Solver trials for '4th component' models ... slow/unstable performance in Flow

[EdmundStephens]

[2024.04] The following are three test cases showing slow and/or unstable solver performance in Flow compared to the identical case in tNav and Eclipse (where licences available). All models are run on my laptop which is a MS Surface Pro, 64 bit Windows 10, intel i7, 2.5 GHz with 4 cores.

Comparison summary

• Polymer: Flow successfully reproduces the profile but performs 10 x slower than tNav. (POLYMER licence not available for Eclipse)
• Solvent: Flow takes a number of small timesteps then stops within 1 day of the run
• Thermal: Flow successfully reproduces the temperature profile as Eclipse (accepting slight differences as documented between these models), but performs 10 x slower than tNav. (tNavigator matches Eclipse but has different sign convention for energy flows and weirdly the interface does not export well head or block temperatures - their bug!).

Models including all input, parameter files, summary and grid output:
polymer.zip
solvent.zip
thermal.zip

Comparison plots below from these projects [note log scale on the TCPU comparison]. Intuitively it seems to me these may all be manifestations of a common underlying issue with the solver using the 'additional 4th component' formulation of the equations, which is implicated in all these cases.

These might also be adapted as useful test cases, to which you are welcome.

Solvent model run output

...
Starting time step 31, stepsize 0.0102146 days, at day 0.281044/0.387998, date = 01-Jan-2020
 Newton its= 1, linearizations= 2 (0.0sec), linear its=  3 (0.0sec)

Starting time step 32, stepsize 0.0306437 days, at day 0.291259/0.387998, date = 01-Jan-2020

Problem:
Timestep chopped to 0.010112 days

Starting time step 32, stepsize 0.0101124 days, at day 0.291259/0.387998, date = 01-Jan-2020
[DESKTOP-8VC26F1:00059] *** An error occurred in MPI_Allreduce
[DESKTOP-8VC26F1:00059] *** reported by process [4038262785,0]
[DESKTOP-8VC26F1:00059] *** on communicator MPI_COMM_WORLD
[DESKTOP-8VC26F1:00059] *** MPI_ERR_TRUNCATE: message truncated
[DESKTOP-8VC26F1:00059] *** MPI_ERRORS_ARE_FATAL (processes in this communicator will now abort,
[DESKTOP-8VC26F1:00059] ***    and potentially your MPI job)
[DESKTOP-8VC26F1:00056] 1 more process has sent help message help-mpi-errors.txt / mpi_errors_are_fatal
[DESKTOP-8VC26F1:00056] Set MCA parameter "orte_base_help_aggregate" to 0 to see all help / error messages

Polymer test result

Solvent test result

Thermal test result