High-Level Topology-Oblivious Optimization of MPI Broadcast Algorithms on Extreme-Scale Platforms
MetadataShow full item record
AbstractThere has been a significant research in collective communication operations, in particular in MPI broadcast, on distributed memory platforms. Most of the research works are done to optimize the collective operations for particular architectures by taking into account either their topology or platform parameters. In this work we propose a very simple and at the same time general approach to optimize legacy MPI broadcast algorithms, which are widely used in MPICH and OpenMPI. Theoretical analysis and experimental results on IBM BlueGene/P and a cluster of Grid’5000 platform are presented.
CitationHasanov K, Quintin J-N, Lastovetsky A (2014) High-Level Topology-Oblivious Optimization of MPI Broadcast Algorithms on Extreme-Scale Platforms. Euro-Par 2014: Parallel Processing Workshops: 412–424. Available: http://dx.doi.org/10.1007/978-3-319-14313-2_35.
SponsorsThis work has emanated from research conducted withthe financial support of IRCSET (Irish Research Council for Science, Engineeringand Technology) and IBM, grant number EPSG/2011/188 and Science Founda-tion Ireland, grant number 08/IN.1/I2054.Some of the experiments presented in this publication were carried out us-ing the Grid’5000 experimental testbed, being developed under the INRIA AL-ADDIN development action with support from CNRS, RENATER and severalUniversities as well as other funding bodies (seehttps://www.grid5000.fr)Another part of the experiments were carried out using the resources of the Su-percomputing Laboratory at King Abdullah University of Science&Technology(KAUST) in Thuwal, Saudi Arabia.
PublisherSpringer Science + Business Media