This correspondence shows that there is a flaw in the results presented by G. L. Feng et al. (see ibid., vol.44, no.3, p.1152-63, May 1998). A large family of the codes constructed are not double-byte error correcting codes as originally claimed.

Though many parallel implementations of sparse Cholesky factorization with the experimental results accompanied have been proposed, it seems hard to evaluate the performance of these factorization methods theoretically because of the irregular structure of sparse matrices. This paper is an attempt to such research. On the basis of the criteria of parallel computation and communication time, we successfully evaluate four widely adopted Cholesky factorization methods, including column-Cholesky, row-Cholesky, submatrix-Cholesky and multifrontal. The results show that the multifrontal method is superior to the others.

We present an integrated approach to deal with timestamp consistency, and trace based performance analysis techniques for distributed parallel applications. Our trace generation facility captures message passing and system events such as process dispatch with minimal trace overhead. Trace driven analysis tools are developed for post execution analysis, reporting information such as the time stolen by other processes in each node, and the observed message passing time and local wait time for each message. We then present our techniques to reduce total elapsed times based on observed message passing times and local wait times.