Unpivoted dense LDL

Though Cholesky factorizations are ideal for Hermitian Positive-Definite matrices, unpivoted \(LDL^T\) and \(LDL^H\) factorizations are frequently used for significantly larger classes of matrices, with the most notable example being Symmetric Quasidefinite matrices, which were introduced by [Vanderbei1995] and subsequently analyzed as a simple rescaling of a nonsymmetric positive-definite matrix in [GSS1996]. While the stability properties of unpivoted \(LDL^H\) factorizations of dense quasidefinite matrices are of intrinsic interest, there are significantly more practical applications for sparse matrices, where dynamic pivoting incurs typically substantially increases both the memory requirements and operation count of a factorization.

References

Vanderbei1995

Robert J. Vanderbei, Symmetric quasi-definite matrices, SIAM Journal on Optimization, Vol. 5, No. 1, pp. 100–113, 1995. DOI: http://dx.doi.org/10.1137/0805005

GSS1996

Philip E. Gill, Michael A. Saunders, and Joseph R. Shinnerl, On the stability of Cholesky factorization for symmetric quasidefinite systems, SIAM Journal on Matrix Analysis and Applications, Vol. 17, No. 1, pp. 35–46, 1996. DOI: http://dx.doi.org/10.1137/S0895479893252623

C++11 implementation

C++11 subroutines

C++11 header

C99 wrappers

C99 header

Python wrappers

C++11 test driver

C++11 example driver