ParBilinearForm w/ non-ConstantCoefficients in MassIntegrator taking extremely long to Assemble

[j-signorelli]

Hi,

For H1-conforming elements, I have code that uses two ParBilinearForms: one with just a MassIntegrator and a separate one with just a DiffusionIntegrator. Whenever I use non-ConstantCoefficients (Like GridFunctionCoefficient), the one with the DiffusionIntegrator Assembles quite quickly regardless with no issues, but the one with the MassIntegrator takes an exceptionally long time. When I run with 1 process, it is not too bad, but the more I run with the seemingly expontentially longer it takes. Is this a common problem? Is there anything I can do to speed this up?

Thank you for any assistance. I can provide more information if helpful.

[j-signorelli]

Just to add more info to this: Right before submitting this discussion post, I started a run with 14 procs, and over 4 hours later it still is assembling, even though it only took a couple minutes on 1 proc.

[jandrej]

Can you post a minimal failing example that we can look to possibly reproduce the issue?

[j-signorelli]

Creating a minimal failing example helped me find my issue, thanks! It was nothing MFEM-related.