[QE-users] d3hess.x runs out of memory

[Paolo Giannozzi]

On 05/09/2024 09:55, Jan-Benedikt Weiss wrote:
> [Non ricevi spesso messaggi di posta elettronica da 
> jan-benedikt.weiss at tu-dortmund.de. Per informazioni sull'importanza di 
> questo fatto, visita https://aka.ms/LearnAboutSenderIdentification.]
> 
> Dear Paolo,
> 
> thank you for your help. This is what the output says:
> 
>       Number of cells replicated along each semiaxis:    6   5   4
> 
>       Number of cells used for Hessian allocations:    6   5   4
>       Number of cells in the supercell (Hessian):      1287
>       Number of atoms in the supercell (Hessian):    301158
>       Hessian allocation dimensions: 634238748

634238748 8-byte real numbers, about 5 Gb

Paolo

> Best wishes
> Jan
> 
> On 2024-09-05 16:42, Paolo Giannozzi wrote:
>> On 04/09/2024 06:35, Jan-Benedikt Weiss wrote:
>>>
>>> whenever I try to run d3hess.x after a geometry optimisation, it stops
>>> because it runs out of memory at the start of the calculation (All my
>>> structures contain >200 atoms).
>>
>> I guess the problem is in the allocation of hess_d3. What is the number
>> printed after 'Hessian allocation dimensions:'?
>>
>>>  When I run the calculation on more cores to provide more memory
>>
>> I don't think running on more cores of a same node provides more
>> memory. Qquite the opposite: I don't think matrices are distributed, so
>> you may end up with a copy of all arrays on all cores
>>
>> Paolo
>>
>>> I get the following error:
>>>
>>> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
>>>       Error in routine  fft_type_set (6):
>>>    there are processes with no planes. Use pencil decomposition (-pd
>>> .true.)
>>>
>>> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
>>>
>>>       stopping ...
>>>
>>> Is there any way to get around this? I already disabled three-body
>>> interactions and the following phonon calculation shall only be run at
>>> q=0. According to the input description, d3hess.x doesn't provide
>>> options to help with this issue.
>>>
>>> Best wishes and thank you
>>> Jan
>>> _______________________________________________
>>> The Quantum ESPRESSO community stands by the Ukrainian
>>> people and expresses its concerns about the devastating
>>> effects that the Russian military offensive has on their
>>> country and on the free and peaceful scientific, cultural,
>>> and economic cooperation amongst peoples
>>> _______________________________________________
>>> Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
>>> users mailing list users at lists.quantum-espresso.org
>>> https://lists.quantum-espresso.org/mailman/listinfo/users

-- 
Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
Univ. Udine, via delle Scienze 206, 33100 Udine Italy, +39-0432-558216
X SCUOLA ENERGIE RINNOVABILI 16-21/9/2024