ITERATIVE AND PARALLEL PERFORMANCE ANALYSIS OF NON-BLOCKING COMMUNICATION ALGORITHMS IN THE MASSIVELY PARALLEL NEUTRON TRANSPORT CODE PIDOTS

The PIDOTS neutral particle transport code utilizes a red/black implementation of the Parallel Gauss-Seidel algorithm to solve SN approximation neutron equation on 3D Cartesian meshes. is designed for execution massively parallel platforms and capable using full resources modern, leadership class high performance computers. Initial testing revealed that some configurations PIDOTS’s Integral Transport Matrix Method solver demonstrated unexpectedly poor scaling. Work at Idaho Los Alamos National Laboratories then this inefficiency was result accumulation high-cost latency events in complex blocking communication networks employed during each iteration. That work explored possibility minimizing those inefficiencies while maintaining communications model. While significant speedups were obtained, it shown fully mitigating problem general-purpose highly unlikely code. This continues analysis by implementing deeply interleaved non-blocking model into PIDOTS. new benefits from optimization performed also providing opportunities overlap remaining un-mitigated costs with computation. Additionally, our approach easily transferable other similarly spatially decomposed codes. resulting tested LANL’s Trinity system up 32,768 processors found processor count effectively hide 100% MPI cost – equivalently 20% phase time. It expected implemented interleaving can support far higher counts completely ~50% total iteration