[mvapich-discuss] Buffer limitations

[Dhabaleswar Panda]

Hi,

Thanks for reporting this issue. The version you are using (MVAPICH2
1.4.0)  is very old (released during Oct 2009). Unfortunately, we do not
support this older version any longer. May I request you to try your
application with the latest MVAPICH2 versions (1.6 or 1.7a). These later
versions have many additional features for performance and scalability.
Many bugfixes have also been carried out during the last one and half
year.  If you still see issues with 1.6 or 1.7a version, we will be happy
to take a deeper look into this problem.

Thanks,

DK

On Thu, 2 Jun 2011, Mike Page wrote:

> Hi,
>
> I am trying to get some code to run but am encountering problems using mpi_send/recv to transfer a large amount of data  (just over 40M elements of an MPI_DOUBLE_PRECISON array). The code just hangs when I reach this data transfer.
>
> Am I possibly blowing out buffer limits? How can I be sure?
>
> > [mpage at fe5 source]$ mpiname -a
> > MVAPICH2 1.4.0 2010-01-21 ch3:mrail
> >
> > Compilation
> > CC: pgcc -g -DNDEBUG -O2
> > CXX: pgCC  -DNDEBUG -O2
> > F77: pgf77  -DNDEBUG
> > F90: pgf90  -DNDEBUG
> >
> > Configuration
> > CC=pgcc CXX=pgCC F77=pgf77 F90=pgf90 --with-rdma=gen2 --prefix=/opt/mvapich2/1.4.1-pgi --enable-romio=yes --with-file-system=lustre+panfs
> >
> > [mpage at fe5 source]$ uname -a
> > Linux fe5 2.6.18-194.11.4.el5-perfctr #1 SMP Wed Oct 27 18:35:31 GMT 2010 x86_64 x86_64 x86_64 GNU/Linux
> >
>
>
> Yhanks
>
> Mike Page
> National Center for Atmospheric Research
> Research Applications Lab (RAL)
> Joint Numerical Testbed (JNT)
>
> Ofc: 303-497-2464
> Cell: 303-944-8291
>
>
>
>
>
> _______________________________________________
> mvapich-discuss mailing list
> mvapich-discuss at cse.ohio-state.edu
> http://mail.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss
>