CPU-GPU-coupled acceleration method for point flux calculation in Monte Carlo particle transport

Area of Science:

• Computational physics
• Particle transport simulation
• High-performance computing

Background:

• Point flux tallying is crucial for Monte Carlo simulations, especially for small detectors in source-detector and dose calculations.
• Large-scale point flux tallying on traditional CPUs is computationally intensive, leading to significant time consumption.

Purpose of the Study:

• To develop and validate a CPU-GPU coupled acceleration method for enhancing the efficiency of large-scale point flux tallying.
• To reduce the computational burden associated with particle transport simulations.

Main Methods:

• A hybrid approach was implemented, leveraging both Central Processing Units (CPUs) and Graphics Processing Units (GPUs).
• Complex logic was handled by the CPU, while computationally intensive tasks were offloaded to the GPU.
• The method was validated using the NUREG/CR-6115 Pressurized Water Reactor (PWR) benchmark problem.

Main Results:

• The CPU-GPU accelerated method achieved identical photon point flux estimation results compared to a pure CPU approach.
• The accelerated program demonstrated a speedup of approximately 50 times over the pure CPU program.
• Significant enhancement in the efficiency of large-scale point flux tallies was observed.

Conclusions:

• The proposed CPU-GPU coupled method offers a substantial performance improvement for large-scale point flux tallying in particle transport simulations.
• This acceleration provides significant convenience for subsequent dose calculations and related analyses.
• The method is a viable solution for reducing computational time in demanding simulation scenarios.