[ANAG] [anag-cvs] [chombousers] Re: [Paraview] CHOMBO file converter for PARAVIEW

Tue Apr 25 15:36:16 EDT 2006

Despite the new classes included with VTK my converter may be of use in the 
short term. here is why:

In the process of developing my converter I encountered a number of instances 
where VTK's defaults applied to multiresolution data produce erroneous 
visualizations. One of the issues behind this is that often CHOMBO data is 
cell centered data while many useful VTK algorithms such as contouring 
require node centered data. While this can easily be created by using the 
CellToPointData filter, if left to VTK's defaults the contouring and color 
mapping routines produce erroneous results. In the color maps the boundaries 
between different Levels(and even between boxes depending on how the data is 
created) are clearly visible. for instance see: 
http://quaoar.sr.unh.edu/images/boxes_problem.png
In the case of the contours, VTK's algorithm generates contours that don't 
connect across Levels(and in some case boxes, again depending on how the 
dataset is created). for instance see: 
http://quaoar.sr.unh.edu/images/contours_not_closed.png

In the process of developing my converter I did a case study on these issues 
and developed solutions that solve the problems. I found that the choice of 
cell type and how the point data(required for certain visualizations) are 
created, are the main issues to address. Out of all the VTK cell types 
triangles, and tetra are the least prone to visualization errors, and that 
the best case is to use the cell centers them selves as the point set to 
triangulate/tetrahedralize. Any other choice for a set of points would 
require interpolation of the CHOMBO data and introduce error. Please see the 
slide illustrating these issues: 
http://quaoar.sr.unh.edu/chhdf52FlatVtk/amr_vis_cell_types.png

The chhdf52FlatVtk converter addresses these issues, by using VTK_TRIANGLE 
cell type in 2d and VTK_TETRA in 3d  generated from point data introduced at 
the CHOMBO cell centers. for example see: 
http://quaoar.sr.unh.edu/chhdf52FlatVtk/resolved_issues.png

So the question that comes to mind is how do the new VTK classes handle these 
issues? From a preliminary test using the Hierarchical Fractal Source 
included with PARAVIEW I'm not convinced that they do. As a result there may 
still much work to be done on these before they can be used in some specific 
visualizations, such as iso surface & contours. The following slide was 
created using the Hierarchical Fractal source included with PARAVIEW. It 
shows that VTK uses a naive choice of cell types when constructing the 
hierarchical data, which leads to contour algorithm breaking and the color 
mapping issues described above. In addition in order to generate the point 
set required for these visualizations the CellToPointData filter is used 
introducing interpolation error. 
http://quaoar.sr.unh.edu/chhdf52FlatVtk/hierarch_data_issues.png

I have not done an in depth study of the new classes, and only tested them 
using my current installation of paraview (2.2 release 1) so I cannot say 
weather these issues have been addressed yet. 

Burlen

On Monday 24 April 2006 04:37 pm, Berk Geveci wrote:
> My work is similar to John's work but is now built into VTK proper. It
> is relatively recent, a little over 1 year old. It supports
> rectilinear AMR grids natively and does not convert them to
> unstructured grids. Support for curvilinear grids is not hard to add.
> If you have a current ParaView (as of 2.4, I belive), you can see it
> in action by creating a Hierarchical Fractal from the Source menu.
>
> Also, CTH is structured and supports AMR:
> (http://www.cs.sandia.gov/web9232/cth/).
>
> -Berk
>
> On 4/24/06, Brian Van Straalen <bvs at hpcrdm.lbl.gov> wrote:
> > Well, that would be of interest to us.  I was not aware that VTK had any
> > native AMR data structures.
> > Is this a fairly recent development ?  I thought CTH was an unstructured
> > finite element code.
> >
> > are are we still discussing turning the structured-adaptive data into a
> > fully unstructured grid for
> > visualization ?
> >
> > Is this the converter that John Shalf originally put together, or
> > have their been multiple independent efforts to create HDF5toVTK
> > data converters ?
> >
> > Berk Geveci wrote:
> > >Have you considered writing a native AMR reader for Chombo?
> > >VTK/ParaView support AMR datasets now. I had a simple Chombo reader
> > >working at one point but it is not production quality. The only AMR
> > >reader in VTK is for CTH right now.  I would be happy to help.
> > >
> > >-Berk
> > >
> > >On 4/24/06, Burlen <burlen at apollo.sr.unh.edu> wrote:
> > >>Currently there are a number of options for visualizing CHOMBO data
> > >> including CHOMBOVIS, and VISIT. I'd like to contribute yet another
> > >> option, a file converter that allows the data to be used with PARAVIEW
> > >> or any other application that can read VTK's file format. The
> > >> converter's primary use is to construct a dataset spanning the entire
> > >> problem domain using the highest resolution data available for a given
> > >> spatial location, while avoiding visualization issues associated with
> > >> multi resolution data. Also the converter can extract boxes from a
> > >> specific level, allowing one to visualize CHOMBO's box layout, level
> > >> by level. One nice feature of the converter is that it has a command
> > >> line interface and can batch process CHOMBO's plot files, resulting in
> > >> a numbered sequence of VTK files which PARAVIEW can animate easily.
> > >> Currently the converter only handles 2 & 3D cell centered data, and
> > >> only reads the HDF5 file format used by CHOMBOVIS. It has gone through
> > >> a number of revisions, and may be usefull to those who prefer PARAVIEW
> > >> for visualization. Development and preliminary testing have been done
> > >> using Fedora 2 linux (single proc), HDF5 1.6.3, and VTK 4.4.2 I hope
> > >> to make improvements in the future, and test the converter on more
> > >> platforms. More information and tarballs can be found at
> > >>http://quaoar.sr.unh.edu/chhdf52FlatVtk
> > >>
> > >>--
> > >>
> > >>Burlen Loring
> > >>Space Science Center
> > >>Institute for the Study of Earth, Oceans, and Space
> > >>University of New Hampshire
> > >>39 College Road, Durham, NH 03824
> > >>Phone: 603-862-1140
> > >>_______________________________________________
> > >>ParaView mailing list
> > >>ParaView at paraview.org
> > >>http://www.paraview.org/mailman/listinfo/paraview
> > >
> > >_______________________________________________
> > >Chombousers mailing list
> > >Chombousers at hpcrdm.lbl.gov
> > >https://hpcrdm.lbl.gov/mailman/listinfo/chombousers
> > >_______________________________________________
> > >anag-cvs mailing list
> > >anag-cvs at hpcrdm.lbl.gov
> > >https://hpcrdm.lbl.gov/mailman/listinfo/anag-cvs
> > >_______________________________________________
> > >ANAG mailing list
> > >ANAG at hpcrdm.lbl.gov
> > >https://hpcrdm.lbl.gov/mailman/listinfo/anag

-- 

Burlen Loring
Space Science Center
Institute for the Study of Earth, Oceans, and Space
University of New Hampshire
39 College Road, Durham, NH 03824
Phone: 603-862-1140