search

underworldcode / underworld3

https://underworldcode.github.io/underworld3/
Other
21 stars 10 forks source link

Development #251

Closed lmoresi closed 2 weeks ago

lmoresi commented 1 month ago

Current Dev branch needs to become the 1.0 release for main, (or 3.0 if we want to be consistent with underworld 1 and 2)

This is the last major thing in the way of the JOSS release. If there are further bugs, they can be fixed as we find them, but we really need to be in a position to tidy the code and get it reviewed ASAP.

lmoresi commented 1 month ago

Sorry, that last commit was trying to fix #205 - I don't think it will do anything for the others.

lmoresi commented 1 month ago

@gthyagi — can you check to see if the surface-integral workaround in my recent commit does the job. It seems to work for me - I'm adding a fictitious natural bc (automatically) that appears on all the decomposed nodes but (mathematically) does nothing

gthyagi commented 1 month ago

@gthyagi — can you check to see if the surface-integral workaround in my recent commit does the job. It seems to work for me - I'm adding a fictitious natural bc (automatically) that appears on all the decomposed nodes but (mathematically) does nothing

Recent changes yield consistent results for meshes without internal boundary, but the results are still inconsistent when internal boundary is present.

Serial run
(36, -0.0, -0.403867, 0.403867, -0.0, 0.838121, np.float64(0.139687))

mpirun -np 2 python Natural_BC_Bug_with_Int_BD.py
(36, -0.0001, -0.403967, 0.403768, -0.003584, 0.838121, np.float64(0.139687))

mpirun -np 4 python Natural_BC_Bug_with_Int_BD.py 
(36, -0.007265, -0.526843, 0.512313, -0.261537, 1.190362, np.float64(0.198394))

mpirun -np 10 python Natural_BC_Bug_with_Int_BD.py
(36, 0.006844, -0.588039, 0.522585, 0.24637, 1.058837, np.float64(0.176473))