A better hash method for high-fidelity Monte Carlo simulations on nuclear reactors

Abstract

With the increasing demand for high-fidelity nuclear reactor simulations, the acceleration of Monte Carlo particle transport codes is becoming a core problem. One of the bottlenecks is locating millions or even billions of cells and fetching their associated parameters in the repeated geometry structure. Typically, Monte Carlo codes utilize a hash function to accelerate the cell locating and parameter indexing process. Specifically, they use the “cell vector → hash → parameter” method to accelerate the direct “cell vector → parameter” method. In this work, we propose a better hash method based on the Cyclic Redundancy Check (CRC) mechanism, which has been mathematically proven to be efficient and produce fewer hash collisions. Experimentally, this new hash method has been compared with some other hash functions and showed its superiority in terms of the calculation speed and collision probabilities. This hash method has been integrated into the Reactor Monte Carlo code RMC and worked well in practical applications.