MP2 CPASSERT failed

Vladimir Rybkin rybk... at gmail.com
Wed Nov 1 10:23:24 CET 2017


Dear Katarina,

I still believe it is memory. For a single water molecule you may try not 
to specify memory at all. Then it will likely go through and you it will 
print how much memory it actually needed. The second thing is your job 
script. The memory keyword is for an MPI rank, so the performance depends 
on how you run the job. In the first crashing file I see you are running on 
168 ranks with 1 thread. It may be a good idea to decrease the former 
increasing the latter.

Yours,

Vladimir

среда, 1 ноября 2017 г., 9:25:58 UTC+1 пользователь katarína stančiaková 
написал:
>
> Dear Vladimir,
> thank you for your suggestions! I've tried it and I also played with 
> memory settings, number of cores/nodes used, but still I got exactly the 
> same error. I also tried to run the same calculation on a single water 
> molecule in my unit cell, but the simulation crashes at the same point 
> again. Can be there anything else wrong? 
>
> Thank you for any further suggestions
> Katarina
>
> Dňa pondelok, 30. októbra 2017 16:28:13 UTC+1 Vladimir Rybkin napísal(-a):
>>
>> Dear Katarina,
>>
>> it's likely a memory issue. I would as a first step suggest 
>> 1) to change 
>>  GROUP_SIZE -1 
>> to 
>>  GROUP_SIZE 1 
>> in the &WF_CORRELATION
>> Since this has to do with the integral calculation load over the 
>> processors this can help. In fact, for a system where integral evaluation 
>> cost is not dominant (the larger the system the more it's so) it's better 
>> not to use -1, it will be faster. Your system is relatively large.
>>
>> 2) Your memory specification: MEMORY 60000 looks huge, as those are MiB. 
>> Try to reduce it. I would say 600 would suffice. If you ask for more 
>> memory, than is available on your machine the job will immediately and 
>> inevitably crash.
>>
>> Yours,
>>
>> Vladimir
>>
>> воскресенье, 29 октября 2017 г., 12:50:45 UTC+1 пользователь katarína 
>> stančiaková написал:
>>>
>>> Dear CP2K users,
>>> I am trying to run MP2 calculations. My ultimate goal is to run geometry 
>>> optimization (therefore I chose RI implemetation), however so far I was not 
>>> able to get even calculated energy of my system. At first I run scf 
>>> optimization with HF and then I wanted to use that wavefunction as the 
>>> input for MP2. It seems that at the beginning it works fine, but then the 
>>> job crashes and I dont understand what the error message means. I would 
>>> appreciate any help with that.
>>> Input/output files are attached. Thank you!
>>>
>>> Katarina
>>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20171101/f1db4b2e/attachment.html>


More information about the CP2K-user mailing list