Original Post
Hi all, Currently I'm using an offscreen surface (I call it pick screen), that has the same resolution as my visible screen. I render every pickable object in it using some kind of ID as color. This way, I can pick every face of my objects, and can define special object in separate ID value areas. This is simple and great, but I think, that when I'm reading the picked element (that is, I read the pixel prom this buffer which is under the mouse cursor), the surface travels from videomem to the main mem through the AGP every time. Are you aware of any better solution, or at least some method for decreasing the memory consumption, thus the bandwidth requirement od this method? I think that picking by rays would be very-very problematic, since I don't want to check zillions of faces one by one, and using this solution, the Z buffer makes it easy to determine which object is the closest to me. Thanks for your answers in advance, Peter