CUDA and MPI performances on the computation of Rossi-α distribution from pulsed neutron sources

Abstract

This study presents the methodology to calculate the Rossi-α distribution for a subcritical nuclear reactor core (assembly) driven by an external pulsed neutron source through MCNP6 computer simulations. The Rossi-α distribution is obtained from the signal (e.g. 3 He(n,p) reactions) of a neutron detector placed in one experimental channel of the subcritical assembly. The detector signal is obtained from MCNP6 computer simulations modeling a single pulse of the external neutron source. The MCNP6 output data for a single pulse of the external neutron source must be processed to take into account the effect of delayed neutrons born in all previous source pulses. The calculation of the Rossi-α distribution from a pulsed neutron source is time consuming because of the huge amount of processed data. Consequently, the algorithms introduced in this work use parallel computing platforms, e.g. CUDA (which costs up to a few thousand dollars) or MPI (which costs from a few ten thousand to a few hundred thousand dollars), to reduce the computing time.