Vertex Arrays v Display Lists

[Andrew Jameson]

Like many, I'm looking at ways to optimize OpenGL performance. In our application that we're currently developing, all of our objects are within display lists. We're considering moving all of the vertex/normal data into vertex arrays but we'd like to feel that such an effort would be rewarded. So, what's your view ? Is there likely to be any significant improvement ?

Thanks

Andrew Jameson

[Toriton]

>all of our objects are within display lists. We're considering moving all of the vertex/normal data into vertex arrays but we'd like to feel that such an effort would be rewarded. So, what's your view ?

I think it will depend on implementation.

For GeForce case, display list is faster than most methods of vertex array.

I saw data somewhere, which shows compiled vertex array is slightly faster than display list. But very slight, and to use compiled vertex array, you need to check extensions.
If you use compiled vertex array, then the worst case is that, lack of compiled array extensions in users' system forces them to use other methods of vertex array that are much slower than display list. To avoid such worst case, you need to keep both vertex array data and the data for display list(huge memory?)

I would stick with display list currently.
Just my opinion.

San

[fshana]

I can't speak to the vertex arrays but the biggest thing I've seen improve performance is moving textures to texture objects. Since you're using display lists already, I'd try the texture objects BEFORE changing the vertices.
fs