Hi John,

On Thu, Apr 3, 2014 at 6:36 AM, Biddiscombe, John A. <[email protected]> wrote:
> Greetings
>
>
>
> I've been forwarded some info regarding a hang on hdf5 read from a user of
> enzo.
>
> The problem appears when 512 (or more) processes open a file in serial and
> each independently read a hyperslab (non overlapping, I believe, but this
> may not be correct).

Looking at the stack trace, this is indeed a set of non-overlapping
hyperslabs, as it occurs during what Enzo refers to as
ParallelRootGridIO.

>
> I asked for a stack trace and the attached screenshot was sent to me.
>
>
>
> My advice to them was to open the file in parallel using the mpio driver
> rather than in serial and the problem with probably go away.

I suspect this would probably require a rather large and somewhat
invasive change to the Enzo codebase, particularly as there are
incompatibilities with MPI-enabled HDF5 in other areas of Enzo.

-Matt

>
> But I'd like to ask if from the stacktrace you can tell  if this is a known
> bug?
>
> The problem occurs with 1.8.12 and 1.9.something
>
>
>
> thanks
>
>
>
> JB
>
>
>
> --
>
> John Biddiscombe,                        email:biddisco @.at.@ cscs.ch
>
> http://www.cscs.ch/
>
> CSCS, Swiss National Supercomputing Centre  | Tel:  +41 (91) 610.82.07
>
> Via Trevano 131, 6900 Lugano, Switzerland   | Fax:  +41 (91) 610.82.82
>
>
>
>
> _______________________________________________
> Hdf-forum is for HDF software users discussion.
> [email protected]
> http://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
>

_______________________________________________
Hdf-forum is for HDF software users discussion.
[email protected]
http://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org

Reply via email to