Parallelisierung des Finite-Element Codes SMART-Konvektion

Many science and engineering problems can be described by systems of partial differential equations. Due to the fact that analytical solutions of those systems exist only for some special cases, it is necessary to apply numerical solution strategies such as finite difference or finite element methods. Massively parallel computers allow to reduce the computing times of the comprehensive numerical calculations by subdividing the calculation into parts and assigning them to different processors. In this thesis, the finite element code "SMART-Konvektion" is taken to develop and implement a concept which enables a complex computer code system to be ported to a massively parallel computer system. This concept is based on a decomposition of the considered finite element grid, which allows a parallel organization of the data and the solution of the systems of equations by an efficient parallelly executable conjugate gradient method. Three examples as there are the Rayleigh-Benard-convection, a convective flow with phase change, and a forced convection flow are analyzed with the computer code in cosiderationto valuate its power increase on the parallel computer Intel/Paragon. A good speedup can be recognized for all examples