Parallel calculation on GW correction

You can find here problems arising when using old releases of Yambo (< 5.0). Issues as parallelization strategy, performance issues and other technical aspects.

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

Locked
el16yz
Posts: 23
Joined: Mon Sep 23, 2019 4:34 pm
Location: United Kingdom

Parallel calculation on GW correction

Post by el16yz » Fri Nov 08, 2019 1:10 pm

Dear Yambo Users,

I tried to perform GW correction with parallel calculation. It is fine when NGsBlkXp is below 3 Ry. But when I set this parameter as 3 or larger, the error occurs:
-------------------------------------------------------
Primary job terminated normally, but 1 process returned
a non-zero exit code.. Per user-direction, the job has been aborted.
-------------------------------------------------------
--------------------------------------------------------------------------
mpirun detected that one or more processes exited with non-zero status, thus causing
the job to be terminated. The first process to do so was:

Process name: [[39358,1],26]
Exit code: 104
--------------------------------------------------------------------------


I use Yambo 4.4.0 and 64 processes and 24G per slot for the calculation as my system is a little bit large.

I attach my input file and responding l-* and r-* file, hope they will do some help. It will be very grateful if you can solve this problem for me. Thank you.

Best,
Yang
You do not have the required permissions to view the files attached to this post.
Yang Zhou
PhD student
University of Leeds
United Kingdom

User avatar
Daniele Varsano
Posts: 4209
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: Parallel calculation on GW correction

Post by Daniele Varsano » Fri Nov 08, 2019 1:20 pm

Dear Yang,
it seems the code has not been properly installed to run in parallel.
When running in parallel you should have just one report file indicating the number of used CPU.
Her you have many reports indicating:

Code: Select all

 * CPU             : 1
 * THREADS    (max): 1
which means you are running several serial jobs.

You can check in config/report what is indicated regarding the parallel support:

Code: Select all

 - PARALLEL SUPPORT -
#
# [X] MPI
# [X] OpenMP
in case the MPI support is not marked you should inspect what went wrong in the config.log file.
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/

el16yz
Posts: 23
Joined: Mon Sep 23, 2019 4:34 pm
Location: United Kingdom

Re: Parallel calculation on GW correction

Post by el16yz » Fri Nov 08, 2019 3:27 pm

Dear Daniele,

The parallel setting in my config/report file is regarded as:
# - PARALLEL SUPPORT -
#
# [X] MPI
# [-] OpenMP
Is that mean MPI is available but not OpenMP? I also check config.log file. It seems that there is something wrong in configuring. But I don't know how to fix them. I attach these two files so that the error could be checked in detail. Thank you.
You do not have the required permissions to view the files attached to this post.
Yang Zhou
PhD student
University of Leeds
United Kingdom

User avatar
Daniele Varsano
Posts: 4209
Joined: Tue Mar 17, 2009 2:23 pm
Contact:

Re: Parallel calculation on GW correction

Post by Daniele Varsano » Fri Nov 08, 2019 3:43 pm

Dear Yang,

MPIFC='mpiifort'

MPI parallelization is enough to run Yambo in parallel, OMP is not mandatory, unless you need it.
From your config.log file I can see that parallel compiler is not assigned, check you have mpiifort and mpicc in your path and you can try to assign when configuring Yambo:
./configure MPIFC='mpiifort', MPICC='mpiicc'


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/

Locked