Table of Contents:
00:09 - Optimization of MPI application for MIC architecture
00:20 - Two options to use MPI with coprocessors
00:28 - Option 1: MPI+OpenMP with Offload
00:44 - Option 2: Symmetric hybrid MPI+OpenMP
01:02 - Peer-to-peer communication between MICs
01:16 - Virtualized Ethernet or InfiniBand
01:36 - I_MPI_FABRICS
01:54 - I_MPI_DEBUG=n
02:00 - xeonphi.com/papers/p2p
02:17 - Amount of communication end-points
02:42 - xeonphi.com/papers/sc14