[CP2K-user] [CP2K:13950] Re: gas phase calculations

[Matt W]

There are no timings for the gas phase run but it looks like the pw 
operations are dominating - they don't really care how many atoms are 
involved just the box size (you have nearly 8x more grid points). With the 
wavelet poisson solver you only need around 5A space around the system to 
be OK but having to be cubic means you are limited by the largest dimension 
of your molecule. If it is very much bigger in one direction you might be 
better with analytic poisson solver and a non-cubic box.

Matt

On Tuesday, September 22, 2020 at 3:14:58 PM UTC+1 pier... at gmail.com 
wrote:

> Dear Marcella,
>
> Please find as attached document two output files. The first aco-mo.inp 
> contains the results for a crystal made of 78 atoms. This is a normal 
> condensed phase calculation. The second opt.out is for the gas phase system 
> made of 75 atoms. The systems have nothing in common.
>
> I would like to attract your attention that the grid cutoffs of aco-mo.inp 
> are much larger than those of opt.inp, and yet aco-mo.inp is much faster 
> than opt.inp.
>
> I am also running the gas phase system on a single node of a cluster with 
> 40 MPI tasks. The calculation is obviously faster than on my workstation 
> but still frustratingly slow.
>
> Regards,
> Pierre
>
>
> On 22/09/2020 14:00, Marcella Iannuzzi wrote:
>
> Dear Pierre,  
>
> from the timings in the output you should be able to determine where the 
> calculation is spending most of the time and compare to the condensed 
> matter calculation. Most probably OMP is not the most efficient in this 
> case to parallelise. Have you tried with 8 tasks instead. 
> Regards
> Marcella
>
>
> On Tuesday, September 22, 2020 at 1:53:36 PM UTC+2 pier... at gmail.com 
> wrote:
>
>> Hi Marcella,
>>
>> I am trying to run the calculation on my workstation with just OMP (8 
>> threads). This usually works for condensed phase with a two hundred atoms 
>> so I thought it would be fine for just 75 atoms. Anyway, I reduced the 
>> cutoffs and it seems to be helping a lot but it is still slow and require a 
>> lot of memory. Please find as attached document the input file I am using.
>>
>> Regards,
>> Pierre
>>
>>
>> On 22/09/2020 11:06, Marcella Iannuzzi wrote:
>>
>> Dear Pierre,  
>>
>> Without additional information, like input, output, scaling with the 
>> number of processors ..., 
>> it is not possible fo provide any help.
>> Regards
>> Marcella
>>
>>
>> On Monday, September 21, 2020 at 1:09:03 PM UTC+2 pier... at gmail.com 
>> wrote:
>>
>>> Dear CP2K users, 
>>>
>>> I would like to know how to perform a gas phase calculation with CP2K. I 
>>> tried to follow some of the examples available in the tutorials but they 
>>> lead to very slow and heavy calculations for a system with only 75 atoms.
>>>
>>> I used wavelet for the Poisson solver, I set periodic none for both the 
>>> solver and the cell. The cell is cubic with a size of 26.250 A. I am using 
>>> DFT (PBE) with OT. Should the number of grid be changed compared to a 
>>> condensed phase? What about the cutoffs?
>>>
>>> Alternatively, I tried the periodic approach with a larger cell so that 
>>> the molecule "does not interact" with its periodic image .Yet again, the 
>>> calculation is extremely demanding in terms of memory which makes the OS 
>>> kill the job.
>>>
>>> Is there a solution to these problems?
>>>
>>> Regards,
>>> Pierre
>>>
>> -- 
>> You received this message because you are subscribed to the Google Groups 
>> "cp2k" group.
>> To unsubscribe from this group and stop receiving emails from it, send an 
>> email to cp... at googlegroups.com.
>> To view this discussion on the web visit 
>> https://groups.google.com/d/msgid/cp2k/7e7000c8-68b2-4542-b757-450714178594n%40googlegroups.com 
>> <https://groups.google.com/d/msgid/cp2k/7e7000c8-68b2-4542-b757-450714178594n%40googlegroups.com?utm_medium=email&utm_source=footer>
>> .
>>
>>
>> -- 
>> Dr Pierre Cazade, PhD
>> AD3-023, Bernal Institute,
>> University of Limerick,
>> Plassey Park Road,
>> Castletroy, co. Limerick,
>> Ireland
>>
>> -- 
> You received this message because you are subscribed to the Google Groups 
> "cp2k" group.
> To unsubscribe from this group and stop receiving emails from it, send an 
> email to cp... at googlegroups.com.
>
> To view this discussion on the web visit 
> https://groups.google.com/d/msgid/cp2k/afec8cd6-9834-4e05-bdee-3637cb366dd2n%40googlegroups.com 
> <https://groups.google.com/d/msgid/cp2k/afec8cd6-9834-4e05-bdee-3637cb366dd2n%40googlegroups.com?utm_medium=email&utm_source=footer>
> .
>
>
> -- 
> Dr Pierre Cazade, PhD
> AD3-023, Bernal Institute,
> University of Limerick,
> Plassey Park Road,
> Castletroy, co. Limerick,
> Ireland
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20200922/b94fe259/attachment.htm>