Generally there isn't much to change.

You can however, play with performance by selecting different diagonalizers
and choosing only to calculate a subset of the eigenstates (generally a lot
of states will not be populated, and hence there is no need to calculate
all of them).

Check out these flags:
NumberOfEigenstates
Diag.Algorithm

The MRRR algorithm is quite fast, but may in some cases not be as precise
as D&C, you should test and verify.

Please note that NumberOfEigenstates only influences some of the
diagonalization algorithms (they are specified in the manual).

2017-10-30 19:03 GMT+01:00 I. Camps <[email protected]>:

> Hello,
>
> I would like to know how to use SIESTA for system with (almost) a thousand
> atoms. At this time, my system is molecular and have 850 atoms, but this
> number will be increased.
>
> When I ask "how to use", I mean, what flags should be set up in the input
> file.
>
> Note: My SIESTA is compiled with MPI.
>
> []'s,
>
> Camps
>



-- 
Kind regards Nick

Responder a