[petsc-users] Boundary integral problem

[Azeddine Messikh]

Dear developers
I tried to run ex24.c https://petsc.org/release/src/snes/tutorials/ex24.c.html using the following command line
./ex24 -sol_type quadratic -dm_plex_simplex 0  -field_petscspace_degree 1 -potential_petscspace_degree 1 -dm_plex_box_faces 2,1
I discovered that at
254:     PetscCall(PetscWeakFormSetIndexBdResidual(wf, label, 1, 0, 0, 0, f0_bd_quadratic_q, 0, NULL)); reverses the value of the integrals at the top only. That isThe boundary integral corresponding to node 5 becomes that of 4 and vise-versa. 
Same thing for nodes 5 and 6.
The mesh index is as follows
*4---*5---*6
 |       |      |
 |       |      |*1--- *2---*3
However,  if I use  -dm_plex_simplex 1 there is no problem.
The model is in the form Au = b
the value of b with   "-dm_plex_simplex 0" is[0.25
0.0104167
0.
0.
0.145833
0.
-0.583333
0.177083
0.
0.0833333
-0.28125
0.
0.
-0.6875
0.
-0.75
-0.364583
0.]
and for -dm_plex_simplex 1
[0.0833333
0.0104167
0.
0.
0.145833
0.
-0.583333
0.177083
0.
0.25
-0.260417
0.
1.43404e-16
-0.645833
0.
-0.75
-0.427083
0.]

you can see that the value at node 1 =0.25  and node 4 = 0.0833333 ( simplex 0)
which is reversed, that is,      node 4 =0.25  and node 1 = 0.0833333 (simplex 1)

So, my own calculation shows that at node 1 should be 0.083333 not 0.25.The -dm_plex_simplex 1 gives the correct answer but -dm_plex_simplex 0 gives wrong answer.

Would you please help me in this matter. Sincerely yoursAzeddine M 
 





-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20230925/f2bb14d7/attachment-0001.html>