[Pw_forum] MPIR_Get_contextid_sparse_group(1318): Too many communicators

[Ilya Ryabinkin]

Sorry, colleagues, seems like it is a MKL bug discussed quite recently:

http://qe-forge.org/pipermail/pw_forum/2016-May/109821.html

I apologize for the omission.

--
I.

On Thu, Sep 15, 2016 at 1:45 PM, Ilya Ryabinkin <igryabinkin at gmail.com> wrote:
> I'm  getting received these error messages
>
> **************
> Fatal error in PMPI_Comm_split: Other MPI error, error stack:
> PMPI_Comm_split(557).................: MPI_Comm_split(comm=0xc4000004,
> color=10, key=10, new_comm=0x7fffffffa788) failed
> PMPI_Comm_split(533).................:
> MPIR_Comm_split_impl(270)............:
> MPIR_Get_contextid_sparse_group(1318): Too many communicators
> (3141/16384 free on this process; ignore_id=0)
> **************
>
> I suspect it may be caused by the lack of
>
>  mp_barrier()
>
> call in my code, which is using calbec() routine to get overlap matrix
> between KS orbitals at t-dt and t steps.
>
> Could anyone comment on that?
>
> Thanks in advance,
> Ilya
>
> --
> *******************************************************
>                     Ilya Ryabinkin
>                  Postdoctoral Scholar
>           Physical and Environmental Sciences
>            University of Toronto Scarborough
>   http://www.utsc.utoronto.ca/~aizmaylov/Members.html
> *******************************************************



-- 
*******************************************************
                    Ilya Ryabinkin
                 Postdoctoral Scholar
          Physical and Environmental Sciences
           University of Toronto Scarborough
  http://www.utsc.utoronto.ca/~aizmaylov/Members.html
*******************************************************