[petsc-dev] IMP prototype

[Victor Eijkhout]

On Jan 2, 2014, at 10:50 AM, Jed Brown <jedbrown at mcs.anl.gov> wrote:

> I find simple demonstrations as unconvincing as most patents.  99% of
> the work remains in extending the idea into something practical.  It may
> or may not pan out, but we can't say anything from the simple
> demonstration alone.

Maybe you and I disagree on what I'm demonstrating. My goal was to show that my notion of parallelism generalizes MPI & tasking notions. Not that I have a better notation for VecScatters.

And from this demonstration we can definitely say something: namely that I've shown how one API can address multiple types of parallelism. That's more than any other system I know of. The most interesting question to me is if I can do heterogeneity. There I have to wave my hands somewhat vigorously.

But let's be constructive: I want to use this demonstration to get funding. NSF/DOE/Darpa, I don't know. Now if you can't say anything from this simple demonstration, then what would convince you as a reviewer?

> another layer of callbacks

If you have mentioned that objection before it escaped my attention. Yes, I agree that in that respect (which has little to do with the parallelism part) my demonstration is not optimal. The unification of MPI & tasks is going too far there. For MPI it would be possible to have calls like VecScatterBegin/End and instead of a callback just have the local node code in place. For task models that is not possible (afaik). See for instance Quark, where each task contains a function pointer and a few data pointers.

Victor.
-- 
Victor Eijkhout, 512 471 5809 (w)
Texas Advanced Computing Center, 
The University of Texas at Austin