Event selection for MUCH of CBM experiment using GPU computing

High Throughput Computing plays an important role in the field of computational science. The Compressed Baryonic Matter (CBM) experiment is being planned at Facility for Antiproton and Ion Research (FAIR) accelerator complex, which is under construction at GSI laboratory in Darmstadt, Germany. This experiment is going to produce substantial data which requires high throughput computing to perform reconstruction, analysis, etc. Simulation of such a very large experiment requires computational capabilities which cannot be satisfied by a single processing system. A possible way to solve this problem is to use of heterogeneous computing which refers to the systems that use more than one kind of processors viz CPU; Graphics Processing Unit (known as GPU); Accelerated Processing Unit; Xeon Phi; Field Programmable Gate Array etc. GPU can be used as a general purpose processor as a part of heterogeneous computing. In this experiment a rare particle, J/Psi will be produced. For detection of J/Psi via di-muon decay channel, Muon Chamber Detector System is being designed, where India is playing a major role. Detection of rare particles requires very high interaction rate (upto 107 events per second), an efficient event selection algorithm is required for selecting events which contain probable candidates of J/Psi. A novel algorithm has been developed for online selection of useful events which results in faster computation time with respect to the existing algorithm capable of processing 4×105 events per second using GPU Tesla C2075 card. This paper explains in detail the difference with earlier algorithm and shows that more than 107 events can be processed per second using same GPU card.