Reconstructing the trajectories of charged particles from signals left in tracking detectors is a formidable pattern reconstruction problem, especially at high particle rates. We are developing algorithms for tracking at the Mu3e and P2 experiments which should be both precise and at the same time fast enough to cope with several billion tracks a second. In collaboration with the Mu3e group in Heidelberg we have developed a track fit for the case of dominating scattering uncertainties which is also suitable for running on GPUs.
