Introducing GPU Acceleration into the Python-Based Simulations of Chemistry Framework.

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• High-Performance Computing

Background:

• PySCF is a popular quantum chemistry package.
• Accelerating quantum chemistry calculations is crucial for complex molecular simulations.
• GPU computing offers significant speedups for computationally intensive tasks.

Purpose of the Study:

• Introduce GPU4PySCF, a module for GPU acceleration in PySCF.
• Provide a GPU implementation of two-electron repulsion integrals (ERIs).
• Demonstrate accelerated quantum chemistry workflows using GPU-accelerated ERIs.

Main Methods:

• Developed a GPU implementation of two-electron repulsion integrals (ERIs) using Rys quadrature.
• Integrated GPU-accelerated ERIs into integral-direct Hartree-Fock (HF) calculations.
• Utilized GPU4PySCF for nuclear gradient construction in quantum chemistry workflows.

Main Results:

• Achieved a 2-orders-of-magnitude speedup for Hartree-Fock calculations compared to multithreaded CPU versions.
• Demonstrated performance comparable to established GPU-accelerated quantum chemistry packages (GAMESS, QUICK).
• Successfully applied GPU acceleration to ERI computation and HF build.

Conclusions:

• GPU4PySCF provides significant acceleration for quantum chemistry methods within PySCF.
• The module enables efficient use of GPUs for computationally demanding tasks like ERI calculation.
• GPU4PySCF represents a valuable tool for advancing computational chemistry research.