Andrew Ho

I figured out why I was getting test failures. It's because CMake doesn't build the actual test executables unless you run `make tests`. After building the tests, the only failure...

I commented out these lines and the test passes, so it might be ok to remove this now?

I tried HYPRE+UVM+CUDA and now I'm getting many test failures. It looks like they're failing with segfaults. I tried `mpirun -np 1 ./ex1p -pa -d cuda`, and it segfaults at...

I tried with your latest commit, and I still get the same segfault when I try running `mpirun -np 1 ./ex1p -pa -d cuda` with HYPRE+UVM+CUDA. I can try again...

I figured out what was causing my segfaults, it was how I was building HYPRE. I managed to configure HYPRE with `HYPRE_USING_UNIFIED_MEMORY` and `HYPRE_USING_DEVICE_MEMORY` at the same time. I can...

That fixes the issue, and now with HYPRE+UVM+CUDA all the tests pass in debug and release mode. I tried on Tioga with the MI300 and HYPRE+HIP+UVM, and I ran into...

I suspect the problem might be something related to the Tioga system (or at least my setup) rather than mfem. I rebuilt everything and I can get tests to pass...

@dylan-copeland any test which uses AMR derefinement should work close to identically (barring fp rounding) as it did before, though this may be difficult to test since by design the...