Data distribution schemes of sparse arrays on distributed memory multicomputers

Array redistribution is usually needed for more efficiently executing a dataparallel program on distributed memory multicomputers. To minimize the redistribution data transfer cost, processor mapping techniques were proposed to reduce the amount of redistributed data elements. Theses techniques demand that the beginning data elements on a processor not be redistributed in the redistribution. On the other hand, for satisfying practical computation needs, a programmer may require other data elements to be un-redistributed (localized) in the redistribution. In this paper, we propose a flexible processor mapping technique for the Block-Cyclic redistribution to allow the programmer to localize the required data elements in the redistribution. We also present an efficient redistribution method for the redistribution employing our proposed technique. The data transfer cost reduction and system performance improvement for the redistributions with data localization are analyzed and presented in our experimental results.

show abstract

Compressing three-dimensional sparse arrays using inter- and intra-task parallelization strategies on Intel Xeon and Xeon Phi

Lin

Yen

Hung

2016

J Supercomput

View full text Add to dashboard Cite

Data distribution schemes of sparse arrays on distributed memory multicomputers

Abstract: A data distribution scheme of sparse arrays on a distributed memory multicomputer, in general, is

Cited by 4 publications

References 25 publications

Message Clustering Technique towards Efficient Irregular Data Redistribution in Clusters and Grids

Message Clustering Technique towards Efficient Irregular Data Redistribution in Clusters and Grids

A flexible processor mapping technique toward data localization for block-cyclic data redistribution

Compressing three-dimensional sparse arrays using inter- and intra-task parallelization strategies on Intel Xeon and Xeon Phi

Contact Info

Product

Resources

About