Minor improvements to examples/douglas_2024

[prj-]

This needs FreeFem/FreeFem-sources#393. BTW, examples/douglas_2024/modecomputeaug.md does not run, should this file be removed, or fixed?

[cmdoug]

Thanks for the suggestion. A serial code (that did not use MPI) was specifically requested by a reviewer, so I'm going to leave this one as-is.

[cmdoug]

Though not planned, I can confirm that this works with FreeFem/FreeFem-sources#393. The fix for the bug you've reported is addressed in #23.

[prj-]

A serial code (that did not use MPI) was specifically requested by a reviewer, so I'm going to leave this one as-is.

That's plain wrong, though. The code does not run as is (with FreeFem++), because you are using load "MUMPS", which yields:

$ FreeFem++ -v 0 examples/douglas_2024/example1_suppmat.md
 -- CASE #1: Convective BC -- 
Attempting to use an MPI routine (internal_Comm_dup) before initializing or after finalizing MPICH

[cmdoug]

My (seemingly incorrect) understanding was that MUMPS is an alias to MUMPS_seq when run using FreeFem++ instead of FreeFem++-mpi. I am not seeing the error you report, perhaps because I have compiled all 3rdparty libraries when building FreeFEM. When I run FreeFem++, I see:

$ FreeFem++ -v 0 examples/douglas_2024/example1_suppmat.md
init MUMPS_SEQ: MPI_Init
 -- CASE #1: Convective BC -- 
Residual norm = 0.131016, Step norm = 76.3352.
Residual norm = 1.69096e-10, Step norm = 6.73719e-06.
Residual norm = 3.17352e-16.
convective BC: err = 0.000191472, #DOF = 68032.
 -- CASE #2: Free Outflow BC -- 
Residual norm = 0.0200298, Step norm = 109.95.
Residual norm = 0.000913451, Step norm = 18.2839.
Residual norm = 4.56383e-05, Step norm = 0.312913.
Residual norm = 3.09426e-07, Step norm = 0.00100929.
Residual norm = 7.25702e-12, Step norm = 2.06395e-08.
Residual norm = 4.64197e-17.
free outflow BC: err = 0.400172, #DOF = 68031.
 -- CASE #3: Balanced Outflow BC -- 
Residual norm = 0.481056, Step norm = 148.737.
Residual norm = 3.97457, Step norm = 3.70177e-08.
Residual norm = 8.84609e-12, Step norm = 1.2913e-13.
Residual norm = 9.67158e-14.
balanced BC: err = 2.14021e-05, #DOF = 68134.
close  MUMPS_SEQ: MPI_Finalize

Then, with FreeFem++-mpi (and your recent commit on FreeFEM):

$ FreeFem++-mpi -v 0 examples/douglas_2024/example1_suppmat.md
 -- CASE #1: Convective BC -- 
Residual norm = 0.131016, Step norm = 76.3352.
Residual norm = 1.69096e-10, Step norm = 6.73721e-06.
Residual norm = 1.55966e-16.
convective BC: err = 0.000191472, #DOF = 68032.
 -- CASE #2: Free Outflow BC -- 
Residual norm = 0.0200298, Step norm = 109.95.
Residual norm = 0.000913451, Step norm = 18.2839.
Residual norm = 4.56383e-05, Step norm = 0.312913.
Residual norm = 3.09426e-07, Step norm = 0.00100929.
Residual norm = 7.25702e-12, Step norm = 2.06395e-08.
Residual norm = 4.59208e-17.
free outflow BC: err = 0.400172, #DOF = 68031.
 -- CASE #3: Balanced Outflow BC -- 
Residual norm = 0.481056, Step norm = 148.737.
Residual norm = 3.97457, Step norm = 3.70008e-08.
Residual norm = 8.88315e-12, Step norm = 1.63808e-13.
Residual norm = 1.04927e-13.
balanced BC: err = 2.14021e-05, #DOF = 68134.

[prj-]

I am not seeing the error you report, perhaps because I have compiled all 3rdparty libraries when building FreeFEM.

Indeed, I don't have MUMPS_seq, so it picks up MUMPS_mpi instead, and it fails.

[prj-]

The latest commit needs FreeFem/FreeFem-sources@e21a086.

[cmdoug]

Thanks for that. Will rebase with your new commit and give this a test later today.

[cmdoug]

All works on my end. Anything else in the pipeline here or should I merge this after FreeFEM #393 is merged?

[prj-]

All good on my end. There are still a couple of MatConvert() in macros_bifbox.idp which could likely be removed, but it's orthogonal to this PR, and I'm not sure how to trigger these to make sure I'm not introducing any regression.

[cmdoug]

Thanks. Yes, I was looking at those once I saw your latest commit on FF. Its delicate since we need to construct a Mat[int, int] block-by-block, and then pass everything into a pre-setup Mat. Will try some things.

[prj-]

Should be good to go now.