[Wien] NUME parameter

[Peter Blaha]

> NMATMAX            real:    14000    complex:    7000

real: sqrt2*sqrt2=2 --> 2Gb
complex:  7*7=49   *2 (complex) -->  1 Gb !!!!

> NUME                     real:    2800      complex:    1400

I would use such large values ONLY if you are interested in absorption
spectra, where you need that many unoccupied states. For all other
questions of interest reduce it by a factor of 2. (It does not really
hurt, one can always redimension...)

> If one assumes that with an increase of 1 Gb (going from 1 to 2 Gb) allows
> for an increase in NMATMAX of sqrt(2)*10000 (NMATMAX for 1 Gb of memory).
> Have I understood this correctly. Or should I use parameters a bit lower
> just to be on the safe side,
>
> NMATMAX            real:    13000    complex:    6500
> NUME                     real:    2600      complex:    1300


Of course these estimates are only rough numbers and actually we need
more memory. However, this is no problem when you have a paging space
(but it is a problem with queuing systems, which restrict memory allocation).
As long as the matrix fits into memory, pagin is small.



                                      P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------