Re: [Plib-devel] Terrain optimization in ssg?

[Steve B. <sjb...@ai...>]

Gil Carter wrote:
> 
> At 02:15 PM 8/20/00 -0400, you wrote:
> >I seem to remember a while back that someone
> >was talking about adding one of the terrain optimization
> >algorithms (ala ROAM and others) to ssg.  Does anyone
> >know who brought this up and if anything has been done about
> >it?
> 
> Well, I did mention it about a fortnight ago as a possible future
> contribution, but my efforts are currently on the VRML import/export
> followed by documentation and then the ssgStatistics module.  I think that
> it will be a fair while before I get to implementing ROAM for PLIB in my
> spare time!
> 
> That said, ROAM is only one of a number of continuous level of detail
> schemes for real time rendering - albeit a fairly oft-discussed one.

..and it's generally agreed that you don't really want to implement
ALL the subtleties of ROAM because some of the final layers of optimisations
do bad things to memory locality - and are actually somewhat pessimal.

>  If you're thinking of doing some work on a CLOD system for PLIB, you might
> like to also look at some of the other methods of doing it - adaptive
> quadtrees, view independent progressive meshes and bin tri-trees, as well
> as simple distance based switching of meshes.

Distance-based mesh switching is of course already supported in SSG...although
not with alpha blending between levels.

>  Ben Discoe's VTerrain site
> has a good summary of these techniques, with links to their original papers
> at http://vterrain.org/LOD/published.html

There was significant discussion of ROAM and other continuous terrain LOD
mechanisms at SigGraph.  The bottom line conclusion was that the decision
as to whether to go with fixed levels of detail - or continuous LOD - changes
with the prevailing graphics technology - and that currently the availability
of cheap T&L hardware means that the pendulum is swinging back towards multiple
fixed levels of detail - or perhaps no level of detail management at all.

This definitely reduces my enthusiasm for implementing ROAM in SSG - although
if someone did want to write it, I'd be very happy to accept such a contribution
into the codebase.

-- 
Steve Baker   HomeEmail: <sjb...@ai...>
              WorkEmail: <sj...@li...>
              HomePage : http://web2.airmail.net/sjbaker1
              Projects : http://plib.sourceforge.net
                         http://tuxaqfh.sourceforge.net
                         http://tuxkart.sourceforge.net
                         http://prettypoly.sourceforge.net