Fast parallelized kalman filter based reconstruction of charged particle trajectories for the compressed baryonic matter experiment on a many-core server at the laboratory of information technologies, JINR

The charged particle trajectory online reconstruction in the Compressed Baryonic Matter (CBM) experiment (GSI, Germany) is an extremely difficult task. It is conditioned by a high rate of the ion beam-tar-get collisions (up to 107/s), high track multiplicity in each nucleus-nucleus collision (up to 1000 particles), and charged-particle trajectory registration with the coordinate detectors located in a highly inhomogeneous magnetic field. Such a problem could be solved only by using modern high-performance computers. This work scrutinizes a Kalman filter based track reconstruction algorithm implemented using different parallelization approaches. To perform the analysis, a many-core hybrid server with two Intel Xeon X5660 CPUs and a NVidia GTX 480 GPU (JINR LIT) was used.