STOP signal received:[09] Dynamic Dielectric Matrix (PPA)

[feffeficus]

Hi all dear Yamboers!

while doing my cute G0W0 calculations, the code crash in several cases (different calculation with different number of Gblk vectors and bands) with this error message in the l_file

Code: Select all

 <18h-05m-46s> [M 1.411 Gb] Free X_WS (0.232)
[ERROR] STOP signal received while in :[09] Dynamic Dielectric Matrix (PPA)
[ERROR][NetCDF] NetCDF: One or more variable sizes violate format constraints

more info:
running on sp6 cineca, mem 10Gb on 8 core (ST mode)
yambo rev_552 ( I know very old but since now was working...)

I attach the l_report and r_report, the l_dbs, and the job + input I'm using.

Could it be only a simple memory error? since I'm trying to check the convergence of Gblk vector?
I have the same error also when fixed all the variables (bands and G vector to a resonable running and converged set) I calculate the QP correction for 17 kpt of my band structure. it crashes in the same way..
What do you think?

With the best
feffe

[Daniele Varsano]

Dear Feffe,
the reason can be ascribed on the fact that necdf can fail when handling very big
databases (5/6 Gb). Is it your case? But I'm not an expert and may be someone can
give you more details.

Cheers,

Daniele

[myrta gruning]

Hallo Feffe and Daniele,

Normal netcdf (without enable large files option in the configure) can read up to 2 Gb. In that case the solution is either recompiling with that option, or starting the calculation for the dielectric matrix imposing fragmentation (-S option in the command line). In general the latter is always a good/safe option if you are doing large calculations.

Ciao
m

[feffeficus]

ciao belli!

So, I put -S when I construct the database at the initialization of the yambo runs.
From Myrta's post I understand I have to put the -S also in the other step?

Thanks so much for the support!

[feffeficus]

no it doesnt' work and I cant' go on with the calculation..

other ideas?

[feffeficus]

PROBLEM FIXED

The database ndb.pp that yambo was creating during the plasmon pole calculation was larger than 2Gb.

With Daniele support, I've launch yambo -S (with fragmentation option) for the plasmon pole calculation.
It works. In the end I obtain a ndb.pp database fragmented in n-part (less than 2Gb) as the number of q-point I have in my calculation.

So, I can say that this error I got was due to database larger than 2 Gb.
Problem fixed for the moment.
Thanks to all.

[plasmon]

Does Yambo 4.5 still support -S option? I also met with the problem of segmentation fault due to the large database. Thank you!

Hao

PROBLEM FIXED

The database ndb.pp that yambo was creating during the plasmon pole calculation was larger than 2Gb.

With Daniele support, I've launch yambo -S (with fragmentation option) for the plasmon pole calculation.
It works. In the end I obtain a ndb.pp database fragmented in n-part (less than 2Gb) as the number of q-point I have in my calculation.

So, I can say that this error I got was due to database larger than 2 Gb.
Problem fixed for the moment.
Thanks to all.

[Daniele Varsano]

Dear Hao,
fragmentation is now the default,
can you post the error message you encounter together with input/report file and compilation option (config.log)?

Best,
Daniele

[plasmon]

Dear Hao,
fragmentation is now the default,
can you post the error message you encounter together with input/report file and compilation option (config.log)?

Best,
Daniele

Dear Daniele,

Thank you! We encountered error with the message of 'Dynamical Dielectric Matrix *** Process received signal ***'.
I have attached the necessary files here. Could you please tell me how to fix it?

Thank you!
Hao

[Daniele Varsano]

Dear Hao,
actually I'm not sure it is a problem of the size of the database.
From the report file, it seems you are running on top of a previous run.

Code: Select all

[RD./SAVE//ndb.em1d_fragment_2]-----------------------------
    :: Current Q-pt index     : 2
   *WRN* :: X energy range     [ev]:0.1013E-1  13.94
    :: X damping range    [ev]:  0.10000   0.10000
   *ERR* :: Number of frequencies  :  477

Can you repeat the calculation starting from a clean SAVE, i.e. removing the previous ndb.em1d* files?
Next post the error message. If possible, compile the code adding the
--enable-memory-profile
option in the configure in order to monitor the memory usage, as it could be a problem of memory allocation.
Other suggestion is to assign a parallelization in the evaluation fo the response function as:

Code: Select all

X_CPU= "1 1 1 $ncpu 1 "                      # [PARALLEL] CPUs for each role
X_ROLEs= "q g k c v"                    # [PARALLEL] CPUs roles (q,g,k,c,v)

This will allow the distribution of the memory.

Best,
Daniele