Bjarne S. Andersen
The Danish Meteorological Institute
Copenhagen, Denmark
and
John A. Gunnels and Fred G. Gustavson
IBM T.J. Watson Research Center
Yorktown Height, USA
and
John K. Reid
Atlas Centre
Rutherford Appleton Laboratory, UK
and
Jerzy Wasniewski
Technical University of Denmark
Lyngby, Denmark
Abstract:
Two new packed storage data formats for symmetric positive-definite matrices have been developed recently. A paper on the packed recursive data format has already been published in TOMS[1], and one on the block packed hybrid has been submitted to TOMS[2] recently. These two algorithms were explained in the two previous talks.
Here we will explain how we tuned these algorithms and compared their speeds with the packed and full-format Cholesky algorithms of LAPACK. A selection of our results, obtained on six different popular computers, will be displayed.
References:
1. B.S. Andersen, F.G. Gustavson, J. Wasniewski. "A recursive formulation of Cholesky factorization of a matrix in packed storage". TOMS Vol. 27, No. 2, June. 2001, pp. 214-244.
2. B.S. Andersen, J.A. Gunnels, F.G. Gustavson, J.K. Reid, and J. Wasniewski. "A Fully Portable High Performance Minimal Storage Hybrid Format Cholesky Algorithm". It is submitted to TOMS.
3. B.S. Andersen, J.A. Gunnels, F.G. Gustavson, J.K. Reid, and J. Wasniewski. "Fortran 90 Subroutines for the Cholesky Algorithm in Blocked Hybrid Format". In preparation.