Robert Webb (robertw++at++wormald.COM.AU)
Tue, 13 Jun 1995 14:47:03 +1000 (EST)
> > I want an object to stay stationary with respect to the viewer. Here is
> > what I tried:
> > - Put the object under a DCS.
> > - Each frame calculate the matrix for the new viewer
> > position/orientation.
> > - Call pfChanViewMat() with this matrix.
> > - Call pfDCSMat() with this matrix.
> >
> > So the view and object both use the same matrix. This almost works. I can
> > see part of the object within my view, but it jiggles around. I printed out
> > the matrices used for the view and for the DCS each frame in a draw
> > call-back, and sure enough they are different! Surely both the view
> > transform and the DCS transform are passed down the Performer pipeline (ie
> > they should be frame accurate), but it seems one of them is not. If I run
> > in a single process then the matrices printed out are the same, but it still
> > appears to jiggle around!
> >
> > Can anyone shed any light on what's going on? Or how to keep an object
> > completely still with respect to a viewer? I could just use a draw call-back
> > to load some constant matrix onto the stack, but I also want other viewers
> > to be able to see the object at the location of the first viewer.
>
> Perhaps you have some form of collision detection enabled which
> collides with the object and deflects the viewpoint.
I'm afraid I don't understand this sentence at all.
> I am certain both DCS and pfChannel view matrix are propagated properly
> down the pipeline - your printout in the DRAW is misleading because the
> matrix for the DCS is that of the APP's copy which is 1 or 2 frames in the
> future. This is a strange side effect of Performer's multiprocessing but
> we deem it livable because it saves lots of memory and processing time.
>
> The fact that you get jitters in single process leads me to believe
> that something other than multiprocessing wackiness is the problem.
Well, it's time to eat humble pie. I have discovered that the jiggling was
due to rounding error! The coordinates of our viewing position and the
objects in our scene are up around the 100000 mark, but floating point
numbers only have about 6 or 7 significant digits. Somewhere down the line
this caused the viewing position to jump around with respect to the scene,
although nothing in the scene jumped around with respect to anything else in
the scene.
Hmm, don't know how we're going to fix this, but at least now we know what
the problem is.
Rob.
------------------------------------------------------------------------------
Robert Webb. robertw++at++wormald.com.au
------------------------------------------------------------------------------
This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:51:35 PDT