Fischer, Alexander2020-12-142020-12-142020174348268Xhttp://nbn-resolving.de/urn:nbn:de:bsz:93-opus-ds-112059http://elib.uni-stuttgart.de/handle/11682/11205http://dx.doi.org/10.18419/opus-11188Sorting networks are usually bound at a depth of O(log^2 n), since a perfect halver is of at least depth O(log n). However, the AKS Sorting Network, by Ajtai, Komlos and Szemeredi, can sort data with depth O(log n) by using so-called ε-halvers, which are of constant depth. Such ε-halvers are allowed to have some errors and will eventually be corrected by sending elements to a level above. In this thesis, a CPU and CUDA version are implemented following a paper by Vasek Chvatal and the original paper by Ajtai et al. Experiments are run on these versions to observe and improve parameters.eninfo:eu-repo/semantics/openAccess004bachelorThesis