CFD software, especially codes performing DNS and LES simulations, are exceptionally com- putationally expensive. Even small gains in optimization can bring a large benefit allowing larger, more detailed simulations and saving resources. This is compounded by the rapid development of new computing hardware which dictate continual updates for CFD codes to stay relevant. Integrating well known libraries such as the toolkit for scientific computing, PETSc is a way to reduce the effort required to stay on the cutting edge of HPC performance. Optimization potential is explored in the CFD code MGLET by implementing PETSc to solve the pressure correction of Chorin’s decomposition method. The pressure correction was identified as a hotspot in the MGLET code with approximately 40-80 % of time spent, mak- ing it an ideal place to test the functioning of PETSc with MGLET. Due to the limit of the scope of the project and to serve as a starting point for further research, PETSc is only im- plemented for the pressure correction. PETSc Solvers were used to investigate the optimality of the currently implemented solver with regards to robustness, speed, and accuracy over a variety of test cases for laminar and turbulent flow with and without immersed boundaries. The results show that PETSc is a viable alternative to in house solvers, and that an equal or better performance was achieved with a comparably small programming effort. Many factors especially immersed boundaries and case influenced the performance of the solvers; these factors benefit PETSc, as solvers could be exchanged depending on problem type. Overall, the solvers with the best performance were the conjugate gradient method with the algebraic multigrid preconditioner Boomeramg for cases without immersed boundary objects and the GMRES method using the sparse approximate inverse preconditioner Parasails for cases with immersed boundary objects.     «

CFD software, especially codes performing DNS and LES simulations, are exceptionally com- putationally expensive. Even small gains in optimization can bring a large benefit allowing larger, more detailed simulations and saving resources. This is compounded by the rapid development of new computing hardware which dictate continual updates for CFD codes to stay relevant. Integrating well known libraries such as the toolkit for scientific computing, PETSc is a way to reduce the effort required to s...     »