“…Figure 1 displays the sparsity pattern of
on the first four levels of AMG using the AMGToolbox 29 for matrix dimension
. Here, the drop tolerance is set to 1.e
and fill limit to 200 per row using the ILUTP implementation in
30 with pivoting turned off 6 . A very small drop tolerance and large fill‐in are unreasonable choices for an ILU factorization because they can substantially increase the cost associated with computing, applying and, storing the factors.…”