2011 · 19 citations · 9 references
ComputingComputer EngineeringComputer ScienceComputer ArchitectureMessage PassingParallel ComputingBroadcast CommunicationCollective CommunicationDistributed ComputingInterprocess CommunicationLatency OptimizationNonblocking CommunicationBroadcast OperationsCollective Communication OffloadCommunication-computation OverlapInfinibandMessage Passing InterfaceNonblocking Collective CommunicationShared MemorySynchronization In Collective Operations
This paper describes the design and implementation of InfiniBand (IB) CORE-Direct based blocking and nonblocking broadcast operations within the Cheetah collective operation framework. It describes a novel approach that fully offloads collective operations and employs only user-supplied buffers. For a 64 rank communicator, the latency of CORE-Direct based hierarchical algorithm is better than production grade Message Passing Interface (MPI) implementations, 150% better than the default Open MPI algorithm and 115% better than the shared memory optimized MVAPICH implementation for a one kilo-byte (KB) message, and for eight mega-bytes (MB) it is 48% and 64% better, respectively. Flat-topology broadcast achieves 99.9% overlap in a polling based communication-computation test, and 95.1% overlap for a wait based test, compared with 92.4% and 17.0%, respectively, for a similar Central Processing Unit (CPU) based implementation.
9
COMB: a portable benchmark suite for assessing MPI overlap
W. Lawry, Christopher Riley Wilson, Arthur B. Maccabe et al. · 2003 · 57 citations
Efficient reliable multicast on Myrinet
Kees Verstoep, Koen Langendoen, Henri E. Bal · 2002 · 56 citations
MPI collective communications on the blue gene/p supercomputer
Ahmad Faraj, Sameer Kumar, Brian E. Smith et al. · 2009 · 54 citations