[Wien] NUME parameter

[Jorissen Kevin]

To give you an idea, I did a calculation (InGaN) using NAT = 72 and Emax = 3.5 Ry; it needed 1412 bands.  In another calculation (carbon) with NAT = 60, it took only 1100 bands to go up to Emax = 5.0 Ry.  In yet another calculation with NAT = 20 (very large cell of TiO), 2000 bands got me only up to +2.5 Ry.
I think the size of lapw1 (the biggest program) scales approx. like nmatmax * (nmatmax + nume).  So you should 'balance' nmatmax and nume, depending on the available memory.
Let's focus on 'real' ; nmatmax = 20 000 and 8b per real number makes
8 * 20 000 * (20 000 + nume) b.  So nume could go up to 15 000, technically speaking.
You'll never make that much, since my simple formula is probably not accurate, and you maybe don't really have the full 2 Gb available (OS might claim some?).  And you'd need *huge* calculations to use nume 15000 (of course, with nmatmax 20 000, you can do huge calculations) ...

I hope this made sense?

Kevin.


> -----Oorspronkelijk bericht-----
> Van: Michael Gurnett [mailto:michael.gurnett at kau.se]
> Verzonden: woensdag 19 november 2003 15:26
> Aan: wien at zeus.theochem.tuwien.ac.at
> Onderwerp: [Wien] NUME parameter
> 
> 
> Is the size of NUME at all hardware related, or should one 
> increase this
> value from the default for machines with larger memory. Am about to
> compile on a 2 Gb machine. Was going to use the following parameters
> 
> NMATMAX 20000 real ... 10000 complex
> 
> NUME ???? real ... ???? complex
> 
> Anyone have any suggestions or experience if this is ok
> 
> Thanks
> 
> Michael
> 
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> 
>