FFTs and multiple collective communication on multiprocessor-node architectures

Authors:
Andreas Jocksch
Affiliations:
CSCS, Swiss National Supercomputing Centre, Manno, Switzerland
Venue:
PPAM'11 Proceedings of the 9th international conference on Parallel Processing and Applied Mathematics - Volume Part I
Year:
2011

Citing 14
Cited 0

Two and three dimensional FFTS on highly parallel computers

Parallel Computing
Optimum Broadcasting and Personalized Communication in Hypercubes

IEEE Transactions on Computers
Computational frameworks for the fast Fourier transform

Computational frameworks for the fast Fourier transform
Methods and problems of communication in usual networks

Proceedings of the international workshop on Broadcasting and gossiping 1990
Parallel algorithms for personalized communication and sorting with an experimental study (extended abstract)

Proceedings of the eighth annual ACM symposium on Parallel algorithms and architectures
Efficient Algorithms for All-to-All Communications in Multiport Message-Passing Systems

IEEE Transactions on Parallel and Distributed Systems
How helpers hasten h-relations

Journal of Algorithms
The Scalability of FFT on Parallel Computers

IEEE Transactions on Parallel and Distributed Systems
Scheduling Parallel Communication: The h-relation Problem

MFCS '95 Proceedings of the 20th International Symposium on Mathematical Foundations of Computer Science
The Hierarchical Factor Algorithm for All-to-All Communication (Research Note)

Euro-Par '02 Proceedings of the 8th International Euro-Par Conference on Parallel Processing
Fast Collective Operations Using Shared and Remote Memory Access Protocols on Clusters

IPDPS '03 Proceedings of the 17th International Symposium on Parallel and Distributed Processing
Exchanging messages of different sizes

Journal of Parallel and Distributed Computing
Communication analysis of parallel 3D FFT for flat cartesian meshes on large Blue Gene systems

HiPC'08 Proceedings of the 15th international conference on High performance computing
An implementation of parallel 3-D FFT with 2-D decomposition on a massively parallel cluster of multi-core processors

PPAM'09 Proceedings of the 8th international conference on Parallel processing and applied mathematics: Part I

Quantified Score

Hi-index	0.00

Visualization

Abstract

We consider FFTs for networks with multiprocessor nodes using 2D data decomposition. In this application, processors perform collective all-to-all communication in different groups independently at the same time. Thus the individual processors of the nodes might be involved in independent collective communication. The underlying communication algorithm should account for that fact. For short messages, we propose a sparse version of Bruck's algorithm which handles such multiple collectives. The distribution of the FFT data to the nodes is discussed for the local and global application of Bruck's original algorithm, as well as the suggested sparse version. The performance of the different approaches is compared.