error in running yambo when I generate SAVE folder in one version and using it via different versions

[samare]

Dear Yambo Developers,

Thanks for your great works on developing this code and sharing it with us.

I am a new user of Yambo and would like to use a precompiled version of Yambo on Marcon.
I copied my SAVE folder from another computer to Marconi and get this error :

File ./SAVE//ndb.gops; Variable GROT; NetCDF: Start+count exceeds dimension bound

I compiled the last version of Yambo on my computer (Yambo.4.5.1). on Marconi (Yambo.4.2.1) is compiled.
Is this error caused by using different versions for generating SAVE folder and using it?
of course, I am not sure if in my script I call Yambo code correctly or not.

I was wondering if you could send me a script for running Yambo code in parallel on Marconi.

Thanks in advance for your help.

Best regards
Samare

[Daniele Varsano]

Dear Samare,

Yambo-4.2.1 is a very old version, not supported anymore.
In which Marconi partition are you running? It is the skylake (A3)?

The database format changed from the old releases.
You can delete the ndb.gops and ndb.qindx and rerun the setup.

My suggestion, anyway, is to compile Yambo 4.5.1 in Marconi.
I will also ask the Cineca to update the Yambo packages.

PS:
Please post your question in the right subforum, this is not related to yambo post-processing.
Please add your name and affiliation, this is a rule of the forum. You can do once for all by filling the signature in your user profile.

[samare]

Dear Daniele,

Thanks a lot for your reply.
I am using Skylake nodes (skl_usr_pro) in Marconi.
I will try to compile the newest version in my account on Marconi.

Bests,

[Daniele Varsano]

Dear Samare,
yes, I think this is the best option. Anyway I have already asked to update the compiled version of skylake to the latest release, so it should be there in the next days.

Best,
Daniele

[Nicola Spallanzani]

Dear Samare,
I interacted with Cineca's User Support in order to install Yambo v4.5 on Marconi-SKL.
Now you can find it in profile/phys.

Best,
Nicola

[samare]

Dear Developers,

Thanks a lot for your help in compiling the new version on Marconi.

I am working with Yambo v5.4 on Marconi, but I think there is a problem with it.
I tried a simple gw calculation "yambo -x -p p -g n -F gw_ppa.in" and noticed that
after printing ndb.pp, it will not print ndb.pp_fragment files or other outputs.
my input works well using Yambo v4.2.1 on Marconi and Yambo v4.5 on my computer.
I exactly do not know what the problem is and I was wondering if you could help me to fix this issue.
(of course, if the final results of the old and new versions are the same I could use the old version.)

kind regards,

[Daniele Varsano]

Dear Samare,
in order to spot the problem, can you upload your input/report and log files?
IN order to upload them maybe you have to change the suffix of the files (e.g. .txt, or zip them).
Best,
Daniele

[samare]

Dear Daniele,

please find attached my input and output files.
After printing ndb.pp, nothing happens and job will be canceled due to the time limit.
(I prepared this file just for testing )

Bests,
Samare

[Daniele Varsano]

Dear Samare,

looking at your log and reports file, it seems that the code has some difficulties when inverting the X matrix, which is related to the parallel linear algebra.
In the present case actually you are dealing with very small matrix:

Code: Select all

 X matrix size          : 217

so it is not useful to diagonalize them using scalapack.

Can you try to set in your input:

Code: Select all

X_nCPU_LinAlg_INV= 1

Next, even if the code did not arrive at that point, please note that you set:

Code: Select all

SE_CPU= " 1 10 1"       # [PARALLEL] CPUs for each role
SE_ROLEs= "q qp b"         # [PARALLEL] CPUs roles (q,qp,b)

which means that each CPU calculate a GW correction. This number cannot be larger than the number of the total corrections which in your case it is 2:

Code: Select all

%QPkrange                      # # [GW] QP generalized Kpoint/Band indices
1|1|48|49|
%

Best,
Daniele

[samare]

Dear Daniele,

So many thanks for your help.
It works well now and I can get the results.

Bests,
Samare