Randall Mackie writes:
> After profiling our code, we have found that most of the time is spent in
> MatSolve_SeqAIJ_NaturalOrdering, which upon inspection is just doing simple
> forward and backward solves of already factored ILU matrices.
>
> We think that we should be
After profiling our code, we have found that most of the time is spent in
MatSolve_SeqAIJ_NaturalOrdering, which upon inspection is just doing simple
forward and backward solves of already factored ILU matrices.
We think that we should be able to see improvement by replacing these with