For example i have a model ( about 2000 triangles).
After converting this model to 1 strip, model
has 3700 triangles.
What better: draw model as 1 strip but with many degenerate triagnles OR create 1 strip
witch minimum strip length for example 10 and
1 list. How fast hardware works witch degenerate triangles ?
Is it bad when one model icrases polygon count about 1.5 times after converting it to 1 strip ?
I don’t think there is an easy answer to that one.
It all depends.
Generally one big strip is useful when the API overhead associated with sending multiple small primitive groups is too high. But this isn’t much of an issue with the latest version of OpenGL.
I’ve seen apps where a lot of 2 triangles strips were better than one triangle list. So…
Personnaly I prefer not to have any degenerate triangles added from the stripification. They alter the geometry of the model in subtle ways that can have sometimes serious impacts (e.g. a rendering in wireframe).
Which explains some design rationals behind Tri Stripper .
Degenerate triangles are fast enough on any hardware with a post T&L vertex cache, but you’re still wasting some ressources sending useless data.
A factor of 1.5 increase in poly count isn’t always a bad thing, that is if you don’t mind degenerate triangles, but the only way to be sure is to benchmark and test, over and over.
[This message has been edited by GPSnoopy (edited 12-07-2003).]