Heya GF guys!
Well, i’m currently & intensively using triangle strips in the project i’m working on ( http://www.orkysquad.org ) and i was just wondering if some of u have done some benchs about triStrip vs quadStrips ?? Maybe some people from NV could help giving a clue?
For some additional precisions we are actually using the ideal case where 1 prim = 1 strip, depending on the primitive style maybe it could be interesting to pass it as GL_QUAD_STRIP… that’s my question! ;))
(of course, not tested quadStrips yet! just
to know your thoughts)
As I understand it, the problem has gotten easier as GPUs get more advanced. BUT, it all comes down to the fact that quadstrips must be turned into triangles, most advanced GPUs do this on the board which is faster than on the driver as it used to be done, but that is still one step more, no matter how fast it is. So the question comes down, how much time will it save you? depends primarily on A) how large the strips are and B) how many strips you are talking about and C) how much time you can afford to loose. I use tri-strips by preference because triangles are the simplest structure, and I align my own data, so do so in tri-strip form. But if you have to convert your data from a 3rd source from quad to tri on a per frame basis, DO NOT, just send it to the GPU, save yourself some CPU time, some headache, and some code. But that is entirely my opinion, never did any speed tests on the difference.
Well, let’s assume it’s all about raw perfs.
the primitives used are containing 120-200 pts. they’re stored onboard using VAR etc etc.
I dont use quad strips, but someone has already said that there is no difference.
It makes sense because they are virtually identical except that the driver will reject a stray vertex for quad strip, but for a tristrip, they all turn into valid triangles.
Mostly I find that triangles, triangle_strip and triangle_fan is enough. Polygon, quads and quad_strip are unnecessary.