More on scheduling block-cyclic array redistribution

Frédéric Desprez (1) , Stéphane Domas (1) , Jack Dongarra (2) , Antoine Petitet (2) , Cyril Randriamaro (1) , Yves Robert (1)

1 REMAP - Regularity and massive parallel computing
2 Department of Computer Science. University of Tennessee

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. In a previous paper, we have reported how to derive optimal schedules made up of successive communication-steps. In this paper we assume that successive steps may overlap. We show how to obtain an optimal scheduling for the most general case, namely, moving from a CYCLIC(r) distribution on a P-processor grid to a CYCLIC(s) distribution on a Q-processor grid, for arbitrary values of the redistribution parameters P, Q, r, and s. We use graph-theoretic algorithms, and modular algebra techniques to derive these optimal schedulings.

Domaines: Calcul parallèle, distribué et partagé [cs.DC]

Frédéric Desprez, Stéphane Domas, Jack Dongarra, Antoine Petitet, Cyril Randriamaro, et al.. More on scheduling block-cyclic array redistribution. Proc. of 4th Workshop on Languages, Compilers, and Run-time Systems for Scalable Computers (LCR98), 1998, Unknown, pp.275-287, ⟨10.1007/3-540-49530-4_20⟩. ⟨hal-00856853⟩