On Monday 24 January 2005 20:57, Curtis L. Olson wrote: > Hi Durk, > > The terrain elevation system could stand to be looked at a bit. I think > there is still a lurking bug where the wrong elevation can be returned > under some circumstances immediately after a tile boundary is crossed. > There are some optimizations in the current system that assume you are > reading ground elevation from the same tile or same position stream as > before. > > Be aware that looking up terrain elevation from a TIN is an expensive > operation so we don't want to do very many of these lookups per frame. > Perhaps your AI system could do one lookup per frame and rotate through > the model list so the entire list is updated ever "n" frames ... or > prioritize models that are on/near the ground over models that are high > above the ground where it really doesn't matter so much. > > Does that answer your question? > Curt,
What I am really looking for is a hint where I can find the code in FlightGear that actually does these calculations. I tried tracing back through the functions that eventually set the value of the /environment/ground-elevation-m property, but couldn't really figure it out. Thanks for your explanation of the costs and pitfalls, I'll certainly take it into account. For a start, I would only do this for models that are on -or near- the ground, and use the general airport elevation for any other situation. Secondly, it would only be required to do this for traffic that is within a reasonable distance of the user. Thanks, Durk _______________________________________________ Flightgear-devel mailing list [email protected] http://mail.flightgear.org/mailman/listinfo/flightgear-devel 2f585eeea02e2c79d7b1d8c4963bae2d
