application called MPI_Abort

Concerns issues with computing quasiparticle corrections to the DFT eigenvalues - i.e., the self-energy within the GW approximation (-g n), or considering the Hartree-Fock exchange only (-x)

Moderators: Davide Sangalli, andrea.ferretti, myrta gruning, andrea marini, Daniele Varsano

Post Reply
xjxiao
Posts: 47
Joined: Sat Jan 11, 2025 6:30 pm

application called MPI_Abort

Post by xjxiao » Mon Mar 10, 2025 2:10 pm

Dear all,
What is causing this error, and how can I resolve it? According to the backend logs, there is no memory shortage.

Code: Select all

application called MPI_Abort(MPI_COMM_WORLD, 1) - process 61
l-HF_and_locXC_gw0_dyson_em1d_ppa_CPU_61.txt
Thanks!
Xiao
You do not have the required permissions to view the files attached to this post.
Last edited by xjxiao on Mon Mar 10, 2025 4:29 pm, edited 1 time in total.
Xiangjun Xiao
Institute of Semiconductors, Chinese Academy of Sciences

User avatar
Daniele Varsano
Posts: 4198
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: application called MPI_Abort

Post by Daniele Varsano » Mon Mar 10, 2025 2:57 pm

Dear Xiao,

in order to inspect the problem, you should provide more information as the input and report files.

Best,
Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/

xjxiao
Posts: 47
Joined: Sat Jan 11, 2025 6:30 pm

Re: application called MPI_Abort

Post by xjxiao » Mon Mar 10, 2025 4:19 pm

Daniele Varsano wrote: Mon Mar 10, 2025 2:57 pm Dear Xiao,

in order to inspect the problem, you should provide more information as the input and report files.

Best,
Daniele
Dear Daniele,
I have attached the files. Could you please help me? Thank you!
Yours,
Xiao
Last edited by xjxiao on Mon Mar 10, 2025 4:35 pm, edited 1 time in total.
Xiangjun Xiao
Institute of Semiconductors, Chinese Academy of Sciences

User avatar
Daniele Varsano
Posts: 4198
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: application called MPI_Abort

Post by Daniele Varsano » Mon Mar 10, 2025 4:28 pm

Dear Xiangjun,

I have the impression you have a memory problem. A cutoff of 30Ry in the response function can be rather large.
My suggestion is first to verify if you really need such a large cutoff in the dielectric matrix.
Then you can try to optimize the memory distribution using:

Code: Select all

X_and_IO_CPU= "1 1 16 4"                 # [PARALLEL] CPUs for each role
X_and_IO_ROLEs= "q k c v"     
or try:

Code: Select all

X_and_IO_CPU= "1 4 1 8 2"                 # [PARALLEL] CPUs for each role
X_and_IO_ROLEs= "q g k c v"     
Best,

Daniele
Dr. Daniele Varsano
S3-CNR Institute of Nanoscience and MaX Center, Italy
MaX - Materials design at the Exascale
http://www.nano.cnr.it
http://www.max-centre.eu/

Post Reply