I'm still having a lot of issues with the viewport in H13, both in the office and at my setup at home (two entirely different video cards, etc), despite a recent video card upgrade. Home: GeForce 780ti, office: Quadro 4000, both drivers are recent.
“Hidden Line Invisible” in GL 3.3 produces weird n-gon tearing artifacts, see below.
This is how I'd expect it to look (GL 1.1 and GL 2.1), see below.
“Flat Wire Shaded” shades every polygon differently (new behaviour) in GL 1.1 and GL 2.1, see below.
…yet produces a slightly better result in GL 3.3, but still with the “tearing”, see below.
And those are just the tip of the iceberg - I find myself having to switch between “light” and “dark” every few minutes in order to “wake up” the viewport when points decide to stop displaying or polygon selections no longer become highlighted, etc. I find myself doing most of my SOP work in legacy versions (ie. h11) as a result of the frustration.
Viewport/OpenGL problems continue
45959 98 4
-
- stu
- Member
- 246 posts
- Joined: July 2005
- Offline
-
- anon_user_37409885
- Member
- 4189 posts
- Joined: June 2012
- Offline
-
- malexander
- Staff
- 5156 posts
- Joined: July 2005
- Offline
My guess, at least with the GL3 hidden line, is that there is either a multiple point/vertex at the spot where the tears are. The outliner attempts to compensate for these cases, but after convexing the geometry it is sometimes difficult to conclusively determine if the edge is an interior edge or not. If you turn on Disp Opts > Markers > Coincident point detection, you can see if any coincident points exist in yor model.
The H11 renderer will not be available in releases past H13.0. Please submit bugs that you find, especially for the GL3 renderer. The geometry in this post would be a good example.
The H11 renderer will not be available in releases past H13.0. Please submit bugs that you find, especially for the GL3 renderer. The geometry in this post would be a good example.
-
- stu
- Member
- 246 posts
- Joined: July 2005
- Offline
-
- stu
- Member
- 246 posts
- Joined: July 2005
- Offline
twod
My guess, at least with the GL3 hidden line, is that there is either a multiple point/vertex at the spot where the tears are. The outliner attempts to compensate for these cases, but after convexing the geometry it is sometimes difficult to conclusively determine if the edge is an interior edge or not. If you turn on Disp Opts > Markers > Coincident point detection, you can see if any coincident points exist in yor model.
As a comparison, the hidden line geometry displays fine in earlier versions of houdini, and like I had mentioned above, also fine in GL 1.1 and GL 2.1 in H13. That said, those other H13 renderers have their own peculiarities as well (see above, the flat shaded example).
-
- Sadjad Rabiee
- Member
- 1391 posts
- Joined: Dec. 2010
- Offline
I had same problem with your bgeo file too !
Even I can't import it in Maya correctly :!:
So I think this problem is not for Houdini , VGA or is not a BUG.
It's normals are so dirty ,When I use Facet SOP to fix normals ,It's display correctly.
Also I think GL1 , GL2 , GL3 using different method and library for rendering objects.
However I agree H11 Renderer in the Houdini 12 is really cool 8)
Even I can't import it in Maya correctly :!:
So I think this problem is not for Houdini , VGA or is not a BUG.
It's normals are so dirty ,When I use Facet SOP to fix normals ,It's display correctly.
Also I think GL1 , GL2 , GL3 using different method and library for rendering objects.
However I agree H11 Renderer in the Houdini 12 is really cool 8)
-
- Sadjad Rabiee
- Member
- 1391 posts
- Joined: Dec. 2010
- Offline
-
- stu
- Member
- 246 posts
- Joined: July 2005
- Offline
-
- anon_user_37409885
- Member
- 4189 posts
- Joined: June 2012
- Offline
-
- malexander
- Staff
- 5156 posts
- Joined: July 2005
- Offline
-
- stu
- Member
- 246 posts
- Joined: July 2005
- Offline
Hopefully all of the bugs are worked out of the GL3 renderer before the others are depreciated. I find myself switching between the renderers quite often as I'm finding that depending on what I want to see (hidden line, shaded, etc.) some work better than others. I find that switching back and forth between renderers also “wakes up” the viewport and eliminates whatever artifact that might have creeped up while I am working (non-highlighted selections, etc.).
I'd be curious to know if you were able to see any of what I was seeing with my test geometry.
See below, another example of the hidden line “tearing” that I was mentioning earlier - I've emphasized the problem by turning wireframe width up to 10, but you can see what I'm talking about - GL3.3 struggles with n-gons in a way that pre-GL3.3 renderers didn't. Otherwise, it's a simple box SOP with a polysplit SOP.
I'd be curious to know if you were able to see any of what I was seeing with my test geometry.
See below, another example of the hidden line “tearing” that I was mentioning earlier - I've emphasized the problem by turning wireframe width up to 10, but you can see what I'm talking about - GL3.3 struggles with n-gons in a way that pre-GL3.3 renderers didn't. Otherwise, it's a simple box SOP with a polysplit SOP.
-
- stu
- Member
- 246 posts
- Joined: July 2005
- Offline
Update: the “tearing” problem is apparent in any of the display modes that attempt to display edges in conjunction with solid polys (flat wire shaded, smooth wire shaded, hidden line invisible, hidden line ghost). It comes down to the new rendering scheme - it looks like each poly receives a “texture map” now that defines the boundaries with a ramp that creates the illusion of the edges, unintentionally illustrating the behind-the-scenes triangulation in the process in the case of an n-gon.
13.0.288, btw.
FYI his problem also occurs in H12 when the scene renderer is set to GL 3.2. In H12 I'd set the Scene Renderer to H11 in order to avoid these problems, but that option has since been depreciated.
13.0.288, btw.
FYI his problem also occurs in H12 when the scene renderer is set to GL 3.2. In H12 I'd set the Scene Renderer to H11 in order to avoid these problems, but that option has since been depreciated.
-
- stu
- Member
- 246 posts
- Joined: July 2005
- Offline
-
- anon_user_37409885
- Member
- 4189 posts
- Joined: June 2012
- Offline
-
- malexander
- Staff
- 5156 posts
- Joined: July 2005
- Offline
Turns out that these artifacts are occurring on edges on badly tesselated triangles, where the angle between an edge and a non-edge approaches 180°. The distance from the fragment to the edge line is less than the wire width for a portion of the non-edge, as the two are nearly parallel. This causes the bad shading.
-
- stu
- Member
- 246 posts
- Joined: July 2005
- Offline
…which wasn't a problem pre-H12 and the new GL rendering scheme. So what's the solution? Any chance of re-introducing some sort of back door to the H11 behavior? Again, consider that this is just one of the issues that I'm having, it just happened to be one of the most easily reproduceable. And again, consider the fact that I'm seeing all of these issues on two different setups with two different and very capable video cards.
-
- anon_user_37409885
- Member
- 4189 posts
- Joined: June 2012
- Offline
stu
…which wasn't a problem pre-H12 and the new GL rendering scheme. So what's the solution? Any chance of re-introducing some sort of back door to the H11 behavior? Again, consider that this is just one of the issues that I'm having, it just happened to be one of the most easily reproduceable. And again, consider the fact that I'm seeing all of these issues on two different setups with two different and very capable video cards.
Backdoor to H11 viewport is there already - set the environmental variable
HOUDINI_OGL_H11_VIEWPORT = 1
-
- stu
- Member
- 246 posts
- Joined: July 2005
- Offline
-
- anon_user_37409885
- Member
- 4189 posts
- Joined: June 2012
- Offline
-
- stu
- Member
- 246 posts
- Joined: July 2005
- Offline
-
- Quick Links