OpenGL cards and multiple windows / WinNT

[John Biddiscombe]

Lisa,

>Here are a few more questions for you - what type of polygons are you 
>drawing? Triangles, triangle strips, polygons....?  Is there any chance 
>that you are drawing concave polygons? Non-planar polygons? Non-simple 
>polygons? The pattern continually changes as you move - what if you 
>re-render and stay still (for example, hit the "s" key) - does the image 
>change? What if you change the clipping range and re-render without moving, 
>does the image change?

-------------
?? what type of polygons are you drawing? Triangles, triangle strips,
polygons....?  

In the images I've posted. Triangles only. Not strips, just triangles

The pipeline is one of two (depending on source of data)
either
Points->Delaunay2D->DecimatePro->Actor
or
SPReader->StructuredPointsGeometryFilter->WarpScalars->TriangleFilter->Actor

Please note that in reponse to your earlier question "does it work with a
simple sphere", well, if I set the radius of the sphere to 
EarthRadius = 6.3781639E6;  then yes, I get the funny artefacts when zooming.

----------------------
??Is there any chance 
>that you are drawing concave polygons? Non-planar polygons? Non-simple 
>polygons? 

No chance at all. I can reproduce the artefacts using simple planesources,
spheresources, anythingatallSources - only requirement is that the floating
point range approaces 6 figures or so.

-------------------------
??The pattern continually changes as you move - what if you 
>re-render and stay still (for example, hit the "s" key) - does the image 
>change? 

I find I can zoom in better on an image if I tilt it so I'm looking
straight down. Then I zoom in until  it starts to break up, then back off a
little. If I now begin to rotate the image, as the clipping ranges alter
with the changing bbox of the dataset, so the patterns of black spots
flicker and change. Like ripples on the surface of the sea - except more
random.
The effect is similar to problems I've seen in the past with dodgy rounding
errors etc etc, 

Very slight movements of the dataset can cause very large changes in the
image. but rerendering with a fixed viewpoint does not change the
efffects...see below

--------------------------
??what if you 
>re-render and stay still (for example, hit the "s" key) - does the image 
>change? 

Well, suppose I display one of the "broken images" and then hit "w" - we go
into wireframe and the image stays exactly the same except its now
wireframe, black spots still appear blanking out pieces and yes, they do
obscure the wireframes, not just the "holes" in between.

---------------------------
What if you change the clipping range and re-render without moving, 
>does the image change?

This I've not tried, but if you need an answer I'll report back. If I were
to guess, then I'd say "Yes", for example if I display an image with broken
patches all over it, then resize the window to cause it to regenerate the
image at a diferent resolution, the black patches appear relatively
speaking in the same places. In fact in front of me I've got a window with
one of the displays I posted earlier and I can resize it from postage stamp
size up to full screen and the patterns of black spots is absolutely rock
solid.

Its this behaviour that makes me believe that 
The black spots are not random, but are the result of a calculation which
has gone wrong, they are completely reproducible time after time.

-------------------------------
And further to my earlier message.....
The reason I said "recent changes to the clipping range stuff in vtk"
caused the problems was simply that previously the clipping planes were not
recomputed every frame and so when the image weas first displayed and it
was OK, zooming in and out didn't cause problems. The patterns only change
when the depth calculation changes



OK I've just tried another test.....
Displayed an image with dodgy black patches on it...Now holding shift down
and sliding the terrain around the screen from left to right and up and
down. The clipping ranges should not be altering (I don't think - depends
how you've done the calculation - LOS to corner or just range along
boresight - I assume boresight) - 
Anyway....the black splodges follow the dataset around. They are fixed at
particular places. Not changing. Actually, there are a few 'pixellation'
effects as slightly different pixels come in and out of view, but the black
spots are glued to the terrain.

[My diagnosis : Dodgy Depth buffer calculation.]


John "Sherlock" B
        :)




-----------------------------------------------------------------------------
This is the private VTK discussion list.  Please keep messages on-topic.
Check the FAQ at: <http://www.automatrix.com/cgi-bin/vtkfaq>
To UNSUBSCRIBE, send message body containing "unsubscribe vtkusers" to
<majordomo at gsao.med.ge.com>.  For help, send message body containing
"info vtkusers" to the same address.     Live long and prosper.
-----------------------------------------------------------------------------