<div dir="ltr"><div class="gmail_quote">On Thu, Mar 31, 2011 at 02:50, Barry Smith <span dir="ltr"><<a href="mailto:bsmith@mcs.anl.gov">bsmith@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;">
<div id=":1sl">Don't ask me, ask US! I know Matt and Jed did a couple of things but don't know what is left to be done.<br>
<br>
Everyone please chip in what needs to be done and if they can contribute. I think Hong finished the AO so she might be interested in tackling the rest of what needs to be resolved here and Satish can help here since she did much of the original ctable stuff.</div>
</blockquote></div><br><div>I fixed a non-scalable row map in MatIncreaseOverlap_MPIAIJ_Once, there is a new PetscLayoutFindOwner (cheap, binary searches the row map) which makes it very easy. This change needs to be replicated for BAIJ:</div>
<div><br></div><div><div>changeset: 18476:59612080e9b0</div><div>user: Jed Brown <jed@59A2.org></div><div>date: Sun Mar 13 20:34:14 2011 +0100</div><div>summary: Add PetscLayoutFindOwner and eliminate non-scalable "rtable" from MatIncreaseOverlap_MPIAIJ_Once</div>
</div><div><br></div><div><br></div><div>The column map allocated at mpiov.c:1133,1134 and row map allocated at mpiov.c:1177,1178 need to be replaced with a hash (PetscTable). I think those are all the large allocations, can look more carefully another time.</div>
</div>