Real-space density functional theory on graphical processing units: computational approach and comparison to Gaussian basis set methods

TL;DR

This paper demonstrates how real-space density functional theory calculations can be significantly accelerated using GPUs, achieving high performance and competitive results compared to Gaussian basis set methods.

Contribution

The authors developed an efficient GPU implementation of real-space DFT in the Octopus code, exposing data parallelism to enhance computational speed.

Findings

Achieved up to 90 GFlops performance on current GPUs.

Significant speed-up over CPU implementations.

Outperformed GPU Gaussian basis set codes for some systems.

Abstract

We discuss the application of graphical processing units (GPUs) to accelerate real-space density functional theory (DFT) calculations. To make our implementation efficient, we have developed a scheme to expose the data parallelism available in the DFT approach; this is applied to the different procedures required for a real-space DFT calculation. We present results for current-generation GPUs from AMD and Nvidia, which show that our scheme, implemented in the free code Octopus, can reach a sustained performance of up to 90 GFlops for a single GPU, representing a significant speed-up when compared to the CPU version of the code. Moreover, for some systems our implementation can outperform a GPU Gaussian basis set code, showing that the real-space approach is a competitive alternative for DFT simulations on GPUs.