The necessity of ILU0 for MICP and injectivity runs must be
handled by either error message of fix, this just works around
the issue by forcing the tests to run with ILU0.
The PID part of the timestep controller triggers and results in timestep
sizes depending on the state of the simulation. This can lead to different
timestepping for parallel and serial runs. Using only newton iteration count
for timestep calculations addresses this.
Also, tighten CNV tolerances to avoid solution differences that are accepted by
the simulator as within tolerances, but outside comparison tolerances.
Two data files get compared: one with ACTIONX and one with the same commands but without ACTIONX.
For sequential runs, they produce the same output, but for parallel runs they don't, because the when the COMPDAT keyword is used in an ACTIONX, well connections get created there that are not known to the grid partitioner.
This commit adds sequential and parallel regression tests for the
UDQ_REG-01 example model which uses region-level summary vectors in
field-level UDQs. Restarted tests are missing at this time.
In particular, output the per-cell oil, gas, and water fluid in
place vectors using compatible array names and surface condition
unit conversion tags.
While here, also relax certain parallel and restart case regression
test tolerances to make the tests pass once we generate the arrays.
to handle the situation that a well can be `SHUT` in the Schedule.
If stored in PerfData, when a well is `SHUT`, the inforamtion related to
WINJMULT will be discarded.
Before we want to change how the `PerfData` works, storing
`inj_multiplier` in BlackoilWellModelGeneric is the viable approach.
