GW calculation stoping with Dipoles: P, V and iR (T) |###########| [100%]

[Luo Yan]

Dear all
I did a GW calculation, but it stopped as “ Dipoles: P, V and iR (T) |########################################| [100%]" in LOG and "WARNING] [x,Vnl] slows the Dipoles computation. To neglect it rename the ns.kb_pp file [WF-Oscillators/G space/Transverse up loader] Normalization (few states) min/max : 0.52468E-09 1.0000" in report file.

[Daniele Varsano]

Dear Luo,

dipoles database seems that has been written:

Code: Select all

[WR./SAVE//ndb.dipoles]

It is possible that the job stops because of lack of memory.
You can try to assign the parallelisation strategy in input in order to distribute memory among CPUs using 48CPUs:

X_and_IO_CPU= "1 1 1 12 4" # [PARALLEL] CPUs for each role
X_and_IO_ROLEs= "q g k c v"

and also:

SE_CPU= "1 2 24" # [PARALLEL] CPUs for each role
SE_ROLEs= "q qp b" # [PARALLEL] CPUs roles (q,qp,b)

Best,

Daniele

[Luo Yan]

Dear Daniele
when added
SE_CPU= "1 2 24" # [PARALLEL] CPUs for each role
SE_ROLEs= "q qp b" # [PARALLEL] CPUs roles (q,qp,b)

some problems appear
"[ERROR] STOP signal received while in[07] Dynamic Dielectric Matrix (PPA)
[ERROR] Reading File ./SAVE//ndb.dipoles; Variable DIP_iR; NetCDF: Variable not found"

when added
X_and_IO_CPU= "1 1 1 12 4" # [PARALLEL] CPUs for each role
X_and_IO_ROLEs= "q g k c v"
There are no files generated, except for blank LOG.

My submiting pbs file attached

[Daniele Varsano]

Dear Luo Yan,
try to delete the ndb.dipoles file and relaunch the job.

You can add the parallelisation strategy together, they do act in different part of the calculation, screening and self-energy evaluation.
You can also add a parallelisation strategy for the dipoles:

DIP_CPU= "1 12 4" # [PARALLEL] CPUs for each role
DIP_ROLEs= "k c v" # [PARALLEL] CPUs roles (k,c,v)



Best,

Daniele

[Luo Yan]

Dr. Daniele
Thanks. I added
SE_CPU= "1 2 24" # [PARALLEL] CPUs for each role
SE_ROLEs= "q qp b" # [PARALLEL] CPUs roles (q,qp,b)
X_and_IO_CPU= "1 1 1 12 4" # [PARALLEL] CPUs for each role
X_and_IO_ROLEs= "q g k c v"
DIP_CPU= "1 12 4" # [PARALLEL] CPUs for each role
DIP_ROLEs= "k c v" # [PARALLEL] CPUs roles (k,c,v)
but task stopped and files ungegerated

The input is attached

[Luo Yan]

Dr. Daniele
This problem happened in the past time, but I don't kown how to solve. Sometimes, this task has been normal again after long time without any solutions.

[Daniele Varsano]

Dear Luo,

This problem happened in the past time, but I don't kown how to solve. Sometimes, this task has been normal again after long time without any solutions.

I'm sorry, I do not understand what do you mean.

Daniele

[Luo Yan]

Dear Daniele，
Sorry, I haven't solved this problem when I added these parameters together. Just generating blank LOG without informations.

DIP_CPU= "1 12 4" # [PARALLEL] CPUs for each role
DIP_ROLEs= "k c v" # [PARALLEL] CPUs roles (k,c,v)
X_and_IO_CPU= "1 1 1 12 4" # [PARALLEL] CPUs for each role
X_and_IO_ROLEs= "q g k c v"
SE_CPU= "1 2 24" # [PARALLEL] CPUs for each role
SE_ROLEs= "q qp b" # [PARALLEL] CPUs roles (q,qp,b)

Maybe some mistakes happen in my input, so that input and pbs files are attached. Thanks for your attentions.

[Daniele Varsano]

Dear Luo,

it is possible the mpirun option you are considering are not compatible with the requested resources:
it seems you are assign for 192 CPU

Code: Select all

#PBS -l nodes=4:ppn=48

then you are asking for 8 threads:

Code: Select all

export OMP_NUM_THREADS=8

next in the mpirun command I can see:

Code: Select all

 mpirun -np 48 -ppn 6

Actually I do not know the -ppn option in the mpirun command.
Note that the product of the roles in the input file should be equal to the number of the cpus (48 in your input).

Check the consistency of your submission job, maybe trying to set a simple configuration.

Best,
Daniele

[Luo Yan]

Dear Daniele
Sorry, here, in our mpirun option, ppn*nodes=np, ppn*OMP_NUM_THREADS=one nodes CPU (48 CPU). In this calculation, I used 4 nodes (192 CPU). Following your advises,
I have done three tests.
One:
"DIP_CPU= "4 12 4" # [PARALLEL] CPUs for each role; DIP_ROLEs= "k c v" # [PARALLEL] CPUs roles (k,c,v)" ,

Two:
"SE_CPU= "4 2 24" # [PARALLEL] CPUs for each role; SE_ROLEs= "q qp b" # [PARALLEL] CPUs roles (q,qp,b)"

Three:
"DIP_CPU= "4 12 4" # [PARALLEL] CPUs for each role; DIP_ROLEs= "k c v" # [PARALLEL] CPUs roles (k,c,v), SE_CPU= "4 2 24" # [PARALLEL] CPUs for each role; SE_ROLEs= "q qp b" # [PARALLEL] CPUs roles (q,qp,b)"

But the same as task informations stopped at yesterday. "<02m-49s> P1-cu06: Dipoles: P, V and iR (T) |########################################| [100%] " in LOG and "[WARNING] [x,Vnl] slows the Dipoles computation. To neglect it rename the ns.kb_pp file [WF-Oscillators/G space/Transverse up loader] Normalization (few states) min/max : 0.000000 1.000000" in report file.
As your advises at yesterday, they do act in different part of the calculation, but they don't work in my calculations.

Actually I did not try "X_and_IO_CPU= "1 1 1 12 4" # [PARALLEL] CPUs for each role: X_and_IO_ROLEs= "q g k c v"", because I don't understand "1 1 1" means.

Thanks for your further advises. Some files were also attached.