I have a very large point sprite >10 million points and I’d like to figure out what would be the fastest way to rotate and transform this sprite. Right now i’m using glrotate and gltransform and once I get pass about 10 million points my performance drops by quite a bit. Would glulookat implementation be much faster? All I want to do is have the ability to look around the point sprite from all sides.
What makes you think that the computation of the world-to-camera matrix is the bottleneck in your rendering? More than likely, your rendering time would be improved by improving vertex transfer, faster shaders, etc.
I think I’ll have to look through my drawing portion and see what can be improved. I’m using VBO’s to draw the points, would you consider that the most efficient way of drawing points?
There really is no such thing as a “most efficient” way of drawing anything (although I could easily tell you what the least efficient way would be…) - it all depends on your program, how it uses the data, is the data static or dynamic, if dynamic how often does it need to be updated (and how much of it needs updating), what kind of attributes go with your vertexes and are they shared/can they be batched, are you doing things like frustum culling or occlusion, is this even your bottleneck, etc etc etc.
Answering a question like “is a VBO the most efficient way to draw points?” is completely impossible. There are more efficient ways of drawing things given certain sets of circumstances for sure, and it’s frequently a tradeoff between efficiency and flexibility, but only with knowledge of your program’s goals can it be decided which side of that tradeoff is most appropriate for you.
In other words you need to start showing some of your code here.