GPU-Accelerated Computation of Femtoscopic Correlation Functions for Heavy-Ion Collisions

Since its discovery, Hanbury-Brown Twiss (HBT) interferometry has served as an invaluable tool to particle physicists by allowing the precise measurement of spatio-temporal quantities of various particle-emission sources that are otherwise difficult to probe directly. In our study of Quark-Gluon Plasma (QGP), we overcome challenges posed by the small time and length scales (10e-23 s and 10e-15 m, respectively) involved in relativistic heavy ion collisions by analyzing average quantum correlation effects affecting the resulting particle shower from the QGP's hadronization. By computing the expected correlation functions and comparing them to experimental statistics, we are able to test the current theory of QGP's dynamics. However, as HBT calculations often suffer from poor serial performance, we directed our research on using conventional graphics processors to parallelize HoTCoffeeh, a C++ application created for calculating correlation functions (CFs) in 2+1D. Results indicate that our CF calculations decompose into several embarassingly parallelizable subproblems, leading to major computation time savings that allow larger quantities of events to be generated and analyzed than previously feasible in conventional supercomputing and grid computing environments.