Test team: sorry I missed my (self-imposed) deadline today; family obligations....
Sander,
With all this 'construction' going on, would it be possible to programme/automate the population of the CFS2 world with trees and buildings, as in CFS3/FS9/FSX? Use the info that determines what is displayed on certain tiles to put trees and buldings etc?
They would of course have to be of an extremely low poly level, muti-LOD, with low res texs, so would not have too much FR hit.
Shessi
Hi Shessi,
There is one obstacle: we know which Land Class type is placed on each LOD13 tile, but not exactly which texture (the fields/city types that consist of 5 or 7 textures). It can't be a very complex algorithm, but we need to know, in order to have perfectly aligned objects. If we just placed them at random, the visual result would not be worth the effort, IMO. Once somebody reverse engineers that algorithm, we could go a long way....
Then, we could make an equivalent of AGN files (could be as SCASM API's, or tables of object/heading/location).
To populate the whole world, might be a bit too much: In my experiments, I found that you need AT LEAST 20 objects on each LOD13 tile. There are 65536 LOD13 tiles in an LOD5 tile. There are about 1164 populated LOD5's in the FS world. Given that a rough estimate of about 80% of these consist of land, that would be 1.220.523.840 objects.... At minimum each object would consist of: 16 byte GUID + 2x4 byte Coordinate x,y, 4 byte altitude, 2 byte(?) heading, 2 byte(?) scale = 16 bytes. But I suspect it will probably be more. In best case:
19,528,381,440 bytes = 19070685 KB = 18623 MB =
18 GB! So, for each object added to an each LOD13 tile, it's almost an extra GB of disk space (if the whole world is covered)
When I think of it, if 1 object with multiple buildings/trees is made for each texture variation, it could be a lot smaller. Edit: no that wouldn't work because be need to be able to exclude single buildings/trees where there are roads/rivers/rails/water...
Either way, the impact on the FPS would be considerable. The trick for that is to keep the amount of different objects as low as possible; it's much easier for the graphics engine to display 100 instances of 1 single object than to display 25 different objects only once.
Don't even think of putting that all in the GSL.... Maybe with some clever SCASM programming we could get some collision detection working, but that's as far as it could possibly go.
But it would be FUN and COOL for sure!