div3D producing NaNs in some 2D simulations

[mattzett]

appears with glow_beam NS example

[mattzett]

It appears to be trying to compute a derivative with a zero value for dx2... Strange that this never was a problem before?!!

[mattzett]

This issue is that the derivative functions are trying to compute the subgrid size from the input data size rather than from the grid structure itself... This is dangerous because various types of calculations require derivatives with a single row of ghost cells, etc...

[mattzett]

This is really quite subtle - the pointers in the derivative functions need to be computed based on the size of the input data (lx2,lx3 local variables)bso that the correct differentials are available to each function. However, when checked whether we need to differentiate in a particular direction, we need to check the grid size itself (x%lx2,x%lx3) in that direction (only differentiate if non-singleton).

[scivision]

so it seems like e594245 src/numerical/calculus/div.f90 fixed this issue by using the meshgrid x% values as noted above, thanks