[Wien] WARNING elpa_setup during lapw1_mpi

[Gavin Abo]

That might be an issue worth asking about over in the Spack Community [1].


Not sure if I'm interpreting the message correctly, but if I am, the 
"@1.5.4:2" in the message from spack info I believe is indicating that 
thread_multiple will be disabled unless OpenMPI version 1.5.4 [2] to 2 
[3] is being used.


In the Spack source code of its current latest version 0.20.0 [4], the 
openmpi/package.py file has on lines 662-668:


          # thread_multiple
             if version in spack.version.ver("1.5.4:2"):
                 match = re.search(r"MPI_THREAD_MULTIPLE: (\S+?),?", output)
                 if match and is_enabled(match.group(1)):
                     variants.append("+thread_multiple")
                 else:
                     variants.append("~thread_multiple")


Since you are using OpenMPI 4.1.5, I suspect line 663 might need changed 
for example to "if version in spack.version.ver("1.5.4:4.1.5"):".  
Perhaps there are other changes a Spark developer would know about that 
might also be needed for the MPI_THREAD_MULTIPLE to work with Spack when 
using OpenMPI 4.1.5.


[1] https://github.com/spack/spack/blob/develop/README.md#community

[2] https://www.open-mpi.org/software/ompi/v1.5/

[3] https://www.open-mpi.org/software/ompi/v2.0/

[4] 
https://github.com/spack/spack/blob/v0.20.0/var/spack/repos/builtin/packages/openmpi/package.py


Gavin

WIEN2k user


On 7/4/2023 11:59 AM, Ilias, Miroslav wrote:
>
> Hello,
>
>
> sorry for hidden link;
>
>
> the point is that  the "|ompi_info -a | grep THREAD|" says 
> |MPI_THREAD_MULTIPLE: yes|,
>
>
> but  "spack info openmpi at 4.1.5" gives "thread_multiple [off] 
> [@1.5.4:2] on, off Enable MPI_THREAD_MULTIPLE support"
>
>
> Maybe this is the case of the ELPA  "WARNING elpa_setup: MPI threading 
> level MPI_THREAD_SERALIZED or MPI_THREAD_MULTIPLE required but your 
> implementation does not support this! The number of OpenMP threads 
> within ELPA will be limited to 1"
>
>
> I will investigate it deeper, as I need Wien2k parallel as fast as 
> possible.
>
> Intel MPI will come later, again, I need ELPA built with Intel. I 
> guess that IntelMPI+MKL  Wien2k is faster than OpenMPI+Openblas ?
>
> Best, Miro
>
> ------------------------------------------------------------------------
> *From:* Wien <wien-bounces at zeus.theochem.tuwien.ac.at> on behalf of 
> Laurence Marks <laurence.marks at gmail.com>
> *Sent:* 04 July 2023 18:02
> *To:* A Mailing list for WIEN2k users
> *Subject:* Re: [Wien] WARNING elpa_setup during lapw1_mpi
> That is a private site, do I cannot read anything.
>
> All I can suggest is doing a Google search on "missing  
> MPI_THREAD_MULTIPLE". It looks as if you have to enable this in 
> openmpi configure, although there might be some bugs. There could also 
> be some environmental parameters that need to be set.
>
> Or use Intel mpi. While it does not have everything that some sysadmin 
> want, if you are using mkl & ifort (both free) it integrates 
> trivially. I personally dislike openmpi as it keeps changing and is 
> not always user friendly.
>
> --
> Professor Laurence Marks (Laurie)
> Department of Materials Science and Engineering, Northwestern University
> www.numis.northwestern.edu <http://www.numis.northwestern.edu>
> "Research is to see what everybody else has seen, and to think what 
> nobody else has thought" Albert Szent-Györgyi
>
> On Tue, Jul 4, 2023, 09:33 Ilias, Miroslav <M.Ilias at gsi.de> wrote:
>
>     Dear Professor Marks,
>
>
>     concerning
>     https://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/msg22621.html  :
>
>
>
>     I am trying to find out why is ELPA module complaining of missing
>     MPI_THREAD_MULTIPLE .
>
>
>     We have a debate on this
>     https://git.gsi.de/SDEGroup/SIR/-/issues/85#note_55392
>
>
>     If you somebody could comment on this topic as I would like to
>     investigate deeply. Can somebody advice me a short ELPA testing
>     program for this ?
>
>
>     Best, Miro
>
>
>
>
>
>
>
>
>     _______________________________________________
>     Wien mailing list
>     Wien at zeus.theochem.tuwien.ac.at
>     http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>     SEARCH the MAILING-LIST at:
>     http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
>
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> SEARCH the MAILING-LIST at:http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20230704/3b6536bf/attachment.htm>