[petsc-users] questions about PetscSF

[Matthew Knepley]

On Tue, Mar 4, 2014 at 11:59 AM, Fande Kong <fd.kong at siat.ac.cn> wrote:

> Jed,
>
> Thanks,
>
> I readed the node you gave. I still have some questions.
>
> (1) One root can be related with zero, one or many leaves, right? One leaf
> only can be related with zero or one root.
>

Yes, that is what makes it a tree, rather than a DAG.


> (2) In Algorithms section. Could you please give me a very simple example
> to demonstrate how  'Extracting a submatrix from a sparse matrix', '
> Ownership discovery and transfer' and 'Graph distribution'. The description
> is hard for me to understand.
>

These are not simple things. What exactly are you looking for? You can see
graph distribution
done with PetscSF in DMPlexDistribute().

   Matt


> Fande,
>
>
> On Mon, Mar 3, 2014 at 5:54 PM, Jed Brown <jed at jedbrown.org> wrote:
>
>> Fande Kong <fd.kong at siat.ac.cn> writes:
>>
>> > Hi all,
>> >
>> > I was wondering mechanisms of the object PetscSF. What are definitions
>> of
>> > roots and leaves? Do roots/leaves associate with the data we want to
>> > receive/send?  For the function Bcast, it seems that we transfer data
>> from
>> > roots to leaves. But in another function Reduce, it seems that we move
>> data
>> > in the opposite direction (from leaves to roots). These kind of
>> mechanisms
>> > possibly make users confused.
>>
>> The SF graph is asymmetric so that it can have cleaner semantics.  Read
>> my note and reply here if something is still unclear.
>>
>> http://59a2.org/files/StarForest.pdf
>>
>
>


-- 
What most experimenters take for granted before they begin their
experiments is infinitely more interesting than any results to which their
experiments lead.
-- Norbert Wiener
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20140304/78b521a5/attachment.html>