[petsc-dev] About parallel of ILU

[Jed Brown]

"Chen Gang" <569615491 at qq.com> writes:

> Dear Professor,
>
>
> I'm writing about this mail about the ILU algorithm in PETSc. I have noticed that the ILU in PETSs cannot be paralleled, and the Hypre providing the parallel implement of ILU.
>
>
> My questions is : why the PETSc does not provide the parallel implement of ILU directly, instead of linking to the Hypre to obtain the Parallel ILU algorithm?

We have limited time and try to focus on things we think will provide the most value. If the implementation in Hypre works for you, by all means, go ahead and use it.

ILU isn't a great parallel algorithm -- it isn't very strong for large problems (a math issue) except when using fancy (usually expensive) multilevel ILU methods that overlap with some algebraic multigrid methods, and it doesn't parallelize very well (also a math issue, unless you use reorderings that undermine any advantage relative to additive Schwarz).

If you want to contribute a parallel ILU, we'd be happy to review a merge request.