[petsc-dev] attaching field information to PetscLayout?

[Dmitry Karpeev]

On Sat, May 14, 2011 at 3:32 PM, Jed Brown <jed at 59a2.org> wrote:
> On Sat, May 14, 2011 at 22:26, Dmitry Karpeev <karpeev at mcs.anl.gov> wrote:
>>
>> So the splits are identified by the IS (e.g., "give me the subvec
>> corresponding to the split with this IS")?
>
> Yes, that is VecGetSubVector(Vec,IS,Vec*).
>
>>
>> I assume that will require IS matching.  Is that faster than string
>> matching (I guess it depends on the string/IS length/type).
>
> That depends on the Vec implementation and IS. I think there are many places
> where we would like IS matching (all the submatrix stuff, for example). IS
> is (almost) immutable at this point, therefore we can create a unique token
> for each IS. Then looking up an IS becomes a single hash lookup.

Inside PetscLayout? Token generation then has to beat direct IS matching.