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
We should rework and standardize the TranferOperator objects in the package. However, I think there is quite a lot to do before we can tackle that, in order to be efficient:
Move GridapDistributed to PA 0.4. This is necessary in order to get the repartition! functionality.
Figure out dof gid reordering in distributed FESpaces so that redistributed spaces have matching gids. This is needed for the repartition! function to work. We can do this by using the redistribution glue for the redistributed models.
Once we have this, we can use the built-in PA functionality to re-partition the free-values directly for matrix-wise arrays (which should be more efficient than redistributed cell-wise arrays). On top of this, we will end up with cleaner code than can be made modular and re-used.
Finally, I want to create a modular API for transfer operators, so that
redistribution and fine/coarse transfers are independent
we can work with distributed AND serial arrays
functionalities can be reused between operators (like the redistribution code)
work both on arrays and FEFunctions
The text was updated successfully, but these errors were encountered:
We should rework and standardize the TranferOperator objects in the package. However, I think there is quite a lot to do before we can tackle that, in order to be efficient:
repartition!
functionality.repartition!
function to work. We can do this by using the redistribution glue for the redistributed models.The text was updated successfully, but these errors were encountered: