[QE-users] [QE-GPU] nscf calculation problem

[lorenzo bastonero]

Hi,

I’m currently running on Marconi100 nscf calculations (qe v 6.6) for the interface of two monolayers (which I already studied separately and got no issues). 
I need to calculate several bands (i.e. 8000) for GW calculations. Running on 2 nodes with npool 2, the dynamical RAM per process is above 16Gb. I tried out with npool 1, but I encountered the error "S matrix not positive definite”. I decided then to try with 4 nodes and npool 2 with the option diago_david_ndim='2', then the dyn RAM/process is 14.97Gb, so it has to work now. But the calculation stops without the “CRASH” file and in the output I got "2 total processes killed (some possibly by mpirun during cleanup)”. In the slurm-output there are several segmentation fault error messages. 
I also discussed the problem with my supervisor but we are puzzled. 

Any suggestions on what to do?  

You’ll find attached the main outputs.

Thanks for your precious time.

Kind regards,

Lorenzo Bastonero


-- 
------------------------



Indirizzo istituzionale di posta elettronica 
degli studenti e dei laureati dell'Università degli Studi di TorinoOfficial 
University of Turin email address for students and graduates 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: slurm-1280351.out
Type: application/octet-stream
Size: 44683 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20201022/334fed4d/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: hetero.PdPt.AA.2L.soc.nscf.out
Type: application/octet-stream
Size: 15234 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20201022/334fed4d/attachment-0001.obj>