I’m adapting my code to use VBOs. I’m wondering if implementing stripification would be a good optimization. When using client side vertex buffers, stripification can reduce the amount of data sent by about 1/2, which makes rendering speedier. However, if I’m using server side buffers, there’s no per frame transfer, so I’m wondering if there is any rendering speed difference?
Mark McKay
Most cards should have a post-TnL vertex cache, so as long as you’re using indexed geometry you should get equivilent speed (assuming your indices for triangles in the same order your strips would). Given some clever indexing you could even be able to get the vertex transform faster than with a strip (and I think one of nVidia’s geometry tools tries to do just that).
Although I’d suggest a quick test/profile to check first…