2D_jet example broken @ post_process

[sidkamat]

When I try to run the 2D_jet example MFC gets stuck on the post process part. It does not happen with the other cases I've tried. Is this a known problem?

[sbryngelson]

would you add giving us all of your details. how do you build, run, etc. and with what options? what terminal commands? what is your OS and hardware?

[sidkamat]

I am using the instructions provided in the documentation for building MFC. I am using a HPC with AMD EPYC CPUs. Currently I am trying to run the case on a single node of 64 cores. I am using

./mfc.sh run examples/2D_jet/case.py -b mpirun -n 64

in the slurm script

[sbryngelson]

what's an HPC? you mean a cluster? what does the build output look like? what does the slurm script look like? what modules did you load?

[sidkamat]

Yes, a cluster. I loaded openmpi, python, fftw, gcc and anaconda. Aprt from the module loads, the slurm script is a single line that runs the example case (the one I mentioned in my previous comment). The build is not giving me any errors apart from a few warnings while compiling the post_process/m_data_input.f90 file.

[sbryngelson]

please post full outputs here of build and run otherwise not enough info.

[sidkamat]

Here are the ouput files for boht build and run. I have also attached the slurm script I am using.

mfc_run_out.txt
mfc_build_out.txt

workflow_mfc.txt

[sbryngelson]

Ok, this all looks very fine to me. Nothing appears strange and it built all the dependencies fine. I would try just running it again on a smaller case, perhaps run on a different compute node, use fewer cores, etc. Also toggle parallel i/o from true to false (or vice versa) and see what happens. From the full output I see nothing unusual so this should just work. So trying some different cases on different nodes might help figure out what is going on.

[sidkamat]

I turned off parallel i/o. That gave me some errors. Output attached.

mfc_build_out2.txt

I ran the shockdroplet and shockbubble cases. They ran without issue.

[sbryngelson]

I think your problem is the specific case, which appears to be 2D_jet. I think @wilfonba created this case and might know what's going on. It could just be a problem with that specific case.

[sbryngelson]

confirming i can recreate your issue on my machine. it should be specific to this example case, which has some specific properties. I'm curious @okBrian why the test suite didn't catch this and @wilfonba why this is happening more generally. in the meantime unless you need this specific case, you can just move on to the problem youre interesting in! thanks for bringing it to our attention.

[sidkamat]

Thanks for the quick response. Unfortunately, my problem statement is quite similar to this and I'm getting the same errors when I run my case. I'll keep trying for a solution.