Geometric Templates for Improved Tracking Performance in Monte Carlo Codes

One of the most fundamental parts of a Monte Carlo code is its geometry kernel. This kernel not only affects particle tracking (i.e., run-time performance), but also shapes how users will input models and collect results for later analyses. A new framework based on geometric templates is proposed that optimizes performance (in terms of tracking speed and memory usage) and simplifies user input for large scale models. While some aspects of this approach currently exist in different Monte Carlo codes, the optimization aspect has not been investigated or applied. If Monte Carlo codes are to be realistically used for full core analysis and design, this type of optimization will be necessary. This paper describes the new approach and the implementation of two template types in MC21: a repeated ellipse template and a box template. Several different models are tested to highlight the performance gains that can be achieved using these templates. Though the exact gains are naturally problem dependent, results show that runtime and memory usage can be significantly reduced when using templates, even as problems reach realistic model sizes.