[petsc-users] Persistence of sparse matrices.

[Barry Smith]

   What does HDF5 buy you in this situation? It won't buy you any  
speed, disk space, or portability.

    Barry



On Jan 25, 2010, at 5:40 PM, Don Pellegrino wrote:

> On Mon, Jan 25, 2010 at 12:25 PM, Matthew Knepley  
> <knepley at gmail.com> wrote:
>> I think you want to Viewer technology.
>
> Thanks for pointing me at the PetscViewer functions.  I wrote a quick
> test case to write a sparse matrix to an HDF5 file
> [http://cluster.ischool.drexel.edu/~st96wym4/flumap/cgit/cgit.cgi/exp007/tree/src/assign/assign_blast_scores.c?id=899b82178d3bcc365b07447a5686635c5856c39e 
> ].
> Unfortunately I receive the following run-time error:
>
> [0]PETSC ERROR: --------------------- Error Message
> ------------------------------------
> [0]PETSC ERROR: No support for this operation for this object type!
> [0]PETSC ERROR: Viewer type hdf5 not supported by SeqAIJ matrices!
> [0]PETSC ERROR:
> ------------------------------------------------------------------------
> [0]PETSC ERROR: Petsc Release Version 3.0.0, Patch 10, Tue Nov 24
> 16:38:09 CST 2009
> [0]PETSC ERROR: See docs/changes/index.html for recent updates.
> [0]PETSC ERROR: See docs/faq.html for hints about trouble shooting.
> [0]PETSC ERROR: See docs/index.html for manual pages.
> [0]PETSC ERROR:
> ------------------------------------------------------------------------
> [0]PETSC ERROR: ./updator on a linux-gnu named coffee by don Mon Jan
> 25 18:28:31 2010
> [0]PETSC ERROR: Libraries linked from
> /home/don/Apps/Installers/petsc-3.0.0-p10/linux-gnu-c-debug/lib
> [0]PETSC ERROR: Configure run at Mon Jan 25 17:57:35 2010
> [0]PETSC ERROR: Configure options --with-hdf5=1 --with-hdf5-dir=/usr
> --with-shared
> [0]PETSC ERROR:
> ------------------------------------------------------------------------
> [0]PETSC ERROR: MatView_SeqAIJ() line 625 in src/mat/impls/aij/seq/ 
> aij.c
> [0]PETSC ERROR: MatView() line 643 in src/mat/interface/matrix.c
>
> So it seems that writing sparse matrices to HDF5 is not yet
> implemented.  I suppose such an implementation should be possible.
> For now I'll try a binary viewer as suggested by Barry Smith.
>
> If I were to implement the writer using HDF5 where should I start in
> the PETSc source?