Picking and graphic chip troubles
Hello,
I have an opengl application that uses picking and reading the selection buffer to pick the object closest to the camera.
That works well on all tested graphic cards, but not on my own computer at home which uses an intel G33/G31 Express chipset family. In that case, the depth value is always 0 in the selection buffer and the application can't figure out which object is closest to the camera. Now my question is: is that really a problem of my graphic chip? When I turn all graphic accelerations off, the application picks fine, but rendering time is 10 time slowlier!!
Is there some workaround? Or how do professional applications guaranty picking working on all graphic chips?
Thanks!
Fairly few applications use the OpenGL selection mode, for a couple of reasons: it is limited to only 64 objects in many implementations, and driver support can be very spotty.
The next simplest method is to use the gluUnproject function in concert with the depth buffer, which can give you the exact point of the mouse-click in 3-dimensional space, and you can then compare this point with your objects to pick the right one.
The next simplest method is to use the gluUnproject function in concert with the depth buffer, which can give you the exact point of the mouse-click in 3-dimensional space, and you can then compare this point with your objects to pick the right one.
Thanks for the replies!
I actually could find some other people who are having the same kind of troubles with the same graphic chip, so it must be a hardware (or driver) problem.
Swiftcoder: what you are describing is not picking. I actually also extract the 3D position of my cursor by reading the depth buffer (but for some different reasons). Of course I could compare that 3D position with the center position (or vertice position) of each of my objects, but this does not work when objects are close to each other (or you'll end picking the wrong objects). What I mean to say is your solution is not pixel-accurate.
V-Man: I am surprised by what I read there. That means I should implement the picking by doing a rendering pass with specific colors for each object. That works fine for picking, but I also have a "rectangle selection"-mode, where everything inside a rectangle that I draw on the screen gets selected. Using color-coded picking would not pick my hidden objects (hidden, but still inside the rectangle).
I actually could find some other people who are having the same kind of troubles with the same graphic chip, so it must be a hardware (or driver) problem.
Swiftcoder: what you are describing is not picking. I actually also extract the 3D position of my cursor by reading the depth buffer (but for some different reasons). Of course I could compare that 3D position with the center position (or vertice position) of each of my objects, but this does not work when objects are close to each other (or you'll end picking the wrong objects). What I mean to say is your solution is not pixel-accurate.
V-Man: I am surprised by what I read there. That means I should implement the picking by doing a rendering pass with specific colors for each object. That works fine for picking, but I also have a "rectangle selection"-mode, where everything inside a rectangle that I draw on the screen gets selected. Using color-coded picking would not pick my hidden objects (hidden, but still inside the rectangle).
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement