View Full Version : Performance?

06-23-2004, 11:23 PM
I have a model of e.g. 200,000 triangles. Instead of using naive glNormal() and glVertex() call for each vertex, how much performance improvement can be expected if I use glVertexPoinetEXT() and glNormalPointerEXT() extensions? Or any other more performance-useful extensions/ideas?

Thank you.

06-24-2004, 07:51 AM
why not test and share your results with us?

that would be greatly appreciated, especially for those of us who have yet to try but have also wondered.


06-24-2004, 07:58 AM
You can expect *huge* performance gains. Why don't you do it and see for yourself?

06-26-2004, 09:31 AM
As a side note, vertex arrays have been core since OpenGL 1.1 which is fairly ancient. No need for the EXT unless you're running on some old sgi box or something.

06-26-2004, 01:15 PM
If you have a very big mesh, use Vertex Buffer Objects. You will gain a huge boost in speed.
You can find an example of how to use VBOs on NeHe.gamedev.net

06-26-2004, 06:51 PM
If you use VBO, the the Lock extension is of no use, right? Or does it still help speed things up?

Because from what I understand, the Lock extension keeps your vertices and triangle data in video memory until you unlock them which speeds all your glDrawElements calls up a lot. But using VBO, your data is already in video memory, so would using the Lock not really do anything?

06-27-2004, 03:14 AM
Yes, compiled vertex arrays are pointless with VBOs. VBOs are tighter speced as well and should offer greater performance than CVAs.