In computational engineering, one often encounters a so called fix point update formula, which arise naturally in various numerical methods. In some contexts, the corresponding update function contains parallelizable computations. Parallelization enables solving large problems more efficiently but most algorithms entail even more potential to shorten the computation time.
During the so called synchronization parallel threads wait for others to finish and receive updated information before starting the next iteration. This introduces idle time which can be circumvented by applying asynchronous approaches. In principle, synchronization steps are neglected, i.e., threads keep iterating even if other threads are not finished yet.
In order to speed-up 3D simulation of electrical machines, this project aims to study ParaReal and/or iterative solvers for linear systems arising in domain decomposition methods with a focus on asynchronous iterations.