What type of engine would you use for...
Author
(Please ignore the writing errors ;-)
I'm asked to do a demo engine which supported really large open maps (GTA-style, but bigger), but with almost no detail in buildings (almost cubes, no incremental meshes needed I think, even no textures needed). The map would have, however, a lot of hills and mounts (no flat).
Of course I would like to use Ogre or any other suggested library, but I have a lot of doubts. Which level-partitioning-technique is better for engine like that? Does some open source engine support that? Should I write the graphic engine from scratch (which would be much more complex for me)?
Thanks!
Author
Well. Ogre for example supports Octrees and Quadtrees, not sure about ABT. Would Ogre be able to manage such a large map (even if I have to subclass the scene managing)?
Depends how far you will see and how fine a mesh you use (how many mesh quads visible and how many in whole map.
You said no variable LOD mesh..
If your world doesnt have alot of view blocks yo will see to the horizon and then if you have a large map with a fine mesh you
will be displaying ALOT of quads/triangles.
I did a system with 16x16 mesh quads in zones (a 2 step LOD) and a max range of 8 zones out (window of 15x15 zones visible of the map). The heightmap grid has regular XY coordinates.
A bitmapped angle-zone visibility table does fast subsetting of all the relative zones (pieslice) visible for a particular view angle.
My zone grids are fairly coarse heightmaps so out to 7 zones away
most human sized objects are single pixels.
Objects can be placed/limited to the ground altitude by doing a
line triangle intercept calulation (simplified because of regular grid coordinates and right mesh triangles).
Of course my world map wont fit in memory - 64k x 64k zones (and its for a create terrain on the fly scheme as well) so the zones and the objects on them can be rolledin/out of memmory in zone-sized blocks.
You said no variable LOD mesh..
If your world doesnt have alot of view blocks yo will see to the horizon and then if you have a large map with a fine mesh you
will be displaying ALOT of quads/triangles.
I did a system with 16x16 mesh quads in zones (a 2 step LOD) and a max range of 8 zones out (window of 15x15 zones visible of the map). The heightmap grid has regular XY coordinates.
A bitmapped angle-zone visibility table does fast subsetting of all the relative zones (pieslice) visible for a particular view angle.
My zone grids are fairly coarse heightmaps so out to 7 zones away
most human sized objects are single pixels.
Objects can be placed/limited to the ground altitude by doing a
line triangle intercept calulation (simplified because of regular grid coordinates and right mesh triangles).
Of course my world map wont fit in memory - 64k x 64k zones (and its for a create terrain on the fly scheme as well) so the zones and the objects on them can be rolledin/out of memmory in zone-sized blocks.
Author
Quote:Original post by Anonymous Poster
Of course my world map wont fit in memory - 64k x 64k zones (and its for a create terrain on the fly scheme as well) so the zones and the objects on them can be rolledin/out of memmory in zone-sized blocks.
That's more what I ask. If my map doesn't fit in memory, can Ogre manage it? Can I extend Ogre to manage a scene that dynamically loads and unloads zones?
I the other hand, I don't need LOD in meshes because there is no problem with visual artifacts making hills or buildings dissapear. In fact, I would set a limit of how far should they be rendered. I think this is no problem here. But I could even add LOD to meshes as the map is generated by a tool that could automatically produce some versions of each zone with diferent levels of detail.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement