MLMG always reports: Initial rhs = 0 when I pass a vector of MultiFab rhs with a constant value.

[ZiyangHuang]

Greetings,

When I used MLMG to solve ABecLaplacian, I gave a constant rhs via rhs[lev].setVal(val) for all lev, where rhs is a vector of MultiFab. No matter what val I input to setVal() (I tried 6.0 and 100.0), MLMG always reported: Initial rhs = 0, and no iteration was performed. However, if rhs is not a constant, then MLMG reported a non-zero initial rhs. I have no idea what causes the issue.

[zingale]

Are you periodic on all boundaries?

[WeiqunZhang]

periodic or Neumann

[ZiyangHuang]

Yes, the boundary conditions are either periodic or homogeneous Neumann.

[zingale]

In that case, solvability requires that the source integrate to zero over the domain, so I imagine that the mlmg is subtracting this off of the source before doing the solve.

[ZiyangHuang]

I see, thanks for your reply.

[WeiqunZhang]

assuming the A is zero in your ABecLaplacian

[ZiyangHuang]

Yes, ascalar and acoef are zero.