## Re: [Libmesh-users] AMR in 3D (fwd)

 Re: [Libmesh-users] AMR in 3D (fwd) From: John Peterson - 2007-05-07 21:44:52 Hi, I don't have a huge amount of additional information to add to the thread, but I definitely think Shun Wang is on to something... I ran a transient, linear heat conduction problem on two different meshes http://www.cfdlab.ae.utexas.edu/~peterson/mismatched_hexes.png The mesh on the left has a maximum edge mismatch of 1 level while the one on the right (you might have to scroll over) has an edge mismatch level of 2. This is a linear problem, and so should converge in 1 'nonlinear' iteration. Here are the results for the first solve on the 1-level mismatched mesh: nonlinear |R| linear |R| 0.000624475 4.25804e-22 4.91167e-17 It's a symmetric problem and converges in 1 iteration, as expected. For the 2-level mismatch mesh, it's quite a different story: linear convergence of Newton's method (for a linear heat conduction problem!) nonlinear |R| linear |R| 0.000616188 3.60638e-22 6.75975e-06 1.27404e-24 1.88506e-07 3.13132e-25 5.23973e-09 1.29476e-26 1.44318e-10 3.53689e-28 3.93687e-12 1.37323e-30 I'm hopeful we can get this issue sorted out quickly, but I'm in the "don't know enough about build_constraint_matrix()" camp as well. -John Roy Stogner writes: > > > ---------- Forwarded message ---------- > > Date: Sun, 6 May 2007 00:31:12 -0500 > > From: Shun Wang > > To: Roy Stogner > > Subject: Re: [Libmesh-users] AMR in 3D > > > > And how do I check out this new piece of code for level-one mesh refinement > > from cvs? I don't want to check out all the code, because I added some code > > somewhere else in libMesh. Thanks! > > You can do a CVS update on individual files by just giving the > filename at the end of the command; however I'd be very wary of trying > to update libMesh piecemeal. You'd probably be better off doing a > complete cvs update and then manually resolving any conflicts between > the libMesh changes and your own changes. > > Also, if any of your code additions can be merged into the main tree > without adverse effects, send us a patch! If you could also give us a > small test case to use as an example program, there's even a chance > that we won't break your patch with future libMesh changes. ;-) > --- > Roy