[Paraview] pvtr files doesn't work

[Berk Geveci]

Hi Michael,

There are some subtle but important differences between VisIt and
ParaView when it comes to handling structured data:

1. VisIt's approach is to keep separate pieces - or domains or
partitions, whatever you choose to call them - separate after loading
a multi-piece dataset such as yours. They do this because it allows
them several optimization specially on the IO side. On the other hand,
ParaView treats structured datasets as one big thing, even though the
pieces may be in separate files. When reading such datasets in
parallel, ParaView will decide how to partition the dataset and have
some MPI ranks read subsets of files that lie in the boundary, if
necessary. So at the end, the partitioning of the two tools will end
being very similar. ParaView may have a slight IO overhead if it needs
to assign some files to multiple processes. The biggest difference is
that the ParaView pipeline (the VTK pipeline actually) can make
requests at a cell granularity whereas VisIt works at a piece
granularity. Honestly, this is not a big difference for most
scientific visualization application but it has a huge impact when
analyzing medical datasets because of the algorithms involved.

We are working towards integrating VisIt's design in addition to the
traditional VTK way of dealing with structured data. To get a feel for
it, you can replace your parallel rectilinear grid file with a
multi-block file (vtm). You don't have to change the actual vtr files.
I can describe how to do it if you don't already know.

2. In VisIt, the only type of rectilinear grid is vtkRectilinearGrid,
which requires explicit storage of x, y and z coordinates. In VTK,
hence ParaView, uniform rectilinear grids are best stored at
vtkImageData. The advantage of this is minimal when it comes to memory
usage but significant when it comes to algorithms. There are a large
number of algorithms targeted to vtkImageData in VTK even though many
of these are not exposed in ParaView by default. One of these
algorithms is volume rendering. At this time, ParaView supports volume
rendering of image data and unstructured grids only. In the next
release (4.0 not 3.10), we will support volume rendering for more data
types.

I tried to explain this as much as I can without writing a whole book
chapter. I hope it makes sense :-)

By the way, if you are using .pvtr files, you don't need ghost zones
(cells). ParaView will take care of loading ghost cells from neighbor
pieces if necessary.

As for the Tecplot files, are these binary files by any chance? Can
you send me an example?

Best,
-berk

On Thu, Feb 3, 2011 at 7:20 AM, Michael Scheerer <M_Scheerer at web.de> wrote:
> Hi Berk,
>
> Thank you for the answer!
>
> An addition question:
>
> Paraview supports in the case of CGNS files volume rendering.
> Volume rendering is in the list together with "Outline", "Surface" and so on.
> This is not the case if someone tries to load VTK-files.
> Why not?
>
> So I couldn't see, if the ghost zone overlap works in the sense, that all the parts are really merged to one grid and not independent parts - even in the case of volume rendering.
>
> Also PLT files crashes, although there are loadable with Tecplot and VisIt.
>
> Best,
> Michael
> ___________________________________________________________
> NEU: FreePhone - kostenlos mobil telefonieren und surfen!
> Jetzt informieren: http://produkte.web.de/go/webdefreephone
>