Increase default workset size in Albany about albany

[mperego]

I agree on increasing the default workset size (this will help to mitigate the overhead of the Intrepid2 functions) .
However, I think it is best not to put workset size = -1 in all the tests (even for CUDA only builds) because then the case of multiple worksets per process is never tested.

[ibaned]

I think its early to base the default specifically on the GPU use case. I suspect Albany usage of GPU memory is nowhere near optimal yet. A while back I raised the default from 50 to 1000 for the PUMI Discretization; I propose we do the same for STK and see how things go from there. I am certain there are test problems with >1000 elements, so multiple worksets would still get exercised.

[ikalash]

I am fine with 1000. @ibaned , can you please make this change if there are no objections? I'll assign the issue to you. It will be interesting to see how the CUDA test results change with this increase in workset size.

[ibaned]

so my memory was off and PUMI was using ten thousand, but I changed it everywhere to one thousand. I'm closing this for now, lets reopen if it needs changing again or I broke something.

[agsalin]

Dan: I'm getting errors in my nightly build from this change. For Example:
examples/FELIX/FO_GIS/input_fo_gis_unstruct.xml

Error has to do with type -- not clear why you chose to free it from an .

p=0: *** Caught standard std::exception of type 'Teuchos::Exceptions::InvalidParameterType' :

Error! An attempt was made to access parameter "Workset Size" of type "int"
in the parameter (sub)list "Albany Parameters->Discretization"
using the incorrect type "Albany::AbstractMeshStruct::{unnamed type#1}"!

Throw number = 1

[mperego]

Hi Andy, that has been fixed (see #47).