I know the principles behind how QCSAA is supposed to work (as opposed to how, say, regular CSAA works) and subjectively after some testing I like the style it lends to the image enough that I'm considering turning it on by default in certain situations. The question is - I've read that in practice it doesn't work quite as described. Theoretically, it should only be considering coverage samples from the the nearest edges of neighboring pixels along with the local pixel's samples in its calculation, which makes sense. But I've seen it claimed that it's actually just a blur filter over an antialiased image - that is, a 8xQCSAA frame is just rendered using 8xCSAA and then slightly blurred, using the finished pixels as samples.
This is obviously not the same thing. Does anyone know how it works for sure?