Error in run HF calculation

[el16yz]

Dear Yambo users,

I encounter a problem when I performed the runlevel of "yambo -x" for HF calculation. The error message:

[ERROR] STOP signal received while in :[04] Bare local and non-local Exchange-Correlation
[ERROR]Allocation of WF%c failed
application called MPI_Abort(MPI_COMM_WORLD, 1) - process 0
[unset]: write_line error; fd=-1 buf=:cmd=abort exitcode=1
:
system msg for write_line failure : Bad file descriptor

The Yambo I use is 4.4.0. And the initial information on my system:

[RD./SAVE//ns.db1]------------------------------------------
Bands : 320
K-points : 1
G-vectors [RL space]: 8977061
Components [wavefunctions]: 1727701
Symmetries [spatial+T-rev]: 2
Spinor components : 1
Spin polarizations : 2
Temperature [ev]: 0.02585
Electrons : 512.0000
WF G-vectors : 1727701
Max atoms/species : 96
No. of atom species : 5
Exact exchange fraction in XC : 0.000000
Exact exchange screening in XC : 0.000000
Magnetic symmetries : no
- S/N 001920 -------------------------- v.04.04.00 r.00148 -
[RD./SAVE//ns.wf]-------------------------------------------
Fragmentation :yes
Bands in each block : 320
Blocks : 1
- S/N 001920 -------------------------- v.04.04.00 r.00000 -
[RD./SAVE//ns.kb_pp_pwscf]----------------------------------
Fragmentation :yes
- S/N 001920 -------------------------- v.04.04.00 r.00000 -

If will be very appreciated if you can give me some clues to solve my problem. Thank you.

Best wishes
Yang

[Daniele Varsano]

Dear Yang,
it seems a memory issue, can you please post your HF input and complete report file?
You can upload in a post by renaming with a .txt suffix.
Best,
Daniele

[el16yz]

Dear Daniele,

Sure. Thanks for your reply.

[Daniele Varsano]

Dear Yang Zhou,
your input file seems to be correct. Anyway, it seems the code runs out-of-memory.
I can see you are running in serial, do you know how many RAM memory do you have at your disposal?

My suggestions are:
1) Try to reduce your convergence parameters, just to see if indeed is a memory issue.
2) In order to see if it is indeed a memory issue, you can compile the code adding the line --enable-memory-profile in the configure. Hopefully, this will tell you how much memory you need for this calculation.
3) You can try to run in parallel, using a parallelization strategy over the bands:

Code: Select all

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

4) Your cell is very large, you are dealing with an isolated system, note that you can use the modified coulomb potentials to isolate the system and maybe this will allow you to reduce your simulation box.

Best,
Daniele

[el16yz]

Dear Daniele,

Thanks for your solution. The problem is solved after I try to use parallel calculation following your suggest. Thank you very much.

Best,
Yang