Vectorization of GEANT for generic detector research and event simulation

The Monte Carlo method for particle transport simulation is an essential tool in both the design process and the analytical evaluation process for detectors in the current and coming generation of high energy physics (HEP) experiments. Currently, the GEANT [1] Monte Carlo shower simulation code is the primary research tool being used in modeling all aspects of particle transport in HEP detectors. As with most Monte Carlo programs, the effectiveness and range of applicability of GEANT is limited primarily by the excessive run time required to provide statistically reliable results. Since the number of tracks which must be simulated, and hence the computational effort required, increase dramatically as higher energy events are modeled, the need for increased computational speed will become more urgent as the HEP community continues to explore physics at ever increasing energies. Part of this need may be met by efficient utilization of the existing capabilities of current vector supercomputers, such as the Cray YMP/48 and the IBM 3090. Using such resources effectively with Monte Carlo programs involves alteration of the program from a scalar history based algorithm to a vectorized event based algorithm. Results of a preliminary evaluation of the suitability of GEANT for adaptation to a fully vectorizable event based algorithm are presented here. 22 refs., 5 figs