[gpaw-users] Speeding up calculation for DFT/MD?

[Ask Hjorth Larsen]

Hi

2013/4/23 Toma Susi <toma.susi at aalto.fi>:
>
> On 22.4.2013, at 18.09, gpaw-users-request at listserv.fysik.dtu.dk wrote:
>>
>> Message: 2
>> Date: Mon, 22 Apr 2013 15:37:44 +0200
>> From: Ask Hjorth Larsen <asklarsen at gmail.com>
>> Subject: Re: [gpaw-users] Speeding up calculation for DFT/MD?
>> To: Lars Smykalla <lars.smykalla at physik.tu-chemnitz.de>
>> Cc: gpaw-users <gpaw-users at listserv.fysik.dtu.dk>
>>
>> 2013/4/22 Lars Smykalla <lars.smykalla at physik.tu-chemnitz.de>:
>>> Hello,
>>>
>>> sorry for interfering.
>>>
>>>> and be sure to use xc='oldLDA' or 'oldPBE' if you can as
>>>> this is much faster than the default libxc implementation.
>>>
>>> What is the difference to the default PBE?
>>> Will oldPBE give the same energy but just faster?
>>>
>>>
>>> Greetings,
>>> Lars
>>
>> They are just implementations outside libxc.  libxc is extremely slow.
>>
>> There may a very small difference due to numerics.
>>
>> Regards
>> Ask
>>
>
> Hi,
>
> Does this mean one should ALWAYS use oldPBE instead of PBE, or just to speed up calculations?
>
> -Toma

For most systems LDA/PBE doesn't contribute significantly to the time
things take, so one wouldn't care.

With LCAO on small to medium systems the remaining steps will speed up
many times, making it more relevant.  Also in FD mode for sufficiently
small systems.

Regards
Ask