GREAPMC (GPU-optimized REActor Physics Monte Carlo) is a specialized Monte Carlo code designed for the efficient criticality simulations of pressurized water reactors (PWRs) using GPU acceleration. The primary objective behind developing GREAPMC is to conduct comprehensive whole-core calculations within feasible timeframes. Leveraging state-of-the-art GPU acceleration techniques, GREAPMC enables the execution of large-scale reactor physics simulations for full-scale PWRs within practical time constraints and resource limitations. As development progresses, new features are continually integrated to enhance the capabilities of the code, bringing it closer to the standards set by industry-level Monte Carlo codes. At CORE lab, we are actively engaged in developing, testing, verifying, and integrating new algorithms into the code as part of our effort to create a Monte Carlo solution that delivers high-fidelity results with exceptional speed. Our programming decisions and algorithmic improvements are guided by the specifications of GPU architecture and the features of CUDA C++ programming.

Team

• Programming paradigm: GREAPMC is developed in CUDA C++ which serves as an extension of C++ programming language, encompassing additional constructs specifically tailored for GPU programming. The code adheres to object-oriented programming (OOP) and uses advanced C++ features where necessary.
• Cell-based geometry modeling: The conventional CSG modeling approach is sacrificed to gain acceleration by reducing the thread divergence in GPU. The advantage of this modeling approach is that the particle location is always known inside the transport loop, thus eliminating cell search overhead, and ensuring consistent execution times even as the number of cells increases.
• Dynamic history length truncation: Among the array of optimizations integrated into history-based neutron tracking, dynamic history length truncation stands out as the most significant accelerator. With this approach, the length of history is restricted to a predefined maximum number of interactions during cycles. Following this, neutrons are sorted, and the kernel is called upon with only the surviving neutrons. The key aspect of this method lies in the automatic adjustment of the maximum number of interactions during inactive cycles.
• Cross-section treatment: To elevate our efforts in optimizing geometrical modeling and history-based neutron tracking, GREAPMC currently harnesses multigroup macroscopic cross-sections.

• Tally-bin indexing scheme: GREAPMC uses an efficient tally-bin indexing scheme that ensures that the tallies have a minimal impact on the execution time, irrespective of an increase in the number of cells.