[aspect-devel] AGU meeting feedback

[Timo Heister]

Dear Juliane,

> (1) Cedric sent his input file to some of the aspect users and we tested it
> on different machines (among others I ran the benchmark on Wolfgang's
> machine with his deal.ii when I was in Texas, although back then I did not
> know it was the van Keken benchmark), and there was no difference in the
> results. I can attach the .prm file that Cedric sent to us.

Thanks.

> Finally, I thought about which methods were different in aspect compared to
> other codes and changed the timestepping scheme (I just used the old
> solution everywhere where aspect normally uses BDF2 to extrapolate), but
> this did not change the solution either.

I am not surprised.

> I agree this is a problem we should resolve as soon as possible and when I
> get back to work in one week I would like to run some more tests, e.g. I
> would like to try Wolfgang's idea with setting cR to infinity.

You might need to go into the code and change (assembly.cc)
return std::min (max_viscosity, entropy_viscosity);
to
return max_viscosity;
because even with cR=infinity it might happen that residual=0 and so
entropy_viscosity=0 (shouldn't make a big difference though).

> PS: Timo, I heard you might be there at CIDER next year? I considered going
> there as well, as I heard they might want to use aspect.

Yes, I offered to help but this is not fully planned yet.

-- 
Timo Heister
http://www.math.clemson.edu/~heister/