MPI error with tutorial on static non-linear propertie

option, parallelism,...

Moderators: fgoudreault, mcote

Forum rules
Please have a look at ~abinit/doc/config/build-config.ac in the source package for detailed and up-to-date information about the configuration of Abinit 8 builds.
For a video explanation on how to build Abinit 7.x for Linux, please go to: http://www.youtube.com/watch?v=DppLQ-KQA68.
IMPORTANT: when an answer solves your problem, please check the little green V-like button on its upper-right corner to accept it.
Locked
henry924
Posts: 6
Joined: Mon Sep 24, 2012 5:24 am

MPI error with tutorial on static non-linear propertie

Post by henry924 » Sun Oct 14, 2012 5:11 am

I followed abinit tutorial on static non-linear properties.
If i run abinit with mpirun at step 1 with tnlo_2.in, then the process crashed at dataset 5 calculation. This is the message

*** An error occurred in MPI_Recv
*** on communicator MPI_COMM_WORLD
*** MPI_ERR_TRUNCATE: message truncated
*** MPI_ERRORS_ARE_FATAL: your MPI job will now abort
--------------------------------------------------------------------------
mpirun has exited due to process rank 2 with PID 2821 on node HP-Ubuntu exiting improperly.


Could anyone please confirm that you did try the tutorial with parallelism and successfully get the result? And also please tell me how you did that?

Some other questions that may related:
Does parallelism only apply to certain process? (if i run mpi with one processor there's no error)
Why dataset 5 got problem with parallelism?

Here is my environment:
OS: CentOS 6.3
abinit: 6.12.3
openmpi: 1.6.1

User avatar
Alain_Jacques
Posts: 279
Joined: Sat Aug 15, 2009 9:34 pm
Location: Université catholique de Louvain - Belgium

Re: MPI error with tutorial on static non-linear propertie

Post by Alain_Jacques » Sun Oct 14, 2012 9:05 pm

Hi Henry,

question 1: it's expected that the dataset 5 will crash MPI ... with version 6 of Abinit. This input is supposed to be executed sequentially as you noticed.
question 2: yes of course. It's therefore not advisable to "pack" several consecutive computations in the same input file when they have different parallel resources requirements. At best you waste hardware resources and sometimes it plainly crashes.
question 3: I assume here that the number of k points gets modified during the response calculation and this causes havoc in the MPI processes - anyone else with a more convincing explanation?

Kind regards,

Alain

henry924
Posts: 6
Joined: Mon Sep 24, 2012 5:24 am

Re: MPI error with tutorial on static non-linear propertie

Post by henry924 » Mon Oct 15, 2012 4:22 am

That's the expert explanation I am looking for.
Alain, Thank you very much for your quick reply.

Locked