An accurate SVD algorithm for 2 by 2 triangular matrices
Josip Matejaš · University of Zagreb University Computing Centre (SRCE) · 2010
Using a flne accuracy analysis and the results from (9), a new accurate algorithm for computing the singular value decomposition of 2 by 2 triangular matrices is constructed. It is obtained by combining the new algorithm which is derived in (9) and the algorithm which is coded as an xLASV2 computational routine of LAPACK. Relative error bounds for the output data of the hybrid algorithm are equal to or smaller than the same bounds for any of these two algorithms. AMS subject classiflcations: 65F15, 65G05