[Paraview] Bug in interpolation of point data

[Moreland, Kenneth]

Yup. This is a known issue with how pretty much every implementation of OpenGL renders quadrilaterals.

For the most part, ParaView will interpolate the fields on your hexahedra and quadrilaterals correctly. For example, in either data set, if you add a plot over line filter and plot along the Y axis, you will see that the plot over line filter is symmetrically interpolating the cells (see the attached screenshot).

The issue is that both data sets are rendered by sending quadrilaterals to the OpenGL rendering engine. Every implementation of OpenGL I have run into in at least the last 15 years internally breaks the quadrilaterals into triangles (because it makes the rasterization faster). It turns out, the linear interpolation of the two triangles is subtly different than the bi-linear interpolation on a quadrilateral. And, most importantly for your specific question, the interpolation is subtly different depending on which which diagonal is used to split the quadrilateral. In your case, the rendering system is picking different diagonals for the rotationally isomorphic quadrilaterals (ooh, that sounds mathy) and is causing the asymmetric interpolation.

We've had multiple discussions about fixing this. In the end we decided to leave it as it is. The issue is that the fix is to break polygons like quadrilaterals into several triangles that are closer to the bilinear interpolation. The issue is that it would add a large computation and memory overhead for an issue that is almost never noticeable.

-Ken

From: Marco Goetz <marco.goetz at tu-dresden.de<mailto:marco.goetz at tu-dresden.de>>
Date: Wednesday, March 18, 2015 at 4:20 AM
To: "paraview at paraview.org<mailto:paraview at paraview.org>" <paraview at paraview.org<mailto:paraview at paraview.org>>
Subject: [EXTERNAL] [Paraview] Bug in interpolation of point data

Hello,

I tried to interpolate nodal values using ParaView.
But for symmetric problems there are asymmetric results, as can be seen in the figures appended.
I tested it with hexahedron and quad type elements, but both give the same results.

You can find my simplified input file also appended.

Hopefully you can help me to create symmetric interpolation results for symmetric problems.


Thank you in advance, best regard

Marco Götz

----------------------------------------------
Technische Universität Dresden
Faculty of Civil Engineering
Institute for Structural Analysis
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/paraview/attachments/20150318/21b57665/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ParaView001.png
Type: image/png
Size: 180663 bytes
Desc: ParaView001.png
URL: <http://public.kitware.com/pipermail/paraview/attachments/20150318/21b57665/attachment-0001.png>