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

The article focuses on a novel CPU–GPU-coupled acceleration method designed to improve the computational efficiency of point flux tallying in Monte Carlo (MC) particle transport simulations, particularly for large-scale problems. Point flux tallying, a variance reduction technique important for small detectors and dose calculations, traditionally incurs significant computational overhead when many detectors are involved. The proposed method offloads the computationally intensive and parallelizable point flux calculations to GPUs while retaining complex logic on CPUs, employing a particle clustering strategy to optimize data transfer and integrating MPI (message passing interface) with CUDA for hybrid parallelism. Validation using the NUREG/CR-6115 pressurized water reactor (PWR) benchmark problem demonstrated that the CPU–GPU-coupled program maintains calculation accuracy while achieving approximately a 50-fold increase in computational efficiency compared to a single-core CPU implementation. This approach enhances the feasibility of MC simulations involving extensive point flux tallies, although current implementation supports only photon point flux calculations, with further development needed for other particle types.