We present the design and testing of an algorithm
for iterative refinement of the solution of linear
equations where the residual is computed with extra
precision. This algorithm was originally proposed in
1948 and analyzed in the 1960s as a means to compute
very accurate solutions to all but the most
ill-conditioned linear systems. However, two
obstacles have until now prevented its adoption in
standard subroutine libraries like LAPACK: (1) There
was no standard way to access the higher precision
arithmetic needed to compute residuals, and (2) it
was unclear how to compute a reliable error bound
for the computed solution. The completion of the new
BLAS Technical Forum Standard has essentially
removed the first obstacle. To overcome the second
obstacle, we show how the application of iterative
refinement can be used to compute an error bound in
any norm at small cost and use this to compute both
an error bound in the usual infinity norm, and a
componentwise relative error bound.
%0 Journal Article
%1 axb-itref-toms
%A Demmel, James W.
%A Hida, Yozo
%A Kahan, W.
%A Li, Xiaoye S.
%A Mukherjee, Sonil
%A Riedy, E. Jason
%D 2006
%J ACM Transactions on Mathematical Software
%K acm floatingpoint lapack linearalgebra toms
%N 2
%P 325--351
%R 10.1145/1141885.1141894
%T Error bounds from extra-precise iterative refinement
%V 32
%X We present the design and testing of an algorithm
for iterative refinement of the solution of linear
equations where the residual is computed with extra
precision. This algorithm was originally proposed in
1948 and analyzed in the 1960s as a means to compute
very accurate solutions to all but the most
ill-conditioned linear systems. However, two
obstacles have until now prevented its adoption in
standard subroutine libraries like LAPACK: (1) There
was no standard way to access the higher precision
arithmetic needed to compute residuals, and (2) it
was unclear how to compute a reliable error bound
for the computed solution. The completion of the new
BLAS Technical Forum Standard has essentially
removed the first obstacle. To overcome the second
obstacle, we show how the application of iterative
refinement can be used to compute an error bound in
any norm at small cost and use this to compute both
an error bound in the usual infinity norm, and a
componentwise relative error bound.
@article{axb-itref-toms,
abstract = {We present the design and testing of an algorithm
for iterative refinement of the solution of linear
equations where the residual is computed with extra
precision. This algorithm was originally proposed in
1948 and analyzed in the 1960s as a means to compute
very accurate solutions to all but the most
ill-conditioned linear systems. However, two
obstacles have until now prevented its adoption in
standard subroutine libraries like LAPACK: (1) There
was no standard way to access the higher precision
arithmetic needed to compute residuals, and (2) it
was unclear how to compute a reliable error bound
for the computed solution. The completion of the new
BLAS Technical Forum Standard has essentially
removed the first obstacle. To overcome the second
obstacle, we show how the application of iterative
refinement can be used to compute an error bound in
any norm at small cost and use this to compute both
an error bound in the usual infinity norm, and a
componentwise relative error bound.},
added-at = {2007-10-09T06:57:46.000+0200},
author = {Demmel, James W. and Hida, Yozo and Kahan, W. and Li, Xiaoye S. and Mukherjee, Sonil and Riedy, E. Jason},
biburl = {https://www.bibsonomy.org/bibtex/2768d3556a52aa3b08e35f66ecee39609/ejr},
doi = {10.1145/1141885.1141894},
interhash = {462217af6d05a18ad137bf556e12de3e},
intrahash = {768d3556a52aa3b08e35f66ecee39609},
issn = {0098-3500},
journal = {ACM Transactions on Mathematical Software},
keywords = {acm floatingpoint lapack linearalgebra toms},
month = {June},
mrclass = {65F10},
mrnumber = {MR2272365},
number = 2,
pages = {325--351},
timestamp = {2007-10-09T06:57:46.000+0200},
title = {Error bounds from extra-precise iterative refinement},
volume = 32,
year = 2006
}