dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Sun, Sep 05, 2010 at 08:49:53PM +0200, Marie Rognes wrote:
>
>
> There are some changes/improvements in SLEPc 3.1 vs 3.0. Noteably, the
> krylov-schur eigensolver seems to converge more often when dealing with
> singular generalized eigenvalue problems[1]. However, there are some
> interface changes for instance with regard to the ordering of the
> spectrum. So, I imagine that we could do either of the following
>
> (a) Require SLEPc 3.1(-p2) or higher
> (b) Support versions higher and lower (by continuing to add #ifdefs)
> (c) Do nothing
>
> Opinions?

As far as I understand, SLEPc 3.1-p1 is partly broken but Dorsal has
already been updated to p2.

--
Anders