I get the impression he wants to build a series of 2D depth textures, or maybe create single 3D texture… which technically isn’t a problem…
The problem is getting performance out of it any which way he goes about it as he is rendering the scene (I assume) in slivers, meaning for a 64x64x64 map he is looking at 64 rendering passes (less with optimisations of course)…
So, they original method should be one of the faster methods for that particular approach.
Originally posted by solarforge:
[b]I get the impression he wants to build a series of 2D depth textures, or maybe create single 3D texture… which technically isn’t a problem…
The problem is getting performance out of it any which way he goes about it as he is rendering the scene (I assume) in slivers, meaning for a 64x64x64 map he is looking at 64 rendering passes (less with optimisations of course)…[/b]
Some blues clues
I want to construct a 3D texture containing 16bit values, which just happen to be depth, hence the HILO (16bit) format? I’ll be accessing the 3D ‘depth’ map programatically in an NV30 FP (emulation driver of course).
I’ll only be looking at around 5 passes since I only want the first few (5) occluders.