You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
After refreshing my memory on the reduction code in grudge, I think I see why this isn't working. The global reductions call MPI internally, and they do it in a way that's not compatible with pytato (i.e., they call MPI directly instead of creating send/recv nodes). So as currently written they can't be used in compiled functions. IIRC we would need to add MPI reduction nodes to pytato; not sure how much work would be involved with that.
@tulioricci Is there any way this could be rewritten in a more local way? I don't suppose an elementwise integral could be used somehow instead?
Edit: How rapidly does the blowing velocity change? Another option might be updating it once per timestep, outside the RHS. (Could slow things down though, since the reductions have uncompiled code in them.)
I don't see a way to treat it with elementwise operations because I need the actual integral in the entire volume for the current level of approximation/simplification of the problem. If there is a way to make it lazy friendly, I am all ears.
It is such a crude approximation that I don't see a problem making it fixed during the step. We can chat over slack on the particular details on passing it through to the RHS.
For reference, this is what I am trying to prescribe at the boundary:
Getting a strange (platform-independent) error when trying to perform an integral using
integral@gruge
in McKenna burner driver. The driver code is (https://github.com/illinois-ceesd/drivers_mckenna/blob/main/McKenna_phenolics/2D_burner_phenolics.v1a.py) (#900).Reproduce:
python -u -O -m mpi4py 2D_burner_phenolics.v1a.py --lazy
Error looks like this:
The text was updated successfully, but these errors were encountered: