<div class="gmail_quote">On Wed, Dec 14, 2011 at 20:03, Dmitry Karpeev <span dir="ltr"><<a href="mailto:karpeev@mcs.anl.gov">karpeev@mcs.anl.gov</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
If you don't care about the extra storage, you could put the conserved variables and the "constraint variables" in separate splits, make the Jacobian a MatNest and solve with the consevative subblock most of the time (away from the feasible set boundary). That submatrix would retain the nice structure and all. I'm thinking about adding "bounded variables" to libMesh in that way. </blockquote>
</div><br><div>In TS, we formally add variables to reach a standard form, then rearrange and eliminate on paper to get to a computationally convenient form. I wonder if with some thought, we could do the same for SNESVI.</div>