Message from the Workshop Chairs

Abstract

A Benchmark Fast of Multithreaded Communication Performance” new benchmark for multi-threaded communication, of Multi-GPU MPI Collective Communications on Large FFT Computation” the key area of GPU enabled MPI collectives. Improvements to Allreduce” by Bienz et al. explores the topic of collective communications and whether they can be accelerated by taking into account process placement on multiple nodes and taking advantage of local node shared memory communication in an intelligent way to accelerate all reduce. Gawande et al. MPI to provide a high-performance implementation of a PGAS runtime built on top of MPI two-sided communication in “Accelerating the Global Arrays ComEx Runtime Using Multiple Progress Ranks”. Margolin et al’s Library for Collective Operations in MPI” used underlying RDMA network operations to accelerate large scale collective operations in MPI. Finally, Schuchart et al. explore the possibility of using MPI RMA as a basis for an active message solution in MPI-3 RMA for Active Messages”.