The Mathematical Institute, University of Oxford, Eprints Archive

Minimizing synchronizations in sparse iterative solvers for distributed supercomputers

Zhu, S.-X. and Gu, T.-X. and Liu, X.-P. (2013) Minimizing synchronizations in sparse iterative solvers for distributed supercomputers. Computer & Mathematics with Applications . (Submitted)

[img]
Preview
PDF
224kB

Abstract

Eliminating synchronizations is one of the important techniques related to minimizing communications for modern high performance computing. This paper discusses principles of reducing communications due to global synchronizations in sparse iterative solvers on distributed supercomputers. We demonstrates how to minimizing global synchronizations by rescheduling a typical Krylov subspace method. The benefit of minimizing synchronizations is shown in theoretical analysis and is verified by numerical experiments using up to 900 processors. The experiments also show the communication complexity for some structured sparse matrix vector multiplications and global communications in the underlying supercomputers are in the order P1/2.5 and P4/5 respectively, where P is the number of processors and the experiments were carried on a Dawning 5000A.

Item Type:Article
Subjects:D - G > General
Research Groups:Oxford Centre for Collaborative Applied Mathematics
ID Code:1724
Deposited By: Peter Hudston
Deposited On:30 Jul 2013 13:13
Last Modified:29 May 2015 19:25

Repository Staff Only: item control page