The present paper describes an efficient communication optimization technique for Householder tridiagonalization called CAHTR and evaluates its parallel performance. CAHTR is intended to reduce the number of problems in collective communication, especially MPI_Allreduce operations. We demonstrate the optimal version of CAHTR(3) compared with a naive implementation CAHTR(0). The CAHTR algorithms are evaluated on the K supercomputer system, and speedup exceeds x1.4 for the case of N=5000 and P=1024.
IOS Press, Inc.
6751 Tepper Drive
Clifton, VA 20124
Tel.: +1 703 830 6300
Fax: +1 703 830 2300 firstname.lastname@example.org
(Corporate matters and books only) IOS Press c/o Accucoms US, Inc.
For North America Sales and Customer Service
West Point Commons
Lansdale PA 19446
Tel.: +1 866 855 8967
Fax: +1 215 660 5042 email@example.com