Victor, I am wondering if you have any plan to implement Induced Dimension Reduction (IDR(s) by Sonneveld and van Gijzen) in PETSc.
Thanks, ================================ Keita Teranishi Scientific Library Group Cray, Inc. keita at cray.com ================================ From: petsc-dev-bounces at mcs.anl.gov [mailto:petsc-dev-boun...@mcs.anl.gov] On Behalf Of Victor Eijkhout Sent: Thursday, April 22, 2010 2:33 PM To: For users of the development version of PETSc Subject: Re: [petsc-dev] Yet more KSPs On 2010/04/12, at 1:28 AM, Matthew Knepley wrote: Someone has gone the extra mile and made an A-orthogonal Lanczos http://arxiv.org/pdf/1004.1622v1 Is that hard? Basically a slightly different inner product. A-orthogonalization gives L2 minimal residuals. (Under the ordinary inner product the residuals are minimal in the A^{-1} norm) It appears to outperform BiCGStab in enough of their tests That might be true since BiCGS is based on the ordinary Lanczos / BiCG. Hm. Why not do an A-orthogonal BicGstab. Hold on. Let me grab some scrap paper..... Victor. -- Victor Eijkhout -- 865 974 9308 eijkhout at cs.utk.edu http://www.cs.utk.edu/~eijkhout/ -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20100422/7b5c5a00/attachment.html>