Improving Uintah's Scalability Through the Use of Portable Kokkos-Based Data Parallel Tasks

摘要: The University of Utah's Carbon Capture Multidisciplinary Simulation Center (CCMSC) is using the Uintah Computational Framework to predict performance a 1000 MWe ultra-supercritical clean coal boiler. center aims utilize Intel Xeon Phi-based DOE systems, Theta and Aurora, through Aurora Early Science Program by Kokkos C++ library enable node-level portability. This paper describes infrastructure advancements portability improvements made possible integration within Uintah. marks step towards consolidating Uintah's MPI+PThreads MPI+CUDA hybrid parallelism approaches into single MPI+Kokkos approach. Scalability results are presented that compare serial data parallel task execution models for challenging radiative heat transfer calculation, central center's predictive boiler simulations. These demonstrate both good strong-scaling characteristics 256 Knights Landing (KNL) processors on NSF Stampede system, show KNL-based calculation compete with prior GPU-based same calculation.