Should we move the computation to use the new xmap function and thereby using the DiskArrayEngine in the backend? I would hope that this might be a bit faster for the tilewise computations. @meggart