Histogram looking horrible inside paraview

ParaView Support
1

Hi All

I’m using paraview on an Arch linux system with Intel HD 5500 graphics, built from the github commit b99817087bd572aa894b67d67708e4d5b78b0676. Everything seems to work fine, except for the histogram, which for some reason looks like this

paraview_histogram
paraview_histogram1230×346 15.2 KB

It looks like this for any color I choose. E.g, when choosing red, it looks like this:

paraview_histogram2
paraview_histogram21227×347 13.9 KB

Does anyone know how to prevent the histogram from having this patchy appearance?

Many thanks!

2

Same issue here (archlinux) with a today master (a5d1ee6). Note it also impact the color map editor.

Works well when using my NVIDIA GPU instead of Intel graphics.

3

I can’t reproduce this on my machine, however my suspicion is it’s one of these two VTK commits:

  • 2f824a5b59 Rename some OpenGL methods add some docs
  • 13956bf0d4 clean up some old opengl es stuff

Can someone who can reproduce this try to isolate which commit it is and let the author know?

4

Just a quick sidenote: when installing paraview using pacman (paraview 5.8.0-7), it looks fine.

5

I can’t replicate on Linux, Intel graphics, builtin server, disk_out_ref.exo/ Histogram, master as of v5.8.1-RC1-960-ga608a3c

6

I’m on it !

1 Like
7

This seems to be due to this MR : https://gitlab.kitware.com/vtk/vtk/-/merge_requests/6989

Thanks @utkarsh.ayachit for pointing out the commit, it saved some compute time on my machine !

8

I cannot reproduce this (windows nvidia and intel) which isn’t that surprising at it is probably a driver bug. So if someone can help out maybe we can remotely narrow down the issue. A first test would be to update vtkOpenGLRenderWindow around line 950 and 1015

// if we have a MSAA buffer we have to resolve it using a shader as opposed to
    // a normal blit due to linear/gamma colorspace issues
    if (this->MultiSamples > 1 && this->RenderFramebuffer->GetColorAttachmentAsTextureObject(0))
    {

and make this always false e.g. if (false && …) and see if that fixes the issue.

9

Indeed, making this line always false does the trick!

10

can confirm

11

Can someone on an impacted system do Help -> About in paraview and copy paste the last three lines that show the OpenGL version/Vendor/Renderer ?

12

OpenGL Vendor: Intel
OpenGL Version: 4.6 (Core Profile) Mesa 20.1.3
OpenGL Renderer: Mesa Intel® HD Graphics 5500 (BDW GT2)

13

OpenGL Vendor: Intel
OpenGL Version: 4.6 (Core Profile) Mesa 20.1.3
OpenGL Renderer: Mesa Intel® HD Graphics 630 (KBL GT2)

14

https://gitlab.kitware.com/vtk/vtk/-/merge_requests/7100 in testing

15

Any chance someone could try out these changes in their local PV build that shows this issue. Just to make sure I have it correct?

16

Seems to work, histogram looks nice. Thanks so much for the effort!

17

Awesome, merged to VTK master. Will probably be a week or two before it gets pulled into PV master as they update their VTK hash about that often.

1 Like