> There are many issues and tradeoffs with mesh simplification. There are > many algorithms and approaches, each with their own unique strengths and > weaknesses. Challenges include finding a strategy to hide the cracks > between adjacent tiles draw with different levels of details (and > possibly more or less details on the edges.) You also have issues of objects > floating or being partially (completely) burried if the underlying > terrain becomes coarser or more detailed (or you have to move the objects up > or > down which can also be odd looking.)
As I said previously, I wouldn't compute the simplified mesh online - I would compute it offline and store it in addition to the regular mesh on disk - if the simplified mesh has a factor 10 lower linear resolution, it has a factor 100 less data and is hence just 1% more to store on disk. Dropping mesh points offline, you can require that no vertices from the tile edges are ever dropped, hence a multiple tile mesh will stay continuous and never have cracks. Objects should no longer be visible for tiles where we start seriously considering LOD schemes. We can render 40-50 km visibility of hires scenery just fine - so if I hear terrain LOD, I'm thinking how to simplify the mesh more than 50-60 km away, and objects from that distance are really a non-issue. > So there is always the universal principle that it's faster to do less > work. But styles and algorithm popularity is a moving target as hardware > is a moving target In this case, I think it's rather straightforward. I don't have Curt's experience of course, but: * terrain starts to get 'really nice' with rivers actually following valley bottoms and nice curving roads with a factor 10 or so (probably even more) better linear resolution than default terrain. * a realistic visibility from 33.000 ft in clear weather is close to 1000 km, not the 120 km max. we offer If we want to get 'realistic' in the present scheme and just go ahead and render full detail, memory requirements grow parametrically by a factor 10.000 - you need to have that *before* you talk about algorithms to render it. Now, my impression is that currently the memory outfit of new computers grows by about a factor 2 every 4 years. To get the factor 10.000, we need the hardware from about 50 years from now if the trend doesn't slow down (which it will, because there is a lower limit how small structures on a chip can be before quantum effects kick in and your memory chip becomes a random number generator). Which, for me, is just screaming 'LOD scheme!'. Well, just my two cents... * Thorsten ------------------------------------------------------------------------------ LogMeIn Rescue: Anywhere, Anytime Remote support for IT. Free Trial Remotely access PCs and mobile devices and provide instant support Improve your efficiency, and focus on delivering more value-add services Discover what IT Professionals Know. Rescue delivers http://p.sf.net/sfu/logmein_12329d2d _______________________________________________ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
