[Wien] Speedup for NL-vdW functionals in MPI (for WIEN2k_19)

[Gavin Abo]

Thanks, I have added that to my list of WIEN2k 19.1 patches in the 
README at [1] which seems to be becoming quite long now that I have 
brought it up to date.

I know there is at least one other fix that I don't have in my list, 
since there is not enough information available to patch for those.  
Though from the post at [2], it looks as if that fix can likely be 
expected in the next future version of WIEN2k.

[1] https://github.com/gsabo/WIEN2k-Patches/tree/master/19.1
[2] 
https://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/msg19019.html

On 2/21/2020 2:25 AM, Tran, Fabien wrote:
> Dear WIEN2k users,
>
> If you are using the nonlocal van der Waals functionals in parallel MPI mode,
> then you should copy the two attached Fortran files in $WIENROOT/SRC_nlvdw/
> and recompile:
> make all
> cp nlvdw ..
> cp nlvdw_mpi ..
>
> The scaling with the number of processors was extremely inefficient, and
> with these updates this problem is gone. Note that this will work only if you
> are using WIEN2k_19.
>
> F. Tran