[petsc-dev] Should we replace hypre PILUT with hypre Euclid?

[Jed Brown]

Fande Kong <fdkong.jd at gmail.com> writes:

> On Tue, Jul 17, 2018 at 12:58 PM, Matthew Knepley <knepley at gmail.com> wrote:
>
>> On Tue, Jul 17, 2018 at 2:55 PM Jed Brown <jed at jedbrown.org> wrote:
>>
>>> PETSc had a Euclid interface until Barry removed it.
>>>
>>> https://bitbucket.org/petsc/petsc/commits/22b6d1ca80e4f79123036fb6afd7eb
>>> 4463b3692a
>>>
>>> See also
>>>
>>> https://www.mail-archive.com/petsc-users@mcs.anl.gov/msg33196.html
>>
>>
>> I would also add that it is considerate of Hypre to have such a terrible
>> ILUT implementation
>> because it reminds you that you should never ever ever use ILUT.
>>
>
> So in theory, can we have a much better implementation? From my
> understanding, algorithm itself is not suitable for parallel computation,
> and it might not be a implementation problem only.

Performance is likely poor with any implementation, but bugs are bugs.