[CP2K-user] [CP2K:18930] RI_MP2 memory

['Frederick Stein' via cp2k]

 Hi Niamh,

Apart from Augustin's hint, I have some more ideas:
Your machine should have more than enough memory. We run calculations with 
systems of similar size on Piz Daint (with much less memory than you). 
Please provide the output file, without it, we can just guess.
Can you maybe reduce the number of ranks per node?
If it breaks in the integral calculation part, you could decrease the 
available memory for MP2 (MEMORY keyword in the WF_CORRELATION section). 
You will not need that much memory.
But I agree with Augustin, without any further information, it is difficult 
to narrow it down to the actual issue as you apparently have more than 
enough memory available.

Best,
Frederick

Augustin Bussy schrieb am Donnerstag, 15. Juni 2023 um 10:33:24 UTC+2:

> Hi Niamh,
>
> it is hard to know where things go wrong without your output file. Is it 
> during the SCF or MP2? Generally, I see that you are using a RI_METRIC of 
> type identity. This does not bring much for RI_MP2. The use of a RI metric 
> helps to create sparsity, which is only really exploited in low-scaling 
> algorithms.
>
> You can try setting the keyword 
> WF_CORRELATION%CONONICAL_GRADIENTS%FREE_HFX_BUFFER to FASLE. In the past, 
> this was known to create crashes with MP2 gradients. Depending on the CP2K 
> version you are using, this keyword might be in a different input section.
>
> Best,
> Augustin
>
>
> ------------------------------
> *From:* cp... at googlegroups.com <cp... at googlegroups.com> on behalf of 
> Niamh O'Neill <niamh9... at gmail.com>
> *Sent:* Wednesday, June 14, 2023 7:57 PM
> *To:* cp2k <cp... at googlegroups.com>
> *Subject:* [CP2K:18928] RI_MP2 memory 
>  
> Dear CP2K developers,
>
> I am attempting to compute energies and forces for a system containing 6 
> Na Cl ions each and 52 waters at RI_MP2 level. 
> I am having issues in terms of memory, and get a segmentation fault for 
> invalid memory reference.
> The system I am running on has 256 GB RAM per node and I am running on 32 
> nodes . (128 cores per node and have tried both 64 and 32 tasks per node) 
> and I start from a PBE wavefunction.
>
> Do you know why these issues might be happening?
>
> I have attached my input files below.
> Thank you in advance for the help.
> Best wishes,
> -Niamh
>
> -- 
> 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 cp2k+uns... at googlegroups.com.
> To view this discussion on the web visit 
> https://groups.google.com/d/msgid/cp2k/cf2610e9-22be-4829-a355-f04729d1e7e0n%40googlegroups.com 
> <https://groups.google.com/d/msgid/cp2k/cf2610e9-22be-4829-a355-f04729d1e7e0n%40googlegroups.com?utm_medium=email&utm_source=footer>
> .
>

-- 
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 cp2k+unsubscribe at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/5cdf9ab3-d24e-496c-9315-c04b42f475abn%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20230615/ae254a13/attachment.htm>