David Wells

Fixing that bug in SAMRAI fixes the plots and the output: I don't know if we can convince everyone to reinstall SAMRAI though. Our best move is to probably just...

This mainly impacts @marshallrdavey right now so lets just fix SAMRAI and have him rebuild it with the new patch.

It looks like this bug is still present in SAMRAI! I'll write them a patch later too.

I found two more issues with LINEAR_REFINE: 1. It uses the coarse ghost cells to fill the entire fine grid cell - i.e., it sets values both on the coarse-fine...

Here's another case which is easier to see:  SAMRAI seems to be calculating offsets incorrectly - in some cases values don't get filled at all and in others we...

I want to verify that running ```cpp { std::vector comps; comps.emplace_back(u_nc_idx, "LINEAR_REFINE", false, "NONE", "LINEAR", false); Pointer hier_bdry_fill = new HierarchyGhostCellInterpolation(); hier_bdry_fill->initializeOperatorState(comps, patch_hierarchy); hier_bdry_fill->fillData(0.0); } ``` results in a conforming...

Yes, that should be the case - I think that's yet another indexing bug present in what SAMRAI is doing here.

Yes. There's at least one more bug which shows up when SAMRAI computes fill boxes used to populate fine-grid interface values.

I spent most of yesterday trying to get this to work. Based on those pictures above I'm going to guess that no-one noticed LINEAR_REFINE for nodal values was broken until...

I can try but, independent of whether or not this is a problem, there are still indexing bugs in SAMRAI's nodal refinement routines so we will still need our own...