Hi,
A humble request/suggestion to make WorldMachine more Quadtree friendly:
For my - and I assume most - quadtree systems,
the image size is ideally (a power of 2) + 1,
so that the number of quads is a power of 2.
Any chance of suporting this in a future build?
B
Would there be an advantage when the terrain size is 2^x+1?
Would there be an advantage when the terrain size is 2^x+1?Well, If the image is that size, you can map one pixel to one terrain vertex on the quadtree. The rough-most node on the quadtree is a 3x3 vertex "array" the next, more refined, one is 5x5, and so forth..
Long time ago when I played with that sort of stuff, I always started from the final huge image, added the conveneience points at the border, and sliced it into the appropriate pieces. It’s a timeconsuming run-it-once step, but it’s automatic. It was easy for me to do it, as I was coding it I needed to leave one repeated pixel at each adjacent image, for convenience…
I also generated some texture for the terrain at that step. It was nice to see 250MB of files appear out of nowhere. (250 MB was a lot of disk space then)
Hi Fil,
Yes, I do that too and it works fine, until you start creating tiles.
The joins by the “convenience” row/column are noticable.
Hmm, just thought, I wonder if the “convenience” points could be set to match the adjacent tile…
Perhaps (power of 2)+1 images are not neccessary after all ?
Sorry if this was a pointless thread,
I’ll try it out…
well, since I was starting from a huge file and slicing up, I always matched the vertices right for each tile I sliced! the only points I added where those on the edge of the huge terrain (you’d have to travel a lot in the “game” to see those, and if you crossed them you’d discover the End-Of-The-World, literrally :lol: )
Sorry if this was a pointless threadNah! If it was usefull in some way, then it has never been pointless! Besides, everything becomes "pointless" after the "point" is reached ;)