NPAC Technical Report SCCS-094b
Blocked LU Factorization on a Multiprocessor Computer
A Mohamed, Geoffrey Fox, Gregor von Laszewski
Submitted March 30 1992
Abstract
This paper presents new methods of implementing the Level 3 BLAS in
LU factorization used in engineering applications on the
Alliant FX/80 minisupercomputer. Three ways of expressing the LU
factorization in terms of blocked algorithms using Level 3 BLAS
are considered. We also compare the performance of the parallelism
within the computational kernels using a noblock algorithm that
employs Level 1 and Level 2 BLAS with that obtained over the kernels
when using blocked LU with the Level 3 BLAS. (see also SCCS 271)