Accelerating Relativistic Exact-Two-Component Density Functional Theory Calculations with Graphical Processing Units

Area of Science:

• Computational chemistry
• Materials science
• Quantum mechanics

Background:

• Numerical integration of exchange-correlation potentials is computationally intensive.
• Graphical Processing Units (GPUs) offer massive parallel processing capabilities.
• Relativistic, 2-component density functional theory (DFT) requires significant computational resources.

Purpose of the Study:

• To implement and evaluate GPU acceleration for exchange-correlation potential calculations within the GauXC library.
• To assess the performance gains for relativistic DFT calculations.
• To enable more efficient simulations of systems with heavy elements.

Main Methods:

• Implementation of GPU-accelerated exchange-correlation potential integration in the GauXC library.
• Benchmarking calculations on copper, silver, and gold coinage metal clusters.
• Comparison of GPU-based performance against traditional CPU-based calculations.

Main Results:

• Significant speedup achieved with GPU acceleration compared to CPU calculations.
• One GPU card demonstrated computational power equivalent to approximately 400 CPU cores.
• Speedup increases with system size, indicating scalability for larger simulations.
• Support for arbitrary angular momentum basis functions enhances applicability to heavy elements.

Conclusions:

• The GPU-accelerated implementation provides substantial speedup for relativistic electronic structure calculations.
• This advancement enables more efficient and extensive computational studies in DFT.
• The approach holds significant potential for future, more demanding simulations in computational chemistry and materials science.