Sparse Cholesky Factorization

Description

This application computes Cholesky decomposition for sparse, symmetric positive-definite matrices [1]. It stores the sparse matrix as panels. This application performs I/O using synchronous read()/write() operations.

Input Dataset

To drive this application, we used the skirt matrix from NASA that contains over 45 million double-precision nonzeros and 45,361 columns for a total of 437 MB. The dataset was stored in 8 files, one per processor.

Workload

See above.

Traces

You can download the trace files in the following formats:

• Compressed, ASCII format (84 KB).

Source Code

• Sources for IBM SP-2 and AIX 4.1, in compressed TAR format (7.3MB).

References

[1] A. Acharya, M. Uysal, A. Sussman, J. Saltz, J. Hollingsworth, M. Beynon, A. Mendelson, and R. Bennett. Tuning the Performance of I/O-Intensive Parallel Applications . Proceedings of the Fourth Annual Workshop on I/O in Parallel and Distributed Systems, Philadelphia, Pennsylvania, May 27, 1996.

Last updated on Tue May 27 12:37:44 EDT 1997 by Mustafa Uysal (uysal@cs.umd.edu ).