> http://www.youtube.com/watch?v=Irf-HJ4fBls
>
> I'd seen about this thing a few years back, it's apparently legit.

I saw it a few years back too (maybe posted in here?) and was sceptical. 
It certainly looks impressive and I'm still trying to figure out if this 
really could be possible or not. Something like a tile-based LOD system 
could work, games do this already with polygon and texture data (game 
world size only limited by the storage media), no reason why it couldn't 
work with point data.

As for the compression I was thinking along the lines of something 
similar to the way mp3/jpeg drops information that probably won't be 
noticed. You don't need to store every point at every detail level with 
the full precision, and the point data could make use of the coarser 
data (position and colour) from the next level up.

That just leaves how to get the million or so points that are in RAM on 
the screen in the right place in the right order in a fraction of a 
second. Are CPUs fast enough today to just simply draw every one with a 
z-buffer? Seems like it could be made multi-threaded very easily, on the 
rare occasion two threads try to write to the same pixel at the same 
time it probably won't even be noticed if the pixel is the wrong colour.

Post a reply to this message