search

visit-dav / visit

VisIt - Visualization and Data Analysis for Mesh-based Scientific Data
https://visit.llnl.gov
BSD 3-Clause "New" or "Revised" License
405 stars 110 forks source link

Sliced tensor rendering different depending on system (system-gl vs mesa ?) #4919

Open biagas opened 3 years ago

biagas commented 3 years ago

Investigating regression test issue with plots/label.py on Windows. The tests in question involve sliced tensors. I see strange shadows/black sections in the render plots.

I created a tensor plot of grad_tensor from noise.silo. Set stride to 1 in the Tensor plot atts. Add slice operator.

Zoomed in on a cluster of the glyphs.

Here's what I see on Windows and my rhel6 desktop: rhel6-systemgl

Here's the same plot from Pascal: Pascal-mesagl

Not sure this is really a bug, but wanted to document it.

biagas commented 3 years ago

Investigate the poly data being generated by the tensor glyphing. It looks like it's 3d, but the slice should generated 2d.

markcmiller86 commented 3 years ago

Do any of our operators that manipulate geometry (slice, 3-slice, clip, transform) operate on the geometry of glyphs? Glyphs feel like they live down-stream of the operators in the rendering aspect of things and so don't really integrate with the operators as we might assume. For example, a slice of a 3D vector field yields a 2D slice with 3D vectors on it, doesn't it?

Well, I just tried slicing the disp vector variable from globe.silo. Indeed, I do get a 2D slice (projected to 2D) and the vectors too are projected to 2D. However, they all kinda look like they are maybe the same length as if they were in 3D and so I am not entirely sure the vectors are being projected to 2D in this case correctly. And, when I UNcheck the slice operator's Project to 2D option. What I get is a 3D slice with a bunch of vectors that are all confined to the plan of that slice...which seems wrong then. I think I should see those vectors in their original 3D glory but just for the points that are on the slice.