Yes, I understand what you mean.
But as you wrote "that's going to be difficult/impossible if you're sharing vertices between triangles".
Anyway, no matter how we phrase that thing, VS cannot natively do that. It needs to be instructed in doing that using a texcoord. It is just better to use a GS.
The primitive is not rejected. It is degenerated, a 0-fragments triangle. That's not the same as rejecting a primitive, nor the same as rejecting the rasterized fragments. I notice you have used quotes to de-emphatize the word but I find that idea dangerous to let go.
Because of the limitations of this system and the lack of flexibility, I advice against this method. If it works for you, good (implies independent vertices for terrain?). But I'm having difficulties thinking at it as a best practice.
edit: more elaborations on what "reject" means here.
Show differencesHistory of post edits
#1Krohm
Posted 12 October 2012 - 01:10 AM
Yes, I understand what you mean.
But as you wrote "that's going to be difficult/impossible if you're sharing vertices between triangles".
Anyway, no matter how we phrase that thing, VS cannot natively do that. It needs to be instructed in doing that using a texcoord. It is just better to use a GS.
Because of the limitations of this system and the lack of flexibility, I advice against this method. If it works for you, good (implies independent vertices for terrain?). But I'm having difficulties thinking at it as a best practice.
But as you wrote "that's going to be difficult/impossible if you're sharing vertices between triangles".
Anyway, no matter how we phrase that thing, VS cannot natively do that. It needs to be instructed in doing that using a texcoord. It is just better to use a GS.
Because of the limitations of this system and the lack of flexibility, I advice against this method. If it works for you, good (implies independent vertices for terrain?). But I'm having difficulties thinking at it as a best practice.