27 Aug
2019
27 Aug
'19
10:30 p.m.
On 27/08/2019 17:14, Sander Rhebergen wrote:
I think this is related to me using a PeriodicSquareMesh, because when I use a unit square mesh there are no error messages and the solver seems to run. Could you tell me what is happening in
x = dm.getVecClosure(coordsSection, coordsVec, pt).reshape(-1, dim).mean(axis=0)
and whether it can be modified for the PeriodicSquareMesh?
I think this is because of the way firedrake constructs periodic meshes (by first constructing a 3D mesh and then warping it). It looks like PETSc's opinion of the coordinates are the unwarped 3D coordinates. I think if you set dim = 3 it should work. Maybe Matt can chime in with the right code to calculate dim. Patrick