[petsc-dev] Interface for supporting MKL sparse inspector-executor model?

[Richard Mills]

On Mon, Feb 13, 2017 at 8:11 PM, Jed Brown <jed at jedbrown.org> wrote:

> Barry Smith <bsmith at mcs.anl.gov> writes:
>
> >   Richard,
> >
> >    My gut feeling is that the "hint" based system should be used by
> default, who uses linear stuff only once (only people not doing real
> problems :-)). So you could just provide a  single flag to say don't use
> hints, useful for numerical comparisons. (This is all experimental, stuff
> anyways and will evolve).
>
> You could process the hints using command-line options and not make a
> programmatic API at this time.  If there is no good default, we can talk
> about whether the important MKL hints are really MKL-specific or
> something that other PETSc classes might be able to use.  (There are
> uses cases for implicit conversion to ELL or a compressed
> representation.)  But let's not make a programmatic hints API just yet.
>

Thanks for the suggestion, Jed -- I think that's the way I should go.  I'm
probably going to need to do a decent amount of experimentation, and
command-line options are the easiest way to do that.

I'd like to tie these into the PETSc option handling system properly so
that I can specify options for different matrices using prefixes (e.g.,
"-flow_" and "-transport_" in PFLOTRAN), but I don't remember how to do
this.  Can someone point me to where I should look to refresh my memory?

--Richard
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20170214/90e8cd82/attachment.html>