How much angle independant AF would be needed before Summed Area Table is quicker?

[bloodbob]

There's two things that you're missing here:
1. Highly-anisotropic cases will never in practice have texels aligned in a rectangle in pixel space.

Thats nice are you saying since ripmap is even worse at matching the sample foot print will be better than or equal to SAT? because its worse at matching the sample foot print.

2. Even with summation over all of the right samples, you can still get aliasing.

So your saying your blurring which "you can still get aliasing." is as good or better than intergration? I guess nvidia was really onto something when they made quinix.

 

[KimB]

No, I'm saying that even if you use a very suboptimal form of SAT mapping you'll still get aliasing. With RIP mapping you can help prevent this by using a LOD bias if you have a particularly nasty texture.

 

[bloodbob]

No, I'm saying that even if you use a very suboptimal form of SAT mapping you'll still get aliasing. With RIP mapping you can help prevent this by using a LOD bias if you have a particularly nasty texture.

An how hard would it to be to prevent this in SAT by expanded the co-ordinates? This of course would give you a continous control over the blurring where as rip maping your working with discrete LOD levels or linear interpolation between two discrete levels.