[gpaw-users] LCAO performance - HMOR zeolite

[Marcin Dulak]

Ask Hjorth Larsen wrote:
> Hi
>
> On Thu, 9 Feb 2012, dbjra at student.dtu.dk wrote:
>
>   
>> Dear Gpaw users,
>>
>> I need to perform a number of calculations on the H-MOR zeolite and as
>> the first step I tried to optimize the positions of the atoms in the unit
>> cell. I did the calculation both in the finite-difference and LCAO modes.
>> I was a bit surprised that the LCAO mode is only twice as fast. Do you
>> have any comments on that?
>>
>> The relevant part of the code is below. When performing the LCAO
>> calculation I only change mode to "lcao". GPAW output files are attached.
>>
>> Best regards,
>> Dominik
>>     
>
> Try enabling scalapack:
>
>    GPAW(parallel=dict(sl_default=(4, 2, 64)), ..).
>
> (Also, run it on just one node unless there is more than one k-point in 
> the irreducible Brillouin zone.)
>
>   
i would suggest also to install the latest development gpaw version and 
use it for the whole project.
Here are the instructions: 
https://wiki.fysik.dtu.dk/gpaw/install/Linux/Niflheim/Niflheim.html#niflheim
There have been recently some problems discovered with scalapack that 
gives different results from non-scalapack runs.
Just come to my office if you need help installing.

Best regards,

Marcin

> Regards
> Ask
> _______________________________________________
> gpaw-users mailing list
> gpaw-users at listserv.fysik.dtu.dk
> https://listserv.fysik.dtu.dk/mailman/listinfo/gpaw-users
>
>