[petsc-dev] Scalability of removing null spaces

[Jed Brown]

We recommend to people that they just call MatSetNullSpace() and solve the
singular system. What if reductions are very expensive? Are there Krylov
methods that accommodate a pipelined or deferred null space removal?
Shouldn't be hard to do, but I haven't seen papers on it.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: 
<http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20120722/5f03a065/attachment.html>