Re: [Libmesh-devel] Recursive partitioning

2008-07-22 Thread Bill Barth
Having thought about it a little more, the stuff about needing to have the inner domain boundaries line up across processor boundaries is clearly wrong. (I should draw things on the white board first.) Given that, I think a 2-level patitioning should work just fine. Bill. -- Bill Barth, Ph.D

Re: [Libmesh-devel] Recursive partitioning

2008-07-22 Thread Bill Barth
m not sure of the interface. I think that something like you describe, if you can get the inner-domain boundaries to line up across node boundaries, is likely to show some modest improvement. 5-10%, perhaps. Bill. -- Bill Barth, Ph.D., Manager HPC Applications Group [EMAIL PROTECTED]

Re: [Libmesh-devel] Recursive partitioning

2008-07-22 Thread Bill Barth
node do that node's communications (given the overheads). What this really all means is that I should sit down and write some test code to figure it all out. Bill. -- Bill Barth, Ph.D., Manager HPC Applications Group [EMAIL PROTECTED]| Phone: (512) 232-7069 Office: ROC 1.405