[petsc-dev] hdf5 self-describing output?

[Matthew Knepley]

On Sun, Oct 16, 2011 at 10:21 AM, Jed Brown <jedbrown at mcs.anl.gov> wrote:

> On Sun, Oct 16, 2011 at 10:04, Matthew Knepley <knepley at gmail.com> wrote:
>
>> You can however point the xdmf to an HDF5 file with multiple timesteps.
>
>
> Be careful with this. Last I heard, neither VisIt or ParaView handled large
> numbers of time steps well when using a single XDMF file.
>

Thats fine. We can write multiple xdmf files if something goes wrong. So
far, nothing has.


> I don't think the VTK output can ever be a complete solution and I'm
> skeptical that XDMF could be used as a checkpointing mechanism. At this
> point, VTK is just a practical way to get vis in parallel with zero
> dependencies.
>

xdmf is not used for checkpointing, HDF5 its. And it is fine.

   Matt


> I have a dream of eventually having model code (via the DM interface)
> callable by the vis plugin so that we could store only state variables (e.g.
> density, momentum, energy) and compute derived quantities (e.g. velocity,
> pressure, temperature, lift, drag, fluxes, stresses) in a consistent way. As
> it is, you have to write loads of redundant information into files if you
> want to visualize it. That's lame.
>

-- 
What most experimenters take for granted before they begin their experiments
is infinitely more interesting than any results to which their experiments
lead.
-- Norbert Wiener
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20111016/731bc74b/attachment.html>