[Wien] Parallel for spin-polarized 1-kpoint?

[Jeff Spirko]

On Thu, Apr 28, 2005 at 09:51:06PM +0200, Peter Blaha wrote:
> > I was wondering if it would be possible to do a sort of parallelized
> > calculation for spin-polarized cases with a single k-point, for
> > dual-CPU computers.

> > Specifically, can the up and dn spins be run simultaneously for
> > lapw1 and lapw2?

> In principle this is possible and there should be no conflict.
> You have to modify your runsp_lapw. Basically try to run bot spins in 
> background and put a "wait" statement afterwards (but you cannot use 
> "totalexec").

I have actually already tested by backgrounding both lapw1 calls and
inserting the wait statement.  There seems to be no problem even
with totalexec, though the text in the dayfile is a little
scrambled.

> However, I wonder what case is it that you have a magnetic solution and 
> only one k-point ? This might be good only for "atoms", but even for 
> supercell-solids I'd expect a few k-points are necessary.

In our case, there are a few magnetic atoms in a non-magnetic
lattice, and this is a pre-optimization.  But even with many
k-points, there are advantages to this parallelization for
dual-processor machines.  For one, the file structure is easier to
understand.

Best Regards,
-Jeff Spirko

> Regards
>                                       P.Blaha

-- 
Jeff Spirko   spirko at lehigh.edu   spirko at yahoo.com   WD3V   |=>

The study of non-linear physics is like the study of non-elephant biology.

All theoretical chemistry is really physics;
and all theoretical chemists know it. -- Richard P. Feynman