Improved Mcnp Memory Locality by Neutron Grouping Thesis uri icon



  • Thesis (M.S., Mechanical Engineering)--University of Idaho, June 2014 | This research presents new code for Monte Carlo N-Particle (MCNP) to achieve an improved time during criticality calculations. Modifications implementing the grouping and sorting of neutrons takes advantage of memory locality by processing all neutrons in a group to achieve the temporal reuse of cross section data. This prevents unnecessary data lookups. Various groupings and their results are compared. The modified code utilizing neutron energy groups provided the best result of a 16.7% ± 0.5% speedup for a criticality determination of a two slab tank experiment. This is a savings of 2 ½ hours for a system that normally takes approximately 15 ½ hours to execute. The code implemented was chosen to require minimal modifications to the MCNP program thus avoiding the need to rewrite a new version. Verification and validation is still needed in order to show that a speedup using neutron groups can be achieved in all cases.

publication date

  • June 1, 2014