SC²S Colloquium - January 13, 2016
|Date:||January 13, 2016|
|Time:||3:00 pm, s.t.|
Ivan Bratoev : Lattice Boltzmann Simulations on Spacetree-based Adaptive Grids
Over the past two decades there has been extensive research in the field of Computational Fluid Dynamics (CFD) and one method, that has been the focus of comprehensive exploration, is the Lattice-Boltzmann Method (LBM). It is a method that has been developed from the lattice gas automata and its chief ad- vantages are that it is ease of implementation and parallelization. A major focus in the further development of the method is the application of a non-uniform grid with the LBM. This thesis is focused on a newly proposed approach for the application of a quad-(oct-)tree structure on a simulation domain. First we explore different applications for the LBM and a few approaches in the implementation of a non-uniform grid. We then introduce the basic LBM method and further extended it by offering an alternative for the collision step. We then take a look at two methods for the application of non-uniform grids, one that uses time scaling and the other uses different Courant-Friedrichs-Lewy (CFL) numbers for the different parts of the simulation, which is then used in the implementation of the thesis. A core part of the thesis is the use of the Spacetree Library. We use its great flexibility in creating a data structure that allows us to easily implement the modified LBM. Finally we use the Cavity scenario to validate our results against the expected results presented in various papers. To confirm that all parts of the implementation are correct we use a few specific validation approach. One is used to prove the correctness of the implemented uniform grid approach. A second one is used to validate that the communication between different parts of the simulation domain are correct. The final part of this process is to make sure that a non-uniform grid achieves a better level of precision in comparison to the defined results for the Cavity scenario.
Manuel Fasching: JIT compilation to realize flexible data access in simulation software
Input data or scenarios of simulation applications are frequently defined as mathematical functions. Due to simplicity or for performance reasons those functions are often implemented directly in the application code. This causes an adaption of the code for every new scenario. The main goal of the master‘s thesis is to develop a lightweight Just-In-Time compiler which enables flexible scenario defining in a configuration file. The provided performance of the Just-In-Time compiler should be similar to the current solution.