Evaluating the networking characteristics of the Cray XC‐40 Intel Knights Landing‐based Cori supercomputer at NERSC

• DOI: 10.1002/cpe.4297
• Publication Date: 2017-09-12T08:42:37Z
• Authors: Douglas Doerfler; Brian Austin; Brandon Cook; Jack Deslippe; Krishna Kandalla; Peter Mendygral
• Keywords: Xeon Phi; Xeon; Vectorization (mathematics)

Summary There are many potential issues associated with deploying the Intel Xeon Phi TM (code named Knights Landing [KNL]) manycore processor in a large‐scale supercomputer. One particular is ability to fully utilize high‐speed communications network, given that serial performance of core fraction Xeon®core. In this paper, we take look at trade‐offs allocating enough cores Aries network versus dedicated computation, eg, trade‐off between MPI and OpenMP. addition, evaluate new features Cray support KNL, such as internode optimizations. We also one‐sided programming models Unified Parallel C. quantify impact above using suite National Energy Research Scientific Computing Center applications.