Fixing divergence error with AMR (Issue #617)

[tomidakn]

This is to fix the divergence error with AMR reported in #617.

When a MeshBlock contacting a finer MeshBlock is refined, the original code refined normal B-fields on the shared face by interpolating the coarser fields, but they do not exactly match the existing finer fields on the already refined neighbor MeshBlock. With this commit, the shared face fields are copied from the finer MeshBlock, hence enforcing the divergence B constraint.

Prerequisite checklist

• My code follows the Athena++ Style Guide
• My change requires a change to the documentation.
• I have updated the documentation in the Wiki accordingly.
• I have added tests to cover my changes.
• All new and existing tests passed.

[changgoo]

failure from the pgen_compile test:

src/mesh/amr_loadbalance.cpp:1223:47: error: implicit conversion loses integer precision: 'std::vector::size_type' (aka 'unsigned long') to 'const int' [-Werror,-Wshorten-64-to-32]
  const int nf = my_blocks(0)->pmr->pvars_fc_.size();
            ~~   ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~
1 error generated.

[changgoo]

another error:

src/mesh/amr_loadbalance.cpp: In member function ‘void Mesh::PrepareAndSendFaceFieldCorrection(LogicalLocation*, int*, int*, int*, int)’:
src/mesh/amr_loadbalance.cpp:1296:48: error: suggest parentheses around arithmetic in operand of ‘^’ [-Werror=parentheses]
 1296 |             ffc_send_.emplace_back(i, n, xf2^1 + 2, sffc2, -1);
      |                                              ~~^~~
src/mesh/amr_loadbalance.cpp:1326:48: error: suggest parentheses around arithmetic in operand of ‘^’ [-Werror=parentheses]
 1326 |             ffc_send_.emplace_back(i, n, xf3^1 + 4, sffc3, -1);
      |                                              ~~^~~

[felker]

@tomidakn can you check that my latest suggestion is what is intendend? I had to remind myself of the counter-intuitivie operator precedence rules in C re: bitwise operators and arithmetic operators. Interestingly, ChatGPT 4o-mini gets it wrong:

[tomidakn]

I'm afraid I made the same mistake as Chat-GPT.

[felker]

don't worry, I have scolded the AI

[changgoo]

Wow, I also didn't appreciate the pgen_compile test much and often blame that it is too picky. But, now I see why we have that!

[changgoo]

retest this on stellar

[felker]

Wow, I also didn't appreciate the pgen_compile test much and often blame that it is too picky. But, now I see why we have that!

see, I didnt just implement these strict tests just to be annoying! 😂

[felker]

Could you say a bit about why this moved from being a (mostly?) "multigrid-specific" feature?
I see that the LogicalLocationHash and unordered_map's are currently used in Multigrid and meshblock_tree.* Can you say more about the maps / can the same maps be shared across these classes?