SC is the International Conference for
 High Performnance Computing, Networking, Storage and Analysis

SCHEDULE: NOV 13-19, 2010

Scalable Tile Communication-Avoiding QR Factorization on Multicore Cluster Systems

SESSION: Math Library Parallelization


TIME: 10:30AM - 11:00AM


AUTHOR(S):Fengguang Song, Hatem Ltaief, Bilel Hadri, Jack Dongarra


As tile linear algebra algorithms continue achieving high performance on shared-memory multicore architectures, it is a challenging task to make them scalable on distributed-memory multicore cluster systems. The main contribution of this paper is the extension to distributed-memory environment of an existing tile QR factorization algorithm on shared-memory machines. The fine granularity of tile algorithms associated with communication-avoiding techniques presents a high degree of parallelism where multiple tasks can be concurrently executed, computation and communication largely overlapped, and computation steps fully pipelined. A decentralized dynamic scheduler has then been integrated to efficiently schedule tasks across distributed resources. Our experiments performed on two clusters (with dual-core, 8-core nodes, respectively) and a Cray XT5 system with 12-core nodes show that the tile CA-QR algorithm is able to outperform the de facto ScaLAPACK library by up to 4 times for tall and skinny matrices, and has good scalability on up to 3,072 cores.

Chair/Author Details:

Osni Marques (Chair) - Lawrence Berkeley National Laboratory

Fengguang Song - University of Tennessee, Knoxville

Hatem Ltaief - University of Tennessee, Knoxville

Bilel Hadri - University of Tennessee, Knoxville

Jack Dongarra - University of Tennessee, Knoxville

Add to iCal  Click here to download .ics calendar file

Add to Outlook  Click here to download .vcs calendar file

Add to Google Calendarss  Click here to add event to your Google Calendar

The full paper can be found in the ACM Digital Library and IEEE Computer Society

   Sponsors    IEEE    ACM