David Parks wrote:
The old method relied on depth buffer post-processing, so screen space effects were a no-no and there were always ugly artifacts around silhouettes. It sounds like the new method actually runs through the command buffer twice, generating two entirely separate images. Time will tell how well it interacts with multiple render targets and deferred rendering. Since this is implemented in the driver, there's not much
Interesting, I always wondered about the "consumer" stereo. I had to implement a stereoscopic renderer for my thesis (basically a request of my professor) and put a quick hack into the Ogre3D engine so it would support the quad buffer feature of the Quadro series. With this it's a piece of cake, the driver even generates the necessary blue line to drive an external shutter controller.
That approach of course cuts the frame rate in half since the renderer was simply executed twice with 2 different projection matrices.
I mainly used Paul Bourkes pages for reference but he seems to have rearranged them a bit so my old URL no longer works so one might need to search a bit. His other pages are well worth a visit for any graphics buff anyway.
http://local.wasp.uwa.edu.au/~pbourke Regards, Carsten _______________________________________________ Policies and (un)subscribe information available here: http://wiki.secondlife.com/wiki/SLDev Please read the policies before posting to keep unmoderated posting privileges
