Page 1 of 1

Error in Gw calculations

Posted: Tue Feb 09, 2021 10:29 am
by mjakhar7665
Hello sir
i submit the job for gw using ( mpirun -np 16 ~/q-e-qe-6.3/yambo-stable/bin/yambo -F yambo.in > yambo.out )
Then the pc is hanged and after some time the calculation is stopped without showing any error
here are my input and log file etc.
please suggest me the solution and check my yambo.in parameters.
Thanks in advance

Re: Error in Gw calculations

Posted: Tue Feb 09, 2021 12:20 pm
by Daniele Varsano
Dear Mukesh,
here some issues:
1) No need to redirect the output
2) You are running on 16 cpu, but the input is set for 32 cpu
3) Think to assign a parallel strategy also for the dipoles: in general assign cpu on bands "c" and "v"; the same for X, avoid "q" in SE.
4) Check if there is some error message in one of the log files
5) The values in the input for QPkrange are not meaningful, you usually want to calculate the GW corrections for few bands across the gap (here you are asking for more the forty thousands corrections.
6) Think also to update to a newer version of the code that you can download via:
git clone git@github.com:yambo-code/yambo.git yambo

Best,
Daniele

Re: Error in Gw calculations

Posted: Tue Feb 09, 2021 5:27 pm
by mjakhar7665
Thanks a lot for quick responce
i tried as you suggested (attached as yambo1.in) but the system is hanged while i submit this job with this input file and the calcualtion is proccesed.

I make another input file (attached as yambo2.in) in which the calculation is proccesed and it's still going on.
Please suggest me the soultion.

Re: Error in Gw calculations

Posted: Wed Feb 10, 2021 5:42 pm
by Daniele Varsano
Dear Mukesh,
It is not very clear what it is your problem.
From the report file I can see that in one case yambo is calculating the screening, in another it is not clear if it is calculating or it stopped.
The difference is in the number of CPU used and the parallel strategy.
In any case, my suggestion to be more efficient in terms of memory distribution and workload is to parallelise over conduction and valence bands e.g.:

Code: Select all

X_all_q_CPU= "1 1 8 2"                      # [PARALLEL] CPUs for each role
X_all_q_ROLEs= "q k c v"                    # [PARALLEL] CPUs roles (q,k,c,v)
or

Code: Select all

X_all_q_CPU= "1 1 16 1"                      # [PARALLEL] CPUs for each role
X_all_q_ROLEs= "q k c v"                    # [PARALLEL] CPUs roles (q,k,c,v)
Best,
Daniele