high performance computing on graphics processing units: hgpu.org

A collimated emission of hadrons usually called Jet is the experimental counterparts of the partons (quarks and gluons) which are not observed separately. The CMS detector at LHC is ideally designed to study jet tomography which is an important probe to investigate the hot and dense medium formed during the heavy ion collisions. Although CMS analysis package incorporates various techniques for jet finding, the algorithm of successive recombination of particles based on kt jet-clustering has been widely used for jet reconstruction. This algorithm is simple and infrared safe, however it takes a large computing time of the order of N^3, where N is the total multiplicity in a given event and the computation becomes extremely difficult or even impossible when multiplicity increases particularly in the environment of LHC heavy ion colliders. Recently Cacciari and Salam [1] have proposed a fastjet algorithm which uses geometrical approach to find the closest pair and reduces its running time from N^3 to NlogN. Here we are trying to implement the variant of fastjet algorithm where processing is done parallely over multiple threads on the Graphics Processing Unit.