Scaling of MLMG

[ZiyangHuang]

Greetings,

Is there any scaling information about MLMG linear solver?

I am running simulations on two base grids (64X32X32 and 128X64X64) with 3 levels of refinements. The MLMG solver converges with a similar number of iterations, while the time consumed for each iteration is 20 times more when using the fine grid. I have doubled the number of nodes when I run the fine grid simulation, so it could be 40 times more if I use the same amout of nodes for the fine-grid case.