For the LHCb Upgrade, millions of collisions will be reconstructed in software
per second, a 40x increment in data throughput with respect to today. Graphics cards are amenable architectures to processing these kind of data, having produced good results on computationally intense problems during the last ten years at an affordable cost. In this talk, we will present the parallelization of long standing problems in LHCb reconstruction for massively parallel architectures, as well as the use of non conventional techniques.
LHCb