The Community for Technology Leaders
Green Image
<p><b>Abstract</b>—This article is devoted to the run-time redistribution of one-dimensional arrays that are distributed in a block-cyclic fashion over a processor grid. While previous studies have concentrated on efficiently generating the communication messages to be exchanged by the processors involved in the redistribution, we focus on the <it>scheduling</it> of those messages: how to organize the message exchanges into "structured" communication steps that minimize contention. We build upon results of Walker and Otto, who solved a particular instance of the problem, and we derive an optimal scheduling for the most general case, namely, moving from a <tt>CYCLIC(r)</tt> distribution on a <it>P</it>-processor grid to a <tt>CYCLIC(s)</tt> distribution on a <it>Q</it>-processor grid, for <it>arbitrary</it> values of the redistribution parameters <it>P</it>, <it>Q</it>, <it>r</it>, and <it>s</it>.</p>
Distributed arrays, redistribution, block-cyclic distribution, scheduling, MPI, HPF.
Yves Robert, Jack Dongarra, Frédéric Desprez, Cyril Randriamaro, Antoine Petitet, "Scheduling Block-Cyclic Array Redistribution", IEEE Transactions on Parallel & Distributed Systems, vol. 9, no. , pp. 192-205, February 1998, doi:10.1109/71.663945
98 ms
(Ver 3.3 (11022016))