A Kalman Filter Method for the Charged Particles Trajectories Reconstruction in the CBM Experiment and Its Parallel Implementation at the JINR LIT Manycore Server

The high-intensity heavy-ion beams that will be produced by the accelerators at the Facility for Antiproton and Ion Research (FAIR), together with the Compressed Baryonic Matter (CBM) experiment, now in preparation, offer outstanding possibilities for studying baryonic matter at superhigh densities and moderate temperatures under laboratory conditions. The CBM physics program is aimed at studying the structure and behavior of baryonic matter at densities comparable to those in the center of neutron stars. The program includes 1) setting a phase boundary between the hadronic and partonic matter, 2) determining the critical end point, and 3) searching for indications of the origin of chiral symmetry reconstruction at high pure baryonic densities. The task of a charge particle trajectories reconstruction is one of the most important tasks of the CBM experiment. It assumes a full on-line event reconstruction, that requires development of fast algorithms, which utilize the potential of modern CPU and GPU architectures in the most efficient way. In the current work the results of analysis of the Kalman filter based track reconstruction algorithm, which is implemented using different parallelization approaches, are presented and discussed. For the analysis a manycore server with two Intel Xeon X5660 CPUs and a NVidia GTX 480 GPU at LIT, JINR was used.

CBM, OpenMP, OpenCL, NVIDIA, Intel, CPU, GPU, heterogeneous system, Kalman filter, track fitting