Fatal error in PMPI_Barrier

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

shaldar
Posts: 17
Joined: Mon Apr 13, 2015 1:33 pm

Fatal error in PMPI_Barrier

Post by shaldar » Tue Jun 21, 2016 11:17 pm

Dear All,

I am trying to do a gw calculations on MoTe2 unit cell. I am using the latest version of code from svn (Version 4.0.3 Revision 103). The GW calculation is crashing giving an error "Fatal error in PMPI_Barrier: Invalid communicator, error stack: " I have attached the necessary files in the zip file. The same input files run perfectly with the previous version "Version 4.0.2 Revision 96". Will it be possible to let me know where am I doing wrong with the new version ?

Best Regards
Soumyajyoti
You do not have the required permissions to view the files attached to this post.
Soumyajyoti Haldar
Postdoctoral Researcher
Institute of Theoretical Physics
University of Kiel, Germany

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

Re: Fatal error in PMPI_Barrier

Post by Daniele Varsano » Wed Jun 22, 2016 5:49 am

Dear Soumyajyoti,
thanks for reporting we will inspect what is going wrong and let you know.

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/

User avatar
Davide Sangalli
Posts: 641
Joined: Tue May 29, 2012 4:49 pm
Location: Via Salaria Km 29.3, CP 10, 00016, Monterotondo Stazione, Italy
Contact:

Re: Fatal error in PMPI_Barrier

Post by Davide Sangalli » Wed Jun 22, 2016 10:29 am

Dear Soumyajyoti,
thank you for reporting.

Not sure if this solves the issue, but by mistake I've uploaded a version which I was testing (103) instead of the correct one (104).
Could you please do "svn up" (i.e. rev 104) and let us know if the problem is still there ?

Kind regards,
D.
Davide Sangalli, PhD
CNR-ISM, Division of Ultrafast Processes in Materials (FLASHit) and MaX Centre
https://sites.google.com/view/davidesangalli
http://www.max-centre.eu/

shaldar
Posts: 17
Joined: Mon Apr 13, 2015 1:33 pm

Re: Fatal error in PMPI_Barrier

Post by shaldar » Wed Jun 22, 2016 12:17 pm

Dear Davide,

I am getting the same error with rev 104

Regards
Soumyajyoti
Soumyajyoti Haldar
Postdoctoral Researcher
Institute of Theoretical Physics
University of Kiel, Germany

User avatar
Davide Sangalli
Posts: 641
Joined: Tue May 29, 2012 4:49 pm
Location: Via Salaria Km 29.3, CP 10, 00016, Monterotondo Stazione, Italy
Contact:

Re: Fatal error in PMPI_Barrier

Post by Davide Sangalli » Wed Jun 22, 2016 12:33 pm

Ok.

Then I would need more info to try to spot the problem. The best thing would be to reproduce it on a smaller DB, running on fewer processor. Could you try something like a 6x6 grid, running on 8 processors ("2 2 2" in input), even if results would be non physical.

Thank you again for reporting the problem.

Best,
D.
Davide Sangalli, PhD
CNR-ISM, Division of Ultrafast Processes in Materials (FLASHit) and MaX Centre
https://sites.google.com/view/davidesangalli
http://www.max-centre.eu/

shaldar
Posts: 17
Joined: Mon Apr 13, 2015 1:33 pm

Re: Fatal error in PMPI_Barrier

Post by shaldar » Wed Jun 22, 2016 1:02 pm

Sure,

I'll do it and post it as soon as possible.

Regards

Soumyajyoti
Soumyajyoti Haldar
Postdoctoral Researcher
Institute of Theoretical Physics
University of Kiel, Germany

shaldar
Posts: 17
Joined: Mon Apr 13, 2015 1:33 pm

Re: Fatal error in PMPI_Barrier

Post by shaldar » Wed Jun 22, 2016 2:12 pm

Davide Sangalli wrote:Ok.

Could you try something like a 6x6 grid, running on 8 processors ("2 2 2" in input), even if results would be non physical.

Thank you again for reporting the problem.

Best,
D.
Unfortunately I am getting the same error. The files are attached

Regards

Soumyajyoti
You do not have the required permissions to view the files attached to this post.
Soumyajyoti Haldar
Postdoctoral Researcher
Institute of Theoretical Physics
University of Kiel, Germany

User avatar
Davide Sangalli
Posts: 641
Joined: Tue May 29, 2012 4:49 pm
Location: Via Salaria Km 29.3, CP 10, 00016, Monterotondo Stazione, Italy
Contact:

Re: Fatal error in PMPI_Barrier

Post by Davide Sangalli » Wed Jun 22, 2016 3:12 pm

It is a good news that you are able to reproduce the error on the smaller run.
Unfortunately I tried and I did not get your error.

A stupid question: did you try to do make clean_all before recompiling the code after the "svn up" ?

Meanwhile I will go on inspecting the problem.

Best,
D.
Davide Sangalli, PhD
CNR-ISM, Division of Ultrafast Processes in Materials (FLASHit) and MaX Centre
https://sites.google.com/view/davidesangalli
http://www.max-centre.eu/

shaldar
Posts: 17
Joined: Mon Apr 13, 2015 1:33 pm

Re: Fatal error in PMPI_Barrier

Post by shaldar » Thu Jun 23, 2016 12:46 pm

Dear Davide,

Yes I did make clean_all before configuring again.

I'll try to install in a different computer and will see if I can reproduce the error.

Regards
Soumyajyoti
Soumyajyoti Haldar
Postdoctoral Researcher
Institute of Theoretical Physics
University of Kiel, Germany

jinsoo
Posts: 1
Joined: Tue May 03, 2016 2:00 pm

Re: Fatal error in PMPI_Barrier

Post by jinsoo » Wed Jun 29, 2016 1:58 pm

I am having the same error as above using yambo 4.0.3 rev104,
"Fatal error in PMPI_Barrier: Invalid communicator, error stack:"

Previous version of yambo (4.0.2 rev 90)did not give this error.

I have noticed that the configure options of yambo 4.0.3rev104 lacks "--with-blacs-libs" and "--with-scalapack-libs".
Would this be relevant to the above error? Were these options deleted on purpose?
Jinsoo Park
Ph.D student
Department of Applied Physics and Materials Science
California Institute of Technology

Locked