I am testing MKL ilu factorization. As it only works for ilu0, I provide a sparse matrix pattern that corresponds to ilu(k). One thing I observe is that the performance is getting slower when I increase the level of fills (I have compared with matlab for comparison). Is this expected ?
For instance, I have the following results from mkl.
level0, nnz = 9,027,150, time = 0.598093, matlab = 0.766352
level1, nnz = 12,816,050, time = 1.3787, matlab = 1.31876
level2, nnz = 15,450,825, time = 2.57172, matlab = 1.9329
This seems to be odd. Does mkl perform something special ? any pivoting ? or more specialized for sparser patterns ?