding of molecular basis of materials behaviors and eventually open up new possibilities for breakthrough designs on new materials systems.