I don't quite understand the point in this slide. A primitive is made of individual fragments, which are triangles here(?), and fine granularity means the primitive is divided into more triangles than corse granularity. My question is why it makes more sense to toss out whole primitives? Does it refer to the case when the whole primitive is out of view so we just need to do less judgements?
motoole2
Yes, you are correct. Tossing out whole primitives ("coarse granularity") refers to cases where the triangle, or any primitive for that matter, is completely outside of the view frustum. Hence, we should not render these primitives because they are hidden from the camera.
azzheng
Out of curiosity, what is the advantage of clipping? I feel like clipping creates a lot more points which uses more memory and it also takes quite a bit of work which takes time.
motoole2
It is for efficiency reasons! The number of additional points/vertices is generally quite low relative to the number of fragments (samples) associated with regions of the triangle outside of the view frustum. It would be computationally more expensive to process these fragments that are hidden from view. So clipping helps reduce the overall workload by reducing the number of fragments to process.
I don't quite understand the point in this slide. A primitive is made of individual fragments, which are triangles here(?), and fine granularity means the primitive is divided into more triangles than corse granularity. My question is why it makes more sense to toss out whole primitives? Does it refer to the case when the whole primitive is out of view so we just need to do less judgements?
Yes, you are correct. Tossing out whole primitives ("coarse granularity") refers to cases where the triangle, or any primitive for that matter, is completely outside of the view frustum. Hence, we should not render these primitives because they are hidden from the camera.
Out of curiosity, what is the advantage of clipping? I feel like clipping creates a lot more points which uses more memory and it also takes quite a bit of work which takes time.
It is for efficiency reasons! The number of additional points/vertices is generally quite low relative to the number of fragments (samples) associated with regions of the triangle outside of the view frustum. It would be computationally more expensive to process these fragments that are hidden from view. So clipping helps reduce the overall workload by reducing the number of fragments to process.