[petsc-users] Fwd: Error in HDF5 dumps of DMPlex labels

Smith, Barry F. bsmith at mcs.anl.gov
Tue Apr 2 16:58:50 CDT 2019


   Fabian,

     I apologize, your email never got properly processed by someone when it arrived. 

     Thank you for the bug report and in particular for providing a patch. I have submitted a pull request with your patch https://bitbucket.org/petsc/petsc/pull-requests/1494/isview_general_hdf5-crashes-with-a-zero/diff and listed Matt Knepley as a reviewer. The patch looks good to me but I don't understand the exactly subtleties of HDF chunk sizes. Hopefully the patch is acceptable to Matt and he will review it, otherwise he will have to propose an alternative fix. 

     After approval and compile testing it will be put in the maint and master branches and be in the 3.11.1 patch release of PETSc.

   Barry


> On Apr 2, 2019, at 3:49 PM, Fabian.Jakub via petsc-users <petsc-users at mcs.anl.gov> wrote:
> 
> Hi there,
> 
> I'm not sure if this message got lost (have sent it to petsc-dev, trying
> users list now) or if I just did not get the replies.
> Anyway, the issue that is attached below is unresolved in the current
> master branch and I am curious about your input.
> 
> Many Thanks,
> 
> Fabian
> 
> 
> -------- Weitergeleitete Nachricht --------
> Betreff: [petsc-dev] Error in HDF5 dumps of DMPlex labels
> Datum: Thu, 14 Feb 2019 14:53:22 +0100
> Von: Fabian.Jakub via petsc-dev <petsc-dev at mcs.anl.gov>
> Antwort an: Fabian.Jakub <Fabian.Jakub at physik.uni-muenchen.de>
> An: petsc-dev <petsc-dev at mcs.anl.gov>
> 
> Dear Petsc Team!
> 
> I had an issue when writing out DMPlex objects through hdf5.
> 
> 
> This comes from a DMLabel that has only entries on non-local mesh points.
> The DMLabel write only includes local parts of the label and so leads to
> a zero sized write for the index set.
> This seems to be fine except that the hdf5 chunksize is set to zero
> which is not allowed.
> 
> I added a minimal example to illustrate the error.
> It creates a 2D DMPlex in serial, distributes it, labels the nonlocal
> points in the mesh and dumps it via PetscObjectViewer to HDF5.
> Run with:
> 
>   make plex.h5
> 
> I also attached a quick fix to override the chunksize.
> 
> Please let me know if you anything extra and also if this is expected
> behavior... I could certainly with the fact that DMLabel is not supposed
> to work this way.
> 
> Many thanks,
> 
> Fabian
> 
> 
> <0001-fix-hdf5-chunksizes-of-0.patch><Makefile.txt><error_msg.txt><hdf5_issue_nonlocal_only_labels.F90>



More information about the petsc-users mailing list