Rendering problems with per-cell colored surfaces on Surface 7

Development
1

Hello everyone,

I’m having problems with using my VTK + QT application and per-cell coloring on Surface 7 tablet. I use VTK version (a5f938b2fdfefa522439065e040cb42f65d88444) which was used btw for the latest ParaView release (v5.8), which works great on both tablets), VS2017 in x64 mode and here are results:

  1. Everything works fine on Surface 6 (model 1796), Intel UHD Graphics 620, but,
  2. There are rendering problems on Surface Pro 7 (model 1866), Intel i5-1035G4, 8Gb-RAM, Intel Iris Plus Graphics, when using per-cell colored surfaces. When I try per-point colored surfaces, everything is rendered fine.

I created short example by editing ElevationFilter example and captured video of problem on Surface 7. Btw, I built VTK with VS2017 in x64 mode with mostly default parameters (turned on Imaging, Qt, Rendering, Standalone and Views groups). I also tried some earlier versions of VTK, but, bad behavior was persistant). Installing the latest graphics drivers didn’t help. Do you have idea about what I’m doing wrong?

Thanks,
Nenad.

ElevationFilter.zip (2.2 KB)

2

Yes, unfortunately, VTK rendering of per-cell coloring is broken with Intel Iris Plus Graphics (with driver date 2019-12-21, version 26.20.100.7641, found in Surface Book 3).

  • This is reproducible the same way with multiple VTK-based applications that uses OpenGL2 rendering backend, such as ParaView and 3D Slicer.
  • This is not a recent VTK regression, because VTK version about 1-2 years ago has the same issue.
  • The legacy OpenGL backend renders correctly.
  • Rendering works well on other graphics cards.
  • Coloring by point scalars works well.

Example data set with cell scalar to reproduce the problem: https://1drv.ms/u/s!Arm_AFxB9yqHuroxLpY2DfhNA5ppWQ?e=E21iyv

3

Andras,

Thank you very much for your response! Same behavior is with the latest (26.20.100.7870) driver. Probably unfixable …

Thanks,
Nenad.

4

It is possible that a driver bug (or valid but unusual behavior) exposed an error in VTK, which could be fixed easily. If enough people will report this, or somebody pays for it, then I’m sure Kitware will have someone look into this. A driver update might also make the problem go away in any minute.

I’ve reported this on ParaView bugtracker to give a chance that this issue gets noticed: https://gitlab.kitware.com/paraview/paraview/issues/19720

5

tbh it looks much more like an driver issue than anything.

6

Yes, we know that this is a driver issue, but we also know that it can be fixed in VTK (because the legacy OpenGL rendering backend does not have this problem). Maybe someone who knows how such rendering problems can be debugged could make some progress with this in a few hours (potentially find a bug in VTK or come up with a workaround, or just report the error to Intel).

7

Indeed.

8

It is probably a bug in their buffertexture or glPrimitiveId driver code. I suspect the best way to get it fixed is for someone seeing the issue to send intel an apitrace of the issue

and report it on their forums here

https://software.intel.com/en-us/forums/virtualization-software-development

1 Like