Implementace detekce Lagrangian coherent structures v knihovně MESIO

Abstract

This thesis focuses on the design, implementation, and experimental evaluation of a parallel method for computing LCS in large three-dimensional flow fields. The computation of LCS is a computa- tionally demanding task that requires efficient data handling, numerical stability, and well-designed parallelization. The proposed method combines a multiblock domain decomposition, asynchronous inter-process communication, and hybrid parallelization using MPI and OpenMP. The implemen- tation was tested on the Karolina supercomputer, where various scaling scenarios and the impact of individual optimizations on performance were analyzed. The thesis provides practical guidance on how to design an efficient parallel algorithm for FTLE computation in a real HPC environment and which strategies lead to a robust and scalable solution.