A standalone package for on-line event selection in the CBM experiment

finder [3], a Kalman filter based track fitter [4] (both are SIMDized), a primary vertex finder and a procedure for finding short lived decayed particles (both are implemented within the KF Particle package [5]) and a procedure for selecting charm track candidates. These routines exist already in the cbmroot framework, but in order to develop and test the full selection chain on different modern and future CPU/GPU architectures they have been substantially modified and reorganized into a standalone package. Input data consist of detector geometry, simulated data in form of registered hits and Monte Carlo tracks for performance evaluation. Each part of data is stored in a separate text file produced with a special routine. The total reconstruction efficiency (see Fig. 2) for all tracks is about 92%. The reconstruction efficiency clearly depends on the particle momentum. High energetic particles have efficiency of about 97%. Secondary tracks from D decay have momentum larger than 1 GeV/c and, in addition, come from the target region, which can be used during the reconstruction, and, therefore, they have even higher efficiency of 99%. Let us note that only simple Gaussian smearing was used in these investigation, and no double hit effects in the sensor considered. Most of other secondary tracks are low energetic tracks which suffer significant multiple scattering in the detector material. The efficiency of low energetic tracks is about 82%. Particles with momentum lower than 100 MeV/c are mostly out of geometrical acceptance of the STS detector. Splitting of reconstructed tracks into short parts is negligible. The level of wrongly reconstructed tracks (ghost tracks) is about 3% and these tracks are similar to tracks of short low energetic particles.