Optimum block size for the computation of Schur algorithm using a CRAY

Author

Bae, Jinho ; Chun, Joohwan ; Lee, Sungsin

Author_Institution

Tech. Res. Inst., Daeyang Electr. Co. Ltd., Pusan, South Korea

Volume

3

fYear

2001

fDate

2001

Firstpage

1719

Abstract

We present new results on the parallel and pipelined implementation of the Schur (1986) algorithm using supercomputers such as CRAY Y-MP. We are developing Fortran Schur algorithm code which is efficient on the supercomputer, CRAY Y-MP, and provides concurrency as well as vectorization utilizing Level-3 BLAS. We show that the block Schur algorithm using the block submatrices is faster than that using individual matrix elements. The optimum submatrix size for the Schur algorithm on supercomputer is also suggested

Keywords

FORTRAN; geophysics computing; inverse problems; mainframes; matrix algebra; multiprocessing systems; parallel algorithms; CRAY Y-MP; Fortran Schur algorithm code; Level-3 BLAS; block submatrices; concurrency; inverse scattering problems; matrix elements; ocean environment; optimum block size; optimum submatrix size; parallel implementation; pipelined implementation; power series; supercomputer; supercomputers; vectorization; Concurrent computing; Inverse problems; Laboratories; Linear algebra; Oceans; Power engineering and energy; Supercomputers; Symmetric matrices; System identification; Transfer functions;

fLanguage

English

Publisher

ieee

Conference_Titel

OCEANS, 2001. MTS/IEEE Conference and Exhibition

Conference_Location

Honolulu, HI

Print_ISBN

0-933957-28-9

Type

conf

DOI

10.1109/OCEANS.2001.968092

Filename

968092