[Wien] nHOw to accelerate the caculation?

[Stefaan Cottenier]

> I am caculating the relaxation of surface, I use a 22 stoms supercell. The
> caculation is so large, I want to ask, are there any method to accelerate
> this caculation?

You ALWAYS want a calculation to finish as fast as possible, so what you 
always should do is:

* use APW+lo (not LAPW)
* use the smallest k-mesh that still gives you reliable results (needs testing)
* use the smallest Rkmax that still gives you reliable results (needs testing)
* make your Rmt's as large as you can, but WITHOUT breaking the rules that are 
described in the FAQ
* do not use the precompiled version of the code, but compile yourself and 
link with mkl or atlas
* for surfaces: don't make the vacuum larger than needed (needs testing)

I typically need about 20 steps to relax a surface, which indeed takes a lot 
of time. But that's how it is...

Stefaan