Hi Alexander, On Tue, Nov 13, 2012 at 12:53 PM, Alexander Grayver <agrayver at gfz-potsdam.de > wrote:
> > Jack, > > Thanks for the comment. > > Could you comment on or give a reference to the blow up in memory one > observes when running general purpose direct solvers in parallel? > > People tend not to publish these types of negative results. > P.S. Does Clique support pivoting? > > Not yet. It is something that I will eventually make time for, but its importance for solving indefinite equations is somewhat overblown. It is surprisingly easy to directly solve many time-harmonic wave equations with a Cholesky-like LDL^T factorization. In extreme cases, one can add a small imaginary diagonal shift, chosen in the proper direction to damp propagating waves (this depends upon your convention for time-dependence for any radiating boundary conditions, e.g., exp(i w t) or exp(-i w t)) and then wrap the solves in a few iterations of GMRES(k). This is often referred to as iterative refinement. I will likely implement a few more high-performance triangular solve algorithms before I add support for pivoting. Best, Jack -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20121113/c0dda333/attachment.html>
