|
|
title: a great leap in graphics.
authors: gibbs, w. wayt
source: scientific american mind; 2006, vol. 17 issue 4, p80-83, 4p, 11c
"razor's kd-tree also represents the scene at
multiple levels of detail, so that it can quickly draw a distant castle,
for example, without having to fetch data on every one of its bricks."
So, a single user might not fetch data, not only from where he isn't
but also what he can't see in this tree structure.
The word where becomes in where in the kd-tree.
The examples of random mountains
from Angel's book might work for single viewpoint of folks in a theater,
but I believe these people above might be wise about kd-trees
distributed amongst viewoints that might follow the croquet model.
On another front :
"ray-tracing algorithms have meanwhile improved so much that they can
achieve interactive speeds on a single high-end pc. in 2004 slusallek...
(to save on computation, the rays are traced in reverse; the
physics is the same.)"
So, as Ric Moore said, the models for a raytracing graphics card
might only care about integer pixels of a raster projected backward
to, not points, or planes, but smooth surface abstractions. Might I say,
with doubts I still don't know how to think of these things
even after an entire Angel text.
I have to localize my alleged abstract smooth sphere
and that is going to take significant figures.
I've flown way off the croquet virtual worlds to see what would happen
to push the edge to understand what was going on.
I think I saw something evidencing the edge of
what on p.241 Angel called a frustrum.
|
Locked
