out of memory

Total energy, geometry optimization, DFT+U, spin....

Moderator: bguster

Locked
suman
Posts: 13
Joined: Fri Jan 24, 2020 5:25 pm

out of memory

Post by suman » Fri May 29, 2020 6:59 am

Hello abinit users,

I was trying to dope monolayer GeSe with Sulphur atoms, relax it. I have 100 atoms.
But, I got an error saying


--- !ERROR
src_file: m_gstate.F90
src_line: 716
mpi_rank: 28
message: |
out of memory in cg
...

I believe this error is due to the account of more number of atoms than ABINIT permits.

Could you please let me know how many atoms ABINIT allow for relax?

Thank you

ebousquet
Posts: 469
Joined: Tue Apr 19, 2011 11:13 am
Location: University of Liege, Belgium

Re: out of memory

Post by ebousquet » Thu Jun 04, 2020 3:57 pm

Dear Suman
Abinit has almost no limit on the number of atoms, 100 is clearly not a problem.
Could it be that your computer does not have enough RAM memory for this calculation?
Best wishes,
Eric

suman
Posts: 13
Joined: Fri Jan 24, 2020 5:25 pm

Re: out of memory

Post by suman » Sun Jun 07, 2020 6:11 pm

Dear Ebousquet,

Thank you for the reply. I have sent the job of relaxing 100 atoms for 10.5 hours for a given 512 cores in supercomputer and waiting for the result.

Cheers,
Suman

User avatar
gmatteo
Posts: 291
Joined: Sun Aug 16, 2009 5:40 pm

Re: out of memory

Post by gmatteo » Wed Jun 10, 2020 4:23 am

Well, it's difficult to suggest how to address the problem without looking at your input file.
I hope you are using paral_kgb = 1 with band/fft/kpoint parallelism because the default eigensolver (conjugate gradient, paral_kgb=0) can only distribute memory in k-space

suman
Posts: 13
Joined: Fri Jan 24, 2020 5:25 pm

Re: out of memory

Post by suman » Sat Jun 27, 2020 12:02 am

Thank you for the reply.

Sorry for not uploading the input file.

My previous run didn't work out.

I have attached input, log and submit file.

Could you please suggest me what's the issue in input file?

Thank you
Attachments
gese.in
(5.11 KiB) Downloaded 206 times
gese.log
(162.73 KiB) Downloaded 172 times
submit.in
file used to submit to supercomputer
(405 Bytes) Downloaded 178 times

ebousquet
Posts: 469
Joined: Tue Apr 19, 2011 11:13 am
Location: University of Liege, Belgium

Re: out of memory

Post by ebousquet » Mon Jul 06, 2020 2:31 pm

Dear Suman,
A general comment: you use a grid of k-point of 1x8x8 and your cell parameters a, b and c are 28 ,37 and 40 respectively.
This does not make sense becauase the k-point length is inversely proportional to the cell length such that having 1 kpoint for 28 Bohr and 8 for 37 and 40 is inconsistent in this respect. How did you chose these values of k-points?
regarding the memory, I would propose to test this instead of the automatic one (you can increase nband if you need more analysis of the electronic structure later):

Code: Select all

nband 512
npkpt 32
npband 16
npfft.  1
which is using only MPI or, using more MPI+OpenMP:

Code: Select all

nband 512
npkpt 32
npband 512
bandpp 16
Best wishes,
Eric

Locked