GPU-accelerated computation of electron transfer.

Area of Science:

• Computational Chemistry
• Quantum Mechanics
• Scientific Computing

Background:

• Electron transfer is a fundamental quantum mechanical process.
• Simulating electron transfer is computationally intensive.
• Investigating computational acceleration methods is crucial for advancing research.

Purpose of the Study:

• To evaluate the suitability of graphics processing units (GPUs) for accelerating electron transfer simulations.
• To identify time-critical components in existing simulation implementations.
• To test various GPU implementation strategies at increasing abstraction levels.

Main Methods:

• Profiling an existing electron transfer simulation code.
• Implementing and testing several variants utilizing GPUs.
• Employing a publicly available library for GPU-accelerated basic linear algebra operations.

Main Results:

• A 50-fold acceleration in computation was achieved using a GPU-accelerated linear algebra library.
• Performance gains showed minor dependence on the specific problem size.
• Numerical accuracy was maintained despite the significant speedup.

Conclusions:

• GPUs are highly suitable for accelerating computationally intensive electron transfer simulations.
• GPU-accelerated linear algebra libraries provide substantial performance improvements.
• This approach offers significant practical value for researchers studying electron transfer processes.