Closed cgilet closed 3 years ago
Which case runs to completion?
All of them. There might be a problem with the Rayleigh-Taylor regression test with USE_CUDA=TRUE (it uses ppm where most tests use plm), but I haven't looked into this yet. Also note that the create_umac_grown fix changes some answers, even with USE_CUDA=FALSE.
I should add that I haven't tested PLM yet. Want to get that done that before merging.
@drummerdoc I think we should go ahead and merge this and the accompanying PLM PR#113 since they contain bug fixes that can be relevant to CPU only runs. I've checked PeleLM results on CPU for 2d, 3d, EB and non-EB, and these PRs don't introduce changes above what's expected given problem tolerances. I've confirmed that IAMR changes to CPU run are due to the create_umac_grown() bug fixed here. On GPU, IAMR plm agrees with CPU only results. IAMR ppm still needs debugging. PeleLM also needs further debugging.
Godunov on GPU is not quite working yet in IAMR, but at least this fix can be merged so PeleLM can have it.