yambo-4.2.3 install

[joydev de]

Dear Varsano,

In the previous message, I have told you about p2y conversion problem. This problem is resolved.
But now I am running the next step for calculating the absorption spectra of c60 super shell.
When I submit the job, always I get an error message. Even though, I want to run it in different
number of core in different time . Always the same error come.

Error: mpirun noticed that process rank 0 with PID 476 on node node04 exited on signal 9 (Killed).

Also I am attaching the LOG file, job script, yambo input file .
Is this the mpi error or memory distribution error?

One more question: How I decide the number of core require for a particular job?

Thanking you

Sincerely
Joydev De

[Daniele Varsano]

Dear Joy Dev,
it looks me a memory problem:

Code: Select all

 [M  9.253 Gb] Alloc WF ( 6.564)

as you are using more than 9Gb per core, which is a lot.
Anyway I can see discrepancies between your input/job script file and the log file, you ask for 8 cpus but in the logs I can see 16 cpu are used.
Mosto probably they do not belong to the same run.
Anyway you can try to solve the problem in the following ways:
1) Using less cores using the whole node, if you have 4Gb per core and 16 core in the node, if you use 4 cores you should be able to do the calculation.
2) Reduce the required bands, now you have 1600 bands, are you sure you need that number? Usually one starts with few bands and rise the number of bands until convergence.
3) Most probably you can lower the FFT G vectors, this will save memory, but being careful to not lower too much otherwise you will loose accuracy. Not having seen your report file I cannot advise on how much to lower as I do not know the total number of Gvectors.

The number of core are decided by the input and submission job script and they should be compatible. In your input you are asking for 8 cpus (4x2). For the job submission script you can ask to your system administrator.

Best,
Daniele

[joydev de]

Dear Daniele Varsano,

I am calculating the absorption spectra of h-BN(5*5 unit cell) using alda kernel in g-space.
But the job always exit. What is the resign, I don't understand. Please help me.
I am attaching nscf.out, one log file and output error message.

Thanking you.

Sincerely
Joydev De

[Daniele Varsano]

Dear Joydev,

please some recommendation:
1.) Fil your signature with your affiliation. This is a rule of the forum and you can do once for all by filling the signature in your user preferences.
2.) Please post your question in the appropriate subforum. This is not anymore a compilation problem, but a problem with a linear response calculation.
3.) When posting a problem: please include also the report file, from the file you posted I cannot see how many k points are you using, electronic structure etc...

Coming to your question: it looks a memory issue, possible workarounds are:
1) Run with 8cpus instead of 16, in this way each cpu will have the double of RAM available
3) You can try to lower the FFT G vectors, by specifying FFTGvecs (I cannot advise on how to set it as I need to look at the report file).
4) Run the ALDA in transition space (yambo -o b -k ALDA -y d/h)

Best,

Daniele

[joydev de]

Dear Daniele Varsano,

Can I do calculate spin polarised absorption spectra using yambo?

Thanking you.

Regards,
Joydev De
NISER, India

[Daniele Varsano]

Dear Joydev,

yes it is possible, yambo can handle spin-resolved calculations both in the collinear and non collinear way (spinors). The latter is needed when spin orbit coupling terms are taken into account.

Please try to post the questions in the appropriate subforum, this topic has nothing to do with yambo installation. Note also that there is a dedicated subforum on spin dependent and spinorial calculations.

Best,

Daniele