high performance computing on graphics processing units: hgpu.org

The use of the Graphical Processing Unit (GPU) as a general purpose processor is becoming popular. This thesis describes how GPU Computing can be used and can be beneficial in High Energy Physics (HEP) online computation or real time data analysis. This thesis explains that HEP computing is embarrassingly parallel problem therefore by using GPU computing, online event selection and trigger implementation can be achieved with high event rate. The thesis discusses about GPU, Compute Unified Device Architecture (CUDA) with respect to implementation and optimization of event selection trigger algorithm for Muon Chamber (MUCH) detector system at Compressed Baryonic Matter (CBM) experiment. The CBM experiment, at the upcoming Facility for Antiproton and Ion Research (FAIR) center at GSI Darmstadt, Germany aims at the exploration of baryonic matter at high density produced in relativistic heavy ion collisions. The proposed key observables include the measurement of J/Psi (pronounced as japsi and popularly known as charmonia), which can be measured via their decay in to the di-muon channel. However the production of charmonium is extremely small at FAIR energy regime (E=10-40AGeV) which requires an extreme interaction rate (upto 10^7 events per second) to have good events statistics. For detection of charmonium, MUCH detector system is being designed by a collaboration of Russian and Indian institutes. In this report we mainly focus on the development and implementation of an algorithm for an on-line event selection to pick up the interesting physics events using Graphical Processing Unit to process high event rate. CUDA is an efficient and easy way of programming GPU for general purpose problems. We show that desired goal of processing 10^7 events per second and then storing useful events can be achieved with the algorithm developed and implemented on GPU Tesla C2075 card as a part of this work.