From: Derek G. <fri...@gm...> - 2007-07-04 18:22:17
|
Thanks for all of the responses guys. Roy, if you have a patch, let me at it and I'll test it a bit before you commit it. Yeah, I knew mine was pretty quick and dirty... but like you say... when that presentation in 3 weeks is staring you in the face... you just get _something_ done ;-) As for why I did the for loop instead of just assigning... that was mostly just for safety. I didn't know exactly what it would do if I used the assignment, so I did the safest thing possible. I'm not doing anything in 3D (for now) so I'm not worried about that. Derek On 7/3/07, Roy Stogner <roy...@ic...> wrote: > On Tue, 3 Jul 2007, Roy Stogner wrote: > > > On Tue, 3 Jul 2007, Roy Stogner wrote: > > > >> If everyone else agrees, I've had a change of heart, and I think we > >> should commit the patch as-is. > > > > Where by "as-is", I mean "without new MeshRefinement flags", not > > "without fixing it to handle non-level-1 mismatches and to avoid > > realigning nodes that aren't on a hanging node side". > > The latter problem was just an efficiency question, and I suspect the > extra-convoluted code necessary to fix the former problem more than > outweighs any speed gains from trying to skip the redundant > realignments. > > I've got a new patch which should handle level-n mismatches in 2D, and > level-n face mismatches in 3D, but it may still leave mesh tears on > level-2+ edge mismatches in 3D. I'm not going to commit it to CVS > since it adds a new helper method to Elem and I'd rather not make > changes to elem.h that would probably need to be reverted shortly > afterward. I can't think of an easy way to fix the edge mismatch > problem with AMR+smoothing; for now all I can suggest to Derek is that > if you're doing any 3D runs you should first set > MeshRefinement::edge_level_mismatch_limit() = 1. > --- > Roy > |