Hey there !
I have several question regarding occlusion culling with GL_HP_occlusion_test :
from what i read here http://oss.sgi.com/projects/ogl-sample/registry/HP/occlusion_test.txt
I understand that it doesnt matter if the depth buffer is “crowded” with many polygons : that is,the test wont slow down as i render more and more polygons… I just want to make sure about that =)
The other question i have no idea : is the HP_occlusion test faster than the NV_occlusion_query ? i just wondered about that because NV_occlusion_query count the visible pixel, while HP doesnt
I wouldn’t recommend HP_occlusion_test because it’s not allowing asynchronous handling of multiple queries like the NV_occlusion_query and ARB_occlusion_query (which is based on the NV version).
The time for an occlusion query is depending on the time it takes to raster the primitive.
If a hardware advertises both methods, the HP one is probably based on the pixel counter anyway, which means there’s no performance difference of the query in that case.
I don’t know how fast original HP hardware was with that, but you can consider this legacy now that there’s an ARB version.
You can also do more interesting things with the numbers of pixels, like using visibility thresholds for partially occluded geometry.
Ok the multiple queries thing is definetly interesting, i just checked NV_occlusion_query and its working fine. i’ll look into the ARB version too.
thanks for the advices