rivera@stdc01.UUCP (Greg Rivera) (06/29/90)
Wanted:
a) Fast ways to transform plane equations through arbitrary
matrices, OR
b) Fast ways to transform eye-clipping planes so I can do
clipping in some other coordinate space (still in 3D)
Problem:
My view clipping planes (top, bottom, left, right, near, far)
are defined in eye space. But my models (oodles, and moving
all the time) are defined relative to their own coordinate system.
Therefore, to do culling/clipping, I either:
A) Transform every model into eye space, then cull/clip, or
B) transform the clipping planes into the model's space, so I can
cull/clip there.
I would like to have suggestions for how to do (B) quickly.
For thought:
Are there any advantages due to the clipping planes inter-relationships?
For example, near and far planes are usually parallel.
Also, for perspective, all four side planes meet at a common point.
Or are there speedups if the model transformation is of a special
form, like rotates only (orthonormal)?
Currently, I'm stuck with sampling 3 points per plane (some are shared)
and transforming them into model space. After a cross product and
normalization to get the normal to the plane, one extra dot gets
the distance to finish off the plane equation to get Ax+By+Cz+D=0.
Total is 5 transforms, 5 cross products with normalizations, and
6 dots (assumes parallel near/far).
Please email suggestions. Thanks.
--
-- Gregory Rivera Email: ...{!sun}!rti.rti.org!stdc01!rivera --
-- "... and the wisdom to Paper: Graphicon, P.O. Box 13951, RTP, NC 27709 --
-- know the difference." Voice: (919)361-3848 Fax:(919)361-3888 --xanthian@zorch.SF-Bay.ORG (Kent Paul Dolan) (06/30/90)
uucp!rti.rti.org!stdc01!rivera (Greg Rivera) writes: >Wanted: > a) Fast ways to transform plane equations through arbitrary > matrices, OR > b) Fast ways to transform eye-clipping planes so I can do > clipping in some other coordinate space (still in 3D) > >Problem: > My view clipping planes (top, bottom, left, right, near, far) > are defined in eye space. But my models (oodles, and moving > all the time) are defined relative to their own coordinate system. > Therefore, to do culling/clipping, I either: > A) Transform every model into eye space, then cull/clip, or > B) transform the clipping planes into the model's space, so I can > cull/clip there. > > I would like to have suggestions for how to do (B) quickly. > >For thought: > Are there any advantages due to the clipping planes inter-relationships? > For example, near and far planes are usually parallel. > Also, for perspective, all four side planes meet at a common point. > Or are there speedups if the model transformation is of a special > form, like rotates only (orthonormal)? > > Currently, I'm stuck with sampling 3 points per plane (some are shared) > and transforming them into model space. After a cross product and > normalization to get the normal to the plane, one extra dot gets > the distance to finish off the plane equation to get Ax+By+Cz+D=0. > Total is 5 transforms, 5 cross products with normalizations, and > 6 dots (assumes parallel near/far). > >Please email suggestions. Thanks. >-- >-- Gregory Rivera Email: ...{!sun}!rti.rti.org!stdc01!rivera -- >-- "... and the wisdom to Paper: Graphicon, P.O. Box 13951, RTP, NC 27709 -- >-- know the difference." Voice: (919)361-3848 Fax:(919)361-3888 -- I'm a little confused by your description. If you are already doing the transformations to move your models, and to put them together in the same scene, then composing those transformations with the one to get all the data into the viewing frustrum, before you apply the transformations to the models should be fairly cheap. Quite possibly I'm missing the point. Anyway, I implemented dynamic motion with six plane clipping in an oblique perspective view (in assembly language!) from the nice math in: Richard. F. Puk, General Clipping on An Oblique Viewing Frustrum, Pages 229 to 235 in Computer Graphics, A Quarterly Report of SIGGRAPH-ACM, Volume 11, Number 2, Summer 1977, SIGGRAPH '77 Proceedings. Be warned that the math assumes exact arithmetic; in a real computer things are a bit more complex than Dr. Puk portrays them; I ended up with a few flags to assure that I didn't reuse the same clipping plane where his comparisions for equality of quantities computed in floating point two different ways failed from roundoff errors. Still, on the clunky hardware I had in 1981, it worked OK. Kent, the (unemployable) man from xanth. <xanthian@Zorch.SF-Bay.ORG> <xanthian@well.sf.ca.us>
turk@Apple.COM (Ken "Turk" Turkowski) (08/13/90)
In article <628@stdc01.UUCP> uucp!rti.rti.org!stdc01!rivera (Greg Rivera) writes: >Wanted: > a) Fast ways to transform plane equations through arbitrary > matrices, OR > b) Fast ways to transform eye-clipping planes so I can do > clipping in some other coordinate space (still in 3D) > >... See my paper on "Properties of Surface Normal Transformations" in Academic Press's new Graphics Gems book. A plane is very similar to a normal. -- Ken Turkowski @ Apple Computer, Inc., Cupertino, CA Internet: turk@apple.com Applelink: TURK UUCP: sun!apple!turk