Re: [Algorithms] VIPM With T&L - what about roam

[Rui F. <rui...@mo...>]

A static array of vertices with dynamic indexing sounds interesting, and I
thought on that too, especially caching stuff on the lower levels patches,
but the problem is; how do guys handle geomorphing with this scheme ??!

Personally, I dislike roam, except its basic principles like the binary
triangles concept, diamonds, etc, but what you get with roam, because it
doesn't reset its mesh every time, is popping when a new child is split, you
can minimize it with a morph ratio that takes into account the distance to
the "to be split" node, and such, but those are all fake approximations and
you will end up with pop artifacts sooner or later.

True geomorphing can only be done if you set your morphing values *per node*
and based on your split metric. Only then you can have a smooth
interpolation from one node base edge height to the child's real height from
frame to frame.

If you "lock" any subtree and cache away its vertices you just lose
geomorphing.... again i can see ways to minimize this... but... you have a
mesh, you have vertices connecting edges, you either arrange for your
vertices positions to translate smoothly or you don't !...

I'm a one true believer in top-down, recurse, apply metric, "split-only"
algorithms for this very reason.

-Rui Ferreira