Changed the way to pass dot x for staggered scheme
In this merge request (MR), the major changes are:
- pass dot x of all coupled processes to local Jacobian assembers of processes.
- replaced the previous way of passing dot x vis LocalCoupledSolutions with the way of accessing the global vectors of solutions directly.
- apply the known solutions to the initial solutions for the Picard scheme. This gives a reasonable dot x for the first nonlinear iteration of the first time step for the Picard scheme.
Change items 1 and 2 make dot x of all coupled processes being consistency with that from TimeDiscretization. All these changes in this MR make a lightly change in the solutions of the some staggered scheme benchmarks, which looks an improvement but the solutions have to be checked by the corresponding benchmark developers.
Next: remove the stuffs with CoupledSolutionsForStaggeredScheme thoroughly, which are related to computeSecondaryVariableConcrete and postTimestepConcreteProcess.
-
Feature description was added to the changelog? yes.